betaFeedback

WorksPhysics and Astronomy

Effect of scrape-off losses and current flattening on runaway electron dynamics in ARC

Abstract

Disruptions in high-current tokamaks, such as ARC, may result in megaampere-scale runaway electron currents. In this work, we evaluate mitigation strategies via massive material injection, accounting for the combined impact of scrape-off losses due to vertical displacement events and current profile relaxation during the thermal quench. We find that scrape-off losses alone yield only modest reductions in runaway electron currents, but combining them with current flattening creates a beneficial synergistic effect. The latter broadens the current channel outward, shifting efficient runaway electron generation closer to the plasma edge where the dynamically contracting boundary can effectively scrape off the electron beams. This interplay significantly expands the operational window for mitigation toward lower assimilated densities, theoretically reducing runaway currents to manageable levels at experimentally feasible assimilated quantities.

Introduction

One of the challenges in designing a viable tokamak fusion power plant is the mitigation of plasma disruptions; events where thermal and magnetic confinement are rapidly lost ([10, 26]). Disruptions pose severe risks to the structural integrity of a reactor via three primary mechanisms: intense, localized heat loads on plasma-facing components during the thermal quench (TQ); large electromagnetic forces from eddy and halo currents during the current quench (CQ); and the generation of a relativistic beam of runaway electrons (REs) capable of causing damage upon wall impact ([12]).

To mitigate these hazards, baseline disruption mitigation systems (DMS) for next-step devices rely on massive material injection (MMI) to rapidly radiate away thermal energy and control the subsequent current decay. Such mitigation strategies are being extensively studied on present-day tokamaks ([6, 1, 16, 21, 18, 17, 8, 25, 7, 11, 24, 23, 14]). However, scaling them to future reactor-scale devices involves physics that cannot be fully replicated in existing machines. Consequently, predictive numerical modelling is indispensable. Full three-dimensional (3D) magnetohydrodynamic (MHD) codes offer comprehensive insights into these events but are computationally infeasible for wide-ranging parameter scans. This has motivated the widespread use of reduced-dimensionality tools, such as the 1D fluid-kinetic framework DREAM (Hoppe et al. [13]).

Historically, many 1D disruption studies have operated under the conservative assumption of a fixed magnetic flux-surface geometry, neglecting vertical plasma motion and subsequent scrape-off losses. This omission typically leads to an overestimation of the final RE current relative to experiment. Because open field lines cannot sustain relativistic beams, vertical displacement events (VDEs) scrape off the outer layers of the plasma, rapidly exhausting REs to the wall.

Axisymmetric MHD simulations have demonstrated that this vertical motion restricts the variation of the poloidal magnetic flux inside the remaining last closed flux surface (LCFS) (Wang et al. [30]). Given that the avalanche enhancement factor depends exponentially on this flux variation (Rosenbluth & Putvinski [22]), accounting for VDE-driven scrape-off is important for accurate RE prediction. Recent investigations indicate that incorporating a simplified, flux-conservative scrape-off model can radically alter predictions, making RE avoidance theoretically possible in certain regimes (Vallhagen et al. [28]; Votta et al. [29]).

Recent exploratory modelling utilizing multi-objective Bayesian optimization showed persistent megaampere-scale runaway beams in ARC V3A (Sweeney et al. [27]). Throughout this paper, ARC refers to the V3A design. However, these simulations omitted scrape-off effects and current relaxation induced by stochastization of the magnetic field. The latter phenomenon can be captured via a hyperresistive transport term in Faraday’s law (Boozer [2, 4]).

To assess mitigation performance, we use three reference values as general metrics of success: a maximum runaway electron current IRE<150 kAI_{\mathrm{RE}} < 150\ \mathrm{kA}, a conducted fraction of the pre-disruption thermal energy ηcond<10%\eta_{\mathrm{cond}} < 10\%, and a current quench time in the range 11.3 ms<τCQ<142 ms11.3\ \mathrm{ms} < \tau_{\mathrm{CQ}} < 142\ \mathrm{ms}. These are not ARC design requirements or targets. The τCQ\tau_{\mathrm{CQ}} range is motivated by the electromagnetic-load considerations in Sweeney et al. [27]. The ηcond\eta_{\mathrm{cond}} threshold is our own choice; it serves as a simplified proxy for the actual ARC requirement of avoiding melting of plasma-facing components, which has not been expressed as a limit on the conducted or radiated energy fraction. The IREI_{\mathrm{RE}} threshold is chosen by the authors following earlier mitigation studies such as Sweeney et al. [27].

In this paper, we utilize the DREAM code (Hoppe et al. [13]), including a hyperresistive current-relaxation model and a flux-conservative scrape-off framework (Vallhagen et al. [28]), to re-evaluate mitigated disruption scenarios for ARC. Our findings reveal that while scrape-off alone yields modest reductions, its combination with current flattening expands the operational window for successful disruption mitigation towards lower assimilated densities, potentially suppressing runaways to manageable levels. However, within the heat transport model used here, which likely overestimates conducted losses, satisfying the reference value for conducted heat losses would require larger assimilated quantities than are optimal for runaway suppression.

The remainder of this paper is organized as follows. In § 2, we outline the numerical framework, describing the VDE scrape-off model and the hyperresistive current relaxation in DREAM. § 3 presents the results of the simulations for ARC, contrasting cases with and without current relaxation and scrape-off effects. We summarize our conclusions in § 4.

Disruption modelling

To evaluate the combined effects of vertical plasma motion and current-profile relaxation on RE dynamics during disruptions in ARC, we employ the 1D fluid-kinetic framework DREAM Hoppe et al. [13]. The disruption scenario modelled here mirrors the baseline activated ARC V3A simulation setup described by Sweeney et al. [27], modified to account for a current-density relaxation during the TQ and scrape-off effects.

Following the baseline configuration for an activated ARC scenario, we assume an initial plasma current Ip=12I_{\mathrm{p}} = 12 MA and an on-axis toroidal magnetic field B0=11.3B_{0} = 11.3 T Sweeney et al. [27]. The resistive wall time is set to τw=250\tau_{\mathrm{w}} = 250 ms. Disruption mitigation is initiated via MMI, assuming instantaneous and radially uniform deposition of mixed deuterium (D) and neon (Ne) impurities. The prescribed densities represent assimilated material and are input parameters; the injection and assimilation processes are not explicitly resolved. Throughout this work, we distinguish these assimilated densities from the quantities supplied by the injection system, of which only a fraction may be assimilated.

The thermal quench is triggered by prescribing an elevated Rechester-Rosenbluth-type diffusive transport of heat, superthermal electrons and runaway electrons. The radial heat transport of electrons is calculated assuming a parallel heat flux scale of q∥∼nevteTeq_{\parallel} \sim n_{\mathrm{e}}v_{\mathrm{te}}T_{\mathrm{e}} along the perturbed field lines, with vte=2Te/mev_{\mathrm{te}} = \sqrt{2T_{\mathrm{e}}/m_{\mathrm{e}}} the electron thermal speed, nen_{\mathrm{e}} and TeT_{\mathrm{e}} the electron density and temperature, and mem_{\mathrm{e}} the electron mass. We note that kinetic simulations of the thermal collapse of a nearly collisionless plasma along field lines intersecting a cooling region find that, as a consequence of ambipolarity, the bulk of the electron and ion thermal energy is transported at speeds set by the ion sound speed rather than the electron thermal speed Zhang et al. [31]. A small fraction of fast electrons leaves promptly and sets up a potential that confines the rest. The electron-thermal-speed scaling adopted here is common in reduced disruption models, but it is therefore likely to overestimate conducted heat losses. The implications for the conducted heat fraction are discussed in § 3.1.

The diffusion coefficient corresponds to a spatially and temporally constant, normalized magnetic perturbation of δB/B=0.3%\delta B/B = 0.3\%. It is applied until the temperature falls below 20 eV on radial average, which provides our definition of the end of the thermal quench (this time instance is denoted by tTQt_{\mathrm{TQ}}). The conducted heat loss ηcond\eta_{\mathrm{cond}} is defined as the fraction of the thermal energy lost through heat transport during the thermal quench. The loss is evaluated up to the time when the post-TQ temperature is reached. During the TQ, a diffusive runaway-electron transport corresponding to the same value of δB/B\delta B/B is also applied.

The evolution of the plasma species is resolved by tracking the ionization and recombination dynamics of the background ions, the impurities, and the bulk electron populations. To accurately capture the hot-tail seed generation mechanism during the TQ, the electrons are divided into three populations: the bulk thermal background (also referred to as the “cold” population) and the runaway electrons are modelled using appropriate fluid descriptions. The population at intermediate momenta, that includes typical values of the critical momentum during the TQ, referred to as the “hot” population, is modelled using a pitch-angle-averaged kinetic framework, as in Ekmark et al. [9], which provides an accurate model for seed generation, not only with respect to the hot-tail and Dreicer processes, but also the activated seed generation mechanisms. This approach balances computational tractability with the high fidelity required to capture early-stage seed generation. We do not expect a fully kinetic treatment of the runaway population to significantly change the runaway current evolution predicted here. Such a treatment would provide information about the runaway electron energy distribution, which is relevant for assessing wall loads but is beyond the scope of this work. Moreover, the fully kinetic model in DREAM uses a collision operator expanded about a single Maxwellian background. It is therefore less suited to describing the initial temperature equilibration between the cold injected material and the pre-existing plasma, which our treatment accounts for.

VDE scrape-off model

MHD simulations across various reactor designs, including M3D-C1 simulations for ARC [5], demonstrate that during the runaway-generation phase, the poloidal magnetic flux1 ψ\psi at the last closed flux surface (LCFS) remains approximately constant, i.e., ψ(rLCFS,t)≈ψ(a,t=0)\psi(r_{\mathrm{LCFS}},t) \approx\psi(a,t=0), where aa is the minor radius of the plasma. This behaviour stems from the high conductivity of the first wall, especially if the distance between the wall and the plasma is small, as in ARC. Any flux changes induce strong currents in the wall that act to cancel the flux variation. To avoid the substantial computational expense of RE evolution coupled to 3D MHD simulations while retaining the physics of VDEs, we use a reduced scrape-off model in DREAM, described by Vallhagen et al. [28].

At each computational time step, the instantaneous radius of the last closed flux surface, rLCFS(t)r_{\mathrm{LCFS}}(t), is determined by numerically solving the equation:

ψ(rLCFS,t)=ψ(a,0),(1)\psi(r_{\mathrm{LCFS}},t) = \psi(a,0), \tag*{(1)}

where the right hand side represents ψ\psi at the prescribed radius of the last closed flux surface at t=0t=0. The region outside the boundary r>rLCFSr > r_{\mathrm{LCFS}} represents open magnetic field lines intersecting the first wall. REs entering this zone are rapidly lost, which is enforced numerically by a loss term that is finite only for r>rLCFSr > r_{\mathrm{LCFS}}:

∂nRE∂t=−nREτlossΘ(r−rLCFS),(2)\frac{\partial n_{\mathrm{RE}}}{\partial t} = -\frac{n_{\mathrm{RE}}}{\tau_{\mathrm{loss}}}\Theta(r-r_{\mathrm{LCFS}}), \tag*{(2)}

where Θ\Theta is the Heaviside step function, and the exact value of the loss timescale τloss\tau_{\mathrm{loss}} is unimportant as long as it is significantly (in this case several orders of magnitude) shorter than any other time scale of interest; we set it to τloss=5Δt\tau_{\mathrm{loss}} = 5\Delta t, where Δt\Delta t is the longest allowed simulation time step. We also track the cumulative contribution of scrape-off losses, denoted by IRE,lostI_{\mathrm{RE,lost}}, by accumulating the current contributions removed by this loss term over time, while a finite confined region remains. This quantity expresses the accumulated runaway losses in current units; it does not represent a current that continues to flow after the electrons have been lost. In contrast, in simulations with scrape-off enabled (denoted by +SO), IRE,+SOI_{\mathrm{RE,+SO}} denotes the instantaneous confined runaway current.

A similar loss term is also included on the electron thermal energy density as

∂We∂t=−43γshnecsTeL∥Θ(r−rLCFS),(3)\frac{\partial W_{\mathrm{e}}}{\partial t} = -\frac{4}{3}\frac{\gamma_{\mathrm{sh}}n_{\mathrm{e}}c_{\mathrm{s}}T_{\mathrm{e}}}{L_{\parallel}}\Theta(r-r_{\mathrm{LCFS}}), \tag*{(3)}

where γsh=8\gamma_{\mathrm{sh}}=8 is the heat transmission coefficient, csc_{\mathrm{s}} the sound speed and L∥L_{\parallel} the length of the flux tube along the magnetic field [29]. This term approximates parallel heat losses along open field lines in the halo region using a prescribed sheath heat transmission coefficient. The factor necsn_{\mathrm{e}}c_{\mathrm{s}} represents the particle flux scale associated with the ion saturation current, jsat≃enecsj_{\mathrm{sat}} \simeq en_{\mathrm{e}}c_{\mathrm{s}}. The halo current and its effect on sheath heat transmission are not explicitly resolved in this reduced model. For sufficiently smooth radial dynamics and converged current evolution we use Nr=80N_{\mathrm{r}} = 80 radial grid points.

Hyperresistive current profile relaxation

During the TQ, MHD instabilities stochasticize the magnetic field, which drives a rapid relaxation (i.e., flattening) of the current density profile, manifested as a distinct spike in the total plasma current IpI_{\mathrm{p}} and a sharp reduction in the internal inductance lil_{\mathrm{i}}. We model the spatial redistribution of the current density by a hyperdiffusive term in Faraday’s law, which is constructed to conserve the total magnetic helicity ∝∫ψ(ψt,t) dψt\propto\int\psi(\psi_{\mathrm{t}},t)\,d\psi_{\mathrm{t}} ([4, 20]), where ψt\psi_{\mathrm{t}} denotes the toroidal flux. The evolution of the poloidal flux is governed by

∂ψ∂t=−Vloop+∂∂ψt(ψtΛ∂∂ψtj∥B),(4)\frac{\partial\psi}{\partial t} = -V_{\mathrm{loop}} + \frac{\partial}{\partial\psi_{\mathrm{t}}}\left(\psi_{\mathrm{t}}\Lambda\frac{\partial}{\partial\psi_{\mathrm{t}}}\frac{j_{\parallel}}{B}\right), \tag*{(4)}

where j∥j_{\parallel} is the total parallel plasma current density, and Λ(r,t)\Lambda(r,t) represents the anomalous hyperdiffusion coefficient, commonly termed the hyperresistivity, which is nonzero in regions where the magnetic field lines are stochastic. Furthermore, Vloop=η∥(jΩ/B)2π⟨B2⟩/⟨∇φ⋅B⟩V_{\mathrm{loop}} = \eta_{\parallel}(j_{\Omega}/B)2\pi\langle B^{2}\rangle/\langle\nabla\varphi\cdot\mathbf{B}\rangle represents poloidal flux evolution due to ohmic dissipation, with the parallel ohmic current jΩj_{\Omega} and resistivity η∥\eta_{\parallel}, and ⟨⋅⟩\langle\cdot\rangle denoting flux-surface average. The parallel electric field is calculated solely from this term (hence the notation VloopV_{\mathrm{loop}}) as E∥=η∥(jΩ/B)⟨B2⟩E_{\parallel} = \eta_{\parallel}(j_{\Omega}/B)\sqrt{\langle B^{2}\rangle}, since the volume-integrated parallel electric field vanishes in stochastic regions. Consequently, electrons that sample this volume along stochastic field lines experience no net acceleration, as pointed out by Boozer [3]. The hyperresistive term is implemented with zero radial helicity flux at the outer boundary of the fixed computational domain, so that this term conserves the domain-integrated magnetic helicity. This condition is distinct from the resistive-wall boundary condition governing poloidal flux evolution and is not imposed at the moving LCFS used in the scrape-off model.

The magnitude of the hyperresistivity can be estimated following Boozer [4] as

Λ≈1144μ04πVAψt2Nt,(5)\Lambda\approx\frac{1}{144}\frac{\mu_{0}}{4\pi}\frac{V_{\mathrm{A}}\psi_{\mathrm{t}}^{2}}{N_{\mathrm{t}}}, \tag*{(5)}

where μ0\mu_{0} is the vacuum permeability, VA=B/μ0ρV_{\mathrm{A}} = B/\sqrt{\mu_{0}\rho} is a representative Alfvén speed, ρ\rho is the mass density, and NtN_{\mathrm{t}} denotes the number of toroidal transits required for a stochastic field line to sample the entire radial domain. The otherwise unknown parameter NtN_{\mathrm{t}} can be related to the magnetic perturbation strength δB/B\delta B/B used to model the transport of heat and energetic electrons ([29]), yielding

Λ≈μ0πqR02VAψt2288a2(δBB)2,(6)\Lambda\approx\frac{\mu_{0}\pi qR_{0}^{2}V_{\mathrm{A}}\psi_{\mathrm{t}}^{2}}{288a^{2}}\left(\frac{\delta B}{B}\right)^{2}, \tag*{(6)}

where aa and R0R_{0} are the minor and major radii of the plasma, respectively, and qq is the safety factor. To obtain this relation, we identify the characteristic parallel distance travelled by a runaway electron before diffusing out of the plasma with L=2πR0NtL = 2\pi R_{0}N_{\mathrm{t}}. This distance can also be estimated from the characteristic diffusive loss time, τRE∼a2/DRE\tau_{\mathrm{RE}} \sim a^{2}/D_{\mathrm{RE}}, as L∼cτREL \sim c\tau_{\mathrm{RE}}. Using the Rechester–Rosenbluth-type estimate DRE∼ρREc(δB/B)2D_{\mathrm{RE}} \sim\rho_{\mathrm{RE}}c(\delta B/B)^{2} therefore gives Nt∼a2/[2π2qR02(δB/B)2]N_{\mathrm{t}} \sim a^{2}/[2\pi^{2}qR_{0}^{2}(\delta B/B)^{2}], which, when inserted into Eq. (5), yields Eq. (6). Based on the estimate Eq. (6), typical values of Λ0\Lambda_{0} are of order unity in units of Wb3/C\mathrm{Wb}^{3}/\mathrm{C}. At this level of hyperresistivity, the effects of current relaxation are effectively saturated: varying Λ\Lambda by factors of order unity does not appreciably change the results. Furthermore, the radial variations of Λ\Lambda considered in Appendix § 8 have only a small effect on the resulting runaway currents. Finally, we note that Λ\Lambda is set to zero outside the thermal quench in our simulations.

This formulation allows a synergistic interplay: hyperresistive current relaxation flattens j∥j_{\parallel}, broadening the current channel outward, which in turn shifts efficient runaway generation towards the edge, where it is subsequently scraped off by the dynamically contracting boundary.

Runaway generation in the presence of massive material injection

Mitigation efficacy vs. assimilated material quantities

We start by evaluating disruption mitigation efficacy across a parameter space of assimilated deuterium and neon densities, using the reference values for IREI_{\mathrm{RE}}, ηcond\eta_{\mathrm{cond}} and τCQ\tau_{\mathrm{CQ}} introduced in § 1. The current quench time is evaluated as described by Ekmark et al. [9].

Parameter regions satisfying these reference values are highlighted and bounded by coloured contours in Figure 1; red corresponding to IREI_{\mathrm{RE}}, blue to ηcond\eta_{\mathrm{cond}}, and yellow to τCQ\tau_{\mathrm{CQ}}. The gray region in the lower-left corner, with a boundary close to the τCQ=142 ms\tau_{\mathrm{CQ}} = 142\ \mathrm{ms} limit, indicates an incomplete TQ. An incomplete TQ occurs when a substantial fraction of the plasma ohmically reheats after the elevated transport window, sustaining a long-lived ohmic current. Thus, we use the term incomplete TQ henceforth to denote cases in which ohmic reheating also prevents completion of the CQ. The practical criterion we employ for incomplete TQ is τCQ>500\tau_{\mathrm{CQ}} > 500 ms or a remnant ohmic current at the end of the simulation (that is slightly longer than the upper τCQ\tau_{\mathrm{CQ}} limit) higher than 1 MA. The figure panels compare scenarios without (-SO, upper panels) and with (+SO, lower panels) scrape-off effects, alongside cases without (-CR, left panels) and with (+CR, right panels) hyperresistive current density relaxation.

Figure 1. Regions satisfying the reference values (colour shaded areas) for the conducted heat fraction (ηcond<10%\eta_{\mathrm{cond}} < 10\%, blue), current quench time (11.3 ms<τCQ<142 ms11.3\ \mathrm{ms} < \tau_{\mathrm{CQ}} < 142\ \mathrm{ms}, yellow), and maximum runaway electron current (IRE<150 kAI_{\mathrm{RE}} < 150\ \mathrm{kA}, red), in the parameter space of assimilated densities. (a) and (b) are without, (c) and (d) are with scrape-off (-SO and +SO, respectively). (a) and (c) are without, (b) and (d) are with current relaxation (-CR and +CR). The gray region indicates assimilated densities resulting in an incomplete TQ and CQ. The plotted parameter domain is covered by 32×3232 \times32 simulations, equidistant on the logarithmic scales. Estimated upper limits of nDn_{\mathrm{D}} and nNen_{\mathrm{Ne}} at planned MMI capabilities and an assumed 10% assimilation rate are indicated with the dotted rectangular boundary.

The conducted heat fraction reference value is met only in the high-D–high-Ne region of the parameter space, where radiated losses are sufficiently large. The corresponding blue area in the upper right corner shows negligible sensitivity to either scrape-off or current density relaxation effects. Similarly, the 11.3 ms τCQ\tau_{\mathrm{CQ}} limit (upper yellow curve in the figure), which emerges at high values of nDn_{\mathrm{D}} or nNen_{\mathrm{Ne}}, is also fairly insensitive to these effects. The ηcond\eta_{\mathrm{cond}} boundary should be interpreted with caution. If parallel energy transport is limited by the ion sound speed [31], conducted losses would be reduced and radiation would dominate over a larger part of the parameter space. The ηcond<10%\eta_{\mathrm{cond}} < 10\% region would then be expected to extend towards lower assimilated densities, possibly substantially. Sweeney et al. [27] found that a sheath-limited heat-flux boundary condition substantially reduces ηcond\eta_{\mathrm{cond}} while leaving the runaway dynamics largely unchanged. We therefore expect the IREI_{\mathrm{RE}} results presented here to be considerably less sensitive to this model choice than ηcond\eta_{\mathrm{cond}}.

On the other hand, the 142 ms τCQ\tau_{\mathrm{CQ}} limit (lower yellow curve in the figure), which closely follows the incomplete TQ boundary, remains largely unaffected by scrape-off in the absence of current relaxation. However, when current relaxation is included, the region exceeding this limit shifts toward lower assimilated densities. The reason is that reheating predominantly takes place in the initially hotter, inner plasma regions where the current density is highest. Flattening the current profile reduces the core ohmic current density, making the plasma less susceptible to reheating. Furthermore, incorporating hyperresistivity amplifies the effect of scrape-off, which is consistent with a radially broader reheated region resulting from the flatter current density profile.

The most significant changes occur in the parameter space where the runaway current remains below its reference value. Without scrape-off, this value is essentially unattainable within the explored assimilated density ranges, except in the excluded incomplete TQ region. Here, only the simulation at nD=3.2×1022 m−3n_{\mathrm{D}} = 3.2 \times10^{22}\,\mathrm{m}^{-3} and nNe=1×1017 m−3n_{\mathrm{Ne}} = 1 \times10^{17}\,\mathrm{m}^{-3} has IRE<150I_{\mathrm{RE}} < 150 kA and only when current relaxation is neglected. With scrape-off there are successful mitigation zones in the high-D–low-Ne quadrant of the parameter space. In this regime, deuterium efficiently enhances collisional drag on REs, while the relatively slow plasma cooling and moderate post-TQ temperatures suppress hot-tail seeding and avalanches. This benefit diminishes at the absolute highest deuterium densities, where extremely low post-TQ temperatures trigger partial deuterium recombination, driving a rapid increase in IREI_{\mathrm{RE}} as nDn_{\mathrm{D}} increases further. Finally, incorporating current relaxation amplifies the effectiveness of scrape-off, significantly expanding the parameter space that successfully meets the IREI_{\mathrm{RE}} target.

However, even when accounting for both scrape-off and current flattening, no single parameter region simultaneously satisfies all three reference values. Figure 1 reveals a narrow operational point where the red and blue regions nearly intersect. Unfortunately, reaching this point requires a deuterium density that far exceeds the expected injection and assimilation capabilities of ARC’s planned MMI systems. Assuming a 10% assimilation rate, the estimated upper operational limits are nD=1.4×1021 m−3n_{\mathrm{D}} = 1.4 \times10^{21}\,\mathrm{m}^{-3} and nNe=7.4×1019 m−3n_{\mathrm{Ne}} = 7.4 \times10^{19}\,\mathrm{m}^{-3}, indicated by dotted lines in Figure 1.

Runaway-electron current vs assimilated material

Since IREI_{\mathrm{RE}} demonstrates the highest sensitivity to scrape-off and current relaxation, Figure 2 examines this quantity in greater detail. In addition to the 0.15 MA contour, which corresponds to the red boundary in Figure 1, 1 MA and 3 MA contours are included to complement the colour shading. These additions illustrate that for all scenarios, there is a region where IREI_{\mathrm{RE}} drops below 1 MA, though only for high deuterium and low neon densities when scrape-off is not considered.

Figure 2. Map of the predicted maximum runaway electron current as a function of the assimilated densities. (a) and (b) are without, (c) and (d) are with scrape-off (-SO and +SO, respectively). (a) and (c) are without, (b) and (d) are with current relaxation (-CR and +CR). Contours denote specific runaway current levels (0.15 MA, 1 MA, and 3 MA). The square (■) indicates a case with nD=1.40×1021 m−3n_{\mathrm{D}} = 1.40 \times10^{21}\ \mathrm{m}^{-3} and nNe=2.26×1018 m−3n_{\mathrm{Ne}} = 2.26 \times10^{18}\ \mathrm{m}^{-3}. Compared to this, the triangle (▲) case has an increased nNe=5.13×1019 m−3n_{\mathrm{Ne}} = 5.13 \times10^{19}\ \mathrm{m}^{-3}, and the circle (●) case has an increased nD=3.28×1022 m−3n_{\mathrm{D}} = 3.28 \times10^{22}\ \mathrm{m}^{-3}. The gray region indicates assimilated densities resulting in an incomplete TQ and CQ.

Without scrape-off (upper panels of Figure 2), current flattening is primarily beneficial above nNe=1018 m−3n_{\mathrm{Ne}} = 10^{18}\ \mathrm{m}^{-3}, extending the 3 MA contour toward lower nDn_{\mathrm{D}} values, but for nNe<3×1017 m−3n_{\mathrm{Ne}} < 3 \times10^{17}\ \mathrm{m}^{-3}, IREI_{\mathrm{RE}} is increased. The corresponding “bubble”, a region where current flattening is ineffective in the absence of scrape-off, vanishes once scrape-off is enabled. When scrape-off losses are included, current flattening is universally beneficial within the explored parameter space, expanding the regions where IREI_{\mathrm{RE}} lies below any given level. This is particularly clear for the three explicitly marked contours at 0.15, 1 and 3 MA in Figure 2c and Figure 2d.

Current evolution in representative cases

In the following, we will analyze the runaway electron dynamics at three representative points within the assimilated density parameter space. The baseline case, denoted with the square marker in Figure 2, minimizes IREI_{\mathrm{RE}}, utilizing a practically achievable deuterium density (nD≤1.4×1021 m−3n_{\mathrm{D}} \le1.4 \times10^{21}\ \mathrm{m}^{-3}). In this case the TQ is complete, regardless of scrape-off or current relaxation effects. When both current relaxation and scrape-off are accounted for, IRE=349 kAI_{\mathrm{RE}} = 349\ \mathrm{kA}; although this exceeds the 150 kA reference value, modern runaway electron experiments routinely manage currents of this magnitude. Additionally, a technically achievable scenario with higher assimilated neon content (nNe=5.13×1019 m−3n_{\mathrm{Ne}} = 5.13 \times10^{19}\ \mathrm{m}^{-3}, triangle marker) is evaluated, despite offering no IREI_{\mathrm{RE}} mitigation benefits. Finally, a scenario with highly elevated deuterium content (nD=3.28×1022 m−3n_{\mathrm{D}} = 3.28 \times10^{22}\ \mathrm{m}^{-3}, circle marker) serves as a theoretical reference point, even though it exceeds currently planned injection capabilities.

Without current relaxation

Figure 3 illustrates the evolution of the total current and current density for these scenarios without accounting for current relaxation. The high-Ne case, shown in Figure 3a and d, produces a runaway current density that is strongly localized in the core (mind the logarithmic scale of the colour-bar for jREj_{\mathrm{RE}}). Although the outer third of the plasma column is scraped off (represented by the black region in the upper right part of Figure 3d), this has practically no effect on the total runaway current evolution. This can be explained by the scrape-off process being, to some degree, self-reinforcing: The appearance of a RE plateau slows or stops the current decay, as well as the poloidal flux variation inside the plasma that would favour scrape-off. Thus, if a large RE population is generated sufficiently fast, scrape-off can not be expected to reach the core and mitigate a formed RE beam. This parameter point is representative of the low-D–high-Ne region, where scrape-off remains ineffective without current relaxation, consistent with Figure 2a and Figure 2c.

Figure 3. (a)-(c) Time evolutions of the total plasma currents and the runaway electron currents without current density relaxation, for the assimilated quantity combinations represented by the markers in Figure 2. Solid and dashed curves are total plasma current, dash-dotted and dotted are runaway currents. Dashed and dotted are without, solid and dash-dotted are with scrape-off. (a) low-D–high-Ne (▲); (b) low-D–low-Ne baseline (■); and (c) high-D–low-Ne (●) scenario. (d)-(f) Spatio-temporal evolution of the runaway current density with scrape-off, for the respective assimilated densities of panels (a)-(c).

The low-D–low-Ne baseline case, shown in Figure 3b and e, exhibits a slower CQ due to lower resistivity. Without scrape-off the total current increases significantly after reaching a minimum at around 62 ms, due to the longer ohmic time-scales allowing magnetic flux to diffuse across the resistive wall. (Most of this increase is outside the plotted time range.) This diffusion would in reality drive the development of massive RE currents during the nominal RE “plateau” phase, but, since the RE current continues increasing in this simulation (outside of the plotted domain), it does not form a true plateau. However with scrape-off, these longer timescales also provide more time for scrape-off to occur. The effect of scrape-off thus counteracts the RE generation driven by electric field diffusion, limiting the growth of the RE current. Although IREI_{\mathrm{RE}} still reaches almost 4 MA, it does not grow significantly during this phase. By maintaining a constant magnetic flux at the last closed flux surface, the scrape-off process effectively neutralizes the impact of magnetic flux diffusion across the conducting wall.

The high-D–low-Ne case, shown in Figure 3c and f, proves the most effective for RE suppression. The elevated deuterium content sustains a high effective critical electric field; thus even without scrape-off IREI_{\mathrm{RE}} decays rapidly due to the increased collisional drag and peaks below 3 MA. In this scenario, the initial significant RE population emerges closer to the edge. Here, localized cooling facilitates deuterium recombination at post-TQ temperatures, rendering scrape-off highly efficient at depleting the RE current. When scrape-off is enabled, the entire plasma column is exhausted in under 25 ms. This rapid elimination contrasts sharply with the other evaluated cases, where the scrape-off process decelerates significantly once the RE current begins to dominate the total plasma current.

With current relaxation

Next, we consider the same cases with the current relaxation accounted for, shown in Figure 4, corresponding to Figure 2b and d. The first difference to observe is the initial IpI_{\mathrm{p}}-spike, driven by a rapid current density redistribution (occurring faster than the resistive timescale) at a fixed magnetic helicity. This spike is approximately 30 % for the low-D injection cases and 15 % for the high-D case. Recalling that current relaxation is only kept active during the TQ, the reduced spike in the high-D scenario is partly due to the shorter TQ interrupting the relaxation process. In addition, the MMI increases the plasma mass density, which decreases the Alfvén speed and extends the natural timescale of the current relaxation.

Figure 4. (a)-(c) Time evolutions of the total plasma currents and the runaway electron currents with current density relaxation, for the assimilated quantity combinations represented by the markers in Figure 2. Solid and dashed curves are total plasma current, dash-dotted and dotted are runaway currents. Dashed and dotted are without, solid and dash-dotted are with scrape-off. (a) low-D–high-Ne (▲); (b) low-D–low-Ne baseline (■); and (c) high-D–low-Ne (●) scenario. (d)-(f) Spatio-temporal evolution of the runaway current density with scrape-off, for the respective assimilated densities of panels (a)-(c).

The low-D–high-Ne case remains qualitatively similar to the case without current relaxation, see Figure 4a. However, now scrape-off provides a slight improvement by broadening the radial distribution of the runaway current, making it less centrally peaked. The low-D–low-Ne baseline case exhibits a much more significant improvement. Although the RE current density remains core localized, the flattened current profile reduces the induced electric field at the end of the TQ, which weakens the initial drive for RE growth. This delay allows the entire plasma column to be scraped off before a massive core RE current can develop, resulting in a dramatic reduction in IREI_{\mathrm{RE}} from more than 4 MA (without scrape-off; the exact value slowly increases with time) to just 349 kA (with scrape-off). In the high-D case (Figure 4f), the RE current density remains near the plasma edge, enabling its efficient removal via scrape-off. Consequently, the maximum RE current drops even lower than in the case without current density relaxation. Furthermore, comparing Figure 2c and Figure 2d reveals that the extreme high nDn_{\mathrm{D}} regime, where deuterium recombination previously aggravated RE generation, is pushed to even higher values of nDn_{\mathrm{D}}. This favourable shift occurs because current flattening locally increases the edge ohmic current density, maintaining a higher post-TQ edge temperature. As a result, the synergistic combination of current relaxation and scrape-off successfully suppresses IREI_{\mathrm{RE}} to negligible levels in this regime.

If deuterium assimilation is constrained near the assumed limit of nD=1.4×1021 m−3n_{\mathrm{D}} = 1.4 \times10^{21}\ \mathrm{m}^{-3}, maximizing deuterium assimilation becomes the optimal strategy to prevent an incomplete TQ. Similarly, the baseline neon concentration of nNe=2.26×1018 m−3n_{\mathrm{Ne}} = 2.26 \times10^{18}\ \mathrm{m}^{-3} represents a suitable operating point; it is sufficiently high to avoid incomplete TQs while enabling current density relaxation and scrape-off to synergistically reduce IREI_{\mathrm{RE}} to manageable levels. Increasing nDn_{\mathrm{D}} is highly advantageous; in such scenarios, holding nNen_{\mathrm{Ne}} constant provides additional margin below the IREI_{\mathrm{RE}} reference value, whereas slightly increasing nNen_{\mathrm{Ne}} along the 0.15 MA contour in Figure 2d lowers the conducted heat fraction while maintaining controlled RE currents. With this approximate delineation of the interesting assimilation parameter space, simulations incorporating realistic injection, assimilation, and current profile evolution dynamics are required to find optimal operating scenarios, given the sensitivity of both IREI_{\mathrm{RE}} and the incomplete TQ boundary to transient plasma behaviour.

Avalanche gain factor

It is interesting to consider the avalanche gain factors, GavaG_{\mathrm{ava}}, defined as the total runaway current normalized to the current at the end of the TQ, shown in Figure 5. As we have already established, scrape-off has practically no effect in the low-D–high-Ne case of Figure 5a, due to the core localized RE population. However, now it is more clearly seen that the avalanche gain is reduced when current relaxation is accounted for, which is consistent with the reduced ohmic current in the core at the end of the TQ, and corresponding weaker electric field induction.

Figure 5. Time evolution of the avalanche gain GavaG_{\mathrm{ava}} for the (a) low-D–high-Ne (▲); (b) low-D–low-Ne baseline (■); and (c) high-D–low-Ne (●) scenario. Dotted and dash-dotted curves are without, dashed and solid curves are with scrape-off; dotted and dashed curves are without, dash-dotted and solid are with current relaxation.

In the low-D–low-Ne baseline case of Figure 5b, we find a weaker effect of the current relaxation and the scrape-off on the avalanche rate individually. Both effects have comparable slight reductions of the final RE current, due to the still core localized but somewhat wider current density profile being weakly impacted by either of these effects. However their combined effect not only has the reduced rate caused by the current flattening, but the maximum value is also strongly affected by a complete scrape-off of the plasma around the time the no-scrape-off case reaches its plateau.

The effect of current relaxation appears earlier and remains more significant in the high-D–low-Ne case, than in the other two scenarios, translating to a significant reduction of the final RE current from current relaxation only. This early reduction, and corresponding delay, of the avalanche gain does not in itself explain why maximum RE current becomes smaller with current relaxation (without scrape-off), since the GavaG_{\mathrm{ava}} ratios are comparable to those seen in Figure 5a. Here an additional difference is that the RE current maximum is reached at comparable times. Complete scrape-off is achieved independently of the current relaxation, while scrape-off is more efficient when it acts on a plasma with fewer runaways to begin with, since the scrape-off process is, as previously discussed, somewhat self-reinforcing. In the case with both scrape-off and current relaxation we also observe a double-peak in the evolution of GavaG_{\mathrm{ava}}. This is because in this scenario there are two distinct RE populations, one core localized and one closer to the edge. The latter gets scraped off first, while the smaller, core localized population keeps growing a bit longer, but since it has no time to reach macroscopic values and back-react on the scrape-off process, it also gets scraped off eventually.

Scraped-off runaway current

Finally, we consider the evolution of the scraped-off runaway electron current for the three representative scenarios. Accounting for scraped-off runaways is important because their energy is ultimately deposited in the wall. Even at relatively low energies, they therefore contribute to the wall heat load, while sufficiently energetic runaways may penetrate the surface and cause additional subsurface damage. Consequently, considering only the runaway current remaining within the plasma may not fully capture the potential impact on plasma-facing components. The resulting wall loads depend both on the time available for the runaways to gain energy before being scraped off and on the rate at which they are lost. Although our fluid treatment of the runaway population does not provide information on the energy spectrum of the lost runaways, it can nevertheless be used to obtain conservative estimates of their contribution to the wall heat load.

Figure 6 illustrates the varying impact of the scrape-off mechanism for different assimilated densities, by comparing the runaway current without scrape-off (IRE,−SOI_{\mathrm{RE,-SO}}, solid), the scraped-off runaway current (IRE,lostI_{\mathrm{RE,lost}}, dotted), and the runaway current with scrape-off (IRE,+SOI_{\mathrm{RE,+SO}}, dashed). In the low-D–high-Ne scenario, Figure 6a, the combined total of the confined and scraped-off runaway currents follows, but is slightly higher than the runaway current generated in the absence of scrape-off, such that while the confined runaway current reaches approximately 3 MA, the total of the confined and scraped-off runaways (dash-dotted) reaches above 4 MA. On the other hand, in the baseline low-D–low-Ne scenario, Figure 6b, the scrape-off mechanism dramatically reduces the total runaway generation. While the runaway current without scrape-off grows continuously beyond 4 MA, accounting for scrape-off limits the total generated runaways to a plateau near 0.8 MA. Finally, in the high-D–low-Ne scenario, Figure 6c, where high initial deuterium content limits the runaway growth to approximately 2 MA without scrape-off, applying the scrape-off model suppresses the confined runaway current almost entirely and restricts the total scraped-off current to 0.3 MA.

Figure 6. Evolution of the scraped-off (IRE,lostI_{\mathrm{RE,lost}}) and confined (IRE,+SOI_{\mathrm{RE,+SO}}) runaway electron currents for the low-D-high-Ne (a), baseline low-D-low-Ne (b), and high-D-low-Ne (c) scenarios, illustrating the varying efficacy of the scrape-off mechanism, compared to the total runaway current generated without scrape-off (IRE,−SOI_{\mathrm{RE,-SO}}), as well as the sum of the scraped off and confined runaway currents (IRE,+SO+IRE,lostI_{\mathrm{RE,+SO}} + I_{\mathrm{RE,lost}}). Here, current relaxation is accounted for.

In the cases shown in Figure 6b and c, the cumulative lost runaway current, IRE,lostI_{\mathrm{RE,lost}}, exceeds the maximum confined runaway current. This is possible because IRE,lostI_{\mathrm{RE,lost}} accumulates losses over time, whereas IRE,+SOI_{\mathrm{RE,+SO}} represents the instantaneous confined runaway current. For example, if runaway generation in the confined region balances scrape-off losses, the confined runaway current remains constant while IRE,lostI_{\mathrm{RE,lost}} continues to increase and can exceed it. In calculating IRE,lostI_{\mathrm{RE,lost}}, we accumulate losses only while a finite confined region remains. Accordingly, in Figure 6b and c, IRE,lostI_{\mathrm{RE,lost}} stops increasing once the plasma has been completely scraped off and IRE,+SOI_{\mathrm{RE,+SO}} vanishes. A plateau in IRE,lostI_{\mathrm{RE,lost}} therefore indicates that no further losses are being accumulated, rather than the persistence of a runaway beam. Its time derivative, dIRE,lost/dt\mathrm{d}I_{\mathrm{RE,lost}}/\mathrm{d}t, measures the scrape-off loss rate in current-equivalent units and characterizes the temporal concentration of these losses. This rate is relevant to assessing wall loads, together with the energy distribution of the lost electrons and their deposition footprint. A comparison with the wall loads from a subsequent impact of the remaining confined beam would additionally require its loss timescale and impact characteristics, which are not modelled here.

The evolution of the edge safety factor, evaluated at the instantaneous LCFS, is shown in Figure 7 together with the radius of the LCFS. Without scrape-off, the plasma boundary is fixed and qedgeq_{\mathrm{edge}} increases substantially during the current quench as the total plasma current decays. With scrape-off, by contrast, the contraction of the confined cross-section keeps qedgeq_{\mathrm{edge}} low; after an initial phase at qedge≈3q_{\mathrm{edge}} \approx3–4, it decreases as the LCFS moves inward. In the low-D–high-Ne case (▲), the boundary contraction slows down significantly once the runaway current dominates, leaving a confined region of radius rLCFS≈0.55r_{\mathrm{LCFS}} \approx0.55–0.750.75 m, while qedgeq_{\mathrm{edge}} reaches approximately 1.5 (+CR) and 2 (-CR) during these simulations. In the baseline (■) and high-D (●) cases, qedgeq_{\mathrm{edge}} drops below 2 and eventually below unity. The plasma is eventually scraped off completely in these cases, except in the baseline +SO/-CR simulation. This suggests that additional current-driven MHD instabilities could affect the subsequent evolution and runaway losses. The onset and consequences of such instabilities depend on the current profile and plasma–wall coupling and cannot be determined from qedgeq_{\mathrm{edge}} alone. These effects are not resolved by the present reduced model.

Figure 7. Time evolution of (a)–(c) the edge safety factor qedgeq_{\mathrm{edge}}, evaluated at the instantaneous LCFS, and (d)–(f) its minor radius rLCFSr_{\mathrm{LCFS}}, for the (a,d) low-D–high-Ne (▲), (b,e) low-D–low-Ne baseline (■), and (c,f) high-D–low-Ne (●) scenarios. Dotted and dash-dotted curves correspond to simulations without scrape-off, while dashed and solid curves include scrape-off. Dotted and dashed curves omit current relaxation, while dash-dotted and solid curves include it. In panels (a)–(c), the gray solid and dotted horizontal lines mark qedge=2q_{\mathrm{edge}} = 2 and 1, respectively. Without scrape-off, rLCFSr_{\mathrm{LCFS}} remains fixed at the initial minor radius, so only the +SO cases are shown in panels (d)–(f).

Conclusions

By incorporating both a hyperresistive current relaxation model and a flux-conservative scrape-off model in the DREAM fluid-kinetic framework, MMI scenarios were evaluated using varying densities of deuterium and neon in ARC. The simulations demonstrated that operating under the conservative assumption of a fixed magnetic flux-surface geometry leads to an overestimation of the final runaway electron current.

The inclusion of scrape-off effects alone yields modest reductions in runaway currents, particularly in parameter regimes with high deuterium and low neon densities where the critical electric field remains high. However, combining scrape-off with hyperresistive current flattening creates a highly beneficial synergistic effect. Hyperresistive current relaxation broadens the current channel outward, shifting efficient runaway electron generation closer to the plasma edge. This allows the dynamically contracting boundary, dictated by the vertical displacement, to more effectively scrape off the electron beams. This interplay significantly expands the operational window for successful disruption mitigation toward lower assimilated densities and reduces the confined runaway current to more manageable levels.

Despite the improved runaway suppression, simultaneously meeting all three reference values introduced in § 1 remains challenging. No region of the explored parameter space satisfies all three simultaneously: a maximum runaway electron current below 150 kA, a conducted heat fraction below 10%, and a current quench duration between 11.3 ms and 142 ms. Relaxing the runaway current reference value to approximately 200 kA already allows the regions meeting the current and heat-loss reference values to overlap.

Note that these conclusions regarding the conducted heat fraction should be regarded as conservative. As noted in § 2, the electron-thermal-speed scaling of the parallel heat flux used here likely overestimates conducted losses [31], and Sweeney et al. [27] found that the sheath-limited heat flux boundary condition of Kiramov et al. [15] substantially reduces the conducted heat fraction while leaving the runaway electron dynamics largely unchanged. A more restrictive treatment of parallel heat losses could therefore extend the region satisfying the ηcond\eta_{\mathrm{cond}} reference value towards lower assimilated densities. Moreover, ηcond\eta_{\mathrm{cond}} is only a proxy for the relevant engineering criterion, avoidance of melting of plasma-facing components, which also depends on the spatial and temporal distribution of the deposited heat.

The present study uses a fixed representative value of δB/B\delta B/B to isolate the combined effects of current flattening and scrape-off. Although current relaxation is in the saturated regime at this value (Appendix § 8), the total runaway current and conducted heat loss remain sensitive to TQ transport [19]. Changes in δB/B\delta B/B could therefore shift the quantitative mitigation boundaries through their effects on plasma cooling and runaway seeding, even though scrape-off occurs predominantly after the TQ. This sensitivity is not explored here.

Finally, we note that achieving a viable disruption mitigation strategy will require more advanced simulations that incorporate detailed modelling of MMI dynamics, impurity assimilation, and precise current density profile evolution. Furthermore, the scraped-off runaway current can substantially exceed the maximum confined runaway current. Assessing the resulting wall loads will require modelling the energy distribution of the lost electrons and their spatial and temporal deposition on the wall.

Acknowledgements

The authors are grateful to Mathias Hoppe, Max Larsson, Oskar Vallhagen, Lorenzo Votta and Björn Zaar for fruitful discussions. The computations were enabled by resources provided by the National Academic Infrastructure for Supercomputing in Sweden (NAISS), partially funded by the Swedish Research Council through grant agreement no. 2022-06725. The work has been carried out within the framework of the EUROfusion Consortium, funded by the European Union via the Euratom Research and Training Programme (Grant Agreement No 101052200 — EUROfusion). Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them.

Declaration of Interests

The authors report no conflict of interest.

Funding

This work was supported by the Swedish Research Council (Dnr. 2022-02862), the Swedish Energy Agency, and by the Knut and Alice Wallenberg foundation (Dnrs. 2022.0087 and 2023.0249). The authors affiliated with Commonwealth Fusion Systems (CFS) and MIT are funded by CFS.

Effect of radially varying hyperresistivity

To assess the sensitivity of the results to a possible radial variation of Λ\Lambda, we use the profile

Λ(r)=Λ0p+12p+1−1[1+(ra)2]p(7)\Lambda(r) = \Lambda_0 \frac{p+1}{2^{p+1}-1}\left[1+\left(\frac{r}{a}\right)^2\right]^p \tag*{(7)}

which fulfils the constraint that the average ∫0adr rΛ(r)/∫0adr r\int_0^a \mathrm{d}r\,r\Lambda(r)/\int_0^a \mathrm{d}r\,r is constant and given by the estimate Eq. (6), here denoted by Λ0\Lambda_0. Profiles with p=0p=0 are radially constant, whereas profiles with p<0p<0 (p>0p>0) are biased towards the core (edge). The latter may be used to emulate the effect of an outward-increasing magnetic perturbation level due to externally applied magnetic perturbations. For reference, recalling our choice δB/B=0.3%\delta B/B=0.3\%, taking ψt\psi_t at the plasma edge as a representative value, and calculating VAV_A for the actual plasma composition, the low-D–low-Ne baseline case (■) corresponds to a Λ≈0.91 Wb3/C\Lambda\approx0.91\,\mathrm{Wb^3/C}.

Figure 8a shows the dependence of the maximum runaway current on pp for the baseline case. The details of how such a representative value the maximum runaway current is calculated are given in Appendix § 8 of [9]. The dependence is approximately parabolic, with a minimum for a Λ\Lambda profile slightly weighted towards the core. Despite the wide range of pp considered, IRE,maxI_{\mathrm{RE,max}} varies by only about 15%. This weak sensitivity arises because the adopted value of Λ0\Lambda_0 is sufficient to almost completely flatten the current density profile, even when Λ\Lambda varies radially.

Figure 8. Sensitivity of the maximum runaway current to (a) radial variations in Λ\Lambda, parametrized by pp, at fixed Λ0≈0.91 Wb3/C\Lambda_0 \approx0.91\,\mathrm{Wb^3/C}, and (b) the magnitude of a radially uniform Λ\Lambda. Both scans correspond to the low-D–low-Ne baseline case (■). Note the different vertical-axis ranges. Panel (c) illustrates the radial profiles of Λ\Lambda corresponding to different values of pp.

Figure 8b illustrates this saturation by varying the magnitude of a radially uniform Λ\Lambda. Most of the reduction in IRE,maxI_{\mathrm{RE,max}} occurs as Λ\Lambda increases through the range 0.01–0.1 Wb3/C\mathrm{Wb}^3/\mathrm{C}, with progressively weaker sensitivity at larger values. The baseline value, Λ≈0.91 Wb3/C\Lambda\approx0.91\,\mathrm{Wb}^3/\mathrm{C}, lies well within this regime of weak sensitivity.

References

  1. [1]Bakhtiari, M., Olynyk, G., Granetz, R., Whyte, D.G., Reinke, M.L., Zhurovich, K. & Izzo, V. 2011 Using mixed gases for massive gas injection disruption mitigation on Alcator C-Mod. Nuclear Fusion 51 (6), 063007.
  2. [2]Boozer, Allen H. 1986 Ohm’s law for mean magnetic fields. Journal of Plasma Physics 35 (1), 133–139.
  3. [3]Boozer, Allen H. 2017 Runaway electrons and ITER. Nuclear Fusion 57 (5), 056018.DOI
  4. [4]Boozer, Allen H. 2018 Pivotal issues on relativistic electrons in ITER. Nuclear Fusion 58 (3), 036006.DOI
  5. [5]Clauser, Cesar F., Sweeney, Ryan M., Granetz, Robert S., Ferraro, Nathaniel Mandrachia & Rea, Cristina 2025 Assessment of cold and hot vertical displacement events in ARC-like plasmas. In 67th Annual Meeting of the APS Division of Plasma Physics. Long Beach, California, abstract JO04.14.
  6. [6]Commaux, N, Baylor, L R, Combs, S K, EIDETIS, N W, Evans, T E, Foust, C R, Hollmann, E M, Humphreys, D A, Izzo, V A, James, A N, Jernigan, T C, Meitner, S J, Parks, P B, Wesley, J C & Yu, J H 2011 Novel rapid shutdown strategies for runaway electron suppression in DIII-D. Nuclear Fusion 51 (10), 103001.
  7. [7]Commaux, N., Shiraki, D., Baylor, L. R., Hollmann, E. M., Eidietis, N. W., Lasnier, C. J., Moyer, R. A., Jernigan, T. C., Meitner, S. J., Combs, S. K. & Foust, C. R. 2016 First demonstration of rapid shutdown using neon shattered pellet injection for thermal quench mitigation on DIII-D. Nuclear Fusion 56 (4), 046007.
  8. [8]Eidietis, N. W., Izzo, V. A., Commaux, N., Hollmann, E. M. & Shiraki, D. 2017 Poloidal radiation asymmetries during disruption mitigation by massive gas injection on the DIII-D tokamak. Physics of Plasmas 24 (10), 102504.
  9. [9]Ekmark, I., Hoppe, M., Fülöp, T., Jansson, P., Antonsson, L., Vallhagen, O. & Pusztai, I. 2024 Fluid and kinetic studies of tokamak disruptions using Bayesian optimization. Journal of Plasma Physics 90 (3), 905900306.
  10. [10]Hender, T C, Wesley, J C, Bielak, J, Bondeson, D, Boozer, A H, Buttery, R J, Garofalo, A, Goodman, T P, Granetz, R S, Gribov, Y, Gruber, O, Gryaznevich, M, Giruzzi, G, Günter, S, Hayashi, N, Helander, P, Hegna, C C, Howell, D F, Humphreys, D A, Huysmans, G T A, Hyatt, A W, Isayama, A, Jardin, S C, Kawano, Y, Kellman, A, Kessel, C, Koslowski, H R, Haye, R J La, Lazzaro, E, Liu, Y Q, Lukash, V, Manickam, J, Medvedev, S, Mertens, V, Mirnov, S V, Nakamura, Y, Navratil, G, Okabayashi, M, Ozeki, T, Pacagnella, R, Pautasso, G, Porcelli, F, Pustovitov, V D, Riccardo, V, Sato, M, Sauter, O, Schaffer, M J, Shimada, M, Sonato, P, Strait, E J, Sugihara, M, Takechi, M, Turnbull, A D, Westerhof, E, Whyte, D G, Yoshino, R, Zohm, H & et al. 2007 Chapter 3: MHD stability, operational limits and disruptions. Nuclear Fusion 47 (6), S128.
  11. [11]Herfindal, J. L., Shiraki, D., Baylor, L. R., Eidietis, N. W., Hollmann, E. M., Lasnier, C. J. & Moyer, R. A. 2019 Injection of multiple shattered pellets for disruption mitigation in DIII-D. Nuclear Fusion 59 (10), 106034.
  12. [12]Hollmann, E. M., Aleynikov, P. B., Fülöp, T., Humphreys, D. A., Izzo, V. A., Lehnen, M., Lukash, V. E., Papp, G., Pautasso, G., Saint-Laurent, F. & Snipes, J. A. 2015 Status of research toward the ITER disruption mitigation system. Physics of Plasmas 22 (2), 021802.DOI
  13. [13]Hoppe, Mathias, Embreus, Ola & Fülöp, Tünde 2021 DREAM: a fluid-kinetic framework for tokamak disruption runaway electron simulations. Computer Physics Communications 268, 108098.
  14. [14]Jachmich, S., Kreuzi, U., Lehnen, M., Baruzzo, M., Baylor, L.R., Carnevale, D., Craven, D., Eidietis, N.W., Ficker, O., Gebhart, T.E., Gerasimov, S., Herfindal, J.L., Hollmann, E., Huber, A., Lomas, P., Lovell, J., Manzanares, A., Maslov, M., Mlynar, J., Pautasso, G., Paz-Soldan, C., Peacock, A., Piron, L., Plyusnin, V., Reinke, M., Reux, C., Rimini, F., Sheikh, U., Shiraki, D., Silburn, S., Sweeney, R., Wilson, J., Carvalho, P. & JET Contributors 2021 Shattered pellet injection experiments at JET in support of the ITER disruption mitigation system design. Nuclear Fusion 62 (2), 026012.DOI
  15. [15]Kiramov, DI, Lehnen, M, Khayrutdinov, R & Lukash, V 2016 ITER disruption simulations with improved power balance in the halo region. In Proc. 43rd EPS Conf. Plasma Physics (Leuven, Belgium, July 2016), p. P4.
  16. [16]Lehnen, M., Gerasimov, S.N., Jachmich, S., Koslowski, H.R., Kreuzi, U., Matthews, G.F., Mlynar, J., Reux, C. & de Vries, P.C. 2015 Radiation asymmetries during the thermal quench of massive gas injection disruptions in JET. Nuclear Fusion 55 (12), 123027.DOI
  17. [17]Pautasso, G, Fable, E, Bernert, M, Dux, R, Fuchs, C, Giannone, L, Mccarthy, P J & Mlynek, A 2017 Disruption mitigation by injection of small quantities of noble gas in ASDEX Upgrade. Plasma Physics and Controlled Fusion 59, 014046.
  18. [18]Pautasso, G., Mlynek, A., Bernert, M., Mank, K., Herrmann, A., Dux, R., Müller, H.W., Scarabosio, A., Sertoli, M. & the ASDEX Upgrade Team 2015 Assimilation of impurities during massive gas injection in ASDEX Upgrade. Nuclear Fusion 55 (3), 033015.
  19. [19]Pusztai, I., Ekmark, I., Bergström, H., Halldestam, P., Jansson, P., Hoppe, M., Vallhagen, O. & Fülöp, T. 2023 Bayesian optimization of massive material injection for disruption mitigation in tokamaks. Journal of Plasma Physics 89 (2), 905890204.
  20. [20]Pusztai, István, Hoppe, Mathias & Vallhagen, Oskar 2022 Runaway dynamics in tokamak disruptions with current relaxation. Journal of Plasma Physics 88, 905880409.
  21. [21]Reux, C., Plyusnin, V., Alper, B., Alves, D., Bazylev, B., Belonohy, E., Brezinsek, S., Decker, J., Devaux, S., de Vries, P., Fil, A., Gerasimov, S., Lupelli, I., Jachmich, S., Khilkevitch, E.M., Kiptily, V., Koslowski, R., Kreuzi, U., Lehnen, M., Manzanares, A., Mlynár, J., Nardon, E., Nilsson, E., Riccardo, V., Saint-Laurent, F., Shevelev, A.E. & Sozzi, C. 2015 Runaway beam studies during disruptions at JET-ILW. Journal of Nuclear Materials 463, 143–149.DOI
  22. [22]Rosenbluth, M.N & Putvinski, S.V 1997 Theory for avalanche of runaway electrons in tokamaks. Nuclear Fusion 37 (10), 1355–1362.DOI
  23. [23]Sheikh, U., Decker, J., Hoppe, M., Pedrini, M., Sieglin, B., Simons, L., Cazabonne, J., Caloud, J., Cerovsky, J., Coda, S., Colandrea, C., Molin, A. Dal, Duval, B., Ficker, O., Griener, M., Papp, G., Pautasso, G., Paz-Soldan, C., Reux, C., Tomesova, E. & Wijkamp, T. 2024 Benign termination of runaway electron beams on ASDEX Upgrade and TCV. Plasma Physics and Controlled Fusion 66, 035003.DOI
  24. [24]Sheikh, U.A., Shiraki, D., Sweeney, R., Carvalho, P., Jachmich, S., Joffrin, E., Lehnen, M., Lovell, J., Nardon, E., Silburn, S. & JET Contributors 2021 Disruption thermal load mitigation with shattered pellet injection on the Joint European Torus (JET). Nuclear Fusion 61 (12), 126043.DOI
  25. [25]Shiraki, D., Commaux, N., Baylor, L. R., Eidietis, N. W., Hollmann, E. M., Lasnier, C. J. & Moyer, R. A. 2016 Thermal quench mitigation and current quench control by injection of mixed species shattered pellets in DIII-D. Physics of Plasmas 23 (6), 062516.DOI
  26. [26]Sweeney, R, Creely, A J, Doody, J, Fülöp, T, Garnier, D T, Granetz, R., Greenwald, M., Hesslow, L., Irby, J., Izzo, V. A., La Haye, R. J., Logan, N. C., Montes, K., Paz-Soldan, C., Rea, C., Tinguely, R. A., Vallhagen, O. & Zhu, J. 2020 MHD stability and disruptions in the SPARC tokamak. Journal of Plasma Physics 86 (5), 865860507.DOI
  27. [27]Sweeney, Ryan, Riccardo, Valeria, Braun, Anson, Clauser, Cesar, Creely, Alexander J., Eich, Thomas, Ekmark, Ida, Feyrer, Abigail, Hansen, Christopher, Hillesheim, Jon C. & et al. 2026 ARC disruption physics and strategy. Journal of Plasma Physics 92 (3), E68.DOI
  28. [28]Vallhagen, O., Hanebring, L., Fülöp, T., Hoppe, M., Votta, L. & Pusztai, I. 2025 Reduced modelling of scrape-off losses of runaway electrons during tokamak disruptions. Journal of Plasma Physics 91 (3), E78.
  29. [29]Votta, L., Artola, F.J., Nardon, E., Vallhagen, O. & Hoppe, M. 2026 Runaway electron generation in ITER mitigated disruptions with improved physics models. Nuclear Fusion 66 (10), 106048.DOI
  30. [30]Wang, C., Nardon, E., Artola, F.J., Bandaru, V., Hoelzl, M. & the JOREK team 2025 The effect of vertical displacements on the runaway electron avalanche in ITER mitigated disruptions. Nuclear Fusion 65 (1), 016012.DOI
  31. [31]Zhang, Yanzeng, Li, Jun & Tang, Xian-Zhu 2023 Cooling flow regime of a plasma thermal quench. EPL 141 (5), 54002.DOI

Paper details

vlnc:W_d86msry6
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents