Edge-localized mode mitigation enabled by active control of pedestal density gradient with new EAST tokamak divertor

One of the key issues currently in tokamak fusion energy research is the erosion of plasma-facing materials due to excessive transient heat load from large edge-localized modes (ELMs), a violent instability at the plasma edge. ELMs are characterized by quasiperiodic relaxations of an edge transport barrier, known as “pedestal,” accompanying high-confinement-mode (H-mode) plasmas.¹ Significant efforts have been made on the mitigation of ELMs over the past decades, including the use of active control methods such as resonant magnetic perturbation (RMP),² impurity injection,^3,4 triggering frequent smaller ELMs (e.g., pellet pacing⁵), and natural small/no ELM regimes such as quiescent H-mode (QH-mode),⁶ type-II ELM,⁷ and grassy ELM.^8,9 However, it is still an open issue to find a robust method suitable for future fusion reactors.

Recent studies have indicated that the pedestal density profile is a key for access to small ELM regimes^9–14 and even the suppression of ELMs.^15–17 The grassy ELM in EAST^9,13 and DIII-D¹⁴ is characterized by a low pedestal density gradient and wider pedestal width. The small ELM regime in JET achieved at low edge safety factor q₉₅ and low pedestal collisionality has a low pedestal density and similar characteristics of low-density gradient and wide pedestal width.^11,12 ASDEX-Upgrade (AUG)^18–20 and Tokamak à Configuration Variable (TCV)²¹ have recently obtained a high confinement small ELM regime named quasi-continuous exhaust (QCE) regime at high density and strong shaping. The studies suggest that the high separatrix density is critical for the achievement of QCE regime in AUG,^22,23 and the ballooning mode near the separatrix is considered to drive the small ELMs. In modeling, the effect of separatrix density and density gradient on the ELM amplitude has also been investigated with BOUT++ simulations.^24,25

In current low opacity devices, the pedestal density structure is strongly influenced by the neutral fueling.²⁶ It has been observed that pedestal density profile shifts radially outward with increased gas fueling in AUG,²⁷ JET, and DIII-D.²⁸ This outward shift of pedestal density profile has an impact on the pedestal stability, and thus, reduces the pedestal top pressure, which is related to the confinement degradation observed in the case of high-field side high-density region (HFSHD) in AUG.²⁷ The recycled neutrals as well as upstream density would be affected by the divertor geometry. In AUG, a higher separatrix density has been observed under the closed divertor operation.^29,30 DIII-D has systematically compared the pedestal structures in the open lower and closed upper divertor operations,^31–33 showing that the pedestal top density is much lower for closed divertor configuration in contrast to the open one. JET experiments with different strike point positions have also demonstrated that the divertor geometry can affect the neutral recycling and pedestal top density, and even the ELM activity.^34,35 In addition, the effect of divertor condition on the recycled neutrals and fueling has also been investigated with modeling in previous publications.^36–41 

EAST tokamak has been equipped with a new right-angled lower divertor in 2021 campaign, which allows the outer strike point to be positioned on either the vertical or horizontal target plate.⁴² It has been observed in a recent study that the shift of strike point along the horizontal target has an influence on the separatrix density, and thus, affects the ELM behavior.⁴³ In this paper, we systematically study the effects of strike point location on the pedestal structure and ELM behavior with a series of dedicated experiments in the 2021–2024 campaigns. It is found that the pedestal top density and pedestal density gradient change evidently with the strike point position changing from the vertical to the horizontal target, in addition to the separatrix density. As a result, the ELM mitigation has been achieved and could be well reproduced in a broad parameter space. The rest of this paper is organized as follows. Section II describes the typical ELM mitigation experiments through changing the outer strike point position. Section III presents the parameter space applicable to this method. The mechanism behind the change of pedestal density profile is explored in Sec. IV. In Sec. V, the influences of the change in triangularity are discussed. Finally, a summary is presented in Sec. VI.

EAST is a medium-sized superconducting tokamak (major radius R₀ ∼1.9 m, minor radius a ∼0.45 m) with flexible poloidal field (PF) control system to achieve different plasma shapes and upper single null (USN), lower single null (LSN), and double null (DN) divertor configurations. EAST has been installed with an ITER-like vertical target upper tungsten divertor and molybdenum first wall. From the 2021 campaign, the lower divertor has been upgraded from graphite divertor to an actively cooled tungsten divertor with the power handling capability increasing to ∼10 MW m⁻² for the long-pulse H-mode operation.⁴² The new lower divertor is featured by a right-angled closed corner at the outer divertor constituted by the vertical and horizontal target plates, as shown in Fig. 1(d). Such a divertor structure provides an opportunity to study the impact of divertor condition on the pedestal and ELM behavior.

Figure 1 shows the typical ELM mitigation experiments by changing the lower outer strike point from vertical to horizontal target with similar plasma parameters. These experiments are operated at plasma current I_p = 500 kA, toroidal magnetic field B_t ∼ 2.42 T in favorable B_t direction (i.e., the B × ∇B drift toward the primary X-point), edge safety factor q₉₅ ∼ 5.6, poloidal beta β_p ∼ 1.4, under lithium-coated wall, and LSN configuration with the radial distance between the flux surfaces (separatrix) through the upper and lower X-points at the outer midplane, dR_sep ∼ 2.3 cm. The heating power includes 1.4 MW 4.6 GHz lower hybrid current drive (LHCD), 1 MW electron cyclotron resonance heating (ECRH), and 1 MW neutral beam injection (NBI), totally P_total ∼ 3.4 MW. The central-line-averaged density n_el, which is feedback controlled and measured by the POlarimeter-INTerferometer (POINT) diagnostic, is n_el ∼4.2 × 10¹⁹ m⁻³ (Greenwald density fraction f_GW ∼ 0.5) for the three shots. The stored energy W_MHD is ∼210 kJ, and the energy confinement is maintained with H_98y2 ∼1. In the experiments, the movement of strike point from vertical to horizontal target is achieved through the feedback control of lower X-point position in the plasma control system (PCS), while the other point positions of the last closed flux surface (LCFS) controlled by PCS are kept fixed. The separatrix and strike point shown in Fig. 1(d) are obtained from the magnetic equilibrium reconstruction by the EFIT code within the constraints of magnetic measurements along the vacuum vessel walls. The positions of outer strike point from EFIT reconstruction have been verified by the divertor probe measurements. The results show that, when the outer strike point is on the vertical target with the major radius of the lower X-point R_x ∼ 164 cm, large ELMs can be observed in shot #103 751. In contrast, when the strike point is on the horizontal target with R_x ∼ 158 cm in shot #103 745, the ELM amplitude appears to be significantly decreased and is even smaller as the strike point is on the horizontal target closer to the dome with R_x ∼156 cm in shot #103 748. Such significant reduction in ELM amplitude is indicated by the reduction of spikes on the magnetic signal measured by a Mirnov coil at the inner side midplane in Fig. 1(b) and the reduction of perturbations on the line-integrated density n_eL,edge measured by the outermost horizontal chord (Z = 42.5 cm) of the POINT diagnostic, which passes through the pedestal in Fig. 1(c). The D_α emission measurement is not shown here, as it is affected by the change of magnetic configuration in the divertor region that not suitable for assessing the change of ELM amplitude in this work. With the mitigation of large ELMs, the ELM frequency f_ELM increases from ∼120 to ∼300 Hz and further to ∼500 Hz in the three shots. Note that the ELM frequency of small ELMs in shot #103 748 is in the typical ELM frequency range of EAST small grassy ELM regime.⁹ 

The typical pedestal profiles just prior to ELM crashes are shown in Fig. 2. The electron density n_e and temperature T_e profiles are measured by microwave reflectometry and Thomson scattering, respectively, and fitted with a modified hyperbolic tangent (mtanh) function.⁴⁴ There is a significant change in n_e but little change in T_e profile. When the strike point is on vertical target, the pedestal density profile exhibits a steep gradient, a relatively higher pedestal top density, and lower separatrix density with a low-density ratio between the pedestal foot (separatrix) and top, n_e,sep/n_e,ped ∼0.38. In contrast, when the strike point is located on the middle of horizontal target, a flat density profile is obtained with a much higher density at the separatrix and a higher density ratio n_e,sep/n_e,ped ∼ 0.53. In particular, when the strike point on horizontal target is further away from the corner, pedestal density gradient is even further decreased with a high n_e,sep/n_e,ped ∼0.55. The change of separatrix density can also be confirmed by the Langmuir probe measurement in the Scrape-Off Layer (SOL) at the outer midplane, as shown in Fig. 3, which shows that the ion saturation current and n_e profiles in horizontal case are much higher than that in vertical case. Note that the reduction of separatrix density has also been observed with the strike point on horizontal target moving close to the corner in other EAST experiments.⁴³ As a result, the pressure gradient reduces, and the bootstrap current density is significantly lower in the pedestal. These profiles have also been used for BOUT++ simulation to study the ELM dynamic.²⁵ 

With the reduction of pedestal density gradient, the pedestal widths of density and pressure profiles are extended. As illustrated in Fig. 2(e), the pressure pedestal width for vertical case is in agreement with the width scaling of EPED1 model,⁴⁵ i.e., pedestal width $Δ=0.076βθ,ped0.5$⁠, while it is dramatically wider in the two horizontal cases. It is worth mentioning that the relationship between the occurrence of small ELMs and low pedestal density gradient in these shots is consistent with the previous results in EAST under USN configuration^9,10 and DIII-D result of natural grassy ELMs.¹⁴ 

For the wider density pedestal width in the horizontal case, a possible explanation is discussed here. As described in Ref. 45, the pedestal width Δ is proportional to the pedestal poloidal beta β_p,ped divided by the pressure gradient of linear kinetic ballooning mode (KBM) threshold ⟨α_c⟩, i.e., Δ ∝ β_p,ped/⟨α_c⟩, and the ⟨α_c⟩ increases with decreasing magnetic shear. In our case, as shown in Fig. 2(f), the horizontal case has a larger magnetic shear s in the pedestal, and thus, lower ⟨α_c⟩. Therefore, a wider pedestal width might be allowed.

As mentioned above, the change in nonlinear ELM dynamic with strike point position changing has been carefully considered in Ref. 25. Therefore, we only focus on the linear peeling-ballooning (PB) stability analysis calculated with the eigenvalue ELITE code based on the ideal magnetohydrodynamic (MHD) model.⁴⁶ The kinetic equilibrium is generated by the EFIT code through solving the Grad–Shafranov equation,⁴⁷ within the constraints of the pressure, current profiles, and the magnetic measurements along the vacuum vessel walls. Then, the edge current and pedestal pressure gradient are scanned independently around the experimental values to generate a set of model equilibria by the VARYPED code, while keeping the total stored energy, total current, and plasma shape unchanged through adjusting the core profiles to incorporate the pedestal change. ELITE then calculates the ideal PBM instability and obtains the growth rates of different toroidal mode numbers for each equilibrium generated. The calculation region in ELITE code is generally ψ_N ≤ 0.995. The calculation result is shown in Fig. 4, and the vertical axis is half of the sum of peak current density and separatrix current density, normalized by the volume averaged current density ⟨j⟩, i.e., (j_max+j_sep)/2/⟨j⟩. The horizontal axis is the peak value of normalized pedestal pressure gradient α, defined by $α=−2V′2π2V2π2R12μ0p′$⁠, where V is the plasma volume enclosed by the flux surface, p is the pressure, and the prime represents the derivative with respect to the poloidal flux ψ. The stability boundary is defined as the value when γ/(ω_*i/2) is equal to 1, where γ is the growth rate of the most unstable mode and ω_∗i is the ion diamagnetic frequency. The ELITE result indicates that the operational point for the large ELM in vertical case lies on the peeling instability boundary and close to the corner of the PB instability boundary. In contrast, for the small ELM in the two horizontal cases, the edge bootstrap current is much lower, and therefore the operational points are away from the peeling boundary and stay in the stable region with lower pedestal pressure gradient. This is similar to the result of DIII-D natural grassy ELM.¹⁴ The ballooning instability boundary shifts inward evidently due to the lower pressure gradient and weaker diamagnetic stabilization effect with the reduction of pedestal density gradient. The linear result indicates that the pedestal of the horizontal cases is much more stable than that of the vertical case.

In the 2021–2024 EAST campaigns, a series of dedicated experiments have been performed to verify the repeatability of this ELM mitigation solution and explore the parameter space applicable to this method. These experimental results demonstrate that this ELM control method can be well reproduced in a broad range of q₉₅, heating power P_total, and different B_t directions.

It has been found that q₉₅ is one of the critical parameters for ELM activity in EAST.⁴⁸ Our experimental results suggest that the ELM mitigation enabled by the change of strike point location appears to be insensitive to the change of q₉₅. Figures 5(a)–5(c) show the two discharges (#114 678 and #114 671) with the strike point on vertical and horizontal targets at q₉₅ ∼ 4.7. The other plasma parameters are I_p = 500 kA, B_t ∼ 2.2 T, n_el ∼ 4 × 10¹⁹ m⁻³, f_GW ∼ 0.5, β_p ∼ 1.1, W_MHD ∼ 210 kJ, unfavorable B_t, and LSN configuration. The large ELMs are evidently mitigated, and simultaneously the ELM frequency increases from ∼50 to ∼160 Hz with the strike point position changing from vertical to horizontal target. At higher q₉₅ ∼ 7.1 with smaller ELMs, the ELM mitigation effect has also been observed, as shown in Figs. 5(d)–5(f). The two discharges (#100 447 and #100 443) are operated with I_p = 400 kA, B_t ∼ 2.44 T, β_p ∼ 1.5, W_MHD ∼ 145 kJ, favorable B_t and LSN configuration. It also demonstrates that the ELM mitigation can be achieved in both favorable and unfavorable B_t directions.

We have also examined the sensitivity of the mitigation effect to P_total. Figure 6 shows two groups of experiments with I_p = 450 kA, B_t ∼ 2.4 T, q₉₅ ∼ 6, under unfavorable B_t and LSN divertor configuration, but different P_total. The left shots are operated with P_total ∼ 3.7 MW (1.3 MW LHCD, 1.2 MW ECRH and 1.2 MW NBI), and the right shots have higher P_total ∼ 5 MW with 1.3 MW NBI more. The results show that the ELM mitigation can be achieved at different P_total, suggesting it is not sensitive to the change of P_total.

In the series of experiments, we have found that this ELM mitigation effect appears to depend on higher plasma density. Figure 7 shows two groups of experiments with different plasma densities. The left shots are operated at a higher density n_el ∼ 4.5 × 10¹⁹ m⁻³ (f_GW ∼ 0.5), with I_p = 550 kA, B_t ∼ 2.42 T, q₉₅ ∼ 5.3, favorable B_t, and LSN configuration. The right shots have lower density n_el ∼3 × 10¹⁹ m⁻³ (f_GW ∼ 0.36) with I_p = 500 kA, B_t ∼ 2.2 T, q₉₅ ∼ 4.8, unfavorable B_t, and LSN configuration. At the higher density, the large ELMs are evidently mitigated with f_ELM increasing from ∼90 to ∼400 Hz, while there is weak/no ELM mitigation at lower density.

In order to find out the possible reason for the weak/no mitigation at lower density, we have compared the change of pedestal density profile, as it is the key factor for ELM mitigation in this work. As shown in Fig. 8, the change of pedestal density profile in the weak/no mitigation case is much small with a small variation in n_e,sep/n_e,ped. In contrast, for the ELM mitigation case, the density profile change is significantly larger with much higher n_e,ped and lower n_e,sep when the strike point is on vertical target. Note that, similar dependence of upstream density change on the main plasma density has also been observed in the related experiments of L-mode plasma.

Statistical result further verifies the dependences of mitigation effect on the main plasma density. In the 2021–2024 campaigns, we have carried out 18 groups of experiments with strike point position changing in a wide parameter space under LSN configuration and lithium-coated wall. The ELM mitigation result of these experiments is shown in Fig. 9. As the amplitude of some small ELMs in horizontal cases cannot be well evaluated from the measurements of stored energy or edge density n_eL,edge,⁴⁸ the ratio of ELM frequency in horizontal case to that in vertical case, i.e., f_{ELM,Horizontal}/f_ELM,Vertical, is used to assess the change in ELM behavior. It shows that the ratios of ELM frequency are higher than 1.5 in the mitigation cases, while equal to or less than 1 in the weak/no mitigation cases. It clearly shows that the weak/no mitigation cases occur in relatively low plasma density compared to the mitigation cases, as shown in Figs. 9(c) and 9(d). In addition, the statistical result of the change of pedestal density profile between horizontal and vertical target cases is also illustrated in Fig. 10. Here, the change of pedestal density profile is represented by the variation of n_e,sep/n_e,ped, i.e. (n_e,sep/n_e,ped)_H − (n_e,sep/n_e,ped)_V. It shows that the change of pedestal density profile is systematically smaller at lower plasma density, consistent with the density profile example in Fig. 8. The physical mechanism for this phenomenon would be further explored in future work. Note that, the experiment with lower density has not been performed under favorable B_t, due to higher particle confinement in the field direction.

To summarize, the new ELM mitigation solution is effective and reproducible in a broad parameter space of q₉₅ = 4.7–7.1, P_total = 2.3–5 MW, $νe,ped*$ = 1–6, and under both favorable and unfavorable B_t directions. Stronger mitigation effect can be achieved at relatively higher plasma density of n_el > 3.5 × 10¹⁹ m⁻³ and f_GW > 0.47, as summarized in Table I.

It is worth mentioning that the aforementioned experiments are performed with lithium-coated wall in EAST. To support the ITER operation with full metal wall, EAST experiments have been operated under boronized metal wall in the 2023–2024 campaigns. During the campaigns, dedicated experiments have been performed to further examine the repeatability of this ELM mitigation under boronized wall. As shown in Fig. 11, the plasma parameters of the two shots are I_p = 400 kA, B_t ∼ 2.4 T, q₉₅ ∼ 6, β_p ∼ 1.3, W_MHD ∼ 150 kJ, under favorable B_t direction and LSN configuration, heating power includes 1 MW LHCD, 1.6 MW ECRH, and 0.5 MW NBI. The D_α emission here is used to show the change of ELM frequency. With the strike point position changing from the vertical to the horizontal target, the pedestal density gradient reduces with lower n_e,ped and higher n_e,sep, and thus, the ELMs are mitigated. This result demonstrates that the ELM control method is effective under boronized metal wall and does not rely on the lithium coating.

Generally, the pedestal density profile could be regulated by the turbulent transport, ELM-induced particle flux and gas fueling. Our dedicated experiments indicate that the change of pedestal density profile, especially the increased n_e,ped in vertical target case, is very reproducible. In the study, we mainly focus on the effect of divertor geometry on the pedestal profile. To isolate the effect of divertor geometry and exclude the influences of other factors, the SOLPS-ITER⁴⁹ simulation has been performed with all parameter settings being the same except for different strike point positions. In the simulation, the divertor magnetic configurations of shots #103 748, #103 745 and #103 751 are used as input. Only deuterium species is included in the modeling, and impurities are not considered. The D₂ gas is puffed from the outer midplane with the same puffing rate of 1.0 × 10²² s⁻¹. 2.9 MW power enters the simulation region from the core-edge interface (CEI, i.e., the innermost simulation boundary), and this power is equally divided by ions and electrons. The ionization flux at the CEI is fixed. At the SOL and private flux region (PFR) boundaries, the leakage boundary condition is applied. At the target boundary, sheath boundary conditions for electron and ion temperatures, and ion density are used. The recycling rate at the first wall and divertor target plates are assumed as 100%. The albedos at the pumping surfaces are set to 0.94 for the cryopump with a pumping speed of 75 m³·s⁻¹, and 0.995 for the pumping port-end with the pumping speed of 12 m³·s⁻¹, respectively. The particle diffusion coefficient D_n and heat conductivity of ion and electron χ_i,e used in simulation are set to be the same for different strike point positions, as shown in Fig. 12. Although the drift effects have an impact on the pedestal structure,³⁶ the three shots are operated with the same B_t direction and the change of pedestal density with strike point position changing can be observed in both B_t directions, implying that the drift effects are not the basic factor for the pedestal density change. Therefore, the drift effects are not considered in the simulation, and it will be further studied in future work.

Figure 13 shows the distributions of particle ionization source S_D+ for different strike point locations. It indicates that the particle ionization source concentrates in the vicinity of the horizontal target when the strike point is on the horizontal target [Figs. 13(a) and 13(b)]. In contrast, a much stronger ionization source appears in the vicinity of the X-point area, i.e., low field side (LFS) region and private flux region, especially inside the separatrix, when the strike point is on the vertical target [Fig. 13(c)]. The simulation results show that, in the vertical case, the recycled D atom flux Φ_D,in crossing separatrix into the main plasma in the LFS region near the X-point is more than twice as high as the horizontal case. Specifically, Φ_D,in is 0.349, 0.603, and 1.58 × 10²¹ s⁻¹ for shots #103 748, #103 745, and #103 751, respectively. Therefore, the averaged S_D+ profile in the LFS region near the X-point [marked by white dashed line in Fig. 13(a)] for horizontal cases is much lower compared to the vertical case, as show in Fig. 13(d). The lower S_D+ in the pedestal (i.e., pedestal fueling) could contribute to the formation of a lower pedestal density gradient with lower pedestal top density in the horizontal cases, as shown in Fig. 13(e). The simulated SOL density in the horizontal cases is higher, consistent with the experimental result. In addition, the simulated pedestal T_e in the horizontal cases appears to be higher, different from the experimental observation, which could be since that the change of thermal transport is not considered in the simulation.

The possible interpretations for the different plasma fueling are discussed here. When the strike point is on the horizontal target, most of the particles from the upstream SOL region flow into the outer divertor slot and are then trapped in the closed corner region. Therefore, much less recycled neutral particles from the lower divertor can penetrate into the pedestal through the X-point area, leading to a low pedestal fueling. The recycled particles from the lower divertor would transport back to the upstream SOL region along the same flux tube, increasing the SOL density. In contrast, when the strike point is on the vertical target, the particles are reflected by the vertical target plate toward the private flux region. As the n_e and T_e in private flux region are much lower than that in the SOL, the neutral particles cannot be fully re-ionized and tend to diffuse into the pedestal through the X-point area, enhancing the pedestal fueling. In addition, the parallel connection length from X-point region to divertor target may also play a role in plasma fueling. The parallel connection length can be calculated using the field line tracing method. In the cylindrical coordinate system (R, φ, Z), the equation of a field line is given by dR/Rdφ = B_R/B_φ and dZ/Rdφ = B_Z/Bφ, where the B_R, B_Z, and B_φ are the three components of magnetic field. With the integration of the field line equation, the parallel connection length from X-point to the target is obtained, showing ∼20 m in #103 748 and ∼16 m in #103 751. Since the connection length from X-point to the horizontal target is longer than that to the vertical target, the recycled neutrals from horizontal target have to travel a longer distance to reach the X-point area. This could contribute to a lower fueling in the low field side region near the X-point in the horizontal target case.

During the change of strike point position in Fig. 1, we note that the lower triangularity δ_low increases from ∼0.49 to ∼0.64, and accordingly, the average triangularity δ_ave [i.e., δ_ave = (δ_u+δ_low)/2] increases from ∼0.38 to ∼0.49. Previous publications have reported that the triangularity has an effect on the ELMs, pedestal stability and confinement.^48,50–53 The higher triangularity appears to facilitate an increase in ELM frequency,^48,50 and the theory suggests that increased triangularity has a beneficial effect on the pedestal peeling–ballooning stability,⁵¹ and thus, improve the pedestal confinement.^52,53 In this paper, to examine the influence of triangularity change in our cases, dedicated experiments and simulations have been performed.

In experiments, we increased the δ_ave from ∼0.4 to ∼0.43 through increasing the inner gap between the separatrix and the first wall while keeping the X-point location and outer gap fixed. The result shows that the ELM size remains almost unchanged with such a degree of triangularity change. However larger change of triangularity could not be achieved in this way.

To further verify this, we have performed the simulations with the equilibria reconstructed with the same profile but different triangularities. As shown in Figs. 14(a) and 14(b), the result shows that the pedestal stability changes little with different triangularities in these plasma conditions. Furthermore, linear and nonlinear simulations of the equilibria reconstructed with the same triangularity but different profiles are also performed [Figs. 14(c) and 14(d)], showing a significant change in pedestal stability. These results indicate that the ELM mitigation is largely due to the change in the pedestal profiles instead of the triangularity in our cases.

The possible role of edge turbulence on the pedestal profile is also examined. In EAST, the electrostatic fluctuation named edge coherent mode (ECM), commonly observed in experiments, is located in the pedestal steep gradient region and can drive a considerable outward particle flux across the separatrix.^54,55 Fig. 15 shows the ECM intensity measured by electron cyclotron emission (ECE) system. It suggests that the pedestal ECM is much strong in the shot #103 751 for vertical target case, and almost disappears in the horizontal target case, which could be due to the reduction of pedestal pressure gradient. This result indicates that the ECM-driven transport is not responsible for the reduced density gradients in the case. In addition, the microtearing mode (MTM) may also drive a transport that contributes to the evolution of pedestal profile.⁵⁶ However, it is difficult to identify the MTM in experiments due to the lack of a diagnostic system to directly measuring the small magnetic island structure in the narrow pedestal region with high spatial and temporal resolution. Previous EAST studies have reported a high-frequency electromagnetic mode (HFM) that has some similar characteristics with MTM,^57,58 but the nature of the HFM is not clear. In this work, the HFM has not been observed in the three typical shots. The effect of HFM or MTM on the pedestal profile and ELM behavior should be further studies in future work.

It is worth pointing out that changing the X-point location would have many other effects besides the shift of strike point position, such as the plasma shape, magnetic shear, turbulence, pumping, and edge neutral pressure. It cannot be ruled out that other effects may play a role in the ELM behavior. We have tried to assess the other effects or found some regular change in the series of dedicated experiments during the 2021–2024 campaigns. It is found that the most evident and very reproducible change is the variations in the ELM behavior and the pedestal density profile. For the other effects, a consistent change has not been found in the experiments to draw a solid conclusion, or some other effects have a minor contribution to the ELM mitigation.

Recently, EAST has developed a new tangential-viewing visible camera system with high temporal (up to 36 kHz) and spatial (∼1 mm in the divertor region) resolutions, in which the viewing area covering the lower divertor region.⁵⁹ The measurement of the camera could provide an additional benchmark for SOLPS simulation and could also be used to observe the X-point with an appropriate impurity line emission which has the strongest radiation in the X-point area. With this visible camera, the neutral recycling related issues will be further studied in future work.

In addition, the TS and reciprocating Langmuir probe measurements show tiny change in edge T_e profile, despite an evident variation in n_e profile. A possible explanation is proposed that the divertor recycling and pedestal fueling only affect the pedestal n_e profile, while the pedestal T_e profile could be mainly regulated by the thermal transport. Actually, the decoupling of pedestal n_e and T_e profiles has also been observed in previous research.^60–62 

In summary, the mitigation of large ELMs can be achieved by actively reducing the pedestal density gradient through changing the strike point position from the vertical to the horizontal target, with the assistance of a newly developed lower tungsten divertor on EAST. In the dedicated experiments, a flat pedestal density profile with larger pedestal width, lower pedestal top density, higher separatrix density, and thus, high-density ratio, n_e,sep/n_e,ped is achieved with the strike point being on the horizontal target in contrast to that on the vertical target. As a result, low-pressure gradient and low bootstrap current are obtained in the pedestal. Linear simulation indicates that the pedestal in the horizontal case is much more stable than that in the vertical case. These results provide a new solution for ELM mitigation. EAST results indicate that this ELM control solution is highly reproducible in a broad space of q₉₅ = 4.7–7.1, P_total = 2.3–5 MW, $νe,ped*$ = 1–6, under both favorable and unfavorable B_t directions. Stronger mitigation effect can be observed at higher density of n_el > 3.5 × 10¹⁹ m⁻³ and f_GW > 0.47, as the change of pedestal density profile is systematically larger at high plasma density. Moreover, the mitigation effect can be achieved under both lithium-coated and boronized metal wall. SOLPS-ITER simulation results indicate that the particle ionization source in the LFS region near the X-point and private flux region is lower in the horizontal case that could contribute to the formation of a flattened pedestal density.

For the shift of strike point along the horizontal target, the ELM behavior change is relatively small, and a solid conclusion cannot be drawn due to lack of sufficient experiments. Only several typical discharges show that the plasma with strike point on the horizontal target close to the dome has higher separatrix density and smaller ELM size. It cannot yet exclude the possibility that there exists a best strike point position on the horizontal target for ELM mitigation, which needs more experimental data in future work.

Similar experiments with strike point position changing have also been performed in JET to study the impact of divertor geometry on plasma confinement.^34,35 In JET, a confinement improvement was observed with higher pedestal pressure and steeper core pressure gradient when the outer strike point is on the divertor corner near the pump duct entrance, compared to that on the vertical or horizontal target. In EAST, there is significant change in the edge density profile and ELM behavior. With a moderate increase in the pedestal top density in vertical case, only a tiny change is observed in the core plasma confinement. When the pedestal top density is much different, the core plasma confinement would be changed.

Although the ELM mitigation method is repeatable in EAST, for the application in future fusion reactors such as ITER, there are still some disadvantages or gaps can be foreseen toward the ITER-relevant conditions as follows:

1. Heat flux load: In this method, the strike point needs to be positioned on the horizontal target close to the dome, which however cannot withstand too high heat load.⁶³ This would pose a risk of material erosion.

2. q₉₅ window: Currently, EAST has observed the ELM mitigation effect at q₉₅ ≥ 4.7. However, at low q₉₅ space, the mitigation effect is not strong enough to achieve high-frequency small ELMs. It may decrease the applicability of this method to the low q₉₅ operation in future reactors. Nevertheless, this method could be combined with other ELM control methods like impurity injection to achieve the small ELM regime under low q₉₅ operation in future reactors.

3. Pedestal collisionality: Although the ELM mitigation is achieved under relatively high pedestal collisionality $νe,ped*>1$⁠, the small ELM regime in DIII-D¹⁴ and JET¹¹ are obtained at low collisionality (⁠ $νe,ped* ∼ 0.15$ in DIII-D, $νe,ped* ∼ 0.1$ in JET) with a flat pedestal density. It suggests that the collisionality condition would not be the critical obstacle to small ELM regime in future reactors.

4. High opacity: As future reactors will be opaque to neutral particles, the effect of neutrals on the pedestal density profile would be strongly decreased.²⁶ At such high opacity, the effectivity of this ELM control method needs to be further verified.

The authors wish to acknowledge Professor Chaofeng Sang, Dr. Yaowei Yu, and Dr. Xavier Bonnin, for useful discussions. We thank the staff members at EAST (https://cstr.cn/31130.02.EAST) for providing technical support and assistance in data collection and analysis. This work was supported by the National Magnetic Confinement Fusion Energy R&D Program under Grant Nos. 2019YFE03030000, 2022YFE03020004, 2019YFE03080500, and 2022YFE03060004; the HFIPS Director's Fund, Grant No. YZJJ2023QN22; and the National Natural Science Foundation of China under Grant Nos. 12305255, U24A20342, U19A20113, and 12005257.

The authors have no conflicts to disclose.

X. Lin: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – original draft (equal); Writing – review & editing (equal). Q. Q. Yang: Conceptualization (equal); Methodology (equal); Writing – review & editing (equal). G. S. Xu: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Supervision (equal); Validation (equal); Writing – original draft (equal). G. Z. Jia: Software (equal). C. Zhang: Formal analysis (equal); Software (equal). Y. F. Wang: Software (equal). N. M. Li: Software (equal). N. Yan: Data curation (equal). R. Chen: Supervision (equal). X. Q. Xu: Software (equal). H. Y. Guo: Writing – review & editing (equal). L. Wang: Data curation (equal); Supervision (equal). S. C. Liu: Data curation (equal). Q. Zang: Data curation (equal). T. Zhang: Data curation (equal). F. B. Zhong: Data curation (equal). Y. F. Jin: Data curation (equal).

The data that support the findings of this study are available from the corresponding authors upon reasonable request.