Impacts of recycling impurity on the background ELM behavior during repetitive radiative divertor experiments in EAST

A significant change in the background ELM behavior prior to neon (Ne) seeding has been observed in a series of repetitive radiative divertor experiments in EAST. With similar operational parameters, the ELM behavior before Ne seeding changes from large to mixed ELMs, and finally evolves to pure small ELMs. Meanwhile, a significant increase in both Ne and high-Z impurity (W and Mo) emissions has been observed in the bulk plasma, suggesting the retention and recycling of Ne impurity from the wall surface. Experimental results show that the variation in background ELM behavior is highly correlated with the occurrence of high-Z impurity accumulation. The increased accumulation of high-Z impurities leads to a lower electron temperature both in the plasma core and edge, accompanied by a higher and more peaked electron density in the plasma core. Pedestal linear stability analysis reveals that the decreased pedestal electron temperature and thus the lower pressure gradient and lower edge current density are the primary reason for the change of background ELM behavior. The concentration of recycling Ne in the bulk plasma is estimated to be ∼1% in the discharges with pure small ELMs.

(Some figures may appear in colour only in the online journal) * Authors to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Introduction
The control of excessively high stationary power load on the plasma facing components (PFCs) is one of the most critical issues for future fusion devices, such as International Thermonuclear Experimental Reactor (ITER) [1]. The peak heat flux on the divertor targets for steady-state power exhaust should be kept below 15 MW m −2 in ITER [2]. Aiming for it, the radiative divertor scenario with impurity seeding is proposed to be used in ITER as an effective method to reduce the stationary power load [1,3]. However, experiments suggest that the impurity particles (such as nitrogen, neon, helium, etc) could be retained on the PFC surface through the interactions with the wall material, and released back to the main plasma [4][5][6]. TEXTOR, ASDEX Upgrade and JET devices have made significant efforts on the investigation of nitrogen (N) retention in the past decades, finding that nitrogen impurity can be retained on the tungsten (W) material and gradually accumulate with experiments [4,7,8]. The saturation of retention nitrogen impurity has been observed in ASDEX Upgrade [9]. The main mechanisms for nitrogen retention in nitrogen-seeded experiments involve the implantation into wall material, co-deposition with other impurities and the formation of ammonia [10]. The measurements in TEXTOR and ASDEX Upgrade suggest that the nitrogen retention even exceeds 10% of the injected impurity gas [7], and JET experiments show that the nitrogen retention and recycling could strongly affect the radiative divertor operation and increase the consumption of impurity gas during the power load feedback control experiments [11]. Ion cyclotron wall conditioning method can remove the retention nitrogen but part of the nitrogen is transported to other positions of the wall [12]. For the helium (He) retention, ASDEX Upgrade result indicates that W material surface has a much higher retention of He impurity compared to the Carbon (C) material [5]. In addition, the retention of neon (Ne) impurity in the graphite has also been reported in TEXTOR [4,6].
Besides the stationary power load control, impurity also has a strong impact on the edge-localized mode (ELM) behavior, such as ELM mitigation, suppression and even increasing the ELM size [13][14][15][16][17][18]. Extensive researches have been performed for the physics mechanisms behind the effects of impurities on ELMs. It is found that impurity can modify the pedestal density and temperature profiles, and thus pedestal pressure and edge bootstrap current through changing turbulent transport [15,19,20], recycling [13,14], radiative cooling effect [19,21], etc., which significantly affects the pedestal stability and ELM behavior. In addition, impurity seeding will lead to an increase of the effective charge number Z eff . Higher Z eff reduces the ion density through dilution effect and edge bootstrap current at higher collisionality, facilitating the stabilization of peeling-ballooning modes (PBMs) [19,22]. Some simulations have also suggested that the enhanced resistivity due to increased Z eff could also provide a stabilizing effect on the PBMs [23,24].
EAST tokamak has been equipped with an ITER-like tungsten (W) upper divertor, a carbon lower divertor and a molybdenum (Mo) first wall in the 2019-2020 experimental campaign. Lithium wall coating is routinely performed to achieve high-quality H-mode plasmas every morning before plasma operation [25]. Under this wall condition, a series of repetitive radiative divertor experiments with Ne seeding have been conducted to study the impacts of recycling impurity on the background plasma behavior. EAST results show that, at similar operational parameters, the recycling impurity could have a profound effect on the background plasma behavior, especially the ELM activity. The background ELM behavior prior to impurity seeding changes from large to mixed ELMs, and finally to pure small ELMs in these experiments. The rest of this paper is organized as follows. Section 2 describes the EAST device and the main diagnostics used in this work. Section 3 presents the change in the background ELM behavior before Ne seeding during the repetitive radiative divertor experiments. The increase of Ne and high-Z impurities in the bulk plasma and their impacts on the global plasma parameters are reported in section 4. Three typical discharges of large, mixed, and pure small ELMs are described in detail in section 5. Pedestal stability analysis and the underlying mechanism for the change in ELM behavior are presented in section 6. The estimation of the recycling Ne concentration and the reasons for the observed high-Z impurity accumulation are discussed in section 7. Finally, section 8 gives the summary and future work.

Experimental setup and main diagnostics
The series of repetitive radiative divertor experiments was performed in the 2019 EAST autumn campaign. EAST is a medium-sized fully superconducting tokamak with major radius R 0 ∼ 1.85 m and minor radius a ∼ 0.45 m. The flexible poloidal field control system enables the machine to achieve upper single null (USN), lower single null and double null divertor configurations. The supersonic molecular beam injection (SMBI) system has been implemented for routine density feedback control [26].
The main diagnostics used in this work are illustrated in figure 1. The multichannel Faraday-effect polarimeterinterferometer diagnostic has been developed for electron density measurements, including line-averaged density and core density profile [27]. The edge electron density profile can be provided by the microwave reflectometry with high spatial and temporal resolutions [28,29]. The Thomson scattering diagnostic system can provide the electron temperature profile [30,31]. With neutral beam injection (NBI) heating, the ion temperature profile in the plasma core can be measured by the charge exchange recombination spectroscopy [32]. The turbulent fluctuation in the pedestal can be measured by the Doppler backscattering system [33]. For the information on impurity, the core line-averaged effective charge number Z eff can be measured by the visible bremsstrahlung system [34]. The extreme ultraviolet spectrometer has been developed to measure impurity emissions and monitor the evolution of impurity levels in the main plasma [35,36]. The distribution of radiation power can be estimated from the 64-channel absolute extreme ultraviolet photodiode arrays [37]. The filterscope system can measure the D α emission in the divertor region. Furthermore, the divertor Langmuir probe system can measure the plasma parameters in the divertor region, including electron density, temperature, particle and heat fluxes.

Change in the background ELM behavior before Ne seeding in the repetitive radiative divertor experiments
In the series of radiative divertor experiments, one of the most evident phenomena is the change in the background ELM behavior before Ne seeding, as shown in figure 2. These discharges are reproduced with the same plasma current I p = 500 kA, magnetic field B t = 2.4 T, edge safety factor q 95 = 5.5, elongation κ = 1.6, upper triangularity δ u = 0.55, and are all under the USN divertor configuration. The main ion species of the plasmas is deuterium. The plasma density is feedback controlled by the SMBI system with a same preset density target for these discharges. The cryopumps installed underneath the upper and lower outer divertor target are in working condition. The source heating powers are also the same, including 1.4 MW 4.6 GHz lower hybrid wave (LHW), 0.8 MW electron cyclotron resonance heating (ECRH) and 3 MW co-current neutral beam injection (NBI). The ECRH power is deposited in the plasma core for these discharges. The divertor target remained attached in these discharges with electron temperature at the upper outer strike point in the range of 40-60 eV. Figure 2 shows the upper divertor D α emission in the pre-Ne-seeding phase. It can be observed that the ELMs before Ne seeding are large ELMs in the discharges #91614-91616. As the experiments continue, the background ELM behavior gradually changes from large to mixed ELMs, and finally to pure small ELMs. Correspondingly, the ELM frequency of large ELMs decreases from ∼120 Hz to ∼0 Hz, and the ELM frequency of small ELMs increases up to ∼1400 Hz, as shown in figure 3. Note that, the discharges (#91617, #91620, #91622, #91624, #91626) in the pulse sequence #91614-91633 but not shown in figure 2 are the discharges failed in the initial ramp-up phase of plasma current. Moreover, there is no Ne seeding in these discharges. Considering the above reasons, these discharges have not been included in this study.

Increased Ne and high-Z impurity in the bulk plasma and their impacts on global plasma parameters
In this section, the Ne retention phenomenon, the high-Z impurity accumulation in the plasma core, and their impacts on the global plasma parameters are introduced in detail.

Increased Ne in the bulk plasma
Figure 4(a) shows the seeded Ne particle number and figure 4(b) shows the background Ne emission in the bulk plasma in each shot. As shown in figure 4(b), the pre-seeding Ne emission increases gradually as the experiments continue in the shot sequence #91614-91633. It is worth pointing out that the Ne emission can still be measured in the followed several shots even without Ne seeding, as illustrated by the shaded area in figure 4(b). This suggests that part of the seeded Ne particles could be retained in the first wall, and then released back to the main plasma. Similar increase of impurity emission by retention impurity has been observed in [4,11]. As the mass spectrometer measuring the partial pressure of impurity is located close to the entrance of pumping duct at the midplane and does not cover the divertor region, the quantitative estimation of retention impurity particle number cannot be conducted with the particle balance analysis. The Ne retention on the wall could be correlated with the lithium-coated wall [38] and the graphite target of lower divertor [4,6] on EAST. It is also worth mentioning that, during or between the discharges of #91614-91633, none of the wall conditioning methods (such as lithium/boron powder injection, glow discharge, etc) has been used.

Accumulation of high-Z impurity in the bulk plasma
Besides the recycling Ne, the accumulation of high-Z impurities (W and Mo) in the bulk plasma can also be observed. Figure 5 shows the radiated power P rad,main , the tungsten unresolved transition array (W-UTA) emission at 45-60 Å in the main plasma, and the core electron temperature T e,core at normalized poloidal flux ψ N ∼ 0.01 prior to Ne seeding. The P rad,main and W emission increase dramatically, suggesting the accumulation of high-Z impurities. This leads to an evident  decrease in the core electron temperature T e,core , as shown in figure 5(b). T e,core decreases from ∼3.8 keV in the early discharges with large ELMs to ∼2.6 keV in the latter discharges with small ELMs. In contrast, there is only a slight change in the low-Z impurities (such as C, Li) in the bulk plasma during these experiments. It is also notable that, the high-Z impurity accumulation in the discharge #91630 is mitigated and the ELM behavior changes back to mixed ELMs. It is found that a pedestal coherent mode of frequency range ∼7-15 kHz in shot #91630 appears to be stronger than that of the adjacent shots, which may contribute to reducing the high-Z impurity level.

Impacts on the global plasma parameters
The increased Ne and high-Z impurities could also have impacts on the global plasma parameters. The variations in  line-averaged plasma density n el , stored energy W MHD , confinement enhancement factor H 98y,2 , loop voltage V loop and effective charge number Z eff are illustrated in figure 6. As shown in figure 6(a), the plasma density can be well controlled in the early discharges. Nevertheless, in the latter discharges with pure small ELMs, the plasma density n el increases from 3.4 × 10 19 m −3 to 4.0 × 10 19 m −3 (corresponding to Greenwald density fraction f GW from 0.44 to 0.52) due to the occurrence of high-Z impurity accumulation in the plasma core. For the stored energy and energy confinement, there is no evident change in the pre-Ne-seeding phase in these discharges. However, the good energy confinement cannot be sustained for a long time in the latter discharges. The increase of the loop voltage is correlated with the increase of Z eff (figure 6(d)) and the decrease of electron temperature ( figure 5(b)).

Typical discharges of large, mixed and pure small ELMs
Three typical discharges of large, mixed and pure small ELMs in the pre-Ne-seeding phase are shown in figure 7. Consistent with the statistical results, the Ne and W emissions increase as the experiments continue. As shown in figure 7(f ), the radiated power loss increases evidently in the plasma core. The accumulation of high-Z impurity is significantly enhanced in the discharge #91633 with pure small ELMs, and even leads to a back transition to low confinement mode (L-mode), as shown in figures 7(e) and (b). This indicates that the high-frequency small ELMs in our case could be type-III like ELMs, since the plasma is in the marginal state of H-L transition with a high radiation power loss. Furthermore, as shown in figure 8, for the shot #91618 from which small ELMs begin to occur, the ELM behavior is dominated by the small ELMs before the increase of NBI power at t = 3.5 s. With heating power increasing, the small ELM frequency decreases while the large ELM frequency increases. This is consistent with the observations on JET tokamak that Type-III and Type-I ELMs could coexist in an intermediate range of heating power [39]. Moreover, JET results also indicate that the duration of the phases with type III ELMs decreases with increasing heating power. This is also similar to the observations in our study. It seems that the ELM behavior change (figure 2) in the shot sequence #91614-91633 could be a change from type-I ELMs to type-III ELMs with increasing radiated power.

Plasma profiles and pedestal stability analysis
In this section, the plasma profiles of three discharges with large, mixed and pure small ELMs are presented. The change of pedestal fluctuation is discussed, and the linear pedestal stability analysis is performed to study the underlying mechanism behind the ELM behavior change. Figure 9 compares the plasma core and pedestal profiles prior to Ne seeding for large (#91616), mixed (#91623) and pure small ELM (#91633) discharges. In the core region, the plasma profiles of discharges with large and mixed ELMs are similar, while the discharge #91633 with pure small ELMs has a higher and peaked electron density n e and lower electron temperature T e . The ion temperature T i in the plasma core is increased in the discharge with pure small ELMs, which has also been observed in other experiments with external impurity injection [40][41][42][43][44]. The reasons for this are speculated to be the stabilization of ion temperature gradient-driven turbulence induced by impurity [43][44][45][46] and the increased ohmic heating power at higher loop voltage. It is worth mentioning that the peaked core n e and decreased T e have been widely observed in experiments with high-Z impurity accumulation [47][48][49].
In the pedestal region, the density profile shifts inward with tiny change in the density gradient when the ELM size becomes smaller, as shown in figure 9(e). Consistent with the change in core T e , the pedestal T e decreases evidently for the small ELM case. Therefore, the pedestal pressure gradient and edge bootstrap current estimated from the Sauter model [50] reduce significantly.
The change in the pedestal profiles could also affect the edge turbulence behavior. Figure 10 shows the power spectrum of pedestal density fluctuation in the inter-ELM phase prior to Ne seeding for discharges #91616, #91623, and #91633, where the electrostatic fluctuation named edge coherent mode (ECM) can be observed. The ECM is located in the  pedestal steep gradient region [51] and appears to be destabilized by the enhanced pedestal pressure gradient, especially the temperature gradient [52]. This is consistent with the observation in these experiments. The ECM fluctuation intensities in the mixed and pure small ELM discharges with a low pedestal T e are much lower than that in the large ELM discharge, as shown in figure 10.
Linear peeling-ballooning (PB) stability analysis has been carried out to study the mechanism behind the change in ELM behavior using the ideal MHD eigenvalue ELITE code [53]. The kinetic equilibria, as input to the ELITE calculation, are generated with EFIT code [54] within the constraints of the experimental profiles. Figure 11 shows the operational points and PB stability boundaries for the large, mixed and pure small ELM cases. The vertical axis is the normalized edge current density (J max + J sep )/2⟨J⟩, where ⟨J⟩ is the volume-averaged current density. The horizontal axis is the normalized pedes- 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 the growth rate of the most unstable mode normalized to half of the ion diamagnetic frequency γ/(ω * i /2) is equal to 1. The results show that the decreases in pedestal pressure gradient and edge bootstrap current lead to a downward shift of operational point from the unstable to stable region. This could be the primary reason for the disappearance of large ELMs. On the other hand, the moderate inward shift of pedestal density could provide a stabilizing effect on the ballooning instability [14], and thus a slight expansion of the ballooning boundary could be observed in figure 11. This could also play a minor role in the disappearance of large ELMs. For the appearance of high-frequency small ELMs (likely type-III ELMs), although the pedestal is linear stable to the peeling-ballooning modes, the reduced electron temperature T e causes the increase of edge resistivity η (η ∝ 1/T 3/2 e ), therefore the type-III ELMs could be triggered by the resistive ballooning instability [55]. For the mixed ELM case, its operational point lies near the PB stability boundary due to Figure 9. Radial profiles of large ELM discharge #91616, mixed ELM discharge #91623 and pure small ELM discharge #91633, including core profiles of (a) electron density ne, (b) electron temperature Te, (c) ion temperature T i , (d) total pressure ptot, and pedestal profiles of (e) ne, (f ) Te, (g) current density J, (h) ptot.  a moderately decreased pedestal T e . When the pedestal pressure gradient and current density exceed the PB instability threshold, large ELMs could be triggered. When the pedestal is below the PB instability threshold, type-III ELMs may be triggered by the resistive ballooning instability. As a result, a mixture of large and type-III ELMs appears.
Overall, the key mechanism for the change of ELM behavior could be that: as the recycling Ne increases in the bulk plasma, the concentration of high-Z impurities increases and the electron temperature decreases, causing a reduction of pedestal pressure gradient and edge bootstrap current. Consequently, the pedestal becomes more and more stable with respect to PB instabilities and thus the large ELMs gradually disappear. On the other hand, the reduced electron temperature causes the increase of edge resistivity. In this condition the type-III ELMs could be triggered by the resistive ballooning instability. This could be the physics process for the change from large ELMs to mixed ELMs, and finally pure small ELMs.

Discussion
The concentration of recycling Ne in the main plasma has been estimated for the discharges with pure small ELMs. The estimations indicate that the Ne concentration is of the order of ∼1% electron density. Figure 12 shows the Ne and W emissions, P rad,main and divertor D α emission in the first discharge #91614 with large ELMs and the last #91633 with pure small ELMs. It can be found from figures 12(a) and (e) that the Ne emission before Ne seeding in discharge 91633 is close to that in discharge #91614 after Ne seeding, suggesting that the Ne concentrations in the two cases are comparable. In the discharge #91614, the total seeded Ne particle number from divertor gas puffing is ∼4.14 × 10 19 . Therefore, the volume-averaged Ne concentration is estimated to be ∼0.7% of electron density with the similar method used in [15] and a reasonable assumption of ∼5% fuelling efficiency for Ne gas [56]. On the other hand, a similar Ne concentration of ∼0.8% of electron density could also be estimated from the variation of Z eff (increase from ∼2.5 to ∼3.2) in discharge #91614, with the assumption that the increase in Z eff is mainly contributed from Ne impurity and Ne particles are fully ionized.
From figure 12, the impacts of the Ne itself and high-Z impurities on the ELM behavior could also be identified separately. As shown in figure 12(d), the ELM behavior changes moderately after Ne seeding in discharge #91614. In contrast, with higher W emission and higher P rad,main , pure small ELMs could be observed in discharge #91633. This result further indicates that the accumulation of high-Z impurities could be the dominant reason for the change in ELM behavior.
The underlying reasons for the high-Z impurity accumulation are briefly discussed here. (1) Impurity sputtering: The threshold temperature for W sputtering by Ne ions is lower relative to deuterium ions [57]. Therefore Ne particles could cause a stronger W sputtering on the divertor target and guard limiter of LHW antenna as well as stronger Mo sputtering on the first wall, providing the source of high-Z impurities.
(2) Turbulent transport: The intensity of pedestal ECM fluctuation in small ELM case is lower than that in large ELM case (figure 10), this may contribute to the impurity accumulation. (3) Neoclassical inward pinch: The peaked n e gradient and flattened T e profiles in the plasma core (figures 9(a) and (b)) could enhance the inward pinch, which facilitates the high-Z impurity accumulation [58]. (4) Change of ELM type: ELMs can flush the impurity particles out of the pedestal region and reduce the impurity concentration [59,60], therefore in our case the change of ELM type may also play a role in the impurity accumulation. Due to diagnostic constraints on EAST, there is no direct and reliable experimental data to confirm the dominant one for impurity accumulation, this issue still needs to be clarified by more efforts in future work.

Summary and future work
A significant change in the background ELM behavior before Ne seeding has been observed in a series of repetitive radiative divertor experiments in EAST. With similar operational parameters, the ELM behavior prior to Ne seeding changes from large to mixed ELMs, and ultimately to pure small ELMs. Simultaneously, the pre-seeding Ne emission in the bulk plasma increases gradually, suggesting the retention and recycling of Ne in the wall surface. In addition, the higher radiated power and the accumulation of high-Z impurities (W and Mo) have also been observed, leading to a lower electron temperature. In the latter discharges with pure small ELMs, the high-Z impurity accumulation can be extremely serious and even causes a back transition to the L-mode.
Due to the high-Z impurity accumulation, the n e increases and peaks in the core region. The moderate increase of T i , which has been widely observed in other experiments with external impurity injection, has also been found in this experiment with the increase of recycling Ne. In the pedestal region, the density profile shifts inward with tiny change in the density gradient. Accompanied by the reduction of T e in the plasma core, the pedestal T e is also reduced, leading to a significant decrease in the pedestal pressure gradient and edge current density. Linear PB stability analysis suggests that the reduction of T e should be the primary reason for the change of background ELMs.
It is worth pointing out that the Ne retention has been widely observed in the Ne-seeded experiments on EAST. The retention of Ne could be correlated with the lithium-coated wall [38] and the graphite of lower divertor [4,6]. The impacts of metal wall on the Ne retention is still unknown. To clarify this, more efforts need to be made in the future. The influence of impurity retention on the main plasma operation has also been studied in other devices. The divertor experiments with power load feedback control in ASDEX Upgrade show that the retention and release of nitrogen on the tungsten wall surface would increase the nitrogen consumption, and the nitrogen in plasma can still be measured in the followed several shots even without nitrogen injection [11]. Similarly, the retention Ne can still be found in the followed several shots without Ne seeding in our study as shown by the shaded area in figure 4(b).
For the future work on EAST, the particle balance analysis should be carried out with the upgrade of the mass spectrometer system to quantitatively evaluate the retention impurity particle number. The dominant reason for the high-Z impurity accumulation should also be further studied, as ASDEX Upgrade [61] and JET [62] results have suggested that the W impurity investigation is an important issue in the full mental wall operation. Furthermore, it is also necessary to explore effective methods to mitigate the impurity retention and reduce its impacts on the main plasma behaviors, which should be helpful for the high-confinement steady-state operation with radiative divertor in future.