Long-pulse H-mode operation with stored-energy monitoring for detachment feedback control with a new lower tungsten divertor in EAST

One of the key challenges for future fusion research is to mitigate the high steady-state heat load on the divertor target plates, and divertor detachment offers a promising solution. The Experimental Advanced Superconducting Tokamak (EAST) has developed several feedback control methods for divertor detachment. However, when an off-normal event momentarily disturbs the main plasma, impurity seeding may still be conducted by these methods for detachment, which probably drives the main plasma further away from its stable equilibrium or even causes the disruption. These off-normal events include excessive impurity seeding, loss of heating and dust droplets, which are not rare in current tokamak experiments, especially in long-pulse operation. To compensate for the drawbacks of these methods, we propose and develop a module of stored-energy monitoring to ensure stable plasmas in long-pulse operation. The stored energy usually decreases when the main plasma is away from its stable equilibrium, which is suitable for monitoring the state of the main plasma. Once the stored energy falls below a certain threshold, the module actively switches off the impurity seeding system. Without impurity seeding, the main plasma can recover with the increase in the stored energy. Only when the stored energy exceeds another threshold does the module switch on the impurity seeding to continue the detachment operation. The module function has been verified during the EAST radiative divertor experiments in the newly-upgraded lower tungsten divertor. A typical ∼20 s discharge in grassy-ELM H-mode regime with ∼5 MW source heating power is demonstrated with divertor partial detachment and good energy confinement by active impurity seeding (50% neon, 50% D2). The energy confinement factor is maintained at a high level, i.e. H98,y2∼1.1. The electron temperature in the core region only has a slight change after the impurity seeding, while the electron density has a ∼10% increase. Furthermore, the ion temperature near the axis also has a remarkable increase. These achievements provide an important demonstration of the actively controlled radiative divertor mitigating the heat loads with good core confinement, which is an essential step toward steady-state operation of fusion reactors.

One of the key challenges for future fusion research is to mitigate the high steady-state heat load on the divertor target plates, and divertor detachment offers a promising solution. The Experimental Advanced Superconducting Tokamak (EAST) has developed several feedback control methods for divertor detachment. However, when an off-normal event momentarily disturbs the main plasma, impurity seeding may still be conducted by these methods for detachment, which probably drives the main plasma further away from its stable equilibrium or even causes the disruption. These off-normal events include excessive impurity seeding, loss of heating and dust droplets, which are not rare in current tokamak experiments, especially in long-pulse operation. To compensate for the drawbacks of these methods, we propose and develop a module of stored-energy monitoring to ensure stable plasmas in long-pulse operation. The stored energy usually decreases when the main plasma is away from its stable equilibrium, which is suitable for monitoring the state of the main plasma. Once the stored energy falls below a certain threshold, the module actively switches off the impurity seeding system. Without impurity seeding, the main plasma can recover with the increase in the stored energy. Only when the stored energy exceeds another threshold does the module switch on the impurity seeding to continue the detachment operation. The module function has been verified during the EAST radiative divertor experiments in the newly-upgraded lower tungsten divertor. A typical ∼20 s * 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 excessively high heat load on divertor target plates is one of the key challenges for future tokamak fusion reactors such as ITER [1]. In addition, divertor detachment is proposed as a promising solution for steady-state plasmawall interactions [2,3]. In a metal-wall environment, seeding low-Z impurities can greatly enhance the divertor power dissipation, and the electron temperature near the divertor target falls below 5 eV, which is a signature of divertor detachment [3]. However, excessive seeding can degrade the energy confinement, and even lead to a transition from high confinement mode (H-mode) to low confinement mode (L-mode) [2]. To restrict the negative impact of impurity seeding on the main plasma, many tokamaks have adopted detachment feedback control, aiming to achieve detachment with good energy confinement, such as DIII-D [4,5], ASDEX-Upgrade [6,7], JET [8,9] and JT-60U [10]. Several novel techniques, such as feedback control of impurity emission fronts in TCV [11][12][13] or the X-point radiator in the ASDEX-Upgrade [14], have been developed in recent years, and may have strong capacity to avoid excessive impurity seeding.
In the Experimental Advanced Superconducting Tokamak (EAST), several methods have been developed for divertor detachment, including feedback control of divertor particle flux [15,16] or electron temperature [5,17], total radiation power [18] or radiation intensity near the X-point [19], and the surface temperature of the target plate [20]. These methods have been demonstrated in an upper tungsten divertor, but the wall environment in the lower divertor region is still carbon. Moreover, the discharge pulse with high energy confinement is relatively short (<10 s). As a medium-sized fully superconducting tokamak, EAST aims to demonstrate longpulse high-performance operation. Therefore, the lower divertor was upgraded in 2021 with tungsten material, and the wall environment at present is almost all metal, except that a few neutral beam injection (NBI) shine-through dumpers are still covered with water-cooled graphite tiles [21]. The new lower divertor has a right-angle closed corner that consists of a horizontal target and a vertical target, which can induce the 'corner effect' [21,22]. SOLPS-ITER simulation shows that this new divertor has a relatively lower density onset or less impurity seeding for detachment compared to traditional vertical targets, which helps to reduce the negative effect of divertor detachment on the main plasma performance [21].
One important concern is the response of existing detachment feedback control methods in EAST to off-normal events. Off-normal events, such as dust droplets or a sudden drop in heating power, are not rare in present tokamak experiments [23,24], especially in long-pulse high-heating-power discharges. Excessive impurity seeding is also common in EAST detachment experiments. When these off-normal events momentarily push the main plasma away from its stable equilibrium, existing feedback control methods in EAST may continue low-Z impurity seeding for divertor detachment. This probably drives the main plasma further away from equilibrium or even causes disruption, which is intolerable for future reactors. Therefore, it is necessary to develop a monitoring function to communicate the significant changes in the main plasma or control the objective of the detachment controller, aiming to handle the occurrence of off-normal events in EAST detachment experiments.
In this paper, a newly-developed module for stored-energy monitoring is reported, aiming to promote the capacity of existing detachment feedback control methods in EAST to handle the occurrence of off-normal events and ensure stable main plasma in long-pulse discharges. The module function is verified in the EAST radiative divertor experiments with impurity seeding (50% neon, 50% D 2 ) from the newly upgraded lower divertor. These experiments are conducted with high heating power in grassy edge-localized-mode (ELM) H-mode regime, characterized by high-frequency small-amplitude ELMs [25,26]. A typical ∼20 s discharge with ∼5 MW source heating power is demonstrated with divertor detachment, and the energy confinement factor is maintained at a high level, i.e. H 98,y2 ∼ 1.1. The plasma behavior changes in different regions with the impurity seeding is also presented and discussed. The remainder of this paper is organized as follows: the necessity for and the basic logic of the stored-energy monitoring module is described in section 2; the experimental setup and main diagnostics are presented in section 3; the experimental results are presented in section 4; finally, a summary and discussion are presented in section 5.

The necessity to develop the module
In tokamak experiments, off-normal events are not rare, especially in long-pulse discharges [23,24]. During the EAST 2021 summer campaign, there were 53 shots in total with a radiative divertor, as shown in figure 1(a), in which 16 shots (∼30.2%) end up with disruption. The reasons for the disruption are mainly classified into three types: excessive seeding (10 shots, i.e. ∼18.9%), loss of heating (4 shots, i.e. 7.5%) and dust droplets (2 shots, i.e. 3.8%). Even in the remaining 37 normal shots, the main plasma performance is also affected by off-normal events, as shown in figure 1(b). Excessive seeding or loss of heating induces the transition from H-mode to L-mode, and the main plasma in some shots cannot recover to H-mode.
As shown in figure 2, three typical cases during radiative divertor experiments will be illustrated in detail, to indicate that existing feedback control methods in EAST may drive the main plasma further away from its stable equilibrium, or even cause disruption when an off-normal event already disturbs the main plasma.
(1) In shot #101889, excessive seeding induces several transitions between H-mode and L-mode and eventually leads to disruption. Since the degradation of pedestal performance with excessive divertor impurity seeding tends to be accompanied with significantly increased radiation near the divertor X-point [19], the radiation intensity of channel #6 from the absolute extreme ultraviolet (AXUV) array [27], whose chord passes near the X-point, is chosen as the feedback signal. As shown in figure 2(a1), the plasma control system (PCS) [28] opens the piezo valve for mixed gas seeding (50% neon, 50% D 2 ) once the radiation intensity is below the feedback target. Meanwhile, the chordintegrated intensity of Ne IX (1.345 nm) line emission I Ne , normalized by the central-averaged electron density n el [29,30], increases with a certain time delay, as shown in figure 2(a2). However, the main plasma is away from its stable equilibrium after ∼9 s due to excessive neon impurities, and several transitions between H-mode and L-mode occur, as shown in figures 2(a2) and (a3). After 11.2 s, the main plasma stays in L-mode with low radiation power.
Since the feedback signal is always below the target, the PCS opens the valve with continuous impurity seeding, consequently leading to the disruption. (2) The accidental loss of auxiliary heating power occurs in shot #101892, but the PCS continues the impurity seeding for feedback control, which drives the main plasma further away from its stable equilibrium. As shown in figure 2(b1), the initial source heating power is ∼4.8 MW, including ∼1.9 MW 4.6 GHz low hybrid waves (LHWs), ∼1.4 MW electron cyclotron resonance heating (ECRH) and ∼1.5 MW NBI from a system named NBI1L. Then, the source ECRH power suddenly drops to ∼0.5 MW at ∼6.9 s, which momentarily disturbs the main plasma, as shown in figures 2(b2) and (b3). The impact of heating power reduction on electron density and stored energy seems to be insignificant around ∼7.04 s; however, the impurity seeding is still conducted and the normalized neon radiation increases largely after ∼7.12 s, which disturbs the main plasma severely and finally leads to disruption. (3) In shot #101905, an off-normal event momentarily disturbs the main plasma at ∼7.82 s. As shown in figure 2(c1), the normalized iron radiation measured by the extreme ultraviolet (EUV) spectrometer increases sharply, implying that a fleck of dust enters the main plasma from the components of the vacuum chamber. However, the PCS still continues impurity seeding for feedback control, even after the occurrence of dust droplets, which drives the main plasma further away from its stable equilibrium. Finally, the discharge ends up at ∼8.17 s with the disruption.
As shown above, when an off-normal event momentarily disturbs the main plasma, existing feedback control methods in EAST may still conduct impurity seeding for divertor detachment, which drives the main plasma further away from its stable equilibrium and even causes the disruption. Although the three cases are under the feedback control of radiation intensity, this drawback also exists in other feedback control methods in EAST. For instance, feedback control of divertor particle flux or electron temperature also need feedforward impurity seeding [5,[15][16][17]. However, when the chosen feedback signal reaches the improper target when stopping impurity seeding, excessive feedforward seeding may have already significantly degraded the main plasma performance [5] and may even lead to disruption. Therefore, it is necessary to develop a module to prevent impurity seeding after off-normal events, such as excessive impurity seeding, disturb the main plasma severely. As shown in figures 2(a3), (b3) and (c3), the stored energy usually decreases when the main plasma is away from its stable equilibrium, which could be used to monitor the state of the main plasma. This inspires us to develop a module for stored-energy monitoring to compensate for the drawbacks of existing feedback control methods in EAST.

The logic of stored-energy monitoring module in detachment feedback control
In the EAST 2021 summer campaign, the module of storedenergy monitoring was combined with different feedback control methods and implemented in the PCS. Here, we mainly introduce the overall logic of the module of stored-energy monitoring for feedback control of the total radiation power. As shown in figure 3, the stored-energy threshold W threshold is set to W 0 initially, according to the reference shot, and the module monitors the stored energy W MHD in the whole discharge. Once an off-normal event occurs and W MHD drops below W threshold , the PCS switches off the impurity seeding system to avoid driving the main plasma further away from its stable equilibrium, and resets W threshold to W 1 (W 1 > W 0 ). Without impurity seeding, the main plasma recovers with the W MHD increase. Once W MHD exceeds the threshold, the PCS switches on the impurity seeding system with the initializing  Three typical shots with off-normal events under feedback control of radiation intensity. Shot #101889: (a1) the single channel radiation intensity PXUV6 (black) for the feedback control, the feedback target (red) and the voltage signal of the piezo valve for impurity seeding (blue); (a2) the central-averaged electron density n el (black) and the normalized neon radiation intensity I Ne /n el (yellow); (a3) the stored energy W MHD provided by EFIT equilibrium. Shot #101892: (b1) the source auxiliary heating power, including 4.6 GHz LHW (blue), NBI (red) and ECRH (black); (b2) n el (black), (yellow) and the voltage signal of the piezo valve (blue); (b3) W MHD . Shot #101905: (c1) the total radiation power P rad, total (black) and the normalized iron radiation I Fe /n el (red); (c2) n el (black), I Ne /n el (yellow) and the voltage signal of the piezo valve (blue); (c3) W MHD . W threshold , and continues to conduct the detachment feedback control. The reason why we set W 1 > W 0 is to avoid premature impurity seeding disturbing the fragile recovery.
The logic of feedback control for divertor detachment is similar to earlier works [18,19]. First, the electron temperature T et , measured by divertor Langmuir probes (Div-LPs) [31], is used as the sensor to access detachment onset. Here, the threshold value T onset is set manually, and should be higher than the detachment signature (5 eV) by considering the following fact: the total time from the moment that the gaspuffing command is sent from the PCS to the moment that the radiative diagnostic measures a change caused by the impurity seeding is more than 100 ms for the divertor piezo valve [18]. Once T et ⩽ T onset is confirmed, the PCS switches to feedback control of the total radiation power. Here, the feedback target of the radiation power P rad,target is modified when T et falls below T onset , aiming to avoid excessive feedforward impurity seeding.
The actuator is the gas puff through the piezo valve. The valve voltage and the impurity seeding level can be varied smoothly in feedback control; however, this control mode easily causes excessive seeding and often leads to the disruption in EAST experiments. The EAST team usually chooses the pulsed gas puff for impurity seeding [15][16][17][18][19][20]. The value of the piezo value, maximum pulse width and minimum time interval do not change in a shot. The proportional-integral-derivative (PID) controller modulates the pulse width and time interval in the piezo valve duty circle, based on the deviation between the radiation power target level P rad,target and the real-time total radiation power P rad . Initially, the gas is injected with a relatively large pulse width to quickly promote the radiation level. Once T et ⩽ T onset is confirmed by the PCS, the target level P rad,target is modified at a lower value. And the PID controller modulates the pulse width and time interval with appropriate values to control the AXUV signal level around the target level. In the whole process of feedback control, the module of stored-energy monitoring always works to deal with the occurrence of the off-normal event.

Experimental setup and main diagnostics
During the EAST 2021 summer campaign, long-pulse experiments were conducted on a new lower divertor with detachment feedback control. A typical long-pulse H-mode shot (#103777) with radiative feedback control is operated in the lower single-null configuration with the plasma current I p = 400 kA, as shown in figure 4. The magnetic field B t = 2.4 T is in a favorable direction, i.e. the ion B×∇B drift direction toward the lower divertor. The total source power of the auxiliary heating is ∼5 MW, including ∼2.1 MW 4.6 GHz LHW, ∼1.4 MW ECRH and ∼1.4 MW NBI1L during 2.2-11 s and another system (NBI2L) during 11-20 s, as shown in figure 4(b). Although EAST uses several gas puff systems to actively control the central-averaged electron density n el at a set value, the mixed gas seeding induces the fueling effect and tends to increase n el , which is commonly beyond the PCS capacity of n el control in EAST detachment experiments [15][16][17][18][19][20]. After the mixed gas seeding in shot #103777, n el increases from ∼ 4 × 10 19 m −3 to ∼ 4.4 × 10 19 m −3 , i.e. f GW ∼ 65% Greenwald density fraction, as shown in figure 4(c). The stored energy is ∼180 kJ, and the poloidal beta β p is ∼1.8. Both gradually decrease with the slow reduction in auxiliary heating power, seen in figures 4(b), (d) and (e). The ELM type is classified into grassy ELMs, which have strong tungsten impurity exhaust capability [26].
The main diagnostics used in this paper are briefly introduced here, as shown in figure 5. The total radiation power of the bulk plasma P rad is in fast real-time calculated by the PCS [18], based on high-time-resolution measurement of the line-integrated radiation intensity by the AXUV array [27] and the real-time RTEFIT plasma equilibrium [32]. Plasma behavior near the divertor target is diagnosed by the Div-LPs [31], as shown in figure 6(d). An infrared (IR) thermography diagnostic [33] measures the surface temperature on the divertor target plate. The neon and tungsten line emissions in the core plasma are measured using the EUV spectrometer [29,30]. The core electron density is measured by 11-channel polarimeter-interferometers (POINT) [34], and the edge electron density is measured by the microwave reflectometers (Refl.) [35]. The electron temperature is provided by a Thomson scattering (TS) system [36]. The 48-channel heterodyne radiometers [37] detect the electron cyclotron emission (ECE) to diagnose the local electron temperature information. The effective charge number Z eff is measured by a visible bremsstrahlung (VB) system [38].
To utilize the 'corner effect' of the new lower divertor [21,22], we adjust the location of the lower outer strike point to locate on the horizontal target near the corner, as shown in figure 6(d). The gas inlet is between probe #10 and probe #9 on the vertical target in the lower outer divertor [21]. With the favorable magnetic field direction, the radial E × B drift can drive the impurity ions from the scrape-off layer (SOL) side to the private flux region side to cool the target region with high heat load efficiently [39,40]. The ratio of injected neon atoms to D 2 molecules is 50%:50% due to the following reasons: a too large pure neon peak seeding rate usually leads to disruption in present EAST experiments; it is technically practical to use the same piezo valve for the impurity gas and D 2 to alleviate the impurity seeding rate to meet the requirements of detachment feedback control [41]. The maximum voltage of the piezo valve at the gas inlet is set at 3.6 V, i.e. 3.3 × 10 20 particles s −1 . The PID modulates the pulse width and time interval of the voltage to control the impurity seeding, as shown in figure 6(b). The electron temperature measured by probe #11 is used to identify the onset of detachment, which locates on the vertical target near the corner, as shown in figure 6(d). Based on earlier detachment experiment experience, the threshold value T onset is set appropriately as 15 eV to trigger the transition to feedback control of the total radiation power. To monitor the main plasma state, high-timeresolution signals of the stored energy W MHD,real are required, which can be provided by the RTEFIT plasma equilibrium [32] or diamagnetic loop [42]. In the experiments, we choose the former, which is also used in the calculations mentioned above.

Experimental results
The function of the stored-energy monitoring module is successfully demonstrated in the typical shot #103777. According to the reference shot, the threshold value mentioned in section 2 is set as W 0 = 160 kJ and W 1 = 1.1 W 0 . As shown in figures 6(b) and (c), the electron temperature of probe #11 falls below the threshold value 15 eV at ∼4.3 s, and the PCS switches to P rad feedback control by modifying the feedback target at ∼1.62 MW. However, the feedforward impurity seeding is excessive, which leads to the transition from Hmode to L-mode. The stored energy W MHD, real drops below the threshold W 0 at 4.81 s, and the PCS successfully switches off the impurity seeding system. Although the radiation power is below the setting target, there is no impurity seeding anymore, as shown in the yellow range of figure 6(b). Without impurity seeding, the main plasma recovers and the transition from L-mode to H-mode occurs. When W MHD,real increases up to the threshold W 1 at 5.61 s, the PCS switches on the impurity seeding system and continues the normal detachment feedback control. From 5.61 s, the system timely opens the piezo valve with appropriate impurity seeding to increase the radiation power, once the radiation power becomes below the feedback target with the stored energy above the threshold.
In later content, the transitions between L-mode and Hmode are no longer discussed, and the plasma behavior changes with the impurity seeding are described and discussed. We will use the 'condition before or after impurity seeding' to describe the stable H-mode period before or after the transition to avoid redundancy.

Divertor plasma behavior
The Div-LPs show that the divertor plasmas near the lower outer target are kept in partial detachment by the impurity seeding. As shown in figures 7(a)-(e), the electron temperature T e in the SOL side drops largely after 4.4 s. And T e at probe #10 and #11 is successfully maintained below 5 eV, which is In the shot #103777: (a) W MHD, real with the threshold to switch off (magenta) or switch on (green) the impurity-seeding system; (b) P rad (black) and the feedback target P rad, target (red) with the voltage signal of the piezo valve (blue); (c) Tet measured by probe #11 with the threshold value 15 eV to trigger the transition to P rad feedback control; and (d) the structure of the EAST lower outer divertor (black) with the Div-LP distribution (blue), the gas puff inlet (magenta) and the separatrix of shot #103777 (red). The yellow range in (a), (b) is the duration when the stored-energy monitoring module switches off the impurity seeding. a signature of detachment [2,3]. In channel #12, T e ⩽ 5 eV is also maintained most of the time, but becomes ∼10 eV during 6.5-13 s. In the far SOL, T e also drops after 4.4 s at probe #8, but is still above 10 eV.
The IR thermography also shows that the peak surface temperature on the lower outer divertor target becomes much lower after impurity seeding. The location with the peak surface temperature is on the horizontal target near the corner. As shown in figure 7(f ), the peak surface temperature can be exceeded before impurity seeding. Then, the seeded impurities enhance the divertor power dissipation and the peak surface temperature falls below 400 • C. Thus, the heat load on the divertor target is successfully mitigated by the detachment feedback control.

The main plasma performance
The EUV system can provide impurity information in the core region. Due to the measurement saturation of Ne X line emission (1.21 nm), we use the normalized chord-integrated intensity of Ne IX (1.345 nm) line emission to monitor the neon radiation in the main plasma. As shown in figure 8(b), the normalized neon radiation intensity increases largely after the impurity seeding, and almost remains constant during the feedback control experiment. The VB system shows that the effective charge Z eff increases from ∼1.8 to ∼2.6, as shown in figure 8(c). Correspondingly, the loop voltage increases from ∼0.09 V to ∼0.15 V, as shown in figure 8(d). However, there is no obvious decrease in the stored energy. As shown in figure 4(d) or figure 6(a), the stored energy increases slightly from ∼172 kJ to ∼178 kJ after impurity seeding. Although the total radiation power increases from ∼0.6 MW to ∼1.6 MW, as shown in figure 4(a), the energy confinement factor is maintained at a high level, i.e. H 98, y2 ∼ 1.1, as shown in figure 8(a). The core plasma performance has no degradation after the impurity seeding for detachment, which is different from our previous work on the upper tungsten divertor [19]. It is worth mentioning that EAST cannot separate core radiation loss from the total radiation power yet. Besides, the normalized tungsten radiation intensity also increases in the core region after impurity seeding, as shown in figure 8(e).
To further investigate the effect of impurity seeding on plasma behavior, the radial profiles of the electron density n e and electron temperature T e are shown in figure 9. POINT provides the n e profile in the core region at 3 s and 8 s, and the Refl. provides the profile in the edge region, as shown in figure 9(a). Since only fewer TS data are available before impurity seeding in shot #103777, we use some TS data from the nearby shots for the T e profile with the same parameters before impurity seeding. And the profiles after impurity seeding are based on the TS data during 6-10 s in shot #103777, in which the heating power is nearly unchanged. As shown in figure 9(a), the fueling effect of mixed gas seeding seems beyond the active n el control capacity of the PCS, and n e in the core region increases obviously after the impurity seeding. At the magnetic axis, n e is ∼ 5.5 × 10 19 m −3 at 3 s, and increases to ∼ 6.2 × 10 19 m −3 at 8 s (∼13% larger). The impurity seeding does not seem to affect the electron temperature profile in the core region, as shown in figure 9(b). Thus, the electron pressure in the core region actually increases after the impurity seeding, as shown in figure 9(c).
The x-ray imaging crystal spectrometers [43] on EAST can provide the ion temperature in the core region. As shown in figure 10, the ion temperature T i near the axis (ψ N < 0.2) increases after the impurity seeding. At ψ N ∼ 0, T i ∼ 1 keV at t = 2.9 s, but increases to ∼ 1.4 keV at t = 8.5 s after impurity seeding, i.e. ∼40% higher. Unfortunately, reliable ion temperature in the ψ N > 0.2 region cannot be provided in our experiments. The increase in ion temperature caused by the impurity seeding is similar to the neon or argon seeding results in HL-2A [44], in which the suppression of ion temperature gradient (ITG) turbulence plays an important role. Although EAST lacks a reliable diagnosis to observe the behavior of ITG turbulence, existing models [45] show that the growth rate of ITG decreases as the effective charge Z eff increases.

Edge plasma behavior
In the edge region, the impurity seeding tends to increase n e at the pedestal top, while it decreases n e at the separatrix, as shown in figure 9(d). Conversely, the impurity seeding tends to decrease T e at the pedestal top, while it increases T e at the separatrix, as shown in figure 9(e). The electron pressure gradient decreases after impurity seeding with a modest drop in the pedestal top pressure, as shown in figure 9(f ).
Grassy ELMs have the strong capacity for high-Z impurity exhaust and density control in EAST H-mode discharges [26]. Unfortunately, the small amplitude of grassy ELMs can hardly be estimated by the energy loss or the density drop due to the measurement resolution limits. This problem becomes more severe with the impurity seeding (50% neon, 50% D 2 ) from the lower outer divertor, which seriously affects the widely used D α peaks. Here, we only calculate the frequency of grassy ELMs. As shown in figures 11(a) and (b), the frequency of grassy ELMs is ∼2.1 kHz, and decreases to ∼1.6 kHz at 6 s after impurity seeding. The frequency decrease in grassy ELMs may be attributed to the more stable pedestal, as shown in figure 9(f ).  The ion temperature T i in the core region before (t = 2.9 s) and after (t = 8.5 s) the impurity seeding in shot #103777, measured by x-ray imaging crystal spectrometers.
In the H-mode period before the impurity seeding of shot #103777, two channels of ECE in the edge plasma region detect a mode with ∼25 kHz, as shown in figure 11(c). It is classified into the edge coherent mode (ECM), i.e. a quasielectrostatic instability, which has the nature of the dissipative trapped electron mode [46][47][48]. The transport capability Figure 11. In shot #103777, (a) the raw signal of single channel radiation intensity in the upper divertor PXUV58 (black) to calculate the ELM frequency and normalized neon radiation intensity I Ne /n el (blue); (b) the ELM frequency, and (c) an auto-power spectrum of electron temperature fluctuation measured by ECE channel #03, which corresponds to the edge region. of the ECM is strong, which can drive heat and particle flux exhaust through the separatrix [46]. However, the ECM disappears after ∼4.2 s and cannot be detected in the stable Hmode period with the impurity seeding. The disappearance of the ECM may be due to the reduction of the edge electron pressure gradient [48], as shown in figure 9(f ).

Summary and discussion
Several feedback control methods have been developed in EAST for divertor detachment to mitigate the high steadystate heat load on the divertor target plates. However, these methods may cause severe trouble when an off-normal event momentarily disturbs the main plasma, which is not rare in long-pulse operation. Even after the main plasma is already away from its stable equilibrium, these methods may still conduct impurity seeding for detachment, and may drive the main plasma further away from the equilibrium and even lead to disruption, which is intolerable for future fusion reactors. In this paper, we propose a module of stored-energy monitoring to communicate the important changes in the main plasma to the detachment controller, aiming to handle the occurrence of off-normal events and ensure stable main plasma in EAST long-pulse discharges. Off-normal events usually disturb the main plasma with a decrease in the stored energy; thus, the module actively switches off the impurity seeding system once the stored energy falls below a certain threshold. Without impurity seeding, the main plasma can recover and the stored energy increases. Only when the stored energy exceeds another threshold, does the module switch on the impurity seeding system to continue the feedback control experiment for divertor detachment.
The function of the stored-energy monitoring module has been successfully verified in the long-pulse radiative divertor experiments conducted on the new lower divertor during the EAST 2021 summer campaign. A typical ∼20 s shot is demonstrated with the compatibility of divertor partial detachment and good energy confinement, by active impurity seeding (50% neon, 50% D 2 ) from the lower outer divertor. The discharge is operated in the grassy-ELM H-mode regime with high heating power (P source ∼ 5 MW). Although excessive feedforward impurity seeding induces a transition from H-mode to L-mode, the module of stored-energy monitoring timely switches off the impurity seeding until the main plasma recovers. Active impurity seeding successfully keeps the divertor plasmas in partial detachment, and largely reduces the peak surface temperature of the divertor target plates. Although the total radiation power increases from ∼0.6 MW to ∼1.6 MW, the stored energy is maintained at ∼180 kJ with H 98,y2 ∼ 1.1, which indicates good energy confinement. Although the electron temperature at the pedestal top has a slight decrease, the electron pressure in the core region increases mainly due to the increase in electron density, which is similar to the nitrogen seeding results in DIII-D [49]. The impurity seeding also tends to reduce the ELM frequency, which is similar to the results in JET-ILW with nitrogen seeding [9] and JT-60U with argon seeding [10], though the ELM type is grassy ELM in EAST and type-I ELM in other devices. The large increase in ion temperature in the ψ N < 0.2 region after impurity seeding is similar to the neon or argon seeding results on HL-2A [44], and implies the suppression of ITG turbulence [45]. These achievements provide an important demonstration of the radiative divertor as the solution to the divertor heat loads with high energy confinement, which is beneficial toward long-pulse steady-state operation of fusion reactors.
For future fusion reactors, handling the occurrence of offevents in detachment operation will be much more complex compared to the EAST tokamak. Unlike the EAST experiments, switching off impurity seeding completely may be unacceptable for future reactors, even when the main plasma is already away from its stable equilibrium, since it is essential to ensure the survival of the plasma-facing components. A hierarchy is necessary in future reactors' various control systems to strive for a compromise: for instance, a reasonable reduced level of impurity seeding to ensure the survival of both the main plasma and the divertor targets when off-normal events occur. And the compromise needs real-time communication of significant information, such as the stored energy and the divertor plasma state among relevant control systems. Besides, it may be beneficial for other advanced feedback control techniques to develop a similar monitoring module to communicate significant changes in the main plasma or control the objective of the detachment controller, aiming to promote the capacity to handle the occurrence of off-normal events. For instance, feedback control methods of impurity emission fronts in TCV [11][12][13] or the X-point radiator in the ASDEX-Upgrade [14] have been developed in recent years, and seem to avoid excessive seeding in principle. However, the position of impurity emission fronts or the X-point radiator may be directly affected by a sudden loss in the heating power. Combining the monitoring module with these feedback control methods can yield further robustness to handle off-normal events, including a sudden loss in the heating power.