First results of Ne shattered pellet injection for mitigating plasma disruption with full metal wall in EAST tokamak

Disruption mitigation poses a significant and unresolved challenge for ITER and future fusion reactor devices. To mitigate the effect of plasma disruption, a Shattered Pellet Injection (SPI) system has been developed and the first rapid shutdown experiments have been successfully performed in the Experimental Advanced Superconducting Tokamak with full metal wall. The experiments confirmed the system’s capability to inject 5 mm diameter neon (Ne) pellets at velocities ranging from 100 to 400 m s−1, with a maximum injected quantity of 13.2 Pa m3. Compared to unmitigated disruptions, the total radiation power was significantly higher with the respective use of SPI and Massive Gas Injection (MGI). Additionally, the radiation distribution and divertor heat flux were compared between SPI and MGI methods. The results demonstrated that SPI exhibited shorter cooling time, stronger core radiation and more uniform poloidal radiation distribution compared to MGI, indicating deeper deposition using SPI. Furthermore, during plasma shutdown, the electron temperature and peak heat flux near outer divertor strike points were reduced by 40% and 50% respectively, with SPI, in comparison to MGI. These findings serve as a valuable reference for implementing SPI technology as the baseline approach for disruption mitigation in ITER.

(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
Disruption poses a significant risk to the safe operation of tokamak devices such as ITER and future fusion reactors. It can lead to detrimental effects, includeing high heat loads on Plasma Facing Components (PFCs), electromagnetic stress in the vessel, and runaway electrons, causing substantial damages. Therefore, it is crucial to develop effective methods to mitigate the damages caused by plasma disruption.
One innovative approach, called Shattered Pellet Injection (SPI), was initially proposed in 2009 [1]. This method utilizes in-situ cryogenic cooling technology to form impurity pellets within a pipe gun barrel. Subsequently, the pellets are accelerated by high-pressure helium (He) gas, fractured into ice fragments, and injected into the plasma. Preliminary investigations in DIII-D using deuterium material have demonstrated some expected advantages of SPI over the previously favored Massive Gas Injection (MGI). These advantages include strong density perturbation, similar mitigation of halo current and vessel forces, and deeper penetration [2]. Further experiments on DIII-D indicate that Ne SPI, when using similar quantities as MGI, results in faster and stronger density perturbation, much deeper penetration, and a lower conducted energy after optimizing the SPI system [3]. Subsequently, the SPI systems were developed on different tokamak devices and more experimental studies have been carried out or planned to carry out on, including JET [4], HL-2A [5], KSTAR [4,6], J-TEXT [7] and ASDEX Upgrade [8]. Additional experiments on DIII-D have shown that radiation energy increases with the injection of Ne quantities [9]. Experiments on JET have demonstrated that total radiation energy reaches saturation with increasing injected Ne quantities and leads to higher impurity assimilation with further increases [10]. Furthermore, in J-TEXT, adjusting the argon (Ar) pellet velocity can influence the scale of the Thermal Quench (TQ) and Current Quench (CQ) to achieve higher impurity assimilation [11]. These experimental results on different devices have strengthened the reliability of considering SPI technology as the baseline Disruption Mitigation Systems (DMS) technology for the ITER tokamak. However, further experimental research on SPI is needed to verify scientific conclusions, such as disruption mitigation with single SPI or SPIs injected from different toroidal locations in full-metal-wall devices, the impact of injection directions, runaway electron physics, and disruption mitigation at high electron temperature. Therefore, systematic experiments that vary impurity pellet injection parameters, injection locations, injection directions, and working schemes are necessary to investigate the disruption mitigation effect with SPI in Experimental Advanced Superconducting Tokamak (EAST).
The EAST tokamak is a non-circular, fully superconducting device with a major radius (R) of ∼1.83 m and a minor radius of ∼0.45 m. It has been upgraded with all the relevant auxiliary heating and current drive systems for ITER. These systems can deliver >30 MW of total source power, including Lower Hybrid Wave (LHW), Ion Cyclotron Resonance Heating, Electron Cyclotron Resonance Heating (ECRH), and Neutral Beam Injection (NBI), enabling plasma discharges with high electron temperatures (>5 keV). Moreover, EAST features an advanced ITER-like divertor configuration that allows for various divertor plasmas, including Lower Single Null (LSN), Upper Single Null (USN), and Double Null divertor (DN) configurations [12]. Consequently, EAST provides an excellent platform to study runaway electron physics and disruption mitigation under conditions of high T e and full metal PFCs.
The MGI technology in EAST has undergone several years of development and upgrades [13] for studying disruption mitigation [14] and runaway current [15]. In order to further establish the superiority and reliability of the SPI technique for disruption mitigation, a SPI system was successfully integrated into the EAST tokamak in 2022. A series of SPI experiments have been conducted to investigate its effectiveness in mitigating plasma disruption. Additionally, the injection results of SPI and MGI with high T e target plasmas and full metal wall conditions in EAST can provide further verification and identification of the advantages of SPI in supporting ITER DMS. This data support can contribute to the SPI database for future fusion reactor devices. This paper begins by describing the SPI system for EAST, the calibrated results of actual pellet injection quantities and essential diagnostics for disruptions in section 2. Section 3 presents the first rapid shutdown using Ne SPI in EAST. The comparison of disruption characteristics between SPI and MGI are introduced in section 4. Finally, section 5 offers a discussion and summary of the findings.

SPI system for EAST
In order to explore additional methods for disruption mitigation alongside MGI, a SPI test stand was constructed, and a series of bench tests were conducted to determine the pellet parameters [16]. Ne pellets with a diameter of 5 mm were formed at 8 K, and the length of the pellet was adjusted by varying the heating power of the heat sinks to ablate a portion of the pellet, achieving the desired size and the location of heat sinks are shown in figure 1(c). Three types of pellets were produced, consuming Ne gas quantities of 20 Pa m 3 , 25 Pa m 3 , and 30 Pa m 3 , respectively. These pellets were accelerated to velocities ranging from 100 to 400 m s −1 by adjusting the propellant gas quantities at a pressure of 4 MPa. For the fast valve, a solenoid valve used for SMBI on EAST was selected [17]. The pellet velocity could be modified by changing the pulse width of the valve while maintaining a fixed back pressure. To conduct experimental studies on disruption mitigation, the SPI system based on this test stand was integrated into the K port of EAST. The complete SPI system measures 2.3 m in length and 0.45 m in width. It mainly consists of a gas feeding system, a pellet producer, and a two-stage differential pumping system. The gas feeding system supplies Ne material gas for pellet formation and high-pressure He gas for pellet acceleration. Furthermore, the differential pumping system is employed to evacuate the propellant gas and minimize its impact on the plasma. The SPI system is positioned adjacent to the Fueling Pellet Injection system, utilizing a shared buffer tank, as shown in figure 1(a). After exiting the differential pumping system, the pellet travels through an 8 m long transmission tube before entering the shatter tube, which has a 20 • bend, as depicted in figure 1(b). The outlet of the shatter tube is located at R = 2.5 m and Z = 0.38 m. Ultimately, the pellet is fragmented and propelled into the Vacuum Vessel (VV).

Calibration of actual injected quantities into VV
Prior to conducting disruption mitigation experiments using SPI, the actual quantities of pellets injected into VV were calibrated. It should be noted that due to the pellet forming position being 10 m away from the shatter tube outlet and the utilization of transmission tubes with slight bends to guide the pellet into the plasma, there was a loss of pellets during injection. The cold zone was warmed by the heat wire of the cold head to around 20 K before the pellet launch, reducing adhesion between the pellet and the barrel's inner wall. The system then awaited the launch command. During this period, part of the pellet was ablated, but the gate valve remained open, and most of the ablated pellet gas was pumped into the buffer tank through the differential pumping system. Figure 2 shows that the partial pressures of Ne and He measured by the mass spectrometer after the pellets with different propellent gas were injected into the VV. It was observed that the He pressure increased with the increase of propellant gas quantity. However, the maximum pressure ratio of He pressure to Ne pressure is <3‰, which prove that the propellant gas introduced into the vessel is very small and has a minimal effect on the measurement of injected particles.
To calibrate the actual injected quantities of pellets, three types of Ne pellets were used, consuming Ne gas quantities of 20 Pa m 3 , 25 Pa m 3 , and 30 Pa m 3 (section 2.1). These pellets were injected into the VV without plasma, and their actual injected quantities were calculated using the pressure change in the vessel and the vessel volume (∆P·V). The calibrating results shown that the average actual injected quantities of the three types of Ne pellets are 7.5 Pa m 3 , 9.4 Pa m 3 , and 13.2 Pa m 3 , respectively. However, the loss of pellet mass along the guide tube cannot be presently evaluated due to measurement method limitations. Figure 3 illustrates the toroidal and poloidal views of the main diagnostics employed in this study. For disruption mitigation, a fast valve for MGI [13] with a valve response time of ⩽150 µs has been successfully installed on the middle port O (Z = −0.35 m) [14]. The injection duration of the MGI is 1-4 ms, depending on the back pressure and the duration is ∼2 ms for the experiments in this paper. Two fast visible Camera Diagnostic Systems (CCDs) situated at middle ports F and J, with sampling frequencies of 10 kHz and 50 kHz, respectively, are utilized to observe the trajectories of shattered pellets during injection. To measure the radiation distribution and total radiation power, a fast bolometer system covering the poloidal cross-section of the plasma is employed. This system utilizes Absolute Extreme Ultraviolet (AXUV) photodiodes and operates at a sampling frequency of 100 kHz [18]. Additionally, a 32-channel radiometer system, operating at a sampling frequency of 1 MHz, is employed to detect Electron Cyclotron Emission (ECE) from the plasma, providing indications of the TQ [19]. The divertor probes are utilized to measure the electron temperature and particle flux on the divertor target. It can achieve a temporal resolution of up to 0.02 ms [20]. Moreover, routine diagnostics such as Mirnov probes, with a sampling frequency of 50 kHz, are also employed. Figure 4 presents a typical rapid shutdown scenario utilizing Ne SPI. The shot corresponds to a H-mode plasma discharge with a USN divertor configuration. The pre-disruption plasma parameters include a plasma current (I p ) of 400 kA, a toroidal magnetic field (B t ) of 2.6 T, and an edge safety factor (q 95 ) of 6.9. The total auxiliary source heating power is 5.2 MW, comprising contributions from LHW (1.8 MW), ECRH (0.8 MW), and NBI (2.6 MW). The stored energy of the plasma is estimated to be 155 kJ. In this experiment, a Ne shattered pellet with an actual injection quantity of 13.2 Pa m 3 and a pellet velocity of around 340 m s −1 was used. The moment when the edge AXUV signal (channel 57) began to rise was used to indicate the arrival of initial impurity particles at the plasma edge and it was also defined as t arrival . Around 1.3 ms later, the core AXUV signal (channel 33) started to rise and then the strong core radiation occurred. Subsequently, the V loop increased rapidly, followed by a drop, and strong MHD activities were observed in the Mirnov signal. The cooling time is defined as the duration from when impurity particles reach the plasma edge until the plasma current (I p ) exhibits a spike, indicating the start of the CQ [21]. In this case, the cooling time lasted for 4.6 ms. After that, a TQ occurred, during which the core plasma temperature rapidly decreased within 0.1 ms, resulting in the almost complete loss of thermal energy in the plasma. Subsequently, the CQ phase commenced, and its duration was estimated to be around 10.8 ms, calculated from the time extrapolated from 80% to 20% of the plasma current. Notably, another spike in I p was observed during CQ. This spike occurred because the plasma in the early stage of CQ had not been fully disrupted and still possessed sufficient energy to ionize the injected impurity particles, resulting in a high level of core radiation. The impurity radiation accounted for only 20% of the magnetic energy loss during the CQ phase, while the energy from ohmic induction reheating could reach up to 40%, hence the occurrence of multiple spikes in the CQ phase of plasma disruption [22].

First rapid shutdown using Ne SPI in EAST
In order to provide a clearer understanding of the injection dynamics of fragments into the plasma during the triggered disruption, a series of images captured by two CCDs are presented in frames (a)-(g) of figure 5. These images depict key moments indicated by the blue dashed lines in figure 4. The upper images in figure 5 were acquired from the CCDs at the F port with a dt of 100 µs, while the lower images were obtained from the CCDs at the J port near the SPI with a time resolution of 20 µs. When the initial impurities reached the plasma edge, the edge AXUV signal increased gradually as observed in the AXUV data presented in figure 4, and the ablation light at the position of the shatter tube enhanced gradually. As more impurity particles were injected, a light spot was formed at the position of the shatter tube in frame (b), indicating the presence of initial impurity particles, including He propellant gas and Ne gas, which were ionized and confined to the plasma edge. The clear appearance of fragments was first observed at t arrival + 2.9 ms in frame (c). By t arrival + 3.5 ms, larger fragments had penetrated deeper into the plasma, as depicted in frame (d), coinciding with a prominent peak in core radiation resulting from the injection of fragments, as indicated in figure 4. As additional and larger fragments entered the plasma, their bright illumination also persisted in the CCD images, and the core radiation increased rapidly until reaching its peak, concurrent with the collapse of the electron temperature as shown in figure 4. It is worth noting that the impurity particles exhibited helical-structure movement like the results in JET [23], resulting in the helical upward movement of some ablation light form the shatter tube, which eventually distributed around the vessel. The helical transport of impurity particles can be clearly seen in frames (b) and (f ) of figure 5, from t arrival + 2.0 ms to t arrival + 4.6 ms. Finally, frame (g) portrays the image captured near the end of the CQ phase, where numerous small sparks (impurities on the inner wall) were observed to be sputtered from the inner wall of the VV due to the disruption of plasma confinement. Additionally, a shrink process of the plasma was also observed from t arrival + 2 ms to t arrival + 6.42 ms and this shrink became more obvious during CQ. This may be the reason for the decrease of the edge AXUV signal (channel 57) when the main fragments arrived.
A radiation contour map for discharge #116188 was constructed using the AXUV arrays to depict the temporal and spatial evolution of radiation following SPI, as illustrated in figure 6. Initially, a low-level radiation was observed at the plasma edge once a few impurity gases reached that region. Subsequently, as impurity particles accumulated, the radiation at the plasma edge began to rise, consistent with the observations in figure 5(b). Notably, the radiation region gradually moved toward the zone of Z = 0 starting from t arrival + 2 ms to t arrival + 2.8 ms in figure 6. However, the core AXUV signals did not increase significantly and the penetration processes of fragments were not also seen in the CCD images, so it may result from the helical transport of impurities and the edge radiation was picked up by AXUV signals near the zone of Z = 0 due to a chord-integral radiation measurement. And similar observations were also seen between t arrival + 2.8 ms and t arrival + 4.4 ms and the difference was that this higher radiation region extends to the zones of Z = −3.5-0 m, which was because the fragments penetrated into deeper plasma, and these were ablated and ionized, and transported poloidally at the fixed magnetic surface. To further verify that the material was deposited in deeper location, the penetration process of pellet fragments from the plasma edge towards plasma core was clearly observed from t arrival + 3.4 ms to t arrival + 3.48 ms by the CCDs, as shown in figure 7. These successive images overlaid with the Last Closed Flux Surface (LCFS) prior to SPI show that the ablation lights of the fragments were moving away from LCFS towards the plasma core. Meanwhile, this point was also evidenced by the observation that the distance between the light spot and the wall of the high field side on the images got shorter. As more and larger fragments penetrated into the plasma, TQ happened with a burst of the radiation and the peak radiation region were concentrated near the position at Z = 0.2 m. Moreover, a poloidal radiation asymmetry became apparent during the TQ phase. Following the initiation  of the CQ, the radiation intensity gradually decreased until the plasma shutdown.
Furthermore, it was also observed that during the CQ phase, some fragments continued to be injected into the plasma. This phenomenon could be attributed to the presence of a long pellet with a length of ∼15 mm, which contributed to the sustained core radiation observed in the early stage of CQ. Figure 8 illustrates the deposition of fragments closer to the plasma core compared to their distribution before CQ, which can be attributed to the reduced stored energy and the contraction of the plasma. The last fragments arrived at t arrival + 6.42 ms. The entire duration of impurity pellet injection into the plasma was 3.5 ms, with 1.7 ms occurring before the TQ. This indicates that only around 50% of the pellet fragments were injected into the plasma before TQ, which is contrary to the expected outcome. As a next step, we are planning to reduce the L/D ratio of the pellets in order to conduct disruption mitigation experiments and assess whether this approach can improve the ratio of pellet fragment injection before TQ.

Comparison of disruption characteristics between SPI and MGI
Currently, conventional DMS involve actively injecting a certain amount of particles (solid or gas) into the predisruption plasma to generate sufficient radiation power for dissipating the plasma energy [1]. Among the primary methods of impurity injection for disruption mitigation, SPI and MGI are widely utilized in experimental research, particularly SPI. To compare the effectiveness of SPI and MGI in disruption mitigation, two rapid shutdown experiments were conducted during H-mode plasmas using Ne SPI and MGI, respectively. These shots shared similar plasma parameters, including a LSN divertor configuration, pre-disruption plasma current (I p ) of 400 kA, toroidal magnetic field (B t ) of 2.7 T, and edge safety factor (q 95 ) of 6.1. The total auxiliary heating power was 5.1 MW, comprising LHW heating power of 1.5 MW, ECRH power of 0.8 MW, and NBI power of 2.8 MW. The stored energy was 140 kJ. In these experiments, the actual injected quantity of Ne shattered pellets for SPI was 9.4 Pa m 3 , with a pellet velocity of around 280 m s −1 . In contrast, Ne gas was injected into the plasma for MGI at a quantity of 10 Pa m 3 and a flight velocity of around 610 m s −1 . Due to the differences in flight distance and time between the pellets and Ne gas, the impurity particles reached the plasma edge at different times. To facilitate a convenient comparison between SPI and MGI in terms of their influence on the rapid shutdown of the plasma, the edge AXUV signal was also used to determine t arrival . Figure 9 illustrates the time evolution of selected plasma signals to provide insight into the plasma disruption process following the injection of impurity particles. Compared to MGI, SPI exhibited similar shutdown characteristics, and I p spike using SPI was nearly identical to that using MGI. However, the cooling time (3.9 ms) in the SPI shot was shorter than that (4.9 ms) in MGI, as observed in figures 9(a) and (b). Additionally, the CQ duration (7.38 ms) in the MGI shot was shorter than that (9.66 ms) in the SPI shot, indicating a higher decay rate of CQ in the MGI shot. And some experimental results also showed the CQ duration decreases with the increase of Ne pellet velocity [11]. Therefore, the reason that there was a higher CQ decay rate in the MGI shot might be that the pellet with very high velocity can become predominantly gas at shattering, which made it closer to an MGI. However, more experiments are still needed to compare the disruption timescales of MGI and SPI with the same plasma discharges and the effects of different fragment plumes on disruption mitigation. The core radiation using SPI was stronger than that using MGI, suggesting that SPI can deposit its material deeper into the plasma. Notably, the disruption mitigation using MGI also showed a second I p spike and the reason was the same as it using SPI in section 3. Moreover, the post-TQ plasma had a vertical displacement during CQ and touched the divertor, resulting in the heat flux on to it [14]. Figures 9(d) and (e) reveal that the electron temperature and heat flux near the outer divertor strike points were significantly reduced when using SPI. The peak electron temperature in the SPI shot was 50 eV, representing a reduction of 50% compared to the MGI shot (100 eV). Moreover, the peak heat flux in the SPI shot (6 MW m −2 ) was reduced by 40% compared to the MGI shot (10 MW m −2 ).
The deeper deposition of material in the plasma core during SPI led to higher core radiation prior to the CQ. However, contrary to expectations, the radiation power during the SPI shutdown was not consistently higher than that during the MGI shutdown, except for the peak radiation power, as depicted in figure 10(a). This discrepancy can be attributed to only a portion of the cryogenic pellet fragments, with a high injection velocity, being injected into the plasma core, resulting in strong core radiation and triggering plasma disruption. Moreover, fragments continued to be injected into the plasma during the CQ phases, consistent with the observations from the monitored CCD images ( figure 8). Furthermore, the total radiation power during both the MGI and SPI shutdown phases was significantly higher than that observed during an unmitigated disruption. Here, the total radiation power was calculated directly by volume-integral of AXUV measurement with the assumption of toroidal systemic radiation structure [24]. The absolute values of the total radiation power were corrected with the metal foil resistive bolometer by cross-talk calibration [25]. For disruption discharges, the data used for cross-calibration was from the current plateau stage due to the low time resolution of the resistive bolometer. At present, only a AXUV array in toroidal position can be used in EAST, so the results of the 3D-structure radiation were not considered in this paper.
The differences in poloidal radiation distribution between the two methods, MGI and SPI, are compared using radiation contour maps in figures 10(b) and (c). The high radiation regions observed with MGI were primarily concentrated in the zones of Z = ±(0. 18-0.35) m, with a distinct low radiation region in the zone of Z = 0 during the radiation burst. On the other hand, with SPI, the high radiation region was closer to the zone of Z = 0 m, extending from Z = 0.35 m to Z = −0.35 m. Noted that the measured radiation represents chord-integrated values and the contribution of edge radiation cannot be excluded, some evidences from the CCDs were used to show the fragments penetrated towards plasma core in section 3. Thus, these can indicate that the poloidal radiation distribution using SPI is more uniform and closer to the plasma core. This further supports the notion that impurity material using SPI can be deposited closer to the plasma core, facilitating the dissipation of core thermal energy.

Discussion and summary
A SPI system has been developed and integrated into the EAST tokamak for experimental studies on disruption mitigation. Prior to the experiments, the SPI parameters, including pellet velocity and actual injected quantity, were calibrated. The SPI system enabled the injection of impurity pellets with a diameter of 5 mm and velocities ranging from 100 to 400 m s −1 . Following the testing and calibration of the SPI system, the first rapid shutdown experiments using Ne SPI were performed in EAST, allowing clear observation of the entire process, from pellet injection to plasma disruption, through two CCDs. Prior to the CQ, it was observed that the impurity particles exhibited helical-structure movement, resulting in resulting in the helical upward movement of some ablation light form the shatter tube, which eventually distributed around the vessel. The radiation contour map revealed that radiation initially appeared at the plasma edge and then rapidly moved toward the plasma zone of Z = 0 m until the radiation burst during the TQ phase.
The disruption mitigation experiments using SPI and MGI were conducted, revealing several advantages of SPI over MGI. The SPI-triggered disruption exhibited a shorter cooling time, but a longer CQ duration. Moreover, SPI allowed for a closer deposition of impurity particles to the core plasma, resulting in stronger core radiation. The use of SPI also led to a reduction of 40% in electron temperature and 50% in heat flux near the outer divertor strike points compared to MGI during disruptions. While the peak radiation power during the SPI shutdown was higher, the total radiation power did not surpass that of the MGI shutdown. This could be attributed to only a portion of cryogenic fragments with high injection velocity being injected into the plasma core, triggering strong core radiation and disruption. Both SPI and MGI exhibited significantly higher radiation power compared to unmitigated disruptions. Additionally, a comparison of radiation contour maps between SPI and MGI demonstrated that SPI achieved a more uniform and poloidal radiation distribution closer to the plasma core.
In future work, upgrades of the SPI system will be implemented to achieve the formation of mixed pellets using D 2 /Ne or H 2 /Ne, and then disruption mitigation experiments with various SPI velocities and compositions will be conducted in EAST, which can provide valuable reference data for other devices and offering additional support for ITER.