Design of the shattered pellet injection system for ASDEX Upgrade

Effective mitigation of disruptions is the subject of ongoing research in tokamaks to support the design of the disruption mitigation system for the International Thermonuclear Experimental Reactor (ITER). Large machines with high thermal energy are at risk of suffering damage to plasma facing components due to concentrated heat loads. Vertical instabilities, as well as the current quench, give rise to induced halo and eddy currents in the machines structures, which result in strong electromagnetic forces.¹ Furthermore, strong electric fields as result of the current quench can accelerate electrons up to relativistic speeds² depending on the plasma density and current. Through collisions, these fast electrons transfer kinetic energy to slower electrons, causing an avalanche of secondary runaway electrons (REs), which, when losing confinement, can deposit their energy on a narrow area of the first wall, causing significant damage.³ 

The commonly used method of disruption mitigation is a massive gas injection (MGI). Large quantities of hydrogen/deuterium in combination with higher-Z noble gases (neon, argon, etc.) are injected into the plasma, of which a certain amount is assimilated into the plasma. This leads to radiation of the plasma thermal energy and a controlled decay of the plasma current, thus reducing concentrated heat loads, electromagnetic forces, and the generation of REs. This works well in medium and large size tokamaks,^4–7 but the mitigation of an established RE beam has proven difficult in a large tokamak⁸ due to a hot background plasma that prevents the mitigation gas from reaching the RE beam. Furthermore, on a tokamak with the size of ITER, MGI assimilation is not sufficiently effective to reach high free electron densities to suppress RE formation during the current quench.⁹ This problem can be overcome by injecting material in solid state, which can penetrate deeper into the plasma. The necessary material amount for RE mitigation would require the injection of large cryogenic pellets, which themselves can generate REs due to rapid cooling and the ensuing generation of electric fields.¹⁰ Also, beyond a certain volume/surface ratio, the pellet will no longer fully ablate while crossing the plasma. Therefore, the large pellets are shattered before entering the plasma to create a collimated stream of fast cryogenic fragments. These shattered pellet injection (SPI) systems have been investigated thoroughly in Oak Ridge National Laboratory^11,12 and at DIII-D.^13–15 Since the SPI is one of the prime methods for disruption mitigation in ITER, such systems have been implemented on different machines such as Korea Superconducting Tokamak Advanced Research (KSTAR),¹⁶ Huan-Liuqi-2A (HL-2A),¹⁷ J-TEXT,¹⁸ and JET¹⁹ to carry out experiments in preparation for ITER. In the existing experiments, typically one shattering angle and geometry has been used throughout. However, the question remained what the optimum distributions of fragment sizes and velocities to maximize the material assimilation would be. Therefore, ASDEX Upgrade (AUG) has also been equipped with a SPI system to add experimental data from an all-metal machine with the primary goal to deliver different SPI fragment plume characteristics via different shatter heads and a large flexibility in pellet speeds and sizes.

This paper describes the design of the SPI system for AUG with a special focus on the gas preparation system, the vacuum system, and the pellet guiding/shattering geometries. Furthermore, the setup for the laboratory tests and the setup in the torus hall are shown.

The SPI system for AUG can be divided into four main components: the gas supply system, the injector, the pellet guiding system, and the vacuum system. The system is designed to fit into the installation location in toroidal segment 16 in the AUG torus hall (Fig. 1) with the guiding tubes entering the vacuum vessel through the horizontal port Bo16/7. As the installation location is shared with other diagnostics, the installation space and, therefore, the size of the system are constrained. Furthermore, the system was to be commissioned in the laboratory first to carry out an intensive characterization of the fragment plumes created by the shattered pellets. Afterward, it had to be transported into the torus hall and lifted into position without losing the previously determined characteristics. Hence, the system is designed to be dismantled into the four sub-assemblies, which themselves can be positioned without being disassembled.

To move the system from the laboratory to the torus hall, the system was built on a custom aluminum trolley (Fig. 2) with rotatable heavy load wheels. The trolley also mimics the topology of the installation location in the torus hall. Hence, all major adjustments can be made during the laboratory tests, whereas only minor adjustments are necessary in the torus hall.

The entire system is built technically leak tight to allow for operation with potentially explosive gases. Connections between the major components are made by flexible stainless steel bellows or hoses to allow for movement during pumping and baking. Gas connections are all 1/4″VCR with stainless steel gaskets and vacuum connections are all CF standard with copper gaskets, except for the fore vacuum lines. Electrical insulations are placed in the gas feed lines into the torus hall and in the pumping lines to prevent large current loops. The control electronics for the gas supply and the pumps are located in control boxes close to their respective components, whereas the main PLC and trigger electronics are located outside the torus hall. Connections between the main PLC and the peripheral I/Os are made by optical fibers.

The pellet injector, built by PELIN LLC, is responsible for the pellet generation and acceleration. Its design is based on a previous system with 12 barrels built for TAE.²⁰ The AUG injector holds three barrels, which are connected to a cold head of a cryocooler inside a cryostat. The cold head cools a short section of each barrel, the cold cell, to a temperature of 5–8 K (depending on position inside the assembly). Gas that is injected from both ends into the barrel desublimates onto the wall of the cold cell, slowly forming a pellet that plugs the barrel. Heaters on both sides of the cold cell allow the regulation of the pellet length, and the barrel diameter determines the pellet diameter. The injector can be equipped with barrels of 1, 2, 4, and 8 mm diameter, and the cold cell allows the generation of pellets with varying length to diameter ratios of 0.5–1.5.

Each barrel is connected to a fast gas valve with an internal reservoir of 18 cm³ that can be filled with propellant gas up to 5 MPa. The valves are electromagnetically actuated by applying current to a coil around the valve stem. The resulting magnetic field moves the magnetic stem. Hence, the valves have to be protected from stray fields or operate at a safe distance from the tokamak. When a pellet is fired, the gas valve opens within 1 ms and remains open for up to 2 ms. The opening time, which can be adjusted, determines the amount of propellant gas released into the barrel. The gas pulse moves a hollow puncher, which detaches the pellet from the cold cell wall before the expanding gas accelerates the pellet along the barrel until it reaches the gate valve at the end of the barrel. The actuation scheme of the puncher driven by the propellant gas coming from the fast valve is shown in Fig. 3. By using hollow punchers, the detachment of the pellet from the cold cell becomes more reliable and predictable than using the gas to break off the pellet, especially at low propellant gas pressures. It is important to note that the bulk of the pellet acceleration gas enters the barrel through the hollow puncher after it scrapes the pellet off the forming cell, where it may have remained frozen to the walls for up to 30 min waiting to be fired. Depending on the pellet size, pellet material, and propellant gas and pressure, pellet speeds of up to 750 m/s can be achieved. Deuterium pellets can be released by the puncher alone as the mechanical hardness of solid deuterium is much lower than that of the puncher. Pure neon and neon mix pellets require a short pre-heating of the cold cell before opening the fast valves to decrease the mechanical hardness of the solid neon. Without pre-heating, the high hardness of the solid neon leads to increased wear on the cutting edge of the puncher, thus decreasing its lifetime.

The guard vacuum is generated by a Pfeiffer HiCube 80 Eco pumping station, consisting of a turbomolecular pump with a pumping speed of 0.067 m³/s and a membrane roughing pump with 1.8 m³/h. This assembly can ensure a guard vacuum pressure of 10⁻⁵ Pa which is monitored by a Pfeiffer MPT200 vacuum gauge. The cold head is cooled by a helium cryo-compressor Sumitomo F–50H. The helium lines, connecting the cold head to the compressor, are 20 m long, thus allowing the positioning of the compressor far away from the magnetic field of the tokamak, to reduce potential disturbance.

Flight tubes are used to ensure the delivery of the pellets from the three barrels of the injector into the vacuum vessel of AUG and shatter them before they enter the plasma. Furthermore, the system holds gaps for propellant gas removal and in-flight pellet observation, as well as the tokamak gate valve, which serves as a boundary between the tokamak vacuum and the secondary SPI vacuum. The flight tubes up to the torus gate valve are built onto an aluminum frame, so they can be moved as a single unit. The torus gate valve has its own support so that the SPI assembly can be installed without venting the vessel.

Once the pellets leave the gate valves of the injector (PV30–PV32), they enter one of three flight tubes. The flight tubes have a circular cross-section and are assembled in a triangular configuration in order to fit through the Bo16/7 port of AUG. The central axes of the flight tubes are distributed on a circle with a 48 mm diameter. Each flight tube consists of three separate tubes with gaps in between (Fig. 4). The first tube in a flight tube assembly, which is directly connected to the barrel gate valve with a CF16 flange connection, has an inner diameter of 12 mm and a length of 210 mm. After this tube follows a gap with 56.2 mm length in a CF160 double cross. The propellant gas can expand into the double cross and the attached 0.3 m³ vacuum tank, thus removing most of the gas from the flight tube. Following the first gap is the second tube in the flight tube assembly with an inner diameter of 14 mm and a length of 1800 mm. This tube runs through the 0.035 m³ expansion tank and ends in the torus gate valve. Here, the ends of the three tubes are welded into a plate, which is used to align the tubes in the gate valve and, thus, with the third tubes in the flight tube assemblies, which run inside the torus. This makes sure that the pellets can traverse the gap of 35 mm in the torus gate valve without being destroyed. Residual propellant gas expands in the torus gate valve, is largely reflected by the plate that is welded into the bellows behind the torus gate valve, and expands further into the 0.035 m³ expansion tank before being pumped. The bellows is necessary to compensate the vessel expansion during baking. A PEEK gasket between the torus gate valve and the bellows separates the electric potentials of the torus and the SPI. The third tube in the flight tube assembly has an inner diameter of 16 mm and a length of 1630 mm and directs the pellets from the torus gate valve to the shatter heads at the ends (Fig. 5). Inside the torus, the tubes are supported by a PEEK holder, which keeps the tubes apart and prevents current loops through the tubes. The stainless steel casing of the holder is bolted to the vacuum vessel to keep the tube assembly in place. The ends of the tubes are fitted with silver-coated M18 × 0.75 fine threads onto which the shatter heads are screwed. The length of the threads is 50 mm to allow adjustment of the radial shatter head positions and lock nuts to keep the heads in place. All tubes are straight, have a wall thickness of 1 mm, and are electrochemically polished. The tube lengths were chosen to maximize the distance between the injector and the tokamak for the given installation space in order to reduce the influence of the magnetic field on the cold head, the fast valves, and the guard vacuum pumps.

A selection of different shatter heads was manufactured for the SPI. Heads with circular cross-sections have an inner diameter of 16 mm and a wall thickness of 1 mm and were produced with shattering angles of 12.5°, 15°, 25°, and 30°. Long versions of these heads have a 78 mm long shattering section, whereas the short versions have a length of 40 mm. Shatter heads with square cross-sections have an inner side length of 21 mm, a wall thickness of 2 mm, come with angles of 12.5° and 25°, and a length of 77 mm. All heads have welded miter bends with non-penetrating welds to prevent leading edges inside the heads.

The vacuum in the SPI assembly is crucial for the pellet generation and for the operation of AUG. Furthermore, the pumping system has to remove the propellant gas, residual gas from the gas mixing and pellet formation, as well as pellets that sublimate in the barrels. Hence, the pumping system was designed to cope with significant amounts of gas.

Propellant gas is removed primarily in the first gap of the guiding system. It expands into the 0.03 m³ expansion tank from where it is pumped by a roots pump (Pfeiffer Okta 500 ATEX) with a pumping speed of 560 m³/h. The roots pump is connected by a 7 m long CF40 vacuum hose and through a corner valve (VAT 28432-GE11) to a dry screw pump (Leybold DRYVAC DV650 ATEX) with a pumping speed of 650 m³/h, which is installed in the basement of the torus hall. The fore vacuum pressure is monitored by a vacuum gauge (Pfeiffer MPT200). Residual propellant gas is removed in the 0.035 m³ expansion tank, which is evacuated by a turbomolecular pump TMP (Pfeiffer HiPace 300) with a pumping speed of 0.22 m³/s. Since the TMP is sensitive to magnetic fields, it is connected to the small expansion tank through a 2 m long CF 100 vacuum hose. This TMP is also used to reach the base vacuum pressure of 10⁻⁴ Pa, which is required for operation on ASDEX Upgrade. The TMP is also connected to the screw pump by a 7 m CF40 vacuum hose and a corner valve. During pellet injection, the fore vacuum pressure is too high for safe operation of the TMP so its corner valve is closed and pumped gas is compressed in the vacuum hose until the fore vacuum pressure has dropped to a safe value.

In the idle state, the pressure in the small expansion tank is about 10⁻⁴ Pa, whereas the pressure in the large expansion tank is in the order of 10⁻³ Pa due to the lower pumping speed of the roots pump in this pressure range compared to the TMP. The pressure difference is upheld by the low conductivity through the pipes connecting the CF160 double cross with the small expansion tank.

For explosion safety, the pressure of potentially explosive gases like D2 has to remain below the lower explosive pressure of 5000 Pa. The largest possible D2 pellet contains about 100 Pam³ of gas, and each fast valve can release about 100 Pam³ of propellant gas at maximum pressure. If all three pellets were to be fired and lost in the guiding system at the same time, the pressure in the large expansion tank would rise to 2000 Pa.

The AUG SPI is built to inject pellets containing either a single gas or a mixture of two or three gases with a precise mixing ratio. These gases are usually noble gases like neon and argon, as well as fueling gases like D2 and H2. The gas supply system is designed to prepare the gases for the pellet formation and the propellant gas for the acceleration.

For safety reasons, the gas cylinders are set up in storage units outside the torus hall building. For the commissioning and the first operational campaign, D2, neon, and argon were selected as process gases. The cylinders (200 kPam³ content each) are connected to their respective pressure regulator units (GCE MSLH0XDP), which are also installed in the outside storage. Each pressure regulator unit holds a pneumatic valve; a manual shutting valve; a manual blow-off valve and a pressure sensor on the high-pressure side; and a pressure sensor, safety valve and manual shutting valve on the low-pressure side. The unit for hydrogen is equipped with ATEX sensors (GCE H28397401), and the blow-off and safety valves are connected to an exhaust line, which leads into a safe area.

Three stainless steel 6 × 1 mm gas pipes run from the pressure regulator units into the torus and end at toroidal segment 16 of the vacuum vessel. The gas panel is located next to the injector to minimize gas inventories in the connecting tubes. The panel holds three pellet gas lines and three propellant gas lines, one set for each barrel, thus allowing parallel gas mixing, pellet formations, and injections. The pellet gas lines prepare the gases or gas mixes for the pellet formations. The neon and argon lines are directly connected to the pellet gas inlet valves, whereas D2 or H2 is taken from the propellant gas inlet valve. Since the D2/H2 feed is pressurized up to 5 MPa, a pressure regulator reduces the gas pressure to 0.2 MPa before it reaches its pellet gas inlet valve. The different gases are let sequentially into the pellet gas lines. When a pellet gas inlet valve is opened, the line before the electronic pressure controllers (EPC) is filled. Then, the EPCs (Bronkhorst P–802CI) fill the mixing volumes up to a desired pressure. For a gas mix, the line before the EPCs is then evacuated and filled with a second gas with which the EPCs fill the mixing volumes again until the partial pressures of both gases represent the desired mixing ratios. The EPCs have an accuracy of 3 kPa, and the usual filling pressure of the mixing volumes is about 100 kPa. The gas inlet into the barrels of the injector is done over solenoid control valves (Bürkert 2873), which regulate the inlet pressure to 8–10 kPa for D2/H2, 25 kPa for neon, and 40 kPa for argon, using downstream pressure sensors (Keller 23Sy-Ei) for feedback. Two normally open (N–O) valves at the end of each pellet gas line divide the gas stream, and flexible hoses guide the gas to the barrel inlet, which are equipped with 0.3 mm apertures to block backflow of propellant gas during pellet injection. Additional pressure sensors at the front barrel inlets measure the pressure drop during pellet formation when the respective N–O valve is closed. The pressure difference between the measurement before the N–O valves and at the front barrel inlets gives an indication if a pellet has formed in the barrel. Safety valves that open at a pressure difference of 40 kPa located at the front barrel inlets protect the barrels in case of sudden pellet sublimation. The safety valves are connected to the 0.3 m³ expansion tank. Evacuation of residual gas in the pellet gas lines is done by multiport valves after the EPCs and the solenoid valves, which connect the gas lines to the ½″ stainless steel pumping tube and eventually the 0.3 m³ expansion tank. The propellant gas lines have an EPC (Bronkhorst P–812CI) each that fill the internal reservoirs of the fast valves on the injector with the desired propellant gas pressure. These EPCs have an accuracy of 50 kPa, and the maximum propellant gas pressure is 5 MPa. Flexible hoses connect the EPCs with pneumatic valves attached to the fast valves. Pressure sensors between the pneumatic valves and the fast valves allow the measurement of the pressure drop in the fast valves during pellet injection.

The SPI system on AUG is capable of injecting three pellets simultaneously, either to a certain time into the plasma discharge or triggered by a plasma event, detected by the real-time control system. Since the plasma events in ASDEX Upgrade occur on a millisecond timescale, the timing of the injection must have an accuracy shorter than 1 ms. Hence, a fast trigger electronics was built that allows the selection of the trigger signal, delays the pulse start signal in case of a timing trigger, and interlocks the system with diagnostics that may be affected by the pellet injection. Furthermore, the SPI PLC interlocks the trigger electronics with the torus gate valve to prevent pellet launches against the closed gate valve. The PLC is built around a Siemens S7-1500 CPU, which is located outside the torus hall next to the trigger electronics. The SPI setup with the gas preparation system, the vacuum system, and the cryo-compressor in the torus hall is controlled by the Siemens I/O systems ET200iSP on the gas panel and the ET200SP above the screw pump. Optical fibers connect the peripheral I/Os with the central PLC, while a cable bundle transfers signals to the pressure regulator units outside the building.

The program and its interface (Fig. 6) are designed user-friendly with numerous fully automated processes, such as pumping start-up, leak check, and shutdown. The preparation of the pellet gases is also highly automated with the user’s only inputs being the desired gases and the partial pressures of gases to be desublimated. The system prepares the gas mixtures autonomously. The user can then adjust the desublimation time, depending on the pellet gas and size, and the barrel inlet pressure, which depends on the gas, before starting the pellet formation, which also runs without manual interference. After the pellet formation, the fast valves are armed in preparation for the plasma pulse. The triggering of the propellant valves is then done manually, by a timer or temperature trigger when heating up the cold cells on the barrels, or by an external trigger received from the AUG pulse timer system. This high level of automation simplifies the operation of the SPI system significantly, reduces the risk of errors, and enhances the accuracy and reproducibility of experiments.

For the success of SPI experiments, it is vital that the different pellet injections are comparable and reproducible. Therefore, knowing the pellet condition before it is shattered is of great importance. A fast camera (Phantom v2012), which observes the pellets as they traverse the first gap, serves this purpose. In the laboratory setup, the camera is installed above the CF160 double cross looking downward through a window flange into the gap in the three flight tube assemblies (Fig. 7). On ASDEX Upgrade, direct mounting of the camera is not possible, and therefore, a wound fiber bundle (image guide) relays the image into the shielding box holding the fast camera. The videos are recorded with a frame rate of 110 200 fps. With a gap length of 56.2 mm and a typical high pellet speed of about 600 m/s, the video still contains nine frames where the pellet is visible, allowing the determination of the pellet integrity and its trajectory. This high frame rate is only possible if the camera resolution is below 512 × 384 px. Hence, only one flight tube can be observed at a time to get a good view on the pellet. The camera is focused on the different flight tubes remotely using pneumatic cylinders (Festo DGC-32-52) on which the camera is mounted. They can move the camera between two end points in horizontal and vertical directions. As the upper flight tube is in between the two lower ones, a distance piece has to be manually installed in the actuator to focus on the middle flight tube. When launching multiple pellets into the same plasma discharge, a wider field of view is selected with a typical frame rate of 44 940 fps. The camera requires significant illumination at high frame rates. This is realized by a 200 W LED array (AURO Ariah PRO) mounted to a window flange on the side of the CF160 double cross.

A software package analyzes the recorded videos to determine the pellet speeds, sizes, tilt angles, and trajectories within the image frame, as well as provides information about potential debris, pellet break-up, hollow pellets, rotating pellets, etc.

The SPI system and the different shattering heads were characterized in the laboratory before moving to the torus hall. The setups in the laboratory and the torus hall were identical with the exception of the length of the gas tubes between the pressure regulator units and the gas panel (Fig. 7 background). A cubic vacuum tank (Pfeiffer 820KBH0750-G) with an edge length of 750 mm and an acrylic glass door (Fig. 8) was used as recipient. The Bo16 port was represented by a 1069 mm long CF100 tube. This tube length was chosen so that the ends of the guiding tubes reached just far enough into the cubic tank to install the shatter heads, thus giving the maximum space in tank for the observation of the fragment plumes. The backside of the tank was prepared to hold different background plates to improve the contrasts between the pellet fragments and the background. The observation was done by a second fast camera (Phantom v2012) at typical frame rates of 22 000–310 000 fps. Precise positioning of the camera was necessary to optimize the field of view and frame rate. Since the shattering heads with different shattering angles were tested in different distances from the camera, the camera had to be rotated around its optical axis and moved in all three directions. For this purpose, the camera was mounted onto a motorized rotation actuator (Festo ERMS-32-90), using an aluminum support that aligned the axis of the actuator with the axis of the camera. For the linear movement in vertical direction parallel to the tank door, the rotation actuator was fixed to a motorized spindle axis unit (Festo ELGS-BS-KF-60-600), which allowed a vertical lift of 600 mm. This unit was itself fixed to an identical spindle axis unit for horizontal movement. All actuators were equipped with position indicators with an accuracy of 1 mm or 1° and could be remote controlled. This assembly was mounted onto an aluminum frame, which was sitting on 2 m long rails bolted to the laboratory floor. This allowed movement of the camera toward the tank door to adjust the focal plane of the camera, as well as moving the camera out of the way when the tank door was opened. The illumination was done by two 200 W LED arrays (AURO Ariah PRO) mounted onto window flanges on top and at the bottom of the cubic tank. A third light source (DedoLight DLH400D) was later added in front of the tank door. The cubic tank was pumped by a turbomolecular pump TMP_test (Pfeiffer HiPace 300), which was connected to the dry screw pump of the SPI system for roughing vacuum. As the pressure in the tank could increase above a critical threshold for TMP_test during pellet injection, a gate valve between the tank and TMP_test and a corner valve behind TMP_test could be closed for protection, and a bypass valve would connect the tank directly to the screw pump. Pressure increases due to pellets and residual propellant gas occur on a millisecond time scale. Therefore, an analog vacuum gauge (Pfeiffer PCR280) with a measurement range between 10⁻³ Pa and 100 kPa on the backside of the tank was used for monitoring the tank pressure.

The first step of commissioning the SPI system in the laboratory included tests of the cryo and vacuum systems. After successful evacuation and cool down of the cold cells, the pellet formation was tested with neon and D2. These test pellets were launched with helium propellant as the use of high-pressure D2 was limited in the laboratory. No shatter heads were installed for these first pellet launches to observe if the pellets arrived in the test tank intact. After small adjustments in the first gap, all pellets reached the test tank unscathed (Fig. 9).

For the test of the propellant gas removal, the fast valves were triggered with 2 MPa reservoir pressure with a 2 ms opening time. This was done with and without pellets in the barrels. The vacuum gauge at the test tank was connected to a fast oscilloscope to record the pressure signal. The time trace is shown in Fig. 10. As shown, the pressure increase due to the propellant gas alone is below the sensitivity of the vacuum gauge at this pressure, while the pressure increase due to the sublimating pellet is clearly distinguishable from the base pressure. Hence, the propellant gas removal is good enough for operation on ASDEX Upgrade. The graph in Fig. 10 also shows that the pressure increases to a value of about 40 Pa, which corresponds to the injected pellet with 4 mm diameter and 7 mm length, and an inventory of 16 Pam³ neon. Considering measurement uncertainty of 15% in the vacuum gauge at this pressure range, it can be assumed that the pellet material loss is of this order or smaller.

Pellets injected into the tokamak vacuum vessel may miss the plasma due to wrong timing, unexpected disruptions, or for testing purposes. These pellets will impact a region of the inner heat shield of ASDEX Upgrade, which is equipped with ferritic steel tiles.²¹ Since the neon pellets have a significant mechanical strength²² and carry kinetic energies of up to 33 kJ, their impacts may lead to deformations of the tiles. This was investigated by placing a heat shield tile in the flight path of unshattered pellets on the opposite side of the flight tubes in the test tank. Neon pellets with 8 mm diameter and 8 mm length were then fired at the target tile with propellant gas pressures of up to 3 MPa, reaching speeds of 390 m/s. The impacts were observed with the fast camera at the test tank (Fig. 11). Inspections of the tile after its retrieval from the test tank revealed no observable alterations of the tile surface geometry. Only small slightly discolored spots were found, which could be attributed to a change in the surface roughness, where the pellet hit the tile. The test confirms that the injection of pellets into the empty tokamak vessel poses no threat for plasma-facing components.

After the completion of the laboratory tests in mid-October 2021, the injector was disassembled into the four main components, transported to the AUG torus hall, and lifted into position in Sector 16. The cryo-compressor was placed next to the east wall of the torus hall to have the maximum distance to the AUG magnetic field coils. Taking the results of the laboratory test into account, it was decided to equip the ends of the guiding tubes with the square shatter heads with 12.5° and 25°on top and on the left, as well as the short circular shatter head with 25° on the right (Fig. 12). After installation, the exact positions of the heads were measured. Since the access to the vacuum vessel is necessary for the installation of the shatter heads, this is the assembly used for the 2021/22 experimental campaign. The injector is set up with 4 mm barrels on all three lines, although these can be exchanged during the experimental campaign. The setup is summarized in Table I.

The commissioning of the system on AUG took place in early November 2021. The system was successfully leak tested, evacuated, and cooled down. The magnetic field of AUG showed no effect on the SPI systems. Between closing and baking of the AUG vacuum vessel, the first neon pellets were generated and injected into the empty vessel with helium propellant at 2 MPa. The fast valves were also triggered without pellets to check the propellant gas removal. The gas in the vessel vacuum was observed with a residual gas analyzer while the pumping gate valves were closed. The neon pressure in the vessel spiked to a value of 0.3 Pa during pellet injection, which corresponds to the injected pellets with 4 mm diameter and 7 mm length in the 40 m³ vessel, whereas the helium pressure only rose 0.02 Pa. Without pellets, the helium pressure in the vessel reached 0.04 Pa. This corresponds to 0.022% and 0.044% of the propellant gas.

For the final pre-operation test, neon pellets were injected into a helium glow discharge and the events were observed by a fast camera. A bright flash was observed a few milliseconds after the pellet trigger at the location of the shatter heads. Among other flashes in the glow discharge, caused by dust particles, this flash can be attributed to the shattered neon pellet due to its timing and location. This served as proof that the SPI setup on AUG is able to inject shattered pellets, and the vessel baking could proceed.

Further tests following the ASDEX Upgrade vessel baking confirmed the excellent removal of propellant gas. When a single fast valve is fired without a pellet, the torus pressure increase indicates that only about 0.035% of the propellant makes it into the AUG vessel. This shows that the propellant removal works effectively even at high AUG vacuum.

On 2021 December 14, SPI was fired into a plasma discharge for the first time. Pellet release was triggered by the tokamak control system at a pre-programmed time. Each barrel was tested individually, and also dual injection was demonstrated. In AUG pulse No. 39895, two 4 mm diameter, 7 mm length 100% deuterium pellets were injected at speeds of ∼570 m/s (Fig. 13). The first pellet was fired through the 12.5° rectangular head, whereas the second one followed 20 ms later through the 25° short circular head. The second injection leads to a disruption.

A new SPI system for AUG has been designed, built, tested, and commissioned. The SPI can produce and inject three pellets simultaneously through three separate barrels and flight tubes. The pellet diameter can be selected from 1, 2, 4, and 8 mm, and the pellet length can be adjusted between 2 and 11 mm. Pellet speeds between 60 and 750 m/s can be reached. Each flight tube is equipped with its own, exchangeable shattering geometry, allowing the study of the impact of different fragment plume characteristics on the disruption mitigation process. The system was built in a laboratory environment, characterized extensively using a special test setup with a target vessel, and transported to the AUG torus hall after completion of the tests. The SPI was successfully installed and commissioned on AUG.

The authors are grateful to A. Bolland, M. Gruber, D. Bösser, and the ASDEX Upgrade Team for their technical help in assembling, commissioning, and operating the AUG SPI system.

The work has been performed as part of the ITER DMS Task Force program (Grant No. IO/CT/43-2084). The ASDEX-Upgrade SPI has received funding from the ITER Organization. The views and opinions expressed herein do not necessarily reflect those of the ITER Organization.

The authors have no conflicts to disclose.

M. Dibon: Conceptualization (lead); Investigation (equal); Writing – original draft (lead). P. de Marne: Investigation (equal); Software (lead). G. Papp: Investigation (lead); Project administration (lead); Supervision (equal); Writing – review & editing (equal). I. Vinyar: Conceptualization (equal). A. Lukin: Conceptualization (equal). S. Jachmich: Data curation (lead); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Writing – review & editing (equal). U. Kruezi: Conceptualization (equal); Funding acquisition (equal). A. Muir: Conceptualization (equal). V. Rohde: Conceptualization (equal); Writing – review & editing (equal). M. Lehnen: Funding acquisition (equal). P. Heinrich: Data curation (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Writing – review & editing (equal). T. Peherstorfer: Software (equal); Validation (equal). D. Podymskii: Conceptualization (equal).

The data that support the findings of this study are available from the ITER Oranization. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the authors upon reasonable request and with the permission of the ITER Oranization.