Design of a large nonevaporable getter pump for the full size ITER beam source prototype Free

The full-scale beam source of the ITER heating neutral beam injectors¹ (HNBIs) is being developed at the neutral beam test facility.² The neutral beam injector exploits the higher neutralization yield at high energy of a precursor negative ion beam; however, the creation of a negative ion hydrogen beam is rather complex, requiring a plasma with relatively high electronegativity, which this is commonly achieved in a cesiated hydrogen discharge with proper magnetic field configuration.³ The surface conversion mechanism on low work function surfaces,⁴ from hydrogen atoms to negative ions, was found to be the most effective approach.⁵ The targets of such devices, based on the Max Planck Institute for Plasma Physics (IPP) design of RF-driven ion source,⁶ are to demonstrate a stable, high extracted current density for the ion beam (1 h, 355 A/m² H⁻, 285 A/m² D⁻) with a rather uniform distribution over the 1280 apertures, each one accelerating an ion beamlet up to an acceleration energy of 1 MeV. The SPIDER⁷ experiment was constructed and dedicated to the purpose of testing and optimizing the ion source, developing proper design solutions and procedures for obtaining a uniform plasma and cesium layer. This can be achieved before adding the complexity of the high-voltage multielectrode acceleration. A low-energy multiaperture triode accelerator extracts negative ions from the ion source, up to a beam energy of the order of ∼100 keV, aiming at a total negative beam current of about 70 A. A vacuum-insulation concept, in opposition to gas-insulation configurations that exploit SF₆ to maximize the voltage holding (see for instance, Ref. 8), was applied for the first time. This means that the RF-driven source is fully contained in the vacuum vessel; as the ion beam propagates inside the vessel at ground potential, the beam source has to be biased at the acceleration voltage required for the ion beam (with opposite polarity) with respect to the vacuum vessel. Soon after the beginning of operations in 2018, it was found that the hydrogen pressure in the vessel was a direct cause of unwanted RF-induced breakdowns on the outside of the plasma source. As a temporary solution, the conductance between the ion source chamber and the vessel was reduced, by limiting the number of apertures of the accelerating electrodes,⁹ therefore, reducing the vessel pressure for a given source filling pressure. This solution allowed plasma discharges at a high RF-power of 400 kW, finally demonstrating the negative ion beam formation with the use of cesium.¹⁰ This paper describes the conceptual design of an enhanced pumping system, whose performance shall allow the operation of the SPIDER source with all beamlet apertures. The experimental characterization of gas conductance between the ion source and the vessel was carried out,¹¹ determining a conductance of about $C≈30 m 3 / s$ for hydrogen (H₂); at the same time, a pressure threshold of about 40 mPa was derived from dedicated tests characterizing the discharge probability versus pressure and RF power.¹² From those measurements, the required additional pumping speed reads $S a d d=C( p s − p v)/ p v− S 0$⁠, with $S 0$ being the presently installed pumping speed (of the order of 80 m³/s for H₂), and with $p s, p v$ indicating the source and vessel pressure, respectively. An additional effective pumping speed of the order of about 200 m³/s is necessary to allow operation with $p s$ of 0.3 Pa or above.

The nonevaporable getter (NEG) technology was selected for the high affinity for hydrogen, also in terms of the pumping speed, the large specific capacity for hydrogen of recently developed sintered components, and the possibility to design highly modular pumps based on relatively small cartridges, composed of sintered getter disks. The use of NEG pumps in fusion research facilities was considered^13,14 since the 1970s. Scaling NEG pumps to larger assemblies was demonstrated in recent research and development activities,¹⁵ deeply investigating the basic module of one getter cartridge to be replicated in large pumps,¹⁶ after showing that the pumping speed of the newly developed getter alloy, ZAO®,¹⁷ is rather stable over the hydrogen load¹⁸ and does not present embrittlement or peeling even for thousands of sorption/desorption cycles reaching relatively high hydrogen concentrations (>1000 Pa m³/kg). NEG pumps are commonly used in UHV regimes, however, the peculiarity of the ZAO® alloy makes it the best choice for the application to higher pressures in the HV regime, such as those required for our application, with hydrogen pressure in the vessel $p v$ of the order of tens of mPa during operation. Nonideal conditions¹⁹ were also considered for the getter operation, and a pump composed of multiple cartridges was recently tested.²⁰ 

Getter materials store indeterminately the getterable species (O₂, N₂, and H₂O), as they are not released²¹ when the getter is heated to release the stored hydrogen. In comparison to other capture pumps, such as cryogenic pumps—which release hydrogen when regenerated, but in general, also other adsorbed species—this is a unique feature. Such a feature is particularly interesting for an environment in which the passivation of cesium-adsorbed atoms inside the plasma source would occur, in case impurities are released during pump regeneration such as oxygen and water, thus implying an additional conditioning effort for reactivating the cesium layer.²² In addition, NEGs have a very controllable release of absorbed hydrogen in comparison to other capture pumps, as they absorb hydrogen at low temperatures (RT-150 °C/200 °C) and require to be heated to high temperatures for releasing the stored hydrogen (400–600 °C), as shown in Fig. 1(a). However, for this very same peculiarity, the integration into existing facilities of the NEG pumps imposes engineering challenges, to avoid overheating during the NEG regeneration procedures of in-vessel components, which were not designed originally for such thermal loads. The pumping speed of a single cartridge that will be considered here is shown in Fig. 1(b), as a function of the hydrogen load. It was extrapolated from a set of measurements performed at RT on the NEG cartridge, which itself is composed of about 270 sintered ZAO® disks of 2 mm thickness, distanced 1 mm one from the other, piled in six equidistant columns with a heater at the center. A rigid cage with a large open area provides structural support to the assembly, with a minimal decrease of the pumping speed, but a quite effective reduction of the heat exchange by radiation, thus reducing the power required for regenerating the getter disks. The geometrical details and the influence on the pumping performances are discussed in Ref. 16. The mass of NEG alloy per cartridge is about 0.92 kg. The sintered disks based on ZAO® alloy were demonstrated to withstand thousands of loading cycles up to 1300 Pa m³/kg, without causing embrittlement or peeling-off of the disk,¹⁸ and they have, in general, improved mechanical performances as well as good pumping performances with respect to nonsintered NEG elements. The design of the heater, based on a tantalum filament supported by a ceramic material, was verified to allow cyclic operation of the NEG pump with disks reaching 550 °C without filament failure.

Figure 1 shows the three main characteristics for the application of ZAO-based NEG pumps: regeneration, the pumping speed variation with the hydrogen concentration, and the influence of impurities. The regeneration cycle in Fig. 1(a) exemplifies how the sorption phase at low-temperatures moves along a pressure isotherm $p=k(T) c 2$ (dashed lines in the figure), up to a concentration at which the NEG is heated and a different pressure isotherm is reached; if auxiliary pumping is applied, the concentration can decrease, and the NEG is ready for hydrogen absorption after cooling down to a condition that satisfies $p< p v$⁠. The results obtained on a single cartridge for the pumping speed, and the design requirements for equipping large clusters of cartridges, were demonstrated in recent tests,²⁰ from which the curve of Fig. 1(b) is extrapolated. The presence of possible air leaks in the large vacuum system of SPIDER, or in general, of a relatively high background gas pressure with respect to ultra-high vacuum conditions which are the common application for NEG pumps, was also considered beforehand. Dedicated exposure tests of small NEG cartridges were carried out during the recent operation of SPIDER, by connecting a chamber equipped for pumping speed measurements to the SPIDER vessel. The performances of a small NEG cartridge were measured after exposures of various durations, yielding the effect of background gas on the passivation of the getter in terms of pumping speed for hydrogen shown in Fig. 1(c).

During those tests, the single NEG cartridge could be tested for pumping speed in a small vacuum chamber, then connected to the SPIDER vessel, in order to expose the cartridge to the SPIDER gas background, and then again tested for pumping speed for hydrogen in isolated configuration. The pumping performance of the tested cartridge and the background gas composition were used to scale-up the pumping performance of the large NEG pump exposed to the SPIDER background gases, as shown in Fig. 1(c). During these repeated tests, the small vacuum chamber atmosphere was monitored by a residual gas analyzer. Figure 2 reports the spectrum analysis, indicating the presence of a significant air leak in the main vessel, which however, allowed to a certain extent, a successful operation with cesium of the ion source.¹⁰ This condition was anomalous, and the detected leak will be solved in future operations; however, for the design of the pump, we considered the possibility that such an off-normal condition might happen again. In this sense, Fig. 1(c) provides the lower limit to NEG resilience over time to the SPIDER background pressure as included in the new pump design.

The three main characteristic curves describing the NEG alloy shown in Fig. 1 were considered in the design of the large NEG pump for SPIDER, as well as all constraints coming from the pre-existing in-vessel components and of the structures of the facility, as detailed in the following.

The modularity of the SPIDER vessel, which is composed by two cylindrical modules of 4 m diameter and 4.3 m in total plus two lids, was exploited to integrate the pumping system in a new vessel module, to be installed behind the ion source. The length of the extra volume is limited by the extension of the existing bioshield, and the new vessel module was designed with a length of 0.6 m. A water jacket is integrated with the cylindrical part of the new vessel module, to protect from thermal deformation due to overheating and to preserve the O-rings sealing the large ISO-K flanges. The getter pump was conceived with a modular design constructed around the getter cartridges, as shown in Fig. 3(a). From an engineering point of view, modularity simplifies the support structure, the assembly procedure, as well as the electrical connections, as represented in Figs. 3(b) and 3(c). In addition, it allows exploiting the already available characterization of the performance of a single cartridge and cartridge assemblies.

A total of 384 cartridges, protruding from two walls facing each other, are arranged, grouped, and cabled in sectors, totaling 32 precabled panels, as shown in Fig. 4(a). Room is reserved for the possible installation of 128 cartridges, further [as sketched in Fig. 3(b)]. To minimize thermal radiation, intermediate stainless steel sheets with polished surfaces were introduced between the cartridges and the support structure. As the gap between the modules greatly influences conductances toward cartridges that are installed at the outer radial positions, the pump structure was designed with an axial extension greater than the length of the vessel module, protruding inside the lid, as shown in Fig. 4(b). A stiff stainless steel frame, Fig. 4(b), was designed to support the weight of the pump module, guaranteeing at the same time, the expected thermal deformation for the pump structure during regeneration at a high temperature. A thermal shield, composed of a copper plate with brazed stainless steel water-cooled pipes, was designed to protect the beam source from direct thermal radiation, as shown in Fig. 4(c). As the protection from radiative heat exchange also means reducing the gas conductance, the geometry was developed as a compromise. The surface treatment of burnishing was defined to maximize the surface emissivity of the water-cooled shields and to minimize the contribution of reflections on radiation reaching the beam source.

In neutral beam test stands dedicated to technological developments, the occurrence of water leaks is possible. For the protection of investment, the NEG disks will be operated at RT during the beam operation, i.e., during the sorption phase. Due to progressive passivation, occurring due to gas species that are chemisorbed on the getter surface, the pumping speed for hydrogen would be progressively reduced. It is found experimentally that the pumping speed is slightly higher when the getter disks are maintained at relatively high temperatures, i.e., between 125 and 200 °C. However, in the event of water or air leak, the high reactivity of the getter surface obtained in these conditions would also cause deeper surface passivation, corresponding to several controlled passivations at room temperature, or in the worst case, a condition in which the pumping speed for hydrogen cannot be fully recovered by the reactivation procedure. Using the pump at RT in the absorption phases minimizes the potential damage in case of a water leak, which might be a concern at the beginning of operations at high beam power. On the other hand, to minimize the impact of background species [already shown in Fig. 1(c)] and maximize the pumping speed in operation, short heating cycles are envisaged. These short reactivations of the pump shall be carried out overnight, as indicated in Fig. 5, not necessarily aiming at desorbing the full amount of hydrogen absorbed during the daily experimental session. Long regeneration cycles can be carried out during the weekends, to desorb the amount of hydrogen cumulated during the week. During regeneration, the NEG disks are maintained at high temperatures and their reactivity can be very high. Avoiding air or water leaks during these phases is of the utmost importance: for this reason, the cooling circuit for the thermal shields has been designed with particular care on the leak tightness, and double windows are required for diagnostic ports facing the NEG pump, where relatively high temperature is expected. As discussed previously, the NEG pump does not release hydrogen in the case of a power outage; in principle, this might allow to exceed the limit on the hydrogen amount inside a vacuum vessel, which shall be considered to avoid explosive mixtures for instantaneous release of hydrogen and, in the meanwhile, a large air leak occurring in the vessel. In the case of SPIDER, however, to maximize the overall pumping speed and the pumping capacity for all species, the cryogenic system is planned to operate in parallel. With a safety margin, the maximum hydrogen amount for the absorption and regeneration cycles of the NEG pump is taken as 240 kPa m³. The total amount of hydrogen determines the highest specific concentration: for the case of 384 cartridges, it corresponds to about 680 Pa m³/kg, a relatively low concentration for the ZAO® alloy. This gives an important constraint for the operation of a NEG pump, because desorbing hydrogen at low concentration is more difficult (i.e., achievable at a slower rate) than desorbing at high concentration, as indicated by the isotherms in Fig. 1(a). The existing pumping system based on cryogenic pumps can be operated simultaneously with the NEG system; due to the very large NEG capacity for hydrogen, the second stage of cryogenic pumps (i.e., the one kept close to liquid-helium temperature) can be heated at the end of each session, to transfer the adsorbed hydrogen to the NEG pump. Such regeneration of the cryogenic panels will be performed with a controlled temperature increase, with a rather low equilibrium pressure in the vessel, thanks to the large pumping speed for hydrogen of the NEG pump. The second-stage regeneration procedure was tested during the recent operation of SPIDER with cesium, regularly demonstrating a controlled release of 1.5–2.5 kPa m³ during about 8h, with a stepped increase of the cryopanel temperature. The controlled release was bound to maintain a vessel in the order of a few Pa, at which the pumping speed of turbopumps was already lowered to about 3 m³/s. As the pumping speed of the NEG pump is about a hundred times larger, a similar procedure might be shortened greatly, with hydrogen amount to be transferred of the order of 20/30 kPa m³, as the pressure limit for the operation of turbopumps does not apply to the NEG pump.

We will discuss a scenario for the operation of the SPIDER vacuum system, covering four days of operation, with 384 installed cartridges and seven cryogenic pumps of nominal 10 m³/s pumping speed each, with a source filling pressure of 0.3 Pa (corresponding to about 8.4 Pa m³/s of injected throughput). The source is operated on Mondays with short pulses aiming at recovering cesium conditioning and high voltage conditioning, and on the remaining three days, it operates short pulses and two one-hour pulses. Figure 5 shows the gas load on the NEG pump during the first two days of operation; 35 conditioning pulses of 6 min each are performed on the first day, with a duty cycle of one. On the second day, ten short pulses are replaced by two long pulses of 1 h each. At the end of each day, cryogenic pumps (onto which a gas load proportional to their effective pumping speed was delivered) are regenerated; immediately after, the temperature of the getter disks is increased to a regeneration temperature. Due to the very low initial concentration in the pump, equivalent to about 95 Pa m³/kg, the amount of desorbed hydrogen during the first night is very low. As mentioned, the short reactivation at night is envisaged to remove the possible effect of background gases over time (H₂O, outgassing, minor leaks) and avoid the slow decrease in the NEG performance independently of the hydrogen load; therefore, a regeneration temperature of the order of 450 °C can be considered. In the scenario discussed in the following, a temperature of 550 °C has been assumed for both the short regeneration during nights, and for the long regeneration which is aimed at reducing the hydrogen concentration. Such a regeneration temperature has to be considered as an effective temperature, describing the overall condition of the getter disks (i.e., local nonuniformity of the disk temperatures is neglected). For the regeneration phases, a reduction of the auxiliary pumping speed by turbopumps down to 3.1 m³/s from the nominal 9 m³/s was defined, in order to consider their pumping speed characteristics at relatively high pressures. In this sense, a safety margin was taken: only the presently installed turbopumps (9 m³/s in total for H₂) were considered, while the installation of additional four turbopumps is planned (totaling 4.6 m³/s for H₂), see Sec. III D. Furthermore, the peak vessel pressure during regeneration is used to obtain the pumping speed, which is an overestimation as it may rapidly decrease along the getter isotherm when the concentration starts to decrease. During regeneration, a vessel pressure up to 2 Pa is expected in the case of a relatively high hydrogen concentration. NEG regeneration can be carried out by steadily increasing the getter temperature, so as to maintain the vessel pressure at an optimal throughput for the auxiliary pumping system (for instance, around 1 Pa).

Figure 6 shows that the weekly operation of the SPIDER beam source can be sustained even in a high-availability scenario, as long as the weekend is fully dedicated to the regeneration of the NEG pump. At the same time, the pumping speed decreases due to the hydrogen concentration [see Fig. 1(b)] as well as because of the influence of the background species [assumed equal to the reference measurements of Fig. 1(c), which were taken in the presence of a well-known air leak]. It is possible to notice that a certain amount of hydrogen remains present at the beginning of the week: this amount is of the order of 95 Pa m³/kg, close enough to the concentration of 65 Pa m³/kg down to which NEG pumps are commonly regenerated. In the case the vessel has to be opened for maintenance of the beam source, the pump will be dismantled from the vessel and isolated in a dedicated vessel for storage, in a controlled nitrogen atmosphere or a vacuum for protection.

As shown in the bottom part of Fig. 6, an effective pumping speed between 200 and 250 m³/s was estimated. The relation between the installed pumping speed and the effective pumping speed considered in this calculation will be discussed in the following section using gas flow simulations in the molecular regime, to determine the influence of thermal shields.

The described load cycle was determined for an optimistic case of high availability of the SPIDER source, for a rather heavy cycle of conditioning and long pulses. At present, the ion source is operated for no more than one hour of beam pulses per day. However, the maximum regeneration temperature, required to operate the pump without impact on the SPIDER operation even in such an optimistic case, was determined as never exceeding 550 °C. In the following sections, an effective temperature of 550 °C will be considered for the integrated thermomechanical analysis of the pump.

A three-dimensional gas flow model in the molecular regime,²³ based on the view-factor analogy, was used to study the gas pressure distribution in the SPIDER vessel and the effective pumping speed of the NEG pump, which is limited by the presence of thermal shields. A pumping speed of one cartridge of 0.87 m³/s was applied, taken from experimental characterization at RT and a H₂ concentration of 27 Pa m³/kg (i.e., the beginning of gas injection, while at 625 Pa m³/kg, it is expected to be about 25% lower). The case of a filling pressure of 0.283 Pa was simulated by imposing on the accelerator side of the source, the equivalent gas fluxes (over the grounded electrode and the lateral gaps of the accelerator, with the partitioning calculated previously¹¹). At the vessel ports, the boundary conditions relevant to eight cryogenic pumps are considered (and of the six, less effective, turbomolecular pumps). As shown in Fig. 7(a), the structure of the beam source and the thermal shields, as well as the vessel ports, are detailed in the simulation. Arrows of different colors are used to indicate, in the figure, various definitions for an effective pumping speed, depending on the position at which the pressure is defined. As expected, the pressure on the lid where the NEG pump is installed is lower, so the pumping speed S_lid is the highest (255 m³/s). The pumping speed defined on the rear side of the source S_rear confirms the capability of the pump to reduce the pressure below the thresholds, to minimize the probability of RF-induced breakdowns. Finally, the hydrogen pressure on the other side of the vessel could be rather higher, and S_front is found to be as low as 155 m³/s. It has to be mentioned that the rear surfaces of the ion source are realized with a punched sheet, with a certain transparency, which has been neglected here in a conservative approach. Two cases were simulated: one with 384 NEG cartridges, and one with 512. The gas load on the getter cartridges might be nonuniform, depending on their position and mutual shadowing with respect to the preferential flow direction, which should proceed radially into the annulus of the NEG cartridges. Figure 7(b) shows the calculated variation of gas throughput on the cartridges depending on their position, with respect to the average. Thanks to the spacing between the two layers of cartridges, and their loose pattern and distancing, all cartridges contribute equally to the pumping speed, with a slightly higher contribution from the innermost layer.

Figure 8 shows the extrapolation of the simulation results to a varying source filling pressure, showing that the requirement on the vessel pressure could be satisfied even for a source filling pressure up to 0.55 Pa. This will allow a larger window for characterizing the beam source performances even at pressures higher than the nominal value. The relatively higher pressure expected on the front side of the vessel is not an issue for the propagation of the negative ion beam, as the charge state of the beam particles does not modify either the beam optics or the beam power, which are the quantities to be measured by beam diagnostics²⁴ to define the performance of the ion source.

As long regeneration phases are necessary during the weekends, with getter temperatures of the order of 550 °C to be maintained during 20h, there is the risk of overheating delicate parts of the in-vessel components of the beam source (e.g., diagnostics and insulating parts). Each cartridge is protected by an external SS cage which has roughly 50% transparency. Let us use the temperature of the external cage of the getter cartridge and obtain the correct value of radiated power. Experimentally, for a NEG temperature T_NEG = 550 °C, a grid temperature of T_grid= 415 °C was found, when inserted in a large stainless steel vessel maintained at about 50 °C. The surface emissivity of the getter disks was characterized to be ɛ_NEG = 0.8, while the emissivity of the grid can be taken as ɛ_grid = 0.2. As the gross area of the cartridge A_NEG+ A_grid is about 0.058 m², one can calculate the effective emissivity of the cartridge taken as a hexagonal prism,

The effective emissivity ɛ_eff is about 0.65, when T₀ = 50 °C and the nominal regeneration temperature is applied. In a preparatory multi-cartridge setup, at steady state, a heating power of about 320 W per cartridge was experimentally found when the getter temperature was maintained stable at 550 °C (after the initial transient condition at which higher power was required). Thermal simulations of the preparatory setup performed with the finite element code confirmed that the value of effective emissivity provides a reasonable accuracy within 15% of the experimental values. It must be noted that an alternative scheme to calculate the effective temperature might use the equilibrium grid temperature obtained experimentally, and define an effective emissivity on the basis of that result; the use of the NEG temperature will provide results on the safe side for thermal verifications.

An integrated thermal simulation was performed considering the main in-vessel components, the actively cooled thermal shields with varying water temperature along the cooling channels, the 384 getter cartridges at a fixed temperature, and natural convection on the outer surfaces of the vessel, as shown in Fig. 9(a). The heating power on the cartridges depends on the position within the cartridge pattern. As expected, the innermost cartridges require the highest power; the intermediate layer has the lowest requirement, while the external layer—facing the inner surface of the water-cooled vessel—has an intermediate condition, as detailed in Fig. 9(b). The effectiveness of the thermal shield in protecting the beam source (and the vessel) strongly depends on the emissivity of its surfaces, and the optimal configuration was found when high emissivity is applied to the copper shields. In this case, the contribution of reflections to the heat load on the beam source components is minimized: in the presented simulation, a surface emissivity of 0.7 was applied to the thermal shield. This value is lower than the attainable value of about 0.9, which can be achieved by blackening the copper surface by chemical treatments, as it was experimentally verified by IR measurements of the treated samples (kept at about 200 °C for the measurement).

The maximum temperatures expected on the beam sources are also detailed in Fig. 10(a). For the purpose of this simulation, we considered the worst case, in which the rear panel of the beam source is removed (as mentioned, punched sheets cover all sides of the source, to realize a planar equipotential and optimize the high voltage holding when the beam source is biased to accelerate the beam up to 110 kV). This configuration might be desired at the beginning of the operations, to operate the RF drivers without the rear sheet, and to have direct visual access to the rear side of the source and verify the absence of RF-induced breakdowns. In that case, the maximum temperature of the source structures within the electrostatic screen is of the order of about 95 °C. The main design requirement was to protect the beam source component from overheating. A second requirement for the NEG pump design and its thermal shields was to shorten, as much as possible, the cooling down phase of the cartridges: the target is to cool down from >500 °C (regeneration) to <100 °C (sorption) in about 6.5 h.

The resulting total heating power is 111 kW, calculated in the transient simulation after 20h. The heat loads detailed for the various components are reported in Table I and were used for the thermo-hydraulic design²⁵ of the vessel and thermal shield. In the same table, high-level requirements for the cooling system are detailed, for the various cooled elements. An interruption of the normal water supply of the cooling system activates the emergency cooling system, which ensures the cooling for 3h. This period allows cooling down of the NEG pump to a temperature lower than 200 °C, to avoid overheating of the surrounding components.

Figure 10(b) presents the temperature inside the NEG pump, highlighting the role of intermediate structures between the hot cartridges and the thermal shields. A detailed temperature distribution was applied to the thermo-mechanical analysis of the support structure, also detailed in Fig. 11. The design was verified in terms of thermal, mechanical, and seismic loads. An integrated model of the support structure with light trusses was developed including air pressure, water pressure, bolt preloads, weights of panels, and NEGs. The relative thermal deformations between the components at different operating temperatures were a key design input for the detailed mechanical design of the support of 32 panels and connections to the vessel.

The regeneration duration is directly proportional to the available auxiliary pumping speed. In order to maximize the availability of the vacuum system for the SPIDER operation, the overall pumping system based on turbopumps, roots pumps, and primary pumps has been revised. An overview of the system is given in Fig. 12, with also the improvements highlighted. In order to verify the foreline design, and in particular, the influence of its length and diameter on the performance of the connected turbopumps, a series of tests were performed on the present forevacuum system. A hydrogen pressure in the vessel of about 2 Pa was sustained, while a valve on the backing side of the turbopump allowed to reduce the total conductance of the foreline, as sketched in Fig. 13(a). By measuring the pumping speed in the vessel of one turbopump as a function of pressure on the backing side, the curve of Fig. 13(b) was realized. This reduction was measured for a vessel pressure relevant to the NEG regeneration phase, i.e., for a vessel pressure of 1.7 Pa. In these conditions, the turbomolecular pumps (TMPs) work at high throughput, and the long foreline causes a high backing pressure at the TMP, which in turn, would cause a 20% reduction of the TMP pumping speed with respect to the datasheet value [the result expected with the present design of the foreline is indicated by the vertical black line in Fig. 13(b)]. Increasing the conductance of the foreline by doubling the diameter appears to be enough to approach the limit given by the forevacuum pumps (which would provide a TMP pumping speed of 10% less than the nominal value).

The conceptual design of a first-of-its-kind large vacuum system based on nonevaporable getter cartridges has been presented, covering various aspects of the design. For the use of a NEG pump to pump hydrogenic species, the definition of sorption and regeneration cycles is fundamental to derive the required regeneration temperature and the auxiliary pumping speed and maximize the availability of NEG for the sorption phases. However, it was shown that the thermal design is critical for the integration of these pumps in existing facilities, and care had to be taken in the design of thermal shields, as their geometry also determines the conductance of the pump.

SPIDER will have one of the most interesting vacuum systems in Europe, with an installed pumping speed of 420 m3/s for hydrogen, which is about 1/10th of the required pumping speed of the ITER heating neutral beam injector²⁵ (in which, however, a gas neutralizer will also be operated). The capacity of the NEG pump will be used up to 240 kPa m³ while reaching only 660 Pa m³/kg of hydrogen concentration, and it will be used in a regime quite far from the common application of technology, i.e., it will operate with high hydrogen throughputs and high cumulative loads, in contrast to common applications in high- and ultra-high vacuum conditions and without processing gases. A total beam pulse duration of 15 h will be permitted per week, allowing both cesium conditioning and operation with long pulses. The operation of the future SPIDER vacuum system will be a novelty, being a hybrid system, with a very large NEG pump operating simultaneously with cryogenic pumps and turbopumps.

This work has been carried out within the framework of the ITER-RFX Neutral Beam Testing Facility (NBTF) Agreement and has received funding from the ITER Organization. The views and opinions expressed herein do not necessarily reflect those of the ITER Organization. This work has been carried out within the framework of the EUROfusion Consortium, funded by the European Union via the Euratom Research and Training Programme (Grant Agreement No. 101052200—EUROfusion). Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them.

The authors have no conflicts to disclose.

E. Sartori: Conceptualization (equal); Data curation (equal); Investigation (equal); Writing – original draft (equal); Writing – review & editing (equal). M. Siragusa: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (supporting). G. Berton: Data curation (equal); Formal analysis (equal). C. Cavallini: Formal analysis (equal). S. Dal Bello: Project administration (equal); Supervision (equal). M. Fadone: Investigation (equal). L. Grando: Data curation (equal); Investigation (equal). D. Marcuzzi: Project administration (equal). D. Rizzetto: Data curation (equal). G. Serianni: Data curation (equal). P. Sonato: Data curation (equal); Investigation (equal); Writing – review & editing (equal). M. Zaupa: Formal analysis (equal). F. Dinh: Data curation (equal). A. Ferrara: Data curation (equal). E. Maccallini: Data curation (equal). M. Mura: Data curation (equal); Investigation (equal). F. Siviero: Data curation (equal); Investigation (equal); Methodology (equal); Writing – review & editing (equal). V. Toigo: Resources (equal).

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