Table of contents

It is widely recognized that neutral beam injection (NBI), i.e. the injection of high energy, high power, beams of H or D atoms, is a flexible and reliable system that has been the main heating system on a large variety of fusion devices, and NBI has been chosen as one of the three heating schemes of the International Tokomak Reactor (ITER). To date, all the NBI systems but two have been based on the neutralization (in a simple gas target) of positive hydrogen or deuterium ions accelerated to <100 keV/nucleon. Above that energy the neutralization of positive ions falls to unacceptably low values, and higher energy neutral beams have to be created by the neutralization of accelerated negative ions (in a simple gas target), as this remains high (≈60%) up to >1 MeV/nucleon. Unfortunately H^- and D^- are difficult to create, and the very characteristic that makes them attractive, the ease with which the electron is detached from the ion, means that it is difficult to create high concentrations or fluxes of them, and it is difficult to avoid substantial, collisional, losses in the extraction and acceleration processes. However, there has been impressive progress in negative ion sources and accelerators over the past decade, as demonstrated by the two pioneering, operational, multi-megawatt, negative ion based, NBI systems at LHD (180 keV, H⁰) and JT-60U (500 keV, D⁰), both in Japan. Nevertheless, the system proposed for ITER represents a substantial technological challenge as an increase is required in beam energy, to 1 MeV, D⁰, accelerated ion (D^-) current, to 40 A, accelerated current density, 200 A m^-2 of D^-, and pulse length, to 1 h.

At the Fourth IAEA Technical Meeting on Negative Ion Based Neutral Beam Injectors, hosted by the Consorzio RFX, Padova, Italy, 9–11 May 2005, the status of the R&D aimed at the realization of the injectors for ITER was presented. Because of the importance of this development to the success of the ITER project, participants at that meeting were asked if they were interested in rewriting and extending their contributions as a submission to Nuclear Fusion. Technology papers were accepted because of the very nature of the subject. The submissions underwent the regular double-referee peer-review process, and the accepted articles are grouped together in this special issue with a sequence given by the following subjects:

• Beam line and large source development

• Small sources and source modelling

• Source diagnostics development

• ITER N-NBI design and development

I hope that this special issue will document in sufficient detail the present state of the art of negative ion based neutral beam injection systems.

Message from the Editor, F.C. Schüller

Now that the construction of ITER has become a reality the consequence is that fusion research will gradually shift in its focus from plasma physics alone to more technological issues.The Editorial Board of Nuclear Fusion has recognized this trend and therefore wants to give more prominence in the journal to heating methods and related techniques. Therefore we are happy to bring negative ion based neutral beam injection to the foreground with this special issue. We have found a dedicated Guest Editor in the person of Ron Hemsworth.

A negative-ion-based neutral beam injection (NBI) system, in which large caesium-seeded negative-ion sources are utilized, has operated reliably in the Large Helical Device (LHD) since it was operational in 1998. The injection power of the hydrogen beam has been increased up to 13.1 MW with three injectors. In one injector with modified ion sources utilizing a multi-slotted grounded grid, the injection power reached 5.7 MW with an energy of 184 keV, which exceeds the designed value of 180 keV and 5 MW. Individual control of the local arc-discharge with independently divided arc and filament power supplies is effective in improving beam uniformity in a large negative ion source. The injection duration has been extended to 120 s, during which the LHD plasma is sustained by the NBI alone with a reduced power of 0.2–0.3 MW using one ion source with a cooled plasma grid. These results demonstrate that the injection performance of the negative-ion-based NBI is comparable to that of the conventional positive-ion-based NBI. The progress of the negative-NBI in the LHD is reviewed in detail from the point of view of the improvement of the negative ion sources.

The 500 keV negative-ion based neutral beam injector for JT-60U started operations in 1996. The availability of the negative ion based neutral beam injection system has been improved gradually by modifying the ion source and optimizing its operation parameters. Recently, the extension of the pulse duration up to 30 s has been intended to study quasi-steady state plasma on JT-60U. The most serious issue is to reduce the heat load on the grids for long pulse operation. Two modifications have been proposed to reduce the heat load. One is to suppress the spread of beamlet-bundle which may be caused by beamlet–beamlet interaction in the multi-aperture grid due to the space charge force. Indeed, the investigation of the beam deflection, which was measured by the infrared camera on the target plate set 3.5 m away from the grid, indicates that the spread of beamlet-bundle is in proportion to the current density. Field-shaping plates were attached on the extraction grid to modify the local electric field. The plate thickness was optimized to steer the beamlet deflection. The other is to reduce the stripping loss, where the electron of the negative ion beam is stripped and accelerated in the accelerator and then collides with the grids. The ion source was modified to reduce the pressure in the accelerator column to suppress the beam-ion stripping loss. To date, long pulse injection of 19 s of 1.5–1.6 MW at a high energy beam of 360 keV, 9–10 A for D⁻ has been obtained by one ion source with these modifications.

The development of a large-area RF source for negative hydrogen ions, an official EFDA task agreement, is aiming at demonstrating ITER-relevant ion source parameters. This implies a current density of 200 A m⁻² accelerated D⁻ ions at a source filling pressure of ⩽0.3 Pa and an electron-to-ion ratio of ⩽1 from an extraction area similar to the positive-ion based sources at JET and ASDEX Upgrade and for pulse lengths of up to 1 h. The work is progressing along three lines in parallel: (i) optimization of current densities at low pressure and electron/ion ratio, utilizing small extraction areas (<0.01 m²) and short pulses (<6 s), in this parameter range the ITER requirements are met or even exceeded; (ii) investigation on extended extraction areas (<0.03 m²) and pulse lengths of up to 3600 s and (iii) investigation of a size-scaling on a half-size ITER plasma source. Three different test beds are being used to carry out these investigations in parallel. An extensive diagnostic and modelling programme accompanies the activities. The paper discusses the recent achievements and the status in these three areas of development.

Operation of the neutral beam system of ITER for the entire ITER pulse is foreseen, with pulse lengths extending up to 1 h. Operation for such a pulse length is entirely new for neutral beam systems and the associated sub systems. This paper reviews some of the experience gained so far in the operation of the MANTIS test bed where long pulse operation of the ITER reference type of negative ion source is being studied. The three identified adverse effects of long pulse operation—reduced negative ion yield, reduced plasma grid temperature effect (see below) and increased caesium consumption—are still being studied, but some tentative explanations for each effect are given. In addition to the aforementioned effects, some operational difficulties associated with long pulses will be discussed, and some areas where caution will be needed in the future long pulse, high power operation of the ITER injectors are indicated.

The physical phenomena important in volume and surface production negative hydrogen ion sources are reviewed. The phenomenon, designated volume production, was attributed to dissociative electron attachment of low energy electrons to rovibrationally excited molecules. The experimental verification in 2005 of the reality of this mechanism is reported. Recent results obtained by measuring the negative ion temperature in a filament discharge ion source indicate that two negative ion populations are present. Understanding the physics of volume negative ion sources operating in accelerators with H⁻ ion density in the 100 mA cm⁻² range opens the prospect of creating volume caesium-free sources for fusion. The evolution of physics of surface production is also described.

A measurement effected near the plasma electrode (PE) using two-laser beam photodetachment showed the presence of a directed negative ion flow. The ratio of the negative ion and electron currents extracted through the transverse magnetic filter field near the PE can be predicted from a collisional flow model. A specific for two negative species plasmas transport mechanism in the weak transverse magnetic filter field is described.

We report current and past efforts made by our group in the ab initio modelling of different negative ion sources. In particular, we discuss the cross sections of elementary processes relevant to negative ion kinetics, including electron–molecule, atom–molecule and atom/molecule gas surface interactions, particularly emphasizing the role of vibrational excitation in affecting the cross sections. Attention is also paid to the elementary processes involving caesium in both volume and surface sources.

Self-consistent models, which couple the Boltzmann equation and the vibrational kinetics, are used for describing multipole and rf discharges, while a PIC-MC (particle in cell) with Monte Carlo collisions is used to study electron and ion dynamics in a parallel plate reactor in the post-discharge regime. The present theoretical results should encourage further dedicated experimental work in the field.

Trajectories of H⁻ ions are calculated numerically by solving the 3D motion equation, including effects of collisional destruction, elastic collisions and charge exchange collisions. According to these trajectories, the extraction probability of H⁻ ions produced at any location inside the source and the energy relaxation of extracted H⁻ ions are estimated as a function of gas pressure. Effects of production zone and field intensity of the magnetic filter on extraction probability are also discussed. The extraction probability of surface produced H⁻ ions is much higher than that of volume produced H⁻ ions. The kinetic energy of extracted H⁻ ions is reduced mainly by charge exchange collisions with H in higher pressure region and by elastic collisions with H⁺ in lower pressure region.

We also briefly discuss the pressure dependence of extracted negative ion currents combining the present numerical results and the results of the model calculation with the zero-dimensional code.

Over the last few years, the accelerator community requested the development of improved negative hydrogen ion sources. For spallation sources, like SNS or ESS, pulsed high intensity H⁻ ion beams of a few tens of milliamperes, with a duty cycle close to 10%, are required. New facilities like CERN also ask for high performance negative ion beams. Since CEA undertook an electron cyclotron resonance (ECR)-based ion source programme, a European network devoted to high performance negative ion source development has been created.

In this group, several laboratories developing 2.45 GHz ECR sources follow different approaches to increase the extracted ion current. At Saclay, with a solenoidal magnetic structure based on coils, close to 3 mA H⁻ ion beam is now extracted in pulsed mode (2 ms/100 ms). A metallic biased grid separates the plasma bulk from the H⁻ ion production zone and significantly improves the H⁻ extracted current. At Ecole Polytechnique, the source Camembert III operates in continuous wave and pulsed modes. Photodetachment measurements in continuous wave mode show considerable H⁻ ion density (10⁹–10¹⁰ cm⁻³), whether the primary electrons are provided by filaments or small ECR modules inserted into the plasma chamber. Similarities and differences between the ECR-driven and the filamented source are discussed. Representative experimental results from the sources of the two laboratories as well as future plans are reported here.

Pure volume production of D⁻ negative ions is studied in a rectangular arc chamber. Production and control of D₂ plasmas are tested by varying the intensity of the magnetic filter MF. The values of electron temperature T_e and electron density n_e in D₂ plasmas are slightly higher than those in H₂ plasmas. T_e in D₂ plasmas cannot be decreased below 1 eV in the extraction region with the same MF conditions for optimizing H₂ plasmas. A stronger MF field is required for controlling T_e in D₂ plasmas. Therefore, the transport in or the production of H₂ and D₂ plasmas is different, i.e. there is an isotope effect, a consequence of which is that the H⁻ and D⁻ densities have different spatial distributions. Extracted H⁻ and D⁻ currents are primarily determined by H⁻ and D⁻ densities in front of the extraction hole, respectively. According to the estimated rate coefficients and collision frequencies of main collision processes, it is reconfirmed that T_e in the extraction region should be maintained below 1 eV with n_e kept high for the enhancement of H⁻ and D⁻ production. For the enhancement of D⁻ production, preliminary results on VUV emission measurement are presented and discussed briefly.

The development of negative hydrogen ion sources for neutral beam systems is closely linked with an optimization of negative ion formation in hydrogen plasmas, which requires knowledge of the plasma parameters. Emission spectroscopy is introduced as a non-invasive and in situ diagnostic tool for line of sight averaged plasma parameters. Diagnostic lines and simplified analysis methods for a variety of plasma parameters, such as electron density and electron temperature, gas temperature, atomic and molecular hydrogen density, caesium densities (atoms and ions), and negative ion densities are identified and prepared for direct application. Emphasis is laid on results obtained in RF generated negative ion sources. Correlations of plasma parameters with extracted negative ion current densities are discussed. Stripping losses in the extraction system are quantified by using beam emission spectroscopy.

Determining d²I/dV² from a traditional Langmuir probe trace using numerical techniques is inherently noisy and generally yields poor results. We have developed a Langmuir probe system based on a method first used in the 1950s by Boyd and Twiddy (1959 Proc. R. Soc. A 53 250). The system measures the 2nd derivative directly. This paper presents an account of the experimental method, apparatus and software used along with some preliminary results from the KAMABOKO III source including a comparison with conventional probe methods.

A tool for investigating the electron energy distribution function (EEDF) in plasmas was developed and used to study the primary electron behaviour in a bucket type ion source. The spatial variation of the EEDF in filter magnetic fields, which optimize the negative ion production, is extensively studied. It is verified that the high energy component of the EEDF damped over a shorter characteristic length than the low energy component. Similar behaviour is also observed and studied in the cusp regions of the plasma.

The origin of longitudinal beam non-uniformity, which is one of the key issues in large caesium (Cs)-seeded negative ion sources for fusion application, was experimentally investigated in the JAERI 10 A negative ion source. A sufficient caesium was seeded in the negative ion source to enhance the negative ion production. The beam intensity with Cs was typically four times higher than that without Cs, and was relatively high in the upper region where the electron temperature and plasma density were relatively high. This distribution was significantly different from that without Cs. From the correlation between the beam intensity and the plasma parameters, it was found that the beam non-uniformity was due to the localization of the plasma.

The front edge of the filaments was bent towards the plasma grid to suppress the plasma localization. This resulted in the significant improvement of the beam uniformity. The root-mean-square deviation of the beam intensity from the averaged value decreased to a half of that before the modification while the beam intensity integrated along the longitudinal direction was kept constant. This result indicates that the suppression of the plasma localization by the filament modification is effective for improving the beam uniformity in the Cs-seeded negative ion source.

The ITER neutral beam injection is based on the acceleration and neutralization of negative deuterium ions. The target performance for the ITER beam source is to accelerate to 1 MeV a 40 A D⁻ beam, with a current density of 200 A m⁻², with pulse lengths of ⩾1000 s. It was found that in long pulse operation the negative ion yield from the filamented Kamaboko III ion source (a model of ITER ion source) degrades in comparison with short pulse operation, <5 s. This could be linked to the behaviour of caesium (Cs), which is added to the source to increase its negative ion yield and tungsten (W) evaporated from filaments. Cs and W are co-adsorbed on the source walls and the plasma grid and the composition of this coating can vary during long pulse operation. The possible consequences of this changing surface on the negative ion production will be discussed.

Tungsten filaments have a limited lifetime in the ion source and changing filaments and refilling of the Cs oven are the only scheduled maintenance events for the ITER injectors. These are complicated operations as the ITER injectors will be highly activated and all maintenance has to be carried out remotely. Therefore, increasing the filament lifetime and decreasing the Cs consumption are highly desirable. This paper presents results of relative measurements (including spectroscopic and chemical) of the W content of the Kamaboko-III source and reports relevant calculations on Cs consumption and W evaporation.

Major issues of neutral beam (NB) injector power supplies are high-speed switching, regulation and transmission of dc ultra-high voltage (UHV) and suppression of surge energy input to the beam source at high voltage breakdown. A gate turn-off thyristor inverter type power supply where the control is performed on the low voltage ac side was designed for the ITER NB. Based on the remarkable progress of a high power IEGT (injection enhanced gate transistor), the design of the inverter has been modified to increase efficiency and compactness using such new elements. The power loss in the inverter is reduced to 0.5 MW from 1.6 MW by adopting the IEGT inverter system at the same frequency of 150 Hz. Further, power supply performance has been studied for higher frequency operation of the inverter at 300 Hz. It was confirmed that the voltage regulation was stable even when the filter capacitor was reduced to 1/5 of the original value by circuit simulation. For the transmission line of the dc UHV with intermediate voltages, a disc shape multi-conductor bushing with a transmission line test chamber has been developed. The dimensions of the bushing are 1.8 m (diameter) and 140 mm (thickness) at the edge. Electric fields at the conductor surface and insulator surface were designed to be lower than 5 kV mm⁻¹ and 7 kV mm⁻¹, respectively. An electric field at the bottom of the ground potential outer conductor was designed to be lower than 1.2 kV mm⁻¹ to prevent particle levitation which triggers breakdown. The prototype transmission line has passed the lightning impulse test up to 1300 kV. A dc UHV up to 1175 kV was successfully sustained for 300 s. To prevent electric damage of the beam source at high voltage breakdown, core snubbers using Fe-based nanocrystalline soft magnetic materials have been adopted to dissipate the surge energy.

The 1 MV bushing of the ITER NB injector is probably one of the most challenging components of the whole accelerator, since such a high voltage (HV) in a contaminated and radiated environment requires the development of non-standard technologies. This paper intends to provide a guideline for the design of 1 MV bushing for the ITER neutral beam injector test bed. Although mainly focused on the SINGAP alternative concept of the injector, the reported general criteria can be useful for the design of other HV components, such as post insulators of the acceleration grids. For every different kind of required insulation, basic physical models describing breakdown phenomena are reported together with the main technological implications and the practical recommendations necessary to improve insulation performance. The results of the finite element analyses of 1 MV bushing, operating at CEA—Cadarache laboratories, are used to validate general criteria in a typical geometry. The effects of radiation, such as neutrons and gamma rays, on solid insulators and gases have also been taken into account, describing the most important phenomena and the related implications in terms of voltage hold-off capabilities. Finally the design of the bushing in the SINGAP option and the solutions adopted are described.

A local mixing model (LMM) (H.-D. Falter et al 1992 Proc. 17th Symp. on Fusion Technology (Rome, 1992) p 481) has been utilized to compute the evolution of the hydrogen isotope content within the implantation zone of the CuCrZr target material of the beam-stopping elements of the ITER neutral beam test facility (NBTF), together with the beam–target fusion reaction rate calculated by taking account of the slowing-down of the 1 MeV incoming projectile ion within the implantation layer. An important modification of the LMM code is to treat the tritium reaction product ions, resulting from D–D reactions, as a constituent of the incident beam. Although the treatment of tritium in the LMM is not ideal, this and other simplifying assumptions either do not significantly affect the predictions, or ensure conservatism in the results when used as input to the safety analysis of the facility. For example, it is shown that T → D 'beam–target' reactions always dominate over those of D → T, which overcomes the problem of uncertainty, in the model, of the distribution of tritium trapped within the implantation layer; in contrast there is little uncertainty that this region could approach deuterium saturation after <100 full-length (1 h duration) pulses. Using the computed sources as input, neutronics and activation calculations for the NBTF components have been carried out using the codes MCNP/FISPACT. As expected, long-pulse operational requirements lead to neutron activation and tritium production levels which have non-negligible but manageable radiological consequences.

Indirect heating of the neutralizer gas by the beam has been shown to be responsible for the reduced neutralization efficiency observed in positive ion neutral beam systems. The translational gas temperature can be predicted from the neutralizer plasma density and electron temperature and, in general, a higher gas temperature implies a lower neutralization target. A simple one-dimensional model, originally developed for the JET positive ion neutral injection system, has been adapted for ITER relevant negative ion systems. The results imply that gas heating for the ITER heating and diagnostic beams is unlikely to be severe and hence the neutralization target might be expected to be close to the design value.

A prototype negative ion accelerator, based on a pre-accelerator normally used for positive ion acceleration, was used on the Cadarache MV negative ion beam facility up to 2004. Even though this accelerator demonstrated the feasibility of SINgle GAP, SINGle APerture (SINGAP) accelerators it was not possible to produce beams with the optical quality required for ITER. A new 'ITER-like' accelerator, which is a scaled down version of the ITER SINGAP accelerator, has been built and installed on the Cadarache 1 MV test bed. The objective of the 'ITER-like' is to demonstrate reliable D⁻ beam acceleration as close as possible to 1 MeV with a current density j⁻ ≈ 200 A m⁻² with the beam optics required for ITER, i.e. a beamlet divergence of ⩽7 mrad and beamlet steering within ±2 mrad of that specified. High voltage hold-off tests have been performed and 940 kV has been held without breakdowns. Beams up to 850 keV (D⁻, 15 A m⁻²) were obtained after four weeks of experiments and the highest current density that has been obtained so far is 150 A m⁻² (D⁻, 580 keV, 23 mA for one beamlet). The required beam optics for ITER has been obtained at near perveance match (at 120 A m⁻² D⁻, 727 keV, 18.5 mA for one beamlet), but a larger than expected halo has been measured.

The present objective of the 1 MeV electrostatic accelerator R&D for the ITER NB system is the acceleration of ampere class negative ion beams at the required current density (200 A m⁻²) up to the beam energy of 1 MeV. To produce such high current density H⁻ ions, the KAMABOKO negative ion source was operated under high power arc discharge (≤40 kW) and with Caesium seeding for enhancement of the negative ion surface production. The H⁻ ion beams of 146 A m⁻² (total ion current: 0.2 A) have been obtained stably at the beam energy of 836 keV (pulse length: ≥0.2 s). The pulse length was limited due to the power density of the beams, which is more than twice as high as those of the existing negative ion based NB systems (JT-60U and LHD). A preliminary measurement of beam optics showed a beamlet divergence of ≈5 mrad even in the beams of such high power density. The Bremsstrahlung generation as a consequence of electron acceleration was estimated. A discussion then follows for photoelectron production and possible breakdowns in insulators for the acceleration of higher current negative ions.