Beam divergence of RF negative hydrogen ion sources for fusion

Neutral beam injectors (NBI) for fusion facilities have strict requirements on the beam divergence (7 mrad for the ITER NBI at 1 MeV). Measurements of the single beamlet divergence of RF negative ion sources (at lower beam energy < 100 keV) show significantly higher values (9–15 mrad), also larger than filament arc sources at similar beam energies. This opened up questions whether the higher divergence is caused by different measurement or evaluation techniques, whether it is a direct cause of the RF source, e.g. due to a higher temperature of negative ions or an oscillating extraction meniscus, and whether it is a problem at all after full acceleration. In a joint effort between the labs modeling and diagnostic capabilities at the NNBI test facilities have been strongly extended and evaluation methods benchmarked. Particularly challenging is the strong increase in beamlet divergence at a lower filling pressure, seen both in filament arc and RF sources. Beside the source and beam investigations carried out in SPIDER (with selected, isolated apertures rather than the total of 1280 apertures) at Consorzio RFX, the IPP test facilities ELISE (640 apertures) and BATMAN Upgrade (70 apertures) contribute to the physics understanding of the beam optics in RF sources. The determination of the beam divergence is not straight-forward because effects originating from measuring the divergence of multiple beamlets (Beam Emission Spectroscopy) and/or constraints from the individual diagnostic (lateral heat conductance in CFC tiles) lead to difficulties. Still, the divergence requirement is not met at the limited total beam energy available at ELISE and BATMAN Upgrade (< 60kV). However, variation of the beam energy show a decrease of the divergence for higher energies and beam simulation for the ITER NBI accelerator predict that the divergence requirement will be met after full acceleration of the negative ion beam.


Introduction
Powerful plasma heating systems are required for the upcoming ITER fusion experiment in order to generate hot fusion plasma conditions enabling the fusion process of deuterium and tritium.The neutral beam injectors are the most powerful heating system of ITER: two Heating Neutral Beam injectors (HNB) each delivering 16.5 MW heating power are foreseen, with the option to add a third one in a later stage to end up with a total of 50 MW neutral hydrogen or deuterium beams, which are also required to drive the toroidal current in the ITER plasma needed for connement.A sketch of the neutral beam injectors is shown in Fig. 1.The HNBs are large systems (25 m length), consisting of a negative hydrogen ion source (up to 66 A from 1280 apertures) on negative high voltage potential, an extraction and acceleration system (1 MeV in D, 870 keV in H; one 10 kV extraction stage followed by ve acceleration stages), a gas target neutralizer, residual ion dumps, a beam calorimeter and the beam duct towards the ITER torus.The beam must fulll strict requirements to achieve the 16.5 MW heating power [1]: the divergence must be < 7 mrad in order to have a suciently high beamline transmission and a homogeneous beam extraction (+/-10%) is required from the 1280 apertures.In addition, the ion source must run at a lling pressure of equal or below 0.3 Pa in order to minimize stripping losses in the accelerator.
Negative hydrogen ions are created in a large ion source (2 × 1 m 2 ) by conversion of hydrogen atoms and positive ions on caesiated, low work function surfaces.The concept of the ITER HNB is based on Radio-Frequency (RF) driven ion sources, in which the plasma delivering precursors of negative ions is created by inductive coupling.RF sources benet from the long operational time without the need for maintenance compared to lament-arc driven sources.The ITER HNB systems will be the rst NNBI worldwide based on RF ion sources.A stepwise development program towards the full-size injectors driven by 8 RF drivers (1 MHz, up to 100 kW each) at the Neutral Beam Test Facility hosted by Consorzio RFX has been set up: the development started with the 1/8 scale prototype RF ion source developed at IPP and presently in operation at BATMAN Upgrade (up to 45 kV beam energy) [2], the half size ELISE ion source (up to 60 kV) [3], the full size ion source SPIDER (Consorzio RFX, up to 100 kV) and the full beamline MITICA (under construction at Consorzio RFX, up to 1 MeV) [4].The three facilities presently in operation use only reduced acceleration systems (one extraction stage followed by one acceleration stage, i.e. three grids); thus, comparing the divergences achieved at the present facilities to the requirements of the full ITER accelerator is not straight-forward.In particular, the velocity component of negative ions perpendicular to the beam direction (which is linked to the temperature of negative ions inside the plasma) contributes less to the beam divergence for increasing beam energy (i.e.axial velocity component).This paper focuses on the contribution of the two IPP facilities ELISE and BATMAN Upgrade to beam divergence investigations; results from SPIDER are in line with the ones presented herein and can found e.g. in [5].

Beam diagnostics at ELISE and BATMAN Upgrade
Several diagnostic tools are available to investigate the beam.Both facilities are equipped with a diagnostic CW-calorimeter dumping the beam, which oers also diagnostic access with a certain resolution (30 (horizontal) x 20 (vertical) mm 2 at ELISE and 20x20 mm 2 at BATMAN Upgrade) [6,7].However, since it is installed at a distance of more than 2 m to the acceleration systems, individual beamlets are strongly overlapped and thus only a global uniformity of the beam can be accessed.
The installed Beam Emission Spectroscopy (BES) uses a manifold of lines of sights (LOS) in order to determine the beam prole and divergence: At BATMAN Upgrade, horizontal LOS at two dierent axial positions give vertical resolution (BES1 5 LOS at 26 cm behind the Grounded Grid GG, and BES2 11 LOS 129 cm behind the GG) [8].At ELISE, BES is installed with 16 horizontal and 4 vertical LOS at an axial distance of about 2.5 m to the grounded grid [9].For the evaluation of the divergence measured with BES one needs to take into account that a multitude of beamlets usually contribute to the signal of a certain BES LOS [10], originating from the overlap of individual beamlets and the viewing cone size of the LOS.This is in particular critical for neighboring beamlet rows, because the co-extracted electron deection magnets installed in the second grid (Extraction Grid EG, dumping electrons directly on the EG) create a vertical B-eld that changes polarity from beamlet row to beamlet row.This causes an alternating horizontal zig-zag deection of the accelerated negative ions and results in an overestimation of the beam divergence from horizontal BES LOS.For the ITER HNB, a magnetic compensation system has been pioneered [11], which has been succesfully tested in a joint project with ITER at BATMAN Upgrade [12].Two BES spectra for horizontal LOS (one for compensated beamlet rows, and one for uncompensated beamlet rows) at BATMAN Upgrade are shown in Fig. 2 a).The peak for the uncompensated rows is signicantly broader, and thus the beamlet divergence is clearly overestimated.The same eect is seen at ELISE (no zig-zag compensation) when comparing divergences resulting from horizontal and vertical BES LOS (Fig. 2 b)): the divergence of the vertical LOS is signicanlty lower (minimum ≈ 24 mrad) compared to the horizontal LOS (minimum ≈ 31 mrad).It should be noted that during this campaign ELISE has not been operated at settings for reaching best beam optics (i.e. the ratio of acceleration to extraction voltage has not been optimized), thus the divergences are generally rather high.The usage of 1D Carbon Fiber reinforced Carbon (CFC) calorimeters, in which the bers are aligned in one dimension along the beam direction and thus heat conductivity in this direction is up to 20x higher compared to the perpendicular directions, are common tools in the NNBI community to determine properties of an individual beamlets [13,14].The beam footprint is observed via IR cameras either from the front or the backside of the CFC tile.BATMAN Upgrade is equipped with such an 1D-CFC calorimeter (2 cm thickness of the CFC tile) which can be ipped into the beam and is viewed from the backside [15].In order to avoid overlapping of the beamlets, an individual beamlet can be easily isolated by applying plugs in the plasma grid (the rst grid of the extraction system).However, CFC tile analysis is by far not straight forward, because at rst a realistic t of the beamlet footprint needs to be applied in order to determine rst the beamlet size (a 2D rotatable double Gaussian t is applied at BUG).To determine then the beamlet divergence, certain assumptions of the initial beamlet properties at the end of the accelerator (grounded grid) need to be made; typically an elongated point source with a certain diameter is used.Lateral heat conductance in the tile leads to an overestimation of the beamlet size, so a trade-o between a good signal/noise ratio (late time frame after beam start) and low inuence of the lateral heat conductance (early time frame from beam start) needs to be taken [16].
Since dierent, home-made evaluation routines are applied at the dierent NNBI laboratories, it was not clear, whether the results can be directly compared between the labs.ITER initiated a task force between the labs in which the routines have been benchmarked against each other.When using the same input parameters (timeframe and initial beam size), all routines yield a similar value for the divergence [17].Still, when comparing CFC results from one lab to another, special care should be taken which input parameter and which time frame has been used for the evaluation.

Beam divergence of RF H − sources
In order to determine the beamlet divergence, an individual beamlet has been isolated in the upper half of the BATMAN Upgrade PG (central beamlet in the fourth row of the 14 (v) × 5 (h) beamlet group).The inuence of the source lling pressure on the single beamlet divergence is shown in Fig. 3 a).Because variation of the lling pressure changes also the extracted ion current and thus the perveance, whole perveance curves (carried out by variation of the RF power) are compared.For the CFC evaluation the timeframe at 0.5 s after beam start and a point source at the GG is used here; thus, the determined divergence values represent an upper boundary.The following statements can be drawn from the plot: (1) the divergence is reduced for higher perveances, no overperveant regime with increasing divergence is reached (up to 80 kW RF power applied).( 2) there is a strong dependence of the divergence on the lling pressure; in particular a pronounced increase in the divergence from 0.4 Pa to 0.3 Pa (minimum value is increased from 13.5 mrad to 17.1 mrad).( 3) simulations carried out for the BATMAN Upgrade accelerator from IBSimu using a perpendicular temperature of 1 eV for negative ions [18,12] match with the divergence measured for the higher lling pressures around 0.6 Pa.(4) the divergence requirement of the ITER beam source is not reached (but it needs to be taken into account that the total beam energy of 37.5 keV is far from the ITER value of 870 keV for hydrogen).
The result using an earlier timeframe at 0.2 s after beam start and an initial radius of 2.5 mm for the CFC evaluation is shown in addition to the divergence evaluated with the BES for the same isolated beamlet in Figure 3 b) for the cases of 0.3 Pa and 0.4 Pa lling pressure.Using 2.5 mm initial diameter has been agreed in the before mentioned task force initiated by ITER.The strong eect of the initial beamlet size and lateral heat conductance on the evaluation result can be clearly seen: the minimum divergence at 0.3 Pa is reduced from 17.1 mrad to 12.2 mrad.BES does not need to make assumptions on the initial beam properties since it directly measures the angular distribution.The comparison to the divergence of the same beamlet measured with BES shows a better match to the upper boundary CFC evaluation, even though it needs to be taken into account that the signal/noise ratio for the single beamlet BES is rather weak and thus the resulting error bar is quite big.
In the next step, consequences for the ITER HNB are explored.If the lateral velocity of negative ions contribute signicantly to the divergence, the latter should be reduced for higher total beam energy U tot .The inuence of the divergence on U tot is shown in Fig. 4 in which perveance curves are compared for CFC measurements at BATMAN Upgrade as well as IBSimu predictions.The ratio of acceleration to extraction voltage has been kept at the optimum value of 6.5 for the BATMAN Upgrade accelerator.Generally, the divergence is slightly lowered in the accessible U tot range at BATMAN Upgrade (up to 45 kV).A very similar trend (but lower absolute values) for the divergence is predicted by IBSimu for the BATMAN Upgrade accelerator, again using T H − = 1eV.
In order to predict the divergence of the ITER HNB, modeling with IBSimu for the full ITER accelerator has been carried out (more details are given in [19]).The simulations (Fig. 5) show that in the low beam energy regime of several 10 keV (at which the test facilities BATMAN Upgrade and SPIDER show a minimum beamlet divergence of 915 mrad [16,5]) the highest fraction of the divergence comes from the perpendicular velocity of negative ions.Consequently, the divergence is predicted to meet the ITER HNB requirement (< 7 mrad) after full acceleration.Nevertheless, a higher perpendicular velocity distribution might lead to a stronger scraping, and thus loss, of beam particles inside the 7-grid extraction and acceleration system of the ITER NBI.IBSimu modeling predicts scraping in the low single-digit percentage range inside the accelerator for beams with elevated perpendicular temperature [19], which should be still tolerable for the system.Determining the beam divergence from Beam Emission Spectroscopy and CFC tile thermography is far from being straight forward: regarding BES, it needs to be taken into account that a multitude of beamlets can contribute to the signal of a line-of-sight; in particular, if the row-wise horizontal zig-zag deection of the accelerator is not compensated, the divergence is overestimated by horizontal LOS.Beamlets need to be isolated for CFC measurements, and due to the lateral heat conductance and assumption of the initial beamlet properties, the resulting divergences can have signicant uncertainties.A more accurate, but complex to install, tool for determining the beamlet divergence is a beam emittance scanner operated by Consorzio RFX in SPIDER [20].The emittance scanner has been used recently also at BATMAN Upgrade in a joint campaign; evaluation is on-going.CFC measurements at BATMAN Upgrade show that a single beamlet divergence in the range of 1217 mrad is reached at 0.3 Pa and a beam energy of 37.5 keV at standard operational parameters; when optimizing operational parameters for minimum divergence, values down to 915 mrad can be achieved [16].Similar values are also reached at SPIDER [5].This divergence is signicantly larger than the divergences typically achieved in lament-arc driven ion sources, albeit often at higher total beam energy [21,22].A strong reduction of the divergence occurs for higher lling pressure: already a 20% reduction can be achieved at 0.4 Pa lling pressure.
The reason behind the strong lling pressure dependence and the higher divergence RF sources seem to have compared to lament-arc sources is not yet understood but presently strong eort is taken in the whole NNBI community to identify the reason.It may result from meniscus perturbations or a higher temperature of negative ions, possibly resulting from more energetic precursor particles (i.e.hydrogen atoms and positive ions).Both precursor species can give energy to negative ions by the reection of surface-produced negative ions; positive ions transfer energy in addition by coulomb collisions inside the plasma.A possible correlation between T H − and the lling pressure is shown in [19].Consequently, diagnostics for determining the properties of the precursor particles are presently enhanced.Consorzio RFX contributes here with the measurement of the energy distribution function of positive ions using a Retarding Field Energy Analyzer (RFEA), IPP contributes with the measurement of H properties using a Two photon Absorption Laser Induced Fluorescence (TALIF) diagnostics [23]; the investigations are on-going.
If the divergence is dominated by the lateral velocity of negative ions, a strong reduction should occur after acceleration to higher energies.This trend was shown at BATMAN Upgrade in the available range of maximum 45 keV beam energy.Beam Simulation with IBSimu for the full ITER HNB accelerator predict that the divergence requirements will be met for the foreseen beam energies (870 keV in H, 1 MeV in D).However, it might be challenging for the DNB operating at lower beam energies of 100 keV and targeting even lower divergences, so a possible optimization of the ion source to deliver lower divergent beams would be helpful.

Figure 1 .
Figure 1.Sketch of the neutral beam injectors for ITER: two Heating Neutral Beam injectors (HNB) are foreseen rst, a third one is an option for the future and the Diagnostic Neutral Beam injector (DNB).

Figure 2 .
Figure 2. a): Doppler peak of the BES spectrum in case of a compensated and a noncompensated horizontal zig-zag deection at BATMAN Upgrade (horizontal LOS).b): BES divergence measured at ELISE as function of the normalized perveance P/P 0 for a horizontal and vertical LOS, P 0 is the maximum perveance for the grid system.

Figure 3 .
Figure 3. Single beamlet core divergence as function of the normalized perveance for variation of the RF power (perveance) and lling pressure.a): using the CFC evaluation at the timeframe 0.5 s and a point source at GG (upper boundary).b): in addition with CFC evaluation at an earlier timeframe 0.2 s and 2.5 mm initial beam radius (open squares), and with the BES divergence measurement (stars).The turquoise bar represents the range of IBSimu predictions for the same perveance variation using 1 eV perpendicular temperature of H − .

Figure 4 .
Figure 4. Single beamlet core divergence as function of the normalized perveance for variation of the RF power and the total acceleration voltage keeping the ratio between acceleration and extraction voltage constant: CFC measurements and IBSimu predictions (T H − = 1 eV).Hydrogen operation.

Figure 5 .
Figure 5. Simulation results from IBSimu: divergence of the ITER HNB in hydrogen as function of the total beam energy, with separated contributions from the perpendicular H − temperature, the magnetic eld inside the accelerator and the intrinsic divergence of the accelerator.

4 .
Discussion and conclusions