Identification of the extraction structure of H− ions by Hα imaging spectroscopy

Extraction structure of negative hydrogen ions (H−) (i.e. the regions where H− ions are effectively extracted from the plasma) was obtained using hydrogen Balmer-α (Hα) imaging spectroscopy and cavity ring-down spectroscopy. The H− ion density increases after caesium (Cs) seeding through a surface conversion process on a Cs-covered plasma-grid (PG) surface. We found a reduction in the H− density during beam extraction after Cs conditioning, and the same signal reduction appeared in the Hα intensity caused by the reduction in the excited hydrogen (n = 3) population, which in turn is caused by the decrease in the mutual neutralization process between positive and negative hydrogen ions. We clearly observed the reduction structure of the Hα emissions in the extraction region; the structure expands at optimal Cs conditioning. From this result, the H− ions, which are produced at the PG surface, release to the extraction region and widely distribute during arc discharge. We conclude that the reduction of the H− density is caused by the particle loss due to beam extraction from the PG apertures.

probe measurement, a hydrogen ion-ion plasma forms in the extraction region near the PG surface under optimal Cs conditioning [19,20]. It is considered that the charge neutrality of the plasma is maintained by the balance of positive and negative hydrogen ions with few electrons being involved. To determine the H − density, a suitable diagnostic method using cavity ringdown spectroscopy (CRDS) has been developed by Quandt [21], and it has been applied to an RF-driven negative ion source [22] as well as our arc-driven source [23,24]. This tool is capable of showing the H − growth caused by the Cs conditioning in the extraction region close to the PG surface, but it is difficult to determine the H − distribution for a single discharge because the alignment of the cavity mirrors and laser path is delicate.
On the other hand, optical emission spectroscopy (OES) for the H − ion source has been developed at the Max-Planck Institute for Plasma Physics (IPP), Garching [25]. Here the hydrogen Balmer-α (H α ) emission correlates with the excited-state hydrogen (n = 3) population density. According to the study of atomic and molecular processes in hydrogen plasma [26,27], the excitation mechanisms are electron impact collisions (i.e. direct excitation with H, recombination with H + , dissociative excitation with H 2 and dissociative recombination with H + 2 ), proton impact collisions with H and the mutual neutralization process which can be written as H + m + H − → H(n = 3) + H m (m = 1, 2, 3), where m is the number of atoms. In a pure hydrogen discharge at a low electron temperature (∼1 eV), the dominant excitation mechanism for H α emission is the dissociative recombination between an electron and H + 2 . However, as the percentage of negative ions increases with optimal Cs conditioning in a negative hydrogen ion source, the H α light emission created by the mutual neutralization process becomes dominant, because the excitation cross section of the mutual neutralization is of the same order as that of the dissociative recombination. Previously, we confirmed the relevance of the H − density and the H α intensity measured using CRDS and OES, respectively, for the H − ion source at the National Institute for Fusion Science (NIFS) [28]. Both the H − density and H α emission measured in the extraction region increased with Cs seeding. We also found a signal reduction for both signals during beam extraction. Consequently, we considered that a two-dimensional H α measurement has the ability to probe the extraction behaviour of the H − ions, which is the clarification point for the negative ion source. Further, we developed this imaging diagnostic tool for H α emission [29], so as to obtain the reduction structure of H − ions during beam extraction.
In this paper, we present an optical diagnostic method for H α imaging spectroscopy in the extraction region of a negative hydrogen ion source; the images of the H α emission distribution are presented as evidence. We also present an interesting structure of H α reduction caused by beam extraction, which is an aspect of the H − extraction behaviour. These results suggest growth of the H − distribution depends on the method of Cs conditioning. Finally, a relationship between the reduction in the H α emission and the H − density owing to beam extraction is presented and discussed. Figure 1 shows a cross section of the one-third-scaled hydrogen negative ion source at NIFS. Hydrogen plasma is generated in the arc chamber by arc discharge. The extraction region near the PG surface is separated from the discharge region by magnetic filters to reduce the electron temperature and enhance the H − production. We supplied Cs vapour from the backplate of the arc chamber to increase the H − density through surface production. We installed a bias insulator between the magnetic filter flange and the PG flange to apply a bias voltage for the reduction in the electron extraction current, which is optimized at 2-3 V for the negative ion source for the NBI in the LHD. The line of sight (LOS) of the optical diagnostic tool arranged parallel to the PG surface is also shown in figure 1. The imaging system is installed on the sidewall of the bias insulator and its LOS also passes through the extraction region parallel to the PG surface with a gap of 11 mm. This LOS is parallel to the third row of PG apertures from the bottom of the arc chamber, while the Cs evaporator is located 9 cm above this LOS on the y-axis. Other support diagnostic tools (i.e. CRDS, OES and an electrostatic probe) are also installed on the bias insulator. The electrostatic probe is set 34 cm above the imaging LOS and the gap distance to the PG surface is 11 mm, which is same as the imaging system. The LOS for CRDS is arranged at 13 cm above the imaging LOS and the gap distance is 2 mm in this experiment. The H α imaging system consists of three optical filters (i.e. a neutral density filter, an infrared cut filter and a narrow band pass filter), an aspherical lens and a charge-coupled device detector (CCD). A glass-fibre image conduit for the transfer of the optical image is set between the lens and the CCD to insulate the high voltage for beam extraction. This system acquires a 16-bit monochrome image of resolution 1292 × 964 pixels (width × height). Figure 2 shows a photograph of the extraction region taken from the viewport used for imaging spectroscopy. The viewing angle has coverage from the magnetic filter flange to the PG surface. Both sides of the image field are invisible behind the flange. The left-hand side is the arc discharge side, and the H − ions are extracted to the right-hand side passing through the PG apertures. The diameter of the PG aperture is 12 mm and the gap between the PG and extraction grid (EG) is 5 mm. We applied only negative extraction voltage between the PG and the EG which is grounded in this experiment (i.e. without beam acceleration) in order to exclude back streaming positive ions due to an acceleration voltage. A row of the PG apertures appears as a quadrangle shape in the image. To understand the positional relationship, we superimposed a wire frame in the H α image in order to show the containment of the major component inside the ion source. Figure 3(a) shows the waveform of the arc discharge power without Cs conditioning (i.e. a pure volume) and with Cs conditioning. Arc power for these discharges is held constant at 38 kW during beam extraction under the condition of 0.2 Pa hydrogen gas pressure and 0.2 V (i.e. low) bias voltage. The duration of the arc discharge is 9 s for each shot with an interval time of 2 min. The temperature of the PG surface varied in the discharge between 180 and 230 • C. The temperatures of the two Cs ovens were controlled in the range of 185-195 • C in the experiment; the Cs evaporation rate is estimated at 0.17 mg min −1 for each oven. In the pure hydrogen discharge, we confirmed there were no Cs spectrum lines in the arc chamber using the visible spectrometer. Therefore we considered the Cs-free condition in the arc chamber before the Cs conditioning. Figure 3(b) shows the H − density measured by CRDS at z = 2 mm from the PG surface. After Cs conditioning for 1000 min with 500 discharges, the H − density increases to 1.25 × 10 17 m −3 from 0.3 × 10 17 m −3 for the pure volume discharge. We determined the decrease in the H − density during a 1 s beam extraction with −8.1 kV extraction voltage (V ex ), as shown by the grey field in figure 3. The reduction value of the H − density ( n H − ) is 28% of the H − density close to the PG surface for the case with Cs conditioning. Figure 3(c) shows the electron saturation current (I es ) measured by electrostatic probe in the extraction region at z = 11 mm from the PG surface. A large I es current flows into the probe tip for the pure hydrogen discharge. The I es decreases after Cs conditioning, which is a result of maintaining charge neutrality in the extraction region by increasing the number of H − ions. We note that the signal waveform of the probe is influenced not only by electrons but also by H − ions in the case of ion-ion plasma with Cs [20]. We considered that the actual electron density is relatively smaller during Cs operation than is expected by the waveform of I es in the extraction region. We also found that the increase in the I es current during beam extraction is caused by the influx of electrons into the extraction region from the discharge region. Figures 3(d) and (e) show the H α and H β line intensities, respectively, at z = 11 mm. The H α emission was obtained by imaging system with the exposure time of 60 ms. The H β emission is measured by OES. The exposure time in Cs case is 40 ms. However, the time resolution of the H β emission for pure volume is poor with the exposure time of 1 s because of the OES system used for confirmation of Cs lines in the arc chamber. The H α emission intensity is increased by Cs conditioning because of an increase in the excited-state hydrogen population due to the mutual neutralization process between H + and H − . The contribution of the dissociative recombination process between H + 2 and electrons for H α emission appears to be decreased by Cs conditioning, because we found that the decreasing H β intensity is caused by a decreasing electron density. In the case of a pure volume discharge, H α and H β increase with beam extraction, which is consistent with the increased I es signal and is due to the increase in the electron flux. On the other hand, for a discharge with Cs, we find a reduction in the H α emission during beam extraction. This reduction results from the decrease in the excited-state hydrogen population created by the mutual neutralization processes, which in turn results from a decrease in the H − density. Similar changes in the hydrogen Balmer lines were also observed in RF negative hydrogen ion sources [27]. At the same time, we observed constant H β signals during beam extraction in Cs operation because this is due mainly to the electron density. Thus, the influence of the electrons on the process of H α emission during beam extraction is negligibly small in the rich H − ions at optimal Cs condition. Here the reduction value of the H α intensity owing to beam extraction is defined as H α , which is the key value to understand the H − behaviour.

Figures 4(a) and (b)
show the spectrum images of the H α distribution taken by the imaging spectrometer in the extraction region before beam extraction (i.e. only with the arc discharge) and during beam extraction, respectively. Strong reflections of the tungsten-filament radiation and H α emissions from the discharge region appear around the filter flange and the PG apertures. We also confirmed that similar reflection signals only electrified filaments with and without IR cut filter. In the region at the bias insulator at the centre of the LOS, we did not observe such a large reflection. From the spectrometer measurements, we estimate the background filament radiation to be about 5% at the centre of the line of sight (z = 11 mm). The meaningful area for H α emissions is considered as z = 20 mm from the PG surface, shown as the white area in figure 4(c). As shown in the horizontal profile along the z-axis for the LOS centre (figure 4(c)), the H α intensity for the arc discharge only, plotted as a solid line, gradually decreases toward the surface of the PG. The vertical distribution of the H α for the centre of the LOS is uniform in the extraction region, as shown in figure 4(d). As seen in the horizontal profile, the H α line intensity does not change at the filter flange and PG surface during beam extraction. This result indicates that the reflection component of the H α emission and the filament radiation from the discharge area located on the left-hand side of the image before beam acceleration and during beam extraction are the same. For the bias insulator close to the PG apertures, the H α intensity varies depending on the applied extraction voltage shown as dotted lines in figures 4(c) and (d); the H α emission is reduced near the PG surface at the upper side of the image. Figure 5 shows the distribution of H α in the extraction region for the different types of discharges. The distribution images of H α were produced by subtracting the image acquired  before beam extraction from the image acquired during beam extraction. Here the decrease is represented in red, the constant zone in green and the growth in blue. The white area is signal saturation on image pixels caused by the reflection of filament radiation. For a constant background and with constant arc discharge, the observable area for H α distribution expands to z = 35 mm from the PG. Figure 5(a) is the case of a pure hydrogen discharge. The H α strongly increases during beam extraction for the whole area of the extraction region. It indicates either an increase of the hydrogen density or the electron density or the electron temperature. Since the feeding gas pressure and the arc discharge power stay constant during beam extraction, the neutral hydrogen density is considered to be constant in this experiment. We also confirmed the constant electron temperature of 1.3 eV measured by the electrostatic probe. As shown in figure 3(c), the I es signal, which is strongly dependent on the electron density, increased. And also the H β emission, which is mainly due to dissociative recombination process, also increased as shown in figure 3(e). These two results indicate that the electron density in the extraction region increased with beam extraction. We observed the 9 A high extraction current in the pure volume discharge; it is mostly electron currents which include small negative ions. It is reasonable to think that the extracted electrons from the PG apertures flowed from discharge region to extraction region. Therefore the H α distribution shows the contribution of the dissociative recombination process by electrons widely distributed in the extraction region in pure volume discharge; even so, a small hollow structure appeared beside the PG apertures. This is consistent with the small reduction in H − density shown in figure 3(b). Figure 5(b) shows the H α image after 340 min (with 170 discharges) with Cs conditioning. The reduction structure shown in red clearly appeared beside the PG apertures. We found that the reduction area expanded to the upper side of the image near the Cs evaporator. Although the H α reduction appeared near the PG surface, the increasing area remains in the upstream region in the left-hand side of the image. At the reduction area, it is clear that the electron excitation process for H α emission replaces the mutual neutralization process between H − and H + ions, which is owing to the increasing proportion of H − ions. Therefore, the reduction phenomenon comes from the PG where there are many H − ions produced by Cs-covered surface. Figure 5(b) also shows the small increase in H α (in blue) at the inner surface of the PG apertures. This is the reflection signal of the H α emitted around the filter flange area. The mirror image structure caused by the reflection signal appears on the right-hand side of the PG; this structure is in agreement with the reflection image on the photograph shown in figure 2. Figure 5(c) shows the H α image after 1000 min (500 discharges) with Cs conditioning. The reduction area has spread in the direction of the bottom side of the image, and has also expanded to the inside of the plasma farther than 30 mm from the PG surface. The extraction current decreases to 4.4 A with the condition of low I es current (figure 3(c)). The electron temperature of 2.3 eVs is stay constant during beam extraction. We found a strong spot-like reduction in the region close to the PG surface (z < 10 mm) beside the apertures, which have tails to the left side of the image. The result of this H α distribution leads to the speculation that the H − ions are widely distributed in the extraction region because of the optimal Cs conditioning. On the other hand, we also found that the asymmetric H α distribution remains after 1000 min Cs conditioning at the bottom of the image where the last PG aperture row is close to the bottom chamber wall. We considered that the asymmetric H α distribution results from a fraction of the element of negative ions and electrons due to the different Cs conditioning on the PG surface. If the H α distribution reflects reduction structure of the negative ions, this asymmetric distribution correlates with the beam distribution extracted from the aperture. According to the previous study of the uniformity of negative ion beam, the same beam asymmetry appeared at the peripheral region near the chamber wall [9][10][11]. However, it is not evident whether the beam distribution was under the influence of H α distribution, because we did not measure the beam distribution owing to the non beam acceleration in this experiment. The comparison of H α distribution and the beam distribution should be investigated in the future. . The H α observed inside the area of z < 4 mm, as shown by the grey-coloured area, is disabled because of the short integral distance of the LOS that is blocked by the PG surface. We first found that the H α reduction expands widely to 30 mm from the PG surface in three profiles. This result indicates that the H − ions are widely distributed in the extraction region before beam extraction using arc discharge. Then, it is apparent that the H α reduction increases as the plotting point approaches the PG for the cases of the UA and CA profiles passing through the PG apertures. That the negative ions are lost in the region of the apertures during beam extraction is clear. On the other hand, in the SU profile, a weak H α reduction appeared near the PG surface while there is a strong H α reduction in the area 10 mm < z < 20 mm. This indicates that the production locations of the negative ions are present on the PG surface, and the influence of its production expands to 20 mm from the PG surface at least during beam extraction. It is likely that a similar reduction response occurs between the emission and the H − density and the H α emission. The H − ions distributed in the extraction region compensate for the reduction of the H − ions close to the PG aperture. Therefore, more probable reason seems to be that the reduction of the H − density is caused by the particle loss due to extraction as the negative ion extraction is current limited.  reaction of mutual neutralization increases close to the PG surface (z < 15 mm) and is caused by an increase in the H − density owing to surface production. The reduction area expands to the whole area of the extraction region (z < 30 mm) after 1000 min of Cs conditioning. Therefore, it is appropriate to consider the production source of the H − ions existing at the PG surface. Figure 7(b) shows the variation in the H α profile along the y-axis at z = 4 mm from the PG. The horizontal and vertical axes are the H α and the distance from the centre aperture, respectively. We find hollow structures on the H α , which are caused by the reduction in H − ions, at y = 0 and ±18 mm. This structure is consistent with the configuration of the PG apertures, which have a 12 mm diameter and a 19 mm interval. Comparing the pure volume discharge against the discharge with Cs conditioning after 340 min, the positive H α distribution starts to change to a negative value in the upper side of the image near the location of the Cs evaporator (figure 1). Then the negative H α area expands to the bottom after 1000 min of Cs conditioning. The fact that a difference appears in the reduction in H α caused by the reduction in H − ions by location suggests that the efficiency of the negative ion production varies depending on the distribution of Cs on the PG surface. Figure 8 shows the comparison of the H α distribution applied against an extraction voltage of −3.1 kV (in figure 8(a)) and −8.1 kV (in figure 8(b)) after 1000 min of Cs conditioning; here the image of figure 8(b) is same as figure 5(c). The arc discharge power is maintained constant during beam extraction, and the gas pressure and the bias voltage are set the same for both discharges. Comparing the two images, it is obvious that the higher extraction voltage produces a larger H α reduction in the extraction region. However, the reduction structures are similar in both cases. This result indicates that there is a close relation between the reduction in H α and the strength of extraction voltage by way of the H − ion density. Figure 9 shows the reduction profile of H α along the z-axis at y = 0 mm (i.e. through the centre of the PG aperture) for low and high extraction voltages. The deep reduction structure lies closer to the PG aperture in the region of z < 10 mm for the high extraction voltage (V ex = −8.1 kV). To consider the relation between the reduction of H α and the H − density, we compare these values close to the PG apertures z < 5 mm where negative losses occur due to beam extraction. Figure 10(a) shows a plot of the extraction current (I ex ), the absolute value of the reduction H − density (| n H − |) and the reduction in the H α intensity (| H α |) against the absolute value of the extraction voltage (|V ex |). The current density of 10 mA cm −2 for I ex is extracted at |V ex | = 8 kV. Here, I ex is the drain current (i.e. mixed H − ions and electrons) in the circuit between the EG and a direct current extraction voltage power supply; it is linearly dependent on V ex . Indicating data of a negative ion current and electron current is appropriate here, but they cannot be separated because we did not measure the ion current by a beam calorimeter due to non-beam acceleration in this experiment. The | n H − | is observed at 2 mm from the PG surface; those reductions also depend linearly on the extraction voltage. There seems to be a close relation between the extracted current with a negative charge and the reduction in the H − density near the PG apertures. When the extraction current is 10 mA cm −2 , the | n H − | is 3.4 × 10 16 m −3 . The current density J is the product of the charged particle density n and the drift velocity v. As we assume the extraction current is carried by | n H − |, the drift velocity v = J/n = 1.85 × 10 4 m s −1 . The flow energy of the negative ion is estimated as 1.8 eV, a value close to the electron temperature in the extraction region of this source. A further important point is that the absolute value of the reduction in the H α intensity measured at z = 4 mm, which is as close as possible to the position of H − measurement, also increases with the extraction voltage. We found that the increase in | H α | with increasing extraction voltage was the same as | n H − |. Linear dependence appears between | H α | and | n H − | as shown in figure 10(b). This result indicates that the reduction in the intensity of H α is the result of a reduction in the H − ion density by way of mutual neutralization process in a rich H − ion environment after Cs conditioning close to the PG surface.

Influence of the electron impact excitation process
It is important to note the influence of the electron impact excitation process on H α emissions in the extraction region of the negative ion source. Through Cs seeding, H − ions produced at the PG surface are diffused in the extraction region and forming an ion-ion plasma close to the PG surface. Therefore, the charge neutrality is conserved with H + and H − ions in the low electron current condition less than 1% those of the ions [20]. In the high proton ratio (∼80%) hydrogen negative ion source for NBI, the main excitation mechanisms for H α emission in the extraction region are dissociative recombination and mutual neutralization. The dominant reaction is determined by the fractions of H − ions and electrons. Hence, the dominant excitation mechanism for H α emission becomes mutual neutralization in the case of ion-ion plasmas because they have a large H − fraction near the PG surface. Evidence for this scenario can be seen in the H − , H α , H β and I es observations shown in figure 3. The small amount of electron penetration from the discharge region during beam extraction was observed with the electrostatic probe [20], which has been confirmed by particle-in-cell (PIC) simulations [17]. Optical measurement is also consistent with such electron behaviour in the pure volume discharge. This contribution to the hydrogen emission is considered to be negligibly small in the ion-ion plasma in optimal Cs conditions. To obtain more detail about the electron behaviour and the analysis of the H α excitation process, the local electron density needs to be measured. It is also useful to construct a realistic simulation model for ion-ion plasma with beam extraction in a hydrogen negative ion source.

Integral data analysis along the line of sight
Considering the line integral intensity for H α imaging spectroscopy, the measured H α intensity shows not only the mutual neutralization process but also other excitation processes such as dissociative recombination. In the peripheral region outside the extraction area, the electron excitation process might be an influential element owing to the low number of H − ions because of less Cs conditioning, owing to the position of the Cs evaporator being optimized for maximal beam distribution. However, if the line integrated H α intensity contains such elements, the reduction value of the H α reflects the reduction in H − ions only in the extraction area near the PG apertures. Therefore, it is reasonable to suppose that the distribution image of H α applies to the reduction structure of the H − ions at the centre of the negative ion source, which is rich in H − ions after Cs conditioning.

Extraction of negative ions
Surface produced H − ions can reach PG apertures following two processes: one is direct recoil of H − ion, which is considered the major process of H − ion extraction in RF ion sources. Plasma potential of RF ion source is more than 20 V and positive ions accelerate towards the plasma grid. The positive ions are converted to H − on the conical surface surrounding PG aperture with low work function, and the H − ions gather at the PG aperture. This process is supported by numerical simulation of the RF ion source [16]. The other is a sequential process. The H − ion produced on the PG surface moves to the source plasma initially, and changes direction by Lorentz force and charge exchange and/or elastic collisions with hydrogen atoms. The H − ions relax their energies during this process, and some of them are extracted after arriving at a meniscus. In the former case, the H α reduction during beam extraction should concentrate around the conical surface of plasma grid. In our experiment, however, the H α reduction distributes widely in the extraction region. This feature corresponds to the latter case described above, and is consistent with the H − distribution measured with CRDS [30]. The difference of the former and latter mechanisms of H − production is caused by the plasma potentials in the RF ion source and filament-arc source; typically less than 30 V [19] and less than 5 V [20], respectively. Based on the result of H α imaging spectroscopy, we conclude that 'stray' H − ions are extracted from a wide extraction region in the filament-arc source.

Conclusion
The imaging spectroscopy diagnostic tool successfully worked to observe the distribution of H α emission and its reduction structure in the extraction region in the negative hydrogen ion source. We found significant reductions in the distribution of the H α emission, due to a decrease in the H − density caused by the decrease in the mutual neutralization process. This spectrum structure clearly shows that the reduction in the extracted H − ions generated at the PG surface is widely distributed in the extraction region. These results will have considerable impact on numerical analysis for the transport modelling of H − ions in the extraction region in negative ion sources. The diagnostic technique of H α imaging spectroscopy, which is a powerful tool for experimentally determining the extraction behaviour and distribution of negative ions, will strongly contribute to the development of stable high-power operation for NBI.