TALIF at H− ion sources for the determination of the density and EDF of atomic hydrogen

The production of H− ions in negative ion sources relevant for particle accelerator facilities and neutral beam injection systems is based predominantly on the surface conversion of H atoms at a low work function surface covered with caesium (the plasma grid (PG)). Therefore, the H atom density n H and energy distribution function (EDF) close to the PG determine the amount of surface produced H− ions. As a direct method for the density and EDF determination, two-photon absorption laser induced fluorescence (TALIF) on H atoms was implemented at the ion source of the teststand BATMAN Upgrade (BUG) being the first time that this was accomplished at an H− ion source. Several challenges had to be overcome concerning the application of the diagnostic at the complex facility and the evaluation of the fluorescence signals against a bright H α background. The observed line profiles suggest a Maxwellian EDF with an H atom temperature of (2000±500)  K. The presence of highly energetic H atoms (measured by optical emission spectroscopy, (OES)) could not be resolved by the TALIF system due to the insufficient signal-to-noise ratio. Atomic densities were measured for H2 and D2 plasmas for varying ion source parameters at BUG resulting in values between 3×1018  m−3 and 1.1×1019  m−3 for hydrogen. For the operation with deuterium, 30% higher atomic densities are observed for similar ion source parameters which agree well with the previous results obtained with OES.


Introduction
Both particle accelerator facilities and the neutral beam injection (NBI) system for the fusion experiment ITER are based on negative hydrogen ion sources. For the former, negative hydrogen ion sources serve as a source of protons which are * Author to whom any correspondence should be addressed.
Original Content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. produced via charge-exchange injection where the two excess electrons are stripped from the particle within a thin foil [1]. Examples of such sources are the Linac4 H − source which is operated at CERN Switzerland [2], the H − source operated at J-PARC, Japan [3] and the H − ion source operated at the Oak Ridge National Laboratory, USA [4]. For the NBI at ITER, negative hydrogen or deuterium ions are necessary instead of positive ions due to the high energy (870 keV for H and 1 MeV for D) which is required for the ITER operation [5,6]. At such high energies, the neutralization efficiency of positive ions is very limited whereas it is still 60% for H − . Experiments which investigate H − ion sources towards their use for NBI are for example the teststands BATMAN Upgrade (BUG) (equipped with the prototype ion source of ITER) and ELISE at IPP in Garching, Germany [7], the testbeds SPIDER and MITICA operated at Consorzio RFX in Padova, Italy [8] and the H − ion source operated towards the NBI development for the LHD located at NIFS in Toki, Japan [9]. In all of the stated H − ion sources, H − is produced via the surface conversion of H atoms and positive ions at a caesiated low work function surface. These parental particles are produced in a low pressure low temperature plasma which is (except for the source in [9]) formed with the aid of inductive coupling of RF power. In [9] arcs are used in order to generate the low pressure low temperature plasma. Among H atoms and H + x ions (x = 1, 2, 3), the former are considered to account for the majority of produced H − ions [10] such that the flux of H − from the converter surface is determined directly by the H atom flux towards the surface and the conversion yield. The H atom flux is determined by the density n H of H atoms and their energy distribution function (EDF) while the conversion yield depends heavily on the EDF [11,12] and the work function of the surface [13]. Therefore, in order to characterize and optimize a negative ion source, a determination of n H and the EDF at a position close to the converter surface is strongly desired.
Such measurements have been reported for example in at BATMAN (Upgrade) [14][15][16]. They were conducted via optical emission spectroscopy (OES) in combination with collisional radiative (CR) modelling for the determination of n H and by observing the line broadening of Balmer emission lines for the determination of T H (and thus the EDF). The ratio of atomic to molecular hydrogen was also determined in the same way in [17] comparing results from the RF driven ion source at BATMAN and the arc source KAMABOKO III which was then operated at CEA in Cadarache, France. In addition, [18] reports the determination of T H by means of tunable diode laser absorption spectroscopy (TDLAS) on the H α transition (n = 2 → n = 3) at the NIFS negative ion source. However, close to the converter surface a plasma is present in which recombination of molecular ions contributes significantly to the Balmer line emission. The OES analysis is therefore not trivial and relies on many further input parameters resulting in a high uncertainty of the determined n H value. Furthermore, the line profile obtained by OES and therefore T H stems strictly speaking from excited H atoms which could potentially have a different temperature than the atoms in the ground state. The same holds for the TDLAS measurements as H atoms with n = 2 are probed with the laser. A direct measurement of n H and the EDF was not yet carried out at a negative ion source. A method which could potentially accomplish this is two-photon absorption laser induced fluorescence (TALIF) [19] where a strong pulsed laser at λ = 205 nm is used in order to excite ground state atoms into the n = 3 state via the simultaneous absorption of two photons. The subsequent fluorescence in the course of the H α transition is used for diagnostic purposes. This diagnostic exhibits a high complexity due to the high demands to the laser system and further optical components. On the other hand, high voltage and high RF power are used for the operation of H − ion sources which might cause damage and disturbance in the delicate laser device if it is located too close to the ion source. Further, the diagnostic access to the ion source is often limited. This is why TALIF has not yet been performed at an H − ion source. In this work, the implementation of TALIF at the ion source of the BUG teststand is reported and measurements of n H and the H atom EDF are shown being the first time that TALIF is applied at an H − ion source. As an intermediate step of the installation process, TALIF was first applied at a small scale reference inductively coupled plasma (ICP) setup. This allowed for an extensive characterization of the diagnostic without blocking the operation of the teststand.

The BUG teststand
The teststand BUG is only described briefly in this work. A more thorough description can be found in [7,20,21]. It consists of three parts which are the H − ion source, the extraction and acceleration system and the beam dump calorimeter. The ion source as well as the extraction and acceleration system are shown in figure 1. A H 2 or D 2 plasma is generated and operated in a cylindrical driver via inductive coupling. The driving RF generator is operated with up to 100 kW and its RF frequency is 1 MHz. It is connected to the RF coil wrapped around the driver via a matching network containing matching capacitors and a matching transformer [22]. The impedance matching is such that the reflected power almost vanishes such that perfect matching is established. In [22], the power transfer efficiency was determined to be on the order of 50%-60% for all investigated ion source parameters. The maximum plasma pulse length is 100 s which is imposed by the overheating of the present plasma facing grid. In the driver, the H atoms and positive H + x ions required for negative ion production are produced predominantly via electron impact reaction. The plasma containing H, H 2 , e − and H + x expands from the driver into the expansion region of the source towards the plasma grid (PG) where H − is predominantly created via surface conversion. This process is enhanced via the evaporation of caesium from a Cs oven which is attached above the driver. The source is operated at the negative high voltage potential (up to −45 kV). The negative ions are extracted through the PG by the extraction grid (EG) which is kept at up to +10 kV relative to the PG. The ions are fully accelerated by the grounded grid which is kept at ground potential. The PG can be biased positively with a set current reducing the co-extracted electron current [23].
For this study, the ion source is operated at filling pressures between 0.3 Pa and 0.6 Pa while the applied RF power is kept between 20 kW and 70 kW. Both H 2 and D 2 are used as carrier gases. For these parameters, the effective electron temperature in the driver is commonly found to be ∼10 eV while the electron density is close to 10 18 m −3 [15,16,24]. At such high temperatures, H − is destroyed very efficiently by electron stripping [25]. Thus, the electrons need to be cooled after they leave the driver in order to increase the H − mean free path close to the PG where they are created via surface conversion. This is done by creating a horizontally oriented magnetic filter field by driving a current I PG on the order of a few kA vertically through the PG. The resulting magnetic field is typically on the order of a few mT such that the e − are magnetized causing a reduction of both T e and n e [26,27]. Thus, in front of the PG, T e ∼ 2 eV and n e ∼ 10 17 m −3 are achieved [15,16]. For this study, I PG is kept at 1.5 kA (→ 3 mT close to the PG) for the operation with H 2 and at 2.2 kA (→ 4.5 mT close to the PG) in D 2 operation. A higher filter field strength for D 2 operation is necessary to limit the higher co-extracted electron current observed typically in D 2 discharges [28].
As the region in which H − is produced is close to the PG, this is also the place where TALIF should be installed. Here, OES measurements yield n H values between 6 × 10 18 m −3 and 8 × 10 18 m −3 at 0.3 Pa [15] and values between 1 × 10 19 m −3 and 2 × 10 19 m −3 at 0.6 Pa [14]. Concerning the H atom velocity distribution function, a two temperature distribution among H atoms with a cold share at 2200 K and a hot share at 2.5 eV is reported in [16].

The reference ICP setup
In order to characterize the TALIF diagnostic system and to tune it best for the ion source application, a reference ICP experiment was built and TALIF was installed there. A CAD explosion graph of the setup is depicted in figure 2. Its main component is a vacuum vessel that is made of stainless steel with a diameter of 153 mm and a height of 100 mm. It is equipped with eight access ports. The vessel is terminated with a stainless steel bottom plate and an Al 2 O 3 top plate. It is evacuated down to a background pressure on the order of 5 × 10 −7 mbar with a vacuum pump system consisting of a rotary vane pump and a turbomolecular pump. H 2 , D 2 or Kr gas can be fed into the vessel with mass flow controllers. The gas pressure in the vessel can be controlled with a manual gate valve between the plasma vessel and the vacuum pumps. A plasma is driven inductively with a planar RF coil located above the Al 2 O 3 plate which is connected to an RF generator via a matching network. The RF generator has a maximum output power of 600 W and operates at 13.56 MHz. The experiment is operated at pressures between 2 Pa and 10 Pa which is considerably higher than the pressure present at BUG. With this pressure range and with addition of other gases, a broad n H range could be obtained.

Two-photon absorption laser induced fluorescence
For TALIF on H atoms, the excitation scheme as outlined in [29,30] is used. Here, ground state H atoms are excited into the n = 3 state via the simultaneous absorption of two photons at a wavelength of λ = 205.08 nm. Subsequently, the excited atoms decay in the course of the H α transition and emit photons at 656.3 nm which are collected with a photomultiplier tube (PMT) and processed for the density determination. The two-photon absorption cross section σ (2) H of H atoms was calculated in [31] and is small such that a pulsed and focussed laser is necessary in order to generate a sufficient laser intensity.
Per laser pulse, the total number of fluorescence photons N F which reach the PMT is given by [19] where g(λ) is the line profile of the two-photon transition while the laser is operated at the wavelength λ and I(t) is the temporal intensity profile in the probed volume. The time integrated photomultiplier signal is in course proportional to N F itself. Thus, normalizing it with´∞ −∞ I(t) 2 dt gives the TALIF signal S H TALIF (λ) which is proportional to n H · g(λ). The line profile shape yields information about the H atom EDF as Doppler broadening is the dominant broadening mechanism. Pressure broadening can be neglected here due to the low pressure both in BUG and the reference ICP [32] and Stark broadening can be disregarded for electron densities below 10 19 m −3 [33]. Integrating S H TALIF (λ) spectrally gives S H TALIF which is solely proportional to the n H within the probed plasma volume.
The associated proportionality constant between n H and S H TALIF is determined via calibration of the diagnostic setup after [30] using the noble gas krypton at a known density n Kr . This is done due to the spectrally close two-photon absorption transition of ground state Kr at 204.13 nm into the 5p ′ [3/2] 2 state. Fluorescence from this state is emitted at 826.3 nm. TALIF is performed on this transition in order to determine the corresponding TALIF signal S Kr TALIF . With this, the density of H atoms is given by [19,30] In this equation, ξ H/Kr is the sensitivity of the PMT and T H/Kr is the transmission of the fluorescence collection optics at the fluorescence wavelengths of H and Kr respectively. a H/Kr are the respective branching ratios of the fluorescence transitions and is the ratio of the two two-photon absorption cross sections. It was determined in [30] to be 0.62 (±50%). T H/Kr were calibrated while using an Ulbricht sphere as radiation standard while the ratio ξKr ξH was determined by cross calibration against an absolutely calibrated spectrometer.
The branching ratio of a transition i (i = H, Kr) is given by a i = Ai Atot being the ratio of the Einstein coefficient of the observed transition A i and the sum of all Einstein coefficients of the associated upper level A tot . Quenching can be disregarded in this work due to the low pressure that is present both for the operation of the ion source (0.3-0.6 Pa) and the calibration with krypton (6 Pa).
a Kr can be taken from [34] which results in a value of a Kr = 0.977 for the observed Kr transition. For the H α transition, further considerations are necessary. Following the selection rules for a two-photon transition derived in [35], excitation from the H atom ground state 1 2 S 1/2 is only possible into the 3s state (3 2 S 1/2 ) and the 3d states (3 2 D 3/2 and 3 2 D 5/2 ). The associated H α branching ratio is 1 since both 3s and 3d states do not exhibit a dipole transition into the ground state. However, a transfer of excited particles from the 3s and 3d states into the 3p states (3 2 P 1/2 and 3 2 P 3/2 ), so-called statemixing, is observed in [36,37] leading to a branching ratio of a H = 0.441 since the 3p states can decay in the course of Lyman-β decay. Whether the one or the other case is present in an experiment can be determined by evaluating the fluorescence decay time τ H with the PMT which corresponds to the lifetime of H (n = 3) atoms. In the case of no state-mixing, this time is determined by the decay of the 3d states leading to τ H = 15.4 ns. For this, the excitation into the 3s state can be disregarded since the cross section for the 3d excitation is larger by a factor of 7.56 and the Einstein coefficient for the H α transition of the 3s state is smaller by one order of magnitude compared to the 3d states [19]. In the case of state-mixing taking place, the lifetime gets shortened due to the presence of the Lyman-β transition. If the n = 3 fine structure states are populated according to their statistical weights, τ H = 10 ns. This means that the measurement of τ H can be used in order to determine the value of a H required for the experiment.

Diagnostic setup
The laser system that is used for the laser pulse generation is depicted in figure 3. Its core is a tunable dye laser (Sirah Lasertechnik: Cobra Stretch) which is seeded by a frequency doubled Nd:YAG laser (Innolas: SpitLight Compact DPSS 100) which emits laser pulses at 532 nm. The dye laser produces laser pulses within a wavelength range of 606 nm and 627 nm which are frequency tripled using two β barium borate (BBO) crystals within a third harmonic generation unit. This results in a UV laser beam which is tunable within 202 nm and 209 nm. At the laser exit, the energy per pulse of the UV laser is 600 µJ at a repetition rate of 100 Hz. The pulse length of the laser beam was determined with a photodiode to be 6.3 ns. From the TALIF line profile measurements at krypton follows that the spectral profile of the laser can be approximated by a Gaussian with a FWHM of (1 ± 0.1) pm.
The UV laser is attenuated using a variable attenuator which consists of a λ 2 plate and a Rochon polarizer. A plain quartz window is used as a beam splitter which guides a small part of the laser light onto a photodiode (Alphalas: UPD-300-UP) that is used in order to record the temporal intensity profile of the laser. It is protected against overexposure with a diffusor plate which is positioned between the window and the diode. The diode is connected to an oscilloscope (LeCroy: LT264, bandwidth: 350 MHz) in parallel to a 50 Ω resistor. The temporally integrated photodiode signal is further used in order to determine the energy per laser pulse in BUG. For this, it is calibrated with an energy meter (Ophir Photonics: PE50U-DIFH-C).
Five mirrors are used in order to guide the laser beam towards BUG over a distance of 12 m. Over this distance, the laser beam size increases due to a slight divergence of the laser beam. This is corrected by employing a telescope system close to the laser system consisting of two convex lenses of equal focal length which is also depicted in figure 3. The laser beam within BUG is depicted in figure 4. It is guided into the experiment from below in a distance of 25 mm upstream of the PG. A lens with a specified focal length of 750 mm is used in order to focus the laser beam into the vertical center of the PG. The fluorescence light is collected orthogonal to the laser direction at this position with an optics system consisting of an optical fibre, an achromatic lens (focal length: 85 mm), a spectral filter and a slit mask (width: 3 mm). Two specific spectral  filters are used for the observation of the krypton fluorescence (central wavelength: 830 nm ± 10 nm) and the hydrogen fluorescence (central wavelength: 650 nm ± 40 nm). The slit mask is oriented parallel to the laser beam. This is necessary to reduce the amount of H α background radiation compared to the fluorescence. The signal-to-background ratio was observed to be on the order of 0.1.
The optical fibre is connected to a PMT (Hamamatsu: H11706-20) which is used to observe the fluorescence. For this, it is connected to the oscilloscope in parallel to another 50 Ω resistor. The TALIF signal S i TALIF (λ) (i = H, Kr) at laser wavelength λ is determined via where U PMT (t) and U PD (t) are measured time resolved voltages from the PMT and the photodiode respectively.

TALIF signal acquisition
A sufficient resolution of the fluorescence radiation against the H α background produced by the plasma itself is obtained by averaging the PMT signal over 4000 laser pulses. A single measurement is shown in figure 5 after subtracting the background radiation revealing a signal-to-noise ratio of 2-3 in the center of the transition line. This figure marks the first time that TALIF is successfully implemented at a negative hydrogen ion source.
In order to evaluate the recorded signals, no noise filtering techniques such as Fourier analysis or Savitzky-Golay filtering can be applied due to the lack of temporal resolution (1 ns). Therefore, the PMT signal is fitted with an expected fluorescence signal. This emerges as the solution to the differential equation describing the temporal evolution of the density of H (n = 3) atoms excited via two-photon absorption by a laser with a rectangular temporal intensity profile. The differential equation is given in [30]. That solution is given by C is a proportionality constant serving as a fitting parameter, t 0 is the starting time of the laser pulse, t pulse is the laser pulse length and τ H is the fluorescence decay time (i. e. the lifetime of H (n = 3) atoms). This fit is shown in red in figure 5. t pulse is fixed to the experimentally determined value. The value of τ H is also fixed at 10 ns. This is also motivated from measurements at the reference ICP for plasmas with varying n H values [38]. Here, decay times between 11 ns and 14.5 ns are observed for H atom densities between 10 19 m −3 and 2.5 × 10 20 m −3 . The fluorescence lifetime was observed to rise for increasing H atom density which is attributed the presence of state mixing while photon re-absorption of Lyman-β photons causes a prolongation of the upper state lifetime. Such a behaviour was also observed in [37] for TALIF on H atoms and it is investigated deeper in [39,40]. At BUG, n H values at the lower edge of the investigated interval or even lower are expected at the region of interest such that for the evaluation of the TALIF signals, it is assumed that the lifetime of the excited state is 10 ns with the associated branching ratio a H = 0.441. These considerations leave C and t 0 as free parameters of the performed fitting routine. S H TALIF (λ) is obtained by integration of the fitted function.

Depletion correction of the TALIF signals
In order to generate sufficient fluorescence light, the laser energy per pulse was kept at least at 120 µJ. This value is in the depleted regime of TALIF which means that equation (1) does not hold due to the depopulation of ground state atoms, amplified spontaneous emission or ionization of H (n = 3) atoms with a third photon all reducing the resulting observed fluorescence. This effect can be corrected after [41], where the fluorescence signal F (corresponding to´U PMT (t) dt) is described by Here, E L is the energy per laser pulse such that E 2 L translates to´[U PD (t)] 2 dt. α is the proportionality constant between F and the squared laser energy per pulse in the case of no depletion occurring, i. e. the quadratic regime of the laser energy. β corresponds to the TALIF depletion strength. S X TALIF (λ) can thus be corrected with the knowledge of β. This parameter is determined in [41] by measuring F as a function of E L and fitting equation (5) to the result. This is, however, not possible for TALIF on hydrogen based on the TALIF signal obtained at BUG due to the low signal-to-noise ratio leading to a high error bar in β. Therefore, another approach is chosen for this work. TALIF is performed at the reference ICP on H, D and Kr for varying energy per laser pulse. A fit of equation (5) Here, it is assumed that the depletion occurs at the same laser intensity in both experiments. It therefore accounts for the differing conversion factors between laser intensity and photodiode signal which is caused by the differing laser beam paths to the reference ICP and BUG. This assumption is justified by the two possible depletion mechanisms as outlined in [19] being the depletion of the ground state or three-photon ionization. Collisional quenching could cause a depletion at higher energy per laser pulse but as both discharges operate at such low pressure, this process can be neglected. Figure 6 shows the measured´U PMT (t) dt for varying laser intensity and the described fits to the TALIF data. In figure 6(a), this is done for Kr both in the reference ICP and at BUG and in (b) for H and D in the reference ICP. The measured data is obtained for a laser wavelength at the line center for all depicted graphs. The shown fits exhibit good accordance with the measured data.
From these fits and after the application of equation (6) The different values of the depletion factors of the two isotopes means that the quadratic regime stops already at a lower energy per laser pulse for D atoms compared to H atoms which was observed in the reference ICP. This could be caused by one of the depletion mechanisms being more pronounced for D which is, however, not expected from the considerations carried out for example in [19]. This remains therefore unresolved, but the measurement is treated as valid hereafter.
The correction factor is validated in figure 7 which shows values of S H TALIF (λ) obtained at BUG measured for the two different values of laser energy per pulse of 120 µJ and 170 µJ, both clearly not in the quadratic TALIF regime. On the left side of the graph, the data points are not corrected for signal depletion leading to a deviation of the two results while the result obtained for 170 µJ per pulse is lower. Thus, the depletion of the TALIF signal is stronger for 170 µJ per pulse. By applying  the correction as outlined earlier the two data points coincide. Compared to the data point for 120 µJ, the correction leads to an increase of S H TALIF (λ) by 47%. The match of the two results after applying the correction proves the validity of the depletion correction procedure for the investigated energy per laser pulse.

Experimental results
The time which is necessary to acquire the 4000 laser pulses required for one TALIF measurement is roughly 80 s due to time delays within the acquisition routine and during the data transfer from the oscilloscope to the operating PC. For the acquisition of one TALIF signal S H TALIF (λ), the maximum BUG pulse length of 100 s is therefore chosen and the wavelength of the laser is varied between the BUG pulses in order to record a line profile. In order to guarantee the temporal stability of the plasma parameters which is affected by Cs redistribution during long plasma pulses in H − sources Cs is not injected into the source [28]. Subsequently, the amount of co-extracted electrons is very high such that the extraction was switched off.

Two-photon absorption line profile measurement
TALIF line profiles were measured for the pressure variation for H 2 as the main effect on the H atom temperature is expected here. An exemplary profile is depicted in figure 8 for a pressure of 0.3 Pa and 70 kW of forwarded RF power. Due to the high noise level seen in figure 5, the fluorescence signal could not be resolved against the noise at the wings of the profile which is indicated by the gray shaded area. In order to fit the measured data, multiple line profiles are shown in the figure which arise from Maxwellian EDFs at various temperatures including the broadening caused by the laser line width. No change of the line profile was observed for the pressure variation.
The measured line profile corresponds to a H atom temperature of T H = 2000 K ± 500 K. This value is close to the value of 2200 K observed in [16], whereas the second, 'hotter' component (2.5 eV [16]) of the line profile could not be observed by TALIF. The intensity ratio of the 'cold' and 'hot' component ξ c/h of the line profiles observed by OES is between 0.3 and 1 depending on the ion source parameters and the observed Balmer emission line. If ξ c/h = 1 is present among ground state atoms, a 'hot' component would not be apparent in the observed TALIF line profile due to the low signal-tonoise ratio. This holds for ξ c/h values down to 0.5 assuming a 'cold' temperature of 1500 K and a 'hot' temperature of 2.5 eV as observed in [16]. A better signal-to-noise ratio is required for the observation of the fluorescence signal in order to access the wings of the profile in order to resolve the presence of a 'hot' component of H atoms. The main source of the noise present in figure 5 which prevents low fluorescence intensity measurements is the PMT itself. This could be shot noise at the photo cathode or during the electron multiplication process at the dynode stages. The former can be reduced by collecting more light on the cathode for example by using bigger optics for fluorescence collection [42]. The latter could be improved by increasing the gain of the PMT which causes an increase of the voltage between the dynode stages [42] in combination with gating the PMT as was for example done in [43].
The same line profile and thus the same temperature is observed for D atoms at 0.6 Pa and 70 kW. For the TALIF measurements apart from the pressure variation, the laser wavelength was set at the center of the line profile. In order to determine absolute n H values, the Doppler broadened line profile at a temperature of 2000 K is chosen for the data evaluation.

Variation of the source pressure
The absolute densities of H and D atoms after equation (2) throughout the pressure variation is depicted in figure 9. For both isotopes, a rise of the atomic density is observed for increasing pressure. The absolute values range from 4.5 × 10 18 m −3 to 1.1 × 10 19 m −3 for the H 2 operation and from 7 × 10 18 m −3 to 1.3 × 10 19 m −3 for the D 2 operation. Here, the small error bars denote the random error while the big greyed out error bars denote the systematic error.
The systematic error arises from uncertainties in the transmission of the used spectral filters (±10%), the PMT quantum efficiency ratio (±20%), the uncertainty of  The atomic density of deuterium is higher compared to the same BUG pulses with operated with hydrogen by a factor of 1.2-1.5 which compares well with the isotope effect that was observed in [14]. This effect could be caused by two points. First, a higher electron density was measured in the driver for D 2 plasmas at the same pressure and RF power in [22] which leads to an increased dissociation rate of molecules in the driver. Secondly, the diffusion of D atoms to the walls where they are lost due to recombination is slower due to their higher mass compared to H atoms. This is apparent by evaluating the diffusion formula given in [44].

Variation of the forwarded RF power
The result of the RF power scan at 0.3 Pa and 0.6 P for an H 2 plasma is depicted in figure 10. For both pressure values, a rise of n H is observed, which is from 3 × 10 18 m −3 to 4.5 × 10 18 m −3 for 0.3 Pa (factor 1.5) and from 4 × 10 18 m −3 to 1 × 10 19 m −3 for 0.6 Pa (factor 2.5).
In general, the rise of n H can be explained by a rise in n e caused by the higher amount of RF power coupled to the plasma as quantified in [22]. This causes in turn an increase of the H 2 dissociation rate. The fact that this rise is less pronounced for 0.3 Pa compared to 0.6 Pa is attributed to neutral depletion which depends strongly on the pressure as is observed at BUG experimentally [45] and results from model calculations [46,47]. Neutral depletion describes a reduction of the density of neutral particles due to an elevated electron pressure in the discharge center, strong heating of neutral particles or the strong ionization of neutral particles in the discharge [48].

Consequences for the conversion of H atoms
The converted H − flux from the PG is determined by the impinging flux of H atoms and the conversion yield. The latter depends strongly on the incident energy of H atoms [11,12]. The data of these publications can be used in order to check the consistency of the measured H atom density and temperature with the H − current that is extracted from the ion source. For this, the energy dependent yield needs to be weighed and averaged with the EDF imposed by T H in order to calculate the effective yield present in the ion source. Doing that results in a conversion yield below 0.01% [38]. For the ion source parameters 0.3 Pa and 70 kW, the observed n H is 4.5 × 10 18 m −3 which results in an H atom flux of 7.3 × 10 21 m −2 s −1 assuming an isotropic EDF such that the negative ion flux Γ H − from the cesiated surface is Γ H − = 5.9 × 10 17 m −2 s −1 . The total H − current that is released from the PG can be computed by integrating Γ H − over the PG surface (excluding the apertures) [38]. Its value is 0.0026 A. In contrast, the total extracted current at BUG in a well conditioned state at the same source parameters (0.3 Pa, 70 kW) is 1.5 A which is higher by orders of magnitude. This is true despite not even considering the extraction process of H − produced at the PG. The associated extraction probability was calculated in [49] to be on the order of 20%.
A 'hot' share of H atoms as it was reported in [16] with T H = 2.5 eV could potentially resolve this discrepancy as the corresponding conversion yield is above 30% in this case. This highly energetic share of H atoms is also observed in a DSMC model of the ion source [46] and is assumed to be caused by the recombination of positive ions at the vessel walls. Prior to wall recombination, they are accelerated towards the vessel walls in the plasma sheath. In order to verify the presence of those atoms, the TALIF line profile needs to be resolved better in its wings.

Conclusions
TALIF on H atoms was successfully implemented at the H − ion source of BUG without Cs evaporation which marks the first reported implementation of TALIF at a negative hydrogen ion source. The implementation was achieved by first installing the diagnostic at a reference ICP experiment in order to characterize the TALIF system, to gain experience with it and to fine tune the system. A distance of 12 m had to be overcome with the laser beam in order to reach BUG which was achieved with guiding mirrors. TALIF measurements could be performed close to the PG despite a high signal-to-background ratio enabling for measurements of the H atom density and EDF. The determination of the TALIF line profile, however, is limited to the line center by the signal-to-noise ratio. From this data, an H and D atom temperature of (2000 ± 500) K at all investigated source parameters was determined. Absolute H atom densities between 3 × 10 18 m −3 and 1.1 × 10 19 m −3 were found depending on the source pressure and forwarded RF power. For the source operation with D 2 , a higher atomic density (by a factor 1.2-1.5) was observed compared to H 2 operation at comparable source parameters. In order to resolve the presence of energetic H atoms and to further increase the measurement accuracy, however, improvements of the system aiming for a better signal-to-noise ratio are required.
For the measured values of n H and T H , the production of H − is not sufficient in order to account for the H − ion current that is extracted from BUG. The presence of highly energetic H atoms is suspected in order to resolve this discrepancy. A share of highly energetic H atoms is not excluded by the measurements of this work as they could not be resolved in the line profile due to the low signal-to-noise ratio in the wings of the two-photon absorption line profile. An improvement of the TALIF system towards a better signal-to-noise ratio is therefore planned. Further next steps are TALIF investigations in a cesiated ion source in order to relate n H and T H with the H − density in the plasma measured by cavity ring-down spectroscopy (CRDS) and with the extracted H − current.

Data availability statement
The data cannot be made publicly available upon publication because they are not available in a format that is sufficiently accessible or reusable by other researchers. The data that support the findings of this study are available upon reasonable request from the authors.
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 thank Dr Arne Meindl (IPP) for thorough proof reading of this work and the valuable discussion about TALIF.