Enhancing surface production of negative ions using nitrogen doped diamond in a deuterium plasma

The production of negative ions is of significant interest for applications including mass spectrometry, particle acceleration, material surface processing, and neutral beam injection for magnetic confinement fusion. Methods to improve the efficiency of the surface production of negative ions, without the use of low work function metals, are of interest for mitigating the complex engineering challenges these materials introduce. In this study we investigate the production of negative ions by doping diamond with nitrogen. Negatively biased ($-20$ V or $-130$ V), nitrogen doped micro-crystalline diamond films are introduced to a low pressure deuterium plasma (helicon source operated in capacitive mode, 2 Pa, 26 W) and negative ion energy distribution functions (NIEDFs) are measured via mass spectrometry with respect to the surface temperature (30$^{\circ}$C to 750$^{\circ}$C) and dopant concentration. The results suggest that nitrogen doping has little influence on the yield when the sample is biased at $-130$ V, but when a relatively small bias voltage of $-20$ V is applied the yield is increased by a factor of 2 above that of un-doped diamond when its temperature reaches 550$^{\circ}$C. The doping of diamond with nitrogen is a new method for controlling the surface production of negative ions, which continues to be of significant interest for a wide variety of practical applications.

One application of particular interest is the creation of negative-ion beams suitable for MCF neutral beam injection, which has a proposed requirement of accelerating a 40 A current of deuterium negative ions to 1 MeV 16 . This primarily utilises negative ion surface production, as distinct from volume production, to increase the density of negative ions close to the extraction grid [20][21][22] .
Negative ion production from plasma facing surfaces can be enhanced through the application of a low work function alkali metal 22 . Current methods apply a thin layer of caesium to the extraction region of the ion source 23 . This is achieved by injecting caesium vapour into the plasma and allowing it to condense onto the inside of the ion source 24 . There exist some limitations with this approach, such as controlling the application of the caesium so that it condenses in the right locations and at a rate that is sufficient to maintain an optimum thickness at the extraction grid 25 . Additionally, this method introduces complex engineering challenges, eg. equipment maintenance and potential for caesium pollution 26,27 . Alternative materials to caesium are therefore of interest.
Dielectric materials are of particular interest as an alternative to low work function metals 26 .
Generally, for atoms approaching a surface, the affinity level of the atom is gradually downshifted until it overlaps with the surface material's valence band. Electrons can then tunnel from the valence band of the surface to the approaching atom and form a negative ion, this is the so-called resonant charge transfer (RCT) process, as summarised in Ref. 37. For a metal, the conduction band is situated on top of the valence band. When a newly created ion begins to leave the surface, the probability of electron loss through tunnelling back to the conduction band of the surface is high due to the resonance between the affinity level of the negative ion and the empty states of the conduction band.
This means that most metals produce negligible negative ions through surface ionisation processes 38 .
Unlike most metals, caesium can be used to enhance negative ion production because it has a low work function. This increases the distance at which the resonance between the affinity level of the new ion and the empty conduction states occurs, reducing the probability that the electron tunnels back to the surface 37,38 .
In contrast to metal surfaces, where the conduction band lies on top of the valence band, the band gap of dielectrics suppresses the tunnelling of electrons from a newly created negative ion back to the material's surface. This means that a new negative ion can travel a larger distance away from the surface before reaching a point where its affinity level is in resonance with the empty states of the conduction band. The increased distance of the ion from the surface reduces the probability that the electron associated with the new negative ion will tunnel into the empty states of the conduction band, thereby increasing the negative ion yield 26,39 .
One potential drawback to the use of dielectric surfaces is that to generate a negative ion the atom-surface distance for a dielectric must be much smaller than for a metal. This is due to the larger energy gap that the occupied valence band states lie beneath the vacuum level 37 . Fortunately, the atom-surface interaction process is amplified by the Coulomb interaction between a negative ion and a localised hole in the surface material 40 . This can result in a high ionisation efficiency as demonstrated in beam experiments [41][42][43] . For these reasons, dielectrics are of interest as an alternative to low work function metals for the surface production of negative ions.
Carbon surfaces are one prospective category of materials that are of interest for replacing low work function metals where negative ions are to be produced. For instance, DLC has been used to produce negative ions from incoming neutral particles for a spacecraft particle detector, when low work function metals would not have been appropriate 44 . Of the forms of carbon, diamond has particularly beneficial properties: • It is a dielectric with a large band gap (5.5 eV) 45 that suppresses the destruction of negative ions as they leave the material's surface • It can be grown to have 'designer' properties such as the preferential growth of a particular crystal face to alter the electronic structure of its surface 45 • When it is being grown, dopants can be introduced to change its effective work function and electron affinity [46][47][48][49] • It can have a negative electron affinity when the surface is hydrogen terminated 45 , which reduces its effective work function by reducing the energy gap between the valence band and the vacuum level. This is thought to have a positive influence on negative ion production 26 • When heated to 450 • C, diamond has previously been shown to produce five times more negative ions compared to other forms of carbon e.g. graphite 32 As a means of increasing the production of negative ions, previous work with diamond has investigated using single, nano-and micro-crystalline diamond and also p-type doping of micro-crystalline diamond (MCD) using boron 28 . The n-type doping of diamond using nitrogen has not previously been studied in this context and it is thought that it could lead to favourable properties for negative ion production for two reasons. Firstly, previous studies of the electronic properties of nitrogen doped diamond have demonstrated that nitrogen doping creates a deep donor level in the band gap of the diamond at 1.7 eV 50 . This lowers the effective work function to approximately 3.1 eV 51 , which is lower than boron doped diamond (3.9 eV) 51 and un-doped diamond (∼4.5 eV, with hydrogenated surface and negative electron affinity) 52 . Secondly, it is thought that having the aforementioned deep donor level of electrons close to the vacuum level could increase the negative ion production from diamond by creating a source of electrons close to the vacuum level 53 .
In this study, we investigate the production of negative ions from nitrogen doped diamond films in a low pressure deuterium plasma. Comparing micro-crystalline nitrogen doped diamond (MCNDD) with un-doped micro crystalline doped diamond (MCD) and previously investigated micro-crystalline boron doped diamond (MCBDD) 26,33 , we consider 'low energy' (11 eV) and 'high energy' (48 eV) ion bombardment conditions at the surface as a mechanism for increasing the negative ion yield. The experimental methods are described in section 2: plasma source in 2.1, sample holder in 2.2 and the measurement method in 2.3. The micro-crystalline diamond samples are described in section 2.4, with the surface characterisation using confocal microscopy and Raman spectroscopy described in 2.5. The results are presented in section 3.

Method
The experimental setup is shown in figure 1. It consists of a low pressure deuterium plasma source, a temperature controlled sample holder, and a mass spectrometer for the measurement of negative ions produced at the diamond film's surface.

Temperature controlled sample holder
The sample holder is shown in figure 1 (b). It is attached to a DC voltage source (Equipment Scientific Alimentation de Laboratoire CN7C) that can negatively bias the frame that holds the diamond sample.
The voltage applied to the sample is defined as V DC , which is distinct from the voltage at the sample surface, V S . The sample is positioned 37 mm away from, and perpendicular to, the plane of the mass spectrometer orifice. This is the closest distance that the sample can be placed in front of the mass spectrometer orifice, and is assumed to be sufficiently small to achieve minimal negative ion signal loss. It has previously been demonstrated that this distance has negligible effect on the shape of the negative ion distributions measured by the mass spectrometer 28,33 .
It is worth noting that the angular dependence of the NIEDFs for carbon materials has previously been shown to be similar 33,35,36 . Therefore, a single measurement can be used to compare between samples. A misalignment of the sample surface normal to the mass spectrometer would produce spurious results, so to prevent this, the alignment is regularly checked by rotating the sample and maximising the negative ion signal.
As shown in figure 1 (b), a tungsten heating element is built into the sample holder, which is used to heat the back of the sample. The heating element is controlled by a PID controller (designed and built by AXESS tech) using a K-type thermocouple inside the the frame of the sample holder. By fixing a second thermocouple to the surface of the sample, its temperature is calibrated against the temperature measured by the PID. The heating element behind the sample increases the temperature of the sample's surface up to (750±20) • C. The mass spectrometer polariseable orifice potential is calibrated so that a nearly planar plasma sheath is formed in front of the orifice, as determined by a particle-in-cell (PIC) simulation 54 .
The potential on the surface of the samples accelerates any negative ions created through surface interactions away from the sample and through the plasma to the mass spectrometer. The low pressure of the plasma means there are few collisions between the plasma and the negative ions 54,55 . Any collisions that do occur with the deuterium plasma would predominantly be detachment collisions with deuterium molecules which would neutralise the negative ions, thus preventing measurement of negative ions that have undergone collisions 18,54,55 . The plasma potential in front of the mass spectrometer prevents negative ions generated through volume production processes in the plasma from entering the mass spectrometer, therefore the energy of any negative ions that are measured must have been accelerated away from the sample surface 54,55 . The negative ions are detected at an energy corresponding to the energy they possessed when they were created which can then be shifted by the kinetic energy gained between the sample and the mass spectrometer 54,55 . Presented NIEDFs are shifted to present the kinetic energy the negative ions have at the surface of the samples.
The secondary electrons emitted from the surface of the sample are filtered out within the mass spectrometer.
Positive ions impacting the samples are assumed to dissociate during impact 26,56 , splitting the energy of the ion into its component particles (ie. for D 3 + , 3 deuterium nuclei). This means that because the plasma is predominantly composed of D 3 + ions, the modal energy of the ions striking the samples' surface is E M = e(V S +V p )/3, where V p is the plasma potential, giving approximately energy' ion bombardment and 'high energy' ion bombardment, respectively.
The choice of −130 V is made to align with previously published work, whilst −20 V is chosen as this is the lower limit of what can be reasonably used to ensure effective self-extraction of negative ions from the sample surface into the mass spectrometer 28,33 .

Procedure for measurement of negative ion energy distribution functions (NIEDFs)
Measurements were undertaken using the following method: • The plasma was brought to steady state as determined by measurements of the positive ion energy distributions using the mass spectrometer • The negative ion counts for each sample were integrated with respect to energy and then divided by the positive ion current measured neglecting the possible small changes with temperature to the isolated sample to give the relative negative ion yield for the sample 33 . This is given in arbitrary units as the mass spectrometer is not calibrated to count an absolute number of negative ions.
Recent measurements show that an absolute negative ion flux could be measured using a magnetised retarding field energy analyser via the technique described in Ref. 57. A detailed investigation of this topic remains the subject of future work, but to provide some context for the presented results,

Micro-crystalline nitrogen doped diamond
The nitrogen doped diamond films were created using a similar PECVD technique to the MCD and MCBDD samples of Ref. 28 so only a brief summary is provided here. The PECVD process utilised a bell jar reactor with a pressure of 200 mbar, microwave power at 3 kW, substrate temperature of 850 • C, background hydrogen gas mixture with a methane concentration of 5%. The ratio of nitrogen in the gas mixture was set as a means to vary the concentration of nitrogen in the MCNDD film. Each film was deposited on to a (100) orientated silicon wafer.

Surface characterisation
The samples were analysed using confocal microscopy and Raman spectroscopy prior to plasma exposure to characterise their properties.   The spectra are presented with the background subtracted (in order to aid clarity and comparison between samples) and normalised to the carbon sp3 peak, observed here at 1333 cm -1 . The laser wavelength used for the measurements is 514 nm.

Surface morphology and crystal structure
Raman spectroscopy was undertaken to measure the relative concentration of nitrogen dopant that is introduced into the MCD films with respect to the gas phase nitrogen concentration present during the PECVD process. Raman spectra were generated using a Horiba Jobin Yvon HR800 setup. The measurements were undertaken in air using an excitation wavelength of λ L = 514 nm, ×100 objective  The broad peak centred at 2100 cm -1 observed in figure 3 can be attributed to nitrogen vacancy centres (NV 0 ) that have been introduced into the diamond 63 . This broad peak appears, not due to vibrational modes, but due to the electronic signature attributed to nitrogen vacancy centres and in reality lies at an energy level of 2.15 eV. As Stokes Raman spectroscopy is energy loss spectroscopy, this peak appears arbitrarily at 2100 cm -1 when using a 514 nm laser. Using another laser to perform the Raman spectroscopy results in a change in the wavenumber of this peak 64 .
As the measurement configuration is the same for all samples, a relative comparison of the number of nitrogen centres in the diamond can be made using the broad 2100 cm -1 peak. This can then be used to infer relative nitrogen concentration 65 . As shown in figure 3, the ratio of the NV 0 peak to the peak centred at 1333 cm -1 increases with increasing gas phase dopant concentration, for samples 0 ppm to 50 ppm (200 ppm will be discussed below). This is consistent with previous work, which showed a similar increase in the magnitude of the NV 0 characteristic peak with an increase in the gas phase nitrogen doping 66 .
In figure 3, the Raman spectrum of the 200 ppm nitrogen doped diamond has a peak at 1500 cm -1 that has a much higher intensity than the other samples. This peak is of particular interest as it is associated with the sp2 bond of carbon that has previously been associated to graphite-like bonds 62 .
The ratio of the peaks at 1333 cm -1 and 1500 cm -1 implies that there is a higher ratio of graphite in the 200 ppm diamond film compared to the other samples 28,67,68 . The 200 ppm nitrogen doped diamond sample also exhibits a NV 0 centre peak at 2100 cm -1 , which is slightly lower than the 50 ppm sample, suggesting a reduction in the number of nitrogen vacancies, and therefore, a reduction in the concentration of nitrogen in the diamond.
The observed increase in intensity of the NV 0 peaks, increasing from 0 ppm to 50 ppm, may be attributed to both the increase in nitrogen introduced in the gas phase and by a change in the crystal face from a mix of (100) and (111) has with the crystal face orientation and the measurable number of nitrogen vacancy centres. This is an active area of research 71 . However for this study, as nitrogen doping is the main influencing factor that generates the differences between the samples, it is reasonable to suggest that it is possible to associate the nitrogen gas phase doping with the negative ion yield and this is how the samples will be defined in the next section. and ∼150 • C, MCD does undergo a transition, though this is smaller than that seen for MCNDD.

Nitrogen doped diamond: influence of the dopant concentration
In order to understand these trends, it is important to note that the negative ion yield measured For example, the maximum yield for MCD and MCNDD (10 ppm) is at the peak of a gradual increase and decrease in yield as temperature is increased from ∼30 • C to ∼400 • C and then from ∼400 • C to ∼750 • C respectively. This is most clearly observed in figure 4 (b) for the 10 ppm sample. This sample is distinct from the other MCNDD samples as the 10 ppm sample exhibits an increase in yield as the temperature is increased (due to a change in conductivity) then a further smaller increase up to a maximum negative ion yield at∼400 • C. The yield then gradually decreases as the temperature is increased further. The other MCNDD samples also undergo an increase in yield due to a change in conductivity, but no further gradual increase in yield is observed as temperature is increased.
It could therefore be reasonable to suggest that the peak yield conditions are not observed due to a lack of conductivity for samples with more than 20 ppm gas phase nitrogen doping. The trends observed for the 0 ppm and 10 ppm samples suggest that a temperature of approximately 400 • C may be the temperature at which these MCNDD with more than 20 ppm gas phase doping produce the highest yield. A technique to measure negative ions that does not require a conductive surface would be necessary to explore this further.
In figure 4 (a), the effect of the nitrogen doping on the maximum yield is not readily observed when a bias voltage V DC = −130 V is applied to the sample. This is unlike figure 4 (b) in which a bias voltage of V DC = −20 V is used. In this data there is an observed difference between the nitrogen doped and non-doped diamond. The yield in figure 4 (a) for the MCNDD samples and MCD samples is also lower than the yields from all of the samples in figure 4 (b). A higher bombardment energy as a result of the high magnitude bias is associated with an increase in sp2 bond formation in diamond 36 . It is reasonable to suggest that the reduction in yield for the higher magnitude bias creates more sp2 defects which decreases the yield. Additionally, if the yield is not changing with the addition of nitrogen to the diamond, it is also reasonable to suggest that the nitrogen doped diamond may be more susceptible to defect formation due to high energy bombardment which would result in a surface state that does not enhance the negative ion yield through nitrogen doping. Additional work would be necessary to characterise this process.
The apparent influence of nitrogen doping on the measured negative ion yield is observed in compared to MCNDD (50 ppm) sample is therefore consistent with previous work, which observed that an increased number of sp2 bonds is less favourable to negative ion production 28,33 . This work suggests that this is still the case with nitrogen doped diamond samples.

Mechanism for the surface production of negative ions
NIEDFs for MCBDD and MCNDD are presented in figure 5 to compare negative ion production processes between MCBDD, a previously studied material 26  increases as the surface temperature increases. This is because the main contribution to the measured yield is low energy ions, which are predominantly created through the sputtering process, as distinct from backscattering, due to the acceptance angle of the mass spectrometer 26,54 . Previous work has confirmed this interpretation through comparison of experimental results with SRIM simulations 36 .
The physical interpretation for the decrease in the sputtering contribution is that this is due to a decrease in the amount of sub-surface deuterium available for sputtering as a result of out-gassing caused by the increase in temperature. 35,36 . This implies that MCNDD has similar negative ion production properties to MCBDD.
The trends for MCNDD and MCD observed in figure 4 and discussed in the previous section can be explored in the context of figure 5. Figure 4 shows that the yield for MCD increases up to sample temperatures of ∼400 • C and decreases as its temperature is increased further. This is similar to the trends observed for samples of MCNDD when they are conductive. The increase and then decrease in yield as temperature is increased, from ∼30 • C to ∼400 • C and then from ∼400 • C to ∼750 • C respectively, can be attributed to two processes that combine to generate the observed trend in figure 4. The first process is the removal of defects on the sample surface. The heating of the sample results in an enhancement of the etching of sp2 bonds created by the bombarding positive ions resulting in a surface which results in a higher ratio of sp3 bonds 28 . The increased proportion of diamond bonds on the surface increases the negative ion yield, as explored in previous work through Raman spectroscopy 26,28,32,33 . The second process is the previously discussed decrease in the sputtering contribution to the negative ion yield due to out-gassing of deuterium from the sample surface, as observed in the measurements of figure 5. As temperature is increased, the influence of each of these processes on the measured negative ion yield is observed to vary significantly. At temperatures below ∼400 • C, the reduction in defects increases the yield, whilst the outgassing does not cause a significant decrease in the sputtering contribution. At temperatures above ∼400 • C, the decrease in sputtering contribution reduces the yield by a greater extent than the reduction in defects caused by the elevated temperature, causing a reduction in the the measured negative ion yield 28 .
For the samples of nitrogen doped diamond with more than 20 ppm nitrogen added in the gas phase, the MCNDD film is not conductive at temperatures where the previously mentioned reduction in the defects can increase yield, i.e. between 30 • C and 400 • C. A more thorough exploration of the resulting interplay between the reduction of defects and the decreasing sputtering contribution is beyond the scope of this experimental study. However, figure 5 (b) suggests that the decrease in yield due to a decrease in the sputtering contribution is consistent with current understanding of the behaviour of negative ion formation on micro-crystalline doped diamond.

Negative ion yield: comparison between MCNDD, MCBDD, and MCD
The negative ion yield with respect to sample surface temperature of the MCD, MCBDD and MCNDD samples is shown in figure 6 agrees with MCD and MCBDD. In the high energy bombardment regime, the yield from MCNDD is found to be lower than MCBDD and comparable to MCD. This suggests that the higher positive ion bombardment energy is having a larger influence on MCNDD than MCBDD, though a mechanism for such a difference is beyond the scope of this study.
In figure 6 (b) for V DC = −20 V, the trends for MCD and MCBDD are also observed to be qualitatively similar, showing an increase in yield by a factor of 2 and a factor 1.5 from 150 • C to 400 • C respectively, and a gradual decrease in yield above 400 • C, which has been discussed in section 3.1 26,33 . Figure 6 (b) has a similar trend as figure 6 (a) where the yield from MCNDD at temperatures below 400 • C is effectively zero. The yield increases by several orders of magnitude between 400 • C to 550 • C, after which it decreases gradually. At temperatures above 550 • C the general trend of decreasing yield is consistent with both MCD and MCBDD, and agrees with current understanding of these diamond films as discussed in the previous section. Of particular interest is that the yield for MCNDD in this low energy ion bombardment condition is observed to be higher than MCD, and also higher than the previously best performing type of diamond, MCBDD 26 . At 550 • C, the maximum yield observed, MCNDD has a higher negative ion yield than MCD and MCBDD by a factor of 2 and 1.5, respectively.
This therefore suggests that controlled addition of nitrogen during the growth of diamond using the PECVD process could be an avenue for increasing the negative ion yield from diamond.

Conclusion
In this study, we have investigated the nitrogen doping of diamond films as a means of increasing the negative ion yield during exposure to a low pressure deuterium plasma (2 Pa, helicon source at 26 W). For conditions where positive ions from the plasma bulk bombard nitrogen doped diamond film with energies of 11 eV and 48 eV, 'low energy' and 'high energy' bombardment, respectively, mass spectrometry measurements are used to determine the negative ion yield as the film temperature is scanned between 30 • C and 750 • C. For 50 ppm nitrogen doping, introduced in the gas phase during diamond growth using the PECVD technique, the application of low energy ion bombardment is observed to increase the negative ion yield by a factor of 2 compared to un-doped diamond and a factor of 1.5 compared to boron doped diamond.

Acknowledgements
The 6 References