Ammonia production in a dual crossed atom beam experiment

Production of ammonia by surface reactions of H and N atoms on surfaces not wetted by partially ionized plasma may represent an important technological issue in fusion reactors where puffing nitrogen is employed to cool plasma in the divertor region. The H and N atoms are likely to interact on such surfaces forming NH3 molecules. The interaction efficiency was studied in a laboratory setup consisting of two separate sources of either N or H atoms. Both sources enabled experiments with atoms at room temperature in the range of H-atom density of the order of 1021 m−3 and N-atom density of the order of 1020 m−3. The production of ammonia was measured with a calibrated residual gas analyser. The production depended on the fluxes of both atoms onto the surface of selected materials. As a general rule, the higher H-atom flux at a constant N-atom flux caused an increase in ammonia production. The highest efficiency of up to 50% was found for nickel. It was up to 30% for tungsten, whereas for P92 alloy, it was up to about 20%. The accuracy of these results is within about ±20% of the measured values. Methods for suppressing ammonia formation in fusion reactors will have to be invented in order to enable appropriate long-term operation.


Introduction
The formation of ammonia has been studied in mixed N 2 -H 2 discharges since the late 70 s of the previous century, with the aim of finding an alternative to the Haber-Bosch process [1]. Recently the formation of ammonia has become the focus within the field of thermonuclear fusion [2][3][4][5].
Injection of impurities such as N 2 , Ne, Ar, Xe and Kr, also called impurity seeding, is routinely used to mitigate the heat load of the most exposed inner surfaces of several fusion * 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. reactors and to improve the energy confinement in the hot thermonuclear plasma [6,7]. Among the tested impurities, the best results, both in terms of energy confinement as well as heat load mitigation, have so far been obtained with nitrogen [8].
Here it should be mentioned that in present day devices nitrogen is the best readiator, however this will change with the size of the machine and the power to the separatrix. Based on currently available results, nitrogen seeding will be a mandatory part of the high-powered operation in the upcoming ITER thermonuclear reactor [2,9]. Unlike the other seeded impurities, nitrogen is a chemically reactive element. Once injected into the plasma, it interacts with the surfaces of plasma-facing materials [10,11]. The interaction may lead to the formation of nitrides as well as heterogeneous surface association with hydrogen atoms, thus forming ammonia [9,10,12]. In the case of operation with a deuterium-tritium fuel mixture, as planned for ITER, the injection of nitrogen will cause the formation of partially tritiated ammonia, which is a severe operational and security concern. Pumping of tritiated ammonia will thus contribute to the machine tritium inventory [13]. The maximum tritium retention of the in-vessel components is set at 700 g [14]. Here we have to mention also other issues such as corrosion problems of ammonia within the gas processing plant and exhaust pumping lines and the fact that the tritium processing plant is not designed to work with ammonia. A reliable prediction of the amount of ammonia formed in N 2 -seeded discharges is necessary to optimise the ITER plasma operation, as well as develop ammonia removal methods.
As the low gas density in fusion plasmas favour surface reactions over volume reactions, ammonia is likely predominantly produced in surface reactions. The fluxes of nitrogen and hydrogen radicals are the largest at the most exposed (plasmawetted) surfaces, so it is reasonable to expect that this is also where the majority of ammonia production occurs. Beside the plasma-wetted surfaces, ammonia production can also occur on plasma-shaded surfaces, i.e. surfaces without direct contact with the plasma, where only neutral radicals (i.e. H and N atoms) prevail. The fluxes to the plasma-shaded surfaces are significantly lower than to the plasma-wetted surfaces, so the production of ammonia is expected to be lower accordingly. However, the ammonia which is formed at plasma-wetted surfaces enters the partially ionized plasma after desorption. Ammonia is very efficiently dissociated even in cold laboratory plasma [15], so it is likely that ammonia, formed at the plasma-wetted surfaces, is dissociated back into N and H radicals. Therefore, plasma-wetted surfaces cannot significantly contribute to net ammonia production despite the higher radical flux densities. On the contrary, ammonia formed at plasma-shaded surfaces faces very little danger of plasmaphase destruction. Thus, the reactions at the plasma-shaded surfaces are expected to significantly contribute to ammonia production in a fusion reactor.
The formation of ammonia on plasma-shaded surfaces competes with a heterogeneous surface recombination reaction of H and N atoms to form parent molecules. The heterogeneous surface recombination of H and N atoms to H2 and N2 molecules has been studied extensively. A review paper on recombination probabilities has been published as [16]. In general, the recombination coefficient was found to depend on the type of material. Some materials exhibit a high probability of heterogeneous surface recombination to parent molecules. These materials will chemisorb H or N atoms, and the surface atoms will then interact either with another surface atom or an atom arriving from the gas phase to associate to a stable molecule. The first reaction was called Langmuir-Hinshelwood and the latter Eley-Rideal [17]. The reactions are difficult to distinguish since most authors just measured the loss rate of atoms on a surface and then calculated the recombination coefficient. Whatever the reaction, the coefficient is of the order of 0.1 for many metals. For stainless steel, for example, the coefficient is about 0.10 for hydrogen atoms [18], and 0.07 for nitrogen [19]. These values were measured on smooth surfaces. The coefficient is larger at rough surfaces and may approach 1 for materials of extremely rich topography, such as nanowalls [20].
On the contrary, very few results have so far been published on the reactions between neutral atoms of different kinds, especially the surface reaction between N and H atoms, which results in the formation of the NH x radicals and, further to NH 3 molecules [21]. Partially, this is due to the fact that most authors studied ammonia formation in mixed N 2 -H 2 discharges [4,22,23]. With the exception of the system used by Van Helden et al [24], such setups do not allow for a separate study of the contributions of ionic and atomic species. Moreover, the fluxes of N and H radicals in such systems are challenging to characterise. The authors usually give the fraction of nitrogen in the gas mixture as the main experimental parameter. However, changes in the composition of the gas mixture may impact the plasma parameters as well. This, in turn, influences the ionisation and dissociation of the N 2 and H 2 molecules, thereby modifying the absolute densities of the precursors, as well as causing the dissociation of the surfaceproduced ammonia. Thus, the N 2 fraction in the gas mixture impacts the ammonia production indirectly, coupled to the properties of each specific experimental system, and gives only limited information about ammonia production, so any generalisation is questionable.
In order to address these frequent shortcomings of laboratory studies of ammonia production, we developed an experimental system dedicated to the study of ammonia production through surface recombination of neutral hydrogen and nitrogen atoms. By using two independent plasma sources of H and N atoms, the system allows for independent characterisation and control of the neutral atom densities in the vicinity of surfaces. Moreover, with the use of this system, we ensure that only neutral atomic species take part in the reactions and that there is no subsequent plasma-phase destruction of the surface-produced ammonia. This paper presents the results of systematic measurements of ammonia production in the experimental system.

Experimental
The experimental system is shown in figure 1. The reaction chamber is in the shape of a cylindrical tube with an inner diameter of 36 mm. The tube is made from glass of low coefficient for heterogeneous recombination of H or N atoms to parent molecules. The chamber is pumped with a two-stage rotary pump (Edwards E2M80) with a nominal pumping speed of 80 m 3 h −1 , which allows for a base pressure of about 0.1 Pa. Two plasma sources provide jets of partially dissociated H 2 and N 2 at the sub-sonic velocity, typically around 100 m s −1 , at the entrance to the reaction chamber. In each case, the gas was partially dissociated by passing partially ionized plasma, which was sustained using a surface wave discharge powered by microwave power supplies coupled through surfatrons. The discharges were ignited in quartz glass tubes with an inner diameter of 6 mm, which were inserted into the surfatron cavities, and connected to the reaction chamber, as shown in figure 1. The surfatrons were water-cooled while the quartz tubes were cooled by forced air, so the temperature of the quartz tubes close to the exhaust to the reaction chamber remained close to room temperature at all discharge powers. Flow rates of gases passed through the discharge tubes were adjusted with Aera FC-7700 Mass Flow Controllers. The pressure was measured inside the reaction chamber with the MKS Baratron 722 A absolute pressure gauge.
The quartz tubes were positioned in such a way that the gas jets crossed at a distance of 1 cm from the mouth of each quartz tube, as shown in figure 1. A Catalytic Probe (CP) was used to measure the densities of neutral atoms. The probe was mounted to the system as shown in figure 1. The catalytic tip was in the position of the confluence of the gas jets. The partial pressure of ammonia was measured with a differentially pumped Residual Gas Analyser (RGA) Pfeiffer Vacuum PrismaPlus QMG 220. The RGA was connected to the system through a narrowed glass tube which provided the pressure reduction necessary to maintain the pressure in the RGA chamber below 10−5 mbar. Each RGA measurement was performed long enough to allow the ammonia-related peaks in the mass spectra to stabilise. To verify there was no back-streaming of the molecules from the reaction chamber into the discharge regions, an Avantes AvaSpec-3648 Optical Emission Spectrometer (OES), with 0.5 nm resolution in a wavelength range from 200 to 1100 nm, was used. A standard CP with a catalytic tip made from high-purity cobalt, which is described elsewhere [25,26], was used to measure the densities of both types of neutral atoms.

Calibration of the atom sources
The plasmas in both atom sources were characterised by OES. OES was measured approximately 1 cm below each MW surfatron where the plasmas were the brightest. Typical spectra are shown in figure 2. The plasma of the N-atom source (figure 2(a)) emits radiation that arises from excited nitrogen species. The radiation from hydrogen in figure 2(a) is below the detection limit, which proves that no back-streaming of hydrogen from the reaction chamber to the surfatron cavity occurred. The absence of hydrogen in the nitrogen plasma source is explained by the very high speed of gas through the narrow quartz tube. As mentioned earlier, the gas speed is close to 100 m s −1 , so the concentration of hydrogen diffusing against the gas flow is negligible. The same applies to the hydrogen source. Figure 2(b) indicates only hydrogen atomic lines, and any radiation arising from nitrogen-excited species is below the detection limit. The plasma sources, therefore, assure for production of N and H atoms practically free from impurities.
The efficiency of both plasmas as sources of neutral atoms was evaluated by measuring the densities of neutral atoms in the reaction chamber. The measured densities are shown in figures 3 and 4, plotted against the microwave generator forward power and pressure of the working gas, respectively. The error bars in these figures are mainly intrinsic errors derived from the CP.
Generally, higher powers and higher flow rates (i.e. pressure in the reaction chamber) result in higher densities within the limits of experimental parameters. The density of neutral nitrogen atoms is of the order of 10 20 m −3 , while the density of hydrogen atoms is of the order of 10 21 m −3 . The dissociation fraction of nitrogen molecules is thus of the order of 0.01, while it is over 10% for hydrogen. The difference is explained by the different dissociation energies: 4.5 eV for H 2 and 9.8 eV for N 2 . The measured densities of both atoms enable calculation of the partial pressures of either H or N atoms in the reaction chamber using the standard vacuum relation p = nkT, where the p is the partial pressure of particular atoms, n is the density of these atoms in the reaction chamber, k Boltzmann constant and T the absolute temperature. The gas temperature at the cross-section of the gas jets was not measured, but it is assumed to be close to the temperature of the chamber itself. The temperature of the glass tube remained at about 10 K above room temperature even for the prolonged operation of the plasma sources. Figure 5 shows the partial pressures of both N and H-atoms in the reaction chamber at the position of the CP versus the discharge powers. The pressure of each gas in the reaction chamber is the parameter.

Calibration of the RGA
The RGA measures the ion current at a particular ion mass rather than the partial pressure, so it has to be calibrated to enable information about ammonia production. The calibration is normally performed by variation of the gas pressure while measuring the ion current. In the case of probing ammonia, the interpretation of the mass spectra is further complicated due to the significant dissociation of ammonia in the ionisation chamber of the RGA and the presence of water vapour in the differentially pumped RGA. The NH3 ion currents, therefore, appear at ion masses of 17 (NH 3 ), 16 (NH 2 ), 15 (NH), 14 (N) and 1 (H). The N and H atoms may recombine to parent molecules so the ion currents at 28 and 2 may also arise. The water vapour will give ion currents at masses of 18, 17, 16 and 1. The dissociation and ionisation of probed gases in the ionisation chamber of RGA depend on the electron energy as well as the presence of other gases. Our RGA operates at the electron energy of 70 eV, where the ionisation is efficient and the dissociation is suppressed. Still, dissociation is unavoidable. Water molecules arise from the differentially pupped chamber where the RGA is mounted rather than from the reaction chamber, so the peaks at masses 18 and 17 corresponding to water vapour are not influenced by the discharges in plasma sources. The ion current at mass arising from ammonia (NH 3 + , m = 17) should be corrected by subtraction of the contribution of OH + ions.  Figure 6 shows a couple of RGA spectra measured at the same conditions, except that in one case, the discharges of the plasma sources were off and in another on. The nitrogen and hydrogen partial pressures in the reaction chamber were both set to 25 Pa. The insert in figure 6 shows the enlarged range of masses between 15 and 19. Mass 18 in this range prevailed when the discharges were off, but the ion current at mass 17 was not negligible either. The ion current at mass 17 was attributed to OH + ions when the discharges were off. Therefore, this contribution at the given gas composition was subtracted at any attempt to determine the ion current arising from NH 3 + ions only. The ion current arising from these ions is marked in the insert of figure 6.
The ion current at a particular mass does not depend only on the partial pressure of that particular gas but also on the total pressure in the ionisation chamber, which in turn depends on the total pressure of gases in the reaction chamber. The calibration of the RGA, taking into account this effect, was performed at various pressures of ammonia, nitrogen and hydrogen in the reaction chamber. The results of selected measurements are shown in figure 7. The NH 3 + current (with subtracted OH + contribution) was measured versus the ammonia pressure in the reaction chamber. The curve '0 Pa/0 Pa' indicates conditions where ammonia was the only gas leaked into the reaction chamber. Then, hydrogen and nitrogen were introduced in the reaction chamber so that the pressure of either gas was 25 Pa. Ammonia was introduced stepwise to the reaction chamber containing 25 Pa of either H 2 and N 2 and the NH 3 + current was measured versus the ammonia pressure in the reaction chamber. The measured curve is marked '25 Pa/25 Pa' in figure 7(a). Then, the reaction chamber was filled with hydrogen and nitrogen at pressures of either gas 50 Pa and the NH 3 + current was measured versus the ammonia pressure in the reaction chamber again. This curve is marked '50 Pa/50 Pa' in figure 7(a). Although the ion current was measured versus the ammonia pressure, figure 7(a) represents the opposite diagram, i.e. ammonia pressure versus the NH 3 + ion current. This presentation is because the curves, as shown in figure 7(a), were used for calibration of the RGA, i.e. for determination of ammonia production at various N 2 /H 2 mixtures when plasma was on. The measurements were also performed at other pressures of hydrogen and/or nitrogen in the reaction chamber, but they are not present for the sake of the smoothness of this paper.
The results shown in figure 7(a) indicate that the ion current at the same ammonia pressure in the reaction chamber depends on the pressures of other gases. This important fact was taken into account when deducing the ammonia production as a result of interacting H and N atoms in the reaction chamber. One can observe that the slope of the line p NH3 versus NH 3 + ion current is much larger in the case where only ammonia was present in the reaction chamber. The addition of 25 Pa of hydrogen and nitrogen caused a significant drop in the slope of the curve. For example, the NH 3 + ion current of 5 × 10 −10 A corresponds to the ammonia pressure of about 7 Pa. The same ion current corresponds to ammonia pressure of about 4 Pa when the total pressure of hydrogen and nitrogen is 50 Pa, and to about 3 Pa when the total pressure is 100 Pa. Such a behaviour which is not often reported in scientific literature is due to the simple fact that the electrons in the ionisation chamber of the RGA interact with all gases, so the current at a specific pressure of ammonia is influenced by the pressures of other gases (nitrogen and hydrogen in our case).
The same results are also presented a little different in figure 7(b), where the ammonia pressure is presented versus the pressure of each N 2 and H 2 in the mixture for the same NH 3 + ion currents.

Production of ammonia in the empty reaction chamber
The NH 3 + ion currents were measured at various conditions, and the corresponding partial pressures were deduced using  the calibration curves of figure 7(a). The production of ammonia was measured versus the discharge powers used for powering of both nitrogen and hydrogen plasmas. The production of ammonia in the empty reaction chamber (i.e. without any metallic element inserted into the chamber) is shown in figures 8 and 9. Figure 8 was acquired at the pressures of 25 Pa of both N 2 and H 2 in the reaction chamber. The partial pressure of ammonia increases monotonously with increasing discharge power for powering nitrogen plasma. The result is expected since the N-atoms density also increases with increasing discharge power, as revealed in figure 3(a). The higher N-atom density in the reaction chamber, therefore, results in the larger production of ammonia. The ratio in ammonia partial pressures between the nitrogen discharge powers of 100 and 200 W is roughly a factor of 1.6 at the lowest hydrogen plasma power of 100 W, less at higher hydrogen powers. The differences are sound with the behaviour of the N-atom density in the reaction chamber at the nitrogen pressure of 25 Pa (see the middle curve of figure 3(a). Qualitatively, one can deduce that hydrogen atom density plays an important role in the formation of ammonia at a given density of N-atoms. The same applies for the experiments performed at 50 Pa partial pressures of both gases, whose results are presented in figure 9, except that the differences are smaller.
According to the hypothesis, ammonia is produced in the reaction chamber by interacting N and H atoms. The right parameters are, therefore, the densities of these atoms, which are shown in figure 3. Furthermore, since ammonia production is measured in terms of the partial pressure of NH 3 , the parameter that gives the production efficiency will be the partial pressures of both H and N-atoms. Therefore, we plot the partial pressure of ammonia versus the partial pressure of N atoms with the H-atom partial pressure as the parameter. The results are summarised in figures 10 and 11.
The results summarised in figures 10 and 11 are interesting enough and thus worth a discussion. Let us focus on figure 10, which represents ammonia's partial pressure versus nitrogen atoms' partial pressure. The latter expands from about 0.4 to about 0.6 Pa. Such a relatively narrow range is the consequence of the fact that the N-atom density in the reaction chamber does not differ much on discharge power, as revealed in figure 3(a). As expected, the production of ammonia increases monotonously with increasing N-atom density. At the lowest N-atom partial pressure of 0.47 Pa, the NH 3 partial pressure at the lowest H-atom partial pressure of 8.1 Pa is 0.075 Pa, so the efficacy of N transformation to NH 3 is 0.075/0.47 = 0.16. About 16% of N-atoms are therefore converted to ammonia at these conditions. At the highest N-atom partial pressure of 0.585 Pa, the NH 3 partial pressure at the lowest H-atom partial pressure of 8.1 Pa is 0.14 Pa, so the efficacy of N transformation to NH 3 is 0.14/0.585 = 0.24 (or 24%). Even the lowest partial pressure of H-atoms therefore enables the transformation of about Figure 11. Ammonia partial pressure versus partial pressure of nitrogen atoms in the reaction chamber. The partial pressure of hydrogen atoms is the parameter. The partial pressures of both N 2 and H 2 were 50 Pa.
20% of N-atoms to ammonia. At the highest N-atom partial pressure of 0.585 Pa, the NH 3 partial pressure at the highest H-atom partial pressure of 16.2 Pa is 0.20 Pa, so the efficacy of N transformation to NH 3 is 0.2/0.585 = 0.34. The rather high H-atom density, therefore, assures the association of about a third of N-atoms to ammonia molecules. The production of ammonia in the reaction chamber made from the borosilicate glass is, therefore, very efficient. A comparison of the curves in figure 10 indicates saturation, i.e. as the H-atom partial pressure increases, the production of ammonia saturates.
Similar observations are deduced from the results shown in figure 11, which were acquired at larger pressures of both hydrogen and nitrogen molecules in the reaction chamber. Here, the efficacy is between about 16 and 30%. The results summarised in figures 10 and 11 clearly show that a significant fraction of nitrogen atoms is transformed into ammonia at the experiments performed in the reactor of figure 1. Other nitrogen atoms obviously follow heterogeneous surface recombination to N 2 molecules. The same applies for the Hatoms that have not been consumed upon the formation of NH 3 molecules.
Here, it is worth stressing that the formation of measurable amounts of ammonia was only detected when both nitrogen and hydrogen plasma sources were on. When either of these sources was off, no ammonia was produced. Results of such simple observation, therefore, clearly show that both N and H atoms should be present in the reaction chamber in order to produce ammonia molecules.

Production of ammonia in the metal reaction chambers
Even the remote parts of fusion reactors will probably never be made from (or coated with) glass so the experiments as above were also performed with different metal reaction chambers at the confluence of atom sources. The reaction chambers were made out of different metals and were mounted in the experimental system so that the holes fit both discharge tubes. There was also two holes in the metal reaction chambers to keep efficient pumping and for a CP. A schematic of the reaction chamber position and a photo of all the metal reaction chambers used in this study are shown in figure 12.
Several reaction chambers were prepared: high-purity nickel, high-purity cobalt, commercial purity tungsten, reaction chamber with a thin film of sputter-deposited tungsten nitride prepared by physical vapor deposition (PVD) technique, and alloy P92. Pure metals were supplied by Goodfellow Cambridge Ltd, P92 alloy was supplied by Max-Planck-Institut für Plasmaphysik from Garching and tungsten nitride was deposited at Jozef Stefan Institute. The surfaces of the reaction chambers were carefully cleaned with ethanol before mounting into the experimental system.
The measurements of the ammonia production were performed in each metal reaction chamber mounted as in figure 12. The nitrogen atom partial pressure in the reaction chamber was set to 0.49 Pa. The production of ammonia was measured with the RGA calibrated as explained in section 3.2. The production was measured at different hydrogen atom densities and the results are plotted versus the Hatom partial pressure in the reaction chamber. Figure 13 shows the evolution of the ammonia pressure in reaction chambers versus the partial pressure of H-atoms. As expected from the results summarised in figures 10 and 11, the production of ammonia slowly increases with the H-atom partial pressure. The relatively weak increase is explained by the fact that the H-atom partial pressure is over ten-times larger than the N-atom partial pressure, so there are ample H-atoms available for ammonia formation. The trend is practically the same for all materials, but the absolute values are different.
The largest production of ammonia is observed for highpurity nickel reaction chamber. Figure 13 shows that the partial pressure of ammonia is about 0.15 Pa even at the lowest H-atom partial pressure of about 8 Pa. Since the original Natom partial pressure at these experiments was 0.49, one can deduce that about one third of the N-atoms introduced into the reaction chamber made from pure nickel interacts with Hatoms to form ammonia molecules. The production is even more efficient at higher H-atom partial pressure, for example, about 46% at the largest H-atom partial pressure of 18.9 Pa. The rest of the available atoms are recombined to N 2 and H 2 molecules. The production of ammonia is somehow lower for other materials. The lowest was found for metal coated with a thin film of tungsten nitride. For this material, between about 10 and 20% of available N-atoms were converted to ammonia, depending on the partial pressure of H-atoms. This result is somehow lower than for the glass reaction chamber, as revealed by comparing figure 13 with figure 10.
The results shown in figure 13 are rather trustable. The experiments were performed several times. The variation of the ammonia production versus the type of material immersed in the reaction chamber is difficult to explain due to the lack of an appropriate model of N and H adsorption and surface reactions. As mentioned in the Introduction, it is generally accepted that the N and H atoms stick to the surfaces of various materials, but the initial sticking coefficient should depend on the composition of the surface layer. The situation is further complicated since there is a competition between the production of ammonia and heterogeneous surface recombination of atoms to parent molecules. The recombination coefficients are large for high-purity transition metals such as nickel and cobalt and low for borosilicate glass [27,28]. Namely, the N and H atoms chemisorb on the surface of transitional metals and form a layer of rather loosely bonded atoms which interact with the atoms arriving from the gas phase and associate to parent molecules. The recombination definitely represents a sink for both types of atoms so one may assume the formation of ammonia is negligible due to the rapid loss of atoms by surface recombination. The results summarised in figures 10, 11 and 13, however, clearly show that the formation of ammonia is a highly competitive reaction. Even highly catalytic materials such as pure nickel and cobalt enable the formation of ammonia. The adsorbed N-atoms, therefore, assure for significant production of ammonia.
The efficiency of the conversion of neutral hydrogen and neutral nitrogen atoms into ammonia on surfaces of various materials is presented in table 1.
Perhaps the most important result is the efficiency of ammonia production on the surface of stainless steel P92. This  [29]. The rest is iron. The ammonia production on this material was moderate compared to other probed materials. At the lowest H-atom partial pressure, about 15% of N-atoms in transformed to ammonia, and at the highest, it is just over 20%. This value is still significant so one may expect a rather large formation of ammonia in fusion reactors made from this alloy.
In our experiments, the co-adsorption of H and N atoms is the necessary condition for ammonia formation, the same as in the case of Haber-Bosch process [30], except that the latter runs at elevated temperatures and pressures. The influence of the material temperature on ammonia production was studied in the case of nickel. The nickel reaction chamber was heated by a resistance heater experiment and the temperature was kept at 200 • C during the measurements. The result is shown in figure 14. The difference between the heated nickel reaction chamber and the reaction chamber kept at room temperature is evident and is around 10 %, which is much more than the statistical error of these measurements. The increased temperature therefore increases ammonia production.
De Castro et al [4] also observed that conversion into ammonia increased with temperature. They observed that the conversion rate on stainless steel is higher than on tungsten. Ben Yaala et al [3,23] observed exactly the opposite results.
They measured W to be more active than stainless steel and observed ammonia yield to decrease with temperature. They measured a max ammonia yield at 50% N 2 /H 2 RF plasma with W and stainless steel and a max yield at around 25% N 2 /H 2 without a catalyst. Antunes et al [31] observed similar results.

Conclusions and perspective
The formation of ammonia in the afterglow chamber of both hydrogen and nitrogen plasma was studied. Careful calibration of both the RGA and the atom sources enables a rather reliable interpretation of the measured signals. The density of H-atoms in the reaction chamber was of the order of 10 21 m −3 , and the density of N-atoms was an order of magnitude lower due to the restrictions in N-atom formation in the microwavedischarge atom source used for these experiments. Within the limited range of experimental conditions, ammonia production was relatively efficient and dependent on the type of material the reaction chamber was made from. The efficiency of ammonia formation was expressed in terms of the fraction of N-atoms transformed to NH 3 molecules. The rest of N as well as H atoms, are associated to parent molecules. The fraction was between about 10% and 50%. The highest production of ammonia was observed for high-purity nickel, while for the stainless steel P92 it was moderate, between about 15% and 20%. No production of ammonia was detected from one type of atoms only. Namely, some experiments were performed with an H-atom source in molecular nitrogen and another with an N-atom source in molecular hydrogen but the RGA NH 3 signal was always below the detection limit. Same as in Haber-Bosch process, the formation of ammonia was explained by heterogeneous surface interaction between adsorbed N and H atoms.
The formation of ammonia in fusion reactors may represent an obstacle if nitrogen seeding is used for cooling the partially ionized plasma in divertors. The fluxes of N and H atoms onto the surfaces of fusion reactors not in direct contact with partially ionized plasma should be, therefore as low as possible in order to suppress the ammonia formation. Alternatively, a coating that favours the heterogeneous surface recombination of either H or N atoms to parent molecules should be employed. Currently, such coatings' composition and/or structure is unknown, and the research in this niche represents an important scientific challenge. Modelling of co-adsorption and surface reactions upon exposing solid materials to fluxes of N and H atoms likely to be in remote parts of fusion reactors is needed to show the way of suppressing ammonia formation.
the European Commission can be held responsible for them.