Chemically assisted physical sputtering of Tungsten: Identification via the + 6 6 transition of WD in TEXTOR and ASDEX Upgrade plasmas

Tungsten ( W ) is a preferred plasma-facing material for areas of high particle and power impact in present day and future fusion devices. The lifetime of W plasma-facing components (PFCs) under steady-state conditions is limited by erosion processes induced by energetic impinging particles such as impurities (carbon C , oxygen O , nitrogen N , etc.) and hydrogen isotopes above the threshold energy for Physical Sputtering ( PS ). We discovered for the first time, in addition to the bare physical sputtering process of W , a second W erosion mechanism at PFCs in TEXTOR (limiter surfaces) and ASDEX Upgrade (divertor target plates) during deuterium plasma bombardment. The tungsten deuteride molecule WD has been identified spectroscopically via the + transition in parallel to ordinary neutral W line emission of the WI transition (5 d 5 ( 6 S )6 s 7 S 3 →5 d 5 ( 6 S )6 p 7 P 4 ) at = 400.9nm used as measure to quantify the bare PS process or, more precisely, the gross erosion of W . We identified the underlying process for the molecular release as Chemically Assisted Physical Sputtering ( CAPS ) as observed also in the case of beryllium or lanthanum. Measurements in TEXTOR and ASDEX Upgrade showed a dependence of the WD band emission on the surface temperature - connected to the deuterium content in the near W surface - as well as on the flux and energy of impinging energetic particles. A quantification of the released WD is not yet possible due to lack of appropriate molecular data for the conversion of photons into particles, but the spatially resolved ratio of atomic ( WI ) and molecular photon flux ( WD ) indicates that the release takes place at the same time, but differently distributed along the target plates and limiters reflecting the variation in the described critical parameters (impinging ion flux, ion impact energy, and material temperature) along the surface. The plasma conditions in front of the interaction zone are in both cases fully ionising with an electron temperature T e >10 eV , thus, recombination into WD can be excluded.

transition in parallel to ordinary neutral W line emission of the WI transition (5d 5 ( 6 S)6s 7 S 3 → 5d 5 ( 6 S)6p 7 P 4 ) at = 400.9nm used as measure to quantify the bare PS process or, more precisely, the gross erosion of W. We identified the underlying process for the molecular release as Chemically Assisted Physical Sputtering (CAPS) as observed also in the case of beryllium or lanthanum. Measurements in TEXTOR and ASDEX Upgrade showed a dependence of the WD band emission on the surface temperature -connected to the deuterium content in the near W surface -as well as on the flux and energy of impinging energetic particles. A quantification of the released WD is not yet possible due to lack of appropriate molecular data for the conversion of photons into particles, but the spatially resolved ratio of atomic (WI) and molecular photon flux (WD) indicates that the release takes place at the same time, but differently distributed along the target plates and limiters reflecting the variation in the described critical parameters (impinging ion flux, ion impact energy, and material temperature) along the surface. The plasma conditions in front of the interaction zone are in both cases fully ionising with an electron temperature T e > 10eV, thus, recombination into WD can be excluded.

Introduction
With increasing emphasis, plasma-facing components (PFCs) made of tungsten (W) are introduced in different fusion devices (ASDEX Upgrade [1], TEXTOR [2], JET [3,4]) at the location of highest particle and heat flux owing to its good power load capabilities, high melting temperature, and low fuel retention due to implantation in the bulk material. ITER has therefore selected W as first choice plasma-facing material (PFM) for the actively cooled divertor which must operate in the so-called semi-detached regime during the fusion power operational phase in order to sustain the steady-state high power load densities of more than 10MWm 2 [5]. A critical issue for the plasma operation remains the source strength of W, thus the flux of W eroded at the divertor PFCs, which pollutes the plasma core and might cause under unfavourable screening and transport conditions accumulation. Experiments in ASDEX Upgrade and JET with ITER-like wall showed compatibility of W PFCs with plasma operation and concentrations suitable for reactor with concentrations around × c 3 10 W 5 in the case of good W source control, Edge Localised Modes (ELMs) flushing from the pedestal region as well as localised core heating to expel W from the core region [6]. In the present understanding, the W source strength is determined by Physical Sputtering (PS) processes ( Fig. 1a and b) caused by impinging intrinsic or extrinsic impurities (e.g. C, Be, O, B, N etc.) [7] and by high energetic fuel particle ions (H, D, T) hitting the target plates during ELMs [8,9]. A mixture of both PS processes determines in general the gross source of W [10], whereas the inter-ELM source can be mitigated by operation in the (semi-)detached regime as foreseen in ITER with energies of impinging ions below the PS threshold [11,12]. The intra-ELM source cannot be suppressed as long as energetic ions, fuel ions and impurities, are expelled from the pedestal and transported quasi-ballistic to the divertor target plate. Flushing of W from the pedestal region by ELM pacing can widen the operational window, but is not contributing in a prolonged lifetime of the W PFCs. Only minimisation of the W source with support of the socalled prompt re-deposition effect can help to extend the lifetime of W PFCs [13].
Different laboratory studies of W bombardment by hydrogen ions have identified the formation of a thin surface layer with high hydrogen content of up to a few percent in the subsurface [14], thus more than four orders of magnitude higher than the solute in W for comparable temperatures. The formation is a competition between impinging ion flux, outflux, and diffusion as described in [15,16] and depends in particular on the material temperature. The situation is comparable to beryllium (Be) surfaces where molecular dynamics calculations [17] revealed up to 50% D in the Be surface layer -so-called supersaturation. As consequence an additional mechanism for sputtering can set in: Chemically Assisted Physical Sputtering (CAPS). CAPS has been verified by the release of beryllium deuterides, or more precisely by spectroscopic identification and analysis of BeD molecules, in PISCES-B and JET [18,19]. The critical questions are if CAPS appears in the case of W ( Fig. 1c and d) and what is the parameter space of existence? It should be noted, that CAPS has not been observed in any fusion-related plasma device in the case of W so far, though the worldwide focus in Plasma-Surface Interaction (PSI) research is in the qualification of W for ITER and DEMO.
In this contribution, we present the first observation [20] and analysis of the WD molecule in a fusion-relevant plasma by Optical Emission Spectroscopy (OES). Molecular bands of the electronic WD transition + 6 6 were detected close to the W PFC surface during the exposition in a series of D plasmas in TEXTOR used originally in a tungsten vs. graphite PFC comparison study [7]. Detailed analysis of the spectral range between 673 and 678nm, recorded with a high resolution overview spectrometer, revealed the ro-vibrational bands of the WD molecular transition. The band emission is only present when W erosion takes place, but absent during WF 6 injections -used for calibrations to mimic atomic W sources in the plasma -which excludes the formation of WD in the plasma by recombination. This is inline with the ionising conditions in TEXTOR limiter plasmas. The identification as well as analysis is presented in the Section 2 of the paper. Additionally, the WD transition has been observed in deuterium high confinement mode (ELMy H-mode) discharges in ASDEX Upgrade at the outer target plate made of bulk W. Thus, the ASDEX Upgrade experiment gives access to intra-ELM W sputtering events by spectroscopic observation of the WD molecular bands as well as neutral W with the WI transition (5d 5 ( 6 S)6s 7 S 3 → 5d 5 ( 6 S)6p 7 P 4 ) at = 400.9nm with high temporal and good spatial resolution along the outer target plate in the near Scrape-Off Layer (SOL). CAPS takes place in both the intra-and inter-ELM phase with impinging ion fluxes consisting of mainly deuterons and an intrinsic impurity mix of C, B and O ions in the promille range. Details about the experiment including the distribution of the flux ratio of WD and WI is shown in Section 3.
The summary and conclusion Section 4 provides information on the processes involved in the release of atomic and molecular W compounds. The required conditions for the molecular release via CAPS are indicated and first conclusions drawn. The potential role of CAPS for W in the total W source as well as the appearance in other experiments like linear devices PSI-2 or MAGNUM-PSI, and larger tokamaks like JET is elaborated.

Identification of WD molecules in TEXTOR deuterium plasmas
The first spectroscopic identification of the tungsten deuteride molecule was reported by [21] in a deuterium arc discharge close to a tungsten-made injection nozzle with the aid of a Jarrel-Ash monochromator with a resolution of 0.05nm/pix and a photomultiplier as detector. The compressed molecular spectrum between = 673nm and = 678nm was attributed to the electronic transition of WD as indicated in Fig. 2a), whereas it was not clear at the time of the study if the lower state is indeed the electronic ground state. Nevertheless, the rotational P, Q, and R branches were identfied as shown in Fig. 2b) as well as the corresponding spectroscopic molecular constants B, D, r e , and ν provided. Moreover, [21] also observed the corresponding transition of the WH molecule at the predicted spectral location confirming that the identification of the tungsten hydride and its isotopomer was  is marked according to the identification [22]. b) Reference WD spectrum observed in a low pressure arc discharge close to a tungsten nozzle [21]. C) The corresponding tungsten deuteride WD molecular band identified for the first time in a high temperature fusion plasma in front of a W limiter in TEXTOR [20].
correct. The assignment was later confirmed by [22] in complex molecular calculations providing 15 bound electronic states, spectroscopic constants, and potential energy curves. Fig. 2c) shows the corresponding WD molecular band spectrum observed in the TEXTOR tokamak in front of a W limiter. The spectrum was recorded with the aid of a cross-dispersion spectrometer consisting of a double prism and an Echelle grating covering more than 75 orders with an almost constant resolving power of more than = R 20000 over the full covered wavelength span (Δλ span of about 364 715nm).
The system is comparable to [25], but equipped with an one megapixel EMCCD detector by ANDOR which provides higher sensitivity. The price for the simultaneously obtained high resolution spectrum with wide wavelength span is the relative low sensitivity and low temporal resolution of about 200ms (frame transfer time of 1s) used in the presented measurement. The main spectroscopic features of the WD spectrum can clearly be identified: the line-like shape of the Q-branch at = 674.5nm, the returning R-branch at the lower wavelength end at = nm 673 and the long P-branch tail in the higher range up to = 678nm. Complementary studies with WF 6 injection in a comparable plasma through a stainless steel nozzle, providing a proxy for a neutral W gas source and used to determine photon efficiencies [24], have not shown any molecular lines in the corresponding wavelength region. This confirms that the WD is not produced within the plasma by recombination, but due to surface reactions at the W limiter surface and therefore related to sputtering processes.
The TEXTOR tokamak (closed 12/2013) was a medium-size machine optimised for PSI studies with a major radius = R 1.75m and a minor radius = a 0.46m defined by a toroidal belt limiter made of graphite. Dedicated studies on W PFCs were done with the aid of additional limiters installed at variable radial positions in one of two lock systems with optimised diagnostic coverage including image intensified 2D CCD cameras with narrowband interference filters and a number of spatially integrating overview spectrometers as described in [7]. However, only the Spectrelle provides simultaneously observation of the standard WI line at = 400.9nm and the WD band emission range between = 673nm and = 678nm. We revisited previously performed W qualification experiments [7] in deuterium at a plasma current of in and a radiated fraction f rad between 0.33 and 0.40. A twin limiter made of graphite and bulk tungsten (W from Plansee with 99.97% purity), shown in (Fig. 3a), was positioned 5mm behind the last-closed flux surface (Fig. 3b) in the near-SOL and exposed to a number of plasma discharges with four dedicated density phases owing to variation in D fuelling. The four phases can be described by the following constant plasma conditions at the radial location of the twin limiter: phase I [ = × n 0.60 10 m ,  Fig. 4a) from the Balmer photon fluxes (D δ and D γ ) from recycling at the limiter. The main impurities were C (c C ≃ 5.2%) and O (c O ≃ 0.6%) which are also responsible for the W sputtering as the impinging ion energy of + D is below its physical sputtering threshold. Fig. 3b) shows the emission pattern of sputtered W in the light WI and the deuterium recycling flux in the light of D α . The sputtering yield of W based on WI drops in the four phases with decreasing impact energy from 5.2% in phase I, via 2.0% (phase II), and 1.0% (phase III), to finally 0.5% (phase IV) which is consistent with binary-collision approximation calculations with the given impinging impurity mix and ion energy. The limiter temperature was predominantly determined by the active wall heating in TEXTOR to T limiter ≃ 420K with minor temperature increase during the plasma discharge by about 80K. Fig. 4 b) shows the variation of the line-like Q-branch of the WD transition and the standard WI transition in the four density phases described before. The WI emission drops monotonically with T e whereas T e is a measure for the impact energy via and T e ≃ T i with the ion temperature T i . The WD emission shows a different behaviour with a peak at the third phase before a drop at lowest T e occurs. However, in addition to the impact energy, one needs also to consider the ion flux impinging to the limiter which increases almost a factor 4 with the reduction of T e by a nearly a factor 3 starting from the highest T e -value in phase I. Ferro and co-workers [26], based on experiments of [14], calculated by a novel global model based on thermodynamics, kinetics, and density functional theory data a continuous increase in the hydrogen content in the near-surface layer of W with increasing deuterium flux. The modelled fuel content in W rises even steeper, by orders of magnitude, below a critical temperature T crit due to enhanced defect creation and formation of vacancies in W with higher hydrogen trapping in it -the so-called supersaturated layer is formed. One can assume that there is a higher probability to release a WD molecule at higher impinging ion flux as there is more D embedded in this near W surface layer at constant temperature. It should be noted, that the predicted critical temperature for the supersaturation at the ion flux conditions in TEXTOR with a few + D 10 m s 23 2 1 is around = T 550K crit and below the limiter temperature in the experiment. The resulting WD variation in the case of the W limiter at constant material temperature is therefore a combination of two processes: a) reduction of WD with reduction of T e , or more precisely impact energy E in , indicating the physical sputtering character of CAPS, and b) increase of WD with increase of the deuterium flux, or more precisely the near-surface fuel content in W, indicating the chemical sputtering character of CAPS. Further detailed studies at different material or limiter temperatures would have been required to obtain the full matrix or existence diagram of CAPS, but these are due to closure of TEXTOR no longer possible.

Identification of WD molecules in ASDEX Upgrade deuterium plasmas
Complementary to the first studies in TEXTOR with solely CAPS of W caused by impurities, investigations in the W divertor of ASDEX Upgrade in deuterium ELMy H-mode discharges have been carried. These H-mode discharges ( ) were conducted with the outer-strike line located on a bulk W PFC target plate installed in divertor manipulator (DIM II [23]) in the outer leg (Fig. 5b). The corresponding Fig. 5a)  The plasma and surface conditions in ASDEX Upgrade would fulfill the criteria mentioned before to observe CAPS of W at least in the inter-ELM phase as enough deuterium should be retained in the near surface layer. Fig. 6 shows a comparison of emission spectra of a) WD + 6 6 and b) the standard WI transition (5d 5 ( 6 S)6s 7 S 3 → 5d 5 ( 6 S)6p 7 P 4 ) at = 400.9nm recorded in ASDEX Upgrade close to the strike-line and in TEXTOR. The ASDEX Upgrade spectra were recorded in two consecutive discharges with the aid of a high resolution, high throughput spectrometer with narrow wavelength coverage. The ro-vibrational WD spectrum in ASDEX Upgrade is very much comparable with the one from TEXTOR with the most characteristic line-like feature of the Qbranch. The ro-vibrational population is comparable in both plasmas though the detailed plasmas parameters are different. The use of the line-like Q-branch as characteristic quantity is therefore verified and can be used to compare it with the WI emission at = nm 400. 9 as in TEXTOR case. The P-branch seems a bit more pronounced in ASDEX Upgrade and reflects all details shown in [21]. We can conclude that CAPS of W takes place in the outer divertor of ASDEX Upgrade -at least under the experimental conditions described before. The poloidal distribution of the WD line-like Q branch + 6 6 radiance, the WI radiance of the 5d 5 ( 6 S)6s 7 S 3 → 5d 5 ( 6 S)6p 7 P 4 transition, and the ratio of both along the W target plate given in the relative target coordinate s with respect to the outer strike-line position is shown in Fig. 7. The experimental data is recorded in two consecutive self-similar discharges (#34900 and #34901) with appropriate change of the spectrometer wavelength settings. The analysis of OES data is done twofold: a) application of ELM filtering providing the averaged radiance in the inter-ELM phase and exclusion of the high ion impact energy during ELM events and b) total time averaging over the discharge between 3.7s and 7.0s, thus, the mean value of the intra-and inter-ELM phase. The overall profiles of WD and W radiances for both averaging processes are similar following the impinging ion flux distribution with the maximum emission at the strike-line ( = s 0) where also the surface temperature is the highest. However, the ratio of the two radiances reveals a relative increase of WD away from the strike-line where the surface temperature is significantly lower permitting a higher deuterium retention in the thin surface layer. In the inter-ELM averaged case are the impact energies still sufficient at this location to erode W due to ordinary PS by intrinsic impurities and the impinging deuteron fluxes are still sufficiently high to fill up the vacancies. The spectroscopy suggests that in the near-SOL a higher fraction is released via CAPS than at the strikeline assuming that the D/XB-values are comparable in the range of The intra-ELM W phase dominates over the inter-ELM W sputtering phase as both, the line-like feature of WD + 6 6 and the WI radiance at 5d 5 ( 6 S)6s 7 S 3 → 5d 5 ( 6 S)6p 7 P 4 in the mean averaged phase are significantly above the corresponding signals in the inter-ELM phase. This is consistent with previously reported ELM-induced W sputtering observations in ASDEX Upgrade done solely with the aid of WI emission. Thus, we can first conclude that CAPS is also caused in the intra-ELM phase by impurities as well as energetic impinging deuterons. However, the increase in WI is much stronger than in WD as indicated by the ratio of the two radiances which can be transferred into the statement that in the presence of high energetic impinging ions the CAPS seems less prominent in comparison to the intra-ELM phase. The probability to release directly a W atom is in this case higher than the WD molecule by either an impurity ion or a deuteron.

Summary and conclusion
We have identified the WD + 6 6 transition in TEXTOR limiter and ASDEX Upgrade divertor plasmas in front of PFCs made of bulk W. The ro-vibrational spectra in both plasmas is comparable from the point of view of characteristic features as the line-like Q-branch, the returning R-branch, and the long tail of the P-branch which suggests a  ) at = 400.9nm. The TEXTOR spectra are recorded simultaneously in one discharge; the ASDEX Upgrade spectra are recorded in two comparable discharges. band emission, and the ratio of both as function of the relative target coordinate s. The averaged inter-ELM distribution as well as the total averaged distribution including all intra-and inter-ELM phases between 3.7s and 7.0s are depicted. comparable ro-vibrational population. Modelling of the ro-vibrational population, which is required to determine the total photon flux of the electronic transition, is pending and requires slightly higher spectral resolution to resolve properly the rotational lines to simulate the spectrum. Moreover, conversion factors to particle fluxes, the so-called D/XB-values, which provides the ratio of ionisation and dissociation rate coefficients (the D for decay) of the WD molecule over the eXcitation rate coefficient X for the observed + 6 6 multiplied with the Branching ratio B, are currently unknown. Therefore, we can currently only report on relative changes of the intensities of WD as well as relative changes to the best characterised WI transition (5d 5 ( 6 S) 6s 7 S 3 → 5d 5 ( 6 S)6p 7 P 4 ) at = 400.9nm. We assume that the ro-vibrational population is unchanged in the plasma parameter range covered which is with T e > 10eV a fully ionising plasma and recombination can in the covered range in TEXTOR and ASDEX Upgrade be excluded.
The destruction of the WD molecule can undergo two different paths: a) ionisation of WD and formation of the + WD molecule by electron impact ionisation and subsequent dissociation into W and + D or vice versa into + W and D by electron impact dissociation, and b) dissociation of WD by electron impact dissociation into W and D. Comparable destruction paths have been studied in detail in the case of the CD radical with the most probable path via (b) in ionising conditions at lower T e of below 10eV -like in divertor conditions -and via path (a) at higher temperatures of T e > 20eV -like in limiter conditions. However, there is nor a report on the spectroscopic identification of + WD neither any atomic and molecular data exist to verify the paths at this stage, but it is important to state that in both paths a neutral W atom during the dissociation can be produced. Thus, it cannot be excluded that WD contributes via the potential dissociated product W to the line emission of WI at e.g. = 400.9nm. This must be taken as caveat when discussing the distribution of WD and WI as well as the total W sputtering source.
The sputtering of W in molecular form as WD requires a large fraction of D embedded in the top surface of the W material matrix. The fuel content in this interaction layer depends on the impinging ion flux and the bulk temperature which determines the formation of vacancies by ion damage. This process competes with desorption and outflux of D as well as with diffusion of D deeper into the matrix. Recent experimental observations [14] and novel modelling [26] showed the formation of a narrow supersaturation layer in W with high D content in the percent range below a critical surface temperature T crit which would provide most favourable conditions for chemically assisted physical sputtering CAPS of W. The fuel content in the narrow surface layer rises in general with increase of the impinging deuteron flux, roughly one order in retention per one order of flux, and decrease with the material temperature. The ion flux conditions in TEXTOR and ASDEX Upgrade are in the range of × 0.1 2.0 10 Ds m 23 1 2 which corresponds to T crit ≃ 550K and a supersaturated layer below this temperature. The third parameter required for CAPS is the energy of the impinging projectiles which must overcome a characteristic binding energy. The variation in the observed intensity of WD in TEXTOR and ASDEX Upgrade can be attributed to changes in these three parameters which is in contrast to the temperature independent ordinary physical PS. In TEXTOR impurity ions are solely responsible for the release whereas in ASDEX Upgrade also fast deuterons during ELM events contribute on top of the impurity ions to CAPS. Therefore, we can conclude, that the release mechanism is via cascades followed by a release and not via a capture of a loosely bound W at the surface by an impinging deuteron. However, it is unknown what the minimum impact energy is to cause the process and dedicated experiments seem required to determine it. Most relevant is the determination of the threshold in the relevant impinging ion flux range in the order of about 10 10 Ds m 22 24 1 2 present in divertor conditions in fusion devices which limits the number of relevant test devices to high flux experiments like the linear plasma devices PSI-2 and MAGNUM-PSI. Certainly, the inertially cooled bulk W divertor in JET covers the existence range for CAPS and the formation of WD can be expected though a control of the surface temperature will be challenging. As mentioned before, a contribution of the dissociated molecule WD to the neutral W atom is most likely, thus, the measure of WI potentially includes already a fraction of W sputtered by CAPS on top of PS. Furthermore, one needs to consider that the penetration depth of the WD molecule is even shorter than the one of neutral W atom though of course the penetration depth of the potentially resulting W atom from the WD dissociation will be slightly larger due to the two step process than of the bare physically sputtered W atom. Thus, if ionisation takes place via destruction channel (a) or (b), a large fraction will return to the surface within one Larmor radius at the high magnetic fields above B t > 2T present in TEXTOR and ASDEX Upgrade. Therefore, the contribution of WD to the net W erosion source above the threshold energies for bare PS is even smaller than to the gross erosion source determined by OES. Nevertheless, further experiments with variation of all three parameters: impinging ion flux, impact energy, and surface temperature are required to quantify the exact contributions. In particular, the threshold energy for the CAPS process might below the one for bare PS as the binding energy might be reduced. This would permit a potential W erosion source at low impact energies where W erosion is usually assumed to be switched off. Potentially, easier re-erosion and transport of W towards remote areas might follow as observed e.g. in JET in the louvre area of the divertor. Dedicated experiments in JET and ASDEX Upgrade are currently foreseen to investigate this potential way of enhanced W material transport via CAPS. Better understanding in this topic is required in order to improve the W erosion and transport modelling in the divertor in view of ITER and DEMO W PFC life time predictions.