Plasma-wall interaction studies in W7-X: main results from the recent divertor operations

Wendelstein 7-X (W7-X) is an optimized stellarator with a 3-dimensional five-fold modular geometry. The plasma-wall-interaction (PWI) investigations in the complex 3D geometry of W7-X were carried out by in situ spectroscopic observations, exhaust gas analysis and post-mortem measurements on a large number of plasma-facing components extracted after campaigns. The investigations showed that the divertor strike line areas on the divertor targets appeared to be the major source of carbon impurities. After multistep erosion and deposition events, carbon was found to be deposited largely at the first wall components, with thick deposits of >1 μm on some baffle tiles, moderate deposits on toroidal closure tiles and thin deposits at the heat shield tiles and the outer wall panels. Some amount of the eroded carbon was pumped out via the vacuum pumps as volatile hydrocarbons and carbon oxides (CO, CO2) formed due to the chemical processes. Boron was introduced by three boronizations and one boron powder injection experiment. Thin boron-dominated layers were found on the inner heat shield and the outer wall panels, some boron was also found at the test divertor unit and in redeposited layers together with carbon. Local erosion/deposition and global migration processes were studied using field-line transport simulations, analytical estimations, 3D-WallDYN and ERO2.0 modeling in standard magnetic field configuration.


Introduction
Wendelstein 7-X (W7-X) is an optimized three-dimensional helically shaped stellarator with five-fold symmetry designed for steady-state plasma operation of up to 30 min. Since the beginning of W7-X operation with island divertor, there have been two Operation Phases (OP) completed, namely: OP1.2a and OP1.2b. Due to the operational constraints, the total integral energy input during these phases was limited to 80 MJ (OP1.2a) and 200 MJ (OP1.2b) per discharge, respectively. For OP1.2a Plasma-Facing Components (PFCs) made of fine grain graphite were installed: ten inertially cooled discrete island Test Divertor Units (TDUs), baffles, toroidal divertor closures and wall protection heat shield tiles. The outer wall panels, divertor pumping gap panels, and poloidal closures were made of stainless steel (see figure 1). In addition, for OP1.2b graphite scraper elements were installed near two TDUs in order to mitigate convective plasma heat loads on the pumping gap panels in certain Table 1. Surface areas of the W7-X graphite and stainless steel plasma-facing components obtained based on CATIA design data [6].   configurations caused by the evolution of the internal bootstrap current [1][2][3]. The whole inboard side is covered with graphite heat shield and baffle tiles, whereas stainless steel (EN1.4429 → AISI316LN) wall panels cover most of the outboard side of the plasma vessel except at certain locations, e.g., the triangular cross section, where the heat shield tiles are installed as well [4,5]. The locations of these components are shown in figure 1 and their numbers together with the surface area are given in table 1. Carbon (C) migration has been extensively studied in fusion devices with 2-dimensional geometry such as tokamaks equipped with graphite PFCs [7][8][9][10]. In stellarators with their 3-dimensional geometry such studies are much scarcer and still ongoing. Due to the profound differences in plasma configuration and the geometry of divertor and other PFCs, the locations of net erosion and net deposition areas could be different. Compared to the Large Helical Device (LHD) [11][12][13][14], W7-X provides unique possibilities to study carbon migration in its specific 3-dimensional geometry with modular plasma configurations and island divertors.
Boronization by using the injection of diborane during glow discharge operation is regularly used in fusion machines to suppress oxygen (O) and carbon impurities, resulting in improved plasma operation [15]. Especially directly after fresh boronizations the plasma-facing surfaces consist mostly of boron rather than carbon and therefore, besides carbon it becomes important to study the plasma-wall interaction and migration of B in these machines. It is also important to investigate the effect of inhomogeneity in the boron distribution on the PFCs on the efficient reduction of oxygen and carbon impurities in the plasma.
Simulations were carried out using Field Line Transport (FLT) [16], WallDYN-3D [17] and ERO2.0 [18] codes in order to interpret the results of the post-mortem surface analyses on selected material probes taken from different positions in W7-X.
The details of the experimental techniques used for the analysis i.e. Elastic Backscattering Spectrometry (EBS), Nuclear Reaction Analysis (NRA), Focused Ion Beam (FIB) cutting, Scanning Electron Microscopy (SEM) and Energy-Dispersive X-ray spectroscopy (EDX) are described in [19], pico-second Laser Induced Breakdown Spectroscopy (ps-LIBS) in [20] and optical reflection measurements in [21].

Carbon balance
Studying carbon migration in a fusion device with carbon PFCs (graphite) is challenging to quantify by the postmortem measurements, in particular, the localized erosions or thin depositions with pure carbon layers. For the PWI studies in W7-X, the following three techniques have been applied: i. Coating the base graphite material with marker layers: 18 targets of the TDU were coated with ca. 0.2 μm of Mo and 5-10 μm C marker layers (see figure 2). In order to have a toroidal distribution of erosion/ deposition, ten of these were chosen at the same location in each TDU. For the poloidal distribution, eight additional targets were selected on the upper and lower TDU in a single W7-X module (module 5). After OP1.2a, all the 18 targets were exchanged with new similar marker targets [19].
ii. Tungsten coating: 3 tiles of a TDU target, 2 scraper tiles and 21 tiles of the heat shield were coated with thin W layer [22].
iii. Thicker redeposited carbon layers on carbon tiles can be identified using depth-profiling methods by the presence of co-deposited hydrogen isotopes or other co-deposited impurities, such as oxygen, boron or traces of metals (especially Fe, Ni, Cr). Thin redeposited carbon layers on carbon substrate are generally difficult to quantify and require the use of different substrates.
Using these techniques, the carbon balance was studied by measuring carbon erosion/deposition at the TDU and deposition at various places after the completion of OP1.2b. The exhaust of volatile carbon-containing molecules via the vacuum pumps was measured by mass spectrometers.
Besides the study of global migration of the carbon eroded from the C-PFCs, to study the local erosion and deposition, a 13 C isotope was injected in the form of 13 CH 4 on the last day of OP1.2b plasma campaign. A total of 4.2×10 22 13 C atoms (ca. 0.9 g) were injected (via gas injection nozzles built in the TDU) during 30 consecutive plasma discharges in standard magnetic configuration with identical plasma conditions. Simulations have been performed using 3D codes WallDYN-3D [17] and ERO2.0 [23] to model the erosion and deposition patterns showing good agreement with the initial measurements. Detailed measurements on a number of TDU targets removed after OP1.2b, are under progress, therefore in the present C-balance investigations, the 13 C experiments are not included.

Sources of carbon impurities
Based on the post-mortem measurements two different sources for eroded carbon were found:

Erosion on TDU targets
All the 36 targets with the marker layers were analyzed by EBS, some targets were also analyzed by ps-LIBS and SEM/FIB measurements. High erosion at the strike lines was found during OP1.2a such that not only the marker C layer but the underlying Mo interlayer was also eroded. This was due to high concentrations of C and O impurities evident from the high Z eff 4.5, resulting in high erosion yields. The total estimated erosion of carbon (for all the TDU) over the whole OP1.2a campaign was 48±14 g yielding a mean erosion rate of 13.9 mg s −1 [19].
During OP1.2b, three boronizations reduced the C and O impurity concentrations by a factor of 10-100 and resulted in Z eff ∼1.5 [24,25]. The erosion profile along a typical target shows comparatively lower erosion than in OP 1.2a (see figure 3). Despite of increased plasma duration from 3775 s (OP1.2a) to 9054 s (OP1.2b), the overall erosion of carbon during OP1.2b was reduced to 20±6 g and the erosion rate to 2.3 mg s −1 .

Glow discharge cleaning
Besides the carbon erosion during the plasma operations, carbon erosion occurred also during the glow discharge cleaning (GDC) operations. For OP1.2, the durations of GDC with He and H gases were 19.64 h and 27.37 h respectively [26]. With the discharge currents of 1 A and 1.5 A, the rate of singly ionized particles for ten GDC electrodes were 6.24×10 19 and 9.36×10 19 ions s −1 for the He and H ions respectively. Considering the sputtering yields of γ He→C =0.08 (for GDC voltage 210 eV) and γ H→C =0.009 (for GDC voltage 305 eV) [27] and ignoring the chemical erosion due to low wall surface temperatures of ca. 30°C, and considering, the ration of carbon surface area to the total area,  . Erosion profile along a TDU target TM2h6 (HM5l) for OP1.2b, the measurements of the marker layer thickness before and after the plasma operations are shown for carbon (upper plot) and Mo (lower plot). The location of plasma strike line is shown by the arrow. ANN represents Artificial Neural Network fittings, details of these are mentioned in [19]. The location of pumping gap (PG) is on the right side and the W7-X outboard (OB) on the left side. deposition of 2.1 g. The glow discharge during boronization is not included here, moreover the GDC was suspended after the first boronization due to improved wall conditioning and to avoid sputtering of the thin deposited boron layer. Considering a homogeneous deposition of the eroded carbon on all the PFC with a total surface area of 192.5 m 2 (see table 1) and a density of deposited C-layer as 6.0×10 28 C-atoms m −3 , the thickness of the redeposited C-layer works out to be about 22.6 nm (= 2.61×10 23 /192.5/6.0×10 28 ), in reality, re-erosion and deposition during the plasma operation will have effects on the observed distribution. However, erosion/ deposition during GDCs can reasonably explain the deposition on the plasma vessel wall in the form of stripes as shown in figure 4, where the possibilities of re-erosion by tokamak plasma ions and neutrals was almost negligible. The stripes were formed due to the direct deposition through the small gaps between the carbon heat shield tiles onto the plasma vessel behind. Assuming a uniform distribution behind all the first wall tiles, the estimated area of the deposition stripes is 3.23 m 2 , the amount of carbon deposition here assuming homogeneous redeposition during GDC is ca. 0.1 g.
The deposition stripes were also found on a TAG number plate mounted on the plasma vessel wall (see figure 4(a)), which was removed and investigated by EBS for the deposited material. Besides minor B and O  deposition, the main deposition was from C atoms with an average surface concentration of ca. 4.33×10 21 atoms m −2 [28], equivalent to ca. 0.3 g total carbon deposition in the stripes behind the tiles. Within the uncertainties of measurements and the estimations of the areas, this number is of the same order of magnitude with the 0.1 g based on the simplified estimation by considering only the carbon erosion during GDC operations as described above.

Deposition of carbon 2.2.1. First wall graphite tiles
Out of 7800 graphite tiles in W7-X, about hundred tiles distributed over the torus were removed after OP1.2 for the post-mortem analysis using EBS, ps-LIBS and SEM/FIB/EDX measurements. In particular, thick deposited layers>1 μm were observed on some baffle tiles installed close to the TDU. As shown in figures 5(a) and (b), some of these deposited layers were released in the form of flakes, others are just attached loosely. On the other hand, the depositions on the heat shield and toroidal closure tiles were less pronounced with thicknesses of about 100 nm (see figures 5(c) and (d)). The positions of baffle, heat shield and the toroidal closures are shown in figure 1. The toroidal closure tiles were installed at one end of the TDU closing the gap along the toroidal direction.
Raman spectroscopy is capable of investigating the chemical bonds, defects and structural changes on the surfaces modified by interaction with the incident ions [29]. The measurements, i.e. G band wavenumber (σ G ) as a function of G band width (γ G ) caused by mainly sp 2 hybridized C, on the relevant parts of six tiles (tile 7 is from the baffle and the rest from the heat shield) from OP1.2b to test homogeneity of the tiles, are shown in figure 6. Comparison with intensively investigated Tore Supra and pristine CFC samples [30], showed deposition on all these tiles. On tile 2a small part of area is found close to the erosion region defined by the Tore Supra samples, which would be followed closely after next campaigns.
Assuming a uniform deposition on a particular component such as baffle, heat shield and the toroidal closures, the carbon amount deposited on these tiles are estimated using the following relation: tiles (considering the density of deposited C-layers ρ=1.
Here, ρ is the density of deposited C-layer 1.4 g cm −3 , t −1 is the average thickness of the deposited layer and A the surface area.
• Baffle: Deposits on baffle tiles are very inhomogeneously distributed and layer thicknesses range from very thin to more than 10 μm. For an estimate of the amount of redeposited carbon we assume an average thickness of 0.5 μm and the surface area of 33 m 2 (see table 1) → m baffle = 23 g • Heat shield: Deposits on the heat shield were more homogeneous but very thin, which made a precise quantification difficult. We assume an average thickness of 50 nm and the surface area of 47 m 2 → m heat shield = 3 g • Toroidal closure: Deposition on the toroidal closure was inhomogeneous, by assuming an average thickness of 50 nm and the surface area of 3 m 2 → m toroidal closure = 0.2 g Figure 6. Raman spectroscopy measurements on some of the first wall tiles and their comparison with the Tore Supra samples.
The accuracy of these numbers is only within a factor of two. The baffle tiles are located quite close to the plasma. During some configurations some of the tiles experienced higher particle loads resulting into thicker and inhomogeneous depositions. This is discussed also in the preceding section 4.1. The heat shield is located rather away from the plasma and the homogeneous deposition resulted possibly due to charge exchange processes. The depositions on the toroidal closure tiles occurred on the parts closer to plasma experiencing higher particle influx. In general depositions were observed on all these tiles analyzed so far indicating absence of erosion zones on the first wall components in OP1.2a and OP1.2b.

Carbon deposition on stainless steel panels
The thickness of deposited layers was measured in situ by an optical reflection (of RGB colors) measurement on the stainless steel panels of the outer wall, poloidal closures and the pumping gap panels after OP1.2a and OP1.2b [21]. Average thicknesses of 10 nm and 15 nm were measured for OP1.2a and 1.2b, respectively. Since these panels cannot be removed for analysis, no specific measurements could be performed for the material concentrations, however, it is expected to be C rich layer for OP1.2a and B/C rich co-deposited layer for OP1.2b. With the surface area of stainless steel PFCs, 77.1 m 2 (see table 1) and considering the density of the deposition layer, 1.4 g cm −3 , the amount of C works out to be 1.1 g for OP1.2 and B/C co-deposit, 1.6 g for OP1.2b. Assuming equal amounts of B and C deposition, the overall amounts (OP1.2a+OP1.2b) of C would be 1.9 g and B 0.8 g.

Dust and arc traces
In the fusion machines with carbon (graphite) plasma-facing components, deposited carbon layers are typically not diamond-like hard layers but loosely bound a-C:H layers with varying hydrogen content due to chemical bonding of C with the fuel atoms. Over periods of operation, once these layers get thicker, they may get released in the form of flakes or dust particles and may lead to problems in the plasma operation as already observed in Tore Supra [31]. Such dust formations were also observed in other machines i.e. TEXTOR [32] and ASDEX-Upgrade [33]. In W7-X, during the inspection after OP1.2b, 44 samples of dust and loosely bound particles were collected from the TDU, poloidal closures, and the pumping gap at different positions in the torus. However, the amount of dust was found to be very small [28] and does not contribute to the carbon balance significantly. Similarly, the eroded material due to 212 arc traces found on various metal PFCs was also of negligible amount with traces being of micro sizes [28,34].

Exhaust of volatile carbon oxides and hydrocarbons
Volatile hydrocarbons and oxides (mainly CO and smaller amounts of CO 2 ) formed by the chemical processes occurring during the plasma operations were pumped out via twenty pumping ports, located at the low iota and high iota sides of each TDU. These gases were analyzed using a diagnostic residual gas analyzer (DRGA) at a port in half-module 40. The details of the DRGA measurement set-up (see figure 7) in W7-X are presented in [35]. Two additional mass spectrometers were operated in half-modules 21and 40 but since these were not calibrated, for the present analysis the DRGA data have been used. The amounts of pumped CO, CO 2 , CH 2 and CH 4 are shown for a typical plasma shot 20181004.29 in figure 8. As seen in the figure 8, even though the plasma shot duration was ca. 12 s, the pumped gases could be measured even until 120 s due to the time constant of the vacuum system. The gases measured by the DRGA mass spectrometer are presented in table 2. With the subtraction of background signal, most of the N 2 signal was eliminated from the CO (mass 28) signal. It has been considered most of O 2 was pumped out in the form of CO and CO 2 and no free O radicals were available therefore CH 4 denotes mass 16. Assuming the similar amounts in all the ten ports (at the low iota side shown in figure 7) over all the 1256 plasma pulses of OP1.2b, the estimated amounts of gases are shown in table 2. Only the ten ports on low iota side were taken into estimations since more than 80% of the time, the plasma strike lines were on the low iota side of TDU. The chemical erosion on carbon PFCs has been well studied [36][37][38][39] and CH 3 and CH 4 have been found to be potential species. The other gas measured in mass spectrometer could also result due to cracking occurring within the spectrometer, nevertheless the amount of carbon exhaust would still be estimated as the overall amount of pumped out hydrocarbon and oxides of carbon which is about 1 g for OP1.2b.  The amounts of eroded, deposited and pumped out C for OP1.2 are summarized in table 3. It was observed on the TDU marker target measurements that there exists some variation of the erosion pattern along the torus [19] which is due to some misalignment of TDU elements. While these effects are taken into account for the carbon source they could contribute also on the deposition side. However, currently there are not enough data available for any conclusions about the toroidal distribution of deposition. The amounts of carbon exhausted via vacuum pumps during OP1.2a with large C/O impurity concentrations are yet to be estimated. Some amount of carbon deposited in the pumping port and other port walls and recessed areas could not be accounted here.

Source of boron impurities
For OP1.2b, boron was injected in W7-X during following two operations:

Boronization
Three boronizations were carried out in W7-X for OP1.2b via glow discharge, using a mixture of He-B 2 H 6 (diborane) in a ratio of 90%-10%. A total of 147 l mixture was introduced with a cumulative duration of 14 h. The spectroscopic observations have shown that these have helped to reduce the C and O impurity concentrations by a factor of 10 and 100 respectively [24,25], resulting in improved plasma parameters. Considering a density of B 2 H 6 1.16 kg m −3 , the estimated weight of B 2 H 6 was 22.85 g (= 1.16 kg m −3 * (197 l/ 10)/1000 l m −3 * 1000 g kg −1 ). With a mass fraction of B/B 2 H 6 =0.78, the amount of B introduced was 17.86 g (= 0.78 * 22.85 g). It is assumed that most of the diborane was decomposed in the plasma vessel by the glow discharge. Very small amount of pumped out diborane was decomposed inside a thermal cracking unit. The amount of Boron retained in the thermal cracking unit was not measured.

Boron injection experiments
Another technique to carry out a partial boronization during plasma operation is via a boron injector in the midplane manipulator, in which B 4 C powder is injected during the plasma pulse. This technique was introduced in W7-X for OP1.2b and ca. 2 g B 4 C was injected [40]. With a mass fraction of B/B 4 C=0.78, the amount of B works out to be about 1.6 g (= 0.78 * 2 g).

Deposition of boron
The total amount of B introduced with boronization and boron dropper was 19 g and the corresponding number of B atoms were about 1.11×10 24 . Assuming a density of the B/C co-deposit layer 1.4 g cm −3 and the surface area of W7-X PFCs 192 m 2 (see table 1), the estimated average thickness of the deposited layer on the PFCs is about 70 nm. Asymmetries of deposition were observed along the toroidal and poloidal directions.
A couple of first wall tiles removed after OP1.2b were analyzed using ps-LIBS and an average thickness of B rich co-deposition layer of 120 nm was measured [19]. Since 120 nm also include C and O, so the thickness of B is only about 50%-70% of 120 nm. Within the accuracy of these estimates 120 nm and 70 nm are quite comparable. On each tile, the LIBS measurements were carried out at a number of locations along the toroidal and poloidal directions. A typical deposition pattern for H fuel and other impurities i.e. B, C, O, Fe, Mo and Na, is shown in figure 9 along the toroidal direction for a baffle tile TB-Z167. The He emission in figure 9 is not from He GDCs and He fuel, but from the ambient gas that was employed in the ps-LIBS measurement. Besides the H fuel, the source of B, C and O were from the boronization, erosion from TDU targets and desorption of water or air leak respectively. Fe and Mo impurities were originated from the stainless steel panels. The exact source of traces of Na is not yet clear, though alkali-beam diagnostics used Na beam for the measurements but the introduced amount of Na was too small to attribute to the observed depositions. The deposition pattern both i.e. the toroidal and poloidal directions shows non-uniformity, due to presence of facets on the tile surfaces (see figures 5(a) and (b)) which changes the angle of incident particles. The intensities of signals due to O and Mo appeared to be uniformly distributed, were very low and close to the signal threshold (3σ). Detailed investigation will be carried out by ERO2.0 modeling.
Considering a homogeneous B deposition layer of 70 nm on all the first wall tiles, the estimated mass of B deposited on various graphite PFC components are (assuming the density of layer, 1.44 g cm −3 ): The comparison of B amounts injected during boronization and by using the boron dropper as well as the deposited amount on various first wall components are presented in table 4. A major fraction i.e. ca. 60% of the injected B was traced with the measurements on some of the first wall components. The remaining amount could perhaps be deposited in the port ducts and other inaccessible areas of the plasma vessel. The small amount of B pumped out during the boronizations was completely retained in the thermal cracking unit and the chemical decomposer installed in this circuit, this amount is not known.

Simulations of erosion and impurity migration
4.1. Field line tracing modelling of impurity transport Field line transport code (FLT) is a web-service interface of W7-X, which allows for following field lines, constructing Poincaré maps, estimating heat fluxes to the wall, evaluating magnetic coordinates, etc. The details of this interface were presented in [16]. FLT was used to estimate the deposition zones of the particles eroded from the TDU strike lines in a single step process ignoring re-erosion and deposition. About 108 000 particles were injected at the strike lines at the horizontal targets of the TDU in half-module 3 lower (HM3l) (see figure 10(a)) in the standard magnetic configuration, the sticking coefficient was taken as 1. It was found that ca.

Erosion/deposition estimations of carbon
In order to analyze the experimentally measured erosion/deposition values for OP1.2b, the estimations of erosion/deposition were carried out using the established relations for the physical and chemical sputtering and re-depositions [36,43]. The net erosion is given by the relation [44] ( ) Here, G , ero net G : ero gross Net, gross erosion. Gross erosion is the overall erosion due to physical and chemical sputtering, net erosion is the remaining erosion after taking the redeposition (by the incident ion flux and prompt redeposition) into account.
G : e Electron flux density s: Sticking probability of redeposited impurities P redep : Probability of redeposition Y self /Y i : Sputtering coefficients due to self/ion sputtering The incident H and impurity (C, O) ion fluxes were estimated from electron density and plasma temperature data measured by the Langmuir probes installed on the nearby targets of TDU 3l (lower). Based on the following relation for the ion flux [44]: Here the units are, The surface temperature of the target was taken from the measured infrared (IR) data. The concentrations of C and O impurities were varied in the range of 4%-10% and 0.1%-10% respectively. Keeping in mind the surface roughness (see figure 11(a)), the incident angle was varied from 40°-60°. The estimated plots of net erosion and deposition along the target for the plasma shot 20180807.014 with incident angles of 40°and 60°are shown in figure 11(b). The rate of difference between deposition and erosion at 60°is plotted in figure 11(c). The location of highest erosion indicating the location of strike line shows good agreement with the strike line observed with the IR cameras.

ERO2
.0 modelling of carbon erosion and transport ERO2.0 is a Monte-Carlo code for modelling PWI and 3D plasma edge impurity transport by calculating the test particle trajectories including the full gyro-orbits motions [45]. By parallelizing and including large simulation volumes it is capable of simulating erosion, deposition and material migration for a full 3D fusion device and has been used to simulate PWI in WEST [46], PSI-2 [47], LHD [48] and JET [18]. ERO2.0 is being adapted to interpret the erosion/deposition measurements for 12 C in W7-X [23]. The input plasma parameters such as ion/electron temperatures, ion/heat fluxes were used from EMC3-EIRENE estimations for the standard configuration [49]. The plasma background was assumed to be pure H. The ERO2.0 simulation considered C impurities only, with charge states Z=0-6, and C concentrations being a self-consistent result of the  simulation. No high-Z impurities were considered at this stage. Self-sputtering and prompt redeposition were taken into account.
The simulation results (see figure 12) show a reasonable match to the measured erosion and deposition pattern on a TDU horizontal targets TM2h6 shown in figure 3. Please note the pumping gap sides are in opposite direction in figures 3 and 12. The plasma strike line at ca. 150 mm from PG and slight deposition at ca. 300 mm appear at similar locations for TM2h6. Relatively larger erosion close to PG in ER2.0 deviate from the experimental values presumably due to chemical erosion being over estimated, however further estimations are required to understand the deviations. This is being incorporated in ERO2.0 further. Similar behavior was also observed for TM1h3 and TM3h6 with the marker layer (see figure 12).
ERO2.0 was used to simulate the depositions observed by EBS measurements on the tile on toroidal closure shown in figures 5(c) and (d). The maximum deposition on the closure tile is 6 × 10 19 C-atoms s −1 m −2 (see figure 13). For the 4809 s of operation in standard configuration in OP1.2b [23], the cumulative deposition works out to be 2.9×10 19 C cm −2 , which is in very good agreement with the black curves for the carbon deposition shown in the figure 5(c). However, this estimation is without considering OP1.2a plasma exposure on the modelling side, where standard configuration had 2481 s plasma time, probably with higher erosion due to large C and O impurities without boronization. In this case, the modelling might somewhat overestimate the carbon deposition. On the other hand, the above maximum deposition of 6.0×10 19 C-atoms s −1 m −2 is localized on a small spot on the very edge, on average it's about 3−4×10 19 C-atoms s −1 m −2 at the tile edge, so the agreement appears to be fairly good.

Summary
First efforts were made to understand the carbon balance in W7-X operated with inertially cooled divertor units in OP1.2. At the divertor targets, areas of high net-erosion as well as deposition dominated regions are observed, but net-erosion prevailed considerably. In particular, the strike line regions act as the dominant source of the carbon impurities in the plasma. Rather thick deposited layers with thicknesses >1 μm were observed on some baffle tiles adjacent to the divertor targets. On the heat shield tiles as well as in remote areas, for example on the plasma vessel behind the divertor, only very thin deposits with only few tens of nm thickness were found. This behavior, especially pronounced for OP1.2a, is a profound difference to divertors in tokamaks with carbon plasma-facing components, where the eroded material (from the divertor but also due to erosion at the wall surfaces in the main chamber by charge exchange neutrals) is typically redeposited back onto the divertor targets as well as in remote divertor areas. This may be due to the fact that in W7-X the divertor is not toroidally closed as in tokamaks and higher particle losses can occur during the parallel transport between the different divertor units. This behavior was determined on the basis of limited surface analyses, where different, selected plasma facing components have been taken out after OP1.2b. Moreover, the divertor and the PFCs were not actively cooled and cumulative effects were seen after OP1.2a with high C and O impurity levels and OP1.2b with reduced impurities due to boronizations. Therefore, it would be too early to conclude on a general picture of carbon migration in W7-X. For OP2, a new carbon fiber composite (CFC) divertor is being installed and all the PFCs will be actively cooled, and the input energy to the plasma will stepwise increased to 18 GJ over successive campaigns, the carbon balance will be followed up by in situ spectroscopic measurements and post-mortem analysis of first wall tiles.