The dependence of tungsten fuzz layer thickness and porosity on tungsten deposition rate and helium ion fluence

Fuzz formation on a heated tungsten surface in the presence of a helium-containing plasma and tungsten deposition source was investigated. Tungsten samples were exposed at 1123 K to pure helium plasma with ion incident energy of 76 eV, W/He ion flux ratio of ∼0.4×10−4 , and varied helium ion fluence from 0.18 to 3.4×1026 m−2. Fuzz thickness was measured by cross-sectional scanning electron microscopy to increase from 0.22 to 15 µm with increasing helium ion fluence. No indication of saturation in fuzz thickness at high fluence was observed, in contrast to fuzz produced on a tungsten surface without tungsten deposition. Additional tungsten samples were exposed at 1123 K to pure helium plasma with ion incident energy of 76 eV, helium ion fluence of ∼3.4×1026 m−2, and varied W/He ion flux ratio from 0.26 to 3.0×10−4 . Fuzz thickness increased from 7.5 to 120 µm with increasing W/He ion ratio. A final sample exposed at 1123 K to a mixed helium-deuterium plasma with ion incident energy of 76 eV, helium ion fluence of 0.18×1026 m−2, and W/He ion flux ratio of 2.2×10−4 developed nearly identical fuzz structures to that developed in a pure He plasma. As a function of deposited tungsten fluence, all results were found to trace out a single layer-growth curve given by a power law relation, indicating that fuzz thickness is independent of the W/He ion flux ratio in the range investigated and independent of any deuterium present in the plasma. As a result, for tungsten plasma facing walls in magnetic fusion devices at 1000–2000 K with 10−4 W/He ion flux ratio, fuzz with thicknesses greater than hundreds of microns may form in as little as 104 s (in the absence of ELM-induced erosion or annealing), and may more significantly affect its thermophysical properties than fuzz generated without a tungsten deposition source.


Introduction
In magnetic fusion devices, the plasma-facing walls can be sputtered by energetic ions or neutrals from the hydrogen isotope fuel, helium (He) ash, seeded gases, or other impurities.The sputtered material can be promptly re-deposited or transported in the scrape-off layer and deposited on other areas of the device.For example, ERO-2.0 modeling predicts W erosion (largely by charge-exchange D) at the first wall and W fluxes up to 10 17 m −2 s −1 at the divertor outer vertical target for a neon-seeded D plasma in a fusion device with all metal W walls [1].Note that, at the divertor outer vertical target, conditions for fuzz formation are expected to be met, including surface temperatures between 1000-2000 K and ion energies greater than 20 eV [2].In fact, both fuzz formation and W deposition were recently observed at the divertor of ASDEX Upgrade [3].
In the absence of deposition, the W fuzz layer increases with He ion fluence from ∼10 24 m −2 and saturates to a thickness of ∼7 µm at ∼10 28 m −2 [4,5].However, the fuzz layer thickness is enhanced when exposures are conducted in the presence of a W deposition source [6][7][8][9].Initial experiments by Petty et al [9], used a magnetron sputtering source with W deposition normal to the sample surface, and found that the He fluence threshold for fuzz formation was reduced below 10 24 m −2 .Later experiments by McCarthy et al [8] also used a magnetron sputtering source and found enhanced fuzz thickness for samples with W deposition versus without at the same He fluence.For example, a 7.5 µm-thick fuzz layer developed with a W/He flux ratio of 9 × 10 −3 at a He ion fluence of 9 × 10 24 m −2 (without deposition and at the same fluence, only a ∼400 nm-thick fuzz layer develops [4]).In contrast, Kajita et al [6] utilized a W sputtering wire placed upstream of a sample in a linear plasma device.Since the sample was oriented parallel to the plasma flow and W deposition, the fuzz grew from the leading edge of the sample, resulting in mm-thick fuzz at the leading edge (at a He ion fluence of 3 × 10 26 m −2 and W/He ratio of 10 −4 − 10 −3 ) and in nonuniform fuzz thickness which varied widely across the face of the sample.
This work revisits the growth of fuzz during W deposition, and systematically investigates how the fuzz layer thickness is affected by W to He flux ratio, W He , and He ion fluence, Φ He .Since He peak ion fluxes of 10 21 -10 22 m −2 s −1 are expected at the ITER divertor during D-T operation (depending on divertor neutral pressure) [10], and as mentioned above peak W deposition/re-deposition fluxes up to 10 17 m −2 s −1 are predicted for an all-W device, W He ≲ 10 −4 may be expected in the ITER divertor region.While higher W He on the order of 10 −3 were previously investigated in [8,9], this work focuses on W He = (0.26-3.0) × 10 −4 , characteristic of conditions in all-W devices.In contrast to [6], we use a deposition source normal to the sample surface to avoid the complication of non-uniform fuzz thickness.

Experimental setup
Bulk W samples (Midwest Tungsten, 99.95% purity, 25.4 mm diameter) were polished to a mirror-like finish (R a = 650 nm), cleaned ultrasonically in acetone and ethanol, and exposed to He or He-D 2 plasma in the PISCES-RF linear plasma device [11].Samples were exposed to pure He plasma at an ion flux, Γ He , of ∼9 × 10 22 m −2 s −1 , ion energy, E i , of 76 eV, and sample temperature, T s , of 1123 K in the presence of a W grid (ThermoShield, 99.95% purity, 0.38 mm diameter wire, 72% open area) positioned approximately 20 mm upstream (see figure 1).One batch of samples was exposed at a fixed W He ion ratio of ∼4 × 10 −5 and Φ He was varied from (0.18-3.4)×10 26 m −2 .A second batch was exposed at fixed Φ He ∼ 3.4 × 10 26 m −2 and W He was varied from (0.26-3.0)×10 −4 .A further sample was exposed in mixed 0.1He-0.9Dplasma at Φ He = 0.18 × 10 26 m −2 and W He = 2.2 × 10 −4 .Conditions for all samples are summarized in table 1.
Plasma properties, including plasma potential, V p , plasma density, n e , electron temperature, T e , and Γ He were measured approximately 64 mm upstream of the grid with a fast-reciprocating RF-compensated Langmuir probe (n e = 7.8 × 10 18 and 2.0 × 10 18 m −3 , T e = 15 and 9 eV for pure He and He-D plasmas, respectively).Since the W grid was positioned close to the W sample to maximize deposition, a Langmuir probe could not be inserted between the grid and sample to measure plasma properties downstream of the grid.Since the sample floating potential did not change significantly when the grid was unbiased or biased, it was deduced that T e , and hence V p , remained unchanged downstream of the biased grid.A 2 × drop in the He I (447.1 nm) emission line intensity averaged over the region in front of the sample was measured using an Avantes spectrometer.Hence, the volume averaged n He and Γ He downstream of the grid were also inferred to drop by the same factor.Although the geometric open area of the W grid is 72%, the effective open area for He ions is reduced to approximately 50% by the electrostatic sheath (thickness ∼0.1 mm from the Child-Langmuir law) around the grid wires.Therefore, it is assumed that Γ He to the open regions of the sample is the same as upstream of the grid, while Γ He is much lower to the regions of the sample shadowed by the grid wires.Note that the ion-neutral collision mean free path ∼50 mm can be taken as the disturbance length, which is more than 2 × larger than the distance between the grid and the sample.Therefore, negligible ions move in the transverse direction from the open to the shadowed region.The He + ion fraction in the mixed He-D plasma was determined using a spectroscopic technique [12].At the grid and at the sample, the He ion energy was calculated to be E i = V p − V bias , where V bias is the bias applied independently to the grid and the sample.Mass gain on the sample after plasma exposure was measured with a microbalance and used to calculate the W flux, Γ W , at the sample assuming uniform deposition on the sample and no reflection or re-erosion of deposited W. The latter can be assumed since for E i = 76 eV at the sample, reflection of W + from W is negligible at ∼1 × 10 −4 , sputtering of W by He + or D + is below the threshold, and sputtering of W by W + is negligible at ∼1 × 10 −3 [13].Sample temperature was measured with a thermocouple pressed against the back of the sample and was controlled by forced-air cooling of the sample manipulator.
After plasma exposure, the top surface of samples was imaged with scanning electron microscopy (SEM, FEI Apreo) and the fuzz layer thickness, L fuzz , was measured with crosssectional SEM after samples were mechanically fractured.Laser confocal microscopy (LCM, Olympus LEXT OLS4100) was also used to measure the height map and confirm L fuzz after fuzz was removed from one section of the sample using a sharp instrument [4].For sample I, with the thickest and most fragile fuzz layer, L fuzz was determined solely from LCM.Although the cross-section for sample I was viewed with SEM after a cut was made with a focused ion beam (FIB, FEI Scios), this provided only a measure of the height localized to a 5 µm-wide region, and as will be shown in section 3, structures (and their height) varied significantly across the surface of this sample.

Results and discussion
SEM images are shown in figure 2 of the surface of samples B-E generated at a W He ∼ 4 × 10 −5 and various Φ He .As can be seen from the high magnification images, a background of fuzz structures appears at all Φ He , yet fuzz tendrils become longer and more intertwined at higher Φ He .Superimposed are what look to be nanotendril bundles (NTBs) [14], which become larger and more densely distributed on the surface with increasing Φ He such that the NTBs become interconnected at the highest Φ He (see the low magnification image of sample E).Hence, the fuzz structure formed during deposition is not identical to that of conventional fuzz formed without W deposition.Note that in [14], it was reported that NTBs formed on the surface of samples when exposed to impurity-seeded plasma, due to erosion and re-deposition at the W surface, suggesting that deposition effects are potentially linked to NTB formation.
Figure 3 shows SEM images of the cross-section of these samples.A relatively uniform fuzz layer is observed on all samples, with interconnected arch-like NTBs at the highest Φ He .Likewise, similar arch-like NTBs were observed in NAGDIS-II for W exposed to a He plasma with impurity seeding [14].The uniform part of the fuzz layer thickness increases from 0.22 to 15 µm as Φ He increases from 0.18 to 3.4 × 10 26 m −2 .This is plotted in figure 4(a), along with a power law fit.Also included in figure 4(a) are previous data from Petty et al [9] (E i = 40 eV, T s = 1100 K), McCarthy et al [8] (E i = 80-100 eV, T s ∼ 1050-1150 K), and Kajita et al [7] (E i ∼ 60-70 eV, T s ∼1320 K) for W fuzz formed in the presence of W deposition normal to the sample surface.Comparison of this work and that of [7][8][9] with data from Petty et al [4] of conventional fuzz (i.e.W He = 0) demonstrates that fuzz thickness is enhanced when deposition is present.
Images of samples A, E-F, and H-I after plasma exposure at Φ He ∼ 3.4 × 10 26 m −2 and various W He ratios are found in figure 5.For samples A and E that were exposed at low W He ratio = (2.6-5.5)× 10 −5 , the fuzz layer generated was uniform across the surface.In fact, SEM images of the surface of sample A look very similar to those of sample D with nearly double the W He but half the Φ He (see figure 6(A) versus figure 2(D)).For samples F, H, and I that were exposed at higher W He ratio = (0.93-3.0) × 10 −4 , the fuzzy surface developed a pattern that reflects the geometry of the grid deposition source (i.e. with 10 wires/inch).SEM images of the top surface of these three samples are shown in figure 6, He ratio.The shadowed regions of sample F and I do not show such long intertwined tendrils, but instead show a relatively uniform fuzz layer since lower Γ He is expected at the shadowed regions (as discussed in section 2).The shadowed region of sample H contains many sheets of solid W tens of microns in size, in addition to the fuzz tendrils.The sheets are thin (i.e. less than 100 nm), curved inwards, and have fuzz tendrils stretching out of the edges.Similar sheets/membranes were observed in [6,15] and in [8] (at high W He and high Φ He ).The reason for the formation of the sheets is currently unknown since the sheets were primarily observed on samples H and G with W He = 1.5 × 10 −4 and not at higher or lower W He , however in [15] it was suggested that the sheets/membranes grew between existing fuzz tendrils.Note that the existence of tendrils in the shadowed regions of samples suggests that some He + are in fact implanted here.
Cross-sectional images of samples A, E-F, and H-I are shown in figure 7.For sample I, a 5 µm-wide FIB cut on the surface was made (instead of mechanically fracturing) and the  He value for sample J between that of samples H and I, sample J formed a uniform fuzz layer (i.e.no pattern reminiscent of the upstream W grid) since interconnected NTBs and long intertwined tendrils have yet to form at low Φ W .
From figure 4(a), L fuzz at low W He and low Φ He agrees with that of conventional fuzz (i.e.W He = 0) [4], as expected, but is significantly larger than conventional fuzz at high W He or high Φ He .In fact, unlike conventional fuzz where the thickness saturates with Φ He , L fuzz in this work does not show any indication of saturation.The lack of saturation is due to a continuous supply of W being deposited and supply of incident He, in contrast to conventional fuzz formed without deposition where W is supplied from the bulk and is limited by He being able to penetrate the fuzz layer to reach the bulk [5].
When L fuzz is plotted versus Φ W as in figure 10(a), all data from this work collapse onto a single curve, independent of W He ratio, that can be approximated by L fuzz [µm] = 4.4 × 10 −28 (Φ W ) 1.28 , which again shows no indication of saturation.Good agreement is observed between the fuzz measured here at W He = (0.26-3.0) × 10 −4 (shown as closed circles) and fuzz formed by Kajita et al in Magnum-PSI at W He = 8 × 10 −5 [7] (shown as open square).Independence of L fuzz on W He may be explained by saturation of He retention in the deposited layer, similar to saturation of He retention observed in bulk W [16,17].According to [17], He retention in bulk W at 1107 K (similar to T s utilized here) saturates at Φ He ≳ 2 × 10 24 m −2 .Assuming the retained He lies within a 30 nm depth in bulk W [18], this corresponds to a W He ratio ≲ 9.4 × 10 −4 , in agreement with the values used here.Note that for sample J (with mixed He-D plasma), the He flux was reduced by ∼20× compared to the other samples.Yet the fuzz layer thickness for sample J still lies along the curve given by the equation in figure 10(a), since the W He ratio was still ≲9.4 × 10 −4 .Figure 10(a) also shows that L fuzz is independent of D ions in the plasma.This is not surprising since D retention in bulk [19] and co-deposited [20] W decreases exponentially with increasing sample temperature and it is negligible at the temperature used here to form fuzz (i.e.1123 K).
In figure 10(a), previous data from Petty et al [9] and McCarthy et al [8], which were generated with low Γ He using a magnetron sputtering device, do not lie along the curve traced out by this work.Differences in E i and T s are not the cause of this deviation since data from Petty et al and McCarthy et al, which agree with each other, were taken with E i and T s values lower and higher than that used in this work.To help explain the deviation, included in figure 10(a) is the thickness of a fully dense W film, which is linearly proportional to Φ W .Note that if a fully dense W film uniformly coated fuzz trendrils on a pre-existing fuzz layer, the thickness of the fuzz layer would also increase with Φ W at the same rate.However, figure 10  , respectively).Therefore, it is likely that film deposition dominated over fuzz formation during their experiments at low fluence, as was suggested in [8].Additionally, fuzz annealing may have reduced fuzz growth as was suggested in [9], since fuzz annealing has been observed as low as 1100 K [21].
The density of W in the fuzz layer, n fuzz , can be calculated by dividing Φ W by L fuzz , and is related to the fuzz porosity, p fuzz , by 1 − p fuzz = n fuzz nW = ΦW/L fuzz nW , where n W is the density of bulk W. This is plotted in figure 10     From figure 10(b) we find that 1 − p fuzz decreases, i.e. p fuzz increases, with increasing Φ W for fuzz generated at low W He (as in this work and [7]).In fact, all of the fuzz samples generated in this work have calculated porosities of 95-99%.In contrast, at high W He (as in McCarthy et al [8] and et al p fuzz decreases with increasing Φ W and approaches values for fully dense W. Therefore W He ratios ≲1 × 10 −3 are required to develop low-density/high-porosity fuzz in a co-deposition scenario.Higher W He ratios still lead to layers with forest-like morphology, but with higher density/lower porosity.While this conclusion is based on only two studies in low Γ He and Γ W magnetron devices, it is interesting that the W He ratio threshold of ∼1 × 10 −3 agrees with the value of ∼9.4 × 10 −4 below which He concentration in W is estimated to saturate, as calculated above.Therefore, at lower He concentrations, corresponding to higher W He ratios, fuzz formation is limited by the availability of He. As discussed above, Γ He at regions of the sample shadowed by the W deposition grid is expected to be lower than at the sample open regions, although non-zero.From the thickness of the fuzz layer in the shadowed region of samples, it is estimated that Γ He is ∼10× lower than in the open regions.For samples A-E with low Γ W such that W He remains ≲1 × 10 −3 in the shadowed regions, no difference in fuzz thickness in the shadowed versus open regions is expected and no pattern is observed on the sample surface.However, for samples F-I with high Γ W such that W He may be >1 × 10 −3 in the shadowed regions, the availability of He may once again limit fuzz formation.For sample I for example, L fuzz = 16 µm and p fuzz = 0.90 in the shadowed region, in contrast to L fuzz = 120 µm and p fuzz = 0.99 in the open region.While, the sheath thickness around the deposition grid was larger for samples F, H, and I (with larger grid bias voltage), this resulted in a negligible change in the effective open area which cannot explain why the shadowing was observed on these samples but not samples A-E.Note that although most of the W atoms emitted from the grid can be ionized before reaching the sample (mean free path = 2-8 mm), the angular distribution of sputtered W atoms can result in nearly uniform W deposition onto the sample.Thus, the thinner fuzz layer in the shadowed regions of samples F-I is not thought to be caused by non-uniform W deposition.

Summary
Fuzz growth on a W surface for He plasma with additional W deposition normal to the surface was investigated.Samples were exposed in the PISCES-RF linear plasma device at T s = 1123 K and E i = 76 eV.One batch of samples was exposed at a fixed W He ratio of ∼4 × 10 −5 and varied Φ He = (0.18-3.4) × 10 26 m −2 , while a second batch was exposed at fixed Φ He ∼ 3.4 × 10 26 m −2 and varied W He = (0.26-3.0) × 10 −4 .A further sample was exposed in mixed 0.1He-0.9Dplasma at Φ He = 0.18 × 10 26 m −2 and W He = 2.2 × 10 −4 .Surfaces formed were found to be comprised of uniform fuzz with NTBs that formed arch-like features when connected.Fuzz thickness was observed to increase with Φ He and showed no indication of saturation at Φ He up to 3.4 × 10 26 m −2 , in contrast to the growth of conventional fuzz [4].Thickness also increased with W He ratio for both the pure He and mixed He-D exposures, up to 120 µm.When plotted against Φ W , fuzz thickness was found to increase as Φ 1.28  W , independent of the range of W He ratios investigated here and in [7].Much smaller fuzz thicknesses were previously generated in magnetron sputtering devices with higher W He ratios, likely due to lower He concentrations in the deposited films.
In the case in which the ITER first wall would be covered by W, W fluxes up to 10 17 m −2 s −1 are predicted at the divertor [1].For ITER and other devices with high He flux fraction in the plasma on the order of 10 21 -10 22 m −2 s −1 [10], this results in W He ≲ 10 −4 .Our results show that under such conditions, fuzz with thickness of 1 µm may form in as little as 63 shots and of 100 µm may form in ∼2300 shots.While this is only an upper limit on fuzz thickness (since ELM-induced erosion and annealing, impurity seeded-erosion, prompt redeposition, glancing magnetic field, etc are not considered here), results show that fuzz thickness can be greatly enhanced by W deposition. Hence, the enhanced fuzz layer may more significantly affect the thermo-mechanical properties (including thermal conductivity [22], elastic modulus [23], and arcing susceptibility [24]) of W plasma-facing components than conventional fuzz.

Figure 1 .
Figure 1.Schematic of the experimental setup, including sample and grid separation (not to scale), in the PISCES-RF linear plasma device.

Figure 2 .
Figure 2. SEM images of the top surfaces of samples exposed at Ts = 1123 K to He plasma with E i = 76 eV, W He ∼ 4 × 10 −5 , and Φ He = (B) 0.18, (C) 0.86, (D) 1.7, or (E) 3.4 × 10 26 m −2 .Left, center, and right images correspond to low, medium, and high magnification, respectively.The scale bars indicated in the top row of images apply to all rows of images below.
cross-section was viewed with the sample at either 45 or 5 • .All samples with the exception of sample H show typical fuzz layers.For sample H, the open region shows fuzzy layers mixed with W sheets, while the shadowed region shows an increase in the density of W sheets.Despite the complex fuzz structures observed in the sample open regions in figure 6, the fuzz layer thickness was found to be approximately uniform for all samples except sample I. Since the height of sample I in the open region varied significantly across the surface, the height was determined from LCM (which measured a larger lateral area) instead of FIB/SEM.The height map of sample I in the open region is shown in figure 8.In determining the fuzz thickness for samples F, H, and I, only the open regions are considered since the fuzz layers in the shadowed regions most likely had lower Γ He , and thus resulted in lower thicknesses.Thicknesses for samples at Φ He ∼ 3.4 × 10 26 m −2 are plotted in figure 4(b) and found to increase from 7.5 to 120 µm with increasing W He ratio.Finally, SEM images of the top and cross-section of sample J (Φ He = 0.18 × 10 26 m −2 , W He = 2.2 × 10 −4 ) are shown in figure 9.The surface is similar to samples C and D, with the size and density of NTBs lying in between.The crosssection is also very similar, with L fuzz = 2.4 µm (i.e. between that of samples C and D).As expected, this is much larger

Figure 3 .
Figure 3. SEM images of the cross-section of the samples shown in figure 2.

(a) 4 .
Fuzz thickness, L fuzz , as a function of (a) Φ He and (b) W He flux ratio for W samples exposed at Ts = 1123 K to He-containing plasma at E i = 76 eV, W He = (0.26-3.0) × 10 −4 , and Φ He = (0.18-3.6) × 10 26 m −2 .Also included in (a) are fuzz measurements from [7-9] with W deposition normal to the sample surface at W He = (0.8-90) × 10 −4 , and measurements of conventional fuzz without W deposition [4].In (a) and (b), the fit to the data from this work is shown by the dashed lines.
(b) as a function of Φ W .

Figure 6 .
Figure 6.SEM images of the top surfaces of samples exposed at Ts = 1123 K to He plasma with E i = 76 eV, Φ He ∼ 3.4 × 10 26 m −2 , and W He = (A) 0.26, (E) 0.55, (F) 0.93, (H) 1.5, or (I) 3.0 × 10 −4 .Images for samples F, H, and I are of open and shadowed regions on the surface.Left, center, and right images correspond to low, medium, and high magnification, respectively.The scale bars indicated in the top row of images apply to all rows of images below.

Figure 7 .
Figure 7. SEM images of the cross-section of the samples shown in figure 6.

Figure 8 .
Figure 8. Height map measured by LCM of sample I exposed at Ts = 1123 K to He plasma with E i = 76 eV, Φ He = 3.4 × 10 26 m −2 , and W He = 3.0 × 10 −4 .Note that fuzz was removed (near the center of the image) with a sharp instrument to expose the bottom of the fuzz layer.

Figure 9 .
Figure 9. SEM images of the (a) top and (b) cross-section of sample J exposed at Ts = 1123 K to 0.1He-0.9Dplasma at E i = 76 eV, Φ He = 0.18 × 10 26 m −2 , and W He = 2.2 × 10 −4 .Left, center, and right images in (a) correspond to low, medium, and high magnification, respectively.

Figure 10 .
Figure 10.(a) Fuzz thickness and (b) fuzz porosity as a function of Φ W for W samples exposed at Ts = 1123 K to He-containing plasma at E i = 76 eV , Φ He = (0.18-3.6) × 10 26 m −2 , and W He = (0.26-3.0) × 10 −4 .Also included are fuzz measurements from [7-9], with W deposition normal to the sample surface at W He = (0.8-90) × 10 −4 .The fit to the data from this work is shown by the dashed line and the equation provided, while the gray dotted line corresponds to fully dense W.

Table 1 .
Conditions at the grid and sample during fuzz formation: E i = ion energy, Γ He = He ion flux, Φ He = He ion fluence (sample temperature, Ts, was kept at 1123 K).Also shown are fuzz thicknesses, L fuzz , formed.