RF sheath induced sputtering on Proto-MPEX part 2: Impurity transport modeling and experimental comparison

Linear plasma devices offer a cost-effective method to perform PMI research to understand critical material issues (end-of-life, synergistic effects of neutron/plasma irradiation, etc.) for future fusion devices.^1–4 The proposed Material Plasma Exposure eXperiment (MPEX) is a linear plasma device that is planned to have the capability to perform such PMI research. The Prototype Material Plasma Exposure eXperiment (Proto-MPEX) is the pulsed linear magnetized plasma device with the purpose of performing the research and development (R & D) for the plasma heating sources for MPEX.⁵ As the source R & D finishes operation on Proto-MPEX, it has an opportunity to perform PMI studies of its own. The device has already demonstrated divertor-like plasma parameters and ion fluxes to its target,⁶ but it is also important for Proto-MPEX and subsequently for MPEX to understand the impurity generation and transport. After the first W exposure in Proto-MPEX with pure deuterium as the working gas, the target had a deposition layer of mainly oxygen (O) and aluminum (Al).⁷ The hypothesis given was that the aluminum nitride (AlN) helicon window used in Proto-MPEX was the initiation point of Al, and the O comes from water vapor in the vacuum chamber. The base pressure in Proto-MPEX was $6.67×10−4$ Pa, and residual gas analyzer measurements indicate $30%$ was water vapor. Other works have noted that impurities can enhance tritium retention within impurity dust, change the erosion/redeposition behavior of the first wall, or enhance W fuzz growth.^8–12 This paper focuses on the impurities that come from the helicon window. In part 1 of this work, a model was created of the Proto-MPEX helicon window in COMSOL Multiphysics^®,¹³ with a coupled rectified DC sheath potential model^14–17 to get the sheath potentials on the helicon window.¹⁸ The potentials are fed into the F-TRIDYN binary collision approximation code¹⁹ to calculate reflection and sputtering yields and then Global Impurity TRansport (GITR) code²⁰ tracks sputtered particle transport within Proto-MPEX. In this study, experiments on quantifying the source erosion/redeposition and target deposition are performed to compare to additional modeling of Proto-MPEX using the described methodology.

A detailed overview of the Proto-MPEX device has been given elsewhere,²¹ but the important parameters are listed here. Proto-MPEX uses a helicon antenna at 13.56 MHz as the primary plasma ionization source with deuterium as the working gas. The helicon antenna is shown in Fig. 1. Pulses are 0.5 s in length with a repetition rate of every 3 min to allow for component cooling. The magnetic field profile and plasma radius up to the last uninterrupted flux surface (LUFS) are shown in Fig. 2. The helicon antenna has an AlN radio frequency (RF) window as the vacuum boundary which has the plasma limiting radially on it. The plasma terminates at either the dump plate (a 0.4 m diameter stainless steel plate with 0.0015 m thickness) at Z = 0.5 m or the target (the desired location for the plasma to stream to) that is either a 0.05715 × 0.05715 m W plate with a thickness of 0.003175 m or an exchangeable MAPP target. The target region is a vacuum chamber that has multiple ports which include the double Langmuir Probe (DLP),²² the W plate target, or the Materials Analysis and Particle Probe (MAPP) target and target holder system.⁷ The DLP and MAPP occupy the same space within the target region of Proto-MPEX and cannot be used simultaneously. To know the plasma parameters at the target, the DLP measurements are taken for a plasma pulse and then the MAPP system is inserted for a plasma pulse(s). The experimental conditions for the helicon and target are shown in Fig. 3. The helicon power coupled to the plasma was 115 kW. The helicon power was applied between 4.17 and 4.67 s. The gas fueling starts at 4.0 s, and pressure rises near the helicon until power is applied, after which the helicon mode plasma²³ is established, and the neutral pressure drops as ionization increases. The plasma after being transported to the target recombines in the target region which leads to an increase in the gas pressure as evidenced by the neutral gas pressure increase at the target.²⁴ The electron density on Proto-MPEX starts low, ramps with the helicon power, and then settles at $5×1019m−3$ while the electron temperature starts at 8 eV and comes down to 2.5 eV.

The MAPP W target was exposed to eight shots with a D ion flux of $2.8×1023 m−2 s−1$ and a total fluence of $2.25×1024 m−2$⁠. The exposed W target is shown in Fig. 4(a). The color pattern is indicative of a changing layer thickness. To determine the layer composition and thickness, a depth profile was taken. An argon ion gun was used to mill the surface, and x-ray photoelectron spectroscopy (XPS) measurements were taken as the surface was milled to get the composition. The depths were determined from a SiO₂ standard for depth milled as a function of sputtering time from the Ar ion beam. The standard error for the XPS composition data is given by the thickness of the plotted data. A Thermo Scientific K-Alpha system was used for the XPS data. The location of the XPS and milling on the surface is shown in Fig. 4(b). The width of the exposed target area was 0.012 m, and the XPS location was taken to be the center of the plasma column. The results of the XPS depth profiling are shown in Fig. 5. The predominant impurity was Al and O, with some trace stainless steel components (Fe and Cr). The 1195 nm depth is the crossover point for the O and W signal, indicating the bulk W has been reached. From the O and Al ratios, the target deposition was likely an amorphous aluminum oxide layer.

The AlN window after the experiment was finished was cut in half to observe the inner surface, which is shown in Fig. 6. The helicon antenna pattern has been overlayed on the window, which shows that the clean areas are under the antenna and the discolored areas are regions between the location of the helicon antenna around the helicon window. The blackened area was also analyzed with XPS (location of analysis shown in Fig. 6) and is shown in Fig. 7. The bulk of the composition was found to be similar to the composition of the targets, i.e., predominantly Al and O which leads to the helicon window being suspected to be the source of the impurities observed on the target. Profilometry measurements were taken from the clean areas under the helicon antenna to the blackened areas to test if the region is an erosion or deposition area. The profilometry showed a several micrometer thick jump in height indicating the blackened areas are deposition regions. Figure 8 shows an additional profilometry scan over 5.7 mm that was taken across a location with a “blister”-like surface (location of analysis shown in Fig. 6). The height jumps as it moves off the blister-like region, indicating that this was also a location of deposition, and the region has flaked off, leaving behind the blister-like location. Upon replacing the AlN helicon window, a silicon nitride (⁠$Si3N4$⁠) helicon window was used in Proto-MPEX for performing thermomechanical testing as it is the chosen material for the MPEX helicon window.

The coupling of the COMSOL, sheath model, and GITR code is given in detail in part 1 of this work,¹⁸ but an overview of the models and differences will be given here. The motivation of this modeling effort is to understand the erosion/redeposition pattern on the helicon window and to understand the impurity accumulation at the target.

The COMSOL RF module is used to solve the 3D axisymmetric full-wave simulation in the frequency-domain for a one-meter in length plasma column centered at the Proto-MPEX helicon antenna. The simulation geometry includes the helicon antenna, the AlN helicon window, a 5 mm wall thickness dielectric sheath layer as the interface between the plasma and the helicon window, and the 0.12 m diameter plasma column. The helicon window has a density of 3300 $kg/m3$⁠, a dielectric constant of 9, and an electrical conductivity of 0 S/m. The dielectric sheath layer starts with a dielectric constant of 1 and electrical conductivity of 0 S/m. The sheath model is an iterative impedance matching that outputs the rectified RF sheath potential on the helicon window where the sheath impedance is assumed to be known.^16,17 Matching the displacement current to the helicon window in the COMSOL model to the sheath potential, a new dielectric constant is calculated and used in the COMSOL model. This is repeated until the solution converges. The sheath model has two bounding cases based on the fact that the plasma conditions in Proto-MPEX are between an unmagnetized and magnetized electron boundary limit. This condition arises when the electron Larmor radius is comparable to the sheath width.

To model both the startup of the plasma and the steady-state region of the plasma, two density profiles were modeled in COMSOL. To get the electron density and temperature to model, additional DLP measurements were taken at Z = 1.5 m. A scan of the DLP through the plasma radius was done to get the edge and core densities. The start of the plasma pulse is taken to be during the first 70 ms, and the steady-state conditions are taken at 100 ms into the pulse. The two radial DLP electron density and temperature profiles were taken at t = 4.165 s for the plasma startup and at t = 4.175 s for the steady-state time are given in Fig. 9. The startup part of the plasma has an edge density of $3×1017 m−3$ and a maximum core density of $1.5×1018 m−3$⁠, and is referred to as the low-density case, while the model for the steady-state time frame has an edge density of $3×1018 m−3$ and a maximum core density of $5×1019 m−3$⁠, and is referred to as the high-density case. The average electron temperature used in the COMSOL simulations was 4 eV for both density cases. This gives rise to four total cases to be modeled; the low-density and high-density cases and the unmagnetized and magnetized sheath electron cases. The magnetic field profile from Fig. 2(a) was used in the simulation, and the input profiles for the density followed the flux surfaces.

The unrolled 3D converged sheath potentials on the helicon window from the coupled COMSOL and sheath model are shown in Fig. 10. The location of the helicon antenna around the helicon window is shadowed in the image for reference. The hot spots are seen underneath the helicon antenna at varying locations, and the potentials are higher for the magnetized case due to the higher sheath impedance. The low-density cases also show a larger RF sheath potential as expected from the Trivelpiece Gould (TG) mode²⁵ that is coupled near the window putting additional power on it.²⁶ This power near the helicon window increases the current imposed on it and thus the increased RF potential. These four cases set the bounds for the RF potential on the window to be used in the sputtering calculations.

The converged RF sheath potentials are used to calculate the reflection and sputtering yields from a pure deuterium flux to the surface with F-TRIDYN.¹⁹ F-TRIDYN is a binary collision approximation code that calculates the reflection and sputtering yields of a single element surface by incoming ions of a single element as a function of the incoming ion's energy and angle. The surface of the helicon window was set as pure Al to simplify the reflection and sputtering rate calculations, not include chemical processes as GITR does not include them, and the N is assumed to be lost to the vacuum due to experimental observation of low concentration on the target. Observations of Al enhancement on a D plasma exposed AlN sample were found²⁷ which is also likely occurring on the helicon window within Proto-MPEX. To better capture the ionization mean free path within Proto-MPEX, the electron temperature profiles from Fig. 9(d) was used within the GITR simulation. A 3D map of the sputtering rates is then generated and used as a weighted distribution to assign the starting location of one million particles. These particles are initialized at the surface of the helicon window in GITR, and their transport is tracked in the simulation geometry.

The geometry of the GITR simulation was from Z = 0.5 to Z = 4.14 m with a 0.135 m diameter. This geometry is extended to track the particles that leave the helicon region and make their way to the target region of Proto-MPEX. The temperature profile was mapped to magnetic flux contours, giving constant temperatures for each contour throughout the simulation geometry. No axial variation in electron temperature is considered here because no auxiliary heating is included. The density profile was also assigned to magnetic field lines, which are mapped throughout the simulation geometry. The plasma is a compressible fluid, and as the magnetic field changes throughout the device, the density is scaled to account for the compression. The direction of the background plasma flow velocity is along the field lines, and the magnitude was set as a piece-wise function. Underneath the helicon antenna is a stagnation zone with little measurable flow, but the Bohm criterion requires that the Mach number is one at flux line ending point. Based on previous B2-EIRENE modeling constrained by experimental measurements,²⁸ the piece-wise function for the Mach number was assigned to be 0.1 at the center of the helicon window and increases linearly to one at the ends of the simulation geometry.

The GITR simulation is run for a minimum δt of $1×10−10$ s with a new adaptive stepping feature that allows for variation in the time step depending on force magnitudes on the impurities. The maximum time step is $1×10−7$ s, and the maximum number of time steps is $5×105$⁠. $1×106$ particles are introduced to the simulation geometry, and particle movement is tracked via gyro orbits, Lorentz forces, Coulomb collisions, and particle drifts until they reach a surface. The helicon antenna self-sputtering was tracked by adding additional weight to particles that hit the window with sufficient energy to cause sputtering. The weight is the fraction of the total sputtered impurity flux that is assigned to the $1×106$ introduced particles. The energy of the Al ions coming to the helicon window remains a function of the RF sheath potential for the GITR simulations. The rest of the simulation geometry is a boundary with a sticking coefficient of one. The steady-state erosion and deposition profiles on the helicon window for the four cases, along with the ionized Al impurity flux at the target, are shown in Figs. 11–14. Each case shows net erosion from under the helicon antenna, and gross deposition is between the helicon antenna structure. The impurity flux at the target is peaked at the core for each case, with the highest flux being for the high-density magnetized case. The magnetized cases have a higher peak impurity flux at the target than the unmagnetized cases as expected from the higher RF potentials seen on the window. The two density cases, however, show a similar impurity flux at the target. From the RF sheath potentials on the helicon window being higher for the low-density case, the impurity flux to the target would have also been thought to be higher, but the lower D flux to the window would reduce the sputtering rates.

The RF sheath potential on the helicon window is thought to be the main cause of the erosion. Proto-MPEX operates with a 0.5 s plasma pulse consisting of the startup and the steady-state time. To measure the RF potential, a capacitive probe measurement²⁹ was taken at Z = 2.75 m at a radial point 4 cm off the machine center-axis. The RF potential measured in Proto-MPEX plasma pulse 31439 is shown in Fig. 15 along with the helicon power time history. The RF input power was 90 kW for the capacitive probe measurement plasma pulse. The beginning of the pulse is seen to have the highest magnitude but as the pulse continues the RF potential grows. The end of the pulse as the RF is cut off does not have an additional spike in RF potential, giving agreement to using two density cases in the model.

To investigate if the measured RF potential leads to a change in the impurity deposition seen on the target, a fluence scan was performed. The $Si3N4$ helicon window was in place during these exposures, and the input helicon power was the same as for the AlN window pulses. However, the magnetic field used for these exposures was ramped during the start of the RF pulse to reach a higher field in the helicon region. This ramp was needed to maintain the high-density plasma pulse operations. Pulse lengths of 0.137, 0.238, and 0.383 s were used in single plasma pulse MAPP W target exposures. The impurity layer thickness (again measured as the crossover point for the O and W) for the three exposures are shown in Fig. 16. The fluence scan corroborates the capacitive probe measurements and indicates a slope change for the pulses that enter the steady-state period. However, there is a time delay from looking at the second point in the fluence scan which occurs after the initial RF potential spike, 4.40 s vs 4.23 s, respectively. The second point contributes to the majority of the accumulation at the target, and this time delay occurs due to the helicon region magnetic field ramping that Proto-MPEX used for these experiments. The low-density startup phase of the plasma pulses is a TG plasma which has high power loss at the plasma edge near the helicon window,²⁶ and the steady-state period has more power absorption in the plasma core which reduces the RF potential on the helicon window. The ramping magnetic field has a longer beginning transient phase which has both an impurity creation and transport delay before getting to the steady-state time.

From looking at the low-density and high-density cases, the higher densities have a reduced RF sheath potential because of the lower impedance and also having the power deposited in the plasma core and not near the helicon window. The plasma temperature in the helicon region contributes to the plasma fluxes on the helicon window but to a lesser extent than the density. The electron temperature also needs to be sufficient to cause ionization, so a balance between fueling and electron temperature should be maintained so that the electron temperature does not rise to levels that can cause an enhanced sputtering on the helicon window over that of the RF sheath potential alone.

Eight pulses were used for the exposure, resulting in the XPS profile in Fig. 5 where the pulses can be seen by variations in the signals. The locations at greater depths on the sample show a smoothed profile due to diffusion from additional plasma heating during subsequent pulses. Carbon being near the surface shows that carbon is deposited between pulses and the oxygen signal increases opposite to carbon, meaning that carbon is brought to the target surface during the plasma pulse. Room temperature W after 1 h of O exposure will grow a 1 nm oxide³⁰ which would not be able to explain the thickness of the measured O, also indicating that O is brought to the target during the plasma pulse.

The composition of the helicon window and the targets show that the helicon window is responsible for impurity generation, but to check this, a $Si3N4$ helicon window was used with a MAPP W target exposure. The magnetic field mapping was the same as for the AlN window exposure, and 90 kW of helicon input power was used. The XPS results are shown in Fig. 17 and again show a high O content while having Si as the next most predominant element. This confirms that the helicon window is the source of impurities and not another location.

The limiting location on the helicon window includes flux surfaces that terminate on the stainless steel helicon flanges and the Teflon O-ring between the flange and the window. Additional sputtering of the stainless steel flange and the O-ring leads to Fe, Cr, Ni, and F observed on the window. For the $Si3N4$ helicon window, the Teflon O-rings were shielded from the plasma in the new stainless-steel flanges, and no F was seen on the target. Carbon will always be present within a device due to material impurities and pump oils and is also introduced in Proto-MPEX during previous graphite target operations⁶ and on the MAPP W targets from rough polishing. The N from the window is at very low concentrations and is assumed to be lost to the vacuum. The Gibbs free energy of formation at room temperature, $ΔfG°$⁠, for NH₃ is $−16.4 kJ/mol$⁠, while NO is $86.6 kJ/mol$⁠, and NO₂ is $51.3 kJ/mol$⁠.³¹ The more negative the value, the more likely the reaction is to occur while a positive value will reverse formation to create more of the parent elements. The formation of NH₃ is the most likely compound to form with N and is expected to be lost to the vacuum. The Gibbs free energy of formation at room temperature for $Al2O3$ is $−1582.3 kJ/mol$⁠, while AlO₂ is $−91.7 kJ/mol$⁠, AlN is $−287 kJ/mol$⁠, and AlH is $231.3 kJ/mol$⁠. This indicates that $Al2O3$ is the compound that is most likely to occur with Al, and NH₃ will form with N as O is going to Al. The ratio Of O to Al from the XPS profiles matches the compounds expected to form from the formation energies which indicates that $Al2O3$ is the predominant species. The O is likely initially from the water vapor within the device which $H2O$ has a value of $−273.2 kJ/mol$⁠. $Al2O3$ likely forms with O which leaves N to form NH₃ This also matches with the Si compounds. The Gibbs free energy of formation at room temperature for SiO is $−127.3 kJ/mol$⁠, while SiO₂ is $−853.6 kJ/mol$⁠, and $Si3N4$ is $−647.3 kJ/mol$⁠. SiO₂ will be the most likely to form on the targets, and N will form with the H to again form NH₃.

The experimental profile on the helicon window shows a deposition pattern between the helicon antenna that sits over the window. The modeling shows hot spots that arise on the helicon window for the various conditions, but all of them are beneath the helicon antenna matching that of the experiment. The simulation case with the magnetized electron sheath limit shows patterns that most closely resemble the experimental pattern, and both density cases have been shown with the helicon window in Fig. 18. The plots show agreement with the deposition being between the helicon antenna straps and the regions of erosion being under the antenna. The profilometry measurements also indicate deposition between the antenna location and the blister-like location that was analyzed has flaked off as indicated by the change in surface texture and the build-up in thickness before a sudden drop-off. The helicon window in question has been used on Proto-MPEX for 3+ years under several different heating scenarios and varied magnetic field profiles which are the likely cause for any differences seen, but the primary features from the model match the experimental helicon window deposition pattern. As the helicon window was sputtered initially and redeposits back on the window, the material properties change, i.e., the sputtering yield, surface composition, and the sheath impedance as the thickness of individual locations changes. The modeling does not reflect some of these changes, and assumptions were used about them, i.e., the pure Al surface for sputtering yields and uniform thickness. The varied magnetic field profiles used on Proto-MPEX will also change the limiting location of the plasma on the helicon window leading to large changes in the D flux to the window. Moving the plasma limiting location off the helicon window is the plan for MPEX to reduce the particle and heat flux to the window.³² 

Chemical leaching of AlN and $Al2O3$ was investigated on PISCES-A and not found to be significant.²⁷ The chemical erosion rates of AlN and $Al2O3$ are not known and can be a factor in steady-state operations as the helicon window temperature increases and could reach a temperature where chemical erosion has a higher sputtering rate. Based on the results of this paper, it is advised that the limiting location should not be on the helicon window and that the helicon window should be cooled to keep impurity production low and keep it on the low end of any chemical erosion regimes (if any).

To compare the experimental deposition on the target, the flux to the target needs to be converted to a fluence. The time taken for the startup of the plasma pulse is 0.07 s, and the remainder of the 0.5 s pulse is assumed to be at a steady-state. Eight plasma pulses yielded the deposition and XPS profile in Figs. 4 and 5, respectively. This puts the exposure at 0.568 s of starting density and 3.432 s of steady-state density. The density of the amorphous layer is assumed to be $2.35 g/cm3$⁠, and the monolayer thickness is assumed to be $1.15×10−10 m$⁠.^33,34 This gives a total impurity fluence of $1.7×1022 atoms/m2$⁠. From the XPS profiles, Al accounted for 35%, bringing the Al impurity flux to the target to $5.8×1021 atoms/m2$⁠. To compare to this, the impurity flux from the GITR model at the target is taken. The flux is taken at the center 0.012 m diameter of the simulation geometry. The fluxes and fluences (using the starting density timing and the steady-state density timing) from the four modeled cases are given in Table I. Using a combination of a low-density and high-density case gives a difference from the experimental range of 172 to 13 times lower. To further investigate the cause for this difference, the role of O is investigated.

Oxygen is seen throughout the deposition thickness from the XPS results, and the role of O within the device has not been explored yet. Some oxidation of the target surface will always occur and is typically within the first 10 nm. The layer thickness is much thicker than this, indicating that O is present throughout the pulse. The source of O is from water vapor as measured with residual gas energy analyzer on Proto-MPEX near Z = 0.5 m which has a partial pressure of 30% of the $(6.67×10−4$ Pa base pressure. Proto-MPEX does not employ a baking operation of the vacuum vessel to outgas the surfaces. Oxygen being heavier than deuterium will increase the sputtering yield and lead to an increase in impurity generation. To model this, the RF sheath potentials were used with F-TRIDYN to get the sputtering yields for O onto Al. Figure 19 gives the sputtering yields used for the O onto Al as a function of the incoming ion impact angle and energy. Additional GITR simulations were run for the four cases to get the Al erosion, deposition, and impurity flux at the target from O ion impact. The background plasma was left as a pure D plasma. The flux and fluence for Al ion impurities at the target due to the O are given in Table II. Pure O gives a difference of 38–0.9 times lower than the experiment. Pure O is not reasonable as it will be a fraction of the D fluence; however, up to 10% could be possible within a Proto-MPEX due to poor vacuum quality. The thick layer on the helicon window is also a source of O from the plasma pulse that will contribute to increased O in the plasma. Using a combination of 90% D and 10% O yields a range of 127–5.6 times lower than the experiment. The magnetized electron sheath cases provide the upper bounds which provide the closest comparison to the experiments. The magnetized electron sheath condition is thus what is occurring at the helicon window within Proto-MPEX.

The discrepancy from the model to the experiment could come from the measured electron density and temperature profiles near the helicon being extrapolated to the target, the pure crystalline Al surface assumption for the sputtering yields, the assumptions in the density of the impurity layer thickness, or from not including the trace impurities observed on the target on the sputtering yields. The assumed SiO₂ sputtering rate from the XPS is also only valid for the targets when the $Si3N4$ helicon window was installed. Up to a factor of 2× was found in the sputtering rates for $Al2O3$ from SiO₂³⁵ which would reduce the measured thickness and the number of monolayers moving the model to further agreement with the experiment. Additionally, the modeled GITR geometry has a fixed radius of 0.06 m which is not true for Proto-MPEX, and some of the particles were lost to the wall at points where the field flares out. Figure 20 shows the final location of the particles in the high-density magnetized GITR simulation geometry. Blue particles have been deposited on a surface and orange have not been deposited. Less than 500 particles remain undeposited in all simulations. The particles are predominantly on the helicon window, the dump plate, and the target, but some reside before the dump plate and between the helicon and the target at regions where the magnetic field flairs out and in reality are larger than the 0.06 m geometry used in the simulation, allowing for additional impurities to make their way to the target. However, this simpler geometry case still gives the trend of the radial impurity velocity. These results suggest that decreasing the oxygen content in MPEX may be beneficial to reducing the impurities at the target especially O from the helicon window.

In Ref. 18 the helicon region was modeled with GITR and the magnetized cases showed the Al impurity flux was mainly at the edge plasma. Due to the plasma density and temperature gradient in the radial impurity convection velocity, the particles should move toward the core of the plasma.³⁶ This is seen when looking at the profiles at the target in the current study. The current study was for a helicon-only plasma with a hollow temperature profile but with higher core temperature impurities will be driven to the plasma edge.

A radial scan was performed with the MAPP W targets, while the $Si3N4$ helicon window was installed to collect exposures at the on-axis, 1 cm off-axis, and 2 cm off-axis locations to experimentally investigate the radial ion locations. Figure 21 gives the normalized experimental results for the three axial locations along with the normalized GITR axial Al impurity profile for the high-density magnetized case at 1 cm in front of the target. The profile shows good agreement in the falloff length and indicates that the extended GITR geometry accurately captures the radial impurity ion convection velocity. The Al and Si are comparable in mass (27 vs 28 amu) and ionization energies (6 vs 8 eV), so they are compared here because they will be transported similarly. The impurity convection velocity arises from a diamagnetic drift and the friction force in the Fokker–Planck equation. The following equation gives the classical impurity drift velocity, $v→CLs$⁠:

where s denotes the impurity species and p is the plasma species.^36,37D_CL is the classical diffusion coefficient, q is the charge of the respective species, n is the density, T_e is the electron temperature, m is the mass, and m_sp is the reduced mass $msmp/(ms+mp)$⁠. The inward radial impurity drift will be similar for Al and Si due to their similar mass, charge states, and the plasma conditions being similar. The primary factors which will change the impurity movement are the density gradient, the temperature gradients, and the charge of the impurity species.

To explore the temperature screening on Proto-MPEX and to reduce the impurities deposited on the target, two additional experiments were performed with electron cyclotron heating (ECH), while the AlN helicon window was installed. The ECH launcher is located at Z = 3.25 m and launched 50 kW at 28 GHz via electron-Bernstein wave (EBW).³⁸ During the ECH the electron temperature increases from 2.5 to 8 eV and the electron density decreases to $2×1019 m−3$⁠. The second exposure included 3% krypton gas seeding. The Kr concentration was measured with partial pressures when pre-mixing the fueling gas. The ECH was on for 100 ms for each shot in both exposures from 4.5 to 4.6 s. The ECH exposure with a MAPP W target was exposed to 9 plasma pulses with an average D ion flux of $2.57×1023 m−2 s−1$ and a total fluence of $2.5×1024 m−2$⁠. The ECH exposure including Kr gas was exposed to 18 plasma pulses with an average ion flux of $1×1023 m−2 s−1$ and a total fluence of $1×1024 m−2$⁠. The XPS depth profiles for the two exposures are shown in Fig. 22. The depth profiles show a reduction in impurity layer thickness when ECH is introduced and the Kr addition further reduces the impurities on the surface. It is hypothesized that the ECH enhances the temperature gradient in the classical impurity drift velocity leading to impurities being screened from making their way to the plasma core. A 44% reduction in the impurities on the target for the 100 ms ECH pulse can be increased by increasing the pulse length and the input power. The higher power will also lead to additional sputtering of impurities that do make their way to the target. Kr with ECH reduces the impurity thickness to 7% of the original thickness which comes from the heavier Kr atoms sputtering the deposited material that is on the target. This heavier element will sputter more of the helicon window but the ECH being closer to the target gives the Kr more energy to sputtering the target than the window.

Experiments performed on Proto-MPEX have shown that both the AlN and $Si3N4$ helicon window are being eroded and the impurities from the window are transported to the target region. The Al impurity flux is reduced when ECH and Kr gas is added due to the increase in energy of the particles at the target to sputter the deposited material. The pattern on the helicon window was clean beneath the helicon antenna location and had depositions between the antenna straps. The composition of the deposition on the helicon window and the exposed W targets was found to be mainly aluminum oxide where the O is from water vapor within the device. MPEX will employ baking of the vacuum vessel to reduce O and make use of steady-state ECH and ion cyclotron heating to screen impurities from the plasma core.

A COMSOL simulation with a sheath equivalent dielectric layer was performed to obtain the rectified RF sheath potential on the helicon window. Four cases were modeled to match the experimental plasma pulse conditions at the startup and the steady-state times and for the bounds of the sheath model at the unmagnetized and magnetized electron boundary limit. GITR simulations tracking the eroded particles through Proto-MPEX matching, within reason, the deposition pattern on the helicon window with the area under the helicon antenna being eroded and depositions being between the helicon antenna area. The GITR model also showed a peaked on axis impurity profile at the target region and the role of O in the sputtering was investigated. Oxygen has a larger sputtering yield than deuterium, leading to a high portion of the generated impurity flux, even as a small fraction of the plasma species. The model slightly underpredicts the impurity accumulation rates at the target which could be due to the measured electron density and temperature profiles being measured near the helicon region which could be higher under the helicon antenna leading to a higher impurity generation in the experiment than what is captured by the model. The trace impurities are also ignored within this work which will increase the impurity generation rates and lead to the model better matching the experimentally observed rate. Future work should include the trace impurities and their contribution to the total impurity generation, an additional verification measurement of the impurity layer thickness, and the inclusion of the axial temperature profiles to the models to reflect the role of ECH. Experiments with steady-state ECH should also be performed to measure the temperature screening of impurities to the target.

This manuscript was authored by UT-Battelle, LLC under Contract No. DE-AC05–00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/ downloads/doe-public-access-plan).

The authors would like to thank the Proto-MPEX team at ORNL for their continued assistance in getting experimental data.

The data that support the findings of this study are available from the corresponding author upon reasonable request.