Sputtering of Mo and Ag with xenon ions from a radio-frequency ion thruster

Today, electric propulsion (EP) is part of nearly every space mission. In particular, if long mission duration and comparably low propellant consumption go hand in hand—which is primarily the case for telecommunication satellites—electric thrusters offer inherent advantages over their chemical pendants. Those advantages are based on a comparably very high specific impulse I_sp. Additionally, EP offers high scalability of thrust levels, enabling multiple maneuvers with only one type of thruster [e.g., electric orbit raising (EOR) followed by station-keeping (SK)]. Detailed information on EP (especially on the two major concepts of Hall and ion thrusters) can be found in Refs. 1 and 2, respectively.

To ensure long mission duration, not only propellant consumption plays a role. Erosion and degradation of thruster components often define the lifetime of a thruster. In particular, gridded ion thrusters such as radio-frequency (RF) ion thrusters (RITs) suffer from severe degradation of the extraction grid system.^3,4 In the context of EP, nearly every degrading effect is caused by material sputtering due to high-energetic ion bombardment. On a larger scale, those sputtering effects are also present due to ion beam (plume) divergence. Most prominently, EP may damage solar panels or other satellite components in the vicinity of the thrusters. To put it crudely, the effects of contamination in the context of chemical propulsion⁵ are similar to those of sputtering in the context of electric propulsion.

During the EP mission design, those sputtering effects are taken into account using special frameworks such as ESA’s SPIS (Spacecraft Plasma Interaction System). In theory, the above-mentioned contamination effects can be studied as functions of the points of thruster operation, thrust vector, geometry of thruster, satellite, and thruster suspension, etc., using those frameworks. There exist some (mostly crude) models, but there is a huge void concerning experimentally reliable data of typical and non-typical sputter phenomena, which are subject to multiple uncertainties.⁶ 

With this work, we want to emphasize the need of data in the context of ion sputtering for space applications. In this paper, we show an experimental approach to obtain these data using fairly simple experiments and non-exclusive off-the-shelf equipment. We hope that we can reach out to many more groups and sensitize the EP community to rid the void. Our ultimate goal is to build up a database available to the public with exactly the experimental data needed for (analytical, empirical, and heuristic) model validation.

To exemplarily show our proposed strategy, we chose molybdenum and silver as sputtering targets as they are found in thruster and satellite fabrication, respectively, and, hence, are prone to ion impingement on spacecraft equipped with EP.⁷ We first measured the spatial ion beam distribution of a typical RF ion thruster (RIT-10) and modeled them using an analytical approach. Finally, spatially resolved deposition rates of the chosen sputtering materials with respect to angle of incidence and (primary) energy of the ion beam are obtained using a quartz crystal microbalance (QCM). Another common method for the direct analysis of sputter yield data is a weight loss approach.⁸ It was not chosen for this study as it does not give information about the ejection angle of sputtered particles, which, in turn, is crucial regarding the redeposition of the material on a spacecraft. In the post-process, we validated the results with the established sputtering software framework SDTrimSP; see Sec. IV B.

An ion impinging on a surface, subsequently called projectile, potentially transfers its energy and momentum in such a way that an atom is ejected from the target. This process is called sputtering, in contrast to pure ion implantation or backscattering. A projectile undergoes statistical collisions and experiences an energy loss per traveled unit distance. This loss is called stopping power and originates from nuclear (elastic) or electronic (elastic or inelastic) collisions. For low ion energies, nuclear stopping dominates and can be determined with an elastic binary collision approximation (BCA). Particles are hereby treated as hard spheres with Bohr radii, undergoing different binary collisions depending on the mass, energy, and collision cross section.⁹ There are various integration methods based on classical scattering theory to evaluate the scattering angle of atoms after a collision, specified in Sec. IV. Experiments have so far validated computational first order approximations treating polycrystalline materials as amorphous.¹⁰ The energy of sputtered atoms depends on the surface binding energy (SBE), which, in turn, influences the ejection angle.⁹ Considering the total sputtering yield, maxima are generally to be expected for angles of projectile incidence between 50° and 80°.¹⁰ 

A QCM makes use of a frequency shift that an oscillator undergoes when a mass is deposited onto its surface. The quartz crystal disk is, therefore, excited into thickness shear mode vibrations by using an external oscillator at a frequency of several MHz. Deposition of sputtered material onto the crystal surface increases the mass of the oscillating system and, hence, lowers its frequency. Knowing a material’s density ρ (g/cm³) and acoustic impedance z (10⁵ g cm⁻² s⁻¹) allows for the deduction of a deposition rate over time.^11,12 Equation (1) after Williams et al.¹³ allows for the conversion of the measured deposition rate R(α, ϕ) (g/s) to a differential sputtering yield γ(α, ϕ) (atoms/ion/steradian), making use of the QCM distance to the target (radius r_QCM) and its position on the hemisphere (polar angle α and azimuthal angle ϕ),

where N_A is Avogadro’s constant, q is the elementary charge, M is the molar mass of the sputtered atoms, J (A) is the ion beam current, and A_s is the sensor area of the QCM. Equation (1) is applicable for certain geometrical setups only, which are not present in this study, but it will later be used to deduce a theoretical deposition rate from numerical differential sputtering yield values.

The electric RF ion thruster (RIT-10, 37 extraction apertures), developed and built at Giessen University, Germany, was set up and operated following conventions and empirical data gained by the respective work group.^2,14 The radio-frequency generator (RFG) and the negative and positive high voltages (NHV and PHV) were operated, as stated in Table I. The performance mapping delivered the expected 1/x curve characteristic of an ion thruster when adjusting the RFG power over the mass flow. Grid losses increased noticeably above mass flow rates larger than 1 SCCM. Furthermore, the findings support the usage of a mass flow above 0.6 SCCM to minimize the input power, which led to choosing the operational point of 0.8 SCCM. Being equipped with a rotary vane pump, a roots pump, and a turbo pump, the vacuum chamber is evacuated to 2 × 10⁻⁶ mbar. The pressure rises to about 6 × 10⁻⁵–7 × 10⁻⁵ mbar in the case of a sustained plasma and effective ion extraction (see Table I). This corresponds to a pumping speed of around 600 l/s.

Grids of the RIT-10 used in this study are made of titanium (Ti), while alternative materials are molybdenum (Mo) or carbon composites. Grid erosion caused by charge-exchange (CEX) ions has to be taken into account as soon as it contributes a significant amount of atoms to the extracted beam, as this potentially alters the behavior of the beam. For the present thruster, no critical fraction of charge-exchange ions has been observed so far when simulating comparable operational points.¹⁵ Small currents, a low mass flow, and a reasonable background pressure in the 10⁻⁵ mbar range support this behavior.

In order to determine the ion beam current during experiments, extracted currents and losses were measured directly on the thruster grids. The evaluation of the grid currents was done with Octave by interpolating the data over time and subtracting the current on the accelerator and the ground grid from the one given by the screen grid. The resulting beam current was averaged, excluding major beam events (peaks in the diagram) by masking the data (see Fig. 1). Grid losses lie around 7%, but thruster performance was stable and the standard deviation less than 0.1 mA for total beam currents between 7 and 9 mA.

A beam analysis with a Faraday cup (Model FC-71A, Kimball Physics, Inc.) was carried out to verify the horizontal beam profile. A rotational stage with extended mount (a lower swivel arm of the QCM setup, Fig. 2) was used, resulting in an absolute distance of 56.6 cm of the Faraday cup to the thruster grid in perpendicular position. According to this setup, horizontal 180°-scans were conducted along the corresponding radius of 26 cm. Test conditions were varied to gain knowledge about the influence of an active neutralizer (NTR) in contrast to a pure ion beam.

Faraday cup scans of the ion beam with and without an active NTR showed a Gaussian current distribution over the radial distance from the beam center line. The maximum of the distribution is slightly eccentrical for both cases by less than two degrees. The Gauss fit of the distribution without the NTR in use gives a standard deviation (σ) of 3.75° (1.5 keV) and 3.42° (1.8 keV), which results in a divergence angle of 9.18° and 8.37°, respectively, for a 95% (2.4477 σ) interval (see Fig. 3). The NTR (imitated by two conventional incandescent lamps centered below the extraction grid) influenced the beam current inconsistently and exhibited a slightly less uniform beam profile, which is why its use was dismissed during sputtering experiments. Furthermore, a dip in the beam current measured at the thruster grids was observed parallel to the movement of the Faraday cup through the beam profile.

Lining of the stainless steel chamber walls was realized with graphite plates (Sigraflex^® expanded graphite) using adhesive Kapton^® tape (polyimide tape with silicone adhesive). Both materials promise to decrease background contamination during firing of the ion thruster considering their overall lower sputtering yield compared to stainless steel and aluminum.

Sputtered particles were detected with a QCM sensor head (Intellemetrics Global Ltd.) scanning across a hemisphere over a target’s surface. This was realized with a combination of two pivoting arms made of aluminum, each centered on a rotational stage (two-phase stepper motor PRS-110 by PI miCos at the base position and URS50-BPPV6 by Newport at the upper position), both controlled by using one motion controller (ESP301 by Newport). The bi-directional repeatability of incremental motion of the base stage (azimuthal angle, QCM toward and away from the thruster) was given with 0.01° and observed accordingly. The top stage that addresses the polar angles (QCM up and down) did not exceed an accuracy of ±1° for its movement but was reset prior to each measurement series. The QCM setup was covered with Kapton tape where feasible in order to decrease background sputtering. A platinum resistor (Pt-1000) was used for temperature observation. It was glued to the QCM mounting bracket, which is in direct contact with the crystal plate, with a two component epoxy thermal conducting glue.

QCM measurements were conducted with sheets of molybdenum (Mo, 99.95% purity, thickness 0.5 mm) and silver (Ag, 99.99% purity, 1 mm), each with the size of 10 by 10 cm. This covers ∼95% of the beam profile (see Sec. V B 1) in a perpendicular target position. However, target materials were positioned under an angle of ion incidence of 60° ± 2° (defined as off-normal) as this setting promises elevated sputtering yields^7,10 and reduced risk of thruster backsputtering contamination. Due to the Gaussian shape of the beam profile, the reduced area of the tilted target (2.5 cm along the radial profile of the beam) did not significantly decrease the percentage of incident ions from the initial beam (less than −10%). The target was located in front of the thruster (see Fig. 4), centered at a distance of 30.6 cm right above the rotational center of the QCM setup. The QCM sensor was situated at a radial distance r_QCM of 25.1 cm from the target center. A parking position was arranged to shield the sensor from the direct ion beam and target sputter material when necessary (Fig. 4). Both targets were measured under the 1.8 keV ion energy operational point, while silver was additionally scanned at 1.5 keV (Table I). Operational points were named after the estimated peak ion energy, which corresponds to the applied PHV.¹⁶ 

The QCM input parameters of the deposited material’s density and acoustic impedance for Mo and Ag were 10.2 and 10.5 g/cm³ along with 34.3 × 10⁵ and 16.7 × 10⁵ cm² g⁻¹ s⁻¹, respectively (according to the manual of the Intellemetrics QCM monitor IL150). The QCM monitor makes use of these values to internally convert the detected frequency shift of the quartz crystal to an accumulated thickness of material per second. The crystal was a plano-convex quartz disk, 14 mm in diameter and 3 mm thick, which is excited at 6 MHz.

Angles of particle ejection (azimuthal and polar angles relative to the target normal) were adapted to the convention used in the SDTrimSP computations (Sec. IV B). The forward (40°), perpendicular (90°), and backward (150°) azimuthal sputtering regions for a range of respective polar angles under 10° steps were chosen to be examined (compare Fig. 5). Polar angles were sufficiently determined ranging from 0 (perpendicular ejection) up to just below 90° (shallow ejection) as the ion beam was approximated as symmetrically identical on the upper and lower half of the target and so was the created sputtering profile. Due to excessive heating of the sensor, the above-mentioned points were measured for periods of several seconds only until a sufficient accumulated mass was observed. The corresponding movements were conducted manually via the GUI of the QCM monitor. Subsequently, the increasing absolute thickness signal was plotted over time and a deposition rate was derived from its slope.

Additionally, silicon wafer substrates (5 × 5 mm²) for later element analysis of the deposited sputter material by Rutherford Backscattering Spectrometry (RBS) were glued to the inside of the upper arm of the QCM setup with double sided aluminum adhesive tape. The first series of five substrates was installed during pre-testing and was exposed to the ion beam and potentially sputtered thruster and chamber material under various positions for ∼40 h. Main waiting times elapsed with the QCM setup in the parking position (compare Fig. 4). The second series, installed with the Mo target, was sampled after 6 h and 20 min and the third series along with Ag after 2 h and 50 min of thruster activity, after performed QCM scan routines in each case.

To estimate the radial current density (J in A/m²) for a 2D Gaussian beam distribution over radius r at a given distance z from the thruster exit plane, the following expression approximated for an ion beam expanding from a point source was used:¹⁴ 

with

Here, I_b represents the beam current extracted from the thruster, while σ gives the according standard deviation. The divergence angle in degrees is noted as β. The current density calculation is based on Eq. (2). The standard deviation value (2.4477) is chosen according to 95% of a 2D Gaussian profile projected onto a plane perpendicular to the direction of propagation. A cylindrical integration gives the total current J in A over radius r [Eq. (4)]. The current up to the boundaries implied by the 95% interval is obtained by setting R = r − 2.4477σ. The calculation of the current entering the Faraday cup with an orifice of diameter d requires an integration limit of R = r − d/2,

The generation of CEX ions, along with inelastic collisions of beam ions with neutrals from the background gas present in a vacuum chamber, differs considerably compared to space conditions but is, however, not considered here for the sake of simplicity. For calculations, I_b and β were approximated by empirical values of 8.5 × 10⁻³ A and 9°, respectively (Sec. III A). The distance between the thruster and the target and the Faraday cup allow for a far-field approximation¹⁷ as they are ten times the grid diameter d (d ≈ 3.1 cm for the RIT-10/37) or more apart.

The simulation program SDTrimSP (SD = static–dynamic; SP = sequential and parallel processing) was initialized to run sputtering simulations. The program was published by Mutzke et al.¹⁸ from the Max-Planck-Institute of Plasma Physics, Germany, and is based on the Trim code.¹⁹ Theoretical foundations have been well elaborated in Eckstein’s adaptations of analytical sputtering yields for computer simulations.⁹ Models and approximations were chosen based on the findings of Hofsäss et al.²⁰ and consistently applied for all the SDTrimSP simulations:

• Fully dynamic (Monte Carlo) simulations

• Ion built-up considered

• KrC interaction potential

• Gauss–Legendre interaction method

• Surface binding model: isbv = 1

• Bulk binding energy: 0

The beam current density (3.45 A/m²) and the resulting predicted incident ion fluence onto a target area (0.22 × 10¹⁶ atoms/cm²) were chosen according to the values estimated at the center of the target with the beam current model in Octave. Simulations were run for molybdenum (thickness 150 Å) and silver (170 Å), with xenon ion energies of 1.5 and 1.8 keV, as well as angles of incidence of 0° and 60° for each material. It should be noted that the ion incident angle in SDTrimSP is defined as off-normal and is not equal to the azimuthal angles assigned to sputtered particles. The target thickness values were set slightly larger than implantation depths estimated with the program SRIM (Stopping and Range of Ions in Matter by Ziegler and Biersack), in order to avoid transmitted projectiles, as realistic thickness values significantly exceed the simulated thickness for the considered applications.

During test runs without a sputtering target, the QCM temperature was stable around 19 °C over a long period including sensor positions away from or facing toward the thruster for different ion energies. Test scans with the target material exhibited excessive heating of the water-cooled QCM sensor, which is why holding times were kept to a minimum. This is especially true for the forward region, where the QCM temperature eventually exceeded the limit of the Pt-element (105 °C). Due to limitations of the temperature sensor, the target could only be observed for some minutes after thruster ignition, as it heated up from 30 to 105 °C in less than 4 min. This observation is not crucial, though, as polycrystalline and amorphous materials show mostly no dependence of their sputtering yield on the bulk temperature.¹⁰ 

The deposition rates (Fig. 7) generally decrease when moving from perpendicular, i.e., small polar angle, to shallow angles of ejection. Few outliers in the perpendicular azimuthal direction possibly indicate uncertainties of the measurements, which were conducted only once due to time restrictions. In the case of silver, the perpendicular azimuthal region exhibits clearly higher deposition rates than the backward and forward regions, while exceeding computational results by up to a factor of 4. For molybdenum, the effect is less pronounced as backward and forward sputter rates exhibit roughly the same trend. As the QCM temperature did not rise as quickly in the backsputtering region, two measurements of Ag under the same conditions were feasible and showed a reasonable repeatability (Fig. 6).

The RBS analysis of coated wafer substrates revealed the composition of the deposited films, but as the setup with the wafer substrates was moved during deposition, no dependence on the spatial position could be deduced from these findings. The film thickness t_RBS in 10¹⁵ atoms/cm² is derived from the fits applied in the RBS analysis software SIMNRA²¹ and calculated to an actual thickness t_calc in nm from the atomic density in atoms/cm³. Mo and Ag deposits exhibited a noticeable thickness close to 10¹⁸ atoms/cm², which corresponds to a few 100 nm. Table II lists the range of film thickness with an exemplarily assumption of deposited and oxidized Mo and Ag (MoO₂ and Ag₂O, respectively). The oxidized films were exemplary fitted in two layers with varying oxygen content. Overall, four samples were measured in each case.

A wafer subjected only to chamber background sputtering and xenon ion beam irradiation during pre-testing revealed equally small amounts of xenon and titanium, smaller than 1% of the atoms deposited during sputtering. An exemplary conversion of the 1.3 nm titanium deposit with an atomic radius of 140 pm gives around ten atomic layers.

The beam current model [Eq. (2)] gives a linearly decreasing beam current density with a value of 3.45 A/m² at the center line and z = 30.6 cm, representing the position of the target center. The beam distribution in this plane has σ of 1.98 cm, resulting in an area of 94.09 cm² impinged by 95% of the beam current.

A comparison of the current calculated at z = 56.6 cm (position and orifice size of the Faraday cup, initial ion beam current 11.7 × 10⁻³ A, and 1.5 keV thruster setting) to the current measured with the Faraday cup under said conditions allows for an estimation of the acceptance of its orifice. The computational beam current entering the sensor is 27.2 × 10⁻⁶ A, while the experimental value measured in the center line of the beam gives 2.99 × 10⁻⁶ A, resulting in a low transmission of ∼11%.

The angular distribution of ejected target atoms was divided into counts per solid angle with a two-dimensional binning function across the surface of a quarter-sphere (radius r_QCM = 25.1 cm, symmetrically identical to the second half of the hemisphere). The corresponding polar contour plots are available in the supplementary material (Fig. S1). All case studies tend toward very low sputtering along the contour of the polar plot, which equals an ejection under large and hence shallow polar angles. The distribution is shifted to the forward sputtering region for an angle of incidence of 60° in each case.

In order to compare simulation results to experimental data, the binning process described above was further used to calculate a differential sputtering yield per steradian from the computational data. Therefore, the counts within bins that correspond to the experimentally measured points on the hemisphere (compare Fig. 5; the QCM sensor head equals 4.44 × 10⁻⁴ sr) were used. The differential yield, in turn, was used to deduce the deposition rate [comparable to R in Eq. (1)] according to the experimental setup, using the impinging ion current at the center of the target (3.05 mA at 1.5 keV and 3.45 mA at 1.8 keV). The resulting angular dependences of the deposition rates are plotted in Fig. 7 in comparison to experimental data. As the backsputtering region could not be measured up to a perpendicular polar angle (0°) in the experiment, the computational data are left out accordingly.

Two stable operational points of the RIT-10/37 were iteratively determined, and sputtering measurements were successfully conducted with a silver target and a molybdenum target. Experimental and computational results lie in the same order of magnitude, following similar trends. However, the RBS analysis of the sputter deposit revealed the formation of oxides, and excessive heating of the QCM sensor head was observed.

Experimental data follow the trend of computational SDTrimSP results well, with rates decreasing toward large polar angles. Overall, higher experimental values have been observed for perpendicular azimuthal regions. For non-normal ion incidence (60° in this study), an over-cosine distribution tilted in the forward direction is expected, with sputter rates increasing from backward, to perpendicular, to forward ejection.²² Simulations also give an estimate of a sputtering plume shifted toward forward sputtering in comparison to perpendicular ion incidence (see the supplementary material, Fig. S1). Williams et al.,¹³ however, found that for Mo, this behavior reverts to a nearly symmetric distribution around the target normal.

Contradictions of differing values from literature, simulation, and experiment demand for a rigorous evaluation of contributing effects. Resputtering due to the direct ion beam, which is favored in the forward sputtering region, is one possible reason that may have artificially lowered the measured deposition rates in the forward region. Additionally, a decrease in the effective sticking coefficient of sputtered metallic particles can be caused by thermal desorption, which is facilitated if the substrate (here, QCM crystal) is heated due to incoming energy flux.²³ This demands for a more sophisticated cooling system. Moreover, the effective sticking coefficient is lowered when resputtering or reflection occurs, which, in turn, is favored for high-energy particles.

The energy of sputtered atoms is based on various factors, but mainly on their angle of ejection for incoming ion energies above 1 keV. In general, off-normal ejection results in higher mean energy, while peak ejection energies rise only slightly compared to perpendicular or shallow ejection.⁹ Furthermore, the heavy mass of Xe ions, in general, favors the generation of secondary knock-on atoms, while the shallow ion incidence used in this study (60°), in turn, promotes high-energy primary knock-on atoms.⁹ Hence, the effective sticking coefficient is subject to change depending on the region of detection and to the experimental setup, in general.

In the context of SDTrimSP, care should be taken regarding the assumption of the surface binding energy (SBE) as the program does not take into account an initial surface roughness or a change in SBE during the sputtering process. In terms of experimental investigations, an analysis of the target surface before and after experiments could reveal the formation of patterns, which can, in turn, favor preferred sputtering orientations.²⁴ In order to achieve a good comparison of experiment and simulations, a thorough knowledge of the ion beam impacting the target surface is crucial. Furthermore, the geometrical setup could be improved in the test chamber to create conditions that allow for a conversion of measured deposition rates to a sputtering yield per solid angle.^13,25

RBS analysis was intended for qualitative analysis of the sputter deposit to adapt the settings of the QCM monitor, which relies on an accurate preset film density and acoustic impedance. The findings of deposited oxides instead of pure sputtered metals indicate that measured deposition rates are presumably too low regarding an oxide film thickness. The formation of oxides revealed by RBS analysis was to be expected, as atomic oxygen formed by the RF-plasma inside the ground-based vacuum chamber favors this process.²⁶ 

Background chamber pressure in the low 10⁻⁵ mbar range was achieved during thruster activity, which is equivalent to comparable experiments found in the literature.^7,13 However, the performance of the facility could be further improved by reducing the base pressure in order to reduce the process of oxide formation. Beam events ceased to appear regularly after a longer run time of the thruster and can most probably be explained not only with discharge due to grid contamination but also with artifacts of the membrane in the xenon mass flow controller or insufficient grounding of the facility.

The ion beam characterization with a Faraday cup revealed the expected Gaussian beam profile, but the measured profile might be artificially narrowed by the acceptance angle of the Faraday cup, so the actual divergence angle could be slightly bigger than determined. Furthermore, a dip in grid currents observed during the Faraday cup scans emphasizes the invasiveness regarding beam properties when positioning setup parts in the beam. A non-invasive beam characterization would be needed in order to conduct simultaneous measurements during sputtering experiments.

This study addressed the need of the EP community to generate spatially resolved sputtering data, by implementing an experimental setup with readily available tools along with a computational validation using the program SDTrimSP. The targeted QCM-analysis was successfully conducted on molybdenum and silver, sputtered by the xenon ion beam of a RIT-10. The results seem reasonable as experiments and simulations lie in the same order of magnitude while exhibiting comparable trends over the analyzed spatial angles. Deviations, especially in the perpendicular azimuthal regions, demand future investigations. Moreover, a temperature calibration and sufficient cooling of the QCM sensor are needed with regard to the significant heating observed. The latter is likely to cause frequency instabilities and influence the sticking coefficient and is most commonly induced by bombarding ions when operating a QCM in a sputtering setup.²⁷ Therefore, not only the quantitative but also the energetic distribution of sputtered particles would be of interest here. Information on both is available in the program SDTrimSP.

The findings of oxides by RBS analysis demand an initial deposit characterization (composition, density, and acoustic impedance) in order to adapt the QCM monitor settings. Furthermore, the sampling procedure must be improved to rule out dependencies on sampling orders. A stable thruster operation was realized, creating a symmetric, roughly centered beam, ruling out dependencies in this matter.

In order to examine realistic scenarios for satellite applications, a more accurate knowledge of the ion beam distribution over a (tilted) target surface is needed, especially if the target itself alters the beam properties by inhibiting its free expansion. Another relevant factor is to understand the dependence of the aforementioned surface roughness on the sputter yield and distribution, particularly if the surface structure changes significantly over longer time scales. Optical coatings as they are used in solar cell technologies often have a special design in terms of surface patterning, which should be taken into account when testing or simulating the sputtering behavior of such surfaces.

Overall, the subject of sputtering in space applications is prone to multiple uncertainties,⁶ ranging from the approximations made in computational approaches, over experimental data acquisition, to the assessment of actual applications. We hope that this work emphasizes the need and advocates the research of experimental sputtering data for space applications by presenting the corresponding difficulties and possibilities in terms of experimental work and computational supplements.

See the supplementary material for the polar contour plots of the SDTrimSP angular sputtering distribution simulation (Sec. V B 2).

This work was conducted at DLR (German Aerospace Center), Institute of Aerodynamics and Flow Technology, Spacecraft Department in Göttingen, Germany. The construction of the sensor mount and preliminary positioning unit was developed at DLR as preparatory work by Jonas Schäfer. Advice for electronical matters was gratefully received from C. Geile at DLR. The RBS analysis of the wafer substrates was conducted at the University of Göttingen at the Institute of Nuclear and Particle Physics, supported by H. Hofsäss, who also provided a helpful introduction and kind support for SDTrimSP simulations.

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