Implementation of in situ diagnostics for sputter yield measurements in a focused plasma

Propulsion and Energy Forum July 25-27, 2016, Salt Lake City, UT 52nd AIAA/SAE/ASEE Joint Propulsion Conference AIAA 2016-4841 Implementation of in situ diagnostics for sputter yield measurements in a focused plasma Gary Z. Li, Christopher A. Dodson, and Richard E. Wirz University of California, Los Angeles, CA, 90024, USA Taylor S. Matlock The Aerospace Corporation, El Segundo, CA, 90245, USA Dan M. Goebel Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, 91109, USA The Plasma-interactions test facility at UCLA features a high current hollow cathode plasma source that delivers a cylindrical plasma through solenoids to a target downstream. The near-target plasma is focused to a spot size of 1.5 cm for localized sputtering. The target is then biased negative with respect to the plasma potential to provide energetic ion bombardment. To perform in situ sputter yield measurements, a quartz crystal microbalance and optical emission spectroscopy have been implemented. Preliminary results for an argon plasma on molybdenum sputtering experiment are presented to validate these diagnostics. The quartz crystal microbalance method along with the traditional weight loss method produce sputter yields that agree well with published results. Angular distributions and emission spectra for these test cases are also provided. N s Number of sputtered particles m Atomic weight, amu N i Number of incident ions J t Target current, A Y Averaged sputter yield, atoms/ion A s Sensor area, m 2 E Characteristic energy, ev E Ion energy, ev α Polar angle, Nomenclature I. Introduction Electric propulsion devices produce harsh plasma environments where a boundary material may be eroded by energetic ion bombardment. A primary failure mode of ion thrusters is the ion-induced erosion of the accelerator grids over long periods of operation. 1 Additionally, conventional Hall thrusters suffer from channel wall erosion, while magnetically shielded Hall thrusters experience pole piece erosion. 2, 3 In general, ion-induced sputtering is a life limiting factor in all plasma-facing materials. In order to understand and mitigate ion-induced sputtering, the UCLA Plasma interactions (Pi) test facility has been developed to investigate the behavior of advanced aerospace materials under plasma bombardment. 4 The erosion of a material is quantified by the sputter yield which indicates how many surface atoms are removed per incident ion. Many experiments have been performed to measure the sputter yields of plasma-facing materials, Ph.D. Candidate, Mechanical and Aerospace Engineering Department, UCLA Associate Professor, Mechanical and Aerospace Engineering Department, UCLA Member of Technical Staff, Propulsion Science Senior Scientist, Advanced Propulsion Group Copyright 2016 by the, Inc. All rights reserved. 1 of 13

but the spread of such measurements has remained large. This is primarily due to the variety and complexity of sputtering experiments. 5 Ion sources may be chosen between ion guns and plasma discharges, and a variety of diagnostics exist for measuring the quantity of sputtered particles, some of which include weight loss, 4, 6 optical emission spectroscopy (OES), 6 quartz crystal microbalance (QCM), 7 and Rutherford backscattering. 5 A plasma discharge was chosen for the Pi ion source in order to simulate the plasma conditions in the devices of interest, and also to generate sufficiently high current densities for achieving the desired ion fluence. In this experiment, a hollow cathode plasma source and cylindrical anode produce a partially ionized plasma that is transported downstream to a biased target via guide magnets. A detailed description of the plasma source operation metrics has been presented in a previous paper. 4 In order to perform sputter yield measurements over the widest range of sputtering conditions, we are in the process of implementing three distinct methods of measuring the sputter yield in the Pi facility: weight loss, QCM, and OES. This paper begins by describing the Pi facility experimental set-up, the near-target plasma, and the implementation of the diagnostics used for testing. The sputter yield measurement techniques are discussed and the limitations are presented. Finally, preliminary measurements for a low-energy argon on molybdenum sputtering experiment are presented and compared to published results from literature. II. Experimental Set-up A. Plasma-interactions Test Facility a) Cathode Anode Chamber Magnets Anode Gas Injector Chamber Magnets b) Sample Magnets Target Sample Magnets Anode Target Figure 1: The images above show a) the Pi facility set-up and b) the dashed region with the plasma on. The Pi facility consists of a 1.8 m diameter by 2.8 m long cylindrical vacuum chamber and a dual CT-10 pumping system that allows base pressures of 5 10 7 Torr with a combined pumping speed of 5000 L/s on argon. A 250 A lanthanum hexaboride (LaB 6 ) hollow cathode generates an argon plasma via electron bombardment and connects to a 30 cm cylindrical anode held at ground potential. An additional gas injector is located between the cathode and anode which has been shown to reduce the oscillations in the cathode plume and minimize the production of energetic ions which may sputter the cathode keeper face. 8, 9 The plasma is subsequently transported along the axes of two water-cooled, grounded solenoidal magnets to a biased target downstream to provide energetic ion bombardment. The plasma typically floats within an electron temperature (<10eV) of ground potential 4 and thus provides a stable reference potential for biasing the target negative to draw ion current. The target floats at approximately -40 V when 2 of 13

unbiased and can be biased down to -300 V using our current power supply. The Pi facility has been well-characterized in previous papers 4 to determine operating ranges for different mass flow, magnetic configuration, and discharge conditions. Gaps between the transport solenoids and target magnets allow for easy access to the plasma for probing. The hollow cathode creates a partially ionized argon and xenon plasma of moderate electron temperature (2 10 ev) and plasma density (10 16 10 19 m 3 ), allowing a wide selection of deliverable plasmas to the target. Since the goal of this facility is to study plasma-material interactions, the plasma near the target must be understood to correctly interpret sputter yields. 1. Focused Plasma Configuration Target Chamber Magnets (a) r z (b) Sample Magnets Figure 2: The images above show a) the near-target plasma and b) the magnetic field simulation at the target. As shown in Figure 2a, the current plasma configuration utilizes a magnetic field structure that funnels the plasma into a relatively narrow beam impinging upon the target. Figure 2b shows a magnetic field simulation of the near-target region using ANSYS Maxwell, indicating a lensing effect at the location of the target. This operating mode localizes the erosion region on the target with a beam spot size of around 1.5 cm. A small spot size was chosen to facilitate the interpretation of sputtering from a point source rather than analyze sputtering across the entire target. This simplication makes interpretation of quartz crystal microbalance data significantly easier, and allows an accurate angular distribution to be extracted. In order to maintain consistency during materials testing, we fix the near-target plasma conditions to the values shown in Table 1. The mass flow was divided between the cathode (15 sccm) and anode (12 sccm) gas injectors to produce a working pressure of 6 10 5 Torr, corrected for argon and measured at the roughing port located at the back of the chamber. Note that the measured working pressure will be lower than the neutral pressure near the plasma. The plasma discharge was operated in current-limited mode so the voltage would float to the equilibrium value for completing the circuit. The anode is grounded in this configuration so current is collected by every grounded component in the chamber. These conditions produced an average axial ion flux of 10 17 cm 2 s 1 located 5 mm in front of the target. These conditions are replicated to within 10% deviation for each experiment. Table 1: Near-target Plasma Conditions Parameter Typical Values Mass flow (total) 27 sccm Discharge Voltage 36 39 V Discharge Current 45 A Working Pressure 10 4 Torr Plasma density 10 18 m 3 Electron temperature 7 ev Ion energy to target 10 to 300 ev Ion flux to target 10 17 cm 2 s 1 Target area 7 cm 2 3 of 13

10 17 Ion Flux (ions s -1 cm -2 ) 2 1.5 1 Data Fit 0.5 (a) -1-0.5 0 0.5 1 Lateral Position, X (cm) (b) Figure 3: The plots above show a) ion flux contours and b) an ion flux profile across the target center-line for the plasma. The plasma is offset by 4 mm in the vertical direction relative to the target. In order to map the plasma structure, a planar Langmuir probe with a guard ring biased to -100 V and insulated backside was used to measure the incident ion flux in the ion saturation regime. The axial ion flux is measured in the target plane (X-Y) 5 mm in front of the target. The 2D contour is shown in Figure 3a while a lateral profile is shown in Figure 3b. The plasma at this location is azimuthally symmetric and has a roughly Gaussian profile. The full-width half-maximum (FWHM) of the plasma is roughly 1.5 cm. Compared to the typical target diameter of 5 cm, the plasma beam is small enough that it can be reasonably approximated as a point source. In previous tests, the plasma profile was closer to a flat-top which implied significant sputtering across the entire target. In the present case, the sputtering is concentrated at the center of the beam and decreases towards the edge of the target. In the ideal sputtering experiment, the ions would be normally incident, singly ionized, and monoenergetic. In a plasma, this is never perfectly the case as there will always be some spread in the incidence angles, charge states of the ions, and energy. Fortunately, the plasma sheath allows the ions to be accelerated perpendicular to the target which typically results in near-normal incidence angle. Doubly-ionized argon should be minimal in the delivered plasma beam since spectra of the argon plasma reveal no detectable Ar ++ transitions. An ExB probe is under development to verify the percentage of doubly-charged argon ions. Lastly, the energy spread of the ions at the target is typically much lower than the sheath voltage due to the biasing of the target. A retarding potential analyzer is being developed to characterize this energy spread in the ion energy distribution function. Even though the plasma conditions are roughly consistent from test to test, the boundaries and thermal conditions result in transients during each test. After biasing, the target also increases in temperature to about 500 C 700 C as measured by a thermocouple. As the system approaches a steady state over a period of a few hours, the discharge voltage steadily increases which increases the ion current delivered to the target over time. The variable ion current is monitored through a 10 Ω shunt resistor connected between the target and bias power supply. In the next section, the target current is combined with sputterant data to obtain sputter yields. III. Sputter Yield Diagnostics A suite of diagnostics is being developed to measure sputter yield using three established methods: by weight loss, by quartz crystal microbalance, and by optical emission spectroscopy. This allows cross-comparisons for each sputtering experiment, and enables measurements in sputtering regimes where one technique may fail. The current procedures for measurement by weight loss and QCM are described below, while the implementation of OES is briefly presented. A. Weight Loss The measurement of sputter yield by weight loss is often used in the sputtering community due to its relative simplicity. Before testing, the sample is weighed using a precision microbalance and examined for contamination using a Scanning Electron Microscope (SEM) and Energy-dispersive X-ray Microscopy (EDS). The day before testing, the sample is inserted into the biased sample mount and the chamber is pumped down to base pressure (< 5 10 6 4 of 13

Figure 4: Overview of the Pi facility diagnostics suite. Torr) to lower the partial pressures of background gases and outgas the sample. To begin the test, the hollow cathode is ignited, and the discharge and chamber magnet settings are adjusted to deliver a previously characterized plasma of known ion flux to the target. The target is then biased to bombard the target with plasma ions extracted through the sheath. The target current is monitored for irregularities and recorded throughout the entire test via the current shunt. After a pre-determined period of time, the target bias is removed and the plasma turned off. The sample is allowed to cool to prevent oxides from forming after exposure to atmosphere. At least one hour later, the chamber is returned to atmosphere and the target is removed and weighed. The mass loss and target current data provide sufficient information to calculate the sputter yield for the target material at a given ion energy. Assuming a material with singular composition, the total number of particles sputtered can be calculated as follows: M NA, (1) Ns = m where Ns is the total number of sputtered atoms, M is the mass loss in grams, m is the atomic weight of the target material in amu, and NA is Avogadro s constant equal to 6.02 1023 atoms/mol. For materials with multiple consituents, the atomic weight m can be replaced with an effective mass meff. The total ion fluence over the burn can be found by integrating the target current from beginning to removal of the target bias: Z 1 t1 Ni = Jt (t)dt, (2) e t0 where Ni is the total ion fluence, e is the elementary charge (1.6 10 19 C), t0 and t1 are the bias initiation and removal times respectively, and Jt (t) (A) is the time-varying target current. The ratio of these two terms gives the averaged total sputter yield in atoms per ion: Y = Ns. Ni (3) There are several assumptions that we make to produce this sputter yield value. We assume that no ion implantation occurs, a valid assumption considering the low ion energies involved and the verification using EDS post-testing. We approximate the target as flat such that all ions are assumed to be normally incident on the target. This assumption fails near the edges of the target where fringe fields exist,10 but the effect should be negligible due to the low ion flux at the edges of the target (see Figure 3b). Finally, we assume that plasma redeposition can be neglected due to low plasma density in front of the target. Previous experiments have corrected for redeposition by estimating the ionization mean free path and determining what fraction of sputtered atoms ionize within the sheath and redeposit 5 of 13

on the target. 4, 6 In order to avoid this complication, we adjust the plasma parameters to lower the plasma density in front of the target such that the ionization mean free path is much larger than the pre-sheath size (λ mfp L). B. Quartz Crystal Microbalance The QCM concept is based on the piezoelectric sensitivity of quartz crystals to added mass, which for our application can be used to measure the incoming rate of sputtered particles. As material is deposited on the crystal, the resonance frequency decreases linearly in proportion to the deposited mass as discovered by Sauerbrey in 1959 11 An INFICON front load single sensor was employed in the chamber and routed to an STM-2 deposition monitor. The AT-cut quartz crystal in the sensor has a resonance frequency of 6 MHz and an exposed area of 0.55 cm 2 as seen in Figure 5a. We assume all incident particles are adsorbed onto the silver-coated crystal surface (a sticking coefficient of order unity). 12 The crystal frequency can be monitored until about 5.7 MHz before frequency mode hops rendering the crystal un-usable. To ensure stable operation, the crystal is water-cooled to prevent thermal drifts in the resonance frequency and protected by a shutter to prevent thermal transients due to hot incident particles. The QCM crystal holder is mounted to a rotary stage at a distance of 25.4 ± 0.2 cm from the center of rotation, which is aligned to the target center to enable measurements at varying polar angles in the lateral plane. The coordinate system for the QCM measurements is shown in Figure 5b with the polar angle, α, being defined from the target normal (z-axis) in the y-z plane. For the case of normal incidence, we assume the angular distribution of sputtered particles to be azimuthally symmetric so the differential sputter yield is a function of polar angle only. Since ions are accelerated perpendicular to the plasma sheath, this assumption holds true in all regions except the edges where fringe fields result in a non-normal incidence angle. 10 By assuming azimuthal symmetry of the sputtered particle distribution, we can effectively measure the differential sputter rate, R(α), over the full hemisphere above the target (shown in Figure 5b). The quantity R(α) has units of ng cm 2 s 1 and is defined as the mass flux to the QCM sensor. Shutter Crystal Target (a) QCM sensor (b) Coordinate System Figure 5: QCM apparatus with definition of geometry. To perform the polar scan, we rotate the QCM in 5 increments from α = 90 to 30, beyond which the QCM becomes blocked by the transport solenoids. The QCM vibrates when moving and causes oscillations in the data, so we record for 1 minute at each location and neglect the first 5 and last 5 seconds of each measurement. The differential sputter rate R(α) recorded at each polar angle is averaged over the 50 seconds of measurement. The differential sputter rate integrated over the hemisphere with a radius of the QCM-target distance gives the total sputter rate, Γ s (atoms/s), of the target. Provided the total ion flux, Γ i (ions/s), we can calculate the instantaneous sputter yield of the target material, Y (atoms/ion) as follows: Y = Γ s Γ i. (4) 6 of 13

An equivalent method of calculating this same value involves rearrangement of the right-hand-side in terms 7, 13 of a differential sputter yield. Similar to the methodology of Yalin, the differential sputter yield in units of atoms/ion/steradian can be defined as follows: ( ) ( ) ( ) dy(α) N A Jt As = R(α)A s / / dω m e rqcm }{{}}{{}}{{}, (5) sputtered atoms incident on QCM ions incident on target solid angle subtended by QCM where A s is the QCM sensor area (0.55 cm 2 ), m (amu) is the atomic weight of sputtered particles, N A (6.02 10 23 atoms mol 1 ) is Avogadro s constant, J t (C/s) is the target current, and r qcm (25.4 ± 0.2 cm) is the distance from the QCM to the target center. To calculate the total sputter yield, the differential sputter yield is integrated over all solid angles in the hemisphere in front of the target: Y = 2π π/2 0 0 dy(α) sin(α) dα dφ (6) dω In order to perform the integration in Equation 6, the differential sputter yield must be extrapolated to the interior polar angles where data cannot be obtained. This can be done by fitting the data to a physical sputtering profile. The sputtering profile is known to have a cosine shape for high energies (> 1 kev). However, for low energy sputtering (< 1 kev), the angular profile tends be more diffuse with a relative maximum near 45. A profile based on sputtering theory is derived by Zhang and Zhang 14 and modified by Yalin 13 to provide an accurate fit for QCM data. Yalin showed that the original equation could be expressed in terms of two free-fit parameters as shown in Equation 7: dy MZ dω = 1 Y cos(α) E E cos(β) π [ 1 1 4 γ(α) = 3 sin(α)2 1 sin(α) 2 + cos(α)2 (3 sin(α) 2 + 1) 2 sin(α) 3 ln ( E E cos(β)γ(α) + 3 ) ] 2 π sin(β)sin(α)cos(φ) (7) ( ) 1 + sin(α), (8) 1 sin(α) where Y is the total sputter yield, E is a characteristic energy, E is the ion energy (ev), β is the incidence angle relative to normal, α is the polar angle, and φ is the azimuthal angle. The two free-fit parameters, Y and E, are determined from least squares fitting to the data while the remaining parameters are given by the experiment. For the present experiment, all ions are assumed to be normally incident so β = 0 and the azimuthal (φ) dependence is removed. Equation 7 is simplified to: dy MZ dω = Y 1 E E cos(α) π [ 1 1 4 E E γ(α) The Modified-Zhang (MZ) expression given above is applicable for low energy sputtering of planar samples. The parameter E is inherently related to the threshold sputtering energy, but has been found to disagree with experimental results. 14 Therefore, we will use this expression as an empirical fit only and not attempt to interpret the characteristic energy. The other free-fit parameter Y is the total sputter yield and is the quantity desired for this study. Several assumptions go into the QCM sputter yield analysis. The QCM is approximated as a point source since the area of the sensor (0.55 cm 2 ) is much smaller than the area of the hemisphere that the QCM subtends (2πr 2 qcm 4000 cm 2 ). The target is also approximated as a point source since the spot size of the plasma beam (1.5 cm diameter) is much smaller than the QCM distance. The quantity that varies for the finite target picture (as seen in Figure 6) is the polar angle from the target normal to the QCM. For a fixed QCM position, a differential area at the edge of the target would see the QCM at a different polar angle than from the center. The measured R(α) profile would then contain a convolution of all these individual contributions across the target. Hence, the angular distribution of sputtered particles must be assumed from a model or previous experiment. Another complication for a finite target is the radial profile of the plasma beam; this variation would also need to be accounted for in the integration. By approximating the target as a point source, the measured R(α) can be taken as the true angular distribution of sputtered particles and the radial dependence of ion flux can be ignored. We have also performed calculations with a finite target and assumed angular distribution and found the error to be less than 5%. ] (9) 7 of 13

Figure 6: The finite target integration components with local polar angle α α. C. Optical Emission Spectroscopy The spectroscopy set-up consists of two in vacuo shadow-shielded collimating lenses leading to an FHR1000 1.0 m focal length monochromator through a multi-fiber optic bundle. The output light from the FHR1000 is projected onto a 2048 x 512 pixel Synapse Charge-Coupled Device (CCD) camera. Figure 7b shows the experimental configuration of the collimating lenses and their respective collection volumes. The lenses are oriented parallel to the target plane and are located 3.0 cm and 5.5 cm axially from the target. By combining the two fibers in a bundle, we can resolve the two projected light signals on the CCD and thereby capture data from two spatial locations. A 2400 grooves/mm grating is installed to scan the wavelength range from 300 nm to 600 nm, where the majority of neutral molybdenum and ionized argon lines appear. The spectrometer is absolutely calibrated using the 404.656 nm line in an ArHg lamp, with the wavelength error increasing to 0.1 nm at the edges of the spectral range. (a) Collimating optics in Pi (b) Design of Shadow Shield Lengths Figure 7: Shadow shields were constructed to prevent coating of collimating lenses. Sputtered particles can deposit on collection optics or viewports over the course of an experiment, resulting in a 8 of 13

decreased spectral signal over time. This artificial decrease in emission intensity can complicate subsequent analysis that requires relative comparison of line intensities or absolute calculations for plasma density and temperature. We use shadow shields to prevent sputtered particles from seeing the collimating lenses (see Figure 7b). The shadow shields are sheets of aluminum foil that are sized based on the ray extending from the edge of the target to the edge of the given collimating lens, ensuring all sputtered particles from the target will be blocked by the shield. This addition removes the complication of coated optics present in many sputtering experiments and allows plasma and sputterant intensity changes to be isolated as purely physical phenomena. The two signals can be contaminated by systematic errors such as cosmic rays, stray light, CCD column bleeding, and collection optics misalignment. The cosmic rays can be eliminated by averaging two scans for the wavelength range and removing any outliers specified by a 10% rejection threshold. Stray light is minimized by covering the spectrometer and injection optics with opaque materials. CCD column bleeding results in light from one signal bleeding into the other as shown in Figure (go take an image of bleeding CCD), due to saturation of signals or low-frequency ADC conversion. To prevent column bleeding, we increased exposure time while minimizing slit size to prevent saturation and used 1 MHz high dynamic range ADC conversion. These settings were found to produce bleeding of less than 10% across the two signals when using a calibrated ArHg light source. Collection optics misalignment is minimized by rotating the input ferrule of the fiber-optic bundle to align the two light signals vertically with the entrance slit; a high-precision rotary fitting will allow better alignment in the future. IV. a) Preliminary Results b) c) 100 m Figure 8: The images above show the molybdenum sample a) before sputtering, b) after sputtering, c) an SEM image of the sample showing no large surface features. In order to benchmark the Pi facility for future tests on advanced micro-architectured materials, the sputter yield measurement techniques must be validated. Preliminary measurements are carried out for 150 ev and 300 ev ions incident on a flat molybdenum sample. The molybdenum sample used for this experiment is a 2 in diameter flat disk with 99.99% purity as verified with energy dispersive spectroscopy (EDS). The sample was mounted using six stainless steel bolts with molybdenum pieces tac-welded to the head to fill in the holes in the target. Before testing, the sample was verified to be flat, i.e. devoid of surface features and contamination, as shown in Figure 8. After plasma exposure, the sample developed a localized erosion region in the center corresponding to the incident plasma beam, and remained flat with no signficant surface features. For these tests, the plasma was operated in the focused plasma configuration as described in the previous sections with the same plasma conditions. Each test lasted approximately 1 hour, from applying the bias voltage to the sample to returning the sample to floating potential. The discharge voltage was observed to increase by 5% over the first 10 minutes of each test but remained generally constant over the remainder. The target current was recorded via the shunt resistor over the course of the experiment. The emission spectra in the visible range (300-600nm) is shown in Figure 9. The emission lines of neutral molybdenum are shown to be much more prominent for 300 ev than 150 ev due to the dependence of intensity on neutral density. Further development of the spectroscopic theory is required to extract sputter yields. Sputter yield meaurements were obtained using both the weight loss and QCM methods. The mass of the sample was recorded before and after each test using a precision microbalance. The QCM scan was performed for a total scan duration of 13 minutes after the plasma discharge voltage began to level off (around 10 minutes). Noise in the 9 of 13

(a) (b) Figure 9: Two spectra taken during sputtering with a) 150 ev and b) 300 ev ions showing the argon plasma lines and sputtered molybdenum lines. The molybdenum lines for 150 ev ions are barely visible with a SNR=5, but are strongly distinguishable for 300 ev ions. Note that the NIST notation is used: Mo I corresponds to neutral molybdenum while Ar II corresponds to singly ionized argon. QCM data was minimal for both the 150 ev and 300 ev test cases. The angular distributions were obtained using the conversions described in Section IIIB and is shown in Figures 10a and 10b. The Modified-Zhang (MZ) equation is empirically fit to the data and the curve is plotted over the original data points. The MZ fit appears to match the data reasonably well, and is used to obtain the integrated total sputter yield. The total sputter yields are plotted in Figure 11 against published results. Weijsenfeld et al. 16 and Laegreid et al. 15 performed weight loss measurements in a low-density plasma discharge and completely immersed the target. Zoerb et al. 17 performed QCM measurements using an ion beam at various incident angles. None of these authors corrected for redeposition or secondary electron emission as well. Overall, the measurements matches previous experiments fairly well within the error limits. The diversity in ion sources and measurement techniques justify this large spread in reported sputter yields. 1. Error analysis For the weight loss measurement, the main sources of error are the target current measurement, due to both secondary electron emission (SEE) and excess current collection, and potential plasma redeposition. For argon on molybdenum ion sputtering, the SEE coefficient is approximately 0.1, which should result in an overestimation of current by 10%. However, the uncertainty in these measurements is large and highly dependent on the surface state of the sample (e.g. oxygen or nitrogen monolayers). Since the surface was not characterized in this manner, the secondary electron 10 of 13

(a) 150 ev (b) 300 ev Figure 10: Angular distributions of differential sputter yield for a) 150 ev ions and b) 300 ev ions. Sputter Yield (atoms/ion) 1 0.8 0.6 0.4 0.2 Weijsenfeld Laegreid Zoerb Present Study (QCM) Present Study (WL) 100 1000 Ion Energy (ev) Figure 11: The sputter yields for Ar + Mo are compared to results from literature. 15 17 emission is accounted for as a 10% error in the target current measurement. In addition to SEE, the current collected by the sample is not well defined since the fringe fields near the edges allow the sample collect more current at higher bias voltages due to sheath expansion. 10 At the maximum bias voltage of -300 V, the sample collects around 10% more current than at -100 V as measured with a guard ring test. Since a correction would be difficult considering the field geometry at the edges, we treat the fringe effects as a 10% error. Finally, the plasma redeposition analysis showed a worst case scenario of 10% uncertainty in the yield which is also propagated for the weight loss. 4 For the QCM sputter yield, the analysis suffers from the same target current uncertainties, but is not complicated by redeposition. Therefore, the error bars are smaller in terms of target current uncertainty but may be complicated by the experimental setup (e.g. misalignment) or assumptions in the analysis. These other sources of error are not included in the plotted error bars. 11 of 13

V. Conclusion The Pi facility at UCLA has been used to generate a focused plasma for an argon ion on molybdenum sputtering experiment to validate two methods measuring the sputter yield. The plasma is transported from a LaB 6 hollow cathode through several solenoidal magnets to deliver a small spot-size, high flux plasma to a biased molybdenum target. The plasma discharge parameters and magnet strengths were chosen to produce a near-target plasma of sufficiently high flux for large ion fluences, but also low plasma density to prevent redeposition of sputtered particles. By avoiding redeposition, the weight loss method of calculating the sputter yield is simplified greatly. A planar Langmuir probe was used to measure the incident ion flux profile in front of the target, but the ion current used for the analysis is measured directly from the conducting target. A quartz crystal microbalance is installed with a rotary stage centered on the target to perform polar scans of differential sputter rates in the lateral plane. Two collimating lenses located at different axial distances from the target and oriented radially inward collect plasma and sputterant emission spectra. The following three methods of measuring the sputter yield were discussed: the weight loss method, the quartz crystal microbalance method, and the optical emission spectroscopy method. The sputter yield was measured for argon on molybdenum at ion energies of 150 ev and 300 ev and the sputter yields for the weight loss and QCM methods agreed well with published results. The emission spectra displayed the increase of sputtered molybdenum intensity with increased bias, but the sputter yields have not yet been extracted. The angular distributions at the different ion energies show the diffuse peaked profile characteristic of low energy sputtering as predicted and measured by models and other experiments. We conclude that the two methods of measuring the sputter yield for plasma-material interactions experiments have been validated within experimental error. The OES technique is still under development but shows promising results for the in situ measurement of sputter yield. The next step forward is to investigate the transient erosion of micro-architectured materials using the in situ diagnostics that have been validated. Acknowledgments This work was made possible by AFOSR Grant Nos. FA9550-14-10317, FA2386-13-1-3018, FA9550-11-1-0282, the National Defense Science and Engineering Graduate Fellowship and the UCLA Henry Samueli School of Engineering and Applied Sciences. The authors would also like to thank Ben Dankongkakul, Lucas Garel, and John Hayes for assisting with experiments. References 1 Wirz, R. E., Anderson, J. R., Goebel, D. M., and Katz, I., Decel Grid Effects on Ion Thruster Grid Erosion, IEEE Transactions on Plasma Science, Vol. 36, No. 5, oct 2008, pp. 2122 2129. 2 Goebel, D. M., Jorns, B. A., Hofer, R. R., Mikellides, I. G., and Katz, I., Pole-piece Interactions with the Plasma in a Magnetically Shielded Hall Thruster, Propulsion and Energy Forum, 2014, pp. 1 7. 3 Sekerak, M. J., Hofer, R. R., Polk, J. E., Jorns, B. A., and Mikellides, I. G., Wear Testing of a Magnetically Shielded Hall Thruster at 2000 s Specific Impulse, 34th International Electric Propulsion Conference, 2015, pp. 1 37. 4 Matlock, T. S., Goebel, D. M., Conversano, R., and Wirz, R. E., A dc plasma source for plasmamaterial interaction experiments, Plasma Sources Science and Technology, Vol. 23, No. 2, 2014, pp. 025014. 5 Nakles, M. R., Experimental and Modeling Studies of Low-Energy Ion Sputtering for Ion Thrusters, Master s thesis, 2004. 6 Doerner, R. P., Whyte, D. G., and Goebel, D. M., Sputtering yield measurements during low energy xenon plasma bombardment, Journal of Applied Physics, Vol. 93, No. 9, 2003, pp. 5816 5823. 7 Rubin, B., Topper, J. L., and Yalin, a. P., Total and differential sputter yields of boron nitride measured by quartz crystal microbalance, Journal of Physics D: Applied Physics, Vol. 42, No. 20, 2009, pp. 205205. 8 Chu, E., Goebel, D. M., and Wirz, R. E., Reduction of Energetic Ion Production in Hollow Cathodes by External Gas Injection, Journal of Propulsion and Power, Vol. 29, No. 5, 2013, pp. 1155 1163. 9 Goebel, D. M., Jameson, K. K., Katz, I., and Mikellides, I. G., Potential fluctuations and energetic ion production in hollow cathode discharges, Physics of Plasmas, Vol. 14, No. 10, 2007, pp. 103508. 10 Sheridan, T. E., Ion focusing by an expanding, two-dimensional plasma sheath, Applied physics letters, Vol. 68, No. 14, 1996, pp. 1918 1920. 11 Sauerbrey, G., Verwendung von Schwingquarzen zur Wagungdiinner Schichten und zur Mikrowagung, Zeitschrift fur Physik, Vol. 155, 1959, pp. 206 222. 12 Bachmann, L. and Shin, J. J., Measurement of the sticking coefficients of silver and gold in an ultrahigh vacuum, Journal of Applied Physics, Vol. 37, No. 1, 1966, pp. 242 246. 13 Yalin, A. P., Williams, J. D., Surla, V., and Zoerb, K. A., Differential sputter yield profiles of molybdenum due to bombardment by low energy xenon ions at normal and oblique incidence, Journal of Physics D: Applied Physics, Vol. 40, No. 10, 2007, pp. 3194 3202. 12 of 13

14 Zhang, Z. L. and Zhang, L., Anisotropic angular distribution of sputtered atoms, Radiation Effects and Defects in Solids, Vol. 159, No. 5, 2004, pp. 301 307. 15 Laegreid, N. and Wehner, G. K., Sputtering yields of metals for ar+ and ne+ ions with energies from 50 to 600 ev, Journal of Applied Physics, Vol. 32, No. 3, 1961, pp. 365 369. 16 Weijsenfeld, C. H., Hoogendoorn, A., and Koedam, M., Sputtering of polycrystalline metals by inert gas ions of low energy (100-1000eV ), Physica, Vol. 27, 1961, pp. 763 764. 17 Zoerb, K. A., Williams, J. D., Williams, D. D., and Yalin, A. P., Differential Sputtering Yields of Refractory Metals by Xenon, Krypton, and Argon Ion Bombardment at Normal and Oblique Incidences, IEPC, 2005, pp. 1 25. 13 of 13