This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 164.67.192.128

This content was downloaded on 27/04/2017 at 18:17

Please note that terms and conditions apply.

In situ plasma sputtering and angular distribution measurements for structured molybdenum

surfaces

View the table of contents for this issue, or go to the journal homepage for more

2017 Plasma Sources Sci. Technol. 26 065002

(http://iopscience.iop.org/0963-0252/26/6/065002)

Home Search Collections Journals About Contact us My IOPscience

In situ plasma sputtering and angulardistribution measurements for structuredmolybdenum surfaces

Gary Z Li, Taylor S Matlock, Dan M Goebel, Christopher A Dodson,Christopher S R Matthes, Nasr M Ghoniem and Richard E Wirz

University of California, Los Angeles, CA, 90024, United States of America

E-mail: wirz@ucla.edu

Received 10 October 2016, revised 15 March 2017Accepted for publication 31 March 2017Published 27 April 2017

AbstractWe present in situ sputtering yield measurements of the time-dependent erosion of flat andmicro-architectured molybdenum samples in a plasma environment. The measurements areperformed using the plasma interactions (Pi) Facility at UCLA, which focuses a magnetizedhollow cathode plasma to a material target with an exposure diameter of approximately 1.5 cm.During plasma exposure, a scanning quartz crystal microbalance (QCM) provides angularsputtering profiles that are integrated to estimate the total sputtering yield. This technique isvalidated to within the scatter of previous experimental data for a planar molybdenum targetexposed to argon ion energies from 100 to 300 eV. The QCM is then used to obtain in situmeasurements during a 17 h exposure of a micro-architectured-surface molybdenum sample to300 eV incident argon ions. The time-dependent angular sputtering profile is shown to deviatefrom classical planar profiles, demonstrating the unique temporal and spatial sputtering effects ofmicro-architectured materials. Notably, the sputtering yield for the micro-architectured sample isinitially much less than that for planar molybdenum, but then gradually asymptotes to the valuefor planar molybdenum after approximately 10 h as the surface features are eroded away.

Keywords: plasma-material interactions, sputtering, angular sputtering profile,time-dependent sputtering yield, molybdenum

1. Introduction

Ion-induced erosion of plasma-facing materials is importantto the performance and behavior of many modern plasmadevices. In electric propulsion, a variety of different failuremodes have emerged due to sputtering: accelerator grids inion thrusters [1], channel wall erosion in unshielded Hallthrusters [2], and pole piece erosion in magnetically shieldedHall thrusters [3, 4]. Similarly, in fusion reactors, high-Zimpurities are sputtered from divertors, limiting their lifetime[5]. These issues have substantiated the need for studying theplasma-material interactions, such as ion-induced erosion, in awell-controlled test environment.

Sputtering can be difficult to measure directly due tocomplex geometries in devices and complicating plasma effects.Therefore, models are commonly used to simulate device ero-sion [5–7]. These models require accurate knowledge of the

sputtering yield for the given plasma-wall combination, whichis typically obtained from experiment and compared to theory[8]. While many experimental techniques have been developedto measure the sputtering yield of materials, in situ measure-ments in a plasma environment are scarce. Additionally, widevariations in sputtering yield measurements by plasmas havenot been quantified, since plasma and surface characterizationare generally not provided.

To measure sputtering yield, an energetic ion flux on thesurface of a material is provided by an ion source from aplasma discharge or an ion gun. Ion guns have well definedion energies, incidence angles, and charge uniformity but lackhigh current density due to space charge constraints [9]. Onthe other hand, plasma discharges generate much highercurrent densities, but suffer from a spread in ion energies,multiply charged ions, and plasma-generated redeposition[10]. The experimental parameter space of target material

Plasma Sources Science and Technology

Plasma Sources Sci. Technol. 26 (2017) 065002 (11pp) https://doi.org/10.1088/1361-6595/aa6a7d

0963-0252/17/065002+11$33.00 © 2017 IOP Publishing Ltd1

variables, which includes ion energy, flux, fluence, andtemperature, is further complicated by more material-specificparameters, such as surface composition and topography. Toensure a well-defined experiment, all of these parametersmust be characterized in detail.

Adding to the large parameter space for the sputteringexperiment, numerous diagnostics have been used to mea-sure the sputtering yield. Proven techniques include weightloss [10, 11], optical emission spectroscopy [10, 12], quartzcrystal microbalance (QCM) [13, 14], and Rutherfordbackscattering [9], among others. These diagnostics arecategorized based on whether analysis is performed inside avacuum environment (in situ), or outside of the facility(ex situ). It is desirable to perform in situ measurements tothe extent feasible to avoid the time constraints related toremoving the sample from the chamber vacuum, and toreduce the impact of atmospheric contamination on testresults. For the current Pi facility configuration [11], weightloss measurements are performed ex situ, while QCM dataare obtained in situ. An important advantage of the QCM isthe capability to measure both the time-dependent angularsputtering profile and yield.

The objective of this paper is to present in situ mea-surements of the total sputtering yield to study time-dependent erosion of micro-architectured materials. A plasmadischarge is used to generate sufficiently large ion fluxes atthe target to determine the dependence of the erosion rate onion fluence of micro-architectured materials within practicaltest durations (hours to days). Matlock et al [11] presenteddetailed characterization of the plasma source operatingregimes and provided an initial benchmark of a weight losssputtering yield measurement. In this investigation, wedevelop an experimental technique utilizing a QCM with afocused plasma configuration for measuring the time-resolvedsputtering yield of micro-architectured samples.

The paper layout is as follows. First, we describe the Pifacility experimental set-up and diagnostics suite in section 2.Section 3 discusses the weight loss and QCM methods ofmeasuring sputtering yield. Section 4 presents results on thecharacterization of the focused plasma configuration, fol-lowed by a series of two tests: argon ions impacting flatmolybdenum and micro-architectured molybdenum. Section 5ends with a summary and conclusions.

2. Experimental setup

2.1. Pi facility

The Pi facility consists of a 1.8 m diameter by 2.8 m longcylindrical aluminum vacuum chamber and two cryopumpswith a combined pumping speed of 5000 l s−1 on argon. Thebase pressure is ´ -5 10 7 Torr and the typical workingpressure (at 27 sccm argon flow) is ´ -( )6.0 0.5 10 5 Torr.The overall configuration of the Pi source is shown infigure 1. For this study, the Pi facility’s DC discharge uses a250 A lanthanum hexaboride (LaB6) hollow cathode, locatedat the far upstream end of the plasma column. The cathode

orifice is located at the downstream face of the cathodemagnets and is directed into a 5 cm diameter, 30 cm long,cylindrical anode held at ground potential to supply a partiallyionized argon or xenon plasma. The gas injection is dividedbetween the hollow cathode and an external anode injectorjust downstream of the cathode keeper face. The use of anauxiliary anode injector has been shown to reduce cathodeplume oscillations and also minimize the production ofenergetic ions [15, 16]. The plasma discharge is controlled bycommanding the gas flow and discharge current. The Pifacility operating conditions have been previously describedby Matlock et al [11, 17].

The plasma is confined by electromagnetic solenoidssurrounding both the cathode and anode, and transporteddownstream via the water-cooled chamber magnets, also heldat ground potential. This design isolates the cathode from thetarget region to prevent sputtered particles from the cathodefrom depositing on the target (and vice versa) [11]. At thedownstream end of the chamber, the sample magnets pinchdown the plasma to provide focused impingement on a biasedtarget to avoid interaction with other facility surfaces. Theplasma ions are typically formed within an electron temper-ature (<10 eV) of ground potential, providing a stable refer-ence for biasing the target negative to control the incident ionenergy. The target floats at around −40 V with respect to thechamber magnets, which are maintained at facility ground. Toprovide a range of incident ion energies, the target can bebiased independently, as low as −300 V with the presentpower supply.

2.2. Diagnostics

A subset of the diagnostics suite implemented for this study isshown in figure 1, with a detailed list containing the diag-nostic and function presented in table 1.

The Langmuir probe consists of a cylindrical tungstenwire encapsulated by an alumina probe holder. The electrontemperature is calculated from the slope of the I–V curve inthe transition region (a Maxwellian electron population isassumed), while the plasma density is extracted from the ionsaturation region. Based on previous measurements [11], theprobe ratio ( lrp D), where rp and lD are the probe radius andDebye shielding length, respectively, indicates that orbitalmotion limited and thin-sheath approximations are not validin our plasma. Instead, a numerical approach by Steinbruchel[18], based on the theory by Laframboise, is needed. Thistechnique provides a measurement of the electron temperatureand plasma density by accounting for varying sheath effects.

The Faraday probe is a planar Langmuir probe with aconcentric guard ring used to produce a uniform sheath acrossthe probe face. The circular collection electrode has an area of0.14 cm2 and is insulated on the backside with high temper-ature ceramic cement. Both the guard ring and collector arebiased to −100 V to collect ion saturation current, which isrelated to plasma density, ne, by p»I n eA kT m0.6 8e p esat .Consequently, the 1-sided ion flux can be definedas G = ( )I eAi psat .

2

Plasma Sources Sci. Technol. 26 (2017) 065002 G Z Li et al

The retarding potential analyzer (RPA) is constructed outof four electrically isolated molybdenum meshes ( 0.001diameter, 100 wires/in) with a tungsten collection electrode atthe end. The ordering of the grids is as follows: floatingattenuation, electron repelling, ion retarding, and secondaryelectron suppression. The floating attenuation grid minimizesperturbations and reduces the plasma density entering thedevice. The electron repelling grid is biased to −100 V torepel electrons and only allow ions through. The ion retardinggrid is scanned from a typical voltage range of −60 V to+20 V with the upper limit determined by where the ioncurrent disappears. The secondary electron suppression grid isbiased at −9 V relative to the collector, which is at −100 V,to reflect secondary electrons from the other grids. In order tomeasure the spread of ion energies hitting the sample, theRPA is positioned at the location of the target in figure 1.

The QCM provides in situ measurements of the angulardistribution of sputtered particles, which can be integrated toobtain the total sputtering yield. As particles are adsorbedonto the crystal sensor, the resonance frequency decreaseslinearly in proportion to the deposited mass [19]. Thepiezoelectric sensitivity of quartz crystals to added massallows deposition rates as low as - -0.4 ng cm s2 1 to bemeasured.

An INFICON front load single sensor containing a silver-coated AT-cut quartz crystal is employed in the chamber. The

crystals have a resonance frequency of 6MHz and an exposedarea of 0.55 cm2, and are assumed to have a sticking coeffi-cient of order unity for condensable metals [20]. Data islogged using the STM-2 deposition monitor, which alsopossesses an external oscillator for precise frequency mea-surements. To ensure thermal stability, the crystal holder iswater-cooled and protected by a pneumatically actuatedshutter.

The QCM crystal holder is mounted to a rotary stage witha moment arm of length 25.4 0.2 cm to provide measure-ments at varying polar angles, α, in the lateral plane. Thecoordinate system used for the QCM measurements is definedin figure 2. For normally incident ions, we assume the angulardistribution of sputtered particles to be azimuthally symmetricso the differential sputtering rate, a( )R - -(ng cm s2 1), is afunction of α only.

To obtain polar measurements, the QCM is rotated in 5°increments from α=90° to 30°, beyond which the QCMbecomes blocked by the chamber magnets. Due to vibrationsin between measurements, we record data for 1 min at eachlocation and neglect the first and last 5 s of data. The diff-erential sputtering rate, a( )R , is obtained by performing anarithmetic average of the data over the 50 s of measurement.

The remaining diagnostics are described briefly as theyare under development or used to provide auxiliary

Figure 1. Schematic of the front end of the Pi facility vacuum chamber. The ionization gauge is located 120 cm away from the plasma columnat the rear end of the chamber.

Table 1. Pi facility diagnostics suite.

Diagnostic Measured parameters

Langmuir probe n T,e e

Faraday probe 1-sided ion fluxEmissive probea Plasma potentialRetarding potential analyzer Ion energy distributionQuartz crystal microbalance Differential sputtering rateEmission spectroscopya Line emissionResidual gas analyzera Partial gas pressuresIonization gauge Neutral pressureLong distance microscopea In situ surface visualizationEnergy dispersive spectroscopy Surface composition

aDiagnostics that are not specifically used in the study.

Figure 2. Definition of the QCM coordinate system relative to thetarget. The polar angle is defined relative to the z-axis.

3

Plasma Sources Sci. Technol. 26 (2017) 065002 G Z Li et al

information for this study. Emissive probes were used tomeasure plasma potential fluctuations representative of anazimuthal ion instability [21]. A spectroscopy system, con-sisting of collection optics leading to an FHR1000 spectro-meter through an optical fiber, has been implemented foroptical emission spectroscopy as a supplementary method ofsputtering yield measurement, but the analysis is still underdevelopment. A residual gas analyzer (RGA) was used toensure that there were no water leakages or major con-taminants prior to testing. The ionization gauge is used tomonitor the neutral gas pressure at the rear-end of thechamber. A long distance microscope is being implementedfor in situ visualization of surface evolution during sputtering.A scanning electron microscope (SEM) was used in con-junction with energy dispersive spectroscopy (EDS) to ana-lyze surface morphology and inspect for contaminants beforeand after plasma exposure.

3. Sputtering yield measurement techniques

3.1. Weight loss

The measurement of sputtering yield by weight loss is oftenused because of its relative simplicity. The mass of a sampleis recorded before and after plasma irradiation, and the ioncurrent is monitored during testing from a 10Ω resistorconnected directly to the target. A cylindrical Kapton shieldwith 70 mm thickness is wrapped around the target to preventexcess current collection from the sample mount assembly.The mass loss and target current data provide sufficientinformation for calculating the sputtering yield of the targetmaterial at a given ion energy. For a pure-element samplewith negligible impurity concentration, the total number ofparticles sputtered can be calculated as follows:

=D⎜ ⎟⎛

⎝⎞⎠

∣ ∣ ( )NM

mN , 1s A

where Ns is the total number of sputtered atoms, DM is themass loss in grams, m is the atomic weight of the targetmaterial in amu, and NA is Avogadro’s number. For materialswith multiple constituents, the atomic weight, m, can bereplaced with an effective mass meff . The number of ionsincident on the target is calculated by integrating the targetcurrent from onset to removal of the target bias:

ò= ( ) ( )Ne

J t t1

d , 2t

t

ti0

1

where Ni is the total number of incident ions, e is the ele-mentary charge, t0 and t1 are the bias initiation and removaltimes respectively, and Jt(t) is the time-varying target current.The ion fluence, F, introduced in section 4 is equal to Ni

divided by the exposure area. The ratio of the two termsdescribed above gives the total sputtering yield in atoms perion:

= ( )YN

N. 3s

i

The quantity Y is a time-averaged value, obtained overthe duration of plasma exposure. A time-resolved sputteringyield would be difficult to obtain because the bias time wouldneed to be sufficiently short and the sample would need to beremoved from vacuum to be weighed. A solution wouldrequire an in situ mass balance, but the setup would be un-necessarily complex.

Other complications in the weight loss technique includecontamination, ion implantation, obliquely incident ions, andplasma redeposition. The presence of contaminants, such ascarbon deposits, ion implantation, and oxide formation areinspected visually in the SEM and spectroscopically usingEDS prior to and after testing. It is found that the majority ofcontaminants were present from the fabrication process beforetesting. After plasma exposure, the percentage of carbon andoxygen contamination decreased, indicating that the sputter-ing process cleaned the surface. Obliquely incident ions canalso be neglected because the ion flux drops off rapidly nearthe edges of the target (see section 4).

Lastly, plasma redeposition can result in an under-estimation of the sputtering yield. The reduction in sputteringyield is estimated by using a model that determines how manysputtered atoms are ionized in the near-target plasma and areredirected back to the sample due to pre-sheath electric fields[10, 11]. Redeposition is found to increase with higher elec-tron temperature and density, so the operating conditions areadjusted to minimize these two parameters. Following themethodology of Matlock et al [11], plasma redeposition has aless than 10% effect for densities lower than -10 m18 3 andelectron temperature of 7 eV. This effect is the primaryuncertainty in the weight loss technique and is included as a10% worst case error in the reported sputtering yields.

3.2. Quartz crystal microbalance

The in situ QCM method enables time-resolved measure-ments of the angular distribution of sputtered particles andtotal sputtering yield. Following the methodology of Yalin[14], the differential sputtering yield, Wyd d , can be calcu-lated in terms of the differential sputtering rate, a( )R , asfollows:

a a

W=

( ) ( )( )y eR N r

mJ

d

d, 4

t

A qcm2

where m (amu) is the atomic weight of sputtered particles, NA

is Avogadro’s number, Jt (C/s) is the target current, and rqcm

( 25.4 0.2 cm) is the distance from the QCM to the targetcenter. Note that the differential sputtering rate, a( )R isdefined as the deposition flux on the QCM sensor, where thesensor area is 0.55 cm2. To calculate the total sputtering yield,the differential sputtering yield is integrated over all solidangles in the hemisphere in front of the target:

ò òa

a a f=W

p p ( ) ( ) ( )Yyd

dsin d d . 5

0

2

0

2

4

Plasma Sources Sci. Technol. 26 (2017) 065002 G Z Li et al

For a true steady-state experiment, equation (5) wouldprovide the same sputtering yield regardless of when theQCM scan was performed. However, this is not the case foran experiment with a time-dependent sputtering yield. Amicro-architectured material will have a transient erosion ratewhich is a function of the surface features at the erosion site,changing both the overall sputtering yield and angular sput-tering distribution. To capture this time-dependent sputteringbehavior, the QCM scans must be performed on a time-scalemuch shorter than the transient erosion time-scale.

The time-dependent sputtering yield equation can beobtained by combining equations (4) and (5) as follows:

òpaa

a a=p

⎜ ⎟⎛⎝

⎞⎠( ) [ ( )]

[ ( )][ ( )] ( )Y t

eN

mr

R t

J tt2 sin d . 6

t

Aqcm2

0

2

In this equation, the time-dependence is captured in thedependent variable a ( )t which represents the QCM angularlocation. Since the QCM records data over a finite period oftime, the sputtering rate a[ ( )]R t and target current a[ ( )]J tt

are temporally averaged. The sputtering rate will also changedepending on the angle of collection, the present surfacemorphology, and the average ion flux during measurement.The ion flux is highly dependent on the local plasma density,which is a function of the plasma discharge parameters. Theopen-loop discharge control system tends to allow the localplasma density to increase slowly with time. These transienteffects resulted in a changing target current over time, whichis accounted for by monitoring the plasma current and inequation (6) by averaging the measurements over a periodmuch shorter than the plasma variations (on order of hours).

In order to perform the integration in equation (6), thedifferential sputtering yield must be extrapolated to theinterior polar angles where data cannot be obtained. This isdone by fitting the data to a well-known sputtering profile.The sputtering profile for normally incident ions is known tohave a cosine shape for high energies (>1 keV). However, forlow energy sputtering (<1 keV), the angular profile tends bemore diffuse with a relative maximum near 45°. Sigmund[22] showed that the angular dependence of the sputteringyield is given by a( )1 cos f , where f is a function of the massratio of incident-to-target atoms and is generally of the orderof 1. Yamamura et al [8] made a correction to Sigmund’sdistribution, where they developed an empirical formula thataccounts for the drop off in yield near grazing angle ofincidence. Later, Zhang et al [23] developed an empiricalformula that provided an accurate fit to angularly dependentsputtering data. A modified form of this equation [14] is usedto model normally incident, low energy sputtering of planarsamples:

**

ap

g aW

=-

-⎡⎣⎢⎢

⎤⎦⎥⎥

( ) ( ) ( )y Y E

E

d

d 1

cos1

1

4, 7

E

E

MZ

g aaa

a aa

aa

=-

++

´+-

⎡⎣⎢

⎤⎦⎥

( ) ( )( )

( ) ( ( ) )( )

( )( )

( )

3 sin 1

sin

cos 3 sin 1

2 sin

ln1 sin

1 sin, 8

2

2

2 2

3

where Y is the total sputtering yield, E* is a characteristicenergy (eV), E is the ion energy (eV), and α is the polarangle. The two free-fit parameters, Y and E*, are determinedfrom least squares fitting to the data while the remainingparameters are obtained from the test conditions.

Equation (7) does not hold for micro-architecturedmaterials where ions impact the surface with a range of non-normal incidence angles. Therefore, an alternative method ofextrapolation must be used to determine the integrated sput-tering yield for these samples. Ideally, the sputtering profilefor a material with well-characterized surface features wouldbe modeled using local sputtering distributions for individualcollisions. This process would involve a Monte-Carlo simu-lation integrating 3D surface mapping, an accurate sputteringtheory, and the effects of a radial ion flux gradient. For testingof multiple samples with unique surface features, thissophisticated analysis is not feasible.

Instead, we determine upper and lower bounds for theintegrated sputtering yield through a simpler method. Theupper bound is obtained by assuming the differential sput-tering yield is constant from a = 30 to a = 0 , while thelower bound assumes the yield drops to zero linearly froma = 30 to a = 0 . The reported sputtering yield is theaverage of these two values with the differences from themean included as a standard deviation. The error due to thisextrapolation technique is found to be less than 5%.

The error analysis also includes the point sourceassumption for the target and ion current uncertainties. Thetarget is well-approximated as a point source since the dia-meter of the plasma beam (1.5 cm) is much smaller than theQCM distance (25.4 cm). By approximating the target as apoint source, the measured a( )R can be taken as the trueangular distribution of sputtered particles. Realistically, theion flux has a finite radial gradient across the target whichresults in different emission angles to the QCM. Forexample, when the QCM is located at a polar angle of 45°, itwill collect emission from the two opposing edges of thetarget at 41.3 and 49.3 respectively. The typical 5% var-iation in the sputtering rate over 4 is included in the errorbars for the angular sputtering profile, which is propagatedto the final yield uncertainties. Note that while this effectsmears out the measured angular profile, it should not impactthe integrated sputtering yield since the ejected massremains the same.

The ion current measurement suffers from uncertaintiesin the collection area and secondary electron emission (SEE).The sample mount is encapsulated by a Kapton shield toprevent excess current collection, but deviations on the orderof 5% are still detected. Additionally, the impacts of100–300 eV ions on the molybdenum surface result in sec-ondary electron yields on the order of 0.1 electrons/ion[24, 25]. However, SEE yields depend strongly on surfacecharacteristics such as oxide layers and roughness so the SEEbias correction is omitted in favor of a 10% error in thereported yield.

5

Plasma Sources Sci. Technol. 26 (2017) 065002 G Z Li et al

4. Results and discussion

4.1. Plasma characterization

The operating conditions for each test are fixed to thoseshown in table 2 in order to ensure a consistent plasma-sputtering environment. The mass flow was divided betweenthe cathode (15 sccm) and anode (12 sccm) gas injectors toproduce a working pressure of ´ -6 10 Torr5 on argon andmeasured on the chamber wall 4 ft away from the plasmasystem. The discharge current was fixed to 45 A and thecathode potential floated to an equilibrium value of −40 Vrelative to the grounded anode. This configuration producedthe near-target plasma conditions shown in table 2, measuredusing the diagnostics suite described in section 2. The plasmadensity and temperature are replicated to within 10% for eachexperiment.

The near-target plasma must be understood and char-acterized in detail to provide accurate sputtering yield mea-surements. As shown in figure 3, we focus the plasma into a1.5 cm diameter cusped beam in order to provide localizederosion away from the edges of the 5 cm diameter sample.This enables accurate measurement of the angular sputteringprofile using the QCM. Martinez-Sanchez [26] found that theion flux is conserved throughout the axial extent of the cusp.Therefore, we can directly control the delivered ion flux bytuning the discharge parameters.

We map the 1D axial ion flux 5 mm in front of the targetusing a transverse scanning Faraday probe. These scans areperformed before and after target biasing to avoid con-tamination from sputtering of the probe material. The 2Dcontour and lateral profile are shown in figures 4(a) and (b)

respectively. The plasma at this location is azimuthallysymmetric, and is roughly Gaussian with a full-width half-maximum of approximately 1.5 cm, which we define to be theeffective exposure diameter of the plasma. The data for >x 0deviate from the Gaussian fit with elevated wings that don’tappear to converge to zero flux at infinity. It was discoveredpost-test that the Langmuir probe holder developed a con-ducting surface layer of sputtered molybdenum that resultedin extraneous ion collection as the probe holder entered theplasma. Therefore, the left half ( <x 0) of the Gaussianshould be seen as representative of the incident plasma.

An extrapolation of the Gaussian to the radius of thesample (2.5 cm) reveals that there is a non-negligible currentat the edges. The effect of this finite edge current on the QCMmeasurements has been discussed in section 3.2. One benefitof the radial variation of plasma density across the targetregion is to allow the characterization of the plasma materialinteraction behavior (e.g., erosion-deposition) for a range ofion flux and fluence for a single test. This variation in plasmaexposure conditions for a single sample is especially infor-mative for analysis of featured and micro-architectured sur-faces [27].

Using the spectroscopy system, we record broadbandspectra in the 300–900 nm range to guarantee minimal doublyionized argon in the impinging plasma. ++Ar lines withstrong intensity levels in low temperature argon plasmas fromthe NIST database were shown to be lower than the detectionthreshold of the spectrometer. Representative spectra of theplasma is provided in [28]. In addition, the axial energyspread of the ions at the target is measured using the RPAdescribed in section 2. The normalized ion energy distributionfunction is shown in figure 5. The distribution is reasonablywell approximated by a Gaussian with a standard deviation of6 V. The retarding potential provides a measure of whatplasma potential the ions were created at, implying that themajority of the ions were created near ground potential.Therefore, when the target is biased to a voltage <V 0t , theincident ion energy can be described as ∣ ∣V 6 eVt .

Deviations in the nominal operating conditions aremonitored during each test to accurately characterize thetesting environment. The Langmuir and Faraday probesare scanned before and after the target is biased to determinethe ion flux profiles and electron temperatures. The ion cur-rent time-series is used to monitor changes in the incident ionflux and also allows for time-resolved sputtering yield mea-surements as described in the following sections.

4.2. Testing procedure

In this section, we describe the general testing procedure usedfor the following experiments. First, the sample is weighedusing an A&D GR300 precision microbalance and the surfaceis characterized using a SEM and energy dispersive x-rayspectroscopy (EDS). The SEM images are primarily used todocument how the surface morphology changes after plasmaexposure, and is also used to inspect for surface contamina-tion. The EDS spectra indicate the level of oxidation beforeand after testing. The sample is then inserted into the biased

Table 2. Near-target plasma conditions.

Parameter Typical values

Plasma density -10 m18 3

Electron temperature 7 eVIon energy to target 40 to 300 eVIon flux to target - -10 cm s17 2 1

Exposure diameter/area 1.5 cm 1.8 cm2

Figure 3. The near-target plasma is shown impacting a flatmolybdenum sample biased to −300 V. The green glow is due tode-excitation of sputtered molybdenum atoms.

6

Plasma Sources Sci. Technol. 26 (2017) 065002 G Z Li et al

sample mount and the chamber is pumped down to basepressure (< ´ -5 10 6 Torr) to lower the partial pressures ofbackground gases. To begin the test, the hollow cathode isignited, and the discharge and chamber magnet settings areadjusted to deliver a previously characterized plasma ofknown ion flux to the target, which floats to approximately−40 V where sputtering is below the detection threshold ofour diagnostics. The target is then biased to bombard thetarget with plasma ions extracted through the sheath. The ioncurrent to the target is recorded via a current shunt forcalculating the sputtering yield throughout the test, and canalso be used to check for arcing or variations in the near-targetplasma conditions. At the end of the test, the target bias isremoved and the plasma is turned off. Molybdenum oxidizesat approximately 400 °C, so the sample is allowed to cool to100 °C–200 °C (around one hour after testing) to minimizeoxide formation after exposure to atmosphere. In most cases,the sample was allowed to cool for several hours beforeremoval. The chamber is backfilled to atmosphere through a

venting port at the rear end of the chamber, and the target isremoved and weighed.

4.3. Flat molybdenum test

The Pi facility sputtering yield diagnostics were first bench-marked on an argon ion on flat molybdenum sputteringexperiment for ion energies of 100, 150, 200, and 300 eV.The molybdenum sample is a 2 in diameter disk that was cutfrom a rolled finish molybdenum sheet. Subsequently, thesample was exposed to plasma several times, resulting in thesputter-conditioned surface shown in figure 6. Prior to testing,the sample was composed of 99.99% molybdenum by weightand contained few surface features larger than 10 μm.Figure 6 was found to be representative of the sample surfaceafter a survey of the entire disk. The argon plasma wasconfigured into the focused configuration and maintained thenear-target plasma conditions given in table 2. The target wasexposed to ion bombardment for approximately 1 h while the

Figure 4. The plots above show (a) ion flux contours and (b) an ion flux profile across the target center-line for the plasma. The plasma isoffset by approximately 4 mm in the vertical direction relative to the target.

Figure 5. The ion energy distribution function is measured at thetarget using the RPA and fitted to a Gaussian distribution.

Figure 6. SEM image representative of the flat molybdenum sampleat 900× magnification.

7

Plasma Sources Sci. Technol. 26 (2017) 065002 G Z Li et al

ion current was monitored and the QCM was scanned toobtain the angular sputtering distribution.

The angular sputtering profiles for the four different testcases are shown in figure 7. Each data set is fit to a modified-Zhang (MZ) profile using least-squares minimization of thetwo free parameters, Y and E*. The MZ fit underestimates thepeak values near 50° and drops prematurely to zero for100 eV, but otherwise extrapolates well to the interior polarangles. Whether this discrepancy for the peak values is phy-sical or due to experimental error is uncertain. The 100 eVprofile drops to zero at 20° which is clearly unphysical. Whilethe MZ-fit is based on low energy sputtering, the theory maybe invalid near the sputtering threshold. Zhang uses a value of* =E 114 eV for molybdenum, but he comments that there is

disagreement within the literature on the accuracy of thisvalue [23]. For the 100 eV data, the extrapolation to interiorangles is redefined as a line from the data point at 30° to theorigin at 0° for a more physical representation. Theseempirical fits allow for integration of the angular distributionto obtain the total sputtering yield.

To determine the temperature dependence of the sput-tering yield, QCM scans were obtained at 10 min intervals forthe flat molybdenum test with 300 eV argon ions, as thesample heated from 25 °C to approximately 500 °C. Thevariation of the measured sputtering yields were within 5% ofthe value for 300 eV in figure 8, showing that there was noclear dependence of sputtering yield on temperature withinthis temperature range.

The total sputtering yields were measured using both theweight loss and QCM techniques as described in section 3.Results for the four different ion energies are shown infigure 8 and compared to previous measurements. Whencomparing sputtering yields, it is important to note the ionsource, the method of measurement, and surface character-ization. Weijsenfeld et al [29] and Laegreid et al [30] per-formed weight loss measurements in a low-density plasmadischarge with a fully immersed target. Zoerb et al [31]

performed QCM measurements using an ion beam at variousincident angles. All experiments did not correct for SEE anddid not inspect the surfaces for contamination or surfacefeatures. Due to the abundance of experimental factors, adirect quantitative comparison is difficult.

Overall, the sputtering yields from this study fall wellwithin the scatter of published results. The weight loss andQCM results obtained in this study are consistent withinexperimental error. However, the trend in the energydependence for the QCM yield appears to deviate from theothers, particularly influenced by the data point at 300 eV.This discrepancy is attributed to the QCM method havingmore sources of error (see section 3.2) than the weight losstechnique, with several potential problems based on align-ment. To truly determine whether the energy dependencemeasured by the QCM continues on that trend, an additionalvalidation test from 300 to 1000 eV would need to be per-formed. Fortunately, this problem is not critical to the micro-architectured test described in the following section. Since theQCM setup will remain the same throughout the test, thetransient sputtering yield should possess generally the sameshape, potentially with an offset due to misalignment.

4.4. Micro-architectured molybdenum test

After the in situ diagnostics were benchmarked on a flatmolybdenum sample, we proceeded to measure the time-dependent sputtering yield of a micro-architectured molyb-denum sample. Figure 9 shows an SEM image of the sphe-rical surface features generated from the chemical vapordeposition fabrication process. This experiment presents thenext level of validation as the sputtering yield of a texturedmolybdenum sample exposed to plasma will eventuallyconverge to the flat sample result when the surface structurehas eroded away.

The sample was tested for a total duration of 17 h for thesputtering yield to converge to a constant value. The ion energywas fixed at 300 eV and the plasma was operated using thesame conditions as the flat molybdenum test. The plasma

Figure 7. Angular distributions for +Ar Mo with ion energies of100–300 eV. The uncertainties are not shown here to clearly showthe different profiles, but are included in the final error analysis forthe sputtering yield.

Figure 8. The sputtering yields for +Ar Mo are compared toresults from literature.

8

Plasma Sources Sci. Technol. 26 (2017) 065002 G Z Li et al

density and electron temperature were constant to within 10%of the values shown in table 2, and the average ion flux wasmeasured to be ´ - -( )2.3 0.2 10 ions cm s17 2 1.

The QCM was used to measure the angular sputteringdistribution at discrete time intervals during plasma exposure.By using a QCM scan duration (13 min) shorter than thecharacteristic erosion time (about 1 h based on previousexperiments), the time-dependent angular sputtering profilewas measured. The integrated sputtering yield is described asa function of the ion fluence,F, which is defined as the totalnumber of ions incident on the target per unit area:

F ò= ( ) ( )F t td , 9t

t

i0

1

where Fi(t) is the measured ion flux. As shown in figure 10,the ion fluence can be used in place of time as the dependentvariable through the above calculation.

The featured surface is modified based on the total ionfluence and produces a unique sputtering profile that evolves intime. Since the plasma sheath size is much larger than theheight of micro-features, the ions will strike the surface atvarying angles, which also influences the ejection angle ofsputtered particles. When an ejected target atom intercepts anearby surface structure, it re-deposits on the material. Thisgeometric re-trapping of sputtered particles is known as bal-listic deposition and both modifies the angular sputtering dis-tribution and decreases the overall sputtering yield [27, 32].

The evolution of the angular sputtering distribution forthe micro-architectured molybdenum sample is shown infigure 11. The initial measurement at t=0.16 h shows aprofile that is overall lower in magnitude and not significantlypeaked off axis. The relatively flat profile for the interiorangles results from ions striking the surface structures atvarying incidence angles, which depends on the specificgeometry of the features shown in figure 9. The intermediateprofile at t=3.31 h develops a local maximum near 45° andincreases in magnitude relative to the initial profile, indicatingan increase in sputtering yield. For the final measurement att=17.0 h, the angular distribution closely matches the resultspresented for the flat molybdenum sample at 300 eV fora > 50 , but shows some deviation for the interior angles.Due to the presence of the plasma column, the QCM does notrecord data at angles a < 30 , thus data are not available toconfirm whether the discrepancy remains for ejection anglescloser to a = 0 .

The classical picture of the dependence of differentialsputtering yield on the incidence angle for flat surfacesappears to be violated in samples with micro-architecturedsurfaces. The surface structures modify the distribution ofincidence angles from purely normal to a function thatdepends on the surface geometry. As shown in figure 11, theangular sputtering profile is also a function of time, andequivalently the ion fluence, evolving as the surface morph-ology changes. The surface features suppress the a( )1 cos f

Figure 9. SEM image of the micro-architectured molybdenumsample. The average feature size is approximately 100 mm.

Figure 10. The ion flux is monitored over the total test duration andcombined with the exposure time to calculate the ion fluence. Timeis found to be approximately linearly proportional to the fluence dueto the relatively unchanging ion flux.

Figure 11. The differential sputtering yields for micro-architecturedmolybdenum are shown at three different times in comparison withthe flat molybdenum results. The fluences for t=0.16, 3.31, and17.0 h are F = ´1.2 1020, ´2.6 1021, and ´1.4 1022 -ions cm 2

respectively. The final dataset at t=17.0 h, or F = ´1.4 1022

-ions cm 2, is fit to the modified-Zhang (MZ) profile.

9

Plasma Sources Sci. Technol. 26 (2017) 065002 G Z Li et al

dependence at small fluence and give a flat angular depend-ence, up to angles of about 50°. As the surface is eroded, theangular dependence approaches the classical picture, wherethe competition between the two physical mechanisms pro-duce a marked maximum around 50°.

By integrating the differential sputtering profiles, the totalsputtering yield is obtained as a function of time, or fluence,as shown in figure 12. The sputtering yield begins atapproximately 0.24 atoms/ion when the micro-features are allintact. As the surface is sputtered, the yield is measured toincrease until the structures are entirely eroded, leaving a flatsurface sputtering yield of around 0.42 atoms/ion. The con-verged yield closely matches the value measured for the flatmolybdenum tests reported in the previous section, and thesmall deviations are well within the error bars. Since theincident plasma has a Gaussian profile, the center of the targetis eroded at a significantly higher rate than the edges.Therefore, the surface features at the edges may continue toerode and the sputtering yield will increase slightly, althoughthe change will likely be much smaller than the error bars ofthe measurement.

Contamination of the surface layer by oxidation canstrongly affect the surface structure evolution and reduce themeasured sputtering yield. To ensure the ballistic depositionof the surface is due to the sample material, EDS spectra areacquired to provide a measure of the gross contamination.[27] shows that the surface composition was initially 7.58%oxygen by weight as a result of the fabrication process. Aftertesting, the oxygen content was reduced to 1.93%, indicatingthat the surface was cleaned during plasma exposure. RGAmeasurements taken prior to testing revealed water and oxy-gen partial pressures of 10−7 Torr. At such low partial pres-sures and large ion fluxes, the oxide monolayer formed at asample temperature of 500 °C can be expected to be eroded

away relatively quickly compared to the multi-hour exposuretime, minimizing the presence of dynamic contamination.Additionally, the erosion scales are on the order of hundredsof microns, well beyond the scale of any surface con-tamination, which would be on the order of around 10 nm.Therefore, in situ measurements of dynamic contaminationare outside the useful scope of the present experiment.

Matthes et al [27] present SEM images showing thesurface feature evolution of this sample, and compares it toseveral other micro-architectured samples that were tested inour facility. The images were taken between exposures andshow how the surface structures erode under ion bombard-ment and grow via ballistic deposition. After the final plasmaexposure, they note that the effective exposure area of thesample ( <d 1.5 cm) is essentially flat, with lightly erodedstructures just outside the center that transition to almostuntouched structures at the edges, a result expected for theGaussian exposure profile. These observations show theoverall flattening of the micro-architectured surface, whichexplains the convergence of the sputtering yield to the flatsample result.

5. Conclusions

We have demonstrated time-resolved measurements of ion-induced sputtering for studying the plasma-material interac-tions in advanced plasma-facing materials. In particular, wehave introduced QCM measurements in a focused plasma as anovel method of measuring time-resolved sputtering yield forsputter-resistant samples requiring large ion fluxes. The QCMtechnique was successfully benchmarked on a flat molybde-num sputtering experiment, and the QCM’s capability tomeasure time-resolved sputtering yield was subsequentlydemonstrated on a micro-structured molybdenum sample.This demonstration enables future testing of micro-archi-tectured materials with different surface morphologies in thePi facility.

We found that the angular sputtering profile for themicro-architectured material evolved from a flattened profileto one peaked at 50° which closely resembled the profile forthe flat molybdenum sample. The shadowing effects of themicro-architectured surface features are shown to play a verysignificant role in modifying the classical picture of a cosine-shaped sputtering distribution with normally incident ions.The total sputtering yield for the micro-architectured samplewas found to approach those of the flat sample due to thereduction of surface roughness after a total fluence of about

´ -1.4 10 ions cm21 2, after which the roughness is largelyeroded away.

Acknowledgments

This work was made possible by AFOSR Grant Nos.FA9550-14-10317, FA2386-13-1-3018, FA9550-11-1-0282,FA9550-16-1-0444 the UCLA Henry Samueli School ofEngineering and Applied Sciences, the National Science

Figure 12. The time-dependent sputtering yield for 300 eV argonions incident on micro-architectured molybdenum is shownconverging to the flat molybdenum result. The gray lines signifyerror bars for the flat sample measurement.

10

Plasma Sources Sci. Technol. 26 (2017) 065002 G Z Li et al

Foundation (NSF), and the Department of Defense (DoD)under the National Defense Science and Engineering Grad-uate Fellowship (NDSEG) program. The authors would alsolike to thank Ben Dankongkakul, Lucas Garel, and JohnHayes for assisting with experiments.

References

[1] Wirz R E, Anderson J R, Goebel D M and Katz I 2008 IEEETrans. Plasma Sci. 36 2122–9

[2] Peterson P, Manzella D and Jacobson D 2003 39th AIAA/ASME/SAE/ASEE Joint Propulsion Conf. and Exhibit(Huntsville, AL, 20–23, July 2003) AIAA-2003–5005

[3] Goebel D M, Jorns B A, Hofer R R, Mikellides I G and Katz I2014 Pole-piece interactions with the plasma in a magneticallyshielded Hall thruster Propulsion and Energy Forum pp 1–7

[4] Sekerak M J, Hofer R R, Polk J E, Jorns B A andMikellides I G 2015 Wear testing of a magnetically shieldedHall thruster at 2000 s specific impulse 34th Int. ElectricPropulsion Conf. pp 1–37

[5] Brooks J N, Allain J P, Whyte D G, Ochoukov R andLipschultz B 2011 J. Nucl. Mater. 415 S112–6

[6] Gamero-castaño M and Katz I 2005 Estimation of Hall thrustererosion using HPHall Proc. Int. Electric Propulsion Conf.2005 (IEPC05) pp 1–10

[7] Huerta C E, Matlock T S and Wirz R E 2016 J. Appl. Phys. 119113303

[8] Yamamura Y and Tawara H 1996 At. Data Nucl. Data tables62 149–253

[9] Nakles M R 2004 Experimental and modeling studies of low-energy ion sputtering for ion thrusters PhD Thesis VirginiaPolytechnic Institute

[10] Doerner R P, Whyte D G and Goebel D M 2003 J. Appl. Phys.93 5816–23

[11] Matlock T S, Goebel D M, Conversano R and Wirz R E 2014Plasma Sources Sci. Technol. 23 025014

[12] Stuart R V and Wehner G K 1960 Phys. Rev. Lett. 4 409–10[13] Rubin B, Topper J L and Yalin A P 2009 J. Phys. D: Appl.

Phys. 42 205205

[14] Yalin A P, Williams J D, Surla V and Zoerb K A 2007 J. Phys.D: Appl. Phys. 40 3194–202

[15] Chu E, Goebel D M and Wirz R E 2013 J. Propulsion Power29 1155–63

[16] Goebel D M, Jameson K K, Katz I and Mikellides I G 2007Phys. Plasmas 14 103508

[17] Matlock T S, Conversano R W, Goebel D M and Wirz R E2013 The plasma-material interactions facility at UCLA 49thAIAA/ASME/SAE/ASEE Joint Propulsion Conf.

[18] Steinbruchel C 1990 J. Vac. Sci. Technol. A 8 1663[19] Sauerbrey G 1959 Z. Phys. 155 206–22[20] Bachmann L and Shin J J 1966 J. Appl. Phys. 37 242–6[21] Matlock T S, Dodson C A, Goebel D M and Wirz R E 2015

Measurements of transport due to low frequency 34th Int.Electric Propulsion Conf. pp 1–10

[22] Som T and Kanjilal D 2012 Nanofabrication by Ion-BeamSputtering: Fundamentals and Applications (Boca Raton,FL: CRC Press) pp 1–40

[23] Zhang Z L and Zhang L 2004 Radiat. Eff. Defects Solids 159301–7

[24] Ferron J, Alonso E V, Baragiola R A and Oliva-Florio A 1981J. Phys. D: Appl. Phys. 14 1707–20

[25] Mahadevan P, Layton J and Medved D 1963 Phys. Rev. 12979–83

[26] Martínez-Sánchez M and Ahedo E 2011 Phys. Plasmas 181–10

[27] Matthes C S, Ghoniem N M, Li G Z, Matlock T S, Goebel D M,Dodson C A and Wirz R E 2017 Appl. Surf. Sci. 407223–35

[28] Li G, Dodson C A, Wirz R E, Matlock T S and Goebel D M2016 Implementation of in situ diagnostics for sputter yieldmeasurements in a focused plasma 52nd AIAA/SAE/ASEEJoint Propulsion Conf. pp 1–13

[29] Weijsenfeld C H, Hoogendoorn A and Koedam M 1961Physica 27 763–4

[30] Laegreid N and Wehner G K 1961 J. Appl. Phys. 32 365–9[31] Zoerb K A, Williams J D, Williams D D and Yalin A P 2005

Differential sputtering yields of refractory metals by xenon,krypton, and argon ion bombardment at normal and obliqueincidences IEPC pp 1–25

[32] Küstner M, Eckstein W, Dose V and Roth J 1998 Nucl.Instrum. Methods Phys. Res. B 145 320–31

11

Plasma Sources Sci. Technol. 26 (2017) 065002 G Z Li et al