Effects of Limiting Orifice (Anode) Geometries on Charged Particle Characteristics in an Analytical Radiofrequency Glow Discharge as Determined by Langmuir, Current and Voltage Probes† YUANCAI YE AND R. KENNETH MARCUS* Department of Chemistry, Howard L . Hunter Chemical L aboratories, Clemson University, Clemson, SC 29634-1905, USA Fundamental studies were performed to assess the role of detailed analytical performance characterization, but also limiting orifice (anode) diameter on the charged particle requires an amount of fundamental knowledge of the physical characteristics of a radiofrequency glow discharge (rf-GD) processes taking place within the devices.The Langmuir probe atomic emission source. Measurements of the electron and ion is a powerful tool for investigating the charged particle paramnumber densities, electron temperature, average electron eters in low temperature, reduced pressure rf plasmas.10 Several energy and electron energy distribution function were made by papers have reported the measurement of charged particle using an impedance-tuned Langmuir probe. The electrical parameters including electron number density (ne), ion number features of the rf-GD were studied by the simultaneous use of density (ni), electron temperature (Te), average electron energy voltage and current probes. Studies were focused on the effects (e ) and electron energy distribution function (EEDF) in of the limiting orifice geometry on the rf-GD operation under rf-GD sources using Langmuir probes.10–13 A computervarious conditions of operating pressure and probe sampling controlled, impedance-tuned Langmuir probe system has been position.The results show that limiting orifice diameters and designed and implemented for the diagnostic study of analytdischarge pressures have an important role in the excitation ical, rf-GD sources in this laboratory.11 conditions of rf-GD sources. In addition, rf-power generator One important set of discharge parameters in the determisystems can greatly affect electron and ion densities, and to a nation of cathodic sputtering and gas-phase energetics is the lesser extent the electron temperature, average electron energy combination of dc-bias voltage and current passing through and electron energy distribution functions.Some phenomena the discharge. The dc-bias voltages on the cathode surface are previously reported, including conditions that produce self- strongly related to the discharge processes in the sheath region absorption in atomic emission applications and the high degree very near to the cathode.Therefore, measuring the dc-bias of spatial specificity in MS applications, are reasonably voltages is a good way to investigate the fundamental discharge explained. processes within the sheath as they dictate the kinetic energies of the bombarding ions as well as the resulting secondary Keywords: Radiofrequency glow discharge; atomic emission electrons leaving the surface.A number of previous rf-GD spectrometry ; L angmuir probe ; plasma characteristics studies have reported dc-bias voltages under different discharge conditions and geometries.3,4,14–18 The measurement of the current at the cathode is of additional interest as it should Over the past few years, the radiofrequency (rf ) powered glow assist in the understanding of the flow of electrons and ions discharge (GD) has been developed as a viable analytical tool.within the plasma. Because rf-GDs typically run in a constant The primary advantages of the rf-GD over traditional analyt- power mode, most systems are monitored by in-line measure- ical techniques for analysing solid samples such as arc and ment of the cumulative forward and reflected powers. While spark spectrometries lie in the fact that the rf-GD sources can one hopes that these values accurately reflect the discharge be used in the direct bulk and surface analyses of metals and processes, these values cannot be taken literally.As such, it alloys and electrically non-conducting materials such as cer- would be instructive to measure separately the actual input amics, glasses and geological materials, without any prior power (as voltage and current components) as close as possible matrix modification. Additionally, these sources have demon- to the sample surface. The voltage and current, measured with strated high sensitivity in optical emission and mass spec- appropriate probes by an oscilloscope, can be used to calculate trometry studies, with detection limits of less than 0.1 mg g-1 the actual power delivered to the GD and allow evaluation of and excellent stability.1 In addition, the rf-GD has a wide the validity of the power values measured by the commercial dynamic range and lower degrees of matrix effects than arc systems.The theoretical prediction and experimental measure- and spark emission spectrometries, which should lead to more ment of electrical currents have been detailed by Paranjpe convenient use of the source.et al.19 and Hargis et al.,15 respectively. In this laboratory, research efforts have been focused on the The plasma characteristics of rf-GDs are affected by the development of various rf-GD sources.2,3 The devices have configuration of the electrodes and the chamber, materials of been successfully applied to the bulk and surface analyses of construction, external power circuitry, means of coupling different materials, through the use of atomic absorption, between the rf generator and electrodes, and the discharge emission and mass spectrometries.4 –9 The application and gases and working pressure.Initial results of studies, employing evolution of any spectrochemical device necessitates not only various spectrometric sampling modes, have shown that the limiting orifice (anode) geometry is one of the most important † Presented at the 1996 Winter Conference on Plasma Spectrochemistry, Fort Lauderdale, FL, USA, January 7–13, 1996.factors.4,5,9,16 For example, Lazik and Marcus9 found that Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12 (33–41) 33copper sample sputtering rates were inversely related to orifice Pearson Electronics, Palo Alto, CA, USA) is employed here diameter, while also exhibiting lesser amounts of self- to measure the effective sputtering current.This probe has absorption in AES applications. They concluded that atom been shown to have exceptional band-width and linearity, diffusion paths away from the more confined negative glow providing excellent frequency response and current inforregion were probably responsible for the lesser extent of self- mation.15 The rf power cable is passed through a doughnutabsorption. No conclusions could be made with regard to shaped disc (6.3 mm id, 10 mm thick) wherein the magnetic changes in excitation conditions that might be occurring by fields associated with the rf current induce an ac current which changing orifice diameter; which would also be a viable is sampled by the probe.Because the thickness of the RG-238 contributor to the changes in the degree of self-absorption. In power cable is greater than the aperture in the probe, a splice a seemingly contradictory set of experiments, Shick and was placed in the power line such that a metal connector box Marcus20 found that analyte ion signals in rf-GDMS increased would allow mounting of the probe about the power cable.with increases in the limiting orifice diameter for a comparable The bare Cu connector is encased in a glass sleeve to protect source geometry. Those data suggest that while sputtering the current probe from touching the hot lead. The cubic metal rates are the lowest for large orifice diameters (10 mm), the box serves to shield the Langmuir probe measurement circuitry ionization efficiencies are greatest.This paper focuses on the from interference due to the exposed connections at the current investigation of the effects of limiting orifice geometries on the probe. The output of the current probe is connected to one charged particle populations in an analytical rf-GD source by input channel of the above oscilloscope by a coaxial cable. the use of Langmuir, current and voltage probes. Direct Representative temporal waveforms of the applied voltages measurement of electron and ion number densities, as well as and induced currents are plotted with an x–y recorder electron temperatures, distribution functions, and average ener- (Yokagawa, Model 3203, Baxter Scientific Products, McGaw gies are intended to give insights into the cited analytical Park, IL, USA) for storage and subsequent waveform analysis. responses.The rf-GD source has also been described in detail previously. 9 The Ar discharge gas (purity>99.999%) passes through a flow rate meter and a needle valve, with the flow INSTRUMENTATION AND THEORIES rate fixed at a relatively low value of 30 ml min-1 throughout Since previous papers reported in detail the choice of Langmuir this work.The discharge pressure is set by varying the pumping probe theories and the automated probe acquisition system rate of the source through a bellows valve located between the employed in this study,11,12 the specifics need not be reviewed pump and the source.The rf power was supplied by an RF-5S comprehensively here. The operation of the tuned Langmuir generator (13.56 MHz, RF Plasma Products, Marlton, NJ, probe is straightforward. Insertion of a metal wire into the USA) equipped with an automatic matching network (AM-5). rf-GD and applying a voltage (±60 V) produces a response The forward and reflected powers were displayed on the front current drained from the plasma which is recorded at different panel of the supply, with the reflected power consistently bias voltages, generating an i–V curve.Based on the shape of reading less than 1 W throughout these studies. Two vacuum the curve, the charged particle characteristics may be deduced. gauges are used in tandem to monitor the base vacuum quality The simplest and most classical ‘collisionless’ theories,21–23 (usually <4 mTorr) and the operating source pressure originally put forward by Langmuir24 and then further devel- (2–10 Torr) (1 Torr#133 Pa).oped by Laframboise,25 are chosen to process the obtained Prior to each experiment, the sample (oxygen-free, hard data. A C-language program has been written for the evalu- copper) surface is polished to a mirror finish with an alumina ation of data, with the calculation methods employed being slurry, rinsed with anhydrous methanol and dried in air. The very similar to those described by Fang and Marcus.26 probe is cleaned, to remove sputtered material which deposits The computer-controlled data acquisition system is based during the course of each sputtering cycle, after each data set on an Apple Macintosh IIsi computer with an NB-MIO-16XL with dilute (10%) nitric acid, rinsed with dry methanol and interface board (National Instruments, Austin, TX, USA).A dried in air. Finally, in order to ensure proper removal of high voltage operational amplifier (PA08A, Apex residual vapors from the source and/or sample surface, a Microtechnology, Tucson, AZ, USA) supplies a voltage range 30 min plasma stabilization time is adopted to remove residual from -100 to +100 V to the probe, based on a factor of ten vapors from the system.11 amplification of the interface board DAC output.The As will be seen in subsequent sections, one of the aims of impedance tuning portion of the probe circuit consists of a this study is to measure the difference in phase angle between variable capacitor (2–25 pF) and an inductor (2.5 mH) which the voltage and the current components of the delivered power.are connected in parallel and located between the plasma and Phase differences provide a qualitative, and possibly quantitat- the voltage source. The probe is impedance-tuned by the ive, measure of changes in the plasma impedance. To confirm adjustment of the variable capacitor to affect the lowest voltage that the voltage, the current and the difference in phase are produced on the electrically floating probe.The applied voltage measured accurately, the probes were tested by simultaneously and the voltage drop (i.e., current) across a high precision applying 13.56 MHz voltage signals from a waveform function 3.34 kV resistor are measured by and stored in the computer, generator (Wavetek, Model 166, San Diego, CA, USA) through representing the necessary i–V data. An applied LabView 50 and 100 V resistors. The responses were synchronous and program (National Instruments) supports the hardware used in agreement with the sensitivities presented by the commercial to acquire the data from the probe.The time required to manufactures, indicating that phase differences in the applied acquire a complete i–V curve is less than 1 min, where 200 voltage and current could be measured without bias.15 pairs of data are typically obtained, with excellent precision Once the waveforms of the voltage and the current have (less than 7% RSD) of measurement.11 been accurately measured, the instantaneous power input to A 15100 commercial voltage probe (PM 9100, capacitive, the cathode from the RF-5S generator can be accurately 3 pF, John Fluke Mfg., Everett, WA, USA) is employed to calculated by the following equation: measure the peak-to-peak (Vp–p) and dc-bias (Vdc) voltages present on conductive samples.16 The voltage probe is connecp( t)=V(t)×i(t) (1) ted to a 100 MHz digital storage oscilloscope (Model PM3375; Philips, Eindhoven, The Netherlands), and the voltage readings where p(t), V (t) and i(t) are instantaneous power, voltage and are taken from the CRT screen.A commercial, magnetically inductive, current probe (Pearson 2878, sensitivity=0.1 V A-1, current, respectively, at time t. The actual power (P) dissipated 34 Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12in the plasma can be calculated by: P=APT 0 p(t)dtB/T (2) where T is the period (T=7.375×10-8 s in this case) of the waveforms of voltage and current.The electrical characteristics of the discharge can be examined as a function of the electrode potentials and the various regions of the plasma as shown in Fig. 1. These processes can be simplified as shown in the ‘classical’ equivalent circuit depicted in the center of Fig. 1.27 As depicted here, the electrical behavior in GD sources is determined by the resistances and capacitances of the target/ cathode (Rt and Ct) and wall/anode (Rw and Cw) sheaths and the resistance of the bulk plasma (Rp).It is well known that the sheath adjacent to the cathode is primarily capacitive in Fig. 2 Configuration of the anode orifice disc and the positioning of nature.15,19,27,28 Because of the high degree of asymmetry of the Langmuir probe. the current source design (i.e., the cathode is much smaller than the anode), the capacitance of the anode sheath is much larger than that of the cathode.27 The classical equivalent It should be pointed out that the above equations [(3)–(8)] circuit model can be further simplified for the plasma under are only strictly valid under the supposition that no inelastic study here into the form of the right-hand one in Fig. 1 (see collision reactions occur among the electrons, ions, atoms and Electrical Characteristics of the Rf-GD). The instantaneous molecules within the plasma. This obviously contradicts the voltage, current and the dissipated power can be expressed as observed spectroscopic features of rf-GDs, and a small fraction a function of time through the following equations:15,19,27,28 of the dissipated energy is consumed by the above collisions.28 Therefore, the equations are only able to describe electrical V (t)=Vdc+Vrf sin(2pt/T) (3) characteristics of rf-GD sources approximately. Nevertheless, i(t)=irf sin(2pt/T-w) (4) these mathematical predictions of electrical characteristics are useful for our understanding.If the cathode sheath and the P=0.5×Vrfirf cos(w) (5) bulk plasma are perfectly capacitive and resistive, respectively, where Vdc, Vrf, irf and w are the dc-bias voltage, the peak and no inelastic collisional reactions take place in the glow, voltage, the peak current and the difference in phase between the waveforms of V(t) and i (t) should be perfectly sinusoidal the waveforms of current and voltage, respectively. Here the in nature. The differences between ideal (sinusoidal) and actual w-value, which is always less than 90°, is dependent on the shapes of V (t) and i(t) qualitatively indicate how strongly the resistance of the bulk plasma and the capacitance of the above various inelastic collisional reactions take place within cathode sheath by: the GD source.The focus of this study is to evaluate the influence of anode tan(w)=T /2pRpCT (6) geometry on the characteristics of an analytical rf-GD source. where Rp and CT are the resistance of the bulk plasma and In order to facilitate this, six removable orifice discs, as shown the total capacitance of the system, respectively.Both quantities in Fig. 2, are employed.9 Each disc was 3 mm thick, with orifice depend on many factors such as the thickness of the sheath, diameters ranging from 2.5 to 12 mm. In an effort to minimize the area of the cathode, the discharge gas and operating the complexity of this study, a constant power of 20 W was pressure, the geometry of the cathode and the gas flow rate.employed in all experiments. Previous studies have indicated The cumulative plasma impedance, Z, can be obtained by two that while input power influences the charged particle densities, methods, either: it does not appreciably affect electron temperatures, average electron energies and electron energy distribution func- Z=Vrf/irf (7) tions.10–13 As such, it is believed that the influences of the or input power on the charged particle properties would be easy to infer based on the results of those studies.Z2=(Rp)2+(T /2pCT)2 (8) In this work, Vrf, irf and w are measurable. Therefore, Z, Rp RESULTS AND DISCUSSION and CT can be estimated by the above equations. Electron Number Density Previous studies have shown that sputtering rates are strongly dependent on the limiting orifice diameter, presumably due to higher power densities and consequently higher dc-bias voltages. 9 Fig. 3 clearly shows the influence of limiting orifice diameter and probe sampling distance (from the cathode surface) on electron densities at a working pressure of 5.2 Torr.As has been seen in previous rf-GD studies,12,13 electron densities are spatially homogeneous at distances of 5–15 mm from the surface, regardless of the limiting orifice diameter. This phenomenon is not surprising,11–13 and can be ascribed to the high electron diffusion rate in the GD environment. As will be seen in the following section, and has been observed in MS studies,20 positively charged ions are not so evenly distributed.AES and MS studies have shown that limiting orifice Fig. 1 Equivalent circuit of the rf-GD plasma. diameter and discharge gas pressure can play counteractive Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12 35Ion Number Density Ion number density is an important factor as it may influence the cathodic sputtering rate and also serves as an indication of how and where ionization reactions take place in the plasma.This latter information dictates the fact that ion sampling positions must be optimized in analytical GDMS.2,4 Fig. 5 depicts the spatial distribution of positive ions for three different orifice disc sizes at a source pressure of 5.2 Torr (the same conditions as those employed in Fig. 3). Ion densities show steady decreases with increasing sampling distance, having an approximately inverse dependence on distance.As with electron densities, this response is analogous to results of previous ion density maps in rf-GDs.10–13 In comparison with electrons, ions have larger masses, and thus their diffusion Fig. 3 Electron number densities as a function of probe sampling rates are much lower. The negative dc-bias voltage on the position for different limiting orifice diameters: A, 2.5 mm; B, 6 mm; cathode may also hinder ionic diffusion from the region of the and C, 12 mm (source pressure=5.2 Torr, rf power=20 W).cathode dark space/negative glow interface.12 Given the strong dependence of the determined ion number roles in analytical response, where the observations in those density on the sampling position, the effects of orifice diameter studies suggest variations in excitation/ionization con- and discharge pressure were evaluated at a distance of 5 mm ditions.9–11 Fig. 4 graphically demonstrates the dependence of from the cathode surface, rather than the 7.5 mm distance in electron density on working pressure and limiting orifice the electron measurements.As can be seen in Fig. 6, both diameter at a sampling position 7.5 mm away from the cathode. limiting orifice diameter and working pressure have strong The results show that orifice diameters strongly affect the way influences on the determined ion densities. The maximum ion in which the electron densities respond to changes in working densities are seen to occur at lower discharge pressures as the pressure.Electron number densitiesare seen to increase steadily orifice diameters are increased. This trend of increasing densiwith increases in working pressure for the smallest orifice ties with increased orifice diameter reflects the same obserdiameter, 2.5 mm. This response is very similar to that observed vations seen in MS studies by Shick and Marcus.20 The for analyte emission intensities for that orifice size, where influences seen for the 5 mm sampling position are similar to higher pressures produce greater analytical response even those seen for each sampling position, albeit with differences though sputtering rates decrease.A nearly opposite trend is seen for the larger orifice diameters. Each of the orifice discs with diameters between 4 and 10 mm exhibit a maximum in electron number density at a defined pressure range. This optimum pressure shifts from 8 Torr for the 4 mm diameter to 3 Torr for the largest (12 mm) orifice size.The shift to lower pressures suggests a dependence on the discharge voltage, with higher voltages tending to produce higher number densities.12 The highest number densities (7×1010 cm-3) are seen for an 8 mm diameter orifice and a pressure of 4 Torr. Owing to the homogeneity of the electron distributions in the discharge (negative glow) volume, the electron densities obtained for this single sampling position are probably a fair representation of the negative glow region. Fig. 5 Ion number densities as a function of probe sampling position for different limiting orifice diameters: A, 2.5 mm; B, 6 mm; and C, 12 mm (source pressure=5.2 Torr, rf power=20 W). Fig. 4 Electron number densities as a function of limiting orifice Fig. 6 Ion number densities as a function of limiting orifice diameter and discharge pressure (probe sampling position=5 mm, rf power= diameter and discharge pressure (probe sampling position=7.5 mm, rf power=20 W). 20W). 36 Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12in absolute values. Comparison of Figs. 6 and 4 reveals similar immediately from Fig. 7. First, the ratio shows a distinct inverse relationship with sampling distance. These variations trends as both densities show their highest values when the pressures are on the lower end of the range and the orifice in ni/ne reflect more the changes in spatial distribution of ions, as the electron densities are fairly constant as a function of sizes are large.The slight differences in the trends are attributed in part to the effect of orifice diameter on ionic spatial sampling distance. As shown previously, ion densities decrease with increasing distance, with the ion densities near the cathode distribution. It must be stressed that these density maps are not expected to be identical as the negative glow of the plasma sheath being much higher than those at the bulk negative GD for each limiting orifice.This distribution was illustrated is not charge-neutral over any spatial region sampled by the probe in a given experiment. empirically in the early rf-GDMS studies where optimum analyte signals were obtained by sampling the region of the It is necessary to point out here that both the electron and ion densities in this study are as much as 30 times higher than negative glow/dark space interface, albeit at much lower pressures than those employed here.4 The second obvious result is those presented in the two previous studies performed with the same discharge source under the same conditions of power that while the absolute values of ni and ne vary with orifice diameter, the ni/ne ratio is constant.This indicates that changes and working pressure.11,12 This discrepancy results from the different power supply systems employed in these studies. In in orifice diameter do not affect appreciably the spatial location where ionizing events take place within the plasma.This is an the present study, the complete RF-5S power supply and matching system replaces the earlier system, which consisted important point in the analytical application of the devices in GDMS. While changes in orifice diameters are often dictated of independent components including a function generator (Wavetek, Model 166), an rf amplifier (Amplifier Research, by sample sizes, no changes in ion source geometry (i.e., ion sampling distance) need be made.20 Model 50A220, 10 kHz–200 MHz frequency range), a tuner (MFJ Versa Tuner 5, Model MFJ-989C), and an in-line rf An interesting situation exists at the sampling distance of 7.5 mm, where for this set of discharge conditions the ni /ne power meter (Thruline Model 43, Bird Electronics).11,12 Parker and Marcus16 similarly observed that the component system values are approximately unity.Studies of the roles of source pressure and orifice diameter on the ni/ne characteristics were resulted in much lower atomic and ionic absorbances (i.e., sputtering rates), as well as dc-bias voltages, than the present initiated at this position as it was expected that any effects would be the most profoundhere. Fig. 8 depicts the dependence system. A comprehensive study of the roles of power supplies, coupling and plasma design for semiconductor fabrication of the ni/ne ratio on orifice diameter and working pressure. As observed in previous studies in this laboratory,12 these results applications suggests that such differences as seen here are not surprising.15 Intuitively, the larger dc-bias potentials seen here demonstrate that increases in working pressure dramatically compress the ion populations towards the sample surface, indicate better power coupling to the plasma in this case, in comparison with the former studies.It is believed that the which is probably due to the shortening of the cathode dark space as electron mean free paths are decreased.28 It is generally system employed here allows a more accurate measurement of the power delivered to the plasma.The results in this study observed that high operating pressures compress the size of GD plasmas for all geometries. The extent of the compression also indicate that the power supply system affects electron temperatures, average electron energies and electron energy here seems to be greatest for those orifices with smaller diameters, suggesting that the grounded anode orifices confine distribution functions, although to a lesser extent.the plasmas not only in terms of the surface sputtering area but also in terms of confining the negative glow. Ratio of Ion Number Density to Electron Number Density As mentioned previously, the negative glow region of a GD Electron Energy Distribution Function plasma cannot be assumed to be charge-neutral or homogeneous. These qualities not only necessitate careful spatial While the effects of orifice diameter on the charged particle densities give insights into the structure of the plasma negative optimization in GDMS, but also provide clues as to basic plasma processes.Fig. 7 depicts the dependence of the ratios glow and to spatial considerations required for analytical rf-GDMS,4 studies by Ye and Marcus12 have shown that of ion number density to electron number density (ni/ne) on the sampling distance and orifice diameter at a working factors affecting electron energetics seem to dictate analytical pressure of 6 Torr.Two definite sets of relationships emerge Fig. 8 Ratio of ion number densities to electron number densities as Fig. 7 Ratio of ion number densities to electron number densities as a function of probe sampling position and limiting orifice diameter a function of source pressure and limiting orifice diameter (probe sampling position=7.5 mm, rf power=20 W). (source pressure=6 Torr, rf power=20 W). Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12 37responses to a greater extent, i.e., fewer electrons with higher Average Electron Energy energies produce greater analytical emission than more elec- While visual inspection of EEDFs gives insights into plasma trons of lesser energy. For this reason, analytical responses are excitation efficiencies, they are only qualitative in nature and not as much suppressed for non-conductive sample analyses difficult to put into perspective in parametric studies such as as would have been predicted based on the far lower sputtering these.A better measure of the changes in the energies of the rates in comparison with metals and alloys. The dependence overall electron populations is the weighted average electron of electron energy distribution functions (EEDFs) on working energy e . As shown in Fig. 10, e values obtained at a pressures, orifice diameters and sampling distances is pro- sampling position 5 mm from the cathode surface dramatically nounced, particularly for orifice diameters.As is depicted in increase with decreases in limiting orifice diameter. For Fig. 9, the location of EEDF maxima dramatically shifts to instance, at a pressure of 2 Torr the average electron energies higher electron energies with decreases in limiting orifice for the 10 mm orifice diameter are approximately 0.6 eV, while diameters. For example, maxima are observed at energy values a value of #5 eV is found for the 2.5 mm case.As reported of 0.6 and 5 eV for the 10 and 2.5 mm orifice sizes, respectively, previously,11 average electron energies are very sensitive to at a pressure of 2 Torr. This disparity is not as severe for the working pressure.The data depicted here more clearly illustrate 10 Torr discharge pressure, where for the large orifice disc the the (now expected) enhancement in optical emission intensities maximum has shifted to 4 eV, while the first maximum for at high operating pressures and small orifice diameters.Plots the 2.5 mm orifice is #8 eV at the higher pressure. (The of e values as a function of sampling distance show a very negative-going portion of the latter is due to the bi-modal slight increase with increasing distance, as would be expected electron populations and is an artifact of the second-derivative from the EEDF data of Fig. 9. calculation procedure.26) The trend of increasedEEDF maxima with increasing pressure was observed in previous studies.12 The effect of sampling distance on the measured EEDFs is not as pronounced as the effects of orifice diameter and pressure, but there is a definite shift of the maxima to higher energy values at increasing distance.12 This shift probably reflects the lower densities of sputtered atoms at greater distances, removing effective electron energy loss mechanisms.The observed dependences of the EEDFs on limiting orifice diameters sheds new light on early observations in rf-GDAES.Lazik and Marcus9 found that the intensity of the Cu I 324.7 nm resonance emission strongly depended on orifice diameter. It was concluded in those studies that small orifice diameters (2 mm) suffered less from self-absorption effects than orifices of larger diameter, yielding resonant state emission intensities which were more than an order of magnitude greater in some instances. In contrast, emission from high-lying Cu I levels (402.3 nm) improved with increasing orifice diameter while sputtering rates decreased.The data depicted in Fig. 9 suggest that the higher resonant state emission intensities are the result of the combination of higher sputtering rates with Fig. 10 Average electron energy as a function of source pressure and the smaller orifices and the increased excitation efficiencies limiting orifice diameter (probe sampling position=5 mm, rf power=20W). (thus lower self-absorption) resulting from the higher EEDFs.12 Fig. 9 Effects of orifice disc, diameter, source pressure [(a) 10 mm, 2 Torr; (b) 2.5 mm, 2 Torr; (c) 10 mm, 10 Torr; and (d) 2.5 mm, 10 Torr], and probe sampling position (A, 15 mm; and B, 5 mm) on the determined EEDFs. 38 Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12Electron Temperature Electrical Characteristics of the Rf-GD Langmuir probe measurements have been effective in studying The electron temperature derived from Langmuir probe plots should be the most relevant of these data in terms of compari- the gas-phase results of the application of rf potential to the GD sample and its effects on charged particle densities and sons of different spectroscopic excitation sources.The results shown in Fig. 11, obtained at a 5 mm sampling distance from energies. Previous studies of dc-bias potentials have also provided supporting evidence for observed differences in sput- the cathode surface, show the dependence of electron temperatures on orifice disc diameter and working pressure. Basically, tering rates for different sample types and discharge conditions. 9,16 Questions still remain in terms of the partitioning the effects of these parameters are very similar to those seen in the e data (Fig. 10), albeit with lesser degrees of depen- of the applied rf power between the instantaneous voltage and current components, their phase relationships and the dence. It is obvious that both increasing the working pressure and decreasing the limiting orifice diameter increase the elec- impedance of the cathode dark space.It is these parameters that dictate initial energy input to the plasma and the observed tron temperature. The temperatures obtained here are generally in agreement with those obtained from other low pressure steady-state atomization/excitation/ionization characteristics. Fig. 12depicts the waveforms(captured on the digital oscillo- discharge systems,11–13,29–32 although the experimental conditions here differ from those in the antecedent studies dealing scope) of the voltage on the cathode and of the current passing through the cathode at pressures of 2 and 10 Torr for discs with low pressure inductively coupled and sputter deposition plasmas.More pronounced than for average electron energies, with limiting diameters of 2.5 and 10 mm. The first qualitative feature apparent from the comparison of Fig. 12(a) and (b) electron temperatures definitively increase with increasing sampling distances, particularly at higher pressures. with Fig. 12(c) and (d ) indicates that high working pressures not only decrease Vrf values, dc-bias voltages and currents, but A comparison of the derived average energy and temperature values across the range of discharge conditions studied here they also dramatically distort the waveforms of both the voltage and the current. This observation suggests that the provides further insights into rf-GD source operation.It should be pointed out that temperatures do not smoothly change with rf-GD becomes less capacitive in nature at high working pressure as the waveforms deviate far from their original the changes in orifice disc diameters because the actual electron energy distributions are probably comprised of multiple sinusoidal shape. This reflects the fact that the inelastic collisions/ reactions are more prominent in the cathode sheath at Maxwellian functions as shown in Fig. 9(b) and (d ). Non- Maxwellian behavior (i.e., lack of a single Maxwellian function) high working pressure than at low working pressure. In particular,symmetric charge exchange reactions between argon leads to errors in the calculation of electron temperature. Nevertheless, in most cases, the ratios of the e values to Te ions and atoms are more favorable in terms of shorter mean free paths (high Ar atom densities) and lower ion velocities are between 1.2 and 2.0, close to the value of 1.5 which is indicative of an electron energy distribution that is (low dc-bias potentials).The second observation to be made from these waveforms Maxwellian.11,12,33 If this situation is a correct representation, any of the electron energy parameters (EEDF, e or Te) can is that the discs with small diameters do not exhibit as much distortion of the current and voltage waveforms as for the be used to derive the others. As was the case for the charged particle densities, comparison large orifice diameters.This basic observation reveals that the effects of inelastic collisions are more pronounced in affecting of the EEDF, e and Te values obtained with this power supply system with those obtained with component system10,11 the current (ion) flow with the large diameter discs. This relationship follows directly that of the previous paragraph in indicates that the power supply system can affect electron energetics (each value slightly lower in this case).The influences that the large orifice discs operate at lower dc-bias potentials than the smaller orifice discs for the same applied power. Thus, are not as strong as those seen for the electron and ion densities demonstrated above (see Electron Number Density we would expect more extensive charge exchange reactions to take place under these conditions. and Ion Number Density), as was the case in studies of sputter deposition systems.16 This suggests that power supply systems The total impedance (Z) of the plasma can be directly calculated through eqn.(7). The Z-values depend only slightly will influence the optical and mass spectral characteristics as they are dependent on EEDF, e and Te values in a plasma. Fig. 12 Waveforms obtained by voltage and current probes at pressures of 2 [(a) and (b)] and 10 [(c) and (d)] Torr and limiting orifice diameters of 2.5 [(a) and (c)] and 10 [(b) and (d)] mm.(a) dc bias= Fig. 11 Electron temperature as a function of source pressure and 435 V, Vrf=476 V, irf=3.8 A, Z=125 V; (b) dc bias=461 V, Vrf=436 V, irf=4.3 A, Z=101 V; (c) dc bias=284 V, Vrf=288 V, irf=3.0 A, Z= limiting orifice diameter (probe sampling position=5 mm, rf power=20 W). 96 V; and (d) dc bias=206 V, Vrf=232 V, irf=2.8 A, Z=84 V. Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12 39on the working pressure and disc diameters, having values of An interesting aspect of these measurements is that the discharge current is very large (>2 A), comparable to those #100 V.These values follow the general trends of the Vrf values, increasing with applied rf power. Based on the directly reported for sputter deposition systems.15 In addition, the electron flow velocity is very fast in this rf-GD source (peak measured parameters, eqn. (8) can be solved to obtain the total capacitance (CT) by first rearranging eqn. (6) to solve for current#2.5 A for ne=4×1010 cm-3) in comparison with dc-GD sources (current#0.02 A for ne=15×1010 cm-3).26 Rp [eqn.(9)] and then substitution into eqn. (8). By virtue of the fact that the observed This is reflected in the fact that electrons in the rf-GD have higher energies than those in the dc-GD. This is probably the Rp=T /2p tan wCT (9) result of a larger fraction of the input energy being carried by electrons in the negative glow and less of the energy being phase angles are very close to 90° (86–90°), the tan w values (>14) make the resistive component of the total impedance consumed in the cathode sheath for rf-GDs than for dc-GDs, thus the observations that rf-GD-AES sources are more effec- very small (<1%) in comparison with the capacitive component and thus can be ignored to yield total capacitance values tive in their excitation while sputtering less material than their dc counterparts.1,2,6 This makes intuitive sense because elec- which vary from 139 to 94 pF. These values are consistent with those obtained for sputter deposition systems.15 The trons in the rf plasma are continuously accelerated and decelerated by the varying applied potential while the electrons in resistance of the plasma (Rp) is interesting as it is a measure of the input energy transferred into the bulk negative glow; the dc plasma move by diffusion and the slight potential gradient within the negative glow.28 Finally, since the current the higher the resistance to electron flow the higher the power dissipation.This determination is a major goal in this work; (normally 2500 mA) is large in rf-GD sources in comparison with the current drawn by the Langmuir probe (generally unfortunately, the difference in phase angle between the current and voltage waveforms cannot be measured with sufficient 15 mA), the operation of the probe does not appreciably perturb the bulk plasma properties so that the electron precision (#5% RSD here), making these values very prone to large errors.While small variations in the phase angle have densities, etc., should be correctly determined. little effect on the contribution of the resistive component to the total impedance, the resulting large changes in the tangent CONCLUSIONS values lead to very large differences in the calculation of the resistance. Nevertheless, Rp-values can be roughly calculated The measurements of the charged particle parameters including ne , ni, Te, EEDF and e have been made in an analytical to be in the range 1–7 V, with values of #3 V being the norm.Again, such values are in general agreement with those pub- rf-GD atomic emission source by use of an impedance-tuned, computer-controlled Langmuir probe. In addition, voltage and lished for deposition plasmas.15 The above electrical parameters are useful in the successful design and optimization of current probes have been simultaneously employed to yield new insights into the electrical features of this rf-GD.Studies rf-GD sources and power supply systems for different process plasma applications.28 Improvement in the ability to assess the here have been focused on the effects of limiting orifice geometry on the rf-GD under various conditions of working phase differences between the waveforms in this system will hopefully also add new insights in terms of accurately describ- pressure and probe sampling position.Based on the results obtained here, a number of conclusions can be reached and ing the distribution of input power between the sputtering and gas phase processes. summarized as follows: (1) The limiting orifice diameter plays important roles in The electrical parameters obtained here strongly support a simplification of the ‘classical’ equivalent circuit for rf-GD the charged particle densities in rf-GD sources. At lower pressures, both ion and electron number densities increase plasmas to the model depicted on the right-hand side of Fig. 1. First, the Ct-impedance must be much less than the Rt-value incrementally with orifice diameter. At high pressures, electron densities are observed to decrease while ion densities are not in Fig. 1 because each of the observed phase differences (as shown in Fig. 12) are close to 90°. Second, because the present affected by increasing diameters. Interestingly, the conditions that lead to optimum AES and MS signals correlate well with source design has a high degree of asymmetry, with the cathode surface area being #1/100 that of the anode, the capacitance maxima of electron and ion number densities on the pressure– orifice diameter coordinates.While there is a spatial depen- of the anode sheath should be much larger than that of the cathode.27 Quantitatively, the relative capacitances can be dence to the ratio of ion-to-electron number densities (decreasing with distance), these values are only affected at very low calculated from data of the sort depicted in Fig. 12 through the relationship:27 pressures. (2) As reported previously,11 working pressures greatly Vdc=Vrf[(Ct-Cw)/(Ct+Cw)] (10) influence electron temperature, average electron energy and EEDFs. Increasing pressures tend to increase electron tempera- On average, the capacitance of the cathode sheath is #50 times less than that of the walls over the range of discharge tures and average electron energies, also shifting electron energy distribution to higher energies. Small orifice diameters conditions applied here.This value is higher than those found for sputter deposition systems, but not inconsistent given the lead to the shifting of EEDFs towards higher energies (particularly at low pressures) with average electron energies following disparity in discharge pressures. By analogy, the anode sheath resistance, Rw, can be assumed to be infinite in comparison suit. Electron temperatures are less sensitive to changes in orifice diameter, although small orifice sizes yield higher values with the capacitive impedance of the anode sheath.27 Therefore, it can be assumed that most of the discharge current (electrons) at low pressures.(3) The data reported here provide fundamental support for passes through the Ct and Cw components of the equivalent circuit rather than Rt and Rw, and the total impedance of this the observed differences in the analytical responsivity of rf-GD spectrometries. Specifically, the preferable operation of the source is mainly determined by Ct.Therefore, the discharge processes in this source can be simplified to the form of the sources in an atomic emission mode under conditions of high pressure and small orifice diameters9 reflects a combination of right-hand equivalent circuit in Fig. 1. That is, the electrical behavior in this rf-GD source is determined mainly by the higher sputtering rates for small orifice diameters coupled with the trends of higher average electron energy and temperatures. capacitance of the cathode sheath and the resistance of the bulk plasma, without a strong need to consider the resistances This combination of effects would be expected to produce optimum rf-GD-AES response.The observed dependences of of either the cathode and anode sheaths or the capacitance of the anode sheath. MS signal intensities on limiting orifice diameter and source 40 Journal of Analytical Atomic Spectrometry, January 1997, Vol. 1212 Ye, Y., and Marcus, R. K., Spectrochim. Acta, Part B, 1996, 51, 509. pressure are also supported in these studies.20 In fact, both ion 13 Heintz, M. J., and Hieftje, G. M., unpublished work. and electron number densities tend to reach a maximum under 14 Lazik, C., and Marcus, R. K., Spectrochim. Acta, Part B, 1994, conditions of low pressure and large orifice diameter, as 49, 649. observed in rf-GDMS operation.Interestingly, these conditions 15 Hargis, P. J., Jr., Greenberg, K. E., Miller, P. A., Gerardo, J. B., are the opposite to those that produce high energy electrons. Torczynski, J. R., Riley, M. E., Hebner, G. A., Roberts, J. R., Olthoff, J. K., Whetstone, J. R., Brunt, R. J. V., Sobolewski, This dichotomy would support a mechanism by which most M. A., Anderson, H. M., Splichal, M. P., Mock, J. L., Bletzinger, P., electrons present in the negative glow are secondary electrons Garscadden, A., Gottscho, R.A., Selwyn, G., Dalvie, M., resulting from the sputtering process rather than due to gas Heidenreich, J. E., Butterbaugh, J. W., Brake, M. L., Passow, phase ionization events. M. L., Pender, J., Lujan, A., Elta, M. E., Graves, D. B., Sawin, (4) Simultaneous measurement of the applied rf voltage and H. H., Kushner, M. J., Verdeyen, J. T., Horwath, R., and Turner, current waveforms gives extended insights into the collisional T. R., Rev.Sci. Instrum., 1994, 65, 140. 16 Parker, M., and Marcus, R. K., Spectrochim. Acta, Part B, 1995, processes within the plasma. Strong inelastic collision reactions 50, 617. are seen to take place in the cathode sheath, particularly at 17 Heintz, M. J., Galley, P. J., and Hieftje, G. M., Spectrochim. Acta, high pressures. These collisions, observed as distortions of the Part B, 1994, 49, 745. respective waveforms, also seem to increase with increasing 18 Woo, J., Cho, K. H., Tanaka, T., and Kawaguchi, H., Spectrochim. orifice diameters, even for low pressure operation.These trends Acta, Part B, 1994, 49, 915. correspond to those conditions wherein low dc-bias potentials, 19 Paranjpe, A. P., McVittie, J. P., and Self, S. A., J. Vac. Sci. and thus low ion/electron velocities, exist. T echnol. A, 1990, 8, 1654. 20 Shick, C. R., Jr., and Marcus, R. K., Appl. Spectrosc., 1996, 50, 454. (5) Finally, the results of these studies demonstrate the effect 21 Chen, F. F., in Plasma Diagnostic T echniques, ed. Huddlestone, that a particular rf-power supply/matching network system R. H., and Leonard, S. L., Academic Press, New York, 1965, ch. 4. has on electron and ion densities, as well as electron energies. 22 Swift, J. D., and Schwar, M. J. R., Electrical Probes for Plasma These differences, along with differences in sputtering rates, Diagnostics, Elsevier, New York, 1971. can greatly affect the analytical performance of rf-GD sources. 23 Clements, R. M., J. Vac. Sci. T echnol., 1978, 15, 193. 24 Langmuir, I., in Collected Works of Irving L angmuir, ed. Suits, G., Pergamon, Long Island City, NY, 1961, vols. 4 and 5. Financial support of the National Science Foundation under 25 Laframboise, J. G., University of Toronto, Institute for Aerospace grant No. CHE-9420751 is gratefully acknowledged. Studies Report No. 100, 1966. 26 Fang, D., and Marcus, R. K., Spectrochim. Acta, Part B, 1990, 45, 1053. REFERENCES 27 Kohler, K., Coburn, J. W., Horne, D. E., Kay, E., and Keller, 1 Marcus, R. K., Harville, T. R., Mei, Y., and Shick, C. R., Jr., J. H., J. Appl. Phys., 1985, 57, 59. Anal. Chem., 1994, 66, 902A. 28 Chapman, B., Glow Discharge Processes, Wiley, New York, 1980. 2 Marcus, R. K., J. Anal. At. Spectrom., 1994, 9, 1029. 29 Ruzic, D. N., and Wilson, J. L., J. Vac. Sci. T echnol. A, 1990, 3 Lazik, C., and Marcus, R. K., Spectrochim. Acta, Part B, 1993, 8, 3746. 48, 863. 30 Busch, K. W., and Vickers, T. J., Spectrochim. Acta, Part B, 1973, 4 Duckworth, D. C., and Marcus, R. K., Anal. Chem., 1989, 61, 1879. 28, 85. 5 Parker, M., and Marcus, R. K., Appl. Spectrosc., 1994, 48, 623. 31 Wilson, J. L., Caughman II, J. B. O., Nguyen, P. L., and Ruzic, 6 Harville, T. R., and Marcus, R. K., Anal. Chem., 1993, 65, 3636. D. N., J. Vac. Sci. T echnol. A, 1989, 7, 972. 7 Bordel-Garcý�a, N., Pereiro-Garcý�a, R., Ferna�ndez-Garcý�a, M., 32 Walters, P. E., Chester, T. L., and Winefordner, J. D., Appl. Sanz-Medel, A., Harville, T. R., and Marcus, R. K., J. Anal. At. Spectrosc., 1977, 31, 1. Spectrom., 1995, 10, 671. 33 Lai, C., Breun, R. A., Sandstrom, P. W., Wendt, A. E., 8 Shick, C. R., Jr., Raith, A., and Marcus, R. K., J. Anal. At. Hershkowitz, N., and Woods, R. C., J. Vac. Sci. T echnol. A, 1993, Spectrom., 1994, 9, 1045. 11, 1199. 9 Lazik, C., and Marcus, R. K., Spectrochim. Acta, Part B, 1992, 47, 1309. Paper 6/06346B 10 Paranjpe, A. P., McVittie, J. P., and Self, S. A., J. Appl. Phys., Received September 16, 1996 1990, 67, 6718. 11 Ye, Y., and Marcus, R. K., Spectrochim. Acta, Part B, 1995, 50, 997. Accepted September 25, 1996 Journal of Analytical Atomic Spectrometry, January 1997, Vo