Characterization and Performance of Multiple Gridless Ion Sources for Wide-area Ion Beam Assisted Processes Applications

Transcription

1 Characterization and Performance of Multiple Gridless Ion Sources for Wide-area Ion Beam Assisted Processes Applications L. Mahoney, T. Alexander, and D. Siegfried, Veeco Instruments Inc., Fort Collins, CO ABSTRACT Process developers often require the use of multiple gridless ion sources to produce high current flux densities over wide surfaces to facilitate high-rate etch, pre-clean or deposition steps. In this work we provide detailed output characteristics and performance of multiple gridless sources when grouped or arrayed within a single process system. Specifically, downstream ion beam properties in argon and oxygen are reported for both near-zone (10 to 30 cm) and far zone (75 cm to 100 cm) distances. The results illustrate the degree to which the operation of high current flux density gridless sources collectively interact, particularly with respect to operating stability, net beam profiles at the work piece and beam neutralization settings. Spatial mapping of output beam characteristics will be presented for sources that are linearly arrayed or clustered to distribute the beam output as may required for treatment of wide surfaces ( cm). We also empirically examine how well the measured output of any individual gridless ion sources can be superimposed to predict the net performance of the arrayed sources when driven in tandem. INTRODUCTION Gridless end-hall ion sources have been widely used for assisted deposition process including wide-area pre-treatment and assisted deposition of high-index optical films, iridescent and decorative coatings and protective tribological coatings. These processes are typically preformed within batch-loaded coating systems using large orbital domes or other rotational substrate fixtures to facilitate uniform process coverage from relatively localized output of the deposition source (e.g. e-beam evaporator, deposition precursor source) and an assisting ion source. In many of these applications, gridless end-hall ion sources are favored over other high-energy linear sources, such as large area gridded ion sources [1] or gridless anodelayer sources [2] because of their high current flux density (> 1 ma/cm 2 ), relatively low mean ion energy (<E> from 50 to 200 ev) and high input gas utilization. The low energy features end-hall source has a particular advantage as it allows workers to operate near the sputtering threshold of materials with high current flux. This allows one to achieve high pre-treatment etch rates for adhesion or fine control of other deposited material properties such as optical index and film stress over metal, glass, ceramic and engineering thermo-plastic substrates. Recent advances in end-hall ion source component and power control features make it possible to adapt multiple end-hall source assemblies where is it necessary to boost total beam current output and provide wide spatial coverage. Specifically, advances in end-hall ion source design, modularity and the use of indirect-water-cooling have increased gas utilization of these ion sources, reduced part count and maintenance time, and increased operational reliability [3]. Also the introduction of a new family of end-hall ion source controllers with built in closed-loop control and error detection features workers with integration, automation, troubleshooting and data management [4]. Together, these improvements make the use of multiple ion sources arranged in a collective array an appealing option for processing large surface areas. Taking advantage of the developments, we explore technical issues of scaling an array of ion sources for wide area application typically encountered in web-coatings, in-line systems or large-scale, long-duration batch processing systems. While versions of end-hall ion source arrays have been explored in earlier work on a small scale [5], we explore scaling rules and operational performance issues for an array of high-output Mark II+ ion sources (5 to 15 A of discharge current per source) as needed to provide uniform beam current flux over wide surface areas. Through modeling and empirical measurements, we explore several scaling properties including a) ability to regulate and operate individual ion sources in an array, b) assumptions of ion beam divergence and beam expansion for predicting substrate current flux, c) validity of linear superposition of multiple ion source outputs, and d) rules to predict ion source output scaling against collective gas flows, operating pressure and power settings. Also we explore manufacturing issues and risks pertinent to preventative maintenance, control features and repeatable operation of end-hall ion sources when used in the array. EXPERIMENTAL SYSTEM FOR ION SOURCE ARRAY AND TEST OBJECTIVE For this study we used a linear array of three Veeco Mark II+ high-output (i.e A discharge current) gridless end-hall ion sources. The design objective is to demonstrate uniform ion current flux dosage of 15 to 60 ma/cm (or greater) with ev mean ion energy over a process zone of about 60 cm long for a wide-area, in-line system. (The target process zone is 20 to 40 cm wide). Intended applications include either use of argon for pre-cleaning the wide-area surface or oxygen for 2007 Society of Vacuum Coaters 505/ th Annual Technical Conference Proceedings (2007) ISSN

2 assisted deposition of high-index optical films. The total flow budget is constrained to about 100 sccm in order to maintain an operating pressure between 0.1 and 1 mtorr. Figure 1 shows the experimental system used in this study. The vacuum vessel is a 1.5 m long by 0.85 m diameter cylindrical chamber supported by a 6000 standard-liter-persecond (air) cryogenic vacuum pump positioned at the midlength of the vessel. The three Mark II+ ion sources [3] are situation toward the base of the vessel and pointed upward. The source spacing D1 can be set from 20 to 35 cm and the longest source-to-substrate distance D2 is nominally 35 cm. Each ion source can be elevated and tilted off-axis in order to optimize process coverage. The ion sources use hollowcathode electron sources (HCES) to facilitate long, continuous operating times (~100 to 150 hours) at relatively high input power levels (2-3 kw per source). Each source in the array is driven by either a commercial Veeco Mark II+-HO or Mark III+ ion source controller [4]. Figure 1: Wide-area test system for characterizing multiple Mark II+ end Hall ion sources. Within the test vessel, the output ion beam is bounded by side-shields and an automated linear-motion probe system and cooling plate positioned toward the top of the vessel. These boundaries help define a working process zone that is approximately 50 cm wide by 40 cm high by 110 cm long. The ion beam current flux profiler system is positioned at 34.5 cm above the source array and moves along the width of the process zone in 0.5 cm steps. The collection optics of the profiler uses a line of 31 sampling holes in a thin stainless steel plate over a 30 cm length (1 cm separation) with total open area of ~1 cm to a common, DC-biased, graphite collector behind the plate. The profiler allows us to obtain mean beam current flux density measurements over the 30 cm width of the array and assess lengthwise uniformity of in-line beam current dosage. Given the high divergence angle of the end-hall sources, a line-of-sight vignetting function is used to account for under-sampling of ion beam current flux when scanning at far angles from any one ion source in the array. (Such vignetting effects will attenuate the reading of the true beam current flux that actually would be incident on a flat substrate from the gridless, un-collimated end-hall ion sources). By combining the profiler with the attenuating effects of vignetting, we can compare predicted beam uniformity with actual spatial measurements, and, thereby, verify how ion source placement and operation can be optimized for wide-area applications. OPERATION AND CONTROL OF THE ARRAY OF MARK II+ ION SOURCES The three ion sources were operated in tandem to determine if any input power supply interactions could be observed that would make it difficult to reliably operate the array. The preferred mode of operation was to sequentially start each individual HCES and maintain it with respective keeper supply at equally low argon flows (~5 sccm) prior to starting the ion sources. This is recommended to avoid possible interactions between the high-voltage HCES starter supply (~1500V) and other Mark II+ ion source supplies. It was determined that the sources could be reliably started and operated in tandem using either manual control (fixed input gas and discharge power settings) or auto-control modes (automatically adjusted input gas in order to maintain a desired anode current at specified anode voltage). Furthermore it was determined that each ion source controller could be operated independently with no discernable electrical interactions. Observed variation in the control/response values at the controller deviated by only +/-1 to 4% across the three Mark II+ sources packages with the largest variation related to the HCES keeper and emission power supply voltage responses. [Such variability among the HCES responses are expected given typical working component variation of the HCES tips. Such variability is tolerable since the local DC impedance-state of the HCES has extremely little physical or statistical bearing on downstream ion beam properties and process performance.] The most pronounced interaction among the grouped ion sources relates to the collective contribution of the ion source gas flows to the background pressure within the process chamber. When run separately, each single ion source requires a specific input gas flow to maintain a specific discharge impedance state (i.e. a specific discharge voltage, V D, and current, I D, level) for a given background pressure. However, the necessary gas flow for a given V D -I D source impedance state has an inverse relationship to the working background pressure. This is expected since end-hall sources have high input gas utilization and tend to run in a gas flow limited regime wherein the local charged-particle density within the source is very sensitive to variation in neutral gas concentration. Thus when the ion sources are operated in tandem within a single vacuum system, the collective total flow results in a background chamber pressure that is two to three times higher than that experience when the sources are operated individually. As a result the ion sources in the array run at lower gas flows (less 20 to 30%) for a fixed V D -I D state and will exhibit 10 to 240

3 25% lower ion beam current output when compared to the additive output of single source operation. Consequentially, one must scale the predicted or expected output of the array by the observed net reductions in gas flow. Fortunately, this sensitivity to collective background pressure only impacts the amplitude of the beam current output and has little impact on the angular distribution of the output ion beam as projected by individual ion sources in the array. ION BEAM OUTPUT FROM THE ARRAY OF MARK II+ ION SOURCES PREDICTED AND MEASURED Prediction of Ion Source Array Spatial Output Predictions of the ion beam current flux density over the planar working area are made by projecting angularly resolved beam current profile reference data taken for the Mark II+ end-hall sources at a fixed radius from the face of the source. Such reference data has been reported for the Mark II+ ion source as radial distances of 30, 75 and 100 cm [6]. Examples of such reference data, J(r, ), are shown in Figure 2 for both argon and oxygen at R = 30 cm. Comparison of J(r, ) at varying distances has shown that that to first order the current flux drops off as 1/R 2 way from the face of the source. Thus, for relatively long source-to substrate-distances, we can treat the ion sources as point output sources such that the beam flux scales as the expanding surface area away each ion source. This empirically validated property of the end-hall sources allows one to geometrically project referenced angular ionbeam current flux density data to wide-area surfaces such as planar substrates, planetary fixtures and domes, large substrate drums, or spindles. We should note that this approach only provides first order approximations of beam current flux since charge-exchange and momentum transfer collision processes will attenuate ion beam current flux at far distances. Detailed measurements at R=30, 70 and 100 cm have indicated that the attenuation of the beam current flux is more pronounced for argon than in oxygen [6]. This is expected since resonant charge-exchange processes are more pronounced within argon. Using the apparent spherical surface area expansion of the ion beam from the sources output face and additive linear superposition, we can predict the current flux density of the array as projected over wide area surfaces within the test vessel. Figure 3 shows ion current flux contour plots for argon and oxygen for D1 = 30 cm, D2 = 34.5 cm for all sources and with = 0. [These predictions also account for the 10-20% reduction in ion beam current output due to the expected decrease in total input gas flow as discussed earlier. The predictive plots at V D = 150 V and I D = 10 A for each source suggest that ion current flux uniformity can be enhanced by increasing the elevation of the end ion sources and/or their operating discharge currents. The plots also suggests that we can achieve an ion beam current flux in-line dosage of nearly 10 to 60 ma/cm over a process region of 30 cm wide by 60 cm long (or about 3.0 A of total beam current). Moreover, by appropriately scaling D1 and D2 to larger dimensions, we can scale the illustrated performance to even wider areas with the given three Mark II+ units. Figure 2: Examples of angularly resolved reference beam current density Mark II+ data in argon and oxygen at R = 30 cm and V D = 150 V for various I D = 3, 5 and 10 A. Figure 3: Contour plots of predicted ion current flux density over 40 x 60 cm work area with D1 = 30 cm, D2 = 34.5 cm, V D =150 V and I D =10 A at each source in the array: (a) argon and (b) oxygen. 241

4 Beam Profiler Measurements of the Source Array Beam profile measurements were made of the source array to validate modeling assumptions such as linear superposition of ion source outputs, projection of radial reference data to planar substrates and scaling of beam current to net gas flow and back-ground pressure. The key challenge in using beam profile data for highly divergent ion beams is anticipating the attenuating vignetting effects of the planar ion beam collection optics. To illustrate the attenuating properties of the beam profiler, each ion source was profiled when operated alone at D2 = 34.5 and with no tilt angle. These profiles are compared with projected beam profile reference data and spatially compared as shown in Figure 4. All three sources exhibit the same measured current profile and amplitude and, as expected, their off-axis mean current flux is lower than the predicted (projected) flux density due to the limitations of the collection optics of the profiler. If we multiply the projected current flux by a vignetting filter function (derived solely on the basis of line-of-sight geometry from a point-source), we can obtain a predicted attenuated probe signal that is in good agreement measured probe signal. The attenuated signal predicts a sharper attenuation when compared to measured data, but this should be expected since we assumed a point source output distribution of the ion beam output rather than a more realistic spatially distributed source from the nominal 5 cm diameter opening from the Mark II+ anode. The result validates our ability to use beam projection from angularly resolved reference data for predicting beam current flux density, and allows us to correlate beam profiler measurements to actual array output performance since we can readily account for the attenuated ion collection properties of the beam profiler. and flow controllers) are within +/-2% (+/- 1 ) of their mean value. This was relatively low ion source output variance was achieved without sub-system re-calibration or source assembly adjustments and is indicative of the performance repeatability between multiple Mark II+ sources and integrated units. Figure 5 shows a comparison of predicted beam current and current signal with the beam profiler measurements made for the ion source array with D1 = 30 cm and = 0 and each source operated at the same condition in argon. Once again, the amplitude of the predicted mean beam current profile was scaled back by ~15 to 20% over to account for the reduced input gas flow of each source when operating in tandem. We observe good agreement between the predicted signal and measured signal differences being attributed to idealized assumptions about the attenuating vignetting function of the profiler as discussed earlier. [While not shown here, similar results were found for array operation over much of the broad operating domain in both argon and oxygen. The similarity between the predicted signal and measured signal validates our ability to use additive linear superposition in order to scale the output performance the array. However, it further illustrates the complication of using beam profiler data to directly specify or verify the actual output performance of the array due to the widely distributed ion vector angle output properties that are inherent within the end-hall gridless sources. Figure 5: Comparison of predicted beam current flux, predicted beam profiler signal and actual measurements for the ion source array; all sources operating at 150V, 5A in argon with D1 = 30cm and D2=34.5 cm. Figure 4: Comparison of predicted spatial beam current flux, predicted attenuated beam profiler signal and actual measurements for individual ion sources: 150V, 3A in argon. We also observe in Figure 4 that the measured output performance of the three ion source units (including power supply The uniformity of mean beam current density was optimized over a 60 cm surface length by adjusting the end-source elevation and tilt and adjusting relative discharge currents between the sources. Figure 6 shows the comparison of predicted values and measured values. Again we obtain good agreement between the predicted and measured beam current signal. This operating condition demonstrates the array s ability to 242

5 provide upwards of 3.5 Amps of ion beam current over the 30 cm by 60 cm process zone. sputtering yield coefficients that is specific to the substrate and atomic ion [7]. In Figure 7, we multiplied the incident beam current flux as spatially and angularly contributed by each source with an empirical reported measure of the angularly dependent sputter yield coefficient for silicon dioxide [7, 8]. (The angularly dependent sputter yield was normalized to the value observed at 0 degrees of incidence.) Clearly the weighted trend line matches the etch rate profile observed for the glass substrates. Thus, by combining projected beam current properties and sputter yield coefficients for various material systems, it is possible to make first order approximations of spatial etch rates, and, thereby identify optimal ion source array placement, angles and operation to achieve good process uniformity. In a similar manner, if a user has rudimentary knowledge of process dependencies as function of ion beam current, ion energy and incident angle, it is possible to scale and adjust source placement within the array to obtain near optimal ion beam assisted or direct deposition processes uniformity. Figure 6: Predicted (a) and measured (b) beam current flux at 150 V in argon at D2 = 34.5 cm but with the end sources elevated by +5.5 cm and tilted ( =7o); source discharge currents are 10A, 9A and 10A across the array. ETCH RATE EXAMPLE FOR THE LINEAR ION SOURCE ARRAY Pre-clean etch rate studies were made on cleaned, masked glass slides at D2 = 34.5 cm along the centerline of the array to examine how well static etch uniformity correlates with beam uniformity for the arrayed end-hall sources. Figure 7 shows the resulting etch rates alongside predicted beam current profiles for the optimized end-source condition as described in Figure 6. The etch rate uniformity is +/-10% compared to the +/-1% uniformity for the predicted mean current density over the length of the array. The etch rate peaks at the mid-range distance between the middle and two end sources. The lack of direct spatial correlation between etch rate and lateral beam uniformity is not surprising since sputter etch rates are dependent on the incident angle of ions. Thus to properly predict trends in ion sputter etch rates, one must weigh the current flux density by the differential ion Figure 7: Comparison of typical glass etch rate in argon to mean beam current density and current density as weighted by angularly dependent sputtering yield coefficients on silicon dioxide. SCALING OF GROUPED MARK II+ SOURCES FOR TREATMENT OF 1 METER WIDE SUBSTRATES Using the same scaling principles discussed above, we can predict performance of other grouped arrangements of end- Hall ion sources. In Figure 8 we show how two grouped ion sources can be used to treat very large flat diameters surfaces (typically rotated) as may be required over a large optical mirror or large planetary fixtures. In this illustration, two Mark II+ sources operating at I D =12 Amps in oxygen are positioned at about 90 cm from the plane of the work piece and at radial offsets of 25 cm and 40 cm respectively away from the axis of work piece rotation. Each source is tilted inward at slightly different angles (10-35 o ) to improve total 243

6 beam current dosage and coverage. Such off-axis arrangement of the Mark II+ sources is typically done to provide clearance and optimal spacing for a central deposition source (i.e. e-beam evaporator) that is located toward the center of the process system. Typical current fluxes at these distances are on the order of 0.1 to 0.3 ma/cm 2 with total subtended beam current exceeding 2 A. Predicted spatial profiles like the one shown in Figure 8 can be used to evaluate integrated uniformity over the rotated work piece or to design beamshaping masks that may be desired to further improve the net uniformity of the process. Figure 8: Predicted beam current flux over a 1 m by 1 m surface area from two, off-axis Mark II+ end-hall ion sources (200V, 12 A, in oxygen) at source to subtract distance of about 90 cm. MANUFACTURING CONSIDERATIONS FOR ARRAYED MARK II+ END-HALL ION SOURCES Recent control and maintenance improvements incorporated in the Mark II+ end-hall ion source and its power supply make an array of sources a viable option for reliable use in wide-area manufacturing systems that may need to operate continuously. Table 1 provides a summary of manufacturing performance criterion, device maintenance periods and facilities requirements that are often of concern when gauging the risks to reliable, trouble-free operation. The typical meantime-between-maintenance (MTBM) of the Mark II+ ranges between 100 and 150 hours (depending on operating power and chemistry) and is typically limited by three key consumable parts as well as the ongoing build-up of coatings onto the ion source face and body. Given recent enhancements to the Mark II+ source design [4], rapid preventative maintenance (PM) cycle times are on the order of 10 min./source and no longer involve disruption of coolant lines or disconnection of the shielded electrical leads within the vacuum chamber. (These elements typically require inspection or servicing once or twice a year when refurbishing the process chamber.) Thus, maintenance cycle times are typically limited by time required to remove and clean process chamber liners and shields rather than maintenance of the ion sources. Table 1: Typical Mark II+ ion source preventative maintenance times, repeatability measures and control features pertinent to manufacturing. Parameter Feature Typical Value MTBM (DC hours) Typical PM Cycle Time As measured repeatability* (% variation of on axis beam current amplitude at steady state operation) Controller Features (Mark II+HO or Mark III+) Instrument recalibration Modular Anode Sub assembly, indirectlycooled [6] Consumable HCES Components Power Leads and Coolant Lines Source and HCES between Mark II+ production units between source packages (includes supply and MFCs) between Mark II+ PM cycles w/ use of spare modular anode sub-assemblies Ar: hrs (w/ graphite gas plate) O 2 : hrs typ hrs cathode tip and keeper plate Inspected once ever 6-12 months min. +/-2 % typ. (+/-2 ) +/-4 % typ. (+/-2 ) +/-2 % typ. (+/-2 ) Closed-loop control modes Source: discharge current control via flow servo or fixed flow control Cathode: HCES emission current control or Neutralization current control Visual and digital error codes and alarms for fault detection and classification Data Management: Available data logging options embedded within controller memory Controller MFCs Re-calibration recommended during firmware/software upgrades Periodic re-calibration recommended Source performance repeatability has also been studied. When including the source and its supply, the Mark II+ source package has a deviation of about +/-4% (+/-2 ) (as rated by steady-state on-axis beam current density) with no discernable variation in ion beam output shape. Historically this has been tolerable for most manufacturing pre-clean etch and ion assist deposition processes of interest. Like most plasma-assisted processes, performance variance is typically linked to common step-wise or dynamic variations in vacuum process system performance resulting from chamber wall seasoning, surface temperature variation, background moisture or other causes of variation in background pressure. Perhaps one of the more critical factors in repeatable ion source performance is accuracy of the 244

7 mass-flow-controllers (MFCs) and the expected repeatability when using multiple MFCs to support the process. However, we should note that most contemporary, wide-area process systems routinely use multiple MFCs within any one single process vessel or zone in order to control the spatial distribution of input gases and hence process uniformity. In this light, the risks of using multiple MFCs in the end-hall ion source array are no different from those routinely managed within typical wide-area production systems. One noteworthy advantage of an ion source array for wide area applications is its adaptability to different static process recipes and to dynamic regulation through closed-loop control. For instance, one may use relatively simple and robust ion beam current sensors (i.e. Faraday-cup) at offaxis locations within the process zone to sequentially adjust or track the output performance of each source in the array and/or the array s collective output. Such in-situ signals can also be used to monitor the dynamic properties of the array output for fault detection and classification, and, by means of well-conditioned automated feedback, be used to dynamically regulate input gas flow or power settings to maintain steady process performance against unexpected ion source drift, chamber seasoning effects and other potential process system variations. CONCLUSION A working array of three Mark II+ end-hall ion sources has been successfully demonstrated for scaling to wide-area processing applications. The array can provide a wide output of low energy ions with high current flux densities and inline dosage (up to 60 ma/cm over a 30 wide by 60 cm long area) in either argon or oxygen with relatively low total input gas flow (< 100 sccm). Detailed characterization of the ion beam array validates spatial scaling and linear superposition assumptions used to predict the collective output of the array based on individual ion source performance specifications. By applying these assumptions, the spatial arrangement and operation of the array may be scaled and optimized to meet desired current flux density specifications or around process performance such as spatial etch rate uniformity. This flexibility of the end-hall source array makes it possible to adapt a single array configuration to different substrate sizes and process chemistries and to dynamically regulate the output of the array to accommodate process drifts or variation. Typical MTBM of the sources may be as long as 150 hours to serve either extended batch processes times or week-long in-line system cycle times. Furthermore, new modular and robust design enhancements make the Mark II+ end-hall sources easy to maintain while retaining good repeatability of output performance between individual ion source elements. REFERENCES 1. D. Siegfried, B. Buchlholtz, D. Burtner and W. Foster, Radio Frequency linear ion beam source with 6cm x 66cm beam, Rev. Sci. Instruments 71 (2) pp , D. Burtner, R. Blaker, J. Keem, D. Siegfried, E. Wahlin, Linear Anode Layer Ion Sources with 340- and 1500 mm Ion Beams, 46 th Annual Technical Conference Proceedings of the Society of Vacuum Coaters, pp , D. Burtner, V. Zhurin, and D. Siegfried, End-Hall Ion Source Characterization at High Power, 48 th Annual Technical Conference Proceedings of the Society of Vacuum Coaters, pp , L. Mahoney, D. Burtner and D. Siegfried, A New End-Hall Ion Source with Improved Performance, 49 th Annual Technical Conference Proceedings of the Society of Vacuum Coaters, pp , H. Kaufman, J. Kahn, and R. Netherly Modular Linear Ion Source, 47 th Annual Technical Conference Proceedings of the Society of Vacuum Coaters, pp , L. Mahoney, D. Burtner, D. Siegfried and C. Dale, A New End-Hall Ion Source with Improved Performance 53rd Symposium of the American Vacuum Society, Nov. 14, K. Zoerb, J. Williams and A. Yalin, Differential Sputtering Yields of Refractory Metals by Xenon, Krypton, and Ar Ion Bombardment at Normal and Oblique Incidence IEPC , 29 th International Electric Propulsion Conference, Princeton University, Oct. 31, Differential yield sputtering rates of argon ions incident on silicon-dioxide provided by A. Yalin, Department of Mechanical Engineering, Colorado State University, Fort Collins Colorado,