Dual Magnetron Sputtering of Aluminum and Silicon Oxides for Low Temperature, High Rate Processing Christopher Merton and Scott Jones, 3M Corporate Research Lab, St. Paul, Minnesota, USA and Doug Pelleymounter, Advanced Energy Industries, Inc, Northfield, Minnesota, USA Abstract A technique for depositing non-conductive oxides using bi-polar pulsed DC power supplies and planar targets has been shown to achieve deposition rates significantly higher than those from more standard Mid Frequency (MF) AC sputtering techniques. This method involves the use of two bipolar pulsed DC power supplies each connected to a separate cathode and a common anode. In this manuscript, the use of this technique to reactively deposit aluminum oxide and silicon oxide from rotary cathodes using metallic targets will be reported. Hysteresis loops using voltage control of the bi-polar pulsed DC power supply, deposition rates, film properties and process conditions such as system temperatures are presented and compared to those from a standard MF AC deposition process. Background The sputter deposition of oxides usually involves sputtering of metallic targets using a reactive gas and mid frequency (MF) alternating current (AC) power, typically 40 khz. The sinusoidal AC voltage is applied between the two magnetrons which are both isolated from ground and the chamber. This technique has the disadvantage in that 1) high levels of substrate heating limit power levels achievable with polymer webs [1] and 2) requires the use of two magnetrons operating for coating purposes at 50% duty cycles which limit achievable deposition rates without increased machine sizes. To reduce substrate heating, this AC sputtering technique can be altered with the use of a bi-polar direct current (DC) power supply, such as the Advanced Energy Ascent AMS magnetron sputtering supply with an Ascent Dual Magnetron Sputtering accessory (AMS/DMS), to apply a square wave potential between the two magnetrons. Lower substrate temperatures are obtainable with this technique however deposition rates are still limited due to the cycling of the magnetrons as either an anode or a cathode. An alternative method called Dual Reverse Pulse (DRP) has been recently reported [2] with Figure 1 showing a schematic of the system configuration for the DRP technique. With the DRP technique, a floating anode is placed in the chamber while the output of a first bi-polar DC power supply (AMS/DMS) is attached to the first magnetron (output A) and the floating anode (output B). A second bi-polar DC power supply is attached to the second magnetron (output A) and the same floating anode (output B). The phases of the two bi-polar DC supplies are synchronized so that the two magnetrons have the same polarity throughout the cycle and the floating anode is at the same potential relative to each magnetron. In addition to lower substrate heating with the use of DC power, the DRP technique is a potential route to obtain higher deposition rates since each magnetron is sputtering with the duty cycle of the DMS (typically 80%) as compared to sputtering 50% of the time when AC or bi-polar DC is used. In particular, if the cathode power or current is the limiting factor for deposition rate, the DRP technique might be especially appealing.
Figure 1: System configuration for Dual Reverse Pulse (DRP) sputtering. The previous report [2] studied reactive deposition of TiOx from two planar targets with the floating anode placed between the magnetrons in a collinear configuration. The bi-polar DC power supplies were run using the power control mode in the full oxide region of the hysteresis loop. In this current work, the use of the DRP technique has been expanded with application to a web roll to roll coater using rotary magnetrons to reactively deposit AlOx and SiAlOx in the transition region of the hysteresis loop. Experimental For these studies, a R&D roll to roll web coater with a base pressure of 3 x 10-7 Torr was used with chamber pressures of 5.3 mtorr during film depositions. The targets used in these experiments were 6 diameter, 890 cm long, flame sprayed metallic Al (99.99%) or SiAl from Soleras Advanced Coatings. Films were deposited onto a 2 mil one side matte PET substrate. For comparison between techniques, the system was used to deposit AlOx and SiAlOx reactively with MF AC (Advanced Energy Crystal 60 kw power supply) at 16 kw or 8 kw power and the DRP technique with 6-8 kw from each bi-polar DC power supply (12-16 kw total power) set for 40 khz. For the DRP technique, an 80% duty cycle of negative voltage to the magnetrons and 20% to the common floating anode was used. For the DRP technique, a dedicated water cooled floating anode was designed and placed in the system so that it was located between the two rotary targets as shown in Figure 2. The floating anode is in the shape of a long rectangle with a chamfered top edge and its top located approximately at the target centerline. Electrical contact was made between the anode and an electrical feedthrough in the chamber wall using a solid copper bus bar.
Figure 2: System configured for DRP technique. Water cooled floating anode located between rotary targets. The end cap shields are visible on the targets along with the water fitting plastic elbow for the floating anode. Results One of the features of the Advanced Energy AMS/DMS is called FastDAQ which is a built in logging/recording of the power supply outputs. The samples are recorded every 0.2 microseconds, and can be viewed real-time with the Advanced Energy Virtual Front Panel or easily exported. Using the FastDAQ feature, the effect of changing the boost value of the DMS was explored. The boost value is related to a purposeful voltage overshoot during the voltage transition that is designed to square up the current (i.e. raise the current faster at the transition). Figure 3 shows typical plots of output voltage as a function of time for aluminum in metal mode. Since the floating anode did not have any magnetic confinement like the cathodes, the two sides of the output for each power supply are inherently different and result in different sputter voltages. The sputter voltage A (target) and sputter voltage B (anode) are both lower in oxide mode compared to metal mode but with a fixed boost level, the difference between the sputter voltage A and sputter voltage B remains approximately the same. The lower voltage for the oxide mode is typical for what is observed in a hysteresis curve [3]. Table 1 summarizes the voltages of each output and microarc level (microarc/sec) for 5% and 50% boost for sputtering aluminum target in both metal and oxide modes. Because of the lower sputter voltage asymmetry between the A and B output and the lower microarc rate on the B or anode side, a boost of 5% was found to be more beneficial and was chosen to be used for most of these studies. We recognize that with the lower boost percentage, less squaring of the current at the pulse start will lead to slightly lower deposition rates. However plasma instability/arc handling can also lower rates so there is some trade-off between these two effects.
Figure 3: FastDAQ plot of sputter voltage A (target) and sputter voltage B (anode) in metal mode for 5% and 50% boost. Table 1: Sputter voltage and microarcs for 5% and 50% boost for metallic and oxide modes. Mode Boost (%) sputter voltage A (target) (V) sputter voltage B (anode) (V) difference (V) microarc/sec (target) microarc/sec (anode) metal 50-571 -813 242 0.25 12.70 oxide 50-382 -625 243 0.20 0.48 metal 5-576 -721 145 0.41 3.98 oxide 5-378 -490 112 0.29 0.14 AlOx To establish a baseline for film properties, a hysteresis loop was run using 16 kw AC with optical characteristics measured using a J.A. Woollam alpha-se ellipsometer. Analysis was done in the down-web direction and the backside of the substrate was roughened with fine-grit sandpaper prior to measurement to eliminate the second surface reflection. Film samples were fit using a Cauchy layer for the oxide and a model of the PET substrate created by J. Hilfiker (Woollam). The optical properties n (index of refraction at 632.8 nm) and d (thickness) were calculated and shown in Figure 4 along with a plot of voltage vs O2 flow for AlOx films.
Figure 4: Hysteresis loop plot of voltage as a function of O2 flow for standard AC deposition of AlOx at 16 kw. Index of refraction (n) and thickness (d) are displayed for selected points. As expected, with the AlOx films made at a fixed time, the lower the voltage on the hysteresis curve yielded a lower deposition rate. It is important to note that the rate coming up the backside of the curve is lower than the rate when coming down the front of the curve. This behavior agrees with earlier work done characterizing reactive AlOx [3]. The lower index for the lower deposition rates (thinner samples) are related to different Si/O ratios and/or denser films. For the DRP technique, a similar hysteresis loop is shown in Figure 5 where the sputter voltage A (target) is plotted versus the O2 flow with index of refraction and thickness measured for select AlOx films. The power output from each AMS/DMS was set for 8 kw, so the total power to the pair of cathodes was 16 kw. The curve is not as complete as the AC example since the control loop was not optimized for the DRP process. The sputtering voltage as run was not stable in the transition region (from metal as low O2 to oxide at high O2 flow of 130 sccm). Further process control loop optimization or use of an external control system (such as a plasma emission monitor) should allow the DRP technique to operate in the transition region. For the DRP technique on the backside of the hysteresis loop (decreasing voltage) from oxide toward metal, the sample film thickness measured between 24 and 36 nm. For the same region of the hysteresis loop, the 16 kw AC technique samples had a thickness range from 26 31 nm demonstrating similar deposition rates for the same total applied power. Samples were also made at 8 kw AC in the same region of the hysteresis loop (decreasing voltage) and ellipsometry measured AlOx film thickness 15 20 nm. This demonstrates that for the same peak power to the cathode, the DRP technique has a higher deposition rate than AC. Since the deposition rate depends on the operating point of the hysteresis loop and there is currently not a precise way of determining the equivalent operating point between the two techniques, only a general estimate of the DRP rate being in the range of 120 200% of
AC can be made. If a system is limited by the peak power to the cathode, even a 20% increase in deposition rate is meaningful. Figure 5: Hysteresis loop for DRP at 16 kw total power plotted as sputter voltage A vs O2 flow. Select data points show measured index and thickness. X-ray Photoelectron Spectroscopy (XPS) was also done on selected AlOx samples from both the AC technique and the DRP technique with Figure 6 showing the composition as a function of depth for DRP sample A (sputter A voltage of 424 V) and sample B (sputter A voltage of 373 V). These show a relative thickness of 58% which agrees closely with the ellipsometry data of 24 nm / 36 nm = 67%. The Al:O ratio is 0.53 for DRP sample A and 0.51 for DRP sample B, while for AC sample A the Al:O ratio is 0.57 and for AC sample B is 0.55. This shows that for both techniques, more O relative to Al (smaller Al:O ratio) is observed at the lower voltages, as expected. A reference sapphire (Al2O3) sample had a XPS Al:O ratio of 0.484. Full XPS spectra for the DRP samples do not show any indication that other metals are present above the detection limit of XPS of 0.1% atomic %. This supports the idea that the anode made of stainless steel is not being sputtered to any significant degree due to the lack of any magnetic field at the anode to confine the electrons.
Figure 6. XPS depth profiles of DRP Sample A and DRP Sample B. Assuming a uniform sputter rate, sample B is approximately ~58% the thickness of sample A. The optical transmission from 400 900 nm wavelength was measured inline during AlOx coating and is show in Figure 7 as a function of voltage for both techniques. The transmission was normalized by dividing the transmission of the coated substrate by the transmission of the uncoated substrate. The transmittance was the same for the DRP technique as the AC technique for full oxide region samples. The similarity of the transmittance curve in the transition region for both techniques is unexpected given the different voltages when in metal mode.
Figure 7: Transmittance of AlOx on PET. SiAlOx Applying the DRP process for SiAlOx, it was found that in order to achieve stable plasma control using a PEM, the degree of pumping of the process area had to be significantly higher. Figure 8 shows hysteresis loops using DRP or AC sputtering techniques with the AC sputtering done with standard pumping conditions (red) and higher pumping conditions (orange and green). Under the two pumping conditions, the Ar flow was adjusted to obtain a chamber pressure near 2.5mT in the metal mode. The data was taken with increasing oxygen flow unless marked decrease in which the data was obtained by lowering the oxygen flow. Relative light transmission data was taken for samples made under the high pumping conditions and shown in the figure. Comparing the data for the AC sputtering, a strong shift of the transition region of the curves to higher oxygen gas flows is seen with the use of the higher pumping conditions that gave easier control of the Figure 8. Hysteresis loops for SiAlOx depositions using DRP (2 cathodes at 8 kw each) and AC (16 kw) sputtering using different pumping conditions.
Anode Temperature ( C) process. The higher flows are needed to maintain the oxide formation on the target and substrate with the increased diversion of oxygen towards the pump and away from the deposition areas. In addition, the transition is much sharper, (less of a kink) under the higher pumping conditions leading to potentially an increased difficulty in process control in the transition region, although control of the process throughout the transition region of the curve was achievable in these studies. For the higher pumping, AC sputtering conditions, a small hysteresis is visible. Comparing the AC sputtering curve under high pumping conditions to the DRP sputtering curve under the same pumping conditions, a transition at lower O2 flows is observed for the DRP process while the drop in the transition region is of similar steepness. Plasma Heating The previous work with DRP using TiOx [2] showed a decrease in heat load to the substrate for the same deposition rate when compared to standard AC. Due to the physical constraints of the web coater, it was not possible to measure the heat load to the substrate using similar instrumentation. However, the temperature of the top gas bar shield was measured with a thermocouple during sputtering using both techniques and it reached a comparable 93⁰ C. When using the floating anode for the DRP technique for AlOx, a thermal couple was attached to the side of the anode and recorded a temperature as high as 230⁰ C after 170 minutes of operation (220⁰ C after 130 minutes). Figure 9 displays anode temperature data during SiAlOx depositions showing an initial steep rise in temperature followed by a slower linear increase with time. After two hours of operation at both power levels, no stabilization of the anode temperature was observed suggesting the need for increased (significant amounts of) water cooling of the anode. The relatively high temperatures and the fact the plasma stayed ignited during the oxide depositions supports the concept that the anode stays hot and therefore relatively clean of oxide deposition. Only a small coating was observed on the anode after deposition runs, so further work will be needed to demonstrate long term stability and potentially optimizing cooling by adjusting water flow to the anode. 225 200 175 150 125 100 75 50 6 kw 8 kw 0 10 20 30 40 50 60 70 80 90 100 110 120 Time (min) Figure 9. Anode temperatures during SiAlOx depositions at different applied DRP Powers.
In addition to the previously reported reduced heat load to the substrate [2], the DRP technique uses DC 80-90% of the time to the targets and should have less inductive heating of the target and magnetron components compared to using AC. An indirect measurement of the inductive heating and heat load to the target is the heat dissipated to the cooling water of the magnetrons. The supply and return water temperature for each magnetron was recorded for each technique and showed 4⁰ C change for 16 kw total power DRP and a 5⁰ C change for 16 kw AC. The DRP anode return water temperature was not measured. Recent work on heat dissipation in cathodes during reactive sputtering [4] warrants additional work with particular attention to where on the hysteresis loop each technique is being run before firm conclusions can be drawn between the heat loads to the magnetrons for each technique. Summary The DRP technique has been successfully applied to a pair of rotary cathodes using a floating anode located between and slightly below the targets. Both AlOx and SiAlOx were reactively deposited in the transition region of the hysteresis loop. The AlOx deposition rate, index of refraction and composition as determined by XPS were comparable to standard AC when the total applied power to the magnetrons is the same and operating at same position on the hysteresis loop. For the same peak power, the DRP technique has a higher deposition rate than AC, which could prove useful in cases were the rate is limited by peak power (or current) to the magnetrons. Acknowledgements The authors would like to thank Advanced Energy Industries, Inc and 3M Corporate Research Lab management for supporting the experiments, Brandon Pietz of 3M for the ellipsometry, Ali Rafati of 3M for the XPS and Don McClure for helpful discussions. Advanced Energy, Ascent, Dual Reverse Pulse, Crystal, and FastDAQ are trademarks of Advanced Energy Industries, Inc.. alpha-se is a trademark of J. A. Woollam Co., Inc. References 1. H. Proehl, Sputter Coating on Polymer Film - Strategies for Reduction of Substrate Heat Load Proceedings of AIMCAL 2015 and J.R. Plaisted Effects of the anode configuration on substrate heating in dual magnetron sputtering http://www.advancedenergy.com/upload/file/white_papers/sl-white3-270-01.pdf. 2. D.R. Pelleymounter, Raising the Bar on Reactive Deposition Sputter Rates, 58th Annual Technical Conference Proceedings of the Society of Vacuum Coaters, pp 218-222, 2015. 3. P. Morse, B. Meredith, B. Rooney, R. Lovro, M. Rost, J. German Practical Limitations for the Bi-polar Pulse Sputtering of Al2O3 With Rotary Magnetron Targets Using a Reactive Voltage Controller, 56th Annual Technical Conference Proceedings of the Society of Vacuum Coaters, pp 1-5, 2013. 4. D.J. McClure, C. Bedoya, E.J. Anderson, Cathode heating in reactive magnetron sputtering: A previously ignored hysteresis loop provides a direct determination of the secondary electron yield, AIMCAL Web Coating and Handling Conference, Oct 25 28, 2015 or SY-1, D.J. McClure, C. Bedoya, E.J. Anderson, Cathode Heating Hysteresis in Reactive Magnetron Sputtering: A Path to Accurate Values of the Ion-Induced Secondary Electron Yield, SVC 2016