Voltage Control for Reactive Sputtering: Improving Typical Sputter Rate while Dramatically Reducing Input Power Requirements

Transcription

1 Voltage Control for Reactive Sputtering: Improving Typical Sputter Rate while Dramatically Reducing Input Power Requirements C. Gruber, J. German, M. Wade, T. Valinski, J. Valek, and B. Bednar, Cardinal CG, Spring Green, WI and Northfield, MN; R. Bruers and L. Duellman, Xcel Energy, Minneapolis, MN; and F. Morgan, K. Nauman, and D. Pelleymounter, Advanced Energy Industries, Inc., Fort Collins, CO The goals of every thin film process engineer in the world are to make consistently good films and not to get a call at :00 a.m. Once in a while, the actual order of the aforementioned goals is switched. Sometimes, management demands that more good square meters come out of the box and/or that incoming power requirements be reduced to increase margin. There is a plethora of ways to achieve both of these demands out on the market today. Some are inexpensive, and some are very expensive. Examples are the AMAT Lambda closed-loop controller, the Von Ardenne plasma emissions monitor (PEM), voltage control using high-speed MFCs, and Reactive Sputtering, Inc. s IRESS transition-mode controller. The Lambda, IRESS, and PEM systems have flexibility to control not only the usual hysteresis curves, but also to control non-usual curves such as titanium in the oxide mode. Figure 1 is an example of a usual hysteresis curve. The voltage is high in a total argon atmosphere and is low when in a full-reactive atmosphere. much that the power and current are much higher, causing big swings to occur. If major arcs occur in rapid succession, the process goes into a death spiral, and often needs to be stopped and reset. So much for more good square meters out of the box. The high-power Crystal engineers at Advanced Energy have devised a way to stably run the LF power supply in voltage control even during heavy arcing. Instead of running in the full reactive mode, which is nice and stable and SLOW, the Crystal supply can run high on the curve in transition mode with little to no surgery to the machine. But, is the Crystal solution sustainable? Can process engineers do more than dabble with it? Can they live on it? Yes. So, how is the Crystal power supply able to run in transition mode without extra and expensive components on the sputtering zone? Figure shows how the power supply reacts to an arc when in voltage-control mode and running in the transition regime. Figure 1: Usual Hysteresis Curve. For quite some time, process engineers have been fighting to get high deposition rates on doped silicon, aluminum, zinc, and tin all in the reactive mode. Luckily, all of these metals have the usual hysteresis curve. Many power-supply companies have claimed to run in voltage-control mode, so that process engineers can run high on the curve. Indeed, these power supplies are commonly used, and all of them can run in transition mode until arcing happens. When an arc occurs, the power supply reacts to it and shuts down to clear it. The atmosphere then turns rich in reactive gas, and when the supply comes back on, the parameters have changed so Figure : Voltage-Control Arc Recovery. Note that when the Crystal supply powers back on after the arc, it comes back on at full voltage and higher power for a few milliseconds to eat up the excess reactive gas. It then reverts back to the voltage set point and continues on at the desired power, keeping the deposited film stable and accurate. How is this process set up? There is no tangible feedback that an engineer can tweak. The answer is easily, with little 009 Society of Vacuum Coaters 505/ nd Annual Technical Conference Proceedings, Santa Clara, CA, May 9 14, 009 ISSN

2 to no surgery on the chamber or sputter zone. The best way to set this process up is: PRELIMINARY STEPS 1. Run the chamber with argon gas in pressure control mode.. Burn in the cathodes in argon and watch for the normally characterized peak voltage. PROCESS STEPS 1. Set the Crystal power supply in voltage-control mode.. Make certain the argon flow is correct for proper sputtering pressure. The reactive gas should not be flowing at this time. 3. Turn on the Crystal power output. The power will be very low. 4. Turn on and adjust the reactive gas until the desired power is reached. 5. Start the process. Yes, it is that easy to go down the top part of the hysteresis curve and to be in the best part of the curve. As the reactive gas is introduced, the pressure-controlled argon is reduced to keep the sputtering-zone pressure constant. As arcing occurs, the power supply will react as shown in Figure. Cardinal CG in Northfield, MN, began talking with the local power company, Xcel Energy, about energy-consumption reduction at the time Advanced Energy was introducing the voltage-control mode on the Crystal power supply. The two main power hogs on a large-area coating system are diffusion pumps and power supplies that power cathodes in the full reactive mode. The obvious plan to reduce the power consumption drawn by the diffusion pumps is to swap them for turbo pumps. This is a huge chunk of change to shell out for new pumps and for integration into the system controller. Luckily, Cardinal CG in Northfield has Crystal supplies on the coater. Cardinal CG, Xcel Energy, and Advanced Energy teamed up to get these power supplies to run high on the curve in voltage control on three films that are installed on the coating machine. These films are ZnO,, and SiO. The goal was to lower the incoming power requirements for the whole coating line, while running a high deposition rate. The first tests were with zinc, running in the oxide mode. Zinc has the usual hysteresis curve, but the metal-to-oxide difference is small, so it had to be held very steady in voltage control to keep the film thickness and quality consistent. A power-quality analyzer was connected to the three-phase input of a Crystal supply that was running zinc in the oxide mode. Thickness was determined using a Sharpie pen and a stylus profilometer. The power-quality analyzer was used only for this first set of runs to show the incoming power savings. RUNNING ZNO IN POWER-CONTROL, FULL- OXIDE MODE ZnO Power = 65 Set Point When you calculate P=I*E, P=I*E OUT. The Crystal power supply calculates TRUE POWER. This is P=I*E*Phase Angle. There is always a lag in current in relation to voltage. The power supply detects this and adjusts the output accordingly. Input Ch RUNNING ZNO IN TRANSITION VOLTAGE- CONTROL MODE: TEST 1 The next test was to run the same voltage in voltage-control mode and to get close to the same power out. The cathodes were being cleaned of the entrained oxygen from running for a week in full-oxide mode. In order to get close to the same power level, considerable argon was added. ZnO Voltage = 615V Set Point Input Ch This was the same voltage that is used in full-oxide, powercontrol mode, but with an atmosphere of 40% argon. Close to the same power seen during the initial full-oxide run, the thickness was 8 Angstroms compared to the initial 580 Angstroms. This regime was run for about one hour. 154

3 RUNNING ZNO IN TRANSITION VOLTAGE- CONTROL MODE: TEST Production schedules were screaming to get going, so one more run was made, going one step further. The cathodes were still cleaning up, and a higher-voltage set point was used to obtain a higher power level. Running at low powers while running high on the curve got close to the initial ZnO Voltage = 650V Set Point Input Ch Obviously, the cathodes were cleaning up well enough to run well and to give good deposition rate. This regime ran at about half the power input, while getting almost twice the deposition rate. This regime ran for one hour. During all the ZnO runs, arcing was observed. In all arcing events, the Crystal supply performed as advertised, cleared the arcs, and recovered back to the original process settings. RUNNING SI 3 IN POWER-CONTROL, FULL- NITRIDE MODE The same experiment needed to be run on the doped-silicon targets in the machine. No power-quality analyzer was used during these runs, as the PQA readings were already noted during the above ZnO runs. However, a run was made with a complete pumping zone set of three cathode lids to see what the pitfalls might be to run all three Crystal supplies in voltage-control mode, again, without changing any gas feeds or components. PS1 Power = 50 Set Point PS Power = 50 Set Point PS3 Power = 50 Set Point RUNNING SI 3 IN TRANSITION VOLTAGE- CONTROL MODE PS1 Voltage = 588V Set Point PS Voltage = 599V Set Point Figure 3: Comparing ZnO input power, output power, and deposition Based on the findings above, one Crystal power supply can be run in voltage-control mode while the next sputtering zone is off saving considerable incoming power because one power supply in voltage-control mode will have the same deposition rate as two power supplies in power-control mode. PS3 Voltage = 609V Set Point

4 Notice that in power-control mode, 150 of output power was used to get 56 Angstroms. At the same line speed in voltage-control mode, 83.5 of output power was used to get 196 Angstroms. Using the sputtering zones as is made for a bit of a challenge, as these old BOC machines have only one gas line for all three cathode lids. The voltage and power had to be balanced on the whole zone to get the best rate as well as a good transmission number. After considerable work, good and consistently repeatable results were obtained. RUNNING SIO IN TRANSITION VOLTAGE- CONTROL MODE SiO PS1 Voltage = 588V Set Point SiO PS Voltage = 599V Set Point SiO PS3 Voltage = 580V Set Point Figure 4: Comparing power output and deposition And, what about SiO? RUNNING SIO IN POWER-CONTROL, FULL- OXIDE MODE SiO PS1 Power = 85 Set Point In power control in full O atmosphere, a total of 55 was pumped into the process to get 348 Angstroms. In Crystal voltage-control mode, 3 was consumed to get 60 Angstroms. This test was run for three hours, and it was stable and repeatable light after light. Again, the drawback here is the gas line feed on the old machine. The process had to be run conservatively to avoid absorbing films. SiO PS Power = 85 Set Point SiO PS3 Power = 85 Set Point Figure 5: Comparing S 3 power output and deposition CONCLUSION Figures 6a, 6b, and 6c illustrate improvement in achievable results between power- and voltage-controlled processes. In these figures, deposition effectiveness is defined as the thickness of the film produced by a process for the amount of power applied to that process. 156

5 Figure 6a: ZnO deposition effectiveness. Figure 6c: SiO deposition effectiveness. Even on old BOC style machines, high deposition rates were repeatedly sustainable using the Crystal power supply in voltage-control mode without doing anything to the machine architecture. The requirement is to make sure the argon gas delivery is in pressure-control mode, as this takes a lot of variables out of the equation. Make certain that the targets are sufficiently conditioned in an argon atmosphere. Figure 6b: deposition effectiveness. It would be easier to run these tests with the newer-style compartmentalized machines on the market today, as pumping and gas flow are localized for each cathode lid. Still, these tests show that voltage-control mode is a viable option for even the old-style machines. Incoming power requirements were reduced, while depositing the same or even greater layer It is not difficult to achieve these results consistently using the Crystal power supply in voltage-control mode. THANKS AND CREDITS To all of the folks at Cardinal CG for allowing these tests on their coating machine. Production schedules can be tight and very demanding. To the Xcel Energy guys for agreeing to be part of these tests; they lend lots of credibility to this paper. To the AE guys, Ken Nauman and Forrest Morgan, who keep the data focused. To the AE MarCom group for monitoring deadlines and for content verification. 157