From mcvittie at cis.Stanford.EDU Fri Apr 2 16:49:13 2010 From: mcvittie at cis.Stanford.EDU (Jim McVittie) Date: Fri, 2 Apr 2010 15:49:13 -0800 (PST) Subject: STS2 RF Generator Overheating -- Use Forward Power Mode Message-ID: STS2 Users, The rf generator for the coil appears to be overheating. The symptom is that the matching networking stops being able to get a reasonable match after 10 min or so into a run. You can see the problem if you trace the forward and reflected power during a run. For now the solution is to keep the max power at or below 2500W to set the coil generator to "forward" mode. On Monday, I will looking at the generator cooling to see if it is the cause of the problem. Thanks, Jim -- -------------------------------------------------------------- James (Jim) P. McVittie, Ph.D. Sr. Research Scientist Paul G. Allen Building Electrical Engineering Stanford Nanofabrication Facility jmcvittie at stanford.edu Stanford University Office: (650) 725-3640 Rm. 336X, 330 Serra Mall Lab: (650) 721-6834 Stanford, CA 94305-4075 Fax: (650) 723-4659 From mcvittie at cis.Stanford.EDU Tue Apr 20 16:58:48 2010 From: mcvittie at cis.Stanford.EDU (Jim McVittie) Date: Tue, 20 Apr 2010 16:58:48 -0700 (PDT) Subject: STS2 Undate Message-ID: STS2 Users, As most of you know, we lost the 3KW source RF generator a week or so ago. To speed the repair we purchased a used generator, but we unable able to use it because firmware was not programmed to work with the DeviceNet communication which the STS2 uses. In parallel we sent the broken generator back to its manufacturer (MKS/ENI) to do an expedited repair. On getting the generator the first thing ENI told us was that the unit was a prototype and would need a number of upgrades beside replacing the shorted output transistors. Being a prototype unit is surprising since it came from STS on a new etcher. We received the repaired unit this morning. It has been installed and it is working. However, its characteristics are a bit different in that the plasma appears to be more stable. Before the light emission was constantly changing for a 3s/2s etch/dep process whereas now it is more like 2 levels of brightness, as you would expect for a two step process. In addition, the reflected power, as the process switches, appears to be less. I would like to hold off opening up the tool to users for a day, so I can improve the matching. My reasons are follows: 1. The recent generator failure was caused by excessive reflected power. Lowering the reflected power transients with better matching will keep us from experiencing the same problem. 2. Six months ago when I came back to working on the STS2, I detuned the matching network to eliminate a blinking problem we were having at the time. With the improvement to the generator, I think blinking will be less of a problem and we would be better off with better matching. 3. I expect the improved performance of the generator may affect current processes so now is the time to tweak the matching phase/mag detector since users will already have to tweak their processes. 4. I want to work out a procedure for setting better initial matching positions (tune and load) in our recipes to reduce initial reflected power at the start of runs and when process conditions are changed or stopped in the middle of a run. Thanks, Jim -- -------------------------------------------------------------- James (Jim) P. McVittie, Ph.D. Sr. Research Scientist Paul G. Allen Building Electrical Engineering Stanford Nanofabrication Facility jmcvittie at stanford.edu Stanford University Office: (650) 725-3640 Rm. 336X, 330 Serra Mall Lab: (650) 721-6834 Stanford, CA 94305-4075 Fax: (650) 723-4659 From mcvittie at cis.Stanford.EDU Sat Apr 24 05:41:39 2010 From: mcvittie at cis.Stanford.EDU (Jim McVittie) Date: Sat, 24 Apr 2010 05:41:39 -0700 (PDT) Subject: STS2 Update Message-ID: Sts2 Users, Over the last few days, I have done a lot of testing of the STS2 system to figure out what happened and how to prevent it from happening again. My conclusions as detailed below is that we have allow users to run with too high of reflective power and that there a few problems with the matching network which made things worst. As a consequence there will be new restrictions on how you can operate the tool. 1. The repaired generator is working fine except for an initial interlock fault. While my initial impression was the light emission was more uniform in time, this has not been verified so I am not sure if you will any changes in etch rates. I now expect that the rates have not changed. a. If the system has sat unused for a while, you will get an interlock fault when you start a process. The solution is to clear the error and to hit the retry bottom at the error page. Next week, I will try to solve this problem. 2. We loss the generator because of high reflective power, so more attention has to be paid to keeping the reflected power down. In particular, in all the recipes we have do a better job in setting up the plasma striking (tune and load) positions for the matching network, and we CANNOT allow ETCH FIRST recipes or steps. The reason for the etch first restriction is that SF6 is an extremely electronegative gas and as such requires higher voltages to breakdown. The matching network has a hard time quickly going from matching the breakdown or striking condition to the inductive coupling condition without generating excessive high reflective power for SF6. a. To make sure these new restrictions are followed, for the near term, I want to approve/test all recipes before they are run. Starting Monday I will remove all recipes which have not been approved. In the mean time let me know which one or two recipes you want approved. I will change the strike settings and test out the recipe for minimum reflective power. 3. Steve Vargo from STS gave me the method for strike positions: In practice, I usually set the Load position ~5% lower than settled position and Tune position ~5% higher than settled position. The settled position is typically where the Load and Tune positions stabilize during the process and as such would exhibit minimal reflected power. So when plasma is struck using these start positions the load position increases and tune position decreases to the settled values. 4. The matching network/control system has a hysteresis problem for both the load and tune strike positions. In particular for a tune setting of 40% in the recipe during the stabilation step the tune position goes to 42% if it is coming from a higher percentage, such as 60%, and it goes to the 38% position if it is coming from an initial position of 30%. So far STS has not given me a way of reducing this hysteresis. This hysteresis and the use of etch first recipes was the cause for the matching problems we were seeing before the generator failure. In addition the matching network was going into an oscillation mode so it was not able to correct itself. If we restrict ourselves to using etch first and O2 recipes, we should not see the oscillation mode. In addition, this restriction allows the matching network to correct for hysteresis problem such that on a second run of a given recipe the tune and load position will be at the correct positions when the plasma is begun. Thanks, Jim -- -------------------------------------------------------------- James (Jim) P. McVittie, Ph.D. Sr. Research Scientist Paul G. Allen Building Electrical Engineering Stanford Nanofabrication Facility jmcvittie at stanford.edu Stanford University Office: (650) 725-3640 Rm. 336X, 330 Serra Mall Lab: (650) 721-6834 Stanford, CA 94305-4075 Fax: (650) 723-4659