From mcvittie at cis.Stanford.EDU Mon Jun 17 12:15:10 2002 From: mcvittie at cis.Stanford.EDU (Jim McVittie) Date: Mon, 17 Jun 2002 12:15:10 -0700 (PDT) Subject: Report on Recent PlasmaQuest Problem In-Reply-To: <31B8005BBABAD511A71E00A0C9EEA2F52790A0@MAIL> Message-ID: PlasmaQuest Problems Here is a list of the problems reported during the week of June11 and what I think needs to be done. Poor RF tuning - (reported by Vijit Sabnis and others) It appears that the matching network controller needs to be adjusted. I suspect that the gain needs to be increased. Vitit and I tried to do these adjustment on Friday but were not successful. I have another matching network of the same type. I will try to work out the adjustment procedure off line before another attempt. In the mean time, users will have to continue to first do a rough manual tune before switching over to auto tune. Bias Voltage - (reported by Vijit Sabnis) The bias voltage read out is intermittent. We found that the bias voltage cable connector, which supplies dc voltage for the buffer amp for the DC output, was loose. We were able to fine one of the 2 missing screws on the floor under the connector. After putting in this screw, we got a stable DC reading for the rest of our time on the system. Note that Vitit did report poor etch results, which correlated with a zero bias voltage read out. I am hoping that this problem was due to chamber poisoning (see below) and not to a real plasma problem. We will have to keep an eye on this issue. Bias Voltage Calibration error - While investigating the DC bias problem, I hooked up an O-scope onto the RF feed to the cathode. For the process conditions that we were using, we found that actual DC bias voltage was 80 V while the DC readout gave a value of only 43.5V. Although I need to verify the O-scoped calibration, I suspect that the DC readout is where the error is. More work is needed on this issue. I will let you know as we find out more. No recalibrations will be done without telling the users how to translate between old and new reading. Low etch rate - (reported by Daniel Weaver and Yi Dong) - The etch rate of GaAs has dropped by 30 %. I suspect that this problem was caused by F poisoning of the chamber walls. I believe there was a long nitride etch using F based chemistry done in the chamber before the GaAs run. If this is indeed the case, the solution is to keep minimize the use of F based chemistry in the chamber and to use longer seasoning steps after the use of F chemistry. I have a new optical spectrometer that may be useful in determining proper seasoning times. Friday evening after Vijit and I had been running the std GaAs etch conditions for at least a hour, Daniel was able to etch his GaAs sample with good results. This is consistent with the argument that the chamber needed a longer seasoning. So the problem seems to be gone for now. Dummy wafers turning blue - (reported by Yi Dong) - About the time that Yi and Daniel were seeing the low etch, they also observed the dummy wafer coming out blue after their seasoning step. I was not able to see the wafer in question. I suspect that they may have been seeing what is called "black silicon". Occasionally, we see this problem in other Cl based etch tools in the lab. It is due to micro-masking on the Si surface most likely from oxide and etch products on the wafer surface. Note that in the P5000 where we see "black silicon" more often the wafers come up with a light blue color. In the P5000 etcher we get around this problem by using dummy wafers which have an oxide pattern on them. Since this problem for the PQ has not been reported to me in the past, I suspect that the chamber poisoning that was causing the reduced etch rated also played a role in this problem. Loading Problem - (reported by Vijit Sabnis) - The loading system in the PQ is not of the best design and has been heavily modified by us over the years. Because users have recently started to run the system at higher cathode temperatures, we recently changed from teflon load pins to a higher temperature polymer. Perhaps, the loading system needs some adjustments with these new pins. I will ask the maintenance tech about this issue. Jim -------------------------------------------------------------- James P. McVittie Senior Research Scientist Allen Center for Integrated Systems jmcvittie at stanford.edu Stanford University Tel: (650) 725-3640 Rm. 336, 330 Serra Mall Fax: (650) 723-4659 Stanford, CA 94305-4075 From vsabnis at snowmass.Stanford.EDU Tue Jun 18 12:22:45 2002 From: vsabnis at snowmass.Stanford.EDU (Vijit Sabnis) Date: Tue, 18 Jun 2002 12:22:45 -0700 (PDT) Subject: Report on Recent PlasmaQuest Problem In-Reply-To: from "Jim McVittie" at Jun 17, 2002 12:15:10 PM Message-ID: <200206181922.MAA00444@snowmass.Stanford.EDU> Hi: This is an update and follow up to Dr. McVittie's email based observations from my work yesterday. I apologize for the long email, but I hope that current users read this carefully and communicate their comments and concerns with all of us. 1. Poor RF tuning and lack of DC bias voltage: Problems: These problems are flaky, but since I have been using the machine heavily in the last couple of weeks, I have noticed them occurring more and more frequently. RF tuning can be difficult to achieve. Manual tuning is required. Even when the RF power is matched appropriately, the DC bias voltage reads 0 or a few volts, but at some random point can start reading the correct voltage. I hooked up the oscilloscope with high voltage probe to the RF feed to the cathode. I was able to confirm that when the computer screen and RF power supply display 0 V or a very low DC bias that there is indeed no DC bias measured on the scope. This means that there isn't a problem with the voltage readout on the computer screen or RF power supply display, but that there is a real problem with the plasma or RF being supplied to the chamber. Based on this finding we can eliminate any potential problems with the loose connector that we found last Friday. At some point during my investigation the DC bias suddenly was displayed correctly on the computer screen and RF power supply. I then confirmed the presence of the bias on the oscilloscope. I found the same results as last Friday - for a displayed voltage of -46 V, the DC bias read on the scope was about -80 to -90 V. The voltage read on the computer screen and RF power supply is about a factor of 2 smaller than what it should be, indicating a calibration problem or the presence of a voltage divider in the readout circuitry. Some other interesting observations: My recipe uses 35W of RF power. When the DC bias reads 0 V (or something close to that), the scope displays a sine wave with a 400 V peak-to-peak amplitude. When the DC bias is correct, the sine wave shifts down by this bias and the amplitude of the sine wave drops to 100 V peak-to-peak. I intermittently heard strange noises coming out of the RF matching box. When the DC bias reads properly, there is no problem with your etching. I did not have time to investigate what happens to the etch rate when the DC bias reads close to zero, but based on the above results I think you will get poor results. My guess is that something is wrong with the RF matching network and possibly the RF power supply. However, someone more qualified should confirm this. I think this concern should be taken seriously and acted on immediately by the SNF staff. 2. Loading Integrity I have experienced three wafer crashes in the last couple of weeks. This may be due to the new pins that have been installed into the machine. Please also double check that your wafers sit properly in the loading arm platen. I found that even certain prime wafers can be slighty too large and hence do not sit properly in the load arm. 3. Loading - software problem Sometimes the loading and unloading functions become out of phase with proper pin positions, which end up confusing the machine about the wafer locations. Please pay special attention to this since you can easily crash your wafer into the pins or worse yet crash the load arm into the clamp. Should you find yourself in this position I have found that cycling the load/unload functions along with system reset will eventually get the system back into the appropriate state. Using the prompt option in your recipe sometimes works. 4. The load arm sometimes wobbles too much during loading and unloading. 5. The module located just below the keyboard (MC-200 Motor Controller) is making weird noises from time to time. The fan may need to be replaced. 6. Many of us use low Cl2 flow rates (ie. < 3 sccm). I really don't believe the Cl2 flow rate displayed on the screen. I think this particular flow sensor (better yet, all of them) needs to be recalibrated. My overall conclusion is that the tool is pretty unreliable right now and has several issues which need to be resolved. YOu can still get your etches to work, but you need to be particularly careful about the RF matching/bias and loading. I hope that appropriate action will be taken immediately by the SNF staff so that I and others do not have to spend more time stuggling with the tool. - vijit Vijit Sabnis vsabnis at snow.stanford.edu Stanford University Department of Electrical Engineering Solid State and Photonics Laboratory CISX B113-23 650/725-2774 office 723-4659 fax From DWeaver at sirostech.com Tue Jun 18 12:45:47 2002 From: DWeaver at sirostech.com (Daniel Weaver) Date: Tue, 18 Jun 2002 12:45:47 -0700 Subject: Report on Recent PlasmaQuest Problem Message-ID: <31B8005BBABAD511A71E00A0C9EEA2F52873E1@MAIL> Vijit, That is a great overview of the problems on the tool. As I user I really appreciate knowing what problems other users are having. One thought concerning your item #6 (Cl MFC)-- It has been my experience in the past that for an MFC to accurately control the set flow rate, the set flow rate needs to be in the range of 20% to 40% of full flow. The exception to this rule is where the MFC is a more complex one (and much more expensive). My understanding is that the Cl MFC is a 50sccm and with calibration factor bringing it to 43sccm. That means any user, such as myself, who is using this MFC at anything below 8-9sccm is really asking for problems. If there is enough user interest it might benefit all to resize that MFC. Dan -----Original Message----- From: Vijit Sabnis [mailto:vsabnis at snowmass.Stanford.EDU] Sent: Tuesday, June 18, 2002 12:23 PM To: mcvittie at cis.Stanford.EDU Cc: pquest at snf.stanford.edu; AFischer-Colbrie at sirostech.com; DWeaver at sirostech.com; IGonzalez at sirostech.com; YDong at sirostech.com; harris at snowmass.Stanford.EDU; rcrane at snf.stanford.edu Subject: Re: Report on Recent PlasmaQuest Problem Hi: This is an update and follow up to Dr. McVittie's email based observations from my work yesterday. I apologize for the long email, but I hope that current users read this carefully and communicate their comments and concerns with all of us. 1. Poor RF tuning and lack of DC bias voltage: Problems: These problems are flaky, but since I have been using the machine heavily in the last couple of weeks, I have noticed them occurring more and more frequently. RF tuning can be difficult to achieve. Manual tuning is required. Even when the RF power is matched appropriately, the DC bias voltage reads 0 or a few volts, but at some random point can start reading the correct voltage. I hooked up the oscilloscope with high voltage probe to the RF feed to the cathode. I was able to confirm that when the computer screen and RF power supply display 0 V or a very low DC bias that there is indeed no DC bias measured on the scope. This means that there isn't a problem with the voltage readout on the computer screen or RF power supply display, but that there is a real problem with the plasma or RF being supplied to the chamber. Based on this finding we can eliminate any potential problems with the loose connector that we found last Friday. At some point during my investigation the DC bias suddenly was displayed correctly on the computer screen and RF power supply. I then confirmed the presence of the bias on the oscilloscope. I found the same results as last Friday - for a displayed voltage of -46 V, the DC bias read on the scope was about -80 to -90 V. The voltage read on the computer screen and RF power supply is about a factor of 2 smaller than what it should be, indicating a calibration problem or the presence of a voltage divider in the readout circuitry. Some other interesting observations: My recipe uses 35W of RF power. When the DC bias reads 0 V (or something close to that), the scope displays a sine wave with a 400 V peak-to-peak amplitude. When the DC bias is correct, the sine wave shifts down by this bias and the amplitude of the sine wave drops to 100 V peak-to-peak. I intermittently heard strange noises coming out of the RF matching box. When the DC bias reads properly, there is no problem with your etching. I did not have time to investigate what happens to the etch rate when the DC bias reads close to zero, but based on the above results I think you will get poor results. My guess is that something is wrong with the RF matching network and possibly the RF power supply. However, someone more qualified should confirm this. I think this concern should be taken seriously and acted on immediately by the SNF staff. 2. Loading Integrity I have experienced three wafer crashes in the last couple of weeks. This may be due to the new pins that have been installed into the machine. Please also double check that your wafers sit properly in the loading arm platen. I found that even certain prime wafers can be slighty too large and hence do not sit properly in the load arm. 3. Loading - software problem Sometimes the loading and unloading functions become out of phase with proper pin positions, which end up confusing the machine about the wafer locations. Please pay special attention to this since you can easily crash your wafer into the pins or worse yet crash the load arm into the clamp. Should you find yourself in this position I have found that cycling the load/unload functions along with system reset will eventually get the system back into the appropriate state. Using the prompt option in your recipe sometimes works. 4. The load arm sometimes wobbles too much during loading and unloading. 5. The module located just below the keyboard (MC-200 Motor Controller) is making weird noises from time to time. The fan may need to be replaced. 6. Many of us use low Cl2 flow rates (ie. < 3 sccm). I really don't believe the Cl2 flow rate displayed on the screen. I think this particular flow sensor (better yet, all of them) needs to be recalibrated. My overall conclusion is that the tool is pretty unreliable right now and has several issues which need to be resolved. YOu can still get your etches to work, but you need to be particularly careful about the RF matching/bias and loading. I hope that appropriate action will be taken immediately by the SNF staff so that I and others do not have to spend more time stuggling with the tool. - vijit Vijit Sabnis vsabnis at snow.stanford.edu Stanford University Department of Electrical Engineering Solid State and Photonics Laboratory CISX B113-23 650/725-2774 office 723-4659 fax -------------- next part -------------- An HTML attachment was scrubbed... URL: From true at snf.stanford.edu Tue Jun 18 15:24:12 2002 From: true at snf.stanford.edu (Randy True) Date: Tue, 18 Jun 2002 15:24:12 -0700 Subject: no bias = bad etch Message-ID: <3D0FB30C.6A3845D4@snf.stanford.edu> I did some oxygen plasma etching today in the PQuest and I also had RF matching problems and 0 bias. I just finished SEMing the wafer and indeed almost no etching occured (approx. 20% the usual etch rate and very isotropic). So something is definately wrong. --Randy True From mcvittie at snf.stanford.edu Tue Jun 18 17:17:58 2002 From: mcvittie at snf.stanford.edu (Jim McVittie) Date: Tue, 18 Jun 2002 17:17:58 -0700 Subject: no bias = bad etch References: <3D0FB30C.6A3845D4@snf.stanford.edu> Message-ID: <3D0FCDB6.3C6722F@snf.stanford.edu> Randy, Yes, the problem sounds very real. The most likely cause is that the rf signal coming out of the matching network is shorting to ground before the vacuum feed-through into the system. Last year, we saw this problem and the techs were able to fix it. Jim McVittie re is a short to ground Randy True wrote: > I did some oxygen plasma etching today in the PQuest and I also had RF > matching problems and 0 bias. I just finished SEMing the wafer and > indeed almost no etching occured (approx. 20% the usual etch rate and > very isotropic). > So something is definately wrong. > > --Randy True From booth at snf.stanford.edu Wed Jun 19 13:34:55 2002 From: booth at snf.stanford.edu (Len Booth) Date: Wed, 19 Jun 2002 13:34:55 -0700 Subject: Pquest Chlorine flow rates Message-ID: <3D10EAEF.A1DD3E5@snf.stanford.edu> All - One of the on-going problems with our Plasma Quest has been non-reproducable Chlorine flows. With flows around 2 or 3 sccm, some days etches work as expected, and some days not. The nominal full flow for the Chlorine MFC is 43 sccm. I have noticed a significant day-to-day zero-drift with all of the MFCs on this machine, but it has a most noticable process effect on very low flows. I would like to find out from everyone what are the largest Chlorine flows being used. If you use more than 10 sccm of Cl2, please send me an e-mail describing the recipe name & Cl2 flows used. If everyone is using low Cl2 flows, it may be appropriate to change the Cl2 MFC to a smaller size, to improve low flow performance. thank you, Len From seymour at snf.stanford.edu Wed Jun 19 13:42:32 2002 From: seymour at snf.stanford.edu (Ray Seymour) Date: Wed, 19 Jun 2002 13:42:32 -0700 Subject: gas bottle checks Message-ID: <3D10ECB7.EDC4EE2@snf.stanford.edu> I have reserved the Lam for a gas bottle check on Tue. June 25 at 6:00 and the following leak check will require 24 hours to verify its leak tight. If you have any questions please contact me at 4-7425 From seymour at snf.stanford.edu Wed Jun 19 14:51:02 2002 From: seymour at snf.stanford.edu (Ray Seymour) Date: Wed, 19 Jun 2002 14:51:02 -0700 Subject: gas bottle checks Message-ID: <3D10FCC6.D442DD0A@snf.stanford.edu> I have reserved the amtetcher, the pquest, and the p5000 for a gas bottle check on Tue. June 25 at 6:00 and the following leak check will require 24 hours to verify its leak tight. If you have any questions please contact me at 4-7425 From booth at snf.stanford.edu Wed Jun 19 15:14:27 2002 From: booth at snf.stanford.edu (Len Booth) Date: Wed, 19 Jun 2002 15:14:27 -0700 Subject: Pquest Maintenance Shutdown Message-ID: <3D110243.2FD4AB5E@snf.stanford.edu> Pquest Users - Because there are currently many problems that need repair, I will be shutting down the Pquest tomorrow morning (6/20) at 6am. I have reserved the machine thru Tuesday (6/25) morning. If repairs can be complete more quickly, the shutdown will be removed sooner. Len From mcvittie at snf.stanford.edu Thu Jun 20 13:03:06 2002 From: mcvittie at snf.stanford.edu (Jim McVittie) Date: Thu, 20 Jun 2002 13:03:06 -0700 Subject: Report on Recent PlasmaQuest Problem References: <200206181922.MAA00444@snowmass.Stanford.EDU> Message-ID: <3D1234FA.D024439F@snf.stanford.edu> This is a follow up on Vijit list of PlasmaQuest problem: 1. Poor RF tuning and lack of DC bias voltage: >From Vijit measurements after I left on Friday and from Randy lack of bias after running it at high power, it becoming pretty clear what the likely RF/bias problem is. The RF after the matching network has to go through a torturous path in a long 1.5 ? ID pipe to get to the cathode. At this point the Rf is not in a shielded coaxial cable. Instead, a flat copper strip with thin plastic insulation is used. To prevent shutting of the RF to the pipe wall and to the other conductors in the pipe, the RF conductor is placed in a series of short ceramic sections. Note that along this path the pipe has at least one 90 degree bend. It appears some of the ceramic sections have spread apart to the point that the RF is intermittently shunted to ground. Since the RF is in a power control mode, the RF voltage on this conductor goes up and down depending whether RF is making it to the plasma or being shunted in the pipe. This is why Vijit saw the RF voltage change when he was measuring it near the entrance to the pipe. We have seen this problem before. To correct it, the techs will need to disassemble the whole bottom (cathode) electrode. The work should start this afternoon. The system should be back up at least the middle of next week. 2. Loading Integrity After Vijit?s suggestion, Len measured the wafer pocket diameter in the load arm. He found that it is smaller than the largest wafer diameter allowed by the Semi spec for wafer diameters. This explains the intermittent loading problem that the system has always had. While the tool is down for the RF problem the wafer pocket is being enlarged to accept all wafers within the Semi diameter spec. 3. Loading - software problem Len talked to the vendor about this problem and they suggest that a wafer sensor may need adjusting. He will do this adjust while the system is apart. 4. The load arm sometimes wobbles too much during loading and unloading. The vendor suggested that the lead screw may need lubrication. This will be done once the recommended lubrication has come in. 5. Noises coming from controller. The fan will be checked while the system is down. 6. Cl2 MFC problem Most users are operating at the bottom of the MFC range, the MFC appears to have an unusually large zero drift and the readout does not show the true zero reading. The users will be given a procedure for reading the true zero flow readout for the MFCs. For the near term, when the zero has drifted beyond a set value, the techs will readjust the zeros. We are polling the users to see if we can resize the Cl2 MFC so that we are not running at the bottom of its range where it is overly sensitive to zero drift. When we order a new MFC, we will go for one that has a smaller zero drift. Jim McVittie Vijit Sabnis wrote: > Hi: > > This is an update and follow up to Dr. McVittie's email > based observations from my work yesterday. > > I apologize for the long email, but I hope that current > users read this carefully and communicate their comments > and concerns with all of us. > > 1. Poor RF tuning and lack of DC bias voltage: > > Problems: These problems are flaky, but since I have been > using the machine heavily in the last couple of weeks, I have > noticed them occurring more and more frequently. > RF tuning can be difficult to achieve. Manual > tuning is required. Even when the RF power is matched > appropriately, the DC bias voltage reads 0 or a few volts, > but at some random point can start reading the correct voltage. > > I hooked up the oscilloscope with high voltage probe to > the RF feed to the cathode. I was able to confirm that > when the computer screen and RF power supply display 0 V > or a very low DC bias that there is indeed no DC bias > measured on the scope. This means that there isn't a > problem with the voltage readout on the computer screen or > RF power supply display, but that there is a real problem > with the plasma or RF being supplied to the chamber. Based on this > finding we can eliminate any potential problems with the loose > connector that we found last Friday. At some point > during my investigation the DC bias suddenly was displayed > correctly on the computer screen and RF power supply. I then > confirmed the presence of the bias on the oscilloscope. > I found the same results as last Friday - for a displayed > voltage of -46 V, the DC bias read on the scope was about > -80 to -90 V. The voltage read on the computer screen and > RF power supply is about a factor of 2 smaller than what > it should be, indicating a calibration problem or the > presence of a voltage divider in the readout circuitry. > > Some other interesting observations: > My recipe uses 35W of RF power. When the DC bias reads 0 V > (or something close to that), the scope displays a > sine wave with a 400 V peak-to-peak amplitude. When > the DC bias is correct, the sine wave shifts down by > this bias and the amplitude of the sine wave drops to > 100 V peak-to-peak. > > I intermittently heard strange noises coming out of > the RF matching box. > > When the DC bias reads properly, there is no problem > with your etching. I did not have time to investigate > what happens to the etch rate when the DC bias reads > close to zero, but based on the above results I think > you will get poor results. > > My guess is that something is wrong with the RF > matching network and possibly the RF power supply. > However, someone more qualified should confirm this. > I think this concern should be taken seriously and > acted on immediately by the SNF staff. > > 2. Loading Integrity > > I have experienced three wafer crashes in the last couple > of weeks. This may be due to the new pins that have been > installed into the machine. Please also double check that > your wafers sit properly in the loading arm platen. I found > that even certain prime wafers can be slighty too large and > hence do not sit properly in the load arm. > > 3. Loading - software problem > > Sometimes the loading and unloading functions become > out of phase with proper pin positions, which end > up confusing the machine about the wafer locations. > Please pay special attention to this since you > can easily crash your wafer into the pins or worse > yet crash the load arm into the clamp. Should you find > yourself in this position I have found that cycling the > load/unload functions along with system reset will > eventually get the system back into the appropriate state. > Using the prompt option in your recipe sometimes works. > > 4. The load arm sometimes wobbles too much during loading and > unloading. > > 5. The module located just below the keyboard (MC-200 > Motor Controller) is making weird noises from time to > time. The fan may need to be replaced. > > 6. Many of us use low Cl2 flow rates (ie. < 3 sccm). > I really don't believe the Cl2 flow rate displayed on > the screen. I think this particular flow sensor (better > yet, all of them) needs to be recalibrated. > > My overall conclusion is that the tool is pretty unreliable > right now and has several issues which need to be resolved. > YOu can still get your etches to work, but you need to be > particularly careful about the RF matching/bias and loading. > I hope that appropriate action will be taken immediately > by the SNF staff so that I and others do not have to spend > more time stuggling with the tool. > > - vijit > > Vijit Sabnis > vsabnis at snow.stanford.edu > Stanford University > Department of Electrical Engineering > Solid State and Photonics Laboratory > CISX B113-23 > 650/725-2774 office > 723-4659 fax From YDong at sirostech.com Thu Jun 20 16:42:29 2002 From: YDong at sirostech.com (Yi Dong) Date: Thu, 20 Jun 2002 16:42:29 -0700 Subject: Report on Recent PlasmaQuest Problem Message-ID: <31B8005BBABAD511A71E00A0C9EEA2F51E3434@MAIL> Hello Jim, It is great for you to talk PlasmaQuest problem more clearly right now. Thank you very much for your observation! It will be happy to users If you could recommend some operation rule for PQuest users later! It is important to us for keeping this equipment more reliable and more repeatable in our process. Thanks, Yi Dong Process Development Engineer Siros Technologies 101 Daggett Drive San Jose, CA. 95134 Phone: 408-474-3303 Fax: 408-954-9300 ydong at sirostech.com -----Original Message----- From: Jim McVittie [mailto:mcvittie at snf.stanford.edu] Sent: Thursday, June 20, 2002 1:03 PM To: Vijit Sabnis Cc: Jim McVittie; pquest at snf.stanford.edu; Alice Fischer-Colbrie; Daniel Weaver; Irene Gonzalez; Yi Dong; James Harris; rcrane at snf.Stanford.EDU Subject: Re: Report on Recent PlasmaQuest Problem This is a follow up on Vijit list of PlasmaQuest problem: 1. Poor RF tuning and lack of DC bias voltage: >From Vijit measurements after I left on Friday and from Randy lack of bias after running it at high power, it becoming pretty clear what the likely RF/bias problem is. The RF after the matching network has to go through a torturous path in a long 1.5 " ID pipe to get to the cathode. At this point the Rf is not in a shielded coaxial cable. Instead, a flat copper strip with thin plastic insulation is used. To prevent shutting of the RF to the pipe wall and to the other conductors in the pipe, the RF conductor is placed in a series of short ceramic sections. Note that along this path the pipe has at least one 90 degree bend. It appears some of the ceramic sections have spread apart to the point that the RF is intermittently shunted to ground. Since the RF is in a power control mode, the RF voltage on this conductor goes up and down depending whether RF is making it to the plasma or being shunted in the pipe. This is why Vijit saw the RF voltage change when he was measuring it near the entrance to the pipe. We have seen this problem before. To correct it, the techs will need to disassemble the whole bottom (cathode) electrode. The work should start this afternoon. The system should be back up at least the middle of next week. 2. Loading Integrity After Vijit's suggestion, Len measured the wafer pocket diameter in the load arm. He found that it is smaller than the largest wafer diameter allowed by the Semi spec for wafer diameters. This explains the intermittent loading problem that the system has always had. While the tool is down for the RF problem the wafer pocket is being enlarged to accept all wafers within the Semi diameter spec. 3. Loading - software problem Len talked to the vendor about this problem and they suggest that a wafer sensor may need adjusting. He will do this adjust while the system is apart. 4. The load arm sometimes wobbles too much during loading and unloading. The vendor suggested that the lead screw may need lubrication. This will be done once the recommended lubrication has come in. 5. Noises coming from controller. The fan will be checked while the system is down. 6. Cl2 MFC problem Most users are operating at the bottom of the MFC range, the MFC appears to have an unusually large zero drift and the readout does not show the true zero reading. The users will be given a procedure for reading the true zero flow readout for the MFCs. For the near term, when the zero has drifted beyond a set value, the techs will readjust the zeros. We are polling the users to see if we can resize the Cl2 MFC so that we are not running at the bottom of its range where it is overly sensitive to zero drift. When we order a new MFC, we will go for one that has a smaller zero drift. Jim McVittie Vijit Sabnis wrote: > Hi: > > This is an update and follow up to Dr. McVittie's email > based observations from my work yesterday. > > I apologize for the long email, but I hope that current > users read this carefully and communicate their comments > and concerns with all of us. > > 1. Poor RF tuning and lack of DC bias voltage: > > Problems: These problems are flaky, but since I have been > using the machine heavily in the last couple of weeks, I have > noticed them occurring more and more frequently. > RF tuning can be difficult to achieve. Manual > tuning is required. Even when the RF power is matched > appropriately, the DC bias voltage reads 0 or a few volts, > but at some random point can start reading the correct voltage. > > I hooked up the oscilloscope with high voltage probe to > the RF feed to the cathode. I was able to confirm that > when the computer screen and RF power supply display 0 V > or a very low DC bias that there is indeed no DC bias > measured on the scope. This means that there isn't a > problem with the voltage readout on the computer screen or > RF power supply display, but that there is a real problem > with the plasma or RF being supplied to the chamber. Based on this > finding we can eliminate any potential problems with the loose > connector that we found last Friday. At some point > during my investigation the DC bias suddenly was displayed > correctly on the computer screen and RF power supply. I then > confirmed the presence of the bias on the oscilloscope. > I found the same results as last Friday - for a displayed > voltage of -46 V, the DC bias read on the scope was about > -80 to -90 V. The voltage read on the computer screen and > RF power supply is about a factor of 2 smaller than what > it should be, indicating a calibration problem or the > presence of a voltage divider in the readout circuitry. > > Some other interesting observations: > My recipe uses 35W of RF power. When the DC bias reads 0 V > (or something close to that), the scope displays a > sine wave with a 400 V peak-to-peak amplitude. When > the DC bias is correct, the sine wave shifts down by > this bias and the amplitude of the sine wave drops to > 100 V peak-to-peak. > > I intermittently heard strange noises coming out of > the RF matching box. > > When the DC bias reads properly, there is no problem > with your etching. I did not have time to investigate > what happens to the etch rate when the DC bias reads > close to zero, but based on the above results I think > you will get poor results. > > My guess is that something is wrong with the RF > matching network and possibly the RF power supply. > However, someone more qualified should confirm this. > I think this concern should be taken seriously and > acted on immediately by the SNF staff. > > 2. Loading Integrity > > I have experienced three wafer crashes in the last couple > of weeks. This may be due to the new pins that have been > installed into the machine. Please also double check that > your wafers sit properly in the loading arm platen. I found > that even certain prime wafers can be slighty too large and > hence do not sit properly in the load arm. > > 3. Loading - software problem > > Sometimes the loading and unloading functions become > out of phase with proper pin positions, which end > up confusing the machine about the wafer locations. > Please pay special attention to this since you > can easily crash your wafer into the pins or worse > yet crash the load arm into the clamp. Should you find > yourself in this position I have found that cycling the > load/unload functions along with system reset will > eventually get the system back into the appropriate state. > Using the prompt option in your recipe sometimes works. > > 4. The load arm sometimes wobbles too much during loading and > unloading. > > 5. The module located just below the keyboard (MC-200 > Motor Controller) is making weird noises from time to > time. The fan may need to be replaced. > > 6. Many of us use low Cl2 flow rates (ie. < 3 sccm). > I really don't believe the Cl2 flow rate displayed on > the screen. I think this particular flow sensor (better > yet, all of them) needs to be recalibrated. > > My overall conclusion is that the tool is pretty unreliable > right now and has several issues which need to be resolved. > YOu can still get your etches to work, but you need to be > particularly careful about the RF matching/bias and loading. > I hope that appropriate action will be taken immediately > by the SNF staff so that I and others do not have to spend > more time stuggling with the tool. > > - vijit > > Vijit Sabnis > vsabnis at snow.stanford.edu > Stanford University > Department of Electrical Engineering > Solid State and Photonics Laboratory > CISX B113-23 > 650/725-2774 office > 723-4659 fax -------------- next part -------------- An HTML attachment was scrubbed... URL: From booth at snf.stanford.edu Thu Jun 27 15:21:00 2002 From: booth at snf.stanford.edu (Len Booth) Date: Thu, 27 Jun 2002 15:21:00 -0700 Subject: Plasma Quest update June 27 Message-ID: <3D1B8FCC.DE176934@snf.stanford.edu> Pquest users - After finding water/glycol leaking out from under the cathode, I had to remove the cathode completely from the etch chamber. A teflon cooling line had been leaking onto the chuck heater wiring (208 vac powered) and broken one of the leads to one of the 4 cartridge heaters. Parts are being ordered, and I hope to have the machine reassembled by next Wednesday 7/3. Also because of this coolant leak, the problems associated with RF tuning & DC bias measurement may have occurred. We'll know after the cathode is reassembled and the RF can be tested. While trying to track down the cause of seemingly random wafer breakage a week ago, we found that the pocket machined in the end of the load arm had been machined incorrectly. The pocket was not of a large enough diameter to allow all 4" wafers to fit easily. 4" wafers that were near the upper end of the SEMI specs for prime wafer diameter would get jammed in the pocket, and contribute to the wafer being broken/lost when the lift pins & clamp tried to move the wafer. The pocket has been machined & will now accomodate all prime wafer stock. Len