The following is a collection of stories based on real experience in my research group. I hope this is helpful.
Leaks in unexpected places
I was growing AlOx films in a small vacuum system using reactive sputtering of Al in a mixed Ar/oxygen ambient. The vacuum system was evacuated by a 50 l/s turbo through a 2.75" gate valve. In the last several runs, I observed several things indicating a leak somewhere. First, even after baking, the base pressure was well above 1 x 10-7 Torr (the base pressure should be below 1 x 10-7 Torr). Second, the deposition rate dropped to about 10% of the optimized value. Third, the RGA indicated a mass 32 peak substantially higher than that of mass 31. I used standard helium leak-checking, first checking the flanges opened most recently and found nothing. I eventually checked all flanges and feedthroughs, but the results were all negative. In a moment of despair, I decided to check the rate of pressure rise by closing the gate valve. As I closed the gate valve, I noticed that the pressure increased only very slowly. A quick helium check of the gate valve bellows indicated a leak, probably due to fatigue cracks developed after many cycles of closing and opening.
Where has my plasma gone?
I was using a standard magnetron source to deposit Ti and TiN thin films. Over the last few months, it became more and more difficult to start the plasma at the usual pressure (about 20 mTorr in my system). The RGA indicated no leak or unusual residual gas content. I removed the magnetron. Using a Hall probe, I measured the magnetic field 1 cm above the target and found that it is about 40% lower than the new target. This particular magnetron was designed with the magnets immersed in the coolant chamber. Disassembly of the target revealed corrosion of the magnets. Replacement of the magnets immediately restored the discharge performance.
Glowing everywhere
I have been doing boron nitride deposition using reactive sputtering of B4C in an argon/nitrogen plasma for the past few weeks. In the last few runs, I started seeing an extended plasma covering the chamber walls during deposition. At the same time, film properties appeared to be irreproducible. The extended plasma was strongly suggestive of charging due to deposition of boron nitride on the chamber walls (both hexagonal and cubic phases are highly insulating). I opened the chamber and scraped the chamber clean. Everything returned to normal after that.
Gauge anomalies
My UHV system was evacuated by an ion pump. Every time when I turned off the ion pump, I saw an initial decrease in the pressure, followed by a gradual rise. This does not sound reasonable. My advisor suggested using a higher emission setting for the ionization gauge. Lo and behold, this anomalous effect disappeared. Since an ion pump produces stray positive ions, some of these stray ions may reach the ionization gauge and are registered as additional pressure. When the ion pump is turned off, this stray component disappears, resulting in an apparent momentary pressure drop.
Leak valve leaking
We have a scanning Auger microprobe in our facility. One user came to me complaining that the system pressure did not recover after argon ion sputtering. I went over and noticed that the pressure was indeed high (about 1 x 10-7 Torr). A quick inspection of the RGA showed a large amount of argon. The Varian leak valve appeared to be tight. However, to be sure, I reset the leak valve to allow further tightening. In the process of resetting, I noted that the two thumb screws were not tightened against each other. After resetting, the pressure plunged down by two orders of magnitude within a few seconds.
Charging ahead
My Auger system was acting strange. I set up the sample for elastic peak calibration. It looked fine, but I wasn’t getting any Auger signals. Also the noise was unusually low. There was no sample current. It appeared that either the sample is insulating, has an insulating layer, or the ground return wire is broken. Opening the chamber revealed a broken ground return wire.
High background
For the last several runs, my XPS system was giving very high background counts. I checked the electronics. That did not seem to be the problem. One curious feature was that I was getting significant electron counts with energy above 1500 eV, but I was operating an Al Ka source at 1486.6 eV. Opening the x-ray source showed that the Al window in front of the target was gone. Without this window, stray electrons flooded the electron energy analyzer.
Ion pump blues
I came into the lab this morning and saw that my ion-pumped system was tripped off (it was in the mid-10-10 Torr range when I went home yesterday). Not knowing exactly what happened, I did not want to turn on the ionization gauge. I evacuated the gas inlet manifold using a sorption pump down to below 1 mTorr and then opened the isolation valve to the vacuum chamber. The pressure stayed below 1 mTorr. So far so good. Then I put the ion pump power supply in the start mode and turned on the unit. The pump did not start, and the voltage stayed below 500 volts for more than 15 minutes. I shut everything off and unplugged the high-voltage connector to the ion pump. A quick resistance check showed that the ion pump was shorted. Suspecting that this could be due to titanium flakes (this was an old pump), I zapped the high-voltage feedthrough with a tesla coil. The resistance went up to about 10 ohms, but did not improve further even after many zaps. It’s time to open up the whole system. One ceramic spacer was found to be broken. Several other ceramic spacers were heavily coated with titanium. After replacing the broken spacer and cleaning others, the ion pump is now happy (and so am I).
Broken heaters
My research is in the surface chemistry of intermetallics. I designed a heater to heat my single crystal to about 1000K. This sounded simple enough. I used 10-mil tungsten wire coil and placed it in a macor holder. Everything seemed to be okay up to a current of 10A, when the filament broke. This began a three-month nightmare when I tried everything to make this work. Sometimes, the macor holder broke, but most often, the heater broke, usually around 10A. The crystal never got close to 1000K. Careful examination of the heater showed that the tungsten appeared to have melted into the macor. My advisor suggested the possibility of a reaction between tungsten and macor at elevated temperatures. I did notice that during heating, my RGA showed a significant peak at mass 19 (F), probably coming from the decomposition of macor. Not wanting to waste any more time, I redesigned the heater without the use of macor. The heater has been working well for the last two years.
Cubic boron nitride tricks
I have been trying to grow cubic boron nitride for the last few months, using reactive sputtering of boron carbide. The success was intermittent: sometimes it worked, sometimes it didn’t. I varied the substrate temperature, pressure and substrate bias, but with the same irreproducible results. It’s very frustrating. The chamber I was using achieved a base pressure of ~2 x 10-8 Torr. There was a small oxygen leak. I was not able to determine where the leak was, probably because the leak was so small (the RGA read a mass 32 partial pressure in the low 10-10 Torr range). I began to suspect that such a small oxygen leak might be critical, so I moved to another system which was leak-tight. With this system, I discovered that as long as I gave the vacuum chamber a good bake to reduce the water vapor pressure, along with appropriate substrate biasing and ion bombardment, my growth results became much more consistent.
Short circuit
I was using a standard Ti magnetron target to deposit TiN films. On average, the target shorted after three runs. In disasembling the target, I saw powders and flakes. These were probably the culprit. After cleaning, the target worked fine. The short-circuiting, however, continued after three runs, like clockwork. The RGA showed that the chamber had a small oxygen leak, which was eventually traced to a flange where the gate valve was mounted. After fixing this oxygen leak, I no longer have to clean the target so frequently. Although I did not perform any chemical analysis, my guess is that oxygen reacts with Ti to form titanium oxide powers and flakes that caused the short circuit.
Getting wet
Part of my research involves operating with a high background pressure (~1 x 10-6 Torr) of water vapor in my UHV system. At one time, I noticed that the RGA registered a very high mass 32 peak (> mass 31 intensity). This could not be a leak, since I just completed a thorough leak-check for the system. The mass 32 to mass 18 (H2O) ratio is about 0.01. It did not appear to make a difference when I turned off the ionization gauge. However, when I turned off the ion pump (and turned on a turbo), this ratio dropped to about 0.001, suggesting that the oxygen signal may have come from the fragmentation of water by the ion pump.
More ion gauge anomaly
I used a solid-state negative metal ion source for controlled dosing in surface science experiments. Whenever I turned off the ion source, the pressure jumped up. I would have expected the opposite. Turning off the ion source should reduce the outgassing and hence the pressure. What I did not realize is that the ion gauge measures pressure via the collection of positive ions, any stray energetic negative ions may also be collected during the ion dosing experiment. The effect is that the ion gauge collects a smaller net positive current. With the negative ion source off, the negative component disappears, resulting in a larger net positive current and hence higher pressure reading. In fact, the same is true when operating an electron gun.
Clean surface
Part of my research involves cleaning a metal crystal by the usual sputter-cleaning and annealing (at 1000K), and then exposing this clean surface to gas adsorbates at 100K. Usually, after annealing the sample at 1000K for 30 minutes, I turned off the heater and started cooling immediately with liquid nitrogen. I checked that the surface was clean immediately after annealing, but after cooling, the surface became contaminated. Concerned that the pressure may be too high when I started the cooling, I changed to a new strategy. After annealing, I decreased the heating current a little bit to hold the sample temperature at 700~800 K. Because of less radiation heating, the pressure went down steadily. At this temperature, the sticking probability of residual gas contaminents is sufficiently low that I don't have to worry about contamination. After a few minutes, I just turned off the heater and started cooling with liquid nitrogen. I can get a pretty clean surface this way.
Do viton seals leak?
The short answer is: it depends. In my magnetron sputter-deposition system, I have a magnetron source with two viton seals. One viton seal is tightened between two flat sealing surfaces with about 20 bolts, while the other viton seal is a sliding seal. The first seal is fine, while the second one leaks slightly. The leak is subtle – it takes about 3-5 minutes before the helium peak shows an increase after the leak is dosed with helium, but the leak is definitely there. If anyone wants a secure viton seal, make sure that you do not use one single sliding seal – either replace it with metal seal or use double seals with differential pumping between.