Scientists at UTI Instruments, San Jose, CA, sought to minimize this background and developed practical techniques for exploiting the QMA’s increased detection capabilities. ECNC Europe provided additional direction for the group.
Their findings are summarized in the technical note Lower Detection Levels in UTI QualiTorr Process Monitors.
The UTI researchers set out to see if background improvements could be achieved by substituting a cryopump for the QMA’s traditional turbo. Their QMA was the UTI QualiTorr III in a closed-source, short-probe configuration with a close-coupled, fast-response sampling module. The standard 60 l/sec turbopump used with this was replaced by a CTI On-Board 8F cryopump. The test chamber was pumped with another On-Board 8F. Both cryos were run on the same helium compressor. A test gas mixture consisting of 1% hydrogen in argon was introduced into the test chamber via a variable leak valve.
The researchers found that switching to a cryopump produced much lower backgrounds throughout the sampling system.
This finding is consistent with as yet unpublished research conducted by John O’Hanlon and Barry Brownstein of the Univ. of Arizona, Tucson, The Population And Linkages Service, and Dave Fraser of Intel Corp., Santa Clara, CA. They showed that background partial pressures in an ultrahigh vacuum sputtering system can be reduced by an order of magnitude when the system is pumped with a cryogenic pump, rather than a slower turbomolecular pump.
The UTI group also found that the reduced background allowed them to measure much lower residual gas levels at base pressure and at the operating pressures of many PVD, CVD, and plasma-etch processes.
Specifically, they found:
* Partial pressures in the |10.sup.-10~ and |10.sup.-11~ torr range could be measured repeatedly while monitoring argon at a process pressure of 2 x |10.sup.-3~ torr.
* Water vapor partial pressure changes in the |10.sup.-10~ and |10.sup.-11~ torr range could be measured while monitoring argon at the same process pressure. These changes amounted to 15 to 30 ppb concentrations.
* Partial pressure changes in the |10.sup.-12~ torr range could be measured while the process chamber was at a base pressure of 1.5 x |10.sup.-10~ torr.
* Water vapor partial pressure changes in the |10.sup.-11~ torr range were reliably measured at the same process chamber base pressure.
Their technique for measuring low-level contaminants while monitoring a process at 2 x |10.sup.-3~ torr went something like this:
First, they calibrated the QMA at a pressure that was lower than the process pressure, using a Bayard-Alpert gauge. For the sake of argument, let’s say they calibrated it at the |10.sup.-4~ torr level. That put |10.sup.-4~ torr at the top of their CRT screen instead of |10.sup.-2~ torr and let them look at contaminant levels two decades lower than would otherwise be possible.
When they brought the chamber up to a process pressure of 2 x |10.sup.-3~, the 40Ar peak went off the scale. So they turned off the 40 amu monitoring channel. They did this so they wouldn’t get a high current coming from their electron multiplier, since that would have shortened its lifetime.
They didn’t need to monitor 40Ar directly because argon has other, less abundant isotopes whose partial pressures remain less than 40Ar’s by a constant proportion. This enabled them to monitor the process pressure using 36Ar, whose abundance is about 300 times less than that of 40Ar.
The researchers could be confident of the contaminant levels they measured because they had checked the linearity of their system through a series of dynamic scans.
One of these began with the pressure at 3 x |10.sup.-5~ torr. The pressure was then raised in several steps to over |10.sup.-3~ torr and decreased in another series of steps until the argon was finally cut off. All of the peaks–except hydrogen–tracked one another showing steps that were precisely proportional to the major argon peak at 40 amu.
The hydrogen steps followed the pattern at first, but then departed as the hydrogen amount leveled out in the mid-|10.sup.-8~ torr range. This occurred partly because of the difficulty of pumping hydrogen with a cryopump and partly because of the system’s inherent hydrogen background.
About the time the argon was shut off, heaters on the process chamber were turned on for about 4 min, then the argon was turned back on.
The purpose behind this procedure was to test the linearity of the system for water vapor. Monitoring water vapor is a little tricky when argon is present. The problem is that some of the 36-amu argon isotope is doubly ionized, which means that it shows up as a peak at 18 amu, which is where water also is. In this way, 36Ar obscures the presence of water.
So the researchers set up the QMA to monitor water at both 17 and 18 amu by having a 4.25 x multiplier at the 17-amu channel to account for water’s cracking pattern.
With the heaters on, both the 17- and 18-amu levels rose above the baseline and tracked each other fairly closely until the argon was again turned on. At that point, the 18-amu peak rose sharply, due to the influx of 36Ar. Importantly, though, the 17-amu peak remained undisturbed by the pressure rise of argon, even though that rise continued through about seven decades.
This undisturbed measurement of water vapor while the total pressure in the chamber changed greatly dramatically demonstrated the superior linearity of the QMA system and the accuracy of the measured results.