edfas.org 13 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 28 NO. 3 1980 to address moisture in microelectronics issues) that indicated a total lack of confidence in the methodology. Test Method 1018 was being challenged for a variety of reasons. Not only were there differences in the mass spectrometers in use, but there were issues with the software and the retention of raw data. Differences in reported moisture contents were found to vary significantly as the manufacturers attempted to perform round-robin testing between the certified labs. As noted above, many manufacturing processes had to be upgraded. But even as the industries attempted to meet the 5000 ppm(v) limit specified by the test method, many analysts found that parts passing the reject criteria early on failed in later testing. By the mid-1980s, the scientific community was able to clarify some of the problems, one of which was hydrogen. At that time, a primary metal used for device fabrication was kovar (an iron, nickel, and cobalt alloy) as it had a thermal coefficient of expansion similar to that of the glass (Corning 7040 or 7056) feedthroughs. Unfortunately, the fabrication processes for kovar allowed mono-atomic hydrogen to be entrapped within the material. Furthermore, to protect the kovar from the effects of hostile environments (chemistry), the alloy had to be nickel and gold plated. These electrochemical processes cause various levels of hydrogen to be entrapped in the plating as well. Over time and temperature fluctuations in different applications, hydrogen will gradually diffuse though the base metal into the device cavity. The hydrogen entrapped in the plating will do likewise. The result is a hermetic cavity device with hydrogen entrapped inside where it will start a reduction process of the surface oxides and yield water. (Remember: “Chemistry happens.”) Not only did this in situ moisture create problems, but the diatomic hydrogen also became chemically active with the metal systems utilized by the RF community. This realization set off scores of hydrogen studies in an effort to identify the culprits that contributed the most hydrogen. These studies quickly reinforced the axiom that no two devices are truly identical as the hydrogen content of device elements within a given production lot were found to be significantly different. Lastly, but not nearly complete in this discussion, somewhere in this time frame the decision-makers at RADC concluded that acting as the Preparing Activity for the Military Specs was not in their charter. Hence, their technical support was withdrawn. Essentially, the suppliers and users were left on their own to find some means of proving the precision and accuracy of the instrumentation used to acquire moisture data. PROVING THE RELIABILITY OF THE RGA TEST PROTOCOL The beginning of the 21st century found the microeletronics community still mired in mistrust of RGA data, no solution to a lack of a moisture standard, a growing diversity of microelectronic devices, and gradual withdrawal of the Department of Defense in any attempt to resolve the issues. Throughout the decades, NASA remained on the sidelines as an observer, presenting its own needs and requirements only when needed. Fortunately, that changed, and in one of the annual Minnowbrook sessions a new approach to creating a standard was suggested and NASA personnel identified with the concept. The new approach was to present an identical sample to the various analytical labs measuring moisture content. It was to be done via one very large sample volume (in the sense that RGA sample requirements are extremely small) comprising a one-liter gas cylinder with small sample chambers between the values enclosing gas obtained from the main sample/cylinder. The thinking was that this shared cylinder approach would eliminate all questions about differences in makeup, process variations, gas composition, and aging. Furthermore, the very large single sample would be conveniently portable. NASA Goddard Space Flight Center funded the first stage of this effort. Over time, a stainless-steel ~1.5 liter cylinder was fabricated with shutoff valves on each end (Fig. 2). The ends of these valves were each fitted with a second valve, and the space between the two valves was to act as the sample volume. An opening and closing sequence for evacuation and backfill allowed one to sample the same volume over and over again. The single sample cylinder (SSC) was charged with nitrogen (99.99% purity) along with a nominal 5000 ppm(v) water via the NIST dew pointer. Additionally, the sample cylinder included a trace of argon that, when tested via RGA, provided additional insights into the hermeticity of the Fig. 2 The single sample cylinder for residual gas analysis.[4]
RkJQdWJsaXNoZXIy MTYyMzk3NQ==