edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 28 NO. 3 12 requirements were put into place. The military required that high reliability microcircuits were to contain less than 5000 ppm(v) moisture when tested at ~100°C. The requirement was published as Test Method 1018 in MILSTD-883. The original document identified three analytical techniques that could be used: (1) gas chromatography, (2) mass spectroscopy or residual gas analysis (RGA), or (3) moisture sensors. Of the three, RGA was found to be least affected by moisture adsorption and soon became the preferred analytical method to test for moisture within a cavity device. Unfortunately, the analytical community had yet to develop and perfect the analytical test methods and protocols to meet that measurement requirement. Furthermore, there weren’t any standards for certifying the accuracy of moisture at such low levels. Nevertheless, by 1980, Test Method 1018 was “Dated,” i.e., put into effect. The Preparing Activity responsible for this requirement was the Rome Air Development Center (RADC) at Griffiss Air Force Base in Rome, New York. Any discrepancies in the reported moisture levels within a production certification lot would be addressed by the RADC lab. The lab also continued the search for a moisture standard so that the various analytical labs and the device manufacturers could be assured of the data accuracy, vis-a-vis the reject criteria. In the interim, RGA “calibration” was accomplished via a NIST certified moisture generator in conjunction with a three-volume calibrator kept within each analytical facility. PROBLEMS AND LESSONS LEARNED The manufacturing community quickly challenged the 5000 ppm(v) limit as it appeared that the value was a number pulled from a hat. But such is not the case, as that value is the dew point for moisture at -2.4°C. At this temperature, moisture in a typical 1 cc hybrid package will transition directly from ice to water vapor without passing through the liquid phase. In essence, the moisture sublimes and cannot support the transportation of any residual ionic species. This was a point that had to be emphasized quite often, especially to those who looked at chemistry as a form of black magic. At that time, device manufacturers also made the grave mistake of assuming that all product in a given production lot is identical. To put it bluntly, this assumption was as poor then as it is today. Each device has its own variation in physical and chemical makeup, and those variations can become very significant for Class B and H hybrid devices. Another practice within the manufacturing houses that had to be addressed was the proper technique to dry parts prior to hermetically sealing them. At room temperature (20°C or 68°F), water molecules in the air move at an average speed of approximately 590 to 640 m/s. This means the sealing operator cannot take parts from a drying oven and carry them to the sealing machine without water adsorbing at part surfaces. As the parts quickly cool, what was done during the drying process is quickly undone upon exposure to the ambient environment. One naturally questions how much water is re-adsorbed, and a door to a host of what-ifs and other assumptions and arguments is opened. Several papers have been presented that concluded most surfaces will readily adsorb three layers of water molecules. This value equates to a water adlayer thickness of approximately 10 angstroms, a depth more than sufficient to provide a medium for ionic residues to be mobilized. The water molecule is not simply a three-atom linear structure, but it is V-shaped with a 104° angle between the “legs.” Subsequently, the molecular model that forms will demonstrate significant polarity and create stronger intermolecular forces. This allows for the formation of macromolecules, adsorption phenomenon, mobilization of ions, etc., that can wreak havoc on device performance and reliability. All these phenomena warrant more discussion that is left unaddressed here in the interest of space and can be pursued in the Suggested References included at the end of this article. Within a few years of “Dating” the test method, RADC relented on its position of metal and/or ceramic only. Previously, die attach had been accomplished with solders, but now epoxy die attach was to be allowed as its pros greatly outnumbered the cons. This acceptance of nonmetals opened the door to allow other organic materials to be used as well. For example: Die coatings, PIND materials, RF absorbers, and moisture getters were gradually introduced to the menu. Unfortunately, this expanded list of acceptable options included moisture absorbing materials that had to be out gassed prior to sealing a device. Moreover, manufacturing engineers now had to consider the physical effect of the aspect ratio for moisture permeation through adhesively attached large die or other elements when pre-seal bake out was being done. As noted above, RADC committed to continuing its search for proving the accuracy of the moisture analysis via some standard. That effort was far from successful as some of the above molecular phenomena were not properly addressed. The net result was an annual report at the Minnowbrook Microelectronics Conference (an annual gathering of technologists that RADC started circa
RkJQdWJsaXNoZXIy MTYyMzk3NQ==