On the subject of 80/20 Au/Sn. In a previous existance I used to
do a lot of this on PGA and other semiconductor package
assemblies. Our assemblies where gold over nickel over tungsten
on Alumina packages, gold over nickel over Kovar lids with a
gold/tin solder preform tacked to the lid. Gold plating on the
surfaces was specified at 1.5 to 4.5 microns and was typically









         kept at the low end of the specification by the package and lid 
         suppliers. Reflow was done without flux in an inert nitrogen 
         atmosphere in multi zone belt furnaces. Typical peak sealing 
         temperatures where 340 to 350 deg C. I forget reflow times. 
         
              This system will reflow without flux provided that certain 
         conditions are met. The surfaces must not be contaminated to badly 
         with organic or other materials. We stored packages lids and 
         preforms in Nitrogen cupboards before use. The temperature/time 
         exposure of the packages prior to sealing must be controlled to 
         prevent Nickel diffusing through the gold and oxidising. Nickel 
         oxides are not solderable. If nickel oxides do develop on the 
         surface the situation can be recovered by doing a forming gas 
         scrub with N2 12%H2 at 350 deg C. This reduces the Nickel oxides 
         on the gold surface to nickel and makes the surface solderable for 
         a reasonable length of time. We found that if processing 
         temperatures prior to seal were kept below about 350C during die 
         attach then forming gas scrub was not needed. You want to avoid a 
         forming gas scrub due to the extra process step and the potential 
         dangers of hot Hydrogen atmospheres (Kaboom). Don't try sealing 
         parts in an atmosphere containing Hydrogen. Oxides will be 
         converted to water, which will be trapped inside your package, 
         perhaps leading to corrosion problems. 
         
              The process described produced excellent wetting with minimal 
         voiding. The times when we found voiding and poor wetting where as 
         follows:
         
         Oxidised nickel on the surface and/or poor gold plating.
         
         Poor design of seal ring and preform size, leading to there not 
         being enough solder to fully wet the surfaces. 
         
         Contamination, dust, fibres, etc on the sealing surfaces. We 
         didn't find a need for plasma or other cleaning. Typically we 
         found that if we could aluminium wire bond succesfully with high 
         yields we could seal the packages.
         
              I would worry about using a paste to seal an MCM. I assume 
         that there is some sort of flux used as part of the paste vehicle. 
         Residues from this could be sealed within the package, to 
         potentially cause problems latter. A liquid paste is more likely 
         to give voiding as the gas from evaporated liquid has to travel 
         out across a seal width of a mm or two with a thickness of only 
         perhaps 30 to 40 micron. Gas can easily be trapped leading to 
         voids. Another source of gas for voids could be outgassing from 
         materials within the MCM cavity. 
         
         Batch ovens may be got to work, but in my experience would give 
         very long times above liquidus for the solder due to the 
         relatively slow temperature response compared with a belt furnace. 
         Vacuum soldering can also be made to work, but I don't have any 
         experience with it. Choice of process depends partly on what is 
         contained within the MCM. If the contents are temperature 
         sensitive then vacuum seam sealing might be used which 
         concentrates the heat in the lid joint, without heating the whole 
         package.
         
              You do need to minimise voids in the seal. These can cause 
         leaks over time due to thermal cycling stresses. 
         
              Talk to your lid vendor and package vendor. They should have 
         a lot of knowledge on this subject. 
         
         best regards,   Jeremy Drake