Underfill Update: NUF, MUF, WUF, and Other Stuff
Underfill Update: NUF, MUF, WUF, and Other Stuff
Dr. Ken Gilleo, Cookson Electronics
Background
Many decades ago, the Flip Chip was born of necessity in New York, of all places. IBM was pursuing the optimum process for connecting chips to the outside world. The big mainframe computer makers have always been on the leading edge, requiring the highest speed and density long before the other markets. Direct Chip Attach (DCA), or Flip Chip, was the answer. Nothing could be faster, finer or more reliable, so the computer giant took that path.
IBM was using low-expansion ceramic circuits and chip carriers, so there were no concerns about the thermo-mechanical expansion mismatch found with organic circuit boards. But, there was a concern that moisture could cause problems with tiny solder joints, even though the main frame modules were sealed.
The way the story goes, engineers at IBM抯 Endicott lab tried various adhesives and sealants around and under the flipped chips and found some useful sealers. Just to be sure that these “encapsulants” would not reduce the already phenomenal thermocycle performance, they performed routine temperature cycle tests. Unexpectedly, some of the assemblies thermocycled right past the standard values.
What was going on here? A few months later, the lab realized that they had discovered the “underfill effect” — enhanced performance by interposing adhesive between chip and substrate. Now, decades later, the mechanical coupling mechanism for underfills is finally well understood.
Underfill Blues
Soon IBM gave underfills serious attention. Perhaps underfill would let IBM use standard ceramics, instead of the special and expensive very low expansion ceramics. But when underfills were finally ready to move into production, the assemblers complained that the clear materials were hard to inspect. No problem, just add a pigment.
One enterprising engineer decided to use IBM approved pigments, since it could take months of testing to get something new on the approved list. He went to the “card shop” (circuit boards) and asked for the green solder mask ingredients. Finding that the green was made from a mixture of yellow and blue colorant, he selected just blue for simplicity. And this is really why so many underfills have been blue since then.
Off the CUF
What抯 the status of underfills (UF) today, now that Flip Chip has become pervasive and popular? We must admit that dispensing and curing times sometimes make underfills maligned as the “production bottleneck.” The capillary underfills (CUFs) are doing well but the productivity curves are “maxed out”. Fast flow and snap cure, the key productivity parameters, have been close to the practical limits for several years.
A 5 minute cure at 165oC, achieved many years ago, is nearing the limit. Sure, the chemistry can be cranked up further, but there are such risks as auto-pyrolysis and even explosive polymerization. Risking reliability to save another 60 seconds in production time is a dubious bargain. So where do we go from CUF?
Is NUF Enough?
Productivity can often be improved by combining multiple operations into one. Researchers at Motorola, a pioneer in FC on organic boards, hit on the idea of combining UF and flux into a single material. Their process involved applying hybrid underfill-flux on the board, placing the chip and running the assembly through a reflow solder oven. Later, industry and universities pursued the idea, called “no flow” underfill (NUF) and found that it solved many materials and process problems. Georgia Tech thoroughly explored process parameters and published extensive results.
A remaining problem is that the filler added to reduce the thermal coefficient of expansion (TCE) interferes with soldering. Most vendors address this by reducing or eliminate filler altogether, but producing a higher than optimum TCE: 70 to 85 ppm/oC. This is three times higher than the solder joint TCE of about 25ppm/oC. Underfill cracking and delamination have been a problem with NUF
Recently, Cookson Semiconductor Packaging Materials, borrowing from the epoxy molding compound (EMC) heritage that Cookson acquired with Plaskon, introduced a highly toughened NUF. This material has tested well and may have eliminated the “no filler” problem. The jury is still out on how widely NUF will be accepted.
Don抰 Muff It, MUF It!
The same Plaskon transfer molding experts, now part of Cookson, asked, “Why not apply underfill by transfer molding?” About two years later, molded underfill (MUF) was looking good. Modified EMC could be injected under a flip chip using modified molds. But the method is best applied to Flip Chip in Package (FCIP) since molding requires relatively small assemblies or high chip-to-board ratios.
Today, the process is being used to make FCIPs in BGA format. Individual packages, strips and even full arrays (flood molding) work well. When strips or full sheets are used, the molded packages are singulated by sawing. The MUF process also allows overmolding so that the FC can be totally encapsulated in a single step if desired. Predictions are that MUF will be the choice among packaging foundries because of their focus on high productivity, and the ready availability of molding presses.
Did he say WUF?
What抯 left for a UF process? When a large cellular phone manufacturer asked this question several years ago, their CUF vendor replied, “Build the flux and underfill right into the chip”. This company also made wafer-level die attach films and pastes, so the idea was a natural for them. Thermoplastic reworkable die attach adhesives are applied to the back of wafers, hardened and the wafer is then sawn. The chips are ready to bond.
So why not flip the adhesive to the front and call it wafer-level underfill (WUF)? The company developed solid flux-underfills based on their already patented epoxy-based flux. After two years of work, several approaches, including two-layer flux-underfills had been tested with encouraging results. But there was considerable work remaining, and the program was shelved, after laying a minefield of patents.
Today, several companies are pursuing WUF, including a group funded by the US government. But the jury is still out. Will WUF nip the gremlins or is it going to be a dog? Next year may tell.
For More Information:
Gilleo, K. and Blumel, D. “Transforming Flip Chip into CSP with Re-workable Wafer-Level Underfill,” Proceedings Fourth Annual Pan Pacific Microelectronics Symposium, SMTA, February 1999, pp. 159 – 165. Abstract available in Literature, this site. Full text available from SMTA.
Gilleo, K, and Blumel, D. “Wafer-Level Flip Chip: Bumps, Flux, and Underflip.”
HDI, Vol. 2, No. 9, September, 1999 pp. 25-27.
Several other specialized underfill papers are in Literature, this site. Some points discussed above are illustrated in Photo Gallery, this site
