By Paula Doe
WaferNews Contributing Editor
About the only thing clear in the low-k market these days is that there’s no clear solution yet. Adoption of real low-k materials looks to be delayed still further, as candidate materials proliferate.
Projections for the widespread adoption of low-k materials continue to get pushed out. While users have International SEMATECH busy testing materials with k values around 2 this year and next, “That’s much more aggressive than really will be used,” says Ken Monnig, SEMATECH’s associate director of interconnect. Continuing problems with integration and reliability suggest the roadmap for low-k adoption will have to be pushed out again.
Kline & Co. figures the low-k materials market will come to only some $16 million this year, then start to expand in 2003 to 2004 as companies start to ramp their 100nm production, and finally reach $400 million in 2006.
“There’s been a surprising longevity of FSG [fluorosilicate glass],” says John Davis, industry manager for Kline & Co., Little Falls, NJ. “We thought it would die out, but it’s the low-k of choice at 130nm. It’s a way to punt, it’s not very low-k at all.”
FSG has a k value of about 3.5.
One possible sign that chipmakers may be starting to look more seriously at low-k products, though, is a recent upsurge of interest in the materials among packaging makers, who’ve all suddenly in just the last couple of months started asking SEMATECH for all different kinds of wafers to test.
“They suddenly all woke up and saw the light bulb,” relates Monnig. “They’ve gone from saying ‘That’s interesting, SEMATECH, we’ll look at it later in our spare time,’ to ‘We’re desperate.'”
When chipmakers do start actually ramping production of chips using low-k dielectrics, there still probably won’t be any clear winner for the standard material of choice, or even of the competition between spin-on and CVD.
“At 90nm and 100nm people will have to make hard decisions, but we see a role for both [CVD and spin-on processes]. We see project cohabitation,” says Davis. “Even IBM will probably go down the dual road.”
Indeed the number of potential low-k candidates seems to keep growing, not narrowing. SEMATECH says it is evaluating a lot more candidates than it was six months ago. As suppliers develop clever chemistries to strain the crystal lattice more, some CVD materials, apparently of the organosilicate glass (OSG) type, are starting to achieve only slightly higher k values than the spin-on organics.
“Now we’re seeing some CVD-derived materials with low and stable k values, with porosity,” says Monnig.
So far SEMATECH has found one such porous material with k around 2.2 that’s been successful with integration, and now may have a second.
“We’re pushing CVD down the road even to 70nm,” says Davis.
Monnig also notes increased interest in hybrid approaches, where different materials for the via and contact layers could create built-in etch stop layers. Also a hot topic: The problems of making a barrier layer on a porous surface and concerns about penetration of precursors into low-k material.
“Customers are not reducing their interest in having SEMATECH look at all the available options for k values around 2,” says Monnig. “The CVD vs. SOD war is not over.”