Few people take imprint lithography completely seriously, even after it wiggled its way on to the ITRS roadmap. When Mark Melliar-Smith, the CEO of Molecular Imprints Inc. (MII) took the stage at the SPIE Advanced Lithography conference here, he did a credible job of outlining the challenges facing imprint, which uses UV light to harden a monomeric liquid.
He gained attention by saying that in the very near future a chip manufacturer will announce that it is using imprint. Later in his speech, he said IBM is using imprint to develop a non-volatile “storage class” memory technology.
Another customer is using imprint in a FinFET research project that targets sub-30-nm dimensions. Also, there were several presentations at SPIE on using imprint to pattern interconnect dielectric levels.
Even so, the knee-jerk reaction toward imprint is that the 1:1 masks, or “templates” as the imprint folks call them, are the Achilles heel of the imprint approach.
Melliar-Smith, a former Bell Labs manager who was the CEO of Sematech in the late 1990s, said the templates are indeed “critical” but manageable with modern E-beam mask writing tools. Costs can be reduced by creating multi-die templates.
Overlay is equally important, particularly in a mix-and-match situation where imprint is used for some layers and optical for others. MII is at 20-nm overlay now, not the 5-nm where the company needs to be. “We still have work to do. Better temperature control is one thing we are working on. We feel the overlay issue is an engineering problem, not basic physics.”
Defectivity is still far from the target.
Throughput is gated partly by the time it takes for the polymer fluid, which has a low viscosity, to spread and fill in the template pattern. That fill time is now about 10 seconds, and in the lab has been reduced to three seconds. The problem partly can be dealt with by using multiple imprint heads on one machine, and Melliar-Smith said a quad-head design is being studied. He urged the audience to consider throughput in terms of how many wafers can be processed per million dollars of investment.
Melliar-Smith said the imprint lithography field has several ways to get its foot in the door. Bright LEDs for lighting applications are a big market. He wowed the audience by showing photos of next-generation hard disk drive substrates with 500 megabits per square inch of tiny magnetic pillars, patterned with imprint.
If the HDD industry goes to imprint for its highest-density products, MII will gain the financial traction needed to ready higher-throughput systems for chip manufacturing.
That’s the company’s biggest challenge. Until the chip industry takes imprint seriously, MII must run as a startup, extending the time it will take to develop systems that could be used in CMOS fabs. With more money, more engineers, and more infrastructure partners, imprint can compete.
“We have the temerity to think imprint is a better approach than EUV,” he said.