Direct-write e-beam system shows major promise for volume production

By Ron Wilson, Executive Editor -- 5/30/2008 7:00:00 PM

Fujitsu's eShuttle operation has made some waves in the ASIC prototyping area by using an electron-beam system to write patterns directly onto wafers' photoresist coating, bypassing the mask-making steps in manufacturing prototype and small-volume wafers. The company recently announced that it has fabricated 65-nm LSI and SRAM wafers using the technology. But in a paper delivered today at the International Conference on Electron, Ion, and Photon Beam Technology and Nanofabrication (EIPBN), eShuttle General Manager Shinji Sugatani lifted the curtain on a technology shift that could take the e-beam technology far beyond the world of prototyping.

The change hinges on a shift in resist chemistry taking place in EUV development labs, Sugatani explained. Current 193-nm resists work when photons stimulate acid-generator molecules to release an acid, creating a pattern in the resist. But resists for extreme ultraviolet lithography will probably use an entirely different mechanism: release of acid by ionization of a polymer molecule. This shift in resist technology is important because the new EUV resist will also work well with electron beams. In fact both EUV developers and eShuttle appear to have been moving in the same direction to improve the sensitivity of resist materials for their own purposes.

That convergence means that a resist-coated wafer could be patterned by EUV through a mask and by direct e-beam write, either simultaneously or in two consecutive steps, without having to develop the resist, strip it and recoat with a different resist between exposures. The significance of this goes well beyond chemistry. Sugatani pointed out that if both EUV and e-beam exposure can be used on the same resist layer, than a design team could produce a platform chip design combined with a small area of customizable logic. The platform portion of the die would be exposed using EUV masks and printed on an as-yet-hypothetical production EUV stepper. The custom portion of each mask layer would be written by the e-beam system. Then the resist would be developed conventionally.

Sugatani suggested that this combination of fast resist, dual-exposure, and relatively small custom area would permit throughput high enough to qualify this as a full production process, not limited to prototyping. So adding an e-beam column to the lithography cluster—either in the form of a dual-column stepper or as a second exposure stage—would offer customers the ability to produce a standard SoC platform chip at production volumes and costs, and yet customize some of each chip virtually on the fly.

Another intriguing possibility exists. Today the resist eShuttle uses is derived from conventional 193-nm resist formulae. So the e-beam system could possibly share a single resist formula with a 193-nm immersion stepper, albeit at significantly lower throughput. That suggests that, if EUV continues to be the technology of the future, the industry might be able to get one or two more generations at barely acceptable throughput by using 193-nm dual-patterning for the majority of the features on a wafer, and touching up, as it were, the most critical features with an e-beam system included in the cluster. It would be expensive and slow, but perhaps not in comparison to some of the worse-case projections for EUV systems.

In any case, it is certain that eShuttle is now making a viable business out of producing prototype 65-nm wafers using direct-write e-beam technology. And it seems very plausible that e-beam direct write, used in conjunction with either advanced 193-nm lithography or EUV, may have a role in the full-production fab lines of the future.

© 2009, Reed Business Information, a division of Reed Elsevier Inc. All Rights Reserved.