Nikon, the main driver behind EPL developments, said it has achieved a 23-nm critical dimension error, which satisfies an internal R&D milestone set at 25-nm for a 65-nm process node R&D tool. The achievement also opens the possibility of further enhancing the accuracy to the 15-nm level required for commercial 65-nm tools and 45-nm R&D tools, scheduled to be available in 2005.

Nikon disclosed plans to ship a prototype EPL machine to Semiconductor Leading Edge Technologies Inc. (Selete), a Japanese chip-making consortium, (see February 13 story) and the existence of the multi-company EPL Forum (see February story) at the International Forum on Semiconductor Technology (IFST), being held in Monterey, California.

Nikon engineers are expected to provide details of the R&D tool at SPIE's 28th annual International Symposium on Microlithography to be held next week (February 23 to 28) in Santa Clara, California. The Selete system is expected to ship before the end of June of this year. In parallel with the first system, Nikon is developing a second 65-nm tool that will be completed by the end of the year.

Once the stitching issue has been resolved, a remaining issue is how to achieve a 30-nm overlay accuracy of pattern alignment for the 65-nm tool.

"It's different from stitching accuracy. That required new development. The alignment sensor is the same one that is already used for optical steppers and scanners ...we don't think it will present a difficult challenge," said Kunio Kurishima, manager of sales support section of Nikon Precision Equipment Company.

EPL is one of the next generation lithography (NGL) technologies proposed by Nikon. Japan's Selete organization adopted it as one of its next generation lithography development projects in 2001 and has been strongly supporting the development since then.

Traditional electron beam lithography has used a thin beam to "write" details on photo-resist but this approach results in very long write times for complex chips creating unacceptable productivity issues for the volume production of digital chips.

To increase productivity EPL uses an electron beam combined with optical projection to characterize the chip. In the case of Nikon's machine EPL exposes a 0.25-millimeter by 0.25-millimeter square sub-field in one shot, but multiple sub-fields have to be accurately aligned together.

Each sub-field pattern is projected from a 1-mm by 1-mm stencil mask made from 2-micron thick silicon. It is the robustness of the silicon that limits the mask to the 1-mm by 1-mm square size at present.

The exposed field is about 2,500 times larger compared to that of direct writing electron beam exposure systems, but it is still too small to expose one whole LSI by one shot. To cover an area of 20-mm by 25-mm about 8,000 shots have to be stitched together. Thus it becomes essential to control stitching accurately. At the EPL conference held last December, participating engineers' interest focused how accurately each sub-field could be overlapped.

Nikon set its engineers a development schedule for EPL in which 25-nm level stitching accuracy was to be achieved by December 2002. As of December 29, Nikon engineers achieved a stitching variation of 23.1-nm in the X direction and 22.6-nm in the Y direction.

"We cleared the target on time," said Kurishima.

An even more difficult 15-nm accuracy is required for commercial tools but is achievable by building in compensation for systematic errors within the system, said Kurishima.

Nikon engineers co-developed an electron beam exposure unit with IBM Corp. but mechanical units, such as the wafer stage, separately. Those units were assembled into one system last summer. Since then Nikon engineers have been tuning the system. Starting from about 100-nm stitching gap immediately after the assembly of the system, engineers have edged down the variation by adjusting aberration, development of a wafer stage with uniform velocity and development of a control system that feeds data from the wafer stage to the electron optic system without delay.

The final target of stitching accuracy for the first R&D tool to be installed at Selete's lab in June this year is 20-nm. Kurishima said that the level could be achieved by that time.