EUV gains as ASML/Applied venture ends e-beam lithography work

AUSTIN, Texas -- Momentum appears to have swung sharply towardextreme-ultraviolet technology (EUV) as the next-generation lithography solution following a recent decision by ASM Lithography and Applied Materials Inc. to dissolve eLith LLC, a joint venture formed 14 months ago to commercialize the Scalpel electron-beam projection lithography (EPL) technology developed at Bell Labs.

The demise of eLith is the latest in a string of events that may prompt International Sematech to focus its resources on EUV, dropping development work on EPL masks, resists, inspection tools and other supporting technologies, said Gerhard Gross, director of lithography at the consortium. Gross said he will argue in favor of that course at an International Sematech technology strategy meeting later this month.

ASML, based in Veldhoven, Netherlands, was the first eLith partner to withdraw from the venture. The decision was made after key customers said they preferred the extreme-ultraviolet solution being developed at EUV LLC, located in Livermore, Calif., said ASML marketing manager Mark Bigelow.

ASML's decision prompted Applied Materials of Santa Clara, Calif. to abandon the effort, which would have given Applied a toehold in the lithography sector, the only semiconductor-equipment sector in which it is not a major player.

"We opted not to continue after ASML decided to terminate its support for eLith," said an Applied Materials spokesman.

The Scalpel program started at Bell Labs, now part of Agere Systems, but in late 1999 the program took the critical step of enticing two of the largest commercial tool vendors - ASML and Applied Materials - to put their manufacturing muscle behind the concept via the eLith venture. Four semiconductor companies had voiced support for Scalpel: Lucent (now Agere), Motorola, Samsung and Texas Instruments.

Now, as ASML and Applied Materials wind down their joint venture, the future of the Scalpel technology appears to be in limbo.

David Eaglesham, vice president of electron device research at Agere Systems, said Agere will look for another commercial lithography vendor to take up the commercialization effort. "This type of program needs to be transitioned out to a commercial vendor, and we will look for ways to find support from other tool vendors," he said. "We believe there is room in the semiconductor technology road map for both the EUV and EPL approaches, because they have different cost structures. The masks for the Scalpel approach are less costly, for one thing.

"I would say it is a little bit premature to say that this is the end of Scalpel."

Meanwhile, work continues on another EPL approach: the Prevail system being developed by IBM Corp. and Nikon Corp. A source at IBM said a prototype Prevail system, incorporating technologies that improve throughput, is expected to be ready by the spring.

But G. Dan Hutcheson, principal analyst at VLSI Research Inc., predicted that neither the Scalpel approach nor the Prevail system will achieve sufficient commercial support.

And International Sematech's Gross said his personal view is that EPL should be dropped from the consortium's lithography program. Gross also said he believes the 13 members of International Sematech should commence a feasibility study of a maskless form of direct-write e-beam lithography, one that would use many thousands of electron beams arranged in multicolumn, multibeam arrays.

The Scalpel development group, which has never comprised more than about 30 Bell Labs engineers, has been wrapping up its development work on the EPL approach over the last few months. Bell Lab built a proof-of-lithography system several years ago, and eLith was to have built a second-generation Scalpel, with higher throughput. Several suppliers were in the process of building such critical subsystems as the stage and column.

As recently as November, the eLith partners were considering sites for an eLith development center that would have been independent of the Murray Hill, N.J., Bell Labs facility where Scalpel originated. Andover, Mass. - the home base of precision system manufacturer SSC Corp. - was said to be a likely headquarters for eLith, and a management team was being formed.

Bigelow said ASML's view now is that EUV technology is more extensible - being viable through several generations of process technology - than EPL. "And when you look at the mask costs and analyze the EUV and Scalpel approaches, the lower throughput of the e-beam systems is a significant burden," he said. "That is one reason why the customer support never materialized."

While eLith was getting established, the rival EUV approach was gaining support from lithographers worldwide. ASML itself joined EUV LLC in the spring of 1999. At next-generation lithography workshops organized by International Sematech last year, lithographers voiced a growing preference for EUV over the EPL approach, according to Gross.

The EUV engineering test stand is expected to be completed in early 2002, with a demonstration planned for this spring. Roughly 350 people are involved, from three national labs (Sandia National Laboratories, Lawrence Livermore National Laboratory and Lawrence Berkeley National Laboratory) as well as Intel, Advanced Micro Devices, Motorola, Infineon Technologies, ASML and Silicon Valley Group.

About 200 of those scientists and engineers are involved in the Virtual National Lab, made up of the U.S. government's national laboratories. About 50 are from the Extreme Ultraviolet Limited Liability Co., which includes Intel, AMD, Motorola and Infineon. And SVG and ASML each have assigned 50 to 70 engineers and scientists to the effort. (SVG is in the process of being acquired by ASML, pending U.S. government approval.)

"The industry is focusing down on EUV" partly because it is based on photons, as opposed to the more highly charged electrons used in the EPL approach, said Hutcheson at VLSI Research. "Agere has been dropping research projects that went on for decades within Bell Labs back in the days when AT&T was a cash cow . . . There are always people with money that seem to be willing to pursue technologies, so someone may pick Scalpel up. But I think Scalpel is dead."

Hutcheson believes IBM and Nikon will find it similarly difficult to find support for the Prevail system, which he said faces the same throughput problems - measured in wafers per hour - as the Scalpel approach. He further argued that the electron beams in EPL systems hit the wafer with charges that can change the electrical performance of the transistor gate oxide and other critical areas.

Gross, an Infineon Technologies assignee to International Sematech, said he will propose that the consortium "focus on EUV now and phase out EPL. At our next-generation lithography workshop and a second data review, when it came to the 50-nanometer node on the road map and below, there were twice as many votes for EUV as for EPL. Many people do not believe that EPL will work at the 35-nm node. They argue that EUV could enter at the 50-nm node and extend to the 35-nm node."

Gross said progress on 157-nm lithography systems is on track, allowing companies to put 157-nm scanners into manufacturing lines in 2004 and 2005, in time for introduction in the 100-nm (0.10-micron) technology generation. The 157-nm scanners would be the workhorse chip writers during the 70-nm technology node, but most technologists agree that the 157-nm tools cannot be extended to the 50-nm node.

By then, the EUV solution must be ready, including some way to repair minor defects on the masks and a more compact source of the 13-nm EUV radiation.

Because the time required to make an EUV mask set may be prohibitively long for many IC manufacturers, support is growing for direct-write electron-beam systems.

"My personal view is that there may be a collapse of the mask industry as we know it during the 50-nm node," Gross said. "The ASIC manufacturers need a shorter cycle time for mask creation. At the 50-nm node there is a need for a maskless technology, and I will recommend that we study it within International Sematech."

Gross acknowledged that myriad technical issues would need to be overcome for a maskless technology to work. "We are talking about as many as hundreds of thousands of beams in the array, and then the challenge becomes how to transfer the information onto the beams," he said.

"At a workshop on maskless systems a couple of months ago, an estimate was presented that for a 20-wafers-per-hour tool, about 10 terabits of data per second would need to be transferred to the array. That is just one of several major challenges."

Indeed, Hutcheson of VLSI Research said he has little faith that maskless direct-write systems will ever become practical. Even if a data set can be created that correctly describes the chip design, there is no guarantee that the beams will create the design correctly, he argued. Moreover, since each chip has 20 to 25 levels, the odds are against direct-write e-beam systems, Hutcheson said.

But Gross said that even if multibeam direct-write systems prove impractical, the development effort might not be for naught. "If we start investing now, by 2007 or 2008 we could at least have a fast mask-writing technology," he said.