The decision has revealed deep fault lines in the industry's road map for lithography, which has been the key driver behind Moore's Law scaling.

"The lithography vendors are going to study this very, very hard. They are not likely to do 157 unless the semiconductor companies say they will pay for it," said Risto Puhakka, vice president of operations at VLSI Research Corp. (San Jose, Calif.). "To do 157 will require money, and lots of it. To pay off, 157 has to be a three-node technology, just as 193 will be used at the 90-, 65- and 45-nm nodes."

Intel announced May 23 that it plans to extend 193-nm scanners to the 45-nm node, where the half-pitch actually is about 70 nm. After that, Intel will turn to extreme-ultraviolet lithography (EUV) for the 32-nm node expected at decade's end.

Intel's bombshell announcement angered some, who argue that Intel insisted that 157-nm lithography be put back on the industry road map in 1999 as a hedge against EUV's late arrival. These proponents are optimistic that if 157-nm lithography is introduced at the 45-nm node, it may be extended to the 32- and 22-nm nodes with immersion techniques.

Others said the lithography industry simply cannot afford to develop so many new technologies, and that removing 157 nm from the road map would free up resources needed to develop immersion 193-nm technology, EUV, and perhaps direct-write (maskless) e-beam and electron-beam projection lithography.

Even X-ray lithography is still being pursued for critical layers by Jmar Technologies Inc., a San Diego-based company that has financial backing from the Department of Defense.

Puhakka said the lithography industry, in normal times, has an annual revenue of $3 billion to $4 billion, with R&D of perhaps $500 million. "The problem is that once a company starts to take a new technology into a real production environment, it becomes very, very expensive. The industry cannot afford to develop all of these technologies, so choices have to be made. Intel has made its choice," he said.

Following Intel's announcement, Texas Instruments Inc. was quick to state its support for 157-nm technology. TI fellow Allen Bowling said TI agrees that the industry will try to extend 193-nm tools through the 45-nm node "because it always makes the most financial sense to extend a lithography wavelength as far as possible."

Nonetheless, Bowling said, "as with 193-nm lithography, we feel that the industry will find ways to implement 157 nm and extend it to at least two process nodes. The argument was made that 193 nm would only work for one node too, and look how far it has carried the industry."

Unlike Intel, Bowling said, TI thinks there is "significant risk in putting all your eggs in one, expensive basket this early. TI does not feel that EUV can be ready for the 32-nm node, for production in 2009."

To get EUV tools ready for development starting in 2007, significant progress would have to be made in the vacuum operation, reflective optics and reflective masks. "We feel that EUV will require a lot more development, and it will be at least 2011 or later before EUV is ready for production use," he said.

A spokeswoman for ASML Holding NV in Veldhoven, Netherlands, said ASML is still committed to 157-nm lithography, with its first production tool expected as early as next year. "If our customers want 157 nm, we will develop 157 nm," she said.

Nikon Corp. said it is evaluating its tool options in response to Intel Corp.'s decision. Frank Masciocchi, vice president of sales and marketing at Nikon Precision Inc., said Nikon is evaluating a plan to move to 157-nm technology or to 193-mn tools with immersion capabilities. "We're evaluating both technologies, and we'll make a decision by the end of this year," he said. Nikon also is investing heavily in electron-beam projection lithography.

ASML is working on immersion technology with the Rochester Institute of Technology. ASML, Canon, Nikon and others are expected to deliver their progress reports to an International Sematech-sponsored immersion workshop, planned for mid-July at IBM Corp.'s Almaden research center in California. That Sematech meeting follows Semicon West in San Francisco.

If "wet" or immersion 193-nm tools appear to be feasible, that could further squeeze the space where 157-nm tools are expected to come into use. If much of the industry adopts wet 193, it could remove the economic incentive to develop the 157-nm resists, pellicles and other key infrastructure components.

EUV LLC's test stand is a proof-of- concept vehicle for EUV lithography.

Peter Silverman, director of lithography capital equipment at Intel, said the company plans to use dry 193-nm lithography at the 45-nm node and then move to EUV for critical layers of the 32-nm node. "Immersion 193 is at best a one-generation tool that would barely cover the needs of the 32-nm node. The advantage of EUV is that it is an extensible technology that could be used for the 32-nm node and several generations beyond that," Silverman said.

Ken Rygler, a consultant at Rygler and Associates Inc. (Austin), said Intel's position is influenced by the fact that it runs two very high-volume products through its fabs: flash memories and microprocessors. That allows Intel to spend more on each mask set than a foundry could afford.

Companies like TSMC are more interested in immersion because it extends the reticle technology used today, rather than forcing a move to the reflective masks required by EUV lithography.

If 157-nm tools also could be extended by immersion, albeit with a different liquid than the water that seems to work well with 193-nm tools, then "the industry may not need EUV at all," Rygler said, adding he is a member of the "optical is forever" school of thought. Immersion could take 157-nm tools to the 22-nm node, which is near the end of the line for CMOS according to the International Technology Roadmap for Semiconductors, he noted.

Motorola Inc.'s Joe Mogab, a director of the Advanced Process Development Laboratory here, said it appears likely that 193-nm lithography can be extended to the 45-nm node.

Bernie Roman, manager of the ad-vanced-lithography program at Motorola, said the lack of a soft pellicle to protect the masks is a stumbling block for the 157-nm program. "Motorola will not deal with a hard pellicle," Roman said flatly.

The hard pellicle, as currently defined, is an 800-micron-thick glass structure that tends to introduce random defects to the printed pattern. In manufacturing, adjusting the scanner and mask to the pellicle would seriously impair throughput. Thus far, the industry has yet to find an organic material that could be used with 157-nm scanners, Roman said.

A different source, asking not to be identified, said Intel's decision is causing resentment among cash-strapped equipment suppliers. In 1999, Intel executives realized that EUV would be introduced much later than originally hoped. After Intel factory managers demanded that 157-nm development be put back on the table, the major lithography, resist and infrastructure vendors did so. Billions of dollars have been spent thus far on 157-nm development, she argued.

"After all of that, Intel just walks away? Without Intel in, what kind of market is there for 157?" she asked. "Who is going to walk up to the plate?

"I think Intel always wanted EUV, and now that it is starting to look like EUV is going to be a lot more expensive to develop than earlier thought, Intel is trying to get rid of any other options, like 157. It wants the industry to put all of its eggs into the EUV basket," the source said.