Why is the industry always so quick to write off optical photolithography for printing its integrated circuits? It's hard to believe, but back in the stone age, chip makers agreed that optical wavelengths would not be precise enough to print 1.0-micron features.
Even today, it still is difficult, if not impossible, to predict when optical lithography will finally run out of steam. But that hasn't stopped the industry from going all out to come up with the successor technology to optical.
The latest maneuvers were touched off by Sematech's recent attempt to pick the two leading candidates for next-generation lithography. Its anointment of extreme ultraviolet (EUV) technology and the Scalpel electron-beam system set off a chain reaction of announcements from lithography equipment vendors. They all were spending many millions of dollars to come up with the winning post-optical tools (see story in the Feb. 1 publication of SBN).
But just when it seemed like momentum was finally building again for next-generation lithography, Canon Inc. surprised the industry in mid-February by declaring that it had developed an optical extension technique that could keep optical lithography going for at least several more years -- perhaps even push back the need for non-optical tools for an entire decade.
"We see optical having a lot more legs than Sematech does," declares Phillip Ware, director of technical marketing for the Semiconductor Equipment Division of Canon USA Inc.
But the chip makers and their production equipment vendors have long been enamored with the follow-on lithography generation. Indeed, serious efforts have been underway for most of the 30 years that optical lithography has been the wafer fab workhorse. Many millions of dollars, for example, were poured into X-ray in the 1980s. And direct-write e-beam systems were pursued for nearly 25 years.
Despite these major R&D efforts, optical exposure tools managed to keep on improving and continued to push back the time when they would be no longer capable of printing the ever-smaller device feature sizes.
Optical lithography's staying power also is being strengthened now by the soaring costs of new chip plants. These "megafabs" can easily cost $2 billion now, far more than it takes to build an entire "greenfield" steel plant, for example. So to hold down capital costs, cash-squeezed chip makers are going all out to extend the operating lives of their existing 248-nm deep-UV tools.
These leading-edge systems initially were intended to build 0.25-micron ICs, but they are now widely expected to be included in next-generation process technologies for 0.15- or even 0.13-micron chips. Fab managers want to get "more bang from the buck" while their companies wrestle with uncertain market conditions.
That's what's leading them to try new optical tricks now. Motorola Inc. was the latest to come up with a new scheme. It plans to save a bundle of money by combining advanced phase-shift photomasks and optical proximity correction technology with its 248-nm tools to produce 0.10-micron polysilicon gates (see story in the Feb. 1 publication of SBN).
"Part of the puzzle here is that we're just now starting to tap many of these resolution enhancements or 'tricks,'" points out Canon's Ware. "Just as we start to think we're approaching the limits of optical," he says, "more of these sophisticated and innovative concepts keep popping out of the woodwork."
From now on, perhaps, chip makers and their equipment suppliers may not be so quick to write off optical lithography.