Issue



maskmakers surprised by early 0.18-UM demand


11/01/1996







Maskmakers surprised by early 0.18-?m demand

In other mask news, photomask producers are reporting surprising early demand for 0.25- and even 0.18-micron prototype masks, as chipmakers rush to break new technological ground.

"Right now, we`ve got three generations of prototypes in here - 0.18, 0.25, and 0.35 micron," said Preston Adcox, president and chief operating officer of DuPont Photomasks Inc. "I don`t know what it means in terms of production, but we`re doing 0.18 micron right now. We`re seeing it from logic and memory companies. The parts are here and we`re scrambling to make them. Everybody`s aggressively stepping up."

Ed Grady, newly appointed head of KLA Instruments` Reticle and Photomask Inspection Division commented, "The rate of change has increased a generation in the last eight months. At the beginning of the year, people were saying they would want 0.25 micron in mid 1997. Now they`re saying that by early `97 they want 0.18 micron prototypes."

Grady attributed some of the phenomenon to the rapid price drop in 4-Mbit and 16-Mbit DRAMs, which has caused many producers to accelerate their shift to denser designs. All memory-producing areas, including Japan, South Korea, Taiwan, and the US have expressed interest in 256-Mbit (0.25 micron) and even 1-Gbit (0.18 micron) capabilities, he noted. "It`s a very interesting time," he said. "The mask industry had all those years in the doldrums, then for the past two years we were driven by pure capacity. Now it`s clearly shifted to being driven by technology."

One industry observer said that a great deal of time on advanced mask writing tools is being consumed by advanced prototype masks, in part because of high failure rates. One major merchant mask supplier reported that masks with sub-quarter-micron features represented about 5 percent of production at its most advanced facility in one recent month, but consumed some 50 percent of writing time on the facility`s most advanced litho system, with one of the test masks taking a full 31 hours to inspect.

Many of the masks are reportedly being used to produce test wafers for use in developing other processes, such as etch, but aggressive gate-shrinking efforts are under way at major logic providers like Intel and AMD. One additional issue is that mask CD errors will be magnified because the reticles will be used on exposure tools operating beyond their design limits. - P.D.