Lens problems scupper 157nm

22 Aug 2003

Moore’s Law is still holding true, with the insatiable demand for ever-faster semiconductor chips driving the development of lithography tools that operate at increasingly short wavelengths.

From Opto & Laser Europe September 2003

Moore’s Law is still holding true, with the insatiable demand for ever-faster semiconductor chips driving the development of lithography tools that operate at increasingly short wavelengths. The chips that are currently being churned out in their millions rely on the 193 nm argon fluoride lasers used in today’s state-of-the-art stepper tools.

The next major transition was expected to be a shift to 157 nm exposure with fluorine lasers in 2007. This technology had been pencilled in for what is dubbed "the 45 nm node". But now Intel says that it intends to leapfrog the 157 nm wavelength and focus its attention on extreme ultraviolet (EUV) technology instead.

Caught by surprise Few in the industry were prepared for Intel’s decision in May, says Malcolm Gower of UK-based stepper tool developer Exitech. "Intel was one of the original champions of 157 nm technology, so this decision surprised many in the industry," he said.

According to Peter J Silverman - Intel’s director of lithography capital equipment development - the decision was not taken lightly. "We wanted to use [157 nm] very badly," he told Opto & Laser Europe. "But what became apparent earlier this year was that not only would the tools not be ready, but nor would the pellicule or the photomasks. The problems could not be overcome in time for the 45 nm node."

The chief difficulty is with the 157 nm lens material, calcium fluoride (CaF₂). Silverman says that although this material had seen many years of trouble-free use in other industrial applications, its optical properties had not been fully understood.

"It turns out that CaF₂ has an intrinsic birefringence," he said. "Everybody thought that this was due to stress in the crystal and that a zero-stress crystal would be the answer. That was wrong - the birefringence is there anyway."

The birefringence problem can be resolved by using two different orientations of the crystal. However, because CaF₂ lenses of sufficiently high quality take a long time to manufacture - about three months from the powder stage to lens-quality material - the 157 nm development cycle was held up by 6-12 months.

For full-scale production to begin in 2007, Intel needed its development tools in place by the end of next year. The delayed lens production, coupled with some more minor problems, appears to have led the firm to conclude that 157 nm lithography production tools might not be ready in time.

Schott Lithotec, the German manufacturer of CaF₂ lenses for lithography tools, is disputing Intel’s conclusions. Schott says that with ongoing improvements in its higher-grade lens material it will be able to deliver suitable lenses by 2007.

Indeed, Intel’s move has yet to be followed by the rest of the semiconductor industry, and the likes of Texas Instruments and IBM have been saying publicly that they still support 157 nm technology. ASML, the market-leading Netherlands-based lithography systems manufacturer, told Opto & Laser Europe that it still intends to build 157 nm stepper tools, at least until other chip manufacturers begin to follow Intel’s lead.

Production schedules "For the 2007 node, lots of people are hoping that [157 nm lithography] will still be available," admitted Silverman. "However, those companies have the same data that we have and unless their schedule is a year behind Intel’s then the tools just aren’t going to be ready in time."

Although some smaller manufacturers may be able to produce low chip volumes using only one or two steppers, Silverman is sceptical about the overall prospects for the 157 nm wavelength: "Personally, I don’t think anybody will use 157 nm for the 2007 node," he told Opto & Laser Europe. "The door is still open for its use after that, for example in the 32 nm node (due in 2009), but then the issue becomes economic - the longer it gets delayed, then the less generations it can be used for."

So having ditched 157 nm technology, what are Intel’s plans for the 2007 production cycle? Much is expected of the EUV tools scheduled for the 32 nm node and beyond, but EUV certainly won’t be ready by 2007, so Intel now has only one option for 2007 - to extend 193 nm steppers. These are currently being used to make chips at the 90 nm node, and Intel now plans to extend them to the 65 nm node in 2005 and the 45 nm node in 2007.

To do this, the numerical aperture of the optics used to image the patterns must be increased. Silverman says that its suppliers have promised lenses with a numerical aperture of 0.93, which will be good enough for 193 nm steppers to be used in 2007. He admits, however, that to make 193 nm work at such a high resolution, Intel will have to use all the optical tricks available to it. "Even then the process will have a narrow margin, but we are sure that we can do that."