An obscure optical effect that had faded from view for more than a century suddenly has become a hot topic for microelectronics producers. New studies show that this effect, called intrinsic birefringence, could incapacitate the next generation of factory tools for making chips.
Researchers at the National Institute of Standards and Technology (NIST) in Gaithersburg, Md., began circulating this revelation in May. It probably will force engineers to redesign multimillion-dollar machines already slated for production, they say.
No one seems yet to know the degree of the challenge nor a pathway to its solution, says Mordechai Rothschild of the Massachusetts Institute of Technology’s Lincoln Laboratory in Lexington. The NIST researchers expect the problem to crop up in several years in the chip-making step known as lithography.
In that process, a laser beam shines through a pattern, or mask, to make an image of a circuit layer. Then, lenses shrink and focus that image onto a coated silicon wafer where the pattern becomes imprinted. That imprint then guides chemical fabrication of the chip’s components. Machines called steppers repeat this imprinting many times on each wafer to make scores of chips.
To cram more transistors onto a chip, manufacturers shrink wires and other features. But each reduction requires a laser with shorter-wavelength light. The most advanced factories use ultraviolet (UV) radiation of 193 nanometers, says Chris Van Peski of International Sematech, an industry research consortium based in Austin, Texas. Next will be 157-nm lithography, he adds.
However, that move will force the industry up against intrinsic birefringence, says NIST physicist John H. Burnett. In this phenomenon, which becomes more prominent as wavelengths decrease, light doesn’t travel uniformly through a lens. That nonuniformity blurs and otherwise distorts images.
So far, birefringence hasn’t been a problem for the all-glass optics in current steppers. But since glass absorbs more UV radiation as the wavelength shrinks, stepper makers have begun using nonabsorbing crystals such as calcium fluoride in the machine’s optical components. Today’s 193-nm steppers already include some of these crystals, and 157-nm machines will use them nearly exclusively.
While engineers had long observed birefringence in crystals under pressure or tension, no one in industry considered the possibility of intrinsic birefringence as a manufacturing issue, Burnett says. The lack of concern resided in the tacit assumption that if the crystals are as symmetrical as the ones in their steppers, they wouldn’t exhibit birefringence.
Engineers might have been more savvy had they studied the late-19th-century work of physicist Hendrik A. Lorentz, who conjectured that even highly symmetric crystals would exhibit a small intrinsic effect. At the time, Lorentz didn’t have tools sufficient for testing his hypothesis. Because the effect is so small at visible wavelengths, it remained obscure.
Using today’s powerful tools, Burnett and his colleagues found intrinsic birefringence in calcium fluoride. At 157 nm, Burnett says, the material exhibits 12 times the effect tolerable in current stepper designs.
At press time, Burnett was slated to present his results on July 18 at the trade conference Semicon West in San Francisco.
It clearly is a problem, but I don’t think it’s a showstopper, says Rothschild. Fortunately, he notes, the NIST group found the difficulty before manufacturers actually started making and shipping the 157-nm machines.