FISHKILL, N.Y. -- Three advanced systems for producing deep-submicron features on ICs are headed for a showdown that could leave the industry with one clear wi nner by November. At that time, Sematech is scheduled to make the "decision of the century," choosing among soft X-rays, extreme-ultraviolet lasers and electron beams for the 0.13-micron generation of lithography systems. The choice--if Sematech actually makes it--could literally determine the shape of the fab of the future.

Each of the contenders has a powerful sponsor willing to back it to the end. X-ray lithography, accomplished using an electron storage ring to generate synchrotron emissions, has the support of IBM Microelectronics. Projection printing with e-beams is the approach for Lucent Technologies' Bell Labs. And mighty Intel Corp. is behind the use of extreme-UV laser light sources. None of the companies appears willing to budge on its public stance that its solution is the most viable for sub-0.18-micron lithography.

The choice thus falls to Sematech. But some believe the influential Austin, Texas-based consortium might sidestep the issue. "Most people think that Sematech won't make a rea l decision," one source told EE Times. "They'll just list the pros and cons and leave the choice up to the market."

IBM has used its enormous electron storage-ring system both for research and for doing critical layers for the pilot line at its Fishkill fab. With enormous potential resolution, the system has produced intentional features in the 0.10-micron range, and has been observed to accurately reproduce mask scratches 0.02 micron wide. These results, plus the relative ease of adapting the X-ray source to production, have most major Japanese fabs looking closely at X-ray, sources said.

But the system's resolving power is matched by its ability to intimidate. The heart of the device is an electron racetrack, enclosed in a 10-10 torr vacuum. Electrons are bent around the course by two giant superconducting magnets, each cooled by liquid helium and each weighing tons.

X-rays leaving the ring are directed down evacuated columns, through the five-foot-thick concrete walls of the ring's containment room, and into a series of mirrors and filters. There, they pass through proximity masks and strike the wafer in a relatively conventional stepper.

To operate, the storage ring requires a continuous source of liquid helium, which IBM supplies from a rather elaborate refrigerator. Ion pumps help maintain the vacuum. Electrons are injected into the ring by a linear accelerator, which is in turn powered by a bank of klystrons. A TV-broadcast transmitter feeds energy into the ring, and a bank of enormous power supplies drive the magnets.

Compared with this array of equipment, Lucent's Scalpel printer seems almost impossibly simple. Scalpel used a wide beam of focused electrons to scan across a 4:1 projection mask. The e-beam and its optics are relatively conventional--familiar to any electron-microscope aficionado. No unusual power supplies, temperatures, vacuums or magnetic fields are required. The system sits in a corner of a lab at Lucent's Murray Hill, N.J., research center.

Mark Pinto, director of the Silicon Electronics Research Lab at Lucent, said the system has been producing 0.08-micron features. That success, said Pinto, has led to an enthusiastic following. Lucent will soon announce partnerships with both equipment vendors and semiconductor companies, which will join in the development of the next Scalpel generation. With the device technology worked out, only throughput needs to be improved, Lucent said.

Yet Scalpel faces problems as well. The electron-beam technique is not projection lithography in the traditional sense. Ordinary masks, such as those used by the IBM and Intel systems, create a feature on the wafer by blocking the light (or X-ray) beam with an identically shaped feature on the mask.

Because of the nature of electron optics, the masks for the Scalpel system use an entirely different principle: scattering. Scalpel masks don't block electrons with solid shapes, they scatter them with apertures. If you want to leave an area of the resist unexposed, you must create a donut -like or slit-like feature on the masks that will scatter the e-beam.

An aperture in the optical path blocks the scattered electrons, so only electrons that did not pass through such a feature reach the wafer. Producing such masks looms as a technology challenge.

  Go To Previous Story

Go To Next Story  

Back to This Week's News