Nikon enhances optics to extend e-beam lithography*

04/01/2002

New work suggests that using different optical parameters can extend e-beam (EB) lithography down to 0.05μm, and improve its throughput by controlling the electron interactions that usually limit both resolution and throughput. These Coulomb interactions haven't been well understood in the reduction optics we now use, but we're finally getting results from both simulations and experiments showing that these problematic interactions can be controlled by making major changes in the parameters of the EB optics. Using a small convergence angle and a large subfield controls the Coulomb interactions and significantly reduces the space-charge effect. With these new parameters, we can get throughput of 30-40 wafers/hr at the 0.05-0.035μm node (Fig. 2).

Figure 1. Development map for EB infrastructure. The EB optics and the vacuum stage are integrated and ready for testing this year, and delivery to Selete in 2003, according to Nikon. (Source: Nikon)Click here to enlarge image

Not only does this high-energy EB cut a very sharp pattern, but the small convergence angle also delivers depth of focus 10x better than optical lithography. This greater depth of focus maintains more even lines over a wider range, providing a wider process margin. It can make 0.07μm contact holes, 0.045μm lines and spaces, and individual 0.035μm lines. Problems with the development process, though, still cause collapse of these fine resist lines below 100nm.

Figure 2. New approach to EB optics. (Source: Nikon)Click here to enlarge image

Other companies are making good progress developing the masks and mask-data software needed for mass production with EB lithography. Japanese suppliers are currently providing 2μm silicon stencil masks up to 100mm dia., and expect to have 200mm masks available by mid-2002. Since we use a 4x reduction process, everything about making and positioning the mask is easier than with the 1x masks others are trying to use. But if a 1x mask is developed, our new EB optics would still allow higher current doses, and thus, even higher throughput. Several vendors have already developed the necessary mask-data conversion software.

Figure 3. Nikon's roadmap for EB development. Click here to enlarge image

Nikon has also gotten its technology for stitching subfields together down to errors of only 6nm in the x direction, though the y direction still requires correcting with magnetic deflectors. We've developed a high-speed, high-precision air stage that works in the vacuum required for EB, using differential pumping to remove the air in stages.

The servo control of this stage moves at more than 100Hz, and the vacuum within the chamber remains at ≤10^-4Pa. As of the end of 2001, we had integrated the wafer alignment, mask alignment, and autofocus systems already used in Nikon optical scanners with the EB optics. — Takeru Yamaguchi, Nikon Precision Co., Tokyo, Japan