An ever present need exists in the field of semiconductor lithography for delineating smaller features to support the trend towards faster, cheaper, and denser semiconductor chips. Lithography is the most expensive component of the semiconductor manu...
An ever present need exists in the field of semiconductor lithography for delineating smaller features to support the trend towards faster, cheaper, and denser semiconductor chips. Lithography is the most expensive component of the semiconductor manufacturing process, consuming 35%–50% of the total production cost. This cost is expected to escalate over the next several technology generations. For many chips, the mask production alone can be as high as 50% of the lithography cost. However, electron-beam mask production tools in use today are not scalable to the mask-writing requirements of the next decade because present-day tools use a single beam to serially expose an entire lithography mask.
This dissertation describes a multi-electron-beam lithography technology based on a MEMS-fabricated multi-blanker chip. Demonstrations of multi-blanker chips modulating eight beamlets in a 20kV test stand are presented. Images of the beamlets are viewed on a YAG scintillator and captured by CCD camera, and line-scan data are obtained from Faraday cup/knife-edge experiments. A multi-blanker system is proposed that can deliver 1μA of beam current at 50kV into 144 beamlets. A 6-inch mask for the 100nm technology generation can then be exposed in 40 minutes using a 15μC/cm<super>2</super> resist. Results from a novel Coulomb interaction Monte Carlo algorithm indicate that the multi-blanker can support 3μA of beam current while maintaining a 100nm spot size using present-day electron optics technology.
In addition to often reported stochastic blur, electron beams induce a space-charge lens that introduces third-order aberrations that may prevent electron projection technology from becoming a viable replacement to optical lithography at the 70nm technology node and beyond. A technique is presented to dissect the beam blur predicted by Monte Carlo simulators into the respective third-order aberration effects plus residual stochastic beam blur. For example, space-charge induced aberrations in a typical 250mm length column with 4:1 telecentric projection reduction optics is predicted to result in nearly 300nm of blur at the corner of a 250μm image field at 20μm of beam current and 100kV beam energy. Stochastic beam blur contributes an additional 130nm of beam blur.