Invited Paper PS-MoA2
Following Moore’s Law – How Many Knobs are Enough?

Monday, October 20, 2008, 2:20 pm, Room 304

With the introduction of 3x technology, more new materials are introduced into semiconductor devices and their manufacture. Yet, the basic etch challenges do not change qualitatively: selectivity, profile, uniformity, damage, line edge roughness and more. Quantitatively, of course, virtually everything changes: with each shrink, the difficulty in meeting the tolerances increases along with cost. The tightening of tolerances on the die, the wafer, and the system places un-precedented demands on system and sub-system variability reduction – active and passive. For repeatable, high yield, high output manufacturing, minimizing variability is essential. Yet the requirements to shrink stimulate increased demands for more control “knobs.” Some knobs are provided to compensate for asymmetric limitations in design or inherent non-uniformities. Other knobs are provided to expand process flexibility in the hope of facilitating recipe optimization. The irony is that more control knobs means increased variance in the control parameter vector leading directly to more difficulty in matching chambers and narrower process windows, leading to more manufacturing excursions and downtime. More control knobs also means more recipe complexity. Despite impressive advances in simulation capability, plasma process development remains an empirical endeavor. With the most recent advances in etch production equipment, the number of recipes that have a measurable impact on the wafer are already more than 1 trillion. Can any production recipe be said to be truly optimal? Do more control variables help or hinder our ability to meet next-generation technology challenges? And do they save cost or add cost?