SAN MATEO, Calif. A paper presented Friday (Feb. 28) at the SPIE conference disclosed recent work in support of the movement for small-field lithography. Co-authored by researchers from Numerical Technologies Inc. (San Jose, Calif.) and Photronics, Inc. (Dallas), the paper presents quantitative arguments in favor of using only part of the available reticle on advanced steppers.
Some mask shops and photolithographic equipment makers have been arguing for some time that the industry should consider building a line of steppers with reduced field sizes. The basic argument has been that the amount of investment necessary to make the edges of the current large-field reticles useful is disproportionate to their value in production.
But the study, presented by Numerical's senior R&D director Dipankar Pramanik, took a different approach. Pramanik's argument was to not wait for a new generation of equipment, but to save money by using only the center of available reticles on existing steppers.
Pramanik built a cost model for the entire lithographic process, including mask-making cost, inspection cost and printing cost, and reduced these numbers to a per-wafer basis for comparison. His estimates revealed substantial savings in mask-writing and mask inspection costs if a smaller area of the reticle was used.
But using a smaller area of the reticle means having a smaller number of die images on the reticle, which means more steps will be necessary to expose the entire wafer. So these savings had to be compared to the greater number of flashes necessary to expose the full wafer if a smaller number of die images were carried on the reticle.
That comparison was the key point. Pramanik's model showed that for an 8-by-8 mm die, reducing the number of die images on the reticle from the standard nine to just one would slash the total mask cost from $1.2 million to just $300,000. For the typical number of wafers taken by a high-volume SoC customer this savings completely overwhelms the added cost of printing, which Pramanik estimated at from $18,000 to $160,000 depending on process variables.
Assuming operators must do image-to-image inspection rather than the much more difficult image-to-data inspection required to have two die images on the reticle, Pramanik concluded that there would be substantial overall savings in putting only two die images on the reticle.