by Ken MacWilliams, VP, Applied Materials’ Maydan Technology Center
There are several promising candidates for patterning circuits at 22nm and beyond: extreme ultraviolet lithography (EUVL), nano-imprint, direct write, and optical double patterning methods. But during an afternoon of spirited debate at a forum hosted by Applied Materials during SPIE, it became clear that double patterning is the only lithography solution that has the maturity to deliver the necessary balance of performance and cost required for 22nm volume production.
Panel members represented different segments of the industry: Jongwook Kye from AMD, Burn Lin from TSMC, Steve Radigan from SanDisk, Milind Weling from Cadence, Grant Willson from the U. of Texas at Austin, and Ken MacWilliams from Applied Materials as moderator. Audience members also actively participated through interactive survey devices.
Both panel and audience overwhelmingly agreed that double-patterning will be the dominant litho technique for the next few years. This might be surprising given the diverse needs of the various branches of the industry, but Cadence’s Weling pointed out that joint developments up and down the semiconductor food chain and the reduction of players through consolidation have limited the bandwidth to pursue multiple approaches, and one must win out.
The term double patterning (DP) actually encompasses several different process flows, each of which print features in multiple steps in order to double the pattern density. There are two main approaches: double exposure (DEDP) which uses two masks, and self-aligned double patterning (SADP), also known as spacer-based patterning, which generates pairs of features from a single parent. SADP was the audience favorite, but the panel was more evenly split. AMD’s Kye stated that in the short term a variant of DEDP called litho-freeze-litho-etch (LFLE), in which two photomasks are used with the photoresist being stabilized in between, was most useful today, but that SADP would soon become prevalent for logic as well as the design tools mature.
Applied’s Chris Bencher further argued that the equipment and process flows are in place today to allow SADP to be robust and cost-effective for memory and logic below 20nm. “SADP has been described as a stop-gap solution,” commented Bencher, “but it’s a widely-adopted and production-proven stop-gap. Manufacturers can begin their next-generation designs with confidence that they will be manufacturable using SADP.”
SEM of 11nm node SRAM structure (22nm half-pitch) demonstrates scalability of SADP technology. (Source: Applied Materials)
The growing maturity of double-patterning design tools was highlighted by Jason Sweis, who gave a live demonstration of Cadence’s design decomposition tools for 22nm SADP in action. The fascinated audience watched a circuit automatically converted into multiple photomask patterns, and saw the effect of layout adjustments calculated in real-time.
There was also strong agreement that cost-of-ownership is the biggest lithography challenge currently facing the industry, and applies to all litho options. SanDisk’s Radigan was the only dissenter, pointing out that the industry has always been able to engineer cost out — but without a viable, high-fidelity patterning technology there can be no progress at all.
So what technology will dominate lithography in 2016? For TSMC’s Lin, the ultimate foundry patterning technique is direct-write using multiple electron beams (MEB) that would eliminate masks entirely. This mature technique only lacks speed to deliver the flexibility that foundries crave to satisfy their customers’ compressed timescales. UT’s Willson countered that the technical hurdles to speeding up MEB are insurmountable; nanoimprint lithography (NIL) with the right level of development funding, he said, would deliver the pattern fidelity and low cost to meet the patterning challenges for several technology nodes. Radigan and Kye in turn expressed doubts about the robustness of NIL, explaining that once entrenched, double patterning will continue to evolve and remain dominant for many years.
Everyone seems to have doubts that EUVL, once the leading candidate for advanced lithography, will ever be ready for production. As Weling put it: EUVL has gone from “if”, to “when” and now back to “if” again.
Ken MacWilliams is VP of Applied Materials’ Maydan Technology Center.