Toolmakers Ease Double Patterning Throughput Hit

At this year's Sokudo Lithography Breakfast Forum at SEMICON West, toolmakers presented their latest throughput achievements, detailing a variety of enhancements designed to increase productivity.

Aaron Hand -- Semiconductor International, 7/16/2009

These days, double patterning is widely regarded as necessary to bridge the gap between 193 nm immersion and extreme ultraviolet (EUV) lithography. But that doesn't mean chipmakers are any more eager to take the hit in throughput that will inevitably come along with the technology, regardless of the scheme that's used. However, exposure tool and track suppliers are working to boost throughput capabilities as much as possible, to make the transition a little less painful.

At this year's Lithography Breakfast Forum, organized and sponsored again by track supplier Sokudo Co. Ltd., toolmakers presented their latest throughput achievements, detailing a variety of enhancements designed to increase productivity. The forum, titled "Double Patterning Jumps the 200 wph Hurdle," aimed to alleviate some concerns about double patterning for the cost-sensitive industry, as well as to showcase Sokudo's latest 250 wph Sokudo Duo track system.

Essentially, many chipmakers will be forced to move to double patterning schemes, depending on their device technology, around 32 or 22 nm technology nodes. This is because of the k₁ factor, which corresponds to the available process window, and which becomes a limiting factor by k₁=0.28. Tom Wallow of GlobalFoundries outlined the problem from a logic maker's perspective, noting that, at k₁=0.3, essentially logic has already begun to use double patterning of a sort. At the 32 nm node, k₁ would be 0.224, making a pattern split necessary and thereby taking k₁ back up to 0.45.

In his opening overview of the issues, Wallow presented the situation for patterning at an 80 nm pitch (corresponding roughly with what logic considers the 22 nm node). At 80 nm pitch, he said, pitch splitting is fundamentally required for 193 nm immersion lithography. Although chipmakers still have their hearts set on EUV, "it's really a question of when EUV will become a manufacturing reality," he said. In the meantime, the industry is evaluating such double patterning schemes as litho-etch-litho-etch (LELE), self-aligned double patterning (SADP) and litho-freeze-litho-etch (LFLE). Although LELE is the incumbent technique, some sort of LFLE scheme is preferred to cut processing steps and keep the wafer within the litho cell. Spacer technology is very attractive, Wallow said, but has very high process complexity associated with it.

Regardless of which lithography technology is used, patterning processes are going to be more complex, Wallow said. LFLE, though preferred, will require greater track flexibility and configurability over LELE, and any method is likely to require more complex process flows to compensate for a 1:3 duty cycle.

Charles Pieczulewski explains that, even incorporating several different double patterning schemes, Sokudo's track system can maintain a >200 wph throughput.

And regardless of which technique is used and the complexity involved, Sokudo is convinced that it can maintain a throughput of >200 wph with its coat/develop track equipment. Sokudo's Charles Pieczulewski detailed a few different single-etch schemes being worked on by resist suppliers such as JSR, TOK and Fujifilm -- some involving more complexity than others, and all with varying degrees of success. TOK has the easiest, most straightforward approach with a self-freezing material, but results have been weak, according to Pieczulewski. JSR puts a litho-freeze chemical between the first and second resist. It involves a more complex process on the track, but it has seen good results. Fujifilm has a dual-develop approach that could be useful for trench or contact hole applications.

Each of these double patterning schemes will involve two cycles through the litho cell, and it's very important that the wafer not leave the litho cell to go back for its second pass through, Pieczulewski noted. The question then becomes whether a semiconductor fab host reloads the FOUP for the second cycle, or the track internally manages an auto-start of the second cycle, communicated to the scanner via a linked network. "What we're finding out is that both of these are going to be necessary," Pieczulewski said, noting the need for a clear protocol with the fab host and track/scanner on the second cycle.

Individually, each double patterning scheme sounds manageable in a track system, according to Pieczulewski. "But what people really want is something that incorporates all the schemes within the same track system," he said, because chipmakers will likely use different patterning schemes for different types of patterns. Nonetheless, Sokudo maintains that it is possible to support multiple double patterning processes while keeping throughput at 200 wph.

Lithography toolmakers ASML and Nikon are also assuring customers that they are coming up with new ways to keep double patterning throughputs relatively high. Both suppliers detailed advances introduced earlier in the year with their tools aimed at double patterning applications -- including improved overlay and alignment, as well as source-mask optimization (SMO).

Air fluctuation is a dominant source of overlay error, said Steve Renwick, principal engineer at Nikon Precision. So Nikon has developed what it calls Bird's Eye Control, which is not affected by air fluctuation or high scan speeds (700 mm/sec). Existing interferometry metrology capabilities cannot handle 32 nm requirements, Renwick said, so Nikon has coupled interferometers with encoders to enhance positioning accuracy.

An example of the work being done by ASML to improve throughput is its FlexRay programmable illumination system introduced this week at SEMICON West. Working in conjunction with the Brion subsidiary's Tachyon SMO, FlexRay provides more flexibility in pupil shape with a freeform illumination technology based on a programmable micromirror array that creates different sources in minutes rather than weeks.