Double Patterning Leads Race for 32 nm

Laura Peters, Lead Technical Editor -- Semiconductor International, 10/18/2007 7:51:00 AM

At its annual review meeting earlier this week, IMEC announced its intention to install a pre-production extreme ultraviolet (EUV) lithography tool from ASML (Veldhoven, Netherlands) some time in 2010. Kurt Ronse, director of the Advanced Lithography Program at IMEC, provided an update of the status of immersion lithography, double patterning and EUV at the research center. “We have made significant progress with immersion, but the current generation solution with 1.35 NA is not good enough to pattern the 32 nm half-pitch,” he said. According to the Raleigh equation, a numerical aperture (NA) of 1.6-1.7 is needed, which will require high-index lens materials and high-index immersion fluids. Possible candidates for the high-index optics include lutetium aluminum garnet (LuAG) and polycrystalline materials. “Absorption is the most challenging material issue, followed by SBR and homogeneity.” Ronse said that the LuAG material, available from Schott (Mainz, Germany), requires a 10-20× improvement in absorption.

Five candidate materials exist for second-generation fluids (n~1.65), although they are challenged to deliver the proper surface tension and viscosity for immersion lithography. A few high-absorbing fluids exist (n~1.68), but even with those, the achievable NA of around 1.55 is insufficient for 32 nm half-pitch patterning. No third-generation materials have been identified. Ronse suggested that high-index immersion would be delivered late, and would only provide a single-generation solution.

In this light, double patterning, first with water immersion (NA=1.35), is looking like the most likely candidate for 32 nm. Because two critical exposure/develop/etch steps for each layer will be so costly, resist suppliers are working on a pattern-freezing approach that would save one etch step. Other improvements in cost of ownership (CoO) could be realized through higher throughput on the scanners and tighter integration between the scanners and tracks. Ronse suggested that the overlay issue associated with double patterning is an engineering issue that will be resolved. In essence, every nanometer of alignment error leads to a nanometer change in critical dimension. The overlay budget goes from 33% of the CD to 10% of the CD.

IMEC has produced the first images on its alpha EUV tool using a tin (Sn) source (Figure). 33.8 nm lines and spaces (35 nm L/S target) and 37.4 nm L/S (40 nm target) were produced using a 100-nm-thick modified 248 nm resist from Rohm & Haas (Phoenix), 18 mJ/cm² exposure and conventional illumination. High line-edge roughness (LER, 6 nm) at the high sensitivities required with EUV indicates that a fundamental change in resist architecture may be required, away from the tried-and-true chemically amplified resists. Ronse indicated a 3× improvement in LER (to 2 nm) is needed at a 25 nm resolution. In terms of EUV masks, different material stacks are being developed for the mask blanks, but defect printability is still a concern, as is the mask protection issue. Other issues include EUV shadowing — the result of the 6° offset in incident angle, which essentially requires something along the lines of light optical proximity correction (OPC) for compensation. Ronse summarized the status of EUV by stating that it may be too late for a 32 nm introduction, but CoO and its extendibility may make it a worthwhile learning vehicle toward 22 nm and beyond.

First patterns of EUV-exposed 35 nm lines/spaces (L/S) and 40 nm L/S, patterned using 100 nm MET-2-D resist and ASML’s alpha EUV tool with Sn source (Philips Extreme UV), show unacceptable line-edge roughness. (Source: IMEC)