by Franklin Kalk, Toppan Photomasks

September 21, 2011 - The March 11, 2011 Tohuku earthquake and tsunami caused many deaths and much property destruction, bringing parts of Japan to a virtual standstill. One casualty was Photomask Japan 2011 (PMJ), which was originally scheduled to take place in Yokohama in April but was cancelled after the horrific events. PMJ has a sister conference, SPIE Photomask Technology (known in the vernacular as BACUS), annually staged in Monterey, CA in September, and this year BACUS has been extended by one-half day to allow the "10 best" PMJ papers to be presented in a special kick-off session. (The earthquake and tsunami occurred more than a month before the scheduled PMJ conference, a time when many authors had not yet finished their manuscripts, much less their slide presentations; nonetheless, 10 other papers are queued for the September 20^th poster session. This is a testament to the indomitable Japanese spirit.)

Most of the papers focused on mask reliability and defect management, but a few examined the effect of masks on wafer print quality. Kei Mesuda of Dai Nippon Printing began the special session by discussing the impact of mask blank optical properties on printing quality in argon fluoride (ArF) wafer lithography. Surveying a range of available real-world set of optical indices for the mask blank absorber, Mesuda-san and his colleagues used modeling to identify two regions that might improve on current state-of-the-art materials. Interestingly enough, one area had high k (~2.5) and low n (~0.5) -- clearly a binary absorber candidate -- while the other area had low k (~0.5) and high n (~2.5), which is very similar to conventional molybdenum silicide (MoSi)-based attenuators. They then built and tested a material closely resembling MoSi. In 12 percent transmission (%T) form, the material offers better exposure latitude than 6%T MoSi on simple patterns. Mesuda-san promised future reports on this material.

James Oberschmidt of IBM India presented results of his group's evaluation of process variations and their effects on wafer printing accuracy. Using a baseline optimized mask design, they systematically varied several parameters, including mask critical dimension (CD), illuminator partial coherence, resist n and k, and others. They found that mask CD has the strongest effect on print accuracy. I guess there's no rest for the weary.

Itaru Kamohara of Nihon Synopsys explored the use of ellipsometry to characterize topography in double patterning. Using the example of the litho-freeze-litho-etch (LFLE) process, he showed how the detailed topographic information available from ellipsometry can be used to optimize and control the contact hole lithography process.

Mask pattern accuracy drew a couple of ably-presented papers. Erez Graitzer showed how Carl Zeiss' RegC mask registration correction system can be used to improve the image placement accuracy of masks. The technology deforms the mask placement semi-locally by creating small, non-imaged damage sites inside the fused silica substrate with a femtosecond laser. The damage site layout can impose asymmetric deformation on the mask. A nice example is one of a mask with two areas of asymmetric placement errors that are orthogonal to each other (one error in X, one in Y). The tool can render the placement error symmetrical over the entire imaged area; this "scale" error is then readily corrected by the wafer exposure tool.

Aki Fujimura of D2S discussed the use of model-based data prep to improve the accuracy of variable-shape beam e-beam mask writers while simultaneously reducing shot counts (and therefore cost and cycle time). Standard e-beam proximity effect correct (PEC) algorithms can maintain required pattern accuracy over scale lengths greater than, say, 1μm. But for small features and pitches, they don't work. Further, the e-beam scatter signature of EUV masks is broader than that of masks used at longer wavelengths. It appears that we may soon be able to access correction methods that use both dose modulation (now prevalent in PEC algorithms) and feature shape modulation, which is prevalent in optical proximity correction (OPC) methods.

The balance of the papers focused on mask reliability. Anna Tchoukoleva of GlobalFoundries presented a very interesting paper about the effect of MoSi mask absorber surface preparation on radiation durability in the wafer fab. For many years, the mask industry has worked to eliminate sulfates on the mask surface, mainly by replacing the incumbent mask-cleaning method that uses a mixture of hydrogen peroxide and sulfuric acid. Tchoukoleva asserted that removing all sulfates can actually accelerate exposure-induced mask degradation by promoting the formation of molybdate species (especially ammonium molybdate) that coalesce into particulates on the mask surface. She suggested that a new, yet-to-be-defined surface state specification is required to ensure a mask with adequate radiation durability.

Also on the subject of haze, Toppan's Brid Connolly presented data from an OMOG (opaque MoSi on glass -- a MoSi-based binary mask absorber) radiation durability study performed jointly with CEA-Leti. The material performed flawlessly up to a 193nm radiation dose of 90 kJ/cm² in a room air environment and showed a slight drop in CD at 100 kJ/cm², which corresponds to about 50,000 wafer exposures in the fab. The wafer print work was verified on an AIMS tool. The group has yet to investigate the cause of the change, so we hope to see more from them in the future.

Hynix's Tae Joong Ha compared the cleaning durability of MoSi-based attenuated mask absorbers with different resist strip processes. It is known that mask cleaning processes, all of which use wet chemistry, can degrade the phase shift and transmission of these absorbers. Ha found that masks built with dry resist strip (so-called "ashing") develop a surface oxide that is resistant to cleaning damage, exhibiting much better chemical durability than masks that are processed with wet resist strip techniques.

Luca Sartelli of Dai Nippon Printing Europe described the 193nm radiation-induced degradation of a Cr-based binary mask absorber. In recent years, this degradation has become known as the "sun effect" because it is evident as a strong radial CD non-uniformity due to Cr and O² migration to feature edges. Sartelli's observed migration appeared to be subtly different than that of past studies, so he proclaimed it a new degradation mechanism. He also found that ozonated water cleaning methods (i.e., ones without sulfates) can actually accelerate the effect.

And finally, Masato Naka of Toshiba provided a perspective on defect management of advanced binary masks. These Ta-based absorber films have better chemical durability during cleaning than do MoSi-based phase shifting materials, but it is important to characterize their defect printing properties. To that end, Naka-san built programmed defect masks and compared simulation, wafer printing, and AIMS results to see how they compared. It turns out that wafer printing and AIMS match well for various feature and defect types, while simulation is a decent but imperfect predictor of printability. From there, Naka-san determined inspection tool sensitivity across a spectrum of defect types and sizes to identify tool sensitivity settings for a given defect size and to verify that his inspection hardware could capture all printing defects.