by M. David Levenson, Editor-in-Chief, Microlithography World
Intel chose this SPIE conference to present five papers on an apparently abandoned program on pixelated masks that had pre-occupied litho engineers for several years. Because Intel’s mask shop could not fabricate such masks with chrome, and two phase shifts, design starting points had to substitute 0° and 180° phase grating for opaque material, resulting in chromeless masks with 1012 pixels, according to Vivek Singh of Intel. Software then attempted to minimize the difference between the predicted image projected and the target, but the result depended on details of the simulation software.
The “thin mask” approximation was not at all adequate. Early attempts to account for electromagnetic effects resulted in mask clips that did not approach the target when checked by “gold-standard” simulator — which ran 10,000x slower than the model used for optimization. Eventually a new, more accurate, model was implemented and designs optimized for several layers. However, the mask patterns were so different from the desired images that they seemed “encrypted.”
Jeff Farnsworth described the fabrication, inspection, and repair of pixelated masks for the Cedar Mill metal-1 layer. The complex arrangements of 100nm pits 171.4nm deep implied that the inspection had to be performed on the aerial image, and not the mask structure (fortunately, Intel had an appropriate tool). Missing mesas were repaired by etching to twice the nominal pit depth. The repairs validated by silicon print test and, ultimately, with working chips. Today, only Intel could have conducted such an extensive R&D program — learning much, but abandoning the result when something better came along.
The prospects for high-index immersion technology seem to be dimming as well. Lutz Parthier from Schott Lithotec described measurements on the latest 80mm boule of LuAG, the key lens material. Bulk absorption was <0.05 cm-1 when corrected for surface scattering, and small pieces showed attenuation down to 0.03cm-1. Spectroscopic fits implied that the loss was due to impurities, but they were still 10x higher than could be tolerated in a projection lens. Exposure measurements on early samples at Lincoln Labs showed a mysterious quick decline in LuAG transmission, which then stabilized. Harry Sewell reported that the NA=1.55 immersion fluids were already more transparent than water, and Roger French of Dupont noted that a high-NA option was needed to buy time for EUV. Still, modest increases in index would not allow single exposure lithography at future nodes. Higher-index fluids seem very problematic. Martin van den Brink of ASML doubted the business models of high-index material suppliers — while fluids could be consumed in volume for years, potentially providing return on initial investment, the lens material would not be. Only a few very expensive components would be fabricated at the beginning of the high-index immersion tool generation, which would not last long, he noted. The investment to qualify the material would have to made before any payoff was certain; that was not a viable business model, as proved by the 157nm CaF2 debacle.
Development of a medium-throughput electron beam direct-write exposure tool would facilitate many things, not least low-volume manufacturing, but progress remains sluggish. Erwin Slot of Mapper described static exposures where a 110 beam demonstrator simulated exposures of a 13,000 beam scanning system. Mapper uses optical interconnects to control all its blankers independently without cross-talk. They were able to fabricate 40nm half pitch lines and spaces with 3.2nm uniformity, in vertical, horizontal and diagonal orientations. To be viable, even for contacts, throughput has to be above 10 wafers/hour, one expert observed; he wanted 40 WPH, though.
Medium-throughput multiple e-beam direct write might be just the technology needed to enable imprint lithography, according to Franklin Kalk of Toppan Photomasks. He reported that 1X imprint template fabrication with current Gaussian beam tools took upwards of 120 hours. Lowering the cost of replacement templates was essential for imprint, but not likely with current mask-writing technology.
Directed self-assembly promises to divide imprinted features, enhancing resolution further. The trick is to get useful shapes out of a process driven by chemical and surface interactions. Typically, some material must be coated on the surfaces to orient the structures produced when the co-polymer assembly material self-segregates. Joy Cheng of IBM described how an undercoat material could be added to the mix so that a single coat step could result in polymer assemblies that divided 28.8nm trenches in half. Both imprint and self-assembly show promise for producing patterned discs needed for the next generation of hard magnetic disk storage. — M.D.L.
Click here for the rest of the analyses in this package of SPIE writeups: a list of what’s still needed to enable 32nm generation chips printed with double patterning technology; problems, yet promise, in development of EUV; and clever new technologies added to the equation, and what’s sparked an “arms race” among OPC and EDA firms.