SANTA CLARA, Calif. Papers at the Future of Lithographic Systems track at SPIE here Thursday (Feb. 27) revealed a far from settled roadmap for production lithography, with serious alternative paths branching off from the main road as early as the 90 nm process node.
Despite the fact that the industry is moving into 90 nm pilot production with existing 193 nm steppers for critical layers, vendors of alternative technologies are pushing their systems toward production.
In the near term most fabs will be using 193 nm steppers with extensive reticle enhancements to produce features significantly smaller than the lithographic wavelength. But research on alternatives for the near-term is aimed at producing critical layers using other light sources, with wavelengths short enough to eliminate the need for optical proximity correction or phase shifting.
One example cited was the Collimated Plasma Lithography system from JMAR Research (San Diego, Calif.) Edmund Turcu, chief scientist at the organization, said the first beta production X-ray lithography system from his firm had been installed in a facility in Vermont and was running with IBM masks. The JMAR system uses intense laser beams to convert copper film, carried on tape, into a plasma. The excited plasma radiates in the X-ray band. This radiation is picked up by a collimator formed from bundles of hollow tubes, and directed at the mask.
Turcu reported that the system is producing 100 nm features reliably using IBM Silicon Carbide masks with Tantalum absorbers, and with a very conservative resist chemistry only capable of 60 nm resolution. The goal of the initial system is to produce 100 nm critical dimensions in a production environment, without need for reticle enhancement processing.
Two other papers cited progress on the low-energy electron projection lithography program, being driven by LEEPL Corp. (Atsugi, Japan.) LEEPL researcher Akira Yoshida reported that LEEPL has delivered two beta systems for work at 65 nm, and expects to install its first production stepper this spring. The system is expected to hit 52 8-inch wafers or 41 12-inch wafers per hour and to be used on SoC layers and some critical layers of memory wafers.
Hiroyuki Nakano of Sony Corp. (Kanagawa, Japan) reported work on a three-layer resist formula for use with the LEEPL system. Because of the low energy of the electron beam in these systems, the beam penetrates very little around 100 nm into the resist. So it is necessary to have a multi-stage resist in which the top stage is exposed by the e-beam system and developed, and then serves as a mask for etching the rest of the resist. By optimizing the resist layers, Nakano said, the team at Sony demonstrated the capability to achieve sufficient resolution for the 45 nm process node.
But if one of the goals of next-generation lithography was to eliminate OPC and phase shifting, Mentor Graphics manager Frank Schellenberg offered a reality check. Schellenberg presented a summary of a variety of papers on next-generation efforts in which it was clear, he argued, that some sort of polygon manipulation would be necessary in all but one of the technologies. Schellenberg looked at 13.4 nm extreme ultraviolet work, 0.2 nm e-beam work and two non-optical nanoimprint technologies.
In the EUV work, he said, there were flare effects from the reflective optics that caused distortions very similar to today's sub-wavelength problems. Further, shot noise would likely force redundancy in minimum-dimension features. So something very like OPC would be necessary to prepare mask data.
Similarly in e-beam projection techniques, the stencil masks had to be supported by a grid of struts, requiring special treatment of polygons that crossed a strut location. Further, rules often required a feature to be split into several pieces to avoid isolated and unsupported areas in the mask. Once again, then, polygon manipulation would be required to produce the masks.
Schellenberg said even with maskless technologies such as the Micronic light-valve system it was necessary to stitch together exposure regions, and that other pixel corrections were necessary. He added that some of these are performed on the fly, but that others would require preprocessing.
Finally, examined two nano-imprint technologies. One, in which the template is simply stamped into a soft polymer, requires channels through which displaced polymer can flow out of the way. But step-and-flash technology, in which a template is filled with a monomer by capillary action and then flooded with UV, may be the sole candidate not obviously requiring complex polygon manipulation, Schellenberg speculated.