SAN FRANCISCO Many in the semiconductor industry underestimate the technical and economic challenges posed by new forms of lithography, which are enormous, said Phil Ware, a senior fellow for lithography strategy at Canon Inc., speaking Sunday (July 21) prior to the opening of the Semicon West trade show.
Ware made his comments after participating in a two-day meeting of the U.S. lithography development working group, which is involved in the International Technology Roadmap for Semiconductors. Backers of the various candidates for next-generation lithography outlined their progress in 157-nanometer optical lithography, projection e-beam lithography (EPL), and extreme ultraviolet (EUV) lithography at the meeting.
For starters, technical difficulties will make advanced lithography more expensive: a 157-nm optical scanner may cost in the range of $20 million, and estimates for EUV scanners vary from $40 million to $70 million per unit. Compared with the 157-nm wavelength tools, EUV systems use radiation of 13.5-nm wavelength, and thus proponents argue that the use of EUV equipment can be extended over several process generations.
While 157-nm lithography takes an optical approach, Ware said it nonetheless requires as a basic lens material a higher-quality calcium fluoride than is currently available. Calcium fluoride is used in limited quantities for the lenses used in the 193-nm scanners coming to market now, but the 157-nm lenses require a higher quality crystal "which the industry just doesn't have available today," Ware said.
Also, the lack of a pellicle the protective coating for masks which plays a key role in defect prevention and repair may boost 157-nm mask costs.
Mask challenges are more severe for the EUV systems. The reflective lenses, or mirrors, used in the EUV optical path require multiple coatings of molybdenum and silicon 40 layers each. The mask pattern is also created on a similarly precise reflective substrate.
Ware said current thinking is that the EUV masks must have a surface flatness of 50-nm, from edge to edge, which is beyond anything the industry has been able to create thus far.
More importantly, defects on the mask substrate must be limited to 50-nm or less. The multilayer coatings on the projection EUV mask substrates are very difficult to repair. Also, if water vapor comes in contact with the masks, the mirrors are corrupted by oxidation and most likely will no longer be usable.
"Imagine how costly it would be if you invest in a EUV mask set, and do everything it takes to get that perfected, and then would have to essentially throw it away because of oxidation," Ware said.
Nikon Corp. and IBM Corp. have cooperated in the development of an e-beam projection lithography system. Ware said two challenges facing EPL emerged at last week's lithography development working group meeting: the size of the systems, and the masks.
The current EPL prototype systems are quite large, and in order to remain protected from errant magnetic fields that disrupt the electron beam, must be housed in separate rooms. Also, the industry faces unique challenges with the scattering masks used in EPL. A delicate membrane structure is used to scatter the electron beam, in part to prevent overheating of the mask pattern.
The EPL systems also face throughput issues, with current prototypes processing fewer than 20 wafers per hour. Also, the EPL systems work with relatively small field sizes that must be stitched together, which may make it difficult to create irregular circuit patterns or long interconnects, he said.
"I am not trying to criticize the efforts of the industry, but I think people just don't quite understand that the technical challenges facing us are exponentially more difficult than what we have faced thus far," Ware said.