TSMC preps 193nm immersion ‘risk’ production; others still see no technical showstoppers
By J. Robert Lineback
The first wave of production-worthy 193nm immersion scanners now appears to be a shoo-in for at least part of the 65nm process generation and probably the entire 45nm node at the end of this decade, but industry managers and researchers speaking at the annual SPIE International Symposium on Microlithography (Feb. 28-March 4 in San Jose, CA) attempted to rein in some of the unbridled enthusiasm with a dose of reality. With price tags for early 193nm immersion tools expected to be $25-$30 million, slightly less expensive and nearly equivalent dry 193nm scanners certainly won’t fade away anytime soon.
Beyond the anticipated cost of ArF immersion platforms, accuracy and stability of scanners using water to boost depth-of-focus (DOF) and resolution remain areas of intense engineering work, observed Chin-Hsiang (John) Lin, development manager for advanced lithography processes at foundry giant Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC). By the middle of 2005, TSMC — one of the most aggressive advocates of applying immersion lithography sooner than later — plans to begin offering “risk production” runs for 65nm ICs using a Twinscan XT:1250i 193nm (0.85NA) system delivered by ASML Holding NV in 3Q04.
Only several months ago, TSMC and rival IBM Corp. reported the first working 90nm ICs made with one layer of immersion exposures in separate “split-lot” experiments using ASML’s initial beta-prototype AT:1150i (0.75NA) scanner. In a technical paper presentation at SPIE on March 1, TSMC researchers reported a 50% improvement in DOF using immersion vs. equivalent dry 193nm scanners at both 90nm and new 65nm processes. The initial 90nm tests using an “immersion chip” (an SRAM device) helped TSMC pull in yield demonstrations by nine months, and convinced the chip foundry to apply a protective topcoat on standard photoresists to prevent leaching and contamination of the water as well as protection of lens elements, said Jeng-Horng (Jack) Chen, sector manager in TSMC’s advanced tooling operation based in Hsinchu, Taiwan.
“This tool [ASML’s 1250i] is now ready for risk production with DOF generally improving by 50%,” Chen told a packed room at SPIE during his presentation. “We investigated topcoats, overlay, and concluded that bubbles [in water] are not showstoppers, but we still see the need for engineering work.” The world’s largest chip foundry intends to begin offering volume 65nm production using immersion lithography in 2006.
TSMC’s findings match up with other results from tests and experiments presented at SPIE. “There are no technical showstoppers, but there are obviously areas of concern,” said Robert Allen, manager of the lithography materials group at IBM’s Almaden Research Center in San Jose, CA. “Many things that material researchers thought would be absolutely difficult and monumental challenges for immersion lithography have turned out to be fairly easy, with a few exceptions,” he said during a breakfast panel discussion hosted by Rohm and Haas Electronic Materials LLC. He added that topcoat layers preventing interaction between resists and water were one of the relative quick fixes to key concerns, while understanding immersion defects remains an area of engineering focus as initial processes are being prepared for 65nm and 45nm nodes.
The panel, which included managers from IBM, TSMC, Nikon, and Photronics, generally concluded that the first implementation of 193nm argon-fluoride (ArF) immersion lithography, using exposure systems with numerical-aperture lenses up to 1.0NA, would simply involve adapting existing tools and materials from today’s dry 193nm processes. To deal with 45nm node processes, scanner makers now are preparing second-generation 193nm immersion platforms with lens optics in the 1.07-1.30NA range. These systems are expected to begin appearing between 2Q06 and the start of 2007. In addition, Nikon Corp. and Canon Inc. disclosed at SPIE that they were preparing new multistage platforms for 45nm and beyond processes.
“I don’t think there is any question about the ability to build optics at 1.2-1.3NA for immersion,” said Gene Fuller, principal engineer at US-based Nikon Precision Inc., Belmont, CA. He added that the lenses will have to be based on catadioptric optics, as opposed to refractive optical systems in today’s 193nm tools. However, once the industry attempts to move beyond 1.3NA optics to numerical apertures of 1.5 or 1.6 for a possible 193nm immersion run at 32nm or even 22nm technology nodes, the need for new “hyper” refractive-index fluids beyond pure water will raise the “showstopper” question again, predicted Fuller and others on the panel.
One other note of caution was cited in the immersion debate. “Fab managers have just started spending more money to bring in new tools and will want to see a return on those 193nm investments,” noted Christopher Progler, VP and CTO at Brookfield, CT-based photomask maker Photronics Inc. “So, I suspect the transition to immersion will be slower than people expect. We’re essentially looking at moving in a new generation of tools.”
Even one of the biggest proponents of 193nm immersion is conscious of price issues. “If there are no showstoppers, there is no reason for [immersion] to be more expensive,” said TSMC’s Lin, who admitted that he would expect to see scanner and material suppliers attempt to price initial technologies “slightly more” than dry lithography. Photronics’ Progler added that talking about cost is a little premature, since “familiarity and comfort issues” with immersion will be a deciding factor in tool selection in much of the 65nm node. “One [dry 193nm] works; the other [immersion] is expected to work,” he added. — J.R.L.