For EUV, high-NA hopefully arrives on time

High-NA EUV, or high-numerical aperture extreme ultraviolet, was a popular topic at the SPIE Advanced Lithography and Patterning conference held in San Jose. It showed up in keynotes and numerous presentations.

It’s easy to see why: high-NA EUV is the future of the semiconductor industry and could play a role in meeting challenges faced worldwide.

EUV lithography was deployed in high-volume manufacturing at the 5-nm process node in 2019. It was at 0.33 numerical aperture. In contrast, the corresponding number for high-NA EUV is 0.55. A benefit of the higher number is the ability to pattern smaller features. Predictions are that high-NA EUV will be needed for the 2-nm node.

Mark Phillips, director of lithography hardware and solutions at Intel, noted in his Monday afternoon presentation that help would be coming soon. He pointed to a joint release from Intel and ASML about support for production using high-NA EUV in 2025.

“But the first tool won’t be available until 2023,” he noted.

For comparison, it took almost a decade from the appearance of the first 0.33 NA EUV tool to high-volume manufacturing. Phillips noted some key differences that should make this transition faster. One is that there is less of a technology shift. Before EUV, lithography tools used deep UV. The switch to EUV meant the development of new light sources, masks, tools, resists and more. Delays in the technology, particularly for the light sources, meant that the whole rollout slipped.

Phillips doesn’t see that happening with high-NA EUV. Progress appears good on all fronts, with key components falling into place, he reported. What’s more, many of the tools and techniques developed will work for both 0.33 and 0.55 NA EUV, although the performance may not be optimum for both. Still, he noted that the tools must be available starting in 2023 or the timeline is in jeopardy.

“That’s really critical,” he said, citing a lack of timely access to tools as the key risk factor.

The alternative if high-NA EUV is not ready is to do multiple passes through a lithographic tool. But eliminating repeated passes brought significant benefits in the past.

“By transitioning back to single exposures, it results in cost improvement, in cycle time improvement, in yield improvement,” said Luc Van den hove, president and CEO of the international research and development organization imec, in his Monday morning keynote presentation. “It also provides a CO₂ footprint improvement.”

He noted that Zeiss is doing the optics for high-NA EUV, which have to get bigger in step with numerical aperture, all things being equal. In the case of 0.55 NA EUV, this means meter-sized mirrors that are flat to 20 picometers, a ratio of 11 orders of magnitude of width to allowable deviation in height. Zeiss had to develop completely new measurement techniques to characterize those mirrors, according to Van den hove.

For its part, imec is joining with ASML to set up a high-NA EUV lab so that interested companies can work out kinks in the manufacturing process. As much development will be done in parallel and via simulation as possible, with Van den hove saying such an approach is necessary due to a compressed time schedule. The goal is to move to high-volume production in 2026.

“That gives us only three years to debug this technology.” Van den hove noted, a far shorter time than almost 10 years take for the original EUV deployment.

The industry has also started talks about moving to 0.7 NA EUV. That would provide the ability to pattern still finer features. Taking that step will require some more extensive changes. One such might involve the photomasks, which have to get bigger with NA. In moving to 0.7, the masks, which have always been rectangular, might instead be round. That would entail a major upheaval in the photomask industry, but it could bring benefits to 0.55 NA EUV by increasing throughput anywhere from 15 to 40 percent, according to Phillips.

He noted that chip manufacturers buy multiple tools, sizing this purchase by how many chips they need to produce. An increase in productivity would mean fewer tools purchased, at a saving of hundreds of millions for each tool not bought.

“Fifteen to 40 percent on a tool that’s expensive is enormous,” Phillips said of the potential savings.

Hank Hogan is a science writer based in Reno, Nevada.