大尺寸掩膜版从愿景走向行业路线图：eBeam Initiative 专家访谈

The eBeam Initiative conducted its 15th annual Luminaries survey in July and reported the results on September 9, 2026, at its annual reception during the SPIE Photomask Technology + EUV Lithography Conference. 85 luminaries representing more than 45 companies participated.

At the close of the annual eBeam Initiative reception, an expert panel discussed a number of survey results, including large-format masks, equipment investment, High-NA EUV, and curvilinear masks. Moderator Aki Fujimura, CEO of D2S and co-founder of the eBeam Initiative, was joined by Harry Levinson of HJL Lithography, Naoya Hayashi, DNP Honorary Fellow, and Glen Scheid of Micron Technology (pictured in Figure 1). In this blog, I’ll report key insights shared by the panel.

Fig. 1: eBeam Initiative Luminaries Survey Results Discussed by Expert Panel (left to right, Aki Fujimura, Naoya Hayashi, Harry Levinson, Glen Scheid)

eBeam Initiative (Fujimura): 66% of luminaries predicted that the industry will adopt a large-format mask – approximately 6 x 12 inches – within the next decade. The survey was completed before TSMC and ASML announced an industry initiative targeting a 12-inch mask pilot line by 2031. Why is this transition happening now?

Levinson: The wafer fab argument is productivity. High-NA EUV systems are extremely expensive and occupy valuable cleanroom space. A large-format mask can expose a larger field without the process-control complexity of A/B stitching, with the potential for roughly 40% better productivity. That is a significant advantage when each additional system also adds facilities and power requirements.

Scheid: The alignment forming around the 12-inch format is exciting. Mask-size transitions are extremely rare, and this one has a clear productivity benefit that the industry needs.

eBeam Initiative (Fujimura): Harry, you were involved in the industry’s earlier attempt to standardize a nine-inch mask. Why is this different?

Levinson: The nine-inch proposal came from extrapolating historical growth in DRAM die size. The intended customers did not plan to build dies that large, so the format solved no real problem and was rejected. This time, the demand is coming directly from chipmakers. Intel initiated the discussion, and now TSMC, ASML, and other companies are supporting it. The application and the value are clear.

eBeam Initiative (Fujimura): What must happen across the supply chain to meet the 2031 pilot-line target?

Hayashi-san: The whole ecosystem must be ready at the same time. Some critical parts of the mask-making chain have only one supplier. We need to watch the ecosystem carefully and support the areas that could prevent the complete process from being available when it is required.

Scheid: This is a collective schedule, not a set of individual schedules. The combined probability of success falls as the number of required elements grows. One late supplier – especially in a single-source area – can delay the transition. Redundancy can reduce risk, but this is a small industry, and the investment may be too large to support two suppliers in every category.

Levinson: There are financial and technical risks. Some tools may require a larger stage and different handling, while processes such as uniform etching deserve closer attention. Funding risks should be addressed early, especially for smaller suppliers. Companies planning fabs around this timing will worry about whether every element will arrive on schedule.

eBeam Initiative (Fujimura): What new manufacturing challenges come with a rectangular mask that has twice the area?

Hayashi-san: Processes based on spinning become more difficult. Resist coating, development, cleaning, and drying all must remain uniform across a larger rectangular surface. The longer edges and corners can create water-drop issues at the corners, so these processes will need deep evaluation.

Levinson: It took roughly 15 years to make EUV multilayer films suitable for high-volume manufacturing. Much of that learning will transfer, but scaling the multilayer blank will not be as simple as installing a few larger tools.

Scheid: More mask area means more opportunity for particles, while contamination and cleaning already are leading EUV mask-lifetime concerns according to another survey result. A large-format pellicle could help in the fab, but the mask shop still must clean a larger surface with uniform chemistry and without adding defects.

eBeam Initiative (Fujimura): Has High-NA EUV moved from a question of if to a question of when?

Levinson: Yes. Only a year ago, people still asked whether High-NA was real. Now it is in production at Intel along with publicly announced roadmaps from multiple companies, which is a sea change. Samsung announced they will be in production with High-NA by 2028 and TSMC said they intend to use it in high-volume manufacturing for advanced nodes starting in 2030. Beyond High-NA, demand is less certain. Device and interconnect engineers first must show that smaller features provide the required performance.

Fig. 2: 2026 eBeam Initiative survey result predicting equipment purchase in the next three years.

eBeam Initiative (Fujimura): 68% of luminaries expect mask-etching-system purchases to increase over the next three years (Figure 2). What is driving that result?

Scheid: Purchases of multi-beam writers and inspection tools are predicted to increase over the next three years and reflect both capacity growth and capability needs. Etching is different. New High-NA mask materials may not be compatible with existing systems or with .33 NA materials running side by side. The survey prediction that etching tool purchases will increase in the next three years is less about adding etch capacity and more about adding new capability.

Levinson: At the same time, uniformity requirements are tightening. Improving uniformity while introducing new materials makes etching a very active development area.

eBeam Initiative (Fujimura): What about curvilinear wafer targets?

Scheid: Curvilinear masks are already here, enabled by multi-beam writers and supported by formats such as MULTIGON. Full layers with curvilinear wafer targets automatically generated are harder to see within a three-year window.

Levinson: The benefits demonstrated for curvilinear wafer features are compelling, although backside interconnect and other changes make the timing uncertain. On the mask side, curvilinear is here now and will continue to grow stronger.

Hayashi-san: Writing curvilinear masks with a multi-beam writer is straightforward. Qualification remains a challenge, but the level of attention at this conference shows that the industry is working on it.

The panel’s conclusion was optimistic, but not complacent. The productivity case for large-format masks is strong, High-NA is now in production as well as in roadmaps from two more companies, and mask shops are investing in new equipment and curvilinear masks. The difficult part will be coordinating the complete ecosystem. This industry’s history of solving shared technical challenges may be its most important advantage on the road to 2031.

大尺寸掩膜版从愿景走向行业路线图：eBeam Initiative 专家访谈

Reference links: