Luminary Panel coi tiến trình áp dụng màng mỏng EUV là quan trọng đối với EUV

A significant focus of the 2024 SPIE Photomask and EUV conference was on EUV lithography and high-numerical-aperture (high-NA) EUV lithography, offering the potential to drive resolution to new heights. These EUV solutions bring new challenges such as pellicles, mask inspection, and smaller and smaller minimum mask dimensions. Progress has been impressive, according to lithography luminary Dr. Harry Levinson in his video executive summary of the conference. The mood was positive at the eBeam Initiative’s 15th annual reception and meeting held on October 1. In the eBeam Initiative’s annual Luminaries survey, 100% of those surveyed said that photomask revenues would increase or stay the same this year. A panel of semiconductor photomask and lithography experts debated the survey results on the topics of multi-beam mask writers and curvilinear masks, which I covered in my October blog. Below is a recap of a portion of the panel discussion on EUV topics. Aki Fujimura, CEO of D2S, Inc., which is the managing company sponsor of the eBeam Initiative, moderated the 2024 panel discussion. He was joined by Naoya Hayashi, DNP Honorary Fellow; Harry Levinson, Principal Lithographer at HJL Lithography; and Glen Scheid, Operations Manager at the Micron Mask Technology Center (pictured below). The complete panel can be watched in a two-part video: Part 1, Part 2.

Photo (left to right): Jan Willis, Glen Scheid, Aki Fujimura, Harry Levinson, Naoya Hayashi

Fujimura: Let’s talk about pellicles, a topic we’ve been exploring in the survey the past couple of years. Pellicles are protective membranes that shield photomasks from contaminants and defects during the lithographic process. Based on Glen’s comments in his talk last year about EUV mask lifetime, we asked for opinions on the impact of pellicles on mask lifetime. 45% say EUV masks without pellicles have a lifetime that is less than 40% of 193i masks with pellicles (figure 1). What’s your expectation now, Glen?

Scheid: That’s a great question for a DRAM maker who’s still ramping up production of EUV. We have tremendous experience with 193i lithography. We’re pushing mask lifetime to the limits with hundreds of thousands of wafers from a 193i mask. Looking at EUV masks without pellicles, what impacts mask lifetime? Material effects can be studied in terms of acceleration testing and at what point the material fails. Without a pellicle, you also have contamination, which may not be possible to remove. Excessive handling because of frequent inspections can put the mask under duress. There’s quite a variety in the answers to the survey question so it makes me think the answers are situational depending on where they are on EUV adoption.

Fig. 1: 2024 eBeam Initiative Luminaries survey says EUV masks without pellicles have a shorter lifetime than 193i masks (with pellicles).

Hayashi-san: We’re not providing defect-free EUV masks yet, so we don’t have data on mask lifetime with and without pellicles. But I think there’s more cleaning cycles for EUV masks.  These masks are very sensitive to cleaning. Resistivity and critical dimension (CD) change in every cleaning cycle. So, I think that EUV masks without pellicles have a quite short lifetime compared to 193i masks with pellicles.

Levinson: As Glen provided in his great presentation last year, if you don’t have a pellicle, you have a very short lifetime for EUV masks. If you have a pellicle, how much will that help?  We need more data. But from day one, the industry has been concerned about how many clean cycles an EUV mask without a pellicle can go through. While they’ve been under development for years, we’re now seeing a broader adoption of EUV pellicles as the technology matures. The expansion of pellicle use is vital as we continue pushing for smaller and more precise features, and it’s clear that this technology is poised for broader implementation in the next year or two.

Fujimura: The next question is about actinic inspection of EUV production masks. The data shows a prediction of a lower amount today than in last year’s survey and uncertainty for three years from now (figure 2). You need to remember this is an opinion survey so we’re just asking luminaries what they think is happening. It could very well be that the only thing that’s changed is what people think. Glen, what do you think is going on?

Fig. 2: 2024 eBeam Initiative Luminaries survey predictions aren’t clear for actinic inspection usage in three years.

Scheid: It’s a curious result for three years from now. Let’s frame the question this way: why is actinic inspection needed in the first place? It’s been used for through-pellicle inspection, to look for embedded multi-layer defects, and/or to detect printable defects on the patterns you’re using. The answer may be telling as to the progress the industry is making on multi-layer defects where they may not be as much of a problem now. Maybe not as many pellicles are being deployed yet. Actinic is certainly being deployed today for pattern resolution and we’ve heard about the progress on that at this conference. But not every single pattern of EUV requires the resolution of actinic. We still have extendibility of 193i inspection, and we’ve heard about progress on that at this conference as well.

Hayashi-san: For the merchant mask maker, there’s not an urgent need for defect-free EUV masks yet. So, we can use 193i inspection to guarantee the part of performance of the mask. But in the very near future, there is the second group of manufacturers who will be using EUV. In Japan, they will be targeting 2nm production in 2027. That’s a short amount of time so we will need actinic inspection to guarantee 100% defect-free. But it’s very expensive! It’s kind of a headache for us but we need to plan for it over the next 2-3 years.

Fujimura: The whole mask manufacturing supply chain must get ready for high-NA EUV. So, the question of the smallest size on the 4x dimension of the mask for high-NA comes up (figure 3). Precision for everything must get better. But maybe it’s like the need for actinic inspection in the beginning where high-NA masks don’t have to have SRAFs. On the other hand, no one turns down more process window if they can have it. How’s this going to play out?

Fig. 3: 2024 eBeam Initiative Luminaries survey predictions for high-NA EUV minimum mask dimensions.

Levinson: There was a very good paper earlier this year from imec at Photomask Japan. They showed a 4nm SRAF (wafer dimensions) would print using high-NA EUV lithography but a 3nm SRAF (wafer dimensions) wouldn’t. This is a big deal as the industry is struggling with depth of focus for high-NA EUV and needs SRAFs. At the same time, it’s very difficult to do. Everyone is going to have to roll up their sleeves and work on the problem. I suspect we’ll hear more on this at next year’s conference.

Hayashi-san: The survey says <20nm. Getting to <15nm would be very challenging. In the session on multi-beam mask writing at this conference, a supplier showed test patterns being written between 10nm and 15nm with good resolution. It’s a test pattern, but still it’s a good improvement.

Scheid: It’s no surprise that as lithography progresses, the minimum feature size for the mask maker is going to get smaller and smaller. Whether it’s the early days of high-NA, or the later days or even early days of hyper-NA, I think we’ll see the requirement for <20nm feature size on mask. There’s going to need to be quite a bit of material and even machine development to get there. There’s a lot of thought needed to break through the barriers.

Fujimura: Thanks to our panel for their insights. Looking forward already to next year’s survey results!