Laptop business will take a big hit then. You win some and you lose some.
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
Laptop business will take a big hit then. You win some and you lose some.
Then Sir comparable with 7nm AMD and better than Intel 14nm?
You are conflating novelty with complexity.
Only five (5) nations (US, UK, FR, RU, CN) can build jet engines in-house, however, zero (0) nations can build EUVs in-house and only two (2) nations (NL+JPN, soon CN) can build DUVs in-house, therefore EUV and DUV is much harder than jet engines... Nope!
Niche products naturally have fewer suppliers due to their specialized nature. However, the number of suppliers doesn't necessarily determine how complex one product is compared to another. Heck... even Hentai porn must be difficult to make than jet engines since only 1 nation (Japan) makes it... NO, it doesn't work that way. Novelty does not equal complexity.
No. There are multiple ways to skin a cat and a lot of these techniques are non unique (for example the company I work at had to develop a droplet generator for one of the instruments I work with, and generating droplets is an essential mechanism for that category of instrument, but there are several companies in this market for this instrument). Plus since their goal is servicing the China market and they have political protection they can focus completely on the engineering. If you want to go global you can always redesign the components to circumvent foreign IP protection later. Oftentimes the modifications you would need to do to get around them are trivial.
Also not likely imo. Your droplet generator frequency in this case is dictated by the desired frequency of your photon output. Higher is actually probably better if you can manage it because it means for the same level of emissions per droplet you’re getting a higher output over time. The constraint to going with higher frequency is the frequency at which you can operate your drive laser pulses, the laser that’s being used to do excitation of the tin drops. It may be that in China they think they can do better than ASML on this parameter. Or it may be that China’s engineers are working around a different constraint (like excitation efficiency) and thus chose to go down this path. Or it may just be that they’re thinking about the problems differently. Usually these are engineering first rather than IP first decisions. You figure out how to do a thing and then you find a trivial change to get around IP if you have to.
The focus is domestic right now since that’s the guaranteed market. The demand for EUV comes from fabs that do sub 10 nm process nodes. Right now there are only three markets that provide that demand, the US, South Korea, and Taiwan. Whether SMIC builds fabs abroad is not something that is likely to happen for quite a while, and if they do build plants abroad it’s pretty unlikely it would be for advanced nodes. Plus we don’t know where chip fabrication technology will be by the time they do.
Not likely. The physics of how many electrons you have to accelerate dictate your energy consumption and the amount of electrons you need to accelerate for a certain amount of output is a pretty set physical relationship. Plus you have the bunching control systems, which will also use up a lot of energy. The real main benefit of SSMB is facility size since that impacts production logistics, and construction cost, since a larger storage ring means much more infrastructure costs in general.
SSMB if it works has a few significant advantages over LPP. For one the light output is a lot more coherent, which means you need fewer optical passes to clean up and focus the beam for the scanner. This means your optical assembly can be simpler and also means a lot less light loss and thus much higher dosage at scanning point for the same light source output, since EUV reflection is inefficient and you lose 30% of your photons at each optical pass (the quantity of light you can deliver to the scanning point, not the total output at source, is ultimately what matters here). One SSMB facility can also serve multiple scanners, which is very helpful when you have to do more multipatterning. Furthermore the accelerator has much better future scaling potential and a much easier upgrade path, so long as you’re willing to concede designing your fab around the accelerator facility.
X-Ray lithography for semiconductors is for now mostly a pipe dream. We don’t even know if you can design a good fabrication process around it since it is extremely difficult to find usable mask and resist materials given how energetic the photons are. You will probably also see much worse scattershot noise from secondary energization of the materials interacting with the X-rays, which is already a problem for EUV. Given these physical hurdles it’s unclear whether going to 6.7 nm will give any benefit to industrial process efficiency, and that’s half the point, since it’s all about trying to reduce the patterning steps. If it reduces the steps by half but increases the costs by three fold (hypothetical for the sake of making the argument here) then it’s not worth pursuing. This is why the industry chose EUV. They actually looked at X-Ray and decided it wasn’t a realistic path. It’s also unclear whether there’s much shrinking left that transistors can still do, which further casts into question whether it’s worth developing a whole new set of processes around X-Ray resolutions.
Last edited:
Japan can only produce the light source of EUV at the moment，not the entire machine
Pxw is not even in production. Why do you think Huawei n+1 will be in operation?
They have some stored up and will get smic to make more kunpeng chips. Looks like the hyper threading tech that Huawei has is quite powerful.
What are you talking about? There will be kunpeng 930 getting announced later this year. The Huawei people have been talking about this on video for a while now.
Everything will be smic 7nm!
Jet engines are international combustion engines which have been around since the Industrial Revolution. Fundamentally as a system it is far less complex than an EUV lithographic machine that can etch patterns at a microscopic level.
The material science needed for the best turbofans needed for first rate military powers and extreme safety requirement for commercial usage at the level of airliners is through the roof.
But fundamentally it is a far, far less complex system. You can literally work on a jet engine with hammer. You need an extreme clean environment (cleanroom) to open and touch a lithograph machine.
Do you think Huawei only has PXW?
EUV is highly profitable business，if some country can do it，they would have done it already.
As for Hentai porn，who says Japan has monopoly here？Never seen Western porn？There is alot on internet
They had a prototype ready BUT TSMC and Samsung (due to geopolitical issue) want to work with ASML instead. THE feedback they provide help establish ASML monopoly.
Japan plays catch-up on EUV lithography
By Yoshiko Hara 09.02.2003 0Share Post
Share on Facebook
Share on Twitter
TOKYO A consortium of nine Japanese companies working on extreme-ultraviolet (EUV) lithography says it is making steady progress toward a spring 2006 target for fielding an alpha tool that would provide 10 watts of EUV output power. The effort seeks to bring Japan's effort in line with EUV progress made abroad.
The Extreme Ultraviolet Lithography System Development Association (EUVA) is supporting two parallel movements to develop a source of 13.5-nm wavelength EUV radiation. Commercial EUV scanners will require 110 W or more of EUV power for high-throughput systems. The current EUVA source generates only 0.6 W, although it has a frequency of 10 kHz, the highest reported to date.
Based on EUVA's work, commercial vendors expect to develop beta tools in 2007 and production systems by the end of 2009, according to the EUVA road map.
Established in June 2002, the EUVA () was set up under the auspices of Japan's Ministry of Economy, Trade and Industry (METI). The association includes lithography manufacturers Canon Inc. and Nikon Corp.; light source manufacturers Komatsu Ltd., Ushio Inc. and Gigaphoton Inc.; and device manufacturers Fujitsu Ltd., NEC Electronics Corp., Renesas Technology Corp. and Toshiba Corp. (Hitachi Ltd. and Mitsubishi Electric Corp. were each EUVA members before they founded Renesas in April 2003.)
EUV lithography requires reflective optics, employing a series of mirrors with some 40 precisely deposited layers of silicon and molybdenum, rather than the refractive lenses used in today's scanners. Developing a source of EUV radiation is particularly challenging, with the associated task of ensuring that debris from the source does not contaminate the optics.
Japan has capabilities in “all segments of semiconductor production,” from materials and equipment to light sources and device fabrication, said Masashi Ogawa, EUVA's executive director. “Taking advantage of those resources, we believe that our EUV R&D work will catch up and take the lead in overall performance and cost, though we came late.”
EUVA began source development work last year at corporate labs operated by Ushio and Komatsu. The association set up work on exposure tools and metrology at Canon and Nikon in April 2003.
EUVA is pursuing parallel development efforts for the EUV radiation source: a laser-produced plasma (LPP) method, which is being pursued at Komatsu's Hiratsuka R&D center, and a discharge-produced plasma (DPP) method, which is being worked on at Ushio's Gotenba lab.
The LPP method uses lasers to bombard a target, while the DPP approach employs a cathode-anode system.
The two groups are competing to reach a goal of 4 W of output power at the intermediate focus of the reflective optical system by March 2004. The Hiratsuka lab recently reported that it has generated plasma using xenon gas and a pulsed laser at a 10-kHz repetition rate with a pulse width of 30 nm. The lab is now changing to a laser with a pulse width of 7 nm. The engineers are predicting that the reduction in pulse width will raise the EUV power fourfold. To achieve the targeted 4 W at the intermediate focus, the engineers are working to achieve 13.8-W EUV power at emission.
Meanwhile, the Gotenba team working on the DPP method reported 6.3 W of EUV output power at a 2-kHz frequency. The engineers working on the DPP source estimate that 28 W is necessary at EUV emission to achieve 4 W of EUV output power at the intermediate focus. “It is not necessarily true that a higher repetition rate [with the DPP method] is better. We think 5 to 6 kHz is optimum,” said Hiroto Sato, lead researcher of the EUVA Gotenba branch.
By improving the EUV conversion efficiency, which is now a scant 0.1 percent, to a more respectable 0.5 percent, and by increasing the repetition rate to 6 kHz, the team expects to achieve 28 W at emission and 4 W at the intermediate focus by next spring.
“We are at the level of just having succeeded in generating EUV light. But our final target is 100 W, the level required for production systems. We think we'll be able to catch up to competitors that are now ahead of us,” said Sato.
Other companies and research organizations have reported better performance of their prototype light sources. Cymer Inc. (San Diego) has reported an efficiency of 0.5 percent and a repetition frequency of 4 kHz to get nearly 60 W of usable power. Xtreme Technologies GmbH (Gottingen, Germany) achieved about 0.55 percent efficiency and a repetition rate of 6 kHz, but with usable power of roughly 45 W.
EUVA intends to review progress by its two teams next March and may select one method to boost the source of EUV radiation to 10 W.
Separate effort
The National Institute of Advanced Industrial Science and Technology (AIST), a public research organization operating under METI, is also working on the source challenge apart from the EUVA effort. AIST is experimenting with tin (Sn) as the target material, and the researchers claim to have developed a method for controlling the debris generated when tin is used.
Controlling debris is essential to keeping the cost of ownership for EUV systems within the budgets of semiconductor makers. The collector is particularly vulnerable to debris, and the set of mirrors in the collector must be kept free from carbon and other contaminants in order to reflect EUV radiation precisely.
The peak wavelength of Sn plasma is near the 13.5 nm required for EUV lithography, and AIST researchers believe that Sn-based plasma has a much better spectral efficiency than xenon gas-based plasma, which has a peak wavelength of approximately 11 nm. The AIST team hopes to realize 5 percent efficiency this year using tin as the target material.
The effort seeks to bring Japan's effort in line with EUV ... Other companies and research organizations have reported better performance of their prototype light ...
- Status