Chinese semiconductor industry

Status
Not open for further replies.

Xizor

Captain
Registered Member
The Wassenaar Arrangement was basically that China could only buy machines two generations behind the leading edge. This was based on historical trends where they kept reducing the wavelength of the light sources.
Immersion lithography uses the same light source yet is capable of higher detail. Immersion is basically like putting a lens in front of the regular lens a machine has to imprint finer detail. Since they couldn't upgrade the light source for a long time people worked on improving the lenses.
Then there is another factor where increased detail came with multiple-patterning. Modern multiple-patterning techniques also emerged because of the limitation on improving the light source. These are computational lithography techniques where you use multiple passes to enable the production of increased detail. These require additional masks but no extra tools.

Because of all of these new developments the West had to re-examine the export limitations on tools. For a long time China could not import ArF tools for example. But ever since multiple-patterning was developed it meant China could just use available KrF lithography tools with multiple-patterning to produce at similar resolutions in practice as ArF dry. The industry was getting its initial test EUV lithography tools and ArF immersion was widely available. So the ban on ArF dry was relaxed. Eventually China could manufacture ArF dry tools on their own, EUV tools were in actual use, so the West relaxed the ban on ArF immersion tools. Now China can manufacture its own ArF immersion tools.

The idea of that US AI panel with Eric Schmidt to ban the sales of ArF immersion tools to China is pure brain damage.
What is there to prevent China from modifying the optics of the ArF dry machines they buy and turning them into immersion lithography machines? That is why the industry made ArF immersion tools available to China so readily. Just think about it. 65/28nm requires the ArF light source which is the hardest thing to build. 14nm requires the immersion optical apparatus on top of that which is not that hard to build. 10nm/7nm additionally uses a small mechanical device to ease the multiple-patterning mask switching. Nothing about either of those last two steps is hard to produce. You could just rip the light source out and put it inside your own machine with your own optics.
The fact that China can manufacture their own ArF light sources today makes the whole ban on ArF unviable.
Those US AI panel 'experts' knows nothing about the semiconductor tools sector and it shows.

The West just needs to continue to develop EUV technology to maintain an edge over Chinese technology. Next generation EUV light source with increased power and reduced wavelength and improved optics mean the West will retain its edge in the long term.

In short the idea that Eric Schmidt has that because ArF immersion is enough to produce certain AI hardware it needs to be blocked is pure nonsense. For one China can already build these tools and SMEE will ship its initial devices to customers this year. For another if you do sell ArF machines to China, regardless of type, you are basically providing the hardest to manufacture component and the Chinese can replace the rest easily. For another similar sanctions have been put in the past for devices with a certain computational capability. At one time I think they had limitations based on the number of MIPS or FLOPS a device could have. As these eventually filtered to a point they were available in mass produced products the sanctions lost meaning since it would be easy for any country to buy these mass produced devices on the gray market without anyone spotting it and removing the required components for their own military or government use.
This area is not my zone. I've minimal knowledge so thank you for the detailed write up.

Immersion lithography uses a layer of water under the lens, right?
The way I understand -

DUV can use ArF or KrF
Add in the water layer + modification and you get Immersion.
But all this is inferior to the new EUV (obviously).

I'm not disagreeing but if Immersion lithography is relatively easy then why is there no immersion litho from China? Are patent infringement holding back China from doing so?

BTW, 14 of the 15 members of the AI commission. They have voted to expand the ban to 16nm.

1619560072774.png
 

gelgoog

Lieutenant General
Registered Member
...
Immersion lithography uses a layer of water under the lens, right?

The way I understand -
DUV can use ArF or KrF
Add in the water layer + modification and you get Immersion.
But all this is inferior to the new EUV (obviously).

I'm not disagreeing but if Immersion lithography is relatively easy then why is there no immersion litho from China? Are patent infringement holding back China from doing so?

BTW, 14 of the 15 members of the AI commission. They have voted to expand the ban to 16nm.

Yes. Immersion lithography uses a layer of water to act as a secondary lens. You do need new primary lenses but that is not an issue. For what it's worth it took ASML one year to go from a concept proposed by TSMC to produce the first lithography machine. Another year to ship machines to customers.
1619564665940.png1619564718496.png

China had a historical delay with regards to the production of light sources.
Compare that with the time it took to go from ArF to ArF immersion (ArFi) which was like 1-2 years.
As per prior posts on this thread SMEE will be shipping four ArFi tools to customers this year.
SMEE already has ArF dry tools available and is shipping those i.e. SSA600/20.

This should help to get an idea for how long the industry in the West took to adopt each of those technologies.
1619565017578.png

AFAIK patents are not the issue. Even if they were the Chinese government would not let patents enable foreign companies to exclude the whole Chinese market. That goes against the whole idea of having patents in the first place. Most of the obstacles are trade secrets not patents. i.e. the 'secret sauce' the industry uses to make their products. The issue is not knowing what you need to do but how to do it in practice.

What the AI commission wants is immaterial since Chinese industry will produce the tools to go up to 7nm.
The cat is already out of the bag. China has had access to ArFi tools from ASML for a couple of years already.
 
Last edited:

latenlazy

Brigadier
TSMC has used the same tactic on the past to crush Chinese competitors.

I am sure that the CPC will protect SMIC from these kinds of shenanigans by TSMC. If it doesn't then the CPC fully deserves to lose the tech war.

Anyway, there is still a lot of demand so TSMC wont capture the whole market
Why? If TSMC builds a factory in China that gives China more leverage over TSMC. If TSMC decamped from China it doesn’t take its factories or most of its workers with it. Most of that stays in China. TSMC building factories in China is material FDI. It’s basically getting free assets from someone to help build your industrial base.
 

Bellum_Romanum

Brigadier
Registered Member
The goal is economic development and that involves maximizing efficiency. I don't grow my own vegetables or wire my house, etc
If China was to follow @Oldschool's advice or admonitions then it's pretty much copying the disastrous policies of Chairman Mao when he tried to superjump the industrialization of the country to leap frog the Soviet Union resulting in the great famine. The man is just too emotional and not really thinking in a strategic sense.
 

latenlazy

Brigadier
Yes. Immersion lithography uses a layer of water to act as a secondary lens. You do need new primary lenses but that is not an issue. For what it's worth it took ASML one year to go from a concept proposed by TSMC to produce the first lithography machine. Another year to ship machines to customers.
View attachment 71411View attachment 71412

China had a historical delay with regards to the production of light sources.
Compare that with the time it took to go from ArF to ArF immersion (ArFi) which was like 1-2 years.
As per prior posts on this thread SMEE will be shipping four ArFi tools to customers this year.
SMEE already has ArF dry tools available and is shipping those i.e. SSA600/20.

This should help to get an idea for how long the industry in the West took to adopt each of those technologies.
View attachment 71419

AFAIK patents are not the issue. Even if they were the Chinese government would not let patents enable foreign companies to exclude the whole Chinese market. That goes against the whole idea of having patents in the first place. Most of the obstacles are trade secrets not patents. i.e. the 'secret sauce' the industry uses to make their products. The issue is not knowing what you need to do but how to do it in practice.

What the AI commission wants is immaterial since Chinese industry will produce the tools to go up to 7nm.
The cat is already out of the bag. China has had access to ArFi tools from ASML for a couple of years already.
In terms of on the ground realities what makes development of immersion slow is neither inaccessibility of patents nor trade secrets. The real challenge is that controlling a layer of water on top of a wafer to do what you want it to do at high enough speeds to have competitive production performances is really hard, and requires a lot of practice and fine tuning. Mass flow control is a painstaking thing to do in practice. Process knowledge for this sort of thing always takes time. This is the part of technological development that you can’t leapfrog or shortcut. Some technologies have more of this than others. EUV, once you have a sufficiently powerful light source, is in some ways actually easier from a process knowledge threshold standpoint than immersion.
 

latenlazy

Brigadier
If China was to follow @Oldschool's advice or admonitions then it's pretty much copying the disastrous policies of Chairman Mao when he tried to superjump the industrialization of the country to leap frog the Soviet Union resulting in the great famine. The man is just too emotional and not really thinking in a strategic sense.
Not just Mao. People here like to make fun of India for its slow development all the time...and then ironically proceed to recommend, even demand, that China take India’s approach to industrial development policy.
 
D

Deleted member 15949

Guest
In terms of on the ground realities what makes development of immersion slow is neither inaccessibility of patents nor trade secrets. The real challenge is that controlling a layer of water on top of a wafer to do what you want it to do at high enough speeds to have competitive production performances is really hard, and requires a lot of practice and fine tuning. Mass flow control is a painstaking thing to do in practice. Process knowledge for this sort of thing always takes time. This is the part of technological development that you can’t leapfrog or shortcut. Some technologies have more of this than others. EUV, once you have a sufficiently powerful light source, is in some ways actually easier from a process knowledge threshold standpoint than immersion.
On this topic though, me do wonder how much faster development could be if you poach the relevant ASML/Nikon/subsystem MechEs
 

latenlazy

Brigadier
Guys it’s very clear by now that Oldschool cares a lot more about sticking it to the rest of the world and having China so everything on its own than effective, results oriented economic development or maximizing economic welfare. Can we just agree to disagree over these priorities and move on?
 

Xizor

Captain
Registered Member
Yes. Immersion lithography uses a layer of water to act as a secondary lens. You do need new primary lenses but that is not an issue. For what it's worth it took ASML one year to go from a concept proposed by TSMC to produce the first lithography machine. Another year to ship machines to customers.
View attachment 71411View attachment 71412

China had a historical delay with regards to the production of light sources.
Compare that with the time it took to go from ArF to ArF immersion (ArFi) which was like 1-2 years.
As per prior posts on this thread SMEE will be shipping four ArFi tools to customers this year.
SMEE already has ArF dry tools available and is shipping those i.e. SSA600/20.

This should help to get an idea for how long the industry in the West took to adopt each of those technologies.
View attachment 71419

AFAIK patents are not the issue. Even if they were the Chinese government would not let patents enable foreign companies to exclude the whole Chinese market. That goes against the whole idea of having patents in the first place. Most of the obstacles are trade secrets not patents. i.e. the 'secret sauce' the industry uses to make their products. The issue is not knowing what you need to do but how to do it in practice.

What the AI commission wants is immaterial since Chinese industry will produce the tools to go up to 7nm.
The cat is already out of the bag. China has had access to ArFi tools from ASML for a couple of years already.
So China has become third country to produce Immersion lithography along with Japan and Netherlands?

But the page
Please, Log in or Register to view URLs content!


Doesn't mention any Immersion technology or ArF'i'. Whatmore, it stops at 90nm.
 

gelgoog

Lieutenant General
Registered Member
So China has become third country to produce Immersion lithography along with Japan and Netherlands?

But the page
Please, Log in or Register to view URLs content!


Doesn't mention any Immersion technology or ArF'i'. Whatmore, it stops at 90nm.

It does not mention it because it isn't shipping yet.
They made a prototype and tested it and will be shipping (not shipping yet) four machines this year.
That 90nm machine is ArF dry.
 
Status
Not open for further replies.
Top