antiterror13
Brigadier
not only giving up EUV, GF also gave up the development of beyond 14nm, just too hard
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
not only giving up EUV, GF also gave up the development of beyond 14nm, just too hard
what I see here is SMIC absolutely dominating GloFo and UMC in capex. Much of TSMC capex has been for EUV which is far more expensive for same wafer yield as DUV, so there's no real comparison here.
If SSMB EUVL works out, do you expect the cost of Chinese advanced nodes to be lower than those of LPP EUVL? Since a single synchrotron can supply several stations, it seems a distinct possibility.
very possible. the limitation of course will be land use, but that's an infrastructure problem, and infrastructure is what China is good at.
the other thing is that SSMB EUVL allows for independent development of light source and optics, and isn't actually limited to 13.5 nm since synchrotron radiation is arbitrarily tunable.
Like I said before, the wavelength selection for lithography is all about the photochemical interaction with the resist. away from incident site of the EUV radiation. This causes line blurring.
DUV from ArF (193 nm) and KrF (248 nm) on the other hand, as the article above states, directly ionizes molecules to break bonds and cause chemical reactions. That means that they fulfill the condition of only causing chemical changes at exactly the point of incidence. This is especially true for ArF hardmask resists which help with the photoacid diffusion problem.
But why do they choose 13.5 nm then? The actual mechanism of the source is the same as any other gas phase light source: use some mechanism to ionize a gas phase molecule then when it falls back to ground state it emits light. You can't arbitrarily tune atomic energy levels, they are what they are and you have to just work with it.
But what if, you could get wavelengths shorter than 193 nm, with just as high brightness, but tune it to a wavelength that doesn't produce secondary electrons, so resist development is easier? And it was costless to tune the wavelength? That's what SSMB can offer.
You might wonder, OK, how come nobody uses synchrotron EUV?
Is it because of insufficient brightness? Then it was .
Is it cost? and has a worst case scenario That's 2.3 million EUR per beamline per year which is peanuts for a fab.
when you put it like this, why would anyone choose LPP over synchrotron???
capitalist economics. You can build, ship and sell an individual LPP light source but every synchrotron is a custom building that can't be moved. ASML is not in the construction business. Also, the only way a synchrotron light source makes its money back is by having tons of users simultaneously. That sounds like sharing and we all know sharing is communism.
Is it because of insufficient brightness? Then it was .
Is it cost? and has a worst case scenario That's 2.3 million EUR per beamline per year which is peanuts for a fab.
when you put it like this, why would anyone choose LPP over synchrotron???
capitalist economics. You can build, ship and sell an individual LPP light source but every synchrotron is a custom building that can't be moved. ASML is not in the construction business. Also, the only way a synchrotron light source makes its money back is by having tons of users simultaneously. That sounds like sharing and we all know sharing is communism.
You made an EUV as if it's a public utility .
If you took the worst case scenario for a Synchrotron facility, how many EUV machines would it be equivalent to?
Are we saying 1 beamline = 1 EUV machine?
And how many EUV machines each are being placed in new Fabs?
yes, 1 beamline = 1 EUV machine. Here's a diagram of what a synchrotron beamlines looks like.
It would change fab layout and increase fab construction costs. It would also increase the difficulty of multiple EUV steps, since 1 mask = 1 step = 1 layer. However, in exchange you can have economies of scale with multiple EUV optics sets being serviced with 1 synchrotron. How many EUV instruments do you need?
At some point the cost of adding another EUV source (~$50 million per source) will outweigh the cost of having 1 big synchrotron ($172 million per 8 beamlines for the Shanghai synchrotron or $25 million per source if you use all 8).
So basically synchrotrons are extremely attractive to fab companies like TSMC but not to the equipment manufacturers like ASML? That explains why SSMB EUVL was not really explored - one side does not have the skillset to produce such systems, while the other simply does not want them because they are not as profitable. I hope this works out well for China, I am sure they can find enough land if the project overall is successful - make Xinjiang the semiconductor capital of the world and watch MuriKKKa seethe, j/k.
- Status