Chinese semiconductor industry

Status
Not open for further replies.

ansy1968

Brigadier
Registered Member
The physical limit using immersion lithography is 50nm for an NA of 1.35, that is why ASML can export machines to China because in theory their machine do not broke the 45nm of the Wassenaar Arrangement, made for EUV , but ASML can go as low a 38 using super resolution techniques with software and hardware to correct for aberrations, that is old technology. So is not far fetch to think that SMEE can archive 35-28nm.
@tokenanalyst bro for us common layman, can you describe the difference between ASML NXT 1980i , NXT 2000i and NXT 2050i , from what I know from my previous discussion with @WTAN @foofy @krautmeister and @FairAndUnbiased its all about optics, power sources and software? 2nd is NXT 2050i the last of ASML DUVL and the world most advance cause they can't find anymore improvement?
 

nlalyst

Junior Member
Registered Member
90nm is the single feature resolution, it can be described as a simple equation based on the characteristic of the dry SMEE DUV machine, the light source is 193nm wavelength, the NA of the optics is 0.75, so the resolution is k * LS / NA 0.35 * (193 / 0.75) = 90 nm, this is the maximum resolution without any other enhancement like multiple pattern or computational lithography. Add better optics and you can get a better resolution. Add an immersion system and you can increase the resolution even more, but most of the resolution increase in DUVL after 45nm comes from the software, not the hardware, so ASML is right to worry about the boost of lithography from China because computational lithography is an area that the Chinese are really hard at work.

Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
There are two kinds of resolutions:
- pitch resolution
- feature resolution

Nikon's and ASML's numbers refer to half-pitch resolution. The theoretical limit is close to 36nm (for NA=1.35). This is because the lowest the parameter k can go is 0.5 for pitch resolution. Practical limit is slightly higher, which is why they advertise their DUVL immersion resolution as <=38nm. The feature resolution, on the other hand, is always equal or better than half pitch resolution.

In that case the factor k1 is used in the above formula:
Please, Log in or Register to view URLs content!

Another critical metric that I haven't heard anything about regarding SMEE's machines is overlay. Modern IC products can have over 60 layers that need to be accurately stacked one over the other. The errors accumulate and if you don't keep them in control the electric properties of the product degrade, as does the yield.
 
Last edited:

Skywatcher

Captain
Lacking EUV lithography machines and photoresists or/and other chemicals is a BIG DEAL if it cannot source them from a source that is inert to US sanctions and dictates. That is why I posted earlier that TSMC is hoping - and I believe that they are - that China will be able to produce EUVL or other lithography machines comparable in performance to the ones of ASML, with entirely Chinese made components. I will even add the photoresists and other chemicals to that. Japan currently dominates the market share for the production of photoresists used in UVL and especially EUVL, but if the tells Japan not to sell those products to any foreign entity - bar the US of course - Japan would comply... Unless an entity such as China, which can and will essentially do as it pleases with regards to the sales of tech that it posseses the wherewithal to produce - begins to produce photoresists and other chemicals comparable in quality to those of Japan. In fact, China having tech capabilities in semiconductor and IC chip manufacturing equipment and chemicals substances synthesis very comparable to Japan, frees Japan from the yoke of US dictats.
ASML currently has 50+% gross margins and anticipates having that percentage (and overall revenue) to rise further in 2025.

Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!

Even if the Chinese EUVL of 2025 is only say, 50% wafers per hour capability and maxes out at 3-5nm compared to the latest ASML 2nm EUVL, its mere existence will stop ASML from charging monopoly prices to TSMC, Samsung, Intel even if the latter don't buy/make only token purchases of the Chinese first gen EUVL.
 

ansy1968

Brigadier
Registered Member
ASML currently has 50+% gross margins and anticipates having that percentage (and overall revenue) to rise further in 2025.

Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!

Even if the Chinese EUVL of 2025 is only say, 50% wafers per hour capability and maxes out at 3-5nm compared to the latest ASML 2nm EUVL, its mere existence will stop ASML from charging monopoly prices to TSMC, Samsung, Intel even if the latter don't buy/make only token purchases of the Chinese first gen EUVL.
@Skywatcher Sir your take on USG " request" on TSMC & Samsung about customer confidentiality? Will USS Missouri be refurbished for signing the surrender ceremony. :cool: :cool: :cool:
 

Quickie

Colonel
@tokenanalyst bro for us common layman, can you describe the difference between ASML NXT 1980i , NXT 2000i and NXT 2050i , from what I know from my previous discussion with @WTAN @foofy @krautmeister and @FairAndUnbiased its all about optics, power sources and software? 2nd is NXT 2050i the last of ASML DUVL and the world most advance cause they can't find anymore improvement?

Luckily we have a different kind of analyst rather than the "analyst" with a missing "a".
 

Weaasel

Senior Member
Registered Member
So essentially the playing field summary (for my non-familiar pov) of comprehensive semiconductor fab technology and foundry capabilities places Taiwan and South Korea at the top followed by the US and Japan and China trailing those. 5nm, 7nm, and 28nm done and achieved (rest is about scaling which is being done as we speak). China is at 14nm and 7nm at low scale lab level. All these players own and mastered all the tech chains. They can do it all themselves and using equipment, tech, and tooling they have mastered and produce themselves.

Literally no one else in this club. Germany, UK, Netherlands and so on have some niche expertise within a certain part of the supply chain but do not have comprehensive mastery of the entire chain.

India is far behind with only foreign tooling and equipment and that's for 1990s and 2000s era chip fab and without a single mastery of any of the tech where it produces its own equipment that is core to the process as opposed to a rig to handle wafers for example.

On design, China leads with the absolute best of them using universal architectures.

It is estimated by western observers that within 10 years China's own foundries should be at 5nm and it does have diminishing validity of Moore's law working for it.

On the flip side of silicon, China is at the forefront of alternative tech research along with the US and possibly other major players in Taiwan and South Korea.

Taiwan and South Korea are getting forced to transfer tech to the US and build their fabs within the US. Their leaders aren't looking too kindly on this but they have little say in the matter.

So within silicon semiconductors, the top player will be the US within a few years with supply coming from Taiwan reducing dramatically.

China's own sanctioned exporters like Huawei are in a desperate situation for the meantime. They do not have smartphones or computing that can be as competitive once their inventory is depleted. They are diversifying into other industries but Huawei's core of telecommunications is strong and all of it is strong within China except smartphone side of business. They need to wait until the Chinese foundries can provide them with equivalent chip performance which are done on lab level and using alternative technologies which China has been leading such as various stacking techniques (necessity the mother of all invention). But how the gap will be in 3 to 5 years in unknown.
Some other countries have to come up with EUV lithography machines or other types of similar or better performing lithographic machines devoid of any American made parts.... I read of Canon's nanoimprint lithography machines which is stated to be capable of producing IC chips of resolution comparable to that of EUV.
 

ansy1968

Brigadier
Registered Member
Some other countries have to come up with EUV lithography machines or other types of similar or better performing lithographic machines devoid of any American made parts.... I read of Canon's nanoimprint lithography machines which is stated to be capable of producing IC chips of resolution comparable to that of EUV.
@Weaasel bro Canon or Nikon had develop a DUVL comparable to ASML 2050i capable of producing 5nm using multi patterning, Gigaphoton had a 250kw LPP and lastly what you mention about Canon, all well and good but the problem they can't sell it to China which really needed all those product so its a dead end. It really funny how Japan can do seppuku just to please its master cause now they had the opportunity within 2 years the door will close and it will locked for good.
 

Skywatcher

Captain
@Weaasel bro Canon or Nikon had develop a DUVL comparable to ASML 2050i capable of producing 5nm using multi patterning, Gigaphoton had a 250kw LPP and lastly what you mention about Canon, all well and good but the problem they can't sell it to China which really needed all those product so its a dead end. It really funny how Japan can do seppuku just to please its master cause now they had the opportunity within 2 years the door will close and it will locked for good.
It's telling of Nikon's weakness that they can't build a 50-75% EUVL solution to challenge ASML's cartoonish monopoly profit margins (or get interest from the foundries to build one/banks to loan them the money to build it).
 

tokenanalyst

Brigadier
Registered Member
@tokenanalyst bro for us common layman, can you describe the difference between ASML NXT 1980i , NXT 2000i and NXT 2050i , from what I know from my previous discussion with @WTAN @foofy @krautmeister and @FairAndUnbiased its all about optics, power sources and software? 2nd is NXT 2050i the last of ASML DUVL and the world most advance cause they can't find anymore improvement?
The resolution of any lithography system is a simple equation k1 * (LS /numerical aperture)
LS = light source of the system
NA = optics or mirrors and the refraction index of the medium.
k1 = factor for aberration, diffraction and such.
Decreasing the wavelength increase the resolution, increasing the NA increase the resolution and decreasing the k1 factor increase the resolution.
The most straight forward method is to decrease the wavelength of the light source, that is what ASML did by going EUV.
Is possible to use shorter wavelength laser in DUV but lasers with wavelength shorter than 193nm have technical problems.
The NA is hard limited by the refraction index of the medium for air is 1 and water is 1.33 and the optics, for a dry system the theoretical limit NA for a dry system is 0.8-0.9. One option is to use a high n-index fluid to increase the numerical aperture like n=1.6-1.7, but that fluid has to be highly transparent to DUV light.
The only factor left is k and there is were most of the resolution increase in the last decade has come. So i think they don't see more use to tweak the software and the hardware to archive minimal gains in resolution. They are starting fresh again with a with new type of lithography EUV and then tweak that to archive better resolution without too many tricks.
 
Status
Not open for further replies.
Top