looks like SiGe is where industry is going with GAAFet.
TSMC uses Ge
We will see how this goes
If China is determined to shut this off, we will see if it is able to take advantage of having its own Ge resources
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
looks like SiGe is where industry is going with GAAFet.
TSMC uses Ge
We will see how this goes
If China is determined to shut this off, we will see if it is able to take advantage of having its own Ge resources
EUV is kind of tricky, because as now doesn't give the kind of WPH of immersion lithography but because the reduced CD (13nm) it require less layers to make a chip, but still is necessary multiple patterning.
Advance chips can still be make using immersion lithography because the key of making multiple layers of very small features is not necessary the CD but accuracy, is about having pretty good alignment and accurate nano positioning, that is one of the hard parts.
In most areas they are advancing a breakneck speeds, as one guy said when he was in SEMICON China 2023 "I saw deglobalization at Semicon China", they are creating a big semiconductor supply chain that could have a global impact to much of the dismay of the US think tankers.
Lithography is the most opaque, I don't know what is the issue with ASML, but there is a lot of breakthroughs under the hood so the lithography supply chain is advancing quickly with the mass production of sub-systems, materials and components by China big lithography players already happening or starting soon.
Also The EUV supply Chain is starting to heat up so we should expect breakthroughs in the coming years.
Last edited:
Well, 28nm or even 14nm development have been done already from what I can see. Now, fabs (in this case SMIC) still has to validate them and put them in its process. 2023 was supposed to be year where SMIC has ASML + domestic production line. Given the newness of some of these domestic systems, SMIC will likely want to validate them for less advanced process first.
I don't think it matter what chineses fabs are using right now. It matters whether their capacity expansions can be supported by domestic SMEs. I think that should be doable. After all, there is very little 28nm expansion by Chinese fabs. The most advanced outside of SMIC (and a little from HLMC) are 40nm process. There just isn't need for that many fabs to all do 28nm when there are incumbent heavy weights like TSMC and UMC in China. You may ask why China allows Taiwanese fabs to expand so much on mainland. That is a good question, but I think the govt wants as much chip production on mainland as possible, even if its by foreign companies
SiGe is used for nanosheet GAAFET.
This is why the Ge ban is hugely devastating, not just the Ga ban. This makes it cost prohibitive to even do GAAFET research, much less bring it to production.
They already have production tools in most areas, we have a graph in this thread about that.
Depends what you are referring, if it is already installed equipment before 2022, I guess they are not going to throw those into the sea, they are going to continually use those as much they can, although I hear YMTC may very well throw those into the sea and replace it domestic alternatives, because they also bought services from US companies for those tools.
So most of the tools sold in China are probably direct purchases from Chinese semiconductor tools, materials, software and services providers if we consider how fast the companies in the supply chain are growing and the rest are probably coming from non US companies or tools made in South East Asia.
The point of SiGe is to be able to create a miscible alloy of Si and Ge as to tune the strain between epitaxial layers since SiGe alloy and pure Si will have the same chemical bonding, but slightly different lattice spacing in the crystal. The lattice spacing difference cannot be too big or too small. Too big and the interface delaminates from the substrate crystal, too small and nothing happens.
The strained silicon at the epitaxial interface has different electronic properties than the bulk silicon.
This is actually used in most sub-130 nm logic chips so it's even more devastating to lose Ge than just leading edge logic.
How much actual Germanium is used in those SiGe nanosheets? I doubt it is that much. As for strained silicon, that likely uses even less of it. These are likely minute amounts of what is not pure Germanium applied with a CVD process.
Not whole wafers of it like what is done with gallium in some applications.
Not whole wafers of it like what is done with gallium in some applications.
There are 40t export by China and 40t total production by other countries. Now, you take China & Russia's export off the market and you consider that US military will make sure to grab the x ton it needs for its military and national security stuff.
How much do you think there is left for everyone else? I did a search and each 12-inch wafer is 125g. TSMC alone has 120k wpm of capacity for 4/5nm
Let's say bw TSMC/Samsung, they want to produce 120k wpm of 5nm and more advanced wafers a month and need Ge or SiGe. They will need about 120k * .0125 * 12 = 180t of silicon wafers a year for that
Now even if you need 1/20th that mass of Ge for the layers, that would still be 10t needed for advanced chipmaking. Where are all these Ge coming from? Natsec folks will make sure they get their share of Ge before anyone else does.
And even if TSMC/Samsung gets their Ge eventually, it will still be a disruption to their supply chain while they figure out how they can source Ge.
The question is how much of the processes use Ge and whether they can replaced? And if they can get replaced, how much does that hurt yield and production for the near term? Who does it affect the most?
Based on your link, it seems like straining can be done silicon nitride
Unless you can invent a new group 4 element, Ge is kind of irreplaceable. There's nitride strain engineering processes but it's not a 1:1 replacement as the nitride is an insulator, not a semiconductor, and is a stoichiometric compound, not a miscible alloy. So they may have to start new device architecture for anything using Ge.
- Status