|
|
Post by Baron von Lotsov on Jan 20, 2023 2:45:28 GMT
Lithography is the most difficult part of chip production and EUV is the most difficult lithography to accomplish. EUV stands for extreme ultraviolet, which refers to the kind of light used to etch the chips. The shorter the wavelength of light the smaller the detail which can be copied onto the chip and hence the more transistors you can fit on it and so the more powerful the chip.
Chip production was invented in the US in the late 70s and ever since the fist chips there has been the desire to produce better lithography so as to produce more powerful chips and so the wavelength has got progressively shorter and the engineering precision progressively harder. There are other problems too. The light has to be coherent, as per the kind of light emitted from a laser.
Lasers work at a particular wavelength according the energy of an atomic transition. Different molecules and atoms have different sets of traditions so will produce different wavelengths. CO2 lasers for example produce a light in the infrared. It does however get tricky to produce the EUV light and this is done by targeting minute droplets of molten tin with a laser at 50 000 times per second so that it creates a plasma and emits the right light. As you can imagine this is extremely difficult to do. Not only is it difficult to generate but you need a very powerful source of light, where an EUV light source will take a megawatt of energy input. The problem is the lenses used are almost opaque at this wavelength, so you really need to blast them with a lot of power to get enough light through, and the refractive index is very low as well, meaning you need multiple lenses to bend it enough. The mirror used under the wafer to focus the light in the final optical stage is some 50 separate mirrors each spaced an exact distance apart and all to within nanometres of accuracy.
This machine took $6bn to develop and a decade of research. The patents are held by the US and the US has decided China will be barred from getting access to their technology. It started with a ban on Huawei but the ban is spreading and being tightened so it affects anyone in China who needs high end chips, i.e less than 14nm which can be produced by DUV machines, the predecessor to EUV. So the challenge to China is develop a new EUV machine that does not copy the way it is done above by ASML and do it in about 1/10th of the time. Note the original technology was the work of several countries.
Well here is the story so far.
I think this is the tech referred to in the above two links.
|
|
|
|
Post by Equivocal on Jan 20, 2023 5:01:11 GMT
Lithography is the most difficult part of chip production and EUV is the most difficult lithography to accomplish. EUV stands for extreme ultraviolet, which refers to the kind of light used to etch the chips. The shorter the wavelength of light the smaller the detail which can be copied onto the chip and hence the more transistors you can fit on it and so the more powerful the chip.
Chip production was invented in the US in the late 70s and ever since the fist chips there has been the desire to produce better lithography so as to produce more powerful chips and so the wavelength has got progressively shorter and the engineering precision progressively harder. There are other problems too. The light has to be coherent, as per the kind of light emitted from a laser.
Lasers work at a particular wavelength according the energy of an atomic tradition. Different molecules and atoms have different sets of traditions so will produce different wavelengths. CO2 lasers for example produce a light in the infrared. It does however get tricky to produce the EUV light and this is done by targeting minute droplets of molten tin with a laser at 50 000 times per second so that it creates a plasma and emits the right light. As you can imagine this is extremely difficult to do. Not only is it difficult to generate but you need a very powerful source of light, where an EUV light source will take a megawatt of energy input. The problem is the lenses used are almost opaque at this wavelength, so you really need to blast them with a lot of power to get enough light through, and the refractive index is very low as well, meaning you need multiple lenses to bend it enough. The mirror used under the wafer to focus the light in the final optical stage is some 50 separate mirrors each spaced an exact distance apart and all to within nanometres of accuracy.
This machine took $6bn to develop and a decade of research. The patents are held by the US and the US has decided China will be barred from getting access to their technology. It started with a ban on Huawei but the ban is spreading and being tightened so it affects anyone in China who needs high end chips, i.e less than 14nm which can be produced by DUV machines, the predecessor to EUV. So the challenge to China is develop a new EUV machine that does not copy the way it is done above by ASML and do it in about 1/10th of the time. Note the original technology was the work of several countries.
Well here is the story so far.
I think this is the tech referred to in the above two links.
That was interesting; thank you. (Not wishing to be picky, but do you mean transition rather than tradition?)
This is a paper you might find interesting concerning the tin plasma method mentioned in the video.
The thing I found really interesting was the open admission in the video that the reason the west refuse to sell the machines to China is that China would misappropriate the technology given half a chance:
"With China's learning ability, it will soon be possible to understand the production method of [this] lithography machine and the principle of chip production." (About 3:10).
I had some experience of that kind of thing about thirty years ago. It concerned software for PLCs, duck eggs, bills of lading and letters of credit - nothing changes!
|
|
|
|
Post by Baron von Lotsov on Jan 20, 2023 13:46:04 GMT
Lithography is the most difficult part of chip production and EUV is the most difficult lithography to accomplish. EUV stands for extreme ultraviolet, which refers to the kind of light used to etch the chips. The shorter the wavelength of light the smaller the detail which can be copied onto the chip and hence the more transistors you can fit on it and so the more powerful the chip.
Chip production was invented in the US in the late 70s and ever since the fist chips there has been the desire to produce better lithography so as to produce more powerful chips and so the wavelength has got progressively shorter and the engineering precision progressively harder. There are other problems too. The light has to be coherent, as per the kind of light emitted from a laser.
Lasers work at a particular wavelength according the energy of an atomic tradition. Different molecules and atoms have different sets of traditions so will produce different wavelengths. CO2 lasers for example produce a light in the infrared. It does however get tricky to produce the EUV light and this is done by targeting minute droplets of molten tin with a laser at 50 000 times per second so that it creates a plasma and emits the right light. As you can imagine this is extremely difficult to do. Not only is it difficult to generate but you need a very powerful source of light, where an EUV light source will take a megawatt of energy input. The problem is the lenses used are almost opaque at this wavelength, so you really need to blast them with a lot of power to get enough light through, and the refractive index is very low as well, meaning you need multiple lenses to bend it enough. The mirror used under the wafer to focus the light in the final optical stage is some 50 separate mirrors each spaced an exact distance apart and all to within nanometres of accuracy.
This machine took $6bn to develop and a decade of research. The patents are held by the US and the US has decided China will be barred from getting access to their technology. It started with a ban on Huawei but the ban is spreading and being tightened so it affects anyone in China who needs high end chips, i.e less than 14nm which can be produced by DUV machines, the predecessor to EUV. So the challenge to China is develop a new EUV machine that does not copy the way it is done above by ASML and do it in about 1/10th of the time. Note the original technology was the work of several countries.
Well here is the story so far.
I think this is the tech referred to in the above two links.
That was interesting; thank you. (Not wishing to be picky, but do you mean transition rather than tradition?)
This is a paper you might find interesting concerning the tin plasma method mentioned in the video.
The thing I found really interesting was the open admission in the video that the reason the west refuse to sell the machines to China is that China would misappropriate the technology given half a chance:
"With China's learning ability, it will soon be possible to understand the production method of [this] lithography machine and the principle of chip production." (About 3:10).
I had some experience of that kind of thing about thirty years ago. It concerned software for PLCs, duck eggs, bills of lading and letters of credit - nothing changes!
Yes that was supposed to be transition. It was a little late at night when I wrote it!
Regarding things that change though, they say they want to use cyclotron radiation to generate the light source.
So it seems that they are tuning it down in frequency/energy to the low end of the X-ray spectrum.
Most of the references on that page are Western science except for when we get to ref 18 (Gong, Zhengliang; Yang, Yong (2018).
From this radiation being regard as a nuisance, it could save China from economic ruin. It's replacing one load of ultra tricky engineering with the tin laser, to something else which may well be far easier to engineer. There was even a cyclotron near Manchester once which got shut down in the 1980s.
|
|
|
|
Post by Equivocal on Jan 20, 2023 20:48:29 GMT
Regarding things that change though, they say they want to use cyclotron radiation to generate the light source.
From this radiation being regard as a nuisance, it could save China from economic ruin. It's replacing one load of ultra tricky engineering with the tin laser, to something else which may well be far easier to engineer. There was even a cyclotron near Manchester once which got shut down in the 1980s. I read that in the 'media' link. I thought it was more propaganda than science reporting. Still, if the Chinese have solved the optics problem with synchrotron sources it'll be pretty impressive.
I'll watch this space.
|
|
|
|
Post by Baron von Lotsov on Jan 20, 2023 22:09:07 GMT
Regarding things that change though, they say they want to use cyclotron radiation to generate the light source.
From this radiation being regard as a nuisance, it could save China from economic ruin. It's replacing one load of ultra tricky engineering with the tin laser, to something else which may well be far easier to engineer. There was even a cyclotron near Manchester once which got shut down in the 1980s. I read that in the 'media' link. I thought it was more propaganda than science reporting. Still, if the Chinese have solved the optics problem with synchrotron sources it'll be pretty impressive.
I'll watch this space.
The media link is translated Chinese. I mean having read much of what the Chinese say, I would not think that is anything out of the ordinary. It may look like propaganda but I think it is their culture. They are less formal and more expressive. We Brits look far more reserved. Even their company names can appear like jokes.
Yes i think using cyclotron radiation is a very interesting idea. I think you can engineer it to give you any wavelength in that part of the spectrum so you can select the exact wavelength that optimises the system.
I have a link to the principle of the Lloyd mirror as well.
It gives an interference pattern like a double slit but wastes far less light energy.
There is a paper from the Chinese here on it in 2013.
It looks like they have been doing the preparatory science for some time. The first indication of a chip ban from the US was 2015 when their supercomputer scared the US for being the most powerful in the word leading to a super computer chip ban.
|
|
|
|
Post by Baron von Lotsov on Feb 27, 2023 16:09:59 GMT
A little update on this is that I keep hearing that the firm Shanghai Microelectronics has now got a fully functioning and fully made in China lithography machine which works on a 90nm node. They have this in production and have already shipped a couple of units. This is important because it proves that even if the entire chip manufacturing supply chain to China was sanctioned, they can at least currently manufacture basic chips. This means all the chips used for car electronics can be made in China and in fact a large bulk of the chip market. Only a few chips need the higher resolution nodes like CPU processors. The thing is if they can do 90 nm one would expect them to get to higher nodes in a reasonable amount of time. They have got as far as to be self-financing as they have products on the market, so they will be constantly tweaked as they gain the production experience. I mean China has been a bit famous in the past for saying hey guys, look we can do 7nm in this lab, but realistically to get to production is years away. In business it is what you have that you can sell. They also do chip packaging technology as well.
|
|
