Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
China's exports of germanium and gallium products in August plunged to zero，customs data showed on Wednesday, due to new export controls on the two chipmaking metals.
Beijing exported no wrought germanium products last month, compared to 8.63 metric tons in July when volumes more than doubled from June as overseas buyers rushed to lock in supply ahead of the curbs.
There were also no exports of wrought gallium products in August. In July, exports were 5.15 tons and 7.67 tons in the same month in 2022, the data showed.
...
Permit applications take around 45 working days to process, said a Chinese germanium trader and a germanium producer, who declined to be named because of the sensitivity of the matter.
Any semi historians know cost/benefit analysis for ASML going with laser plasma over synchrotron? Seems like at 200million per tool, it's not particularly more economical than cost of synchrotron/SSMB. There's one article behind paywall "EUCLIDES: European EUVL Program" from 1999 with author including Jos Benschop, who as far as I can tell was instrumental to ASML EUV effort. An interview with him a few years ago basically TLDRed to synchrotrons expensive, so we went with laser plasma... followed by many paragraphs of subsequent engineering challenges. Which one would presume would inflate final costs. For reference PRC's latest large scale synchrotron facility HEPS under construction is 1.3km circumference, with 60+ beam lines for 90 tool/experiment stations, for ~700M USD. Meanwhile, Tsinghua SSMB light source conceptual design is about 1/10 the size. I know they're not analogous, but seems questionable if plasma route ended up more economical, only that it works regardless of sunk cost, and new nodes are increasingly less cost effective relative to capital costs. What am I missing here?
Getting a synchrotron big enough to provide meaningful power output for industrial use wasn’t practical. 60 beam lines at lower power output may translate to much fewer beam lines at the higher output you’d need for meaningful scanning speeds. Furthermore the large size of such a facility meant that setting up production would take a lot of construction time, and every time you wanted to expand production you’d need to build a whole new synchrotron which will involve greater time suck. The large facility size would also impose logistics and operations costs for the rest of the production line, since you’d need to cart wafers back and forth across the physical distances around such a large structure.
I don't believe size of the synchrotron would be much of an issue because the vast majority of the machine would be placed in a subfloor and the final beam would be directed upwards into a floor with the lithography machine to minimize floor space usage.
If you’re trying to redirect the beam to a different area where logistics and operations factors are easier to manage you’ll also need to contend with dispersion and light loss at larger distances I reckon, which compounds the sufficient output problem.
From the abstract
From my understanding, in a conventional design (without SSMB) a giant synchrotron would emit broad spectrum radiation, for it to be used as a EUV light source you would need to connect the beamlines to undulators like those found in free electron laser. The electron beam pass through the undulators once and would go to a beam dump. If you want to be energy efficient since it takes a lot of energy to produce the electron beams, or basically generate EUV radiation at high repetition rates you need need to find a way to recover this electron beam. The undulators and recovery mechanism required for generating EUV radiation takes a lot of space.
SSMB manages to compress all the components in a smaller footprint, e.g., I don't think you need a long undulator connected to each beamline.
Cheers, seems like initial EUCLIDES explorations suggest synchrotron would be cheaper, at least for these tiny 8W wiggler designs. I guess reaching kW generation run into space issues implied by latenlazy. So SSMB value maybe size that makes operations workable, and there maybe cost savings in operations. E: as pointed out by coolgod.
- Status