Lithography semiconductor machining precision problem
Appears in 2 lectures.
Appearances across the corpus
Used to motivate why stress relief matters before precision machining — heat-treated material walks during stock removal, an issue Tom describes as a "20-billion-dollar problem" for the semiconductor industry.
A number of years ago this was a big problem in the lithography business for making computer chips. They needed to machine things within tolerances of like five microns. That's 0.0002 inches, five microns. You can't do that even if you stress-relieve the steel or the aluminum — it will still move that much if you machine off a quarter of the thickness out of one-inch plate. You machine it and come back and measure it and it's moved. So you have to machine it again, and then it moves a little bit more. It does sort of start tapering off, but you have to be careful. You try to grind something flat and you start getting very precise.
Continuation of the lithography case — at sub-wavelength feature sizes, photolithography gives way to UV / electron-beam patterning, with chips patterned in sections matched together because tolerance can't be held over a centimeter.
On the chips, you don't etch the ships. We've got chips on one end, we've got ships on the other end. The chips, they use photolithography, although now they're using high ultraviolet lithography, because you get into the problem of the wavelength of light — these dimensions are smaller than the wavelength of visible light. So they're using ultraviolet, and they've talked about using electron beam lithography. I believe they actually pattern one part of the chip and then they match it to another and match it to another, because over a centimeter you can't hold that type of tolerance. At one time this was supposed to be a 20-billion-dollar problem for the semiconductor industry, about 10 or 15 years ago.