When semiconductors were being researched at Bell Labs in the 1940s, mysterious component failures were eventually traced to researchers who had touched copper door knobs; the tiny number of copper atoms that migrated from the door to the workers hands was enough to ruin their work material.
Copper is incredibly destructive when it's in the wrong part of the fab. It's such a problem that most manufacturers typically separate it across physical cleanroom boundaries and use things like skybridges to move the wafers from one to the other. Once a lot touches the copper area, there's no coming back. It's basically a zombie virus.
Amazing and my further question: What degree of scientific discipline and/or stubborness would it take for the Bell Labs folks to actually get to the bottom of this and finally trace it back to the copper and the door knobs?
One of my professors in college was telling stories about finding out the reason yield dropped after lunchtime was someone eating salty potato chips and then going back to work.
The motivation seems easy though, wafers cost a lot of money! No one wants to see low yields. They are constantly testing and measuring everything to try and find root causes. Look at some of these diagrams if something can be isolated they try to do it.
You run split experiments over subsections of the line to find the source. It's essentially like binary search. You can do this pretty quickly with a skilled team.
Comments
Copper is incredibly destructive when it's in the wrong part of the fab. It's such a problem that most manufacturers typically separate it across physical cleanroom boundaries and use things like skybridges to move the wafers from one to the other. Once a lot touches the copper area, there's no coming back. It's basically a zombie virus.
Amazing and my further question: What degree of scientific discipline and/or stubborness would it take for the Bell Labs folks to actually get to the bottom of this and finally trace it back to the copper and the door knobs?
One of my professors in college was telling stories about finding out the reason yield dropped after lunchtime was someone eating salty potato chips and then going back to work.
The motivation seems easy though, wafers cost a lot of money! No one wants to see low yields. They are constantly testing and measuring everything to try and find root causes. Look at some of these diagrams if something can be isolated they try to do it.
You run split experiments over subsections of the line to find the source. It's essentially like binary search. You can do this pretty quickly with a skilled team.
Impressive level of engineering for modern fabs
Ireland doesn't offer the political roadblocks that the US currently faces. Until that changes, progress will continue to be slow in this country.
https://spectrumnews1.com/oh/columbus/news/2025/02/28/intel-...
Atomic Semi used to be working on at-home semiconductor production, but it looks like they got investment to pivot to something else. Oh well.