Confusion about diffusion leads to disillusion.
"Yousef, take 2 devices, one with finer features than the other, but otherwise the same. The smaller one will suffer from heat induced dopant migration at the same bulk rate as the larger one. However migration is like water wicking into news paper(although it is bidirectional) and since the features are a lot smaller on one it will suffer the effects of this migration more than the large feature CPU.
That means there will be a lower max safe operating temperature for the smaller featured devices."
Do you have any references on this? I think you may have discovered a major new failure mechanism in semiconductors.
We know about sodium ion migration (Note that sodium ions are NOT dopant atoms, and the problem is with sodium ions in oxides, not As or P or whatever dopants in the bulk silicon), but dopant atom diffusion during normal temperature operation (albeit at uncomfortable temperatures to humans, e.g., 100 C) is a new one for the literature.
Drive in diffusions are done at temperatures far, far higher than operating temperatures. In fact, the furnaces are "glowing yellow." Ought to give you some idea of the temperature (e.g. 900 C or even higher).
So, how much dopant diffusion will happen in a diffusion process where the the Arrhenius equation puts the temperature in the exponent? The difference between, say, 1300 K and 400 K is a vast amount when in the exponent of a thermally-activated diffusion process.
Answer: Those P, B, or As dopant atoms _ain't_ gonna move much at all. The ordinary temperatures of devices operating are "freezing cold" compared to the 900-1200 C for the drive-in diffusion.
Operating temperature limits are set by a bunch of other factors, notably circuit design. Junction spiking, latchup, and other failure mechanisms are accelerated at high temperature, as well as formation of some intermetallics, but these are well-known phenomena and have nothing to do with P or B dopant atoms migrating in the way you describe.
--Tim May |
