why is p-substrate typically used as wafer in the typical cmos process flow? why not n substrate?with respect to memories, Has it got anything to do with the aplha-paritcle radiation induced errors (soft errors) ? please explain.
1
-
$\begingroup$ Would Electrical Engineering be a better home for this question? $\endgroup$Qmechanic– Qmechanic ♦2015-01-31 14:49:01 +00:00Commented Jan 31, 2015 at 14:49
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
The mobility for electrons is generally higher than for holes, considering typical doping profiles, etc. Therefore, it is easier to achieve higher signal speeds when the conduction flow travels through an n-type silicon. Hence, you would want the substrate to be p-type since you will be doping in the conduction channel.
For the math which allows you to compute electron and hole mobilities in Si, and plots, take a look here.
$\begingroup$
$\endgroup$
Up until the mid-1990's, your could get silicon wafers grown by float-zone (FZ) or Czochralski (CZ), in (100), (110), and (111) orientations, doped with B, P, and the occasional exotic (As, Sb). By the introduction of 200mm (8") wafers, you became limited to (100) CZ B-doped wafers.
Why?
The float-zone process could not scale up to the larger wafer diameter with good throughput and dopant uniformity. CZ became the only way to go at 200mm, and by now it really really is the only way to go.
The dopant uniformity was also a big factor - you can't have the background doping levels change not just across an individual wafer but also from wafer to wafer along the boule. As transistor features decrease, the required doping levels need to be (a) less than before, and (b) more tightly controlled. With a varying background doping that became harder and harder.
Boron became the only available bulk dopant because it behaves much better in the crystal growing process. It diffuses faster in the liquid, so you reduce the center-to-edge doping variation. It also segregates less during solidification, so the distribution along the growth direction stays more constant. This is good crystal growth thermodynamics and kinetics (materials science) driving this selection.
As for direction, (110) and (111) were always niche uses, and nobody wanted to deal with them in the high volume market (since they weren't high volume). (100) won for CMOS.
