One of my favorite examples of places where you can get tripped up is with implant table discontinuities. Chandra Mouli provided me with this recent example. Here we see two slightly different implant energies giving radically different implant profiles.

Basically, what is happening is that the implant moment tables are comprised of data and models from various sources. Here, for energies greater than 12 keV, a dual Pearson model is used. For energies less than this, data from another experiment was used. The message here is simply to be careful, and to know whether you are operating near one of these implant table discontinuities.

And don't forget, as Misha reminded me, that you can (and should) add PRINT.MOM to your implant statement to get more information on which tables were actually used. I have found this to be very helpful. And while we are at it, you can (and should) do something similar in your device simulator - add PRINT to the MODELS statement in Atlas to see which models are being used).

Michael

Update - Jan 5, 1999. I'll call this the case of the tunneling implants. It's really the same problem as above (implant table discontinuity), but manifests itself differently. Here, an 81 keV BF2 implant into poly shows a surprisingly tail in the silicon. Moral: check the range of validity for the default implant tables, and/or check integrated doses in your device to make sure there is no boron, for example, where there shouldn't be.

Update - Jan 6, 1999. Just to make this problem even more interesting, at still higher energies (e.g., 160 keV), the tail in the silicon disappears.

Update - Jan 25, 1999. Although I didn't mention it earlier, this problem seemed to be worse for the SVDP models. The vendor has confirmed that there is a problem with the SVDP models (the default), but that it is corrected in the next release.