Posted by Dan (126.96.36.199) on May 22, 2005 at 21:39:47:
In Reply to: Can you profile a 50um square pattern? posted by Dan on May 22, 2005 at 21:29:28:
: I have a pattern that is 50um square. What kind of resolution can I get on a 3um thick layer?
Unfortunately you can't get a good profile from a pattern that is only 50x50um...
Having 50um width is good - SRA needs patterns at least 20um wide so that both probes can be placed into the pattern without shorting against each other.
With only 50um of length, we have an effective probing area of roughly 30um. That's because we need to stop beveling when we get to 1/2 the probe spacing away from the end of the pattern. And we need to complete the profile by the time we get to 1/2 the probe spacing from the other end of the pattern. If we don't have this "elbow" room we run into the geometric effect, which increases the measured resistance the closer we get to the edge of the pattern. This will translate into artificially lower carrier concentrations.
We carefully condition and shape our probe tips so that they can run with a 20um separation. Lighter load probes can run at a 2um step increment without landing on prior probe marks which have damaged the surface. Therefore the 30um length we have remaining will only allow for 15 data points.
15 points are fine for minimal characterization, but now we have the problem of the angle used to bevel the sample. The problem is is your layer really going to be 3um thick? Using 10:1 with a 2um step, we can get a 0.2um depth increment (200.0nm or 2,000 Angstroms) or 5.0 points to the micron. To reach 3um depth will require 15 points on a surface area of 20um by 50um. That 15th point will be at 3um depth. If your layer was P+ on P-, we would just barely be reaching a constant background which our algorithms need to know about in order to accurately reduce the surface layer. If your layer is P+ on an N background the algorithms donít need to know about the background. But we have the uncertainty of whether we reached the junction in time or did we start seeing compensation from the proximity of the pattern edge? To be sure we may need to use our next steeper angle which would only provide half the resolution. Getting only 7 points in a layer does not allow our reduction algorithms adequate sampling to provide reliable data.