Re: SEM stereo - reflections

[T3D Dr. George A. Themelis <DrT-3d@xxxxxxxxxxxxxxxxxxxxxxxxxxx>]

Hi!  I am back in ACK mode!!!

A few more thoughts and comments on the SEM.  The principle is totally
different than any optical system but the end result is amazingly
similar!

In the SEM a beam of electrons is thrown on your sample.  The beam is
*scanning* a small area of interest.  Different things are emitted
from the sample as the beam of incoming electrons interacts with the
surface.  Things like, secondary electrons, backscattered electrons,
Auger electrons, x-rays, and others.  Different detectors are placed
around the sample to pick up those signals.  An image is formed by
mapping this information on a screen or piece of photographic paper.

Sounds very different than using a camera or a microscope.  For one
thing, here you have an image which is built gradually by scanning
and mapping.  For another, instead of light you are talking about
electrons and electron-mass interactions.

HOWEVER, the end result (when using the signal from the secondary 
electrons) is suprisignly similar to an optical image formed with the 
lens sitting at the top of the sample (where the beam of primary
electrons is emitted) and the light sitting at the place of the
detector.  Instead of thinking of a beam of electrons coming from the 
top, hitting the sample and the results of the interactions being detected
by the detector in the side, you can think of the situation as light being
emitted from the detector, reflected by the sample and directed to the
lens at the top to form the image.

To change the magnification in the SEM, all you need to do is make
the area scanned smaller.  Very simple and convenient because the
beam does not need to be refocused.

When it comes to 3d imaging you can use the model of the central
projection with the very good approximation of orthogonal projection
in sufficiently high magnifications.

There is not much math around the image formation mechanism in the SEM
and photogrammetric analysis.  The area of error analysis mentioned by
Prof. Mugnier mentioned is about the only area where I have seen some 
math.  This involves everything from errors in assuming the orthogonal
projection approximation, errors in measuring the angle of tilt, to 
display distortions, etc.

Back at Northwester I used a mirror stereoscope and a parallax bar to
measure "depth" (z-measurements) for photogrammetric studies on the
orientation of facets, etc.  At LTV we also have 2 mirror stereoscopes
and a parallax bar.  Analysis of stereopairs have been simplified, I
assume, with the use of computers and computerized image digitizing but 
I still find the parallax bar fascinating.  Perhaps someone could talk
about how this bar works.  The basic idea is to measure the parallax
(relative displacement) between any pair of points in your stereo 
pictures.  Assuming orthogonal  projection, there is a *very simple*
formula that will give you the difference in height between the two 
points, based on their measured parallax and known angles of tilt.

I can send a copy of my unpublished report on how to take view and measure
stereopairs in the SEM to those who are interested in the subject.  This
was written in 1990.  For the past few years at LTV I have used SEM stereo
pairs only to obtain a qualitative idea of how the rough surface looks.  
No measurements so far.  It is fascinating to see a 3d SEM image from a
pair of seemingly complex 3d micrographs!

I plan to have a presentation on "3D at LTV" and discuss all potential
stereo imaging applications at my workplace.  I plan to bring my TDC
stereo projector and show some of my regular slides too.  Will let you 
know (and ask for feedback and help!) when I get closer to doing it.

Thanks! -- George Themelis


------------------------------