Re: SEM theory of ops, and stereo

[T3D Edward Gosfield III, MD <gosfield@xxxxxxxxxxxxxxxxx>]

I have managed to locate my E-microscopy notes from a course i took at Cal
State E LA in 1984 or so (a reward to myself for the horrors of Pre Med
courses at the age of 36). I have also hauled out "Introduction to Electron
Microscopy" Third edition, by Saul Wischnitzer, Pergamon Press, 1981.  I
don't have the energy or self discipline at the moment to put together an
organized presentation, but i will try to address some of other peoples'
points, with some additional material as i feel able.  George probably knows
this stuff off the top of his head, but he should be elbow deep in a bunch
of Red Dots...
 
>John Vala wrote:
>"About 3D SEM:  From what I recall about how an SEM actually captures an image
>it seems as though the only way it could work is by tilt!  Correct me if I
>start loosing it here but, what you get from the camera is the result of an
>incredibly slow electron beam scanning over the surface of your specimen at
>the same rate an electron beam is scanning over the phosphor screen of the
>CRT in the camera.  The intensity of the phosphor is directly related to the
>amount of current flowing down the single stream from the electron gun, to
>the sample, out the stage, and into the amplifier to the CRT, or in the case
>of back-scattered, into the detector.  I can't see any element along the way
>that would act as a "lensing" element which we could apply to standard
>geometric optics.  The way I see it, imaging with an SEM is a fortunate
>phenomenon and shouldn't be confused with customary optical imaging.  What I
>see the "tilt" doing is slightly modifying the "infliction" of the electron
>beam with the sample causing the subtle variations in both position and
>surface luminance that we will later perceive as stereo.  Did that make
>sense?"

SEMs have a primary electron beam source, either composed of a filament of,
say tungsten, or a solid rod polished to a (optomistically) atomic scale
point, made of , say, Lanthanum hexaboride.  This is heated by a high
voltage supply, and boils off electrons (like the electron tube cathode,
beloved of audio freaks and designers of Russian MIG avionics)

These PRIMARY electrons (usually protected by a pretty hard vacuum) are
further accelerated by a 'grid' (again, like a vacuum tube) and passed
through a condensor "lens"--a coil whose magnetic field and physical
geometry allow it to create a beam uniform in intensity and shape.  The
condensor focusses the beam onto the specimen/stage.  Here it is scanned
across the specimen, controlled by orthogonally (hey John!) arranged scan
coils (magnetic) or plates (electrostatic).

As the beam scans across the stage, the primary electrons hit the specimen,
and SECONDARY electrons are knocked off and scattered.  The stage is
maintained at an electrical potential in series with the collector such that
the COLLECTED SECONDARY electrons complete a circuit.  These collected
electrons flow through a resistor, and the voltage drop through the resistor
is amplified by the VIDEO AMPLIFIER.  

a large number of SECONDARY electrons will cause a big V drop through the R,
and a relatively larger signal is sent to the video amplifier, and
subsequently applied to the grid of the CRT, causing a relatively large
acceleration of video tube electrons (completely separate from the
microscope e-beam) toward the screen, and therefor a relatively brighter dot.

The number of SECONDARY electrons coming off the specimen is determined by a
number of factors including the electron density of the specimen (hence
fancy staining and Osmicating techniques), the accelerating voltage of the
PRIMARY electrons, the size of the beam (intensity), and other things i forget.

The VIDEO scan is synched to the microscope e-beam scan.  SO the viewing CRT
has a raster frequency directly related to the image beam scan.
HOWEVER--the photography CRT usually has its own characteristics of phosphor
persistence, beam writing speed (controllable and variable by user) and beam
size.  Slower photoscans with finer beams give higher resolution images,
etc.  Photo writing scan speed can be varied from faster than you can
percieve visually, to slower than you can stand to stick around for while
your Polaroid is being 'written'.

The angle of incidence of the PRIMARY electrons with the specimen (and of
course, the tilt of the stage affects this) also affects the number of
electrons being knocked loose at any given moment/specimen detail location.
Lots of ART involved here, folks. Contrast is determined by the shape and
configuration of the surface.

Resolution is most strongly determined by beam diameter, which is determined
by the e-beam source size and condensor  (like Kohler illumination in light
microscopy, i think).  The condensosr is a demagnifier of the e-beam
produced by the source.  Raster spacing in the specimen plane also has some
influence on resolution. 

DEPTH OF FIELD is determined by the relative spread of the focussed e-beam
impacting 'further away' specimen details compared to the 'in focus' e-beam
at the plane of best focus, while the beam scans across the specimen.
Obviously this is affected by both the depth of physical detail in the
specimen, and by the tilt of the specimen stage with reference to the ebeam.

This doesn't produce geometric distortion, just difference in sharpness of
focus.

John B wrote:
> OK, so now we know George is right and you don't get
> distortion in an SEM pair by using a tilt table.

I am not certain that is what George means, and i certainly disagree that it
is true, since i have seen distortion in PAIRS, although tilt in itself does
not necessarily produce distortion in a single image *depending on the form
of the specimen*.

>> WHOA!  Hang on here.  
>> Am I right in assuming that a SEM images from the top of the specimen?

NOT  necessarily--you can focus anywhere on the specimen you like, just like
in regular macro photography.

>> and the tilt table is nominally sitting in the horizontal plane?
>> and the tilt axis is parallel to the SEM camera's imaging plane?

there are two axes of tilt provided, as well as rotation about the
horizontal plane.  For convenience one usually starts out with the stage
perpendicular to the e-beam, but one often doesn't end up there, if you want
to emphasize relief in the specimen.  The degree of tilt creates the
impression of relief and texture in the image.

>> Am I also right in assuming that a SEM images in a similar manner to
>> the way a conventional camera operates (and not for example like
>> the way an Xray is taken)?

see above.  Xray images are shadows cast on a phosphor screen, then
photographed.  

> ---snip-----it simulates transmission with a collimated 
>beam.  Did I hear you incorrectly, George?
> 
>The way I understood it, the rays coming to the image plane 
>were parallel to each other.  So it would be like illuminating 
>a slide transparency with a perfectly collimated light source.  
>If you do that, anywhere you intercept the beam, you get an 
>image.

It is my understanding that the primary electrons are focussed into as small
a spot as one can achieve--since that directly influences resolution.  So
the rays are not parallel at the point the hit the specimen.  Since the
image is built up by a raster, i don't think this matters.

>If I understood George correctly, the SEM makes it as if we were 
>standing infinitely far away when we took these toed in pairs.  In 
>that case, both images will be square.  So when we view the pair, 
>we see a flat square.
> ---snip----
>Thanks,
>John

The mathematics of figuring out the actual geometric point of view in this
situation gives me the willies.  Especially since magnification can be
changed and determined by  simple magnification of the video raster, without
changing any of the geometric elements of the imaging set-up.

and more John B:
>"How about the electron gun?  Is the beam steered from essentially a 
>point source?  If it is, isn't that the moral equivalent of a lens with
>focal length the distance of the "point source" to the sample?

The Primary beam is a point source, scanned across the specimen in a raster
pattern.  The Secondary electrons are collected in synch with the raster
scan, and generate a voltage applied in synch to the video tube, as
discussed above.

>And Jim C wrote:
>
>"I've had to skim the messages up to now (extremely busy at work).  Did
>someone just say that in SEM the projection is essentially orthographic (as
>though the object were at infinity)?  If so, then you're absolutely right;
>translating sideways relative to the object would result in a rigid shift
>of the entire image & no binocular disparity.

Simple translation of the stage produces a rigid shift of the frame of
reference, but if there is no change in stage angle of incidence re: the
e-beam, (ie, no change in tilt) then there is no disparity of lateral
information in the respective views, and presumably, no binocularity. But i
believe this is dependant on the size of the specimen relative to the field
of view of the e-beam, and the size of the specimen features.

more John B:
>We have several posts now saying that SEM images are orthographic
>projections.  So I think probably a flat circle lying on the tilt 
>table is a flat circle when viewed in the stereo viewer.

I think this depends on the specimen size, specimen depth, and the
magnification.

>But what 
>does it mean when you've taken a picture with the equivalent of an 
>infinite focal length lens and then viewed it with a finite-length 
>ocular?  What sort(s) of distortion(s) result?
>John B

I have personally created stereo SEM images with nasty stretch and close up
misery by using too much tilt and other errors unknown to me.  In my case i
was using a non-eucentric stage (center of tilt is not center of specimen)
which introduced a whole raft of other problems.

I hope (A. Woods? George?) someone can discuss this stuff, since they are
actually folks who try to measure depth using stereo SEM (or at least they
want to, judging from their pleas on sci.technique.microscopy.  I directed
them to P3D, but haven't seen their posts).


BTW--the 'colophon' at the end of the tech3d digest specifies unsubscribing
from P3D as the method for leaving T3D.  Heaven knows i can't imagine
unsubscribing from this excellent forum, but is that really the preferred
method? (we do want the frustrated and/or bored to know how to go away)

Ted G


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