Hi all,
We are trying to measure the composition of some synthetic glasses with our JEOL8200 EPMA. They were made by melting synthetic chemical mixtures at high temperature and then quenched into water.
These glasses contain essentially FeO, SiO2 and small amount of Cu2O. Some of these glass samples looked fine and homogeneous under SEM, but some show fine precipitations (sub-micron) of Cu metal which formed during cooling quenching process. (See photo(left) attached)
We initially used 0 probe diametre (roughly 1um) and did a line scan across the sample as shown in red circles in the photo and we found the concentration of Cu/Cu2O scatters depending on the position where measurement was made, that's understandable. Later we tried to increase the probe size to 5um and measurement results of Cu/Cu2O show much less scattering.
What we are really interested is the average composition of this glass phase(including Cu metal precipitates).
Here is the question, to what extent by increasing the probe size for the measurement in this case would give results that are closer to the average composition?
Hi Jeff,
This is an excellent question and difficult to answer. Well, I can give you an answer: do not defocus the beam on heterogeneous materials. I think it was Chuck Fiori who said: "if the interaction volume is heterogeneous, all bets are off!".
Here is a post related to your question that you might find helpful:
http://probesoftware.com/smf/index.php?topic=198.msg896#msg896also this analysis by Julie Chouinard comparing averaging intensities vs averaging concentrations. The latter is what we really want to do!
http://probesoftware.com/smf/index.php?topic=44.msg145#msg145and then this on pixel boundary effects:
http://probesoftware.com/smf/index.php?topic=49.msg159#msg159Ideally you'd want to analyze both the inclusions and the matrix and sample enough of your material to obtain a representative ratio between the matrix and the inclusions and then calculate the average concentrations. But because your inclusions are so small this will be difficult. What exactly is the size range of these inclusions?
I'd probably run at a very low overvoltage, with a highly focused beam and acquire a quant map over a "representative" area that includes enough matrix and inclusions and then calculate the average composition by averaging the concentrations of all the pixels.
But as to your specific question on how much of an error one would get from defocussing the beam... I'd probably need to run a Monte Carlo model with that geometry.
You'd only have two materials to model, but the geometry model would be complex. I'd take a look at this post here:
http://probesoftware.com/smf/index.php?topic=59.msg1340#msg1340and run several simulations each with a different beam "aperture" in Penepma to see the effect of spreading the beam out on the relative intensities.