Kerrick, who wrote the 'bible'? about carbon coating in American Mineralogist eons ago (American Mineralogist, Volume 58, pages 920-925, 1973, The Role of Carbon Film Thickness in Electron Microprobe Analysis) uses the density of the _coating_ (not the graphite rod) as 1.3 g/cc. However, PfE uses 2.1 g/cc. Graphite is listed in the Google encyclopedia of knowledge) as: amorphous: 1.8–2.1 g/cm3; graphite: 2.267 g/cm3; diamond: 3.515 g/cm3. Why is the Google value, maximum, of 2.1 being used? For all I know, Kerrick is wrong, and Mr. Google is correct, but what is the proof? Is there some reference where the carbon coat density has been directly (or indirectly) measured?
Quote from: JohnF on December 04, 2017, 12:01:45 PM
Kerrick, who wrote the 'bible'? about carbon coating in American Mineralogist eons ago (American Mineralogist, Volume 58, pages 920-925, 1973, The Role of Carbon Film Thickness in Electron Microprobe Analysis) uses the density of the _coating_ (not the graphite rod) as 1.3 g/cc. However, PfE uses 2.1 g/cc. Graphite is listed in the Google encyclopedia of knowledge) as: amorphous: 1.8–2.1 g/cm3; graphite: 2.267 g/cm3; diamond: 3.515 g/cm3. Why is the Google value, maximum, of 2.1 being used? For all I know, Kerrick is wrong, and Mr. Google is correct, but what is the proof? Is there some reference where the carbon coat density has been directly (or indirectly) measured?
Hi John,
Interesting. I knew about Kerrick's work with regard to the electron energy loss calculation for the conductive coating, as that is what is used in CalcZAF and Probe for EPMA for correcting for differences in the coating between one's standards and unknowns. But I did not notice his value for carbon density. Interesting that it is so small.
As you say, it's variable, so enter whatever value you prefer. That is why the field is editable!
john
Jurek et al, 1994 (Microchimica Acta) reports sputter coated C films as having ~ the same density as bulk graphite (2.25 g/cc) but vacuum evaporated films have significantly lower densities (1.3 - 2.0 g/cc), closer to but still generally lower than the bulk density value of amorphous C (1.8 - 2.1 g/cc).
Carbon's not the only material that can exhibit lower than bulk densities as thin films. I've just submitted a paper to MiMi which reports Al thin films with a density of ~2 g/cc compared to the bulk value of 2.7 g/cc.
Stephen Reed in Electron Microprobe Analysis and Scanning Electron Microscopy in Geology quotes 2 g cm-3 citing Jurek et al 1994.
Ben
Thank you Mike and Ben for your input. We are quite interested in this as we found that there was enough variation in the carbon coats on a sample vs the standard that the analytical totals were wacky, whereas the stoichiometry was perfect (plagioclase). Inputing different coat thickness values in PfE (from StrataGem k-ratios) saved the day. However, we noted that the thickness in StrataGem (which uses mass thickness) is a variable dependent upon density. So getting an accurate density would be rather useful.
Quote from: JohnF on December 13, 2017, 09:34:25 AM
Thank you Mike and Ben for your input. We are quite interested in this as we found that there was enough variation in the carbon coats on a sample vs the standard that the analytical totals were wacky, whereas the stoichiometry was perfect (plagioclase). Inputing different coat thickness values in PfE (from StrataGem k-ratios) saved the day. However, we noted that the thickness in StrataGem (which uses mass thickness) is a variable dependent upon density. So getting an accurate density would be rather useful.
Hi John et al.,
I ran into this same carbon coat density question yesterday on a metallurgical sample I am running for another university.
Basically it's a Ti-Nb couple and quite interesting because, for this system the diffusitivities (is that the correct word?) are different by around 5 orders of magnitude, so Nb diffuses readily into Ti, but Ti will not diffuse into Nb. Really weird.
Anyway, the sample is uncoated, but the standards are coated with our usual 20 nm of carbon. So when I went to do the quant I got this output:
Un 4 Ti-Nb boundary, Results in Elemental Weight Percents
ELEM: Ti Nb O
BGDS: EXP LIN EXP
TIME: 40.00 40.00 40.00
BEAM: 30.01 30.01 30.01
ELEM: Ti Nb O SUM
210 -.009 101.534 1.173 102.698
211 -.008 102.291 1.260 103.544
212 -.001 101.548 1.245 102.792
213 -.023 101.994 1.244 103.215
214 .026 101.660 1.221 102.908
215 .005 102.474 1.201 103.679
216 .063 102.108 1.287 103.458
217 .022 102.335 1.279 103.636
218 -.006 101.594 1.209 102.797
219 .043 101.464 1.317 102.824
220 .003 101.527 1.241 102.771
221 -.013 102.022 1.305 103.314
222 -.001 101.583 1.252 102.833
223 .030 101.848 1.171 103.049
224 .019 101.799 1.305 103.123
225 .035 102.506 1.309 103.849
This is using an 8 keV electron beam because they want the best spatial resolution they can get (we're doing another run tonight at 6 keV to try and further improve the spatial resolution and also to further reduce the SF effects).
Anyway, the above output clearly shows the need to specify the carbon coat for the standards and turn it off for the unknown. For the unknown this is easily done from the Calculation Options dialog by unchecking the "Use Unknown Conductive Coating" checkbox. For the standards one needs to use the Standard | Edit Standard Coating Parameters menu, but by default they are set to 200 nm of carbon at 2.1 gm/cm3.
Then one has to go into the Analytical | Analysis Options dialog and explicitly turn on the conductive coating correction by checking these two checkboxes:
(https://smf.probesoftware.com/gallery/395_01_05_18_2_48_56.png)
Once that is done we should be OK (or should we?), and we now obtain this analysis:
Un 4 Ti-Nb boundary, Results in Elemental Weight Percents
ELEM: Ti Nb O
BGDS: EXP LIN EXP
TIME: 40.00 40.00 40.00
BEAM: 30.01 30.01 30.01
ELEM: Ti Nb O SUM
210 -.008 95.301 1.024 96.316
211 -.007 96.010 1.099 97.102
212 -.001 95.313 1.086 96.398
213 -.021 95.731 1.085 96.795
214 .024 95.418 1.065 96.508
215 .004 96.183 1.047 97.234
216 .057 95.838 1.122 97.017
217 .020 96.051 1.116 97.187
218 -.005 95.357 1.054 96.406
219 .040 95.232 1.149 96.421
220 .003 95.293 1.083 96.379
221 -.011 95.756 1.139 96.883
222 -.001 95.345 1.092 96.436
223 .028 95.596 1.021 96.645
224 .017 95.547 1.138 96.703
225 .032 96.210 1.142 97.384
So now our analysis is low! Of course these coating effects are significantly amplified when running at low voltage (or low overvoltage), but we're pretty sure that the coating on our standards is close to 20 nm based on the color on polished brass as discussed here:
http://smf.probesoftware.com/index.php?topic=921.msg7063#msg7063
So then there's the density question and the consensus above seems to be that the evaporated deposited coatings are often less dense than their "accepted" values. So I changed the carbon density from 2.1 to 1.4 and voila:
Un 4 Ti-Nb boundary, Results in Elemental Weight Percents
ELEM: Ti Nb O
BGDS: EXP LIN EXP
TIME: 40.00 40.00 40.00
BEAM: 30.01 30.01 30.01
ELEM: Ti Nb O SUM
210 -.008 97.376 1.072 98.439
211 -.007 98.101 1.151 99.245
212 -.001 97.389 1.137 98.524
213 -.022 97.816 1.136 98.931
214 .025 97.496 1.116 98.637
215 .004 98.277 1.096 99.378
216 .059 97.925 1.175 99.159
217 .021 98.143 1.168 99.332
218 -.005 97.433 1.104 98.532
219 .041 97.307 1.203 98.550
220 .003 97.368 1.134 98.505
221 -.012 97.842 1.192 99.022
222 -.001 97.422 1.143 98.564
223 .029 97.677 1.069 98.775
224 .018 97.629 1.192 98.838
225 .033 98.306 1.195 99.534
You gotta admit, that's pretty impressive...
Interesting, as we say, Mr. Donovan.
Back in the carbon film density saddle again.
With easy-to-access articles on the inter webs:
Characteristic Energy Losses of Electrons in Carbon, August 1960, Journal of Applied Physics, Vol 31, No 8 by Lewis Leder and J A Suddeth (NBS folks), their table 3 reports 1.35 density for evaporated carbon, citing the 1958 paper by W. Langbein, in Naturwissenschaften, 45, p. 510 entitled (all in German) Elektroneninterferometrishce Messing des inneren Potentials von Kohlenstoff-Folien, which gives the density (experimental) as 1.35 +/- 0.03.
Quote from: JohnF on May 02, 2018, 07:31:02 AM
Interesting, as we say, Mr. Donovan.
Back in the carbon film density saddle again.
With easy-to-access articles on the inter webs:
Characteristic Energy Losses of Electrons in Carbon, August 1960, Journal of Applied Physics, Vol 31, No 8 by Lewis Leder and J A Suddeth (NBS folks), their table 3 reports 1.35 density for evaporated carbon, citing the 1958 paper by W. Langbein, in Naturwissenschaften, 45, p. 510 entitled (all in German) Elektroneninterferometrishce Messing des inneren Potentials von Kohlenstoff-Folien, which gives the density (experimental) as 1.35 +/- 0.03.
How cool is that! Nice find!
I was just guessing the carbon density based on the totals, but if I specify a carbon density of 1.35 my totals are even closer to 100%:
Un 4 Ti-Nb boundary, Results in Elemental Weight Percents
ELEM: Ti Nb O
BGDS: EXP LIN EXP
TIME: 40.00 40.00 40.00
BEAM: 30.01 30.01 30.01
ELEM: Ti Nb O SUM
210 -.008 97.524 1.075 98.591
211 -.007 98.250 1.155 99.398
212 -.001 97.537 1.140 98.676
213 -.022 97.965 1.140 99.083
214 .025 97.645 1.119 98.789
215 .004 98.427 1.100 99.531
216 .059 98.075 1.179 99.312
217 .021 98.293 1.172 99.485
218 -.005 97.582 1.107 98.684
219 .041 97.455 1.207 98.703
220 .003 97.516 1.137 98.657
221 -.012 97.991 1.196 99.175
222 -.001 97.570 1.147 98.716
223 .029 97.826 1.073 98.927
224 .018 97.777 1.196 98.991
225 .033 98.456 1.199 99.688
You know what this means? It means we should all change our default carbon density in our probewin.ini file to 1.35!
This keyword is found in the [standards] section of the probewin.ini file (usually C:\ProgramData\Probe Software\Probe for EPMA) as seen here:
StandardCoatingFlag=1 ; 0 = not coated, 1 = coated
StandardCoatingElement=6 ; assume carbon
StandardCoatingDensity=1.35
StandardCoatingThickness=200 ; in angstroms
I'm going to have to change all the default coating density parameters in my example config files!
Just to throw another spanner in the works: I'm sure we've all seen ring-shaped carbon contamination formed around a static beam spot, but did you know the centre of the ring can exhibit net carbon erosion? It doesn't even need particularly aggressive analysis conditions: I've measured 1.1nm loss over 60s with a 25nA FEG beam at 5kV. When the starting thickness is 5nm this represents a serious proportional loss. PfE's TDI analysis mode copes admirably well with correcting the changing emitted intensities.
Quote from: Mike Matthews on May 08, 2018, 11:37:46 AM
Just to throw another spanner in the works: I'm sure we've all seen ring-shaped carbon contamination formed around a static beam spot, but did you know the centre of the ring can exhibit net carbon erosion? It doesn't even need particularly aggressive analysis conditions: I've measured 1.1nm loss over 60s with a 25nA FEG beam at 5kV. When the starting thickness is 5nm this represents a serious proportional loss. PfE's TDI analysis mode copes admirably well with correcting the changing emitted intensities.
Hi Mike,
I think I have observed this effect also, but not really knowing what I was seeing:
http://smf.probesoftware.com/index.php?topic=48.msg160#msg160
See the attachment in the post linked above, and notice the slight decrease in carbon intensity for the first few points (at least until the electron beam starts encroaching on the deposited carbon "ring").
john
Off topic slightly, but carbon related all the same - anyone know if peaksight takes any account of carbon coatings? I am not sure - my doubts are raised by the fact you can't specify its presence or absence for an unknown.....