Carbon coating thickness

[Alejandro Cortes]

Hi everyone,

I am trying to reach an agreement here with other labs regarding the c-coating of samples. While I am very used to the typical 20 nm for typical 15-20 kV analyses on the probe. People using EDS-SEM systems like to use 10 nm because they analyse at lower kV (ca. 6kV) and sometimes they tend to analyse PGE so apparently thicker coatings are a problem (?). I would love to hear your opinion on this.

I run some casino simulations to show the effect on the intensity of K lines when reducing the C thickness, getting lower intensities at lower thickness. So I guess this will have an issue on determining elements at low concentrations without changing the current. I know this will also have an effect on the absorption correction that PfE deals with.

[Probeman]

This doesn't seem correct.  I wonder if they actually tried to model this themselves.  I also don't understand why Casino would have produced a lower intensity for a thinner carbon coating.  That doesn't make sense.

If you run CalcZAF and select the Run | Calculate Electron and X-Ray ranges menu dialog.  You can specify Pure carbon, set the density, specify the thickness and calculate both the x-ray transmission efficiency and the electron energy loss. Yes, it's not as accurate as a MC simulation, but should be more than accurate enough (and a lot less subject to statistical variation!).  And quicker!

So here are results for a 15 keV beam with 20 nm (0.02 um) of carbon and a 4 keV x-ray:
Electron energy transmitted at incident electron energy of 15 keV, 2.26 grams/cm^3, thickness of 0.02 um = 14.99345 keV
X-ray transmission fraction at energy 4 keV, thickness of 0.02 um (average u/p = 37.26325) = 0.9998316

And here for a 2 keV x-ray:
X-ray transmission fraction at energy 2 keV, thickness of 0.02 um (average u/p = 307.1905) = 0.9986125

Why would someone think a loss of 0.14% is worth worrying about?

Trying the same but at 6 keV electron beam energy and 4 keV x-ray we get:
Electron energy transmitted at incident electron energy of 6 keV, 2.26 grams/cm^3, thickness of 0.02 um = 5.987882 keV

And finally let's try a 10 nm carbon coat with a 2 keV x-ray:
X-ray transmission fraction at energy 2 keV, thickness of 0.01 um (average u/p = 307.1905) = 0.999306

Is a fractional percent improvement in transmission really worth all that trouble?  For another thing, as you mentioned, technically one should specify the different thicknesses in carbon coating for standards vs. unknowns for the k-ratio as described here:

https://probesoftware.com/smf/index.php?topic=23.msg8997#msg8997

In the example linked above the user had accidentally coated his unknown thin sections with 70 nm of carbon, and yeah, that does make a difference in this case because the overvoltage of Fe ka is quite low even at 15 keV (Eo/Ec ~ 2) and even a small reduction in electron energy is sufficient to affect the ionization efficiency! 

[Ben Buse]

When I've done low voltage or high voltage work - I've not worried about absolute thickness. But at low voltage quantification it's critical to have the same thickness on the unknowns and the standards. I can't see that it matters hugely whether 10 or 20nm as long as it's the same. Can't see why PGE particularly problematic they are just L and M lines around 2-3keV. Slightly thicker coating expect slightly lower intensity, but need to ensure thick enough that good conductive path, which will partly depend on amount of relief

[Probeman]

When I've done low voltage or high voltage work - I've not worried about absolute thickness. But at low voltage quantification it's critical to have the same thickness on the unknowns and the standards. I can't see that it matters hugely whether 10 or 20nm as long as it's the same. Can't see why PGE particularly problematic they are just L and M lines around 2-3keV. Slightly thicker coating expect slightly lower intensity, but need to ensure thick enough that good conductive path, which will partly depend on amount of relief

I agree with Ben but would point out that we should pay attention to the differences between low voltage and low over voltage.   One can be at low voltage and still not have a low over voltage situation with sufficiently low energy emission lines.

And yes, having the same carbon coat on both standards and unknowns is much easier than knowing how they differ.   All that said, Ben is correct, it shouldn't matter whether the carbon coating is 10 or 20 nm, just make sure it's the same by coating them together.

As for PGM, the M edges are all around 3 keV, and since one would prefer to have an over voltage of say 3 times, one should use an electron beam energy of around 9 keV, so I think your SEM people haven't really thought this all through.  But maybe they are trying to limit the interaction volumes by using low voltage analyses?  But again, 20 nm would be fine as long as it's the same on both standards and unknowns.

[Ben Buse]

Looking at DTSAII for Pt Ma at 6kV and 20kV and carbon coat 10nm and 20nm gives following - full results attached.

Transition
Generated 1/msR
Emitted 1/msR
Ratio   (%)

6kV 10nm
Pt M5-N7
5,751.5
5,382.0
93.6%

6kV 20nm
Pt M5-N7
5,468.3
5,128.2
93.8%

20kV 10nm
Pt M5-N7
46,133.0
29,948.6
64.9%

20kV 20nm
Pt M5-N7
44,566.5
28,766.3
64.5%

[Probeman]

Yes, good to confirm that the carbon thickness (10 vs. 20 nm) isn't relevant.

And yes, these are fairly software x-rays, so 20 keV would not be appropriate.  That is why I suggested 9 keV.

Here are pr(z) curves (in um depth) for 15, 9 and 6 keV:







To be honest I think 6 or 9 keV beam energies would be fine, but as mentioned previously, the carbon coat thickness isn't going to be an issue as long as they are the same for both standards and unknowns.