Standard coating material - carbon v iridium

[emma_fisi]

Hi everyone,

We will soon be purchasing/making a couple of new standard blocks for our probe. My question to the group is about whether people have a strong opinion as to what standards should be coated with for use in the probe - carbon, or iridium?

In the past, all of the standards I have used have been carbon coated - in my previous life, this was the best (and only) option really. In my new world however, we have several users who are interested in measuring carbon (using a variety of techniques, not just the probe), and so would prefer to use iridium. Our Ir coater is also much faster, and more reliable, than our carbon coater.

Is there any strong reason for sticking with the carbon coat? Have other labs out there moved to using other materials to coat their standards? Am I being hopelessly clueless in even suggesting a move away from carbon to something else?! Are there downsides to using an Ir coat on samples and standards?

Any help/advice/pointers/suggestions are greatly appreciated!
Cheers,
Em

[Anette von der Handt]

Hi Emma,

from what I learned so far and talking to more experienced users of "non-traditional" coating materials I am ready to make the plunge (after careful vetting of our new coater, see below).

Besides analysis of carbon itself, carbon has also very high absorbance of O ka, N ka, so there are various other applications for non-carbon coatings. I think the absolutely crucial part for all metal coatings will be how thin as well as how accurate and reproducible you can lay down the coating, especially if you don't always want to coat the standards together with the samples. There is much, much less room for error relative to carbon coatings.

Also, as far as I understand, the attraction of iridium specifically for a lot of people is that there is a specific thickness (0.6 nm??) that gives you the same absorbance as carbon so you could mix and match carbon and iridium coating theoretically (PFE also has a coating thickness correction, btw).

You might want to look further into the work of Kat Crispin and John Armstrong about this or contact Kat. From their abstracts at M&M 2013:

"Results of this study indicate that an Ir-coat of 0.4 nm thickness is sufficient to provide sufficient conductivity on an insulating sample for analysis; although for samples containing beam-sensitive materials, a minimum thickness of 1 nm is necessary to avoid element migration (Table 1). Coatings of less than 2 nm require minimal or no additional offline corrections, making this an ideal substitute for traditional carbon coatings and require no special preparations of standards (i.e. uncoated or carbon coated standards can be directly compared). However, the K-lines of boron to oxygen and the L-lines of calcium to chromium are heavily absorbed by iridium, requiring correction for Ir thicknesses over 1 nm."
(Armstrong & Crispin, 2013, Ultra-thin Iridium as a Replacement Coating for Carbon in High Resolution Quantitative Analyses of Insulating Specimens. Microanal. 19 (Suppl 2), 1070)

[aburnham]

The attached paper makes the case for using Os coatings. Would be interesting to hear if anyone has experience of using this technique.

[Probeman]

The attached paper makes the case for using Os coatings. Would be interesting to hear if anyone has experience of using this technique.

I read the paper and I was a little disappointed.

Yes, there is no doubt that a very thin Os (or Ir) coating could be useful for many situations, particularly when beam sensitive samples are involved due to the apparently greater thermal conduction (though someone should do a calculation comparing the thermal conductivity of 20 nm of carbon vs. 3 nm of Os.

On the issue of the greater absorption of oxygen emission in carbon relative to osmium that is true enough as far as the mass absorption coefficients are concerned (though barely) as seen here in this output from the CalcZAF | X-Ray menu:

"Table of MACs (mass absorption coefficients) from C:\ProgramData\Probe Software\Probe for EPMA\LINEMU.DAT"
"Emitting Element: O "
Absorber         ka        
      C    12390.14                                                          
      Os   11236.48                                                          

But if we do a proper calculation including the densities and thicknesses, the absorption is actually a bit less for carbon than osmium as seen here from the CalcZAF Run | Calculate Electron and X-ray Ranges menu:

O  ka, x-ray transmission fraction through thickness 0.02 um of carbon (2.26 density) (average u/p = 12390.14) = 0.9455359

O  ka, x-ray transmission fraction through thickness 0.003 um of osmium (22.4 density) (average u/p = 11236.48) = 0.9272712

In fact they have merely demonstrated that they didn't measure the thickness of their carbon coatings with sufficient accuracy.  The observed color of a coated slide or the oil drop on a white tile may be a common method for estimating thickness, but it's a pretty poor method for accuracy.

There are two better solutions for this. First simply use polished brass as your carbon thickness monitor, as the color change from 15 to 20 nm is quite striking (red to blue) and therefore very reproducible (I generally shoot for a violet color for best reproducibility).  Or second, and even better, simply carbon coat your standards and unknowns at the same time!   You can't get more reproducible than that.  This is what we do when we need the best accuracy for oxygen measurements.
john

[Probeman]

And then I thought: the above MAC values are from Henke/Heinrich, so I'd better check a more recent table, and using the FFAST table from NIST I get:

"Table of MACs (mass absorption coefficients) from C:\ProgramData\Probe Software\Probe for EPMA\FFAST.DAT"
"Emitting Element: O "
Absorber         ka         kb         la         lb         ma         mb   
      C    11705.00                                                          
      Os    9919.23                                                          

So the MAC for O ka in carbon is about the same but the MAC for O Ka in osmium is a bit lower, so running the calculation again we get:

O  ka, x-ray transmission fraction through thickness 0.02 um of carbon (2.26 density) (average u/p = 11705) = 0.9484686

O  ka, x-ray transmission fraction through thickness 0.003 um osmium (22.4 density) (average u/p = 9919.228) = 0.9355158

So about the same, but still slightly better transmission for 20 nm of carbon.
john

[Ben Buse]

so running the calculation again we get:

O  ka, x-ray transmission fraction through thickness 0.02 um of carbon (2.26 density) (average u/p = 11705) = 0.9484686

O  ka, x-ray transmission fraction through thickness 0.003 um osmium (22.4 density) (average u/p = 9919.228) = 0.9355158

So about the same, but still slightly better transmission for 20 nm of carbon.
john

Hi John,

I haven't read the osmium paper yet. But when considering intensity loss by coat - you also need to consider the energy loss of beam electrons passing through the coat. i.e. MAC is not everything. High Z materials can have same MAC but reduce the energy of beam electrons by a greater extent for same thickness. - Thereby reducing counts/transmission. This is most severe for low overvoltages. Reed had a formula for this in his book, and you can see the same thing in monte carlo modeling results

Ben

[Probeman]

so running the calculation again we get:

O  ka, x-ray transmission fraction through thickness 0.02 um of carbon (2.26 density) (average u/p = 11705) = 0.9484686

O  ka, x-ray transmission fraction through thickness 0.003 um osmium (22.4 density) (average u/p = 9919.228) = 0.9355158

So about the same, but still slightly better transmission for 20 nm of carbon.
john

Hi John,

I haven't read the osmium paper yet. But when considering intensity loss by coat - you also need to consider the energy loss of beam electrons passing through the coat. i.e. MAC is not everything. High Z materials can have same MAC but reduce the energy of beam electrons by a greater extent for same thickness. - Thereby reducing counts/transmission. This is most severe for low overvoltages. Reed had a formula for this in his book, and you can see the same thing in monte carlo modeling results

Ben

Hi Ben,
That is absolutely correct.  But if you do read the paper you will see that they primarily discuss the absorption of low energy x-rays such as oxygen Ka by the coating. 

As you say, for low overvoltage situations, the energy loss of incident electrons by the coating needs to be taken into account, though the paper does not discuss issue this if I remember correctly. 

Fortunately CalcZAF also includes this electron energy loss calculation!  Here is the calculation for 20 nm of carbon and 3 nm of Os:

Electron energy transmitted at incident electron energy of 15 keV, 2.26 grams/cm^3, thickness of 0.02 um = 14.99345 keV
Electron energy transmitted at incident electron energy of 15 keV, 22.4 grams/cm^3, thickness of 0.003 um = 14.9941 keV

So, there is a slight benefit for the 3 nm of Os, in the 5th decimal place!   

But not for 4 nm of Os!

Electron energy transmitted at incident electron energy of 15 keV, 22.4 grams/cm^3, thickness of 0.004 um = 14.99214 keV   

john

[Probeman]

Note that these x-ray and electron range calculations available in CalcZAF from the Run menu are quite crude and not strictly quantitative (compared to the results from the Penepma modeling GUI in Standard), but they are useful for quick "back of the envelope" calculations:

[jon_wade]

Osmium is not something I would routinely want in the lab - in particulate form (and, I would guess, thin coatings) it forms osmium tetroxide which is extremely toxic via skin transmission and also volatile.  I think is best to make EPMA labs hide beneath the H&S radar if at all possible!

[knshugart]

Hello,

Em- I'm curious what your lab decided to go with. We've been having the same debate of C vs Ir for our newest standards.

If you did go with Ir, has it worked well for you so far?

Thanks,
Kathleen

[emma_fisi]

Hi Kathleen,

We've just ordered our new standards - I think we will stick with carbon for the new blocks, as an informal poll of my users seems to indicate that is what most of them want to coat their samples with. I'll keep the old mineral standard blocks and Ir coat them though, so they can go into the probe if necessary, for those people who insist on trying to analyse for carbon....

Cheers,
Em