Improving EDS Analysis

[John Donovan]

This board is for general EDS issues, but for the first post I'd like to make a radical suggestion for improving EDS analysis:

Simply eliminate EDS "standardless" analysis!

Too radical for you?  Let's explore this question...

1. Dale Newbury at NIST estimates that 95-98% of all EDS analyses are performed "standardless".

2. Dale Newbury has demonstrated in the past that, although "standardless" EDS can be fairly accurate for some situations (e.g., major and minor elements in steel alloys), there are a very large number of other materials which perform much worse. New comparison studies of "standardless" by NIST are forthcoming, but initial results reported by Dale Newbury suggest that things have *not* improved much if at all.

3.  Dale Newbury and Nicholas Ritchie have demonstrated that in many situation (though not all!), that EDS when performed carefully with standards can approach and often equal WDS analysis. But again, this is only being performed some 5% or less of the time!

4. At the risk of "tooting my own horn" I should mention that EDS analysis in Probe for EPMA can *only* be performed using full standards where the net intensity from the standard is utilized with the unknown net intensity to create an actual k-ratio, which is then cranked through the matrix correction along with the WDS k-ratios. This is very easy because a full EDS spectra is stored with each WDS analysis for all samples (standards and unknowns) and these are automatically acquired using stage automation.

What am I suggesting? Am I suggesting that every EDS user use standards every time?   In a word, yes.  We know it provides the best data by far and heck, EPMA users *always* use standards!

Does that mean that every SEM lab must obtain suitable standards?  Again, yes.

Is this a bad thing?  No.

If every EDS lab needed standards, we would have much more resources and effort applied to obtaining excellent standards as described in more detail here:

http://probesoftware.com/smf/index.php?topic=301.0

What do you all think?

[Les Moore]

Hi John,

People using EDS for quant alloy analysis in steels need their head read!  The alloy levels are typically around the 1% mark and unless you want to count for ages, you need to bump up the countrate which runs you foul of the iron peak which you are not in the least bit interested. 

WDS all the way for steel for me. 

It annoys me when people use the EPMA and acquire EDS maps that look absolute rubbish.  Bump up the current way beyond what the EDS can cope with and look only at the lines you want.

Having said that, EDS is great for initial sample screening - finding what is there.  It is also good (data cube) when finding what you may have missed.

[Probeman]

It annoys me when people use the EPMA and acquire EDS maps that look absolute rubbish.  Bump up the current way beyond what the EDS can cope with and look only at the lines you want.

Having said that, EDS is great for initial sample screening - finding what is there.  It is also good (data cube) when finding what you may have missed.

Hi Les,
I couldn't agree more. In fact it is a similar situation for characterization of thin films (and nano-particles for that matter). See the attached PPT file on using WDS vs. EDS for thin film characterization.

I'll also mention that because we can so easily quant WDS maps using CalcImage, I now get an occasional email saying something along the lines of: "john, I quantified my x-ray maps, but how come my WDS map totals go from 90 to 110 %?"

And I usually then ask: "What pixel dwell time were you using?" and they'll usually respond with "oh, 20 msec".  So then I say: "Now try again, but this time using 200 msec per pixel!" 

[Les Moore]

In fact it is a similar situation for characterization of thin films (and nano-particles for that matter).

Oh yes!
Last year I was doing quant mapping of <100nm SiO2 layers on substrates using 5kV and mapping to see "holidays" or gaps in the coating.  Summing a TAP, and two PETs at 120msec/pixel.  The sample wouldn't take more than 5nA.  Lucky Si Ka is so close to the pole piece on TAP and the solid angle of collection is so high. The two PETs didn't contribute much but I took all they had to offer.

Quote from the SEM (EDS) operator, "I've never detected the layers".

Another job involved mapping trace levels on a sample that "Is always homogeneous when we analyse it with SEM EDS".  I soon put an end to that nonsense... it was riddled with beautiful ring structures like a tree, of trace level Cr an V. The conversation went something like "You mean we will only see this information when we use your machine".  After an answer in the affirmative, I never saw them again.  Ignorance is bliss 
 

[jon_wade]

Does that mean that every SEM lab must obtain suitable standards?  Again, yes.

Is this a bad thing?  No.

What do you all think?

the other end of this riddle is...

"why don't analytical SEMs come with a faraday cup? "

I know of at least two top spec SEMs bought recently that don't (and can't) have one installed.
it does make me wonder how these things get specified.

[Les Moore]

John,

In a recent paper by Carpenter et al, he finished with standardless EDS on a faceted galena grain (I recall but cannot find it to be sure).

The variation in the analysis as a function of grain orientation is enough to forever banish stdless EDS analysis of non flat surfaces (more correctly - unknown orientation) to the technological culdesac that contains the Betamax video recorder.  

Les

[Probeman]

Here is a slide from a presentation that Dale Newbury sent me on their most recent effort evaluating standardless EDS accuracy:



His complete PPT file is attached below (remember you have to be logged in as a member to see attachments). And by the way, I completely agree that EDS can often be as accurate as WDS (though of course not as sensitive as WDS!)- if EDS people were as careful with their EDS measurements as WDS people are with their WDS measurements- but that is a big "if".

I will repeat here what I said at the conclusion of my 2011 M&M 60 years After Castaing session talk: the one advantage that WDS will always have over EDS, is that one *cannot* perform standardless WDS!

[qEd]

Quantitative EDS with standards can return solid results. THis example is an old one as it was published in EPSL in 1998 164:421- and involves three hydrous REE phases. The analyses were good enough to estimate the water content in a 2 samples of florencite–goyazite and rhabdophane.

Measured  florencite–goyazite:
(Ce0.13La0.1Nd0.05Sm0.01)(Al0.97Fe0.03)3((PO4)0.86(SO4)0.14)2(OH)6–(Sr0.41Ba0.21Ca0.1)(Al0.97Fe0.03)3(2(PO4)0.864(SO4)0.14)0.5(PO3OH)(OH)6.
100- total wt%=11.6%
Calculated  florencite–goyazite:
H2O calc from ideal Ce0.4Al3(PO4)2OH6–(Sr0.4Ba0.2)Al3(PO4)(PO3OH)(OH)6. = 12.1%

Measured rhabdophane:
 (Ce0.47La0.32Nd0.21)(PO4)H2O
100- total wt%=6.8%
Calculated  rhabdophane:
H2O calc from ideal (Ce0.3La0.3Nd0.3)(PO4)H2O = 7.1 %

[Nicholas Ritchie]

The Forensics Chemistry journal isn't necessarily on your radar but there is a paper in there https://www.sciencedirect.com/science/article/pii/S2468170920300400?via%3Dihub that I published recently with some GSR colleagues that is relevant to this topic.  It basically discusses validation procedures to ensure that an SEM/EDS system is performing and QC procedures to check and document the performance.

NIST paid the tariff so it is available for free under the Creative Commons license.