ZAF, Phi (pz) and characteristic fluorescence selections

[Ben Buse]

Sorry I'm going to be picky here - its labelled ZAF, phi(pz), but none of them are ZAF, are they not all phi-rho-z?

Ben

[Probeman]

Sorry I'm going to be picky here - its labelled ZAF, phi(pz), but none of them are ZAF, are they not all phi-rho-z?

Ben

Hi Ben,
Note that it says ZAF *comma* Phi (pz) *and* characteristic fluorescence corrections.

It's just saying there are several ZAF correction methods, several phi-rho-z correction methods and also the fluorescence correction options.



What would you prefer it to say?

[Ben Buse]

Thanks John,

Thats helpful, I did not spot some of them use ZAF method for absorption corrections (as oppose to integrating depth slices of phi(rz) curve).

I don't know what I'd prefer to say, I guess the key distinction is between phi rho z, phi rho z and ZAF combo's and true ZAF (such as offered as standard by JEOL)

(Its actually phi(rz), A & F selections - but that would be difficult to understand!). "Matrix corrections"?

Ben

[Probeman]

Thanks John,

Thats helpful, I did not spot some of them use ZAF method for absorption corrections (as oppose to integrating depth slices of phi(rz) curve).

I don't know what I'd prefer to say, I guess the key distinction is between phi rho z, phi rho z and ZAF combo's and true ZAF (such as offered as standard by JEOL)

(Its actually phi(rz), A & F selections - but that would be difficult to understand!). "Matrix corrections"?

Ben

Hi Ben,
What the heck is a "true ZAF"?  All of these ZAF methods are "true ZAFs" as far as I know!  Sounds like a "no true Scotsman" to me!   

https://en.wikipedia.org/wiki/No_true_Scotsman

The only problem with labeling the dialog as "Matrix Corrections" is that these are only a portion of the matrix corrections offered in CalcZAF/Probe for EPMA/CalcImage as seen in the dialog just prior to the one we're talking about:



I would ague that alpha factor methods based on the analytical corrections or the fast Monte Carlo binaries from Penepma k-ratios or the multi-standard calibration curves are all "matrix corrections" of one kind or another.

Sorry if I'm missing something...

[Ben Buse]

Your right!

But there is a distinguish between phi-rho-z (exponential function to experimental measurements), and the original ZAF equations, we know the phi-rho-z models perform significantly better.

https://probesoftware.com/smf/index.php?topic=890.msg5683#msg5683

 The trouble is it gets confusing as phi-rho-z is converted into Z and A factors for application. I guess some things are difficult to simplify for teaching, etc.

I regret the original question (should think before posting!), I don't think the name is the problem, but how to convey the essential information.
Things like the difference between the phi rho z models (gaussian, parabolic fit etc) or their performance on PAP dataset.


Jacky Ruste (GN-MEBA)
http://micro.icaunais.free.fr

[Probeman]

I certainly agree that ZAF and phi-rho-z methods are different!   

In fact the only reason for including the traditional ZAF methods in CalcZAF and Probe for EPMA is for historical comparisons. We can get the same (original) results with the Shaw data set from the 1970's using these ZAF methods!  By the way, I think the Shaw dataset is included with the CalcZAF installer.  Here is an email discussion with Paul Carpenter on this that I posted here with his permission some time ago:

https://probesoftware.com/smf/index.php?topic=924.msg5944#msg5944

As you said, my preferred explanation to the students is to say that phi-rho-z methods combine the Z and A effects, and for that I utilize some graphs.  Some of my, now quite ancient PPT slides and lecture notes, are found here:

https://epmalab.uoregon.edu/lecture.htm

The matrix correction lecture is here:

https://epmalab.uoregon.edu/pdfs/Electron%20Beam%20MicroAnalysis-Geol_619-12.ppt

john

[Probeman]

Here's another technique I use with students. Calculate compositions for measurements using *all* the different matrix corrections. Here's an example using one of the NIST Au-Cu (20/80) standards:

Summary of All Calculated (averaged) Matrix Corrections:
AUCU_NBS-K-RATIOS.DAT, Sample 1
LINEMU   Henke (LBL, 1985) < 10KeV / CITZMU > 10KeV

Elemental Weight Percents:
ELEM:       Cu      Au   TOTAL
     1  20.079  80.283 100.362   Armstrong/Love Scott (default)
     2  19.473  81.423 100.895   Conventional Philibert/Duncumb-Reed
     3  20.358  80.265 100.624   Heinrich/Duncumb-Reed
     4  19.820  80.247 100.066   Love-Scott I
     5  19.908  80.380 100.288   Love-Scott II
     6  18.772  80.589  99.361   Packwood Phi(pz) (EPQ-91)
     7  21.323  79.946 101.269   Bastin (original) Phi(pz)
     8  19.589  80.142  99.731   Bastin PROZA Phi(pz) (EPQ-91)
     9  19.607  80.119  99.726   Pouchou and Pichoir-Full (PAP)
    10  19.738  80.136  99.874   Pouchou and Pichoir-Simplified (XPP)

AVER:   19.867  80.353 100.220
SDEV:     .661    .413    .588
SERR:     .209    .131

MIN:    18.772  79.946  99.361
MAX:    21.323  81.423 101.269

This is output from CalcZAF, but one can also do it from Probe for EPMA on their own measurements. Then ask the students: which one is "correct"?     

Of course the classic example is the SiIr alloy from Armstrong that Paul Carpenter worked on:

Summary of All Calculated (averaged) Matrix Corrections:
#1  approx. Ir45Si55 (atomic) based on Ir3Si5 standard
LINEMU   Henke (LBL, 1985) < 10KeV / CITZMU > 10KeV

Elemental Weight Percents:
ELEM:       Ir      Si   TOTAL
     1  84.502  14.308  98.810   Armstrong/Love Scott (default)
     2  86.715  11.781  98.496   Conventional Philibert/Duncumb-Reed
     3  84.863  12.910  97.773   Heinrich/Duncumb-Reed
     4  84.339  12.819  97.158   Love-Scott I
     5  84.329  12.833  97.162   Love-Scott II
     6  84.208  11.075  95.283   Packwood Phi(pz) (EPQ-91)
     7  86.129  13.997 100.126   Bastin (original) Phi(pz)
     8  84.658  12.708  97.366   Bastin PROZA Phi(pz) (EPQ-91)
     9  84.531  12.836  97.367   Pouchou and Pichoir-Full (PAP)
    10  84.295  12.633  96.929   Pouchou and Pichoir-Simplified (XPP)

AVER:   84.857  12.790  97.647
SDEV:     .858    .929   1.290
SERR:     .271    .294

MIN:    84.208  11.075  95.283
MAX:    86.715  14.308 100.126

Atomic Percents:
ELEM:       Ir      Si   TOTAL
     1  46.325  53.675 100.000   Armstrong/Love Scott (default)
     2  51.820  48.180 100.000   Conventional Philibert/Duncumb-Reed
     3  48.995  51.005 100.000   Heinrich/Duncumb-Reed
     4  49.016  50.984 100.000   Love-Scott I
     5  48.986  51.014 100.000   Love-Scott II
     6  52.630  47.370 100.000   Packwood Phi(pz) (EPQ-91)
     7  47.347  52.653 100.000   Bastin (original) Phi(pz)
     8  49.328  50.672 100.000   Bastin PROZA Phi(pz) (EPQ-91)
     9  49.040  50.960 100.000   Pouchou and Pichoir-Full (PAP)
    10  49.368  50.632 100.000   Pouchou and Pichoir-Simplified (XPP)

AVER:   49.285  50.715 100.000
SDEV:    1.839   1.839    .000
SERR:     .582    .582

MIN:    46.325  47.370 100.000
MAX:    52.630  53.675 100.000

The customer's question was is it the 50:50 atomic phase or the 45:55 atomic phase?