matrix correction comparison

[Ben Buse]

Hi,

Does anyone know of a paper that compares Armstrong matrix correction with PAP and XPHI. In Pouchou & Pichoir (1991) (Heinrich green book) - we have 826 dataset - and is tested on following



and Merlet (1994) tried his XPHI model



But none compare Armstrong

I notice in Calczaf you can do histogram using Pouchouz10 - but this has 91 analyses - what is this?

I'm guessing Armstrong is good for geological samples - as it was developed and tested for them Armstrong (1988) but maybe not as good across the whole range??

Is it possible to change the default matrix correction in PFE.

Thanks

Ben

[Ben Buse]

Following on from my post.

I see you can calculate binaries for pouchouz10 - does this correspond to the 577 analyses z dominant?
pouchouza20 - does this correspond to the 242 analyses absorption dominant.

I've tried pouchou.dat this has 756 analyses how does this compare to 826 analyses of Pouchou & Pichoir (1991)

 and pouchouz2.dat - I just tried this is the 826 dataset


Thanks

Ben

[Ben Buse]

Right so here is Armstrong and FFAST



And here is PAP and FFAST



And PAP with MAC30  - as in table

[Probeman]

Does anyone know of a paper that compares Armstrong matrix correction with PAP and XPHI. In Pouchou & Pichoir (1991) (Heinrich green book) - we have 826 dataset - and is tested on following

and Merlet (1994) tried his XPHI model

But none compare Armstrong

I notice in calczaf you can do histogram using Pouchouz10 - but this has 91 analyses - what is this?

I'm guessing Armstrong is good for geological samples - as it was developed and tested for them Armstrong (1988) but maybe not as good across the whole range??

Hi Ben,
I see Owen answered your question on changing the default correction in PFE and CalcZAF from the probewin.ini file.

Paul Carpenter can provide the exact details but here is what I remember:  the Armstrong phi/rho/z was only published in this paper:

Armstrong, J. T. "Quantitative analysis of silicate and oxide minerals: a reevaluation of ZAF corrections and proposal for new Bence-Albee coefficients." Microbeam Analysis 19 (1984): 208-212.

but it was just a tweak of two existing corrections (Brown and another I think), where John Armstrong noticed that with his silicate (Shaw) k-ratio dataset (see shaw.dat attached below), he found that one phi/rho/z gave slightly "high" results and another gave slightly "low" results, and then he noticed that the only difference between the two corrections was an exponent in one term. So Armstrong "split the difference" with the two exponent values and got good results, at least with his shaw k-ratio data set.

Now you should be careful using any correction method which was "tuned" to a particular dataset, and that goes for the Pouchou k-ratio dataset as well.  Because Pouchou tuned the XPP and PAP corrections to the Pouchou dataset, (using the Bastin MAcs) you will always get the best results on the Pouchou dataset using the XPP or PAP corrections.

Best to try several different datasets, see attached below.  The Pouchou.dat dataset is the original dataset I got from Armstrong many years ago. The Pouchou2.dat dataset is an "extended" version I got from Philippe Pinard and Hendrix Demers.  The POUCHOU2_wo_B_wo_Cu_La.DAT dataset is the Pouchou2.dat dataset *without* the boron and Cu L k-ratios which are affected by chemical shifts.
john

PS Yes, the Pouchouz10.dat dataset is a subset of the Pouchou dataset where the atomic number correction (Z) is greater than 10%.  These types of subset output can be generated in CalcZAF using the filter options in the Analytical | Binary Calculations Options menu dialog as shown here:



The NISTBIN and Pouchou.dat datasets should be included with the normal CalcZAF installation, but maybe not the new Pouchou2.dat.

[Ben Buse]

Thanks Owen and John,

That's a great help. So I guess not using Bastin MAC's explains the slight discrepancy from the published table.

Ben

[Probeman]

So I guess not using Bastin MAC's explains the slight discrepancy from the published table.

Yes.   But I don't have these Bastin MACs, so if anyone does I'd be interested in adding them as a table in CalcZAF.
john

[Ben Buse]

So I've just been looking up the references,

And NISTBIN contains:
(1) the Au-Cu and Au-Ag alloys of Kurt F. J. Heinrich and Harvey Yakowitz "Absorption of Primary X Rays in Electron Probe Microanalysis"
(2) Mg-Al alloys of J. I. GOLDSTEIN F. J. MAJESKE H. YAKOWITZ  PREPARATION OF ELECTRON PROBE MICROANALYZER STANDARDS USING A RAPID QUENCH METHOD
(3) some other stuff. - Including Boron.

Thanks John, I've just downloaded the Shaw file - I guess this corresponds to
Shaw, H.F. & Albee, A.L. (1979)
An empirical investigation into possible nonlinearities of the microprobe correction factors in the system MgO-CaO-Al2O3-SiO2. In: Microbeam Analysis - 1979, Newbury, D.E. (ed), San Francisco Press, San Francisco, 227 - 230.

John - just a question about Shaw.dat - this is a regular input file for calczaf - so I process the regular way (Open input data and calculate and export all) - what are standards 2112,2113,2114,2120 - are these pure metals? Also when I quantify them what are the correct values.

Thanks

Ben

[Brian Joy]

So I guess not using Bastin MAC's explains the slight discrepancy from the published table.

Yes.   But I don't have these Bastin MACs, so if anyone does I'd be interested in adding them as a table in CalcZAF.
john

The tables of MACs that Pouchou and Pichoir used in place of the Heinrich MAC30 MACs are located on p. 63 of EPQ, or on p. 33 of the attached pdf.

[Probeman]

So I guess not using Bastin MAC's explains the slight discrepancy from the published table.

Yes.   But I don't have these Bastin MACs, so if anyone does I'd be interested in adding them as a table in CalcZAF.
john

The tables of MACs that Pouchou and Pichoir used in place of the Heinrich MAC30 MACs are located on p. 63 of EPQ, or on p. 33 of the attached pdf.

Hi Brian,
Thanks, but I should have been clearer.  Yes, I knew they are in the "green book", but I was hoping to find a *digital* version of them!

Does anyone have that?
john

[Probeman]

Thanks John, I've just downloaded the Shaw file - I guess this corresponds to
Shaw, H.F. & Albee, A.L. (1979)
An empirical investigation into possible nonlinearities of the microprobe correction factors in the system MgO-CaO-Al2O3-SiO2. In: Microbeam Analysis - 1979, Newbury, D.E. (ed), San Francisco Press, San Francisco, 227 - 230.

John - just a question about Shaw.dat - this is a regular input file for calczaf - so I process the regular way (Open input data and calculate and export all) - what are standards 2112,2113,2114,2120 - are these pure metals? Also when I quantify them what are the correct values.

Hi Ben,
Sorry I should have explained better.  The standards are simply the pure oxides, MgO, Al2O3, SiO2 and CaO.  Not sure how they measured a pure CaO standard!

I've attached my current standard.mdb file which has them listed.  Note: previous version of this database had them listed as pure metals, this was incorrect.  They are pure oxides.
john

[Brian Joy]

So I guess not using Bastin MAC's explains the slight discrepancy from the published table.

Yes.   But I don't have these Bastin MACs, so if anyone does I'd be interested in adding them as a table in CalcZAF.
john

The tables of MACs that Pouchou and Pichoir used in place of the Heinrich MAC30 MACs are located on p. 63 of EPQ, or on p. 33 of the attached pdf.

Hi Brian,
Thanks, but I should have been clearer.  Yes, I knew they are in the "green book", but I was hoping to find a *digital* version of them!

Does anyone have that?
john

Hi John and Ben,

I've attached a text file with the MACs that Pouchou and Pichoir used for X-rays of the light elements; it should be entirely consistent with their paper in EPQ.

[Probeman]

So I guess not using Bastin MAC's explains the slight discrepancy from the published table.

Yes.   But I don't have these Bastin MACs, so if anyone does I'd be interested in adding them as a table in CalcZAF.
john

The tables of MACs that Pouchou and Pichoir used in place of the Heinrich MAC30 MACs are located on p. 63 of EPQ, or on p. 33 of the attached pdf.

Hi Brian,
Thanks, but I should have been clearer.  Yes, I knew they are in the "green book", but I was hoping to find a *digital* version of them!

Does anyone have that?
john

Hi John and Ben,

I've attached a text file with the MACs that Pouchou and Pichoir used for X-rays of the light elements; it should be entirely consistent with their paper in EPQ.

Hi Brian,
That's very nice and useful, thanks.  I have essentially the same table already in the EMPMAC.DAT file that is distributed with CalcZAF (since these are empirical measurements from multiple keVs fitted using his XMAC code I believe). For example:

  "b"     "ka"    "b"       3400         "Bastin (1992)"
  "b"     "ka"    "b"       3500         "Pouchou (1998)"
  "b"     "ka"    "b"       3068         "Donovan (2011)"
  "b"     "ka"    "c"       6500         "Bastin (1992)"
  "b"     "ka"    "n"       11200        "Bastin (1992)"
  "b"     "ka"    "n"       11000        "Pouchou (1998)"
  "b"     "ka"    "n"       10421        "Donovan (2011)"
  "b"     "ka"    "mg"      59500        "Pouchou (1998)"
  "b"     "ka"    "mg"      54500        "Donovan (2011)"
  "b"     "ka"    "al"      64000        "Bastin (1992)"
  "b"     "ka"    "si"      84000        "Bastin (1992)"
  "b"     "ka"    "ti"      14700        "Bastin (1992)"
  "b"     "ka"    "v"       17700        "Bastin (1992)"
  "b"     "ka"    "cr"      20200        "Bastin (1992)"
  "b"     "ka"    "fe"      27300        "Bastin (1992)"
  "b"     "ka"    "co"      33400        "Bastin (1992)"
  "b"     "ka"    "ni"      42000        "Bastin (1992)"
  "b"     "ka"    "zr"      4000         "Bastin (1992)"
  "b"     "ka"    "nb"      4600        "Bastin (1992)"
  "b"     "ka"    "mo"      4550        "Bastin (1992)"
  "b"     "ka"    "la"      2500        "Bastin (1992)"
  "b"     "ka"    "ta"      22500        "Bastin (1992)"
  "b"     "ka"    "w"       21400        "Bastin (1992)"
  "b"     "ka"    "u"       8200        "Bastin (1992)"
  "b"     "ka"    "b"       3471         "Pouchou (1992)"
  "b"     "ka"    "c"       6750         "Pouchou (1992)"
  "b"     "ka"    "n"       11000        "Pouchou (1992)"
  "b"     "ka"    "o"       16500        "Pouchou (1992)"
  "b"     "ka"    "al"       64000        "Pouchou (1992)"
  "b"     "ka"    "si"       80000        "Pouchou (1992)"
  "b"     "ka"    "ti"       15000        "Pouchou (1992)"
  "b"     "ka"    "v"       18000        "Pouchou (1992)"
  "b"     "ka"    "cr"       20700        "Pouchou (1992)"
  "b"     "ka"    "fe"       27800        "Pouchou (1992)"
  "b"     "ka"    "co"       32000        "Pouchou (1992)"
  "b"     "ka"    "ni"       37000        "Pouchou (1992)"
  "b"     "ka"    "zr"       4400         "Pouchou (1992)"
  "b"     "ka"    "nb"       4500         "Pouchou (1992)"
  "b"     "ka"    "mo"       4600         "Pouchou (1992)"
  "b"     "ka"    "la"       2500         "Pouchou (1992)"
  "b"     "ka"    "ta"       23000        "Pouchou (1992)"
  "b"     "ka"    "w"       21000        "Pouchou (1992)"
  "b"     "ka"    "u"       7400         "Pouchou (1992)"

Too bad your file (and mine) doesn't include all the MACs for all the emitters in the Pouchou k-ratio database (e.g., Au La, Ma, etc.).  I think Ben was hoping to use the same MACs that Pouchou used for "tuning" the PAP correction to all the measured k-ratios in the Pouchou2.dat database.   I guess Ben was hoping to try and reproduce the error distributions Pouchou originally published...
john

[Brian Joy]

Too bad your file (and mine) doesn't include all the MACs for all the emitters in the Pouchou k-ratio database (e.g., Au La, Ma, etc.).  I think Ben was hoping to use the same MACs that Pouchou used for "tuning" the PAP correction to all the measured k-ratios in the Pouchou2.dat database.   I guess Ben was hoping to try and reproduce the error distributions Pouchou originally published...
john

OK, I see what you mean.  I approach it somewhat differently in that I calculate MACs as needed from Heinrich's equations (or by polynomial interpolation when using the FFAST or CXRO models) and then substitute values from tables as required.  So I don't have a convenient table at hand that contains all MAC values used by P&P for processing their database.  I misunderstood what you were looking for.

However, even with such a table, it might still not be possible to reproduce the results of P&P exactly in CalcZAF.  In their paper in EPQ, they imply that they used the characteristic fluorescence correction of Reed (1965).  Although secondary fluorescence effects are small in the binary systems contained in their database, they're not necessarily negligible.  I haven't programmed Reed's early model, and so I'm not sure to what extent discrepancies in my results versus those presented by P&P are due to this or to some other inconsistency (since they don't give calculated k-ratios line-by-line in Appendix 6).

[Probeman]

Too bad your file (and mine) doesn't include all the MACs for all the emitters in the Pouchou k-ratio database (e.g., Au La, Ma, etc.).  I think Ben was hoping to use the same MACs that Pouchou used for "tuning" the PAP correction to all the measured k-ratios in the Pouchou2.dat database.   I guess Ben was hoping to try and reproduce the error distributions Pouchou originally published...
john

OK, I see what you mean.  I approach it somewhat differently in that I calculate MACs as needed from Heinrich's equations (or by polynomial interpolation when using the FFAST or CXRO models) and then substitute values from tables as required.  So I don't have a convenient table at hand that contains all MAC values used by P&P for processing their database.  I misunderstood what you were looking for.

However, even with such a table, it might still not be possible to reproduce the results of P&P exactly in CalcZAF.  In their paper in EPQ, they imply that they used the characteristic fluorescence correction of Reed (1965).  Although secondary fluorescence effects are small in the binary systems contained in their database, they're not necessarily negligible.  I haven't programmed Reed's early model, and so I'm not sure to what extent discrepancies in my results versus those presented by P&P are due to this or to some other inconsistency (since they don't give calculated k-ratios line-by-line in Appendix 6).

Hi Brian,
I use a similar approach in CalcZAF.  One can choose one of 5 different MAC table sources and then load selected MACs as deemed necessary from the EMPMAC.DAT file. Important for quantifying low energy emissions as you know.

Anyway, it's no big deal, I was just trying to help Ben out. But if such a "Pouchou" MAC table was available, I would add it to the CalcZAF distribution.  Just for fun!
john

[Ben Buse]

Hi John and Brian,

Thank you for your replies, that's very interesting - I was curious as two why they did not agree -and you've answered my question. I guess it be possible to modify the MAC30 table with the values from Appendix 5 of green book -but I'm not sure its worth it - plus as Brian says there's difference in fluorescence correction.

As a related question - does the Cameca SX100 PAP routinue use the MAC's as specified by Pouchou and Pichoir - and Reed 1965 fluorescence correction - if so that would explain why the calczaf results never quite match the Cameca software results. I've never been able to find in the Cameca documentation which MAC's they use.

Thanks

Ben