Si in glasses

[bgarcia]

This is derived from a previous recent post about oxygen and silicate glasses. I've decided to calculate O by stoichiometry in order to focus on the greater issue, Si.

I'm analyzing silicate glass samples, and in acquiring and checking standards, I'm finding that many elements are only accurate when analyzing the assigned standard material. I know this is to be expected to a small degree, however, Si is off by 4+ wt% every time. If I can figure out why this is happening to Si, I can adjust the other elements as some are systematically off as well (though to a much lesser degree).

I'm only after major and minor elements (Si Al Ca Mg K Na Ti Fe) here, with Ti approaching trace levels in my unknowns. I believe I have set the basic conditions accurately: manually chosen background points using 200-pt wavescans, daily PHA check for changes, one applicable APF used, 15kv/10nA/10um-beam. I'm using some USGS glasses for SRMs, BIR-1G (Si Al Ca Mg) and NKT-1G (Na), SPI obsidian (K) and synthetic rutile (Ti).

I've attached the standard anaylses for the glass SRMs, and an unkown analysis of each plus that of rutile and K412 glass. I don't see what's going on in my data: can anyone spot anything obvious? Do I need to do intensity scans for Si?

Also, I'm wondering if an empirical APF would be applicable.

[bgarcia]

some more data attachments.

[Probeman]

Silicon is silicate glasses should not require any special treatment for peak shape/shift.  All your silicate glass standards should agree within 1 to 2% relative accuracy. Have you been able to get agreement in the past?

Do you have the NIST mineral glass standards to try (K-411 and K-412)?  They and BIR-G should all agree quite well, I've tested this many times and they are always in agreement.

If you can post your MDB file that would be helpful.

[bgarcia]

I meant "specified" APFs, sorry.

[bgarcia]

Silicon is silicate glasses should not require any special treatment for peak shape/shift.  All your silicate glass standards should agree within 1 to 2% relative accuracy. Have you been able to get agreement in the past?

That's the impression I was under. As far as past work goes, it's been years since I've done this type of analysis. Minerals and metals are a more common sample type for me. I guess the answer to your question is "not sure".

I don't have the k411 glass, but I do have K2450 (Al-rich).
MDB attached.

[John Donovan]

Hi Ben,
I took a look at your MDB file and I agree- none of your standards seem to agree with each other! 

Some other observations (that don't seem to affect things, though turning off the APF and oxygen by stoichiometry did help, but are generally unnecessary):

1. You have the coating corrections turned on, but both your standards and unknowns are specified as carbon coated.
2. You have aggregate mode turned on but no aggregate elements.
3. You have particle/thin film corrections turned on but these seem to be bulk standards.
4. You have blank corrections turned on, but no blank assigned.
5. You had negative k-ratios being forced to zero and no need for it.
6. You have the standards calculated using stoichiometric oxygen, when normally one would just use the value from the standard database.
7. Your PHA baseline setting are *way* too tight.  Just place all your baselines at 1 volt or so, and set your window to 9 volts or so.
8. You are using APF factors and I don't know why.
9. And most unforgivable: your PFE software is about 8 months out of date!     

Once I turn all these unnecessary things off, here is your silicon primary standard:

St  809 Set  11 BIR-1G Basaltic Glass, Results in Elemental Weight Percents
 
ELEM:       Si      Al      Ca      Mg      Na       K      Fe      Ti       O      Mn
TYPE:     ANAL    ANAL    ANAL    ANAL    ANAL    ANAL    ANAL    ANAL    SPEC    SPEC
BGDS:      LIN    POLY     LIN     LIN     LIN     LIN     LIN     EXP
TIME:    15.00   15.00   15.00   20.00   30.00   20.00   20.00   20.00     ---     ---
BEAM:    10.03   10.03   10.03   10.03   10.03   10.03   10.03   10.03     ---     ---

ELEM:       Si      Al      Ca      Mg      Na       K      Fe      Ti       O      Mn   SUM 
  1094  22.068   8.390   9.459   5.755   2.904    .952   7.920   2.154  43.328    .147 103.078
  1095  22.061   8.326   9.404   5.699   2.995    .918   7.757   2.151  43.328    .147 102.787
  1096  22.290   8.126   9.449   5.926   2.771    .956   7.520   2.141  43.328    .147 102.655
  1097  22.217   8.277   9.499   5.648   2.801    .916   7.748   2.127  43.328    .147 102.708
  1098  22.482   8.146   9.567   5.645   2.713    .932   8.000   2.132  43.328    .147 103.092

AVER:   22.223   8.253   9.476   5.735   2.837    .935   7.789   2.141  43.328    .147 102.864
SDEV:     .175    .114    .061    .116    .112    .019    .185    .012    .000    .000    .207
SERR:     .078    .051    .027    .052    .050    .008    .083    .005    .000    .000
%RSD:      .79    1.39     .64    2.02    3.95    1.98    2.38     .55     .00     .00

PUBL:   22.203   8.203   9.505   5.668   1.372    .030   7.804    .623  43.328    .147  98.886
%VAR:    (.09)   (.61)  (-.31)  (1.17)  106.69 3016.53  (-.19)  243.45     .00     .00
DIFF:    (.02)   (.05)  (-.03)   (.07)   1.464    .905  (-.02)   1.518    .000    .000
STDS:      809     809     809     809     810     333     809     340     ---     ---

The weird thing is that the Na, K and Ti concentrations are off by 100, 3000 and 200% respectively.  So I suspect that your standards are entered into the standard database wrong or they have gotten mixed up.

I would start by making sure that your standards are what you think they are. Analysis of silicate glasses should be similar to analysis of silicate minerals. Maybe try analyzing your silicate minerals (and SiO2) as your primary Si standard and then try to analyze some of your silicate glass standards as secondary standards and see what you get.

Finally, if you are going to utilize TDI correction on these glasses (and you probably should), the elements that you want to apply TDI corrections to (in alkali glasses) are generally Si, Al, Na and K. So that means that those elements all need to be acquired first on their respective spectrometers.

[bgarcia]

I've been working all week on this, and I think that the largest source of error for the USGS glasses was that they were Au-coated.    That's it.

Obviously I've done a lot of other things to tighten up my probe work, but taking off the Au and coating with C puts Si-Al-Mg compositions within 3% or less variance. I wish it would have been higher on my "Things To Check" list!

I'm also finding that for Na K, mineral standards seem to work better when analyzing glass SRMs as unknowns.

Now I just have the original challenge of how to analyze a glass. Thanks for all the help and time, John!

[John Donovan]

I've been working all week on this, and I think that the largest source of error for the USGS glasses was that they were Au-coated.    That's it.

Obviously I've done a lot of other things to tighten up my probe work, but taking off the Au and coating with C puts Si-Al-Mg compositions within 3% or less variance. I wish it would have been higher on my "Things To Check" list!

I'm also finding that for Na K, mineral standards seem to work better when analyzing glass SRMs as unknowns.

Now I just have the original challenge of how to analyze a glass. Thanks for all the help and time, John!

Hi Ben,
Well that would indeed explain your quant issues if some standards were carbon coated and some were gold coated.

By the way, Probe for EPMA does include a coating correction for differences in coating between the standards and the unknowns, even different coatings for different standards as described here:

https://probesoftware.com/smf/index.php?topic=23.msg1256#msg1256

The main caveat is that one needs to know the actual coating thickness (and density) of each standard (and unknown). If these are known (and it's not so easy to be sure), the coating correction works quite well.  But if you read through the topic you can see that even determining the density of carbon in a carbon coat is not straight forward.

That is why we usually just make sure that we either coat both the standards and unknowns at the same time (in which case we can ignore the coating correction), or we make sure to use the same coater on both standards and unknowns and use a thickness monitor or other method to ensure reproducibility for the coating.

[bgarcia]

I've definitely been using the coating correction. For the originally Au-coated glasses, it made a difference, but not enough seemingly. That probably stems from the fact that they were coated in another lab with a questionable amount of Au (~5nm estimate apparently not correct).

Getting all standards and unknowns coated in C, by me, has made a huuuge difference. On my unknowns, the stoichiometric vs calculated oxygen is off by less than 2 wt%, so I think I'm doing something right.

And thanks for that link; it clears up a lot of questions I had about the coating correction. Cheers!