Author Topic: Strategies for Improving Accuracy Using Trace Elements Standards  (Read 2640 times)

Probeman

  • Emeritus
  • *****
  • Posts: 1914
  • Never sleeps...
    • John Donovan
This topic is for discussing trace element standards (whatever that term means to anyone), and how they might be utilized for improving the accuracy of our trace element analyses.

I'll present my thesis for improving accuracy of trace elements, but I would like to hear from others on their own strategies and of course we should discuss the advantages and disadvantages of these various approaches.

But first I think it's worth recalling that on the question of accuracy for trace elements we have the following items to consider:

1. The background correction:

Because the background correction is a subtraction, any error in the background is a direct error in the accuracy of the trace element. For example, a 100 PPM error in the interpolated background determination, say due to a small interfering peak on one of our off-peak positions or interpolation across an absorption edge, would cause a corresponding 100 PPM error in the net intensity of our trace element. So if we are measuring a 100 PPM trace level, we would obtain a result of around 0 PPM, and therefore an 100% relative error in our trace determination. The background correction is therefore usually the largest source of error in our trace element analyses.

2. The matrix correction:

On the other hand, the matrix correction is multiplicative. That is, a 1% relative error in the matrix correction is still a 1% error on our trace element analysis. For example if we are measuring a 100 PPM trace level, our 1 % error from the matrix correction results in a trace element value of 99 or 101 PPM depending on the direction of the matrix correction error.  Therefore, only a 1 PPM absolute error. So at trace levels, the matrix correction is generally a much smaller source of error than the background correction.

Now it should be mentioned up front that whatever (secondary) standards we are using to check the accuracy of our trace element measurements, we will still always be utilizing a (primary) standard that is usually a pure metal or a pure oxide. The advantage of a pure element or a pure oxide (primary) standard is that we can be confident of the accuracy based upon purity considerations only. For example if our Ti metal primary standard is 99.99% pure, its accuracy is known to 99.99% accuracy. Of course, in practice surface oxidation raises its ugly head with pure metals, so perhaps we want to use a pure synthetic TiO2 which is also relatively easy to obtain.

3. Next we can ask the question: what (secondary) standards can we be most confident with regard to accuracy of trace element measurements? I would propose that we can be very confident in the accuracy of so called zero blank standards. That is, standards that have a roughly similar matrix to our unknown, but do not contain any of the trace element in question. Of course if the blank matrix causes an interference correction on our trace element and our unknown matrix is similar, we can consider that the blank correction itself will correct for this interference (without the use of a separate interference correction). This is why we usually do not want to perform both a blank correction and an interference correction. Because we will get a double correction.

So why can we be so confident about the accuracy of zero blank standards?  Well for one thing, if the element in question is below the detection limit of our measurement, then the accuracy of that zero measurement is very high. In fact the accuracy of the zero blank is equal to the variance of the blank measurement itself.  In addition, when our blank is below the detection limit, the homogeneity of our blank standard will likewise be equal to our measurement precision. It will be very difficult to obtain or synthesize a non-zero trace element standard with such accuracy and homogeneity.

That is because the accuracy (and homogeneity) of a non-zero trace element has to be determined.  And how is that determined? Only by other techniques, say ICP-MS, which also has its own systematic errors. That is, if our ICP-MS reports 100 PPM, how sure can we be that it is actually 100 PPM and not say 95 or 104 PPM?  Do we have another opinion?  If so, what are the systematic errors of that method? And if these two methods give different results, do we simply assume that these systematic errors are random and therefore can simply be averaged?

Therefore I conclude that while it might be nice to have a high accuracy non-zero trace element standard, in practice this is a very difficult thing to achieve. On the other hand, if the trace level of our blank standard (say also determined by ICP-MS) is always below 1 PPM, then we can easily have confidence that our trace accuracy equal to our blank measurement precision. And synthesis of very pure (homogeneous) materials is quite easy compared to doped materials.

But let's discuss it because I'm sure there are many strategies for trace element characterization depending on the materials.

john
« Last Edit: July 20, 2017, 09:27:17 am by Probeman »
The only stupid question is the one not asked!

Probeman

  • Emeritus
  • *****
  • Posts: 1914
  • Never sleeps...
    • John Donovan
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #1 on: July 20, 2017, 09:21:50 am »
I'm bumping this topic because I fixed some typos and mistakes in the paragraph on background corrections.
john
The only stupid question is the one not asked!

Probeman

  • Emeritus
  • *****
  • Posts: 1914
  • Never sleeps...
    • John Donovan
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #2 on: May 06, 2018, 04:22:37 pm »
Thoughts for the day...

1. A suitable blank secondary standard is not always available (e.g, for monazite), but it should be remembered that it is easier to accurately determine a zero concentration, than a non-zero concentration. So if a suitable blank standard material is available (and hopefully at least somewhat matrix matched to our unknown), we can simply utilize another technique (e.g., ICP-MS or SIMS) which has better sensitivity, but probably less accuracy, in order to determine that an element is definitely below possible EPMA detection limits, that is, a zero concentration (for the purposes of an EPMA secondary standard).

2. In the case of a blank secondary standard with a zero concentration, one can determine the accuracy of a zero concentration with an accuracy equal to the measurement sensitivity.  That is, if the element is significantly below EPMA detection limits, then the secondary blank standard accuracy is equal to the standard deviation of the measurement.

3. Utilizing a non-zero trace element (secondary or primary!) standard, can only *decrease* trace element accuracy, as we cannot truly know the actual value of a non-zero trace element concentration.  That is, is the trace element standard actually 110 PPM or is it actually 115 PM?  Who knows?

4. For best precision we should generally utilize a primary standard with a high concentration of the element in question. This is usually  a pure element or simple oxide standard.  That is, for maximum sensitivity (see the various expressions for calculating sensitivity), we want the highest cps rate per weight percent of the element being measured in the primary standard.

5. For best accuracy we want to measure a secondary standard (ideally at least somewhat matrix matched), with a demonstrated zero concentration of the element.  That is, how accurately can we measure zero in our secondary standard?

6. Once our error from zero concentration is known from our secondary blank standard, we can then apply this "blank correction" in our trace element analyses for best accuracy.  This allows us to correct for sample, detector and continuum spectral artifacts with the best accuracy.
« Last Edit: May 06, 2018, 05:01:06 pm by Probeman »
The only stupid question is the one not asked!

D.

  • Professor
  • ****
  • Posts: 42
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #3 on: July 16, 2018, 02:07:58 pm »
Hi All,

Perhaps this question needs to be a separate thread, but it's sort of relevant to the topic...

Does anyone analyze trace element glasses (SRMs) as secondary standards to check primary trace element calibrations that are intended for analysis of non-glass phases? E.g., say, analyzing a NIST glass with a routine intended for amphibole.

If so, do you find these glass analyses to be sufficiently accurate given the matrix differences between glass and primary standards?

Do you find that the beam diameter/current issues that affects major/minor element migration and signal variation in glasses also affect the accuracy of the traces?

Thanks,
Deon.

Probeman

  • Emeritus
  • *****
  • Posts: 1914
  • Never sleeps...
    • John Donovan
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #4 on: July 17, 2018, 06:04:16 pm »
Perhaps this question needs to be a separate thread, but it's sort of relevant to the topic...

Does anyone analyze trace element glasses (SRMs) as secondary standards to check primary trace element calibrations that are intended for analysis of non-glass phases? E.g., say, analyzing a NIST glass with a routine intended for amphibole.

If so, do you find these glass analyses to be sufficiently accurate given the matrix differences between glass and primary standards?

Do you find that the beam diameter/current issues that affects major/minor element migration and signal variation in glasses also affect the accuracy of the traces?

Thanks,
Deon.

Hi Deon,
Lots to consider here. 

First remember that the accuracy of our modern matrix corrections are on the order of a few percent (and usually much better). If your precision trace element precision is better than that, well, good on you!   :)

In other words the difference in matrix will generally not be an issue for trace elements (in standards). We usually just want to use a primary standard with a high concentration of the elements and, as I've said before, a good secondary *blank* standard. Because the problem with non-zero trace element standards is their accuracy.  How are these non-zero values actually determined?  And how accurate is that technique?   

The most accurate trace element measurement we can make is usually a zero blank measurement where the element in question is *below* EPMA detection limits, so we get to see how well we can measure *zero* with an accuracy equal to our measurement precision.  In making a blank measurement it is best to have a matrix match, not because of the matrix correction accuracy, but because then we can know that we are properly handling all the other problems such as background positions, absorption edges and spectral interferences.

The two posts above lay out my thoughts on the matter.

On the question of beam sensitive samples for trace elements, yes, this can be a significant issue. Here is a discussion on trace elements started by Julien Allaz:

http://probesoftware.com/smf/index.php?topic=186.0

Note the TDI plots of U and Th attached at the bottom of his post.
john
« Last Edit: July 17, 2018, 06:06:18 pm by Probeman »
The only stupid question is the one not asked!

D.

  • Professor
  • ****
  • Posts: 42
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #5 on: July 18, 2018, 12:42:44 am »
Hi John,

Out of interest, does knowing that you can measure zero well, ensure you can accurately measure a well-characterized (with multiple techniques) CRM trace element glass like NIST 610? Have you tested it?

In my original question, what I'm really asking, is if anyone out there routinely uses glass trace element standards successfully as secondary standards in the absence of availability of a  "zero blank" or a matrix-matched secondary standard.

Cheers,
Deon.

Probeman

  • Emeritus
  • *****
  • Posts: 1914
  • Never sleeps...
    • John Donovan
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #6 on: July 18, 2018, 09:12:29 am »
Out of interest, does knowing that you can measure zero well, ensure you can accurately measure a well-characterized (with multiple techniques) CRM trace element glass like NIST 610? Have you tested it?

Hi Deon,
Having a zero (and matrix matched) zero blank ensures that one can measure accurately at the trace level.  See the earlier paper for more info:

http://epmalab.uoregon.edu/pdfs/3631Donovan.pdf

By having such a zero blank (and utilizing it in a blank correction) one now knows that their accuracy is equal to their measurement precision in that matrix for those elements. How nice is that!

I've looked at SRM 610 and 612 years ago and found them not to be very homogeneous. First of all note that they were never intended to used as microanalytical standards. Only for bulk methods. And even there it's problematic. For example the Mn value is 457 PPM +/- 55 PPM which means that one doesn't know whether the value is 400 PPM or 450 PPM or 500 PPM.  A range of 100 PPM is a lot of imprecision to have confidence in one's accuracy. 

Having a zero blank of that matrix glass would be far better in my opinion, as then one could know how well they can measure zero in that matrix with an accuracy equal to EPMA precision which could be only a few PPM for many elements.

In my original question, what I'm really asking, is if anyone out there routinely uses glass trace element standards successfully as secondary standards in the absence of availability of a  "zero blank" or a matrix-matched secondary standard.

I guess it depends on how one defines "successfully".   How would one know if they have been successful? 

What I would do is find a zero blank standard with a matrix as close as possible to the unknown in question, and utilize that in my blank correction in Probe for EPMA.  As I said before, extrapolating the matrix correction is accurate to a few percent or better. That is much better than the 10, 20 or 30% accuracy variance in the SRM trace values.  For oxide matrices even just a pure SiO2 synthetic glass would be more than close enough.  Again, the idea is to know that the chosen background positions, sample/detector absorption edges and spectral overlaps are being dealt with properly.

Here are some other topics on trace elements:

http://probesoftware.com/smf/index.php?topic=980.0

http://probesoftware.com/smf/index.php?topic=204.0

And here is a topic discussing using synthetic SiO2 as a blank standard:

http://probesoftware.com/smf/index.php?topic=130.0

Did I answer your question?  In your original post you mentioned trying to measure trace elements in amphibole. That's a tough material because it has so many elements and there are some spectral interferences, e.g., Mn Kb on Fe, Fe L on F Ka, etc..  Note that nasty as monazite but still tough when one wants to get accuracy below 100 PPM.

As I mentioned the NIST glass isn't certified for microanalysis (that I know of). In fact it's no longer available from NIST but it can still be purchased.  So it might be the next best thing for a secondary standard, but I'd also throw a synthetic SiO2 secondary standard in there because they are cheap and easily obtained.  After that I might look for a material such as a simple (synthetic?) silicate that has been measured for trace elements by more sensitive bulk techniques such as ICP-MS.

This post is getting long so I'm going to post another approach for you to consider next.
« Last Edit: July 18, 2018, 09:54:30 am by Probeman »
The only stupid question is the one not asked!

Probeman

  • Emeritus
  • *****
  • Posts: 1914
  • Never sleeps...
    • John Donovan
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #7 on: July 18, 2018, 11:12:56 am »
And then there's another thing to consider which may or may not be possible for you (where are you again?).  When acquiring standards in Probe for EPMA the default is to acquire all the elements one is analyzing for in all the standards. One can always check the "quick standards" checkbox later to save time once the elements/backgrounds/interferences setup is finalized, but at least to begin with, one really should measure all the elements one is measuring in all of the standards in the run.

Why is this useful?  Because in PFE one can then "analyze" each primary and secondary (and MAN) standard as an unknown, and then one can ascertain whether they have been successful or not in measuring, first, the minor or major elements in any standard that is *not* a primary standard, and second, for our current purposes, any elements that are denoted as zero in those secondary standards.  For example, one might be using synthetic MnO as a Mn standard. In this case, by measuring all our elements (including say Fe), we can evaluate whether we can measure zero Fe in the presence of Mn. Of course we will immediately learn that there is a Mn Kb interference on Fe K which needs to be dealt with, but the good news is that we've *already* calibrated this spectral interference in measuring all our elements in that standard matrix! One simply then assigns the interference with a few mouse clicks and you're good to good (for trace Fe in the presence of Mn at least!).

As another example one might be analyzing the following elements for an amphibole (this is actually a student run for traces in apatite but it makes the point):

ELEM:    ca ka    p ka    s ka   al ka   fe ka   as ka   ba la    k ka
ELEM:    mn ka   na ka   sr la    f ka   sm la   nd la   cl ka    y la
ELEM:    eu la   gd la   mg ka   la la   ce la   si ka

The standards specified for this run were:

160 NBS K-412 mineral glass
285 Ca10(PO4)6Cl2 (halogen corrected)
327 Anhydrite (CaSO4) UC # 5555
336 Nepheline (partial anal.)
835 BaF2 (barium fluoride)
1001 CePO4 (USNM 168484)
1004 EuPO4 (USNM 168487)
1005 GdPO4 (USNM 168488)
1007 LaPO4 (USNM 168490)
1009 NdPO4 (USNM 168492)
1011 SmPO4 (USNM 168494)
1015 YbPO4 (USNM 168498)
1016 YPO4 (USNM 168499)
25 MnO synthetic
251 Strontium titanate (SrTiO3)
662 GaAs (synthetic)

Now once the standards are acquired, we can "analyze" them as unknowns in PFE to check for backgrounds, absorption edges, interferences, etc.  As you know, problems with the background usually manifest themselves as negative concentrations, while problems with interferences usually reveal themselves as concentrations statistically above zero (assuming the trace element in the standard is at zero concentration).  So here is an example of our synthetic chlor-apatite being analyzed as a secondary standard:

St  285 Set   3 Ca10(PO4)6Cl2 (halogen corrected), Results in Elemental Weight Percents
 
ELEM:       Ca       P       S      Al      Fe      As      Ba       K
TYPE:     ANAL    ANAL    ANAL    ANAL    ANAL    ANAL    ANAL    ANAL
BGDS:      LIN     EXP     LIN     LIN     LIN     EXP     LIN     LIN
TIME:    25.00    8.00   25.00   25.00   30.00   20.00   30.00   30.00
BEAM:    30.14   30.14   30.14   30.14   30.14  100.24  100.24  100.24

ELEM:       Ca       P       S      Al      Fe      As      Ba       K   SUM 
    84  37.971  17.987   -.002   -.007    .013    .227    .025   -.003 100.606
    85  38.465  17.962   -.001   -.008   -.007    .011    .002   -.001  99.534
    86  38.991  17.616   -.018    .004    .012    .050    .034    .006 100.038

AVER:   38.476  17.855   -.007   -.004    .006    .096    .020    .001 100.059
SDEV:     .510    .207    .010    .007    .011    .115    .016    .005    .536
SERR:     .295    .120    .006    .004    .006    .067    .009    .003
%RSD:     1.33    1.16 -140.09 -191.57  189.72  120.09   80.55  586.21

PUBL:   38.481  17.843    n.a.    n.a.    n.a.    n.a.    n.a.    n.a. 100.000
%VAR:   (-.01)   (.07)     ---     ---     ---     ---     ---     ---
DIFF:   (-.01)   (.01)     ---     ---     ---     ---     ---     ---
STDS:      285     285     327     336     160     662     835     336

STKF:    .3562   .1623   .2238   .1336   .0637   .5068   .6946   .0401
STCT:    740.7  4652.6  1673.7  6266.2  1409.7   586.0  3117.7   564.0

UNKF:    .3562   .1623  -.0001   .0000   .0000   .0007   .0001   .0000
UNCT:    740.7  4652.4     -.5    -1.3     1.1      .8      .6      .1
UNBG:      6.6    12.8     3.6    31.0    26.3     8.6     5.7    10.0

ZCOR:   1.0803  1.1000  1.1606  1.3187  1.2139  1.3521  1.5926  1.0850
KRAW:   1.0000  1.0000  -.0003  -.0002   .0008   .0014   .0002   .0002
PKBG:   113.58  369.01     .89     .96    1.04    1.09    1.11    1.02
INT%:     ----    ----    ----    ----    ----    ----    ----    ----

TDI%:    1.107    .530    .000   -.701    .617    ----    ----    ----
DEV%:       .3      .2      .0     1.1     2.3    ----    ----    ----
TDIF:  LOG-LIN LOG-LIN LOG-LIN LOG-LIN LOG-LIN    ----    ----    ----
TDIT:    56.33   37.00   56.00   54.67   57.00    ----    ----    ----
TDII:     747.   4666.    2.80    29.7    27.9    ----    ----    ----
TDIL:     6.62    8.45    1.03    3.39    3.33    ----    ----    ----
 
ELEM:       Mn      Na      Sr       F      Sm      Nd      Cl       Y
TYPE:     ANAL    ANAL    ANAL    ANAL    ANAL    ANAL    ANAL    ANAL
BGDS:      LIN     LIN     LIN     LIN     LIN     LIN     LIN     LIN
TIME:    30.00   30.00   30.00   30.00   30.00   30.00   20.00   30.00
BEAM:   100.24  100.24  100.24  100.24  100.24  100.24  100.24  100.24

ELEM:       Mn      Na      Sr       F      Sm      Nd      Cl       Y   SUM 
    84    .002   -.001    .001   -.044   -.018    .004   7.547   -.016 100.606
    85    .018    .003    .005   -.117    .010   -.005   6.302   -.003  99.534
    86    .009    .002    .017   -.126    .037    .007   6.542   -.011 100.038

AVER:     .010    .001    .008   -.096    .010    .002   6.797   -.010 100.059
SDEV:     .008    .002    .008    .045    .028    .006    .661    .007    .536
SERR:     .005    .001    .005    .026    .016    .004    .381    .004
%RSD:    83.76  194.83  107.46  -46.93  281.24  286.57    9.72  -65.92

PUBL:     n.a.    n.a.    n.a.    n.a.    n.a.    n.a.   6.808    n.a. 100.000
%VAR:      ---     ---     ---     ---     ---     ---  (-.16)     ---
DIFF:      ---     ---     ---     ---     ---     ---  (-.01)     ---
STDS:       25     336     251     835    1011    1009     285    1016

STKF:    .7307   .0742   .3872   .1996   .4847   .4796   .0601   .3980
STCT:   4313.6  1769.9  3829.9   577.4  1129.8  3359.2   730.7  1079.3

UNKF:    .0001   .0000   .0001  -.0002   .0001   .0000   .0600  -.0001
UNCT:       .5      .1      .6     -.6      .1      .1   729.5     -.2
UNBG:      5.3    10.1     6.1     1.8     3.9    10.1     8.3     2.5

ZCOR:   1.2419  2.0160  1.3687  4.8155  1.5658  1.5508  1.1324  1.3564
KRAW:    .0001   .0001   .0001  -.0010   .0001   .0000   .9982  -.0002
PKBG:     1.09    1.01    1.09     .67    1.04    1.01   89.12     .92
INT%:     ----    ----    ---- -110.96     .19    -.59     .00    ----
 
ELEM:       Eu      Gd      Mg      La      Ce      Si       O
TYPE:     ANAL    ANAL    ANAL    ANAL    ANAL    ANAL    SPEC
BGDS:      AVG     LIN     LIN     LIN     LIN     LIN
TIME:    30.00   30.00   30.00   30.00   30.00   20.00     ---
BEAM:   100.24  100.24  100.24  100.24  100.24  100.24     ---

ELEM:       Eu      Gd      Mg      La      Ce      Si       O   SUM 
    84    .015    .003    .032   -.015    .009    .008  36.868 100.606
    85    .015   -.006    .033   -.008   -.012    .005  36.868  99.534
    86    .002   -.019    .027   -.016    .000    .006  36.868 100.038

AVER:     .011   -.007    .031   -.013   -.001    .006  36.868 100.059
SDEV:     .008    .011    .003    .005    .010    .001    .000    .536
SERR:     .005    .006    .002    .003    .006    .001    .000
%RSD:    72.59 -158.52   10.40  -36.15-1391.88   24.61     .00

PUBL:     n.a.    n.a.    n.a.    n.a.    n.a.    n.a.  36.868 100.000
%VAR:      ---     ---     ---     ---     ---     ---     .00
DIFF:      ---     ---     ---     ---     ---     ---    .000
STDS:     1004    1005     160    1007    1001     160     ---

STKF:    .4897   .4890   .0789   .4694   .4676   .1626     ---
STCT:   1247.5  5000.7  3048.5 11847.3  3558.3  5668.3     ---

UNKF:    .0001   .0000   .0002  -.0001   .0000   .0001     ---
UNCT:       .2     -.4     7.9    -2.1      .0     1.9     ---
UNBG:      4.5    18.1    19.3    54.1    21.4     5.1     ---

ZCOR:   1.5781  1.6049  1.5052  1.5738  1.5597  1.1215     ---
KRAW:    .0001  -.0001   .0026  -.0002   .0000   .0003     ---
PKBG:     1.04     .98    1.41     .96    1.00    1.37     ---
INT%:    -2.09   -5.66    ----     .18  -46.55    ----     ---

Of course most of the elements are labeled as n.a. for "not analyzed" because I haven't sacrificed any standard material for ICP-MS (I should do that!), but since it's a synthetic I might be able to assume that the starting materials were relatively pure for most contaminants.

So ignoring the Ca, P and Cl channels (which this is the primary standard for), the other trace elements (with the exception of Mg which is probably an uncorrected overlap from 3rd order Ca K), generally look to be within 1 or 2 standard deviations from zero, so this would give us some confidence that one can measure zero in this matrix for the other trace elements.

One could do the same for a very different matrix such as the NIST K-412 glass (which was the standard for Si, Al and Fe so ignore those channels in this case):

St  160 Set   3 NBS K-412 mineral glass, Results in Oxide Weight Percents

ELEM:      CaO    P2O5     SO3   Al2O3     FeO   As2O3     BaO     K2O   SUM 
   205  15.202    .044   -.023   9.442  10.049    .184   -.045    .004 100.427
   206  14.389    .019   -.032   9.253   9.960    .162   -.003    .014  99.458
   207  14.711    .000    .056   9.364   9.861    .203    .004    .008  99.941

AVER:   14.768    .021    .000   9.353   9.957    .183   -.015    .008  99.942
SDEV:     .410    .022    .048    .095    .094    .021    .027    .005    .484
SERR:     .236    .013    .028    .055    .054    .012    .015    .003
%RSD:     2.77  102.96    ----    1.01     .95   11.29 -181.99   59.43

PUBL:   15.250    n.a.    n.a.   9.270   9.960    n.a.    n.a.    n.a. 100.120
%VAR:    -3.16     ---     ---     .90  (-.04)     ---     ---     ---
DIFF:    -.482     ---     ---    .083   (.00)     ---     ---     ---
STDS:      285     285     327     336     160     662     835     336

ELEM:      MnO    Na2O     SrO       F   Sm2O3   Nd2O3      Cl    Y2O3   SUM 
   205    .091    .081    .042    .024    .017   -.007   -.004   -.002 100.427
   206    .094    .067    .049   -.024    .000   -.009    .006   -.009  99.458
   207    .090    .058    .060   -.005   -.026   -.007    .005   -.017  99.941

AVER:     .092    .069    .050   -.002   -.003   -.007    .002   -.009  99.942
SDEV:     .002    .012    .009    .024    .022    .001    .005    .008    .484
SERR:     .001    .007    .005    .014    .012    .001    .003    .004
%RSD:     2.21   17.25   17.21-1404.69 -763.06  -14.53  220.79  -81.53

PUBL:     .099    .058    n.a.    n.a.    n.a.    n.a.    n.a.    n.a. 100.120
%VAR:    -7.97   18.58     ---     ---     ---     ---     ---     ---
DIFF:    -.008    .011     ---     ---     ---     ---     ---     ---
STDS:       25     336     251     835    1011    1009     285    1016

ELEM:    Eu2O3   Gd2O3     MgO   La2O3   Ce2O3    SiO2       O   SUM 
   205   -.030    .007  19.333   -.016   -.026  45.269    .791 100.427
   206    .035    .006  19.291   -.021   -.029  45.432    .809  99.458
   207    .019   -.017  19.283   -.003    .003  45.490    .801  99.941

AVER:     .008   -.001  19.302   -.013   -.017  45.397    .800  99.942
SDEV:     .034    .014    .027    .009    .018    .115    .009    .484
SERR:     .019    .008    .016    .005    .010    .066    .005
%RSD:   416.32-1044.43     .14  -67.81 -102.03     .25    1.13

PUBL:     n.a.    n.a.  19.331    n.a.    n.a.  45.352    .800 100.120
%VAR:      ---     ---  (-.15)     ---     ---   (.10)     .02
DIFF:      ---     ---  (-.03)     ---     ---   (.04)    .000
STDS:     1004    1005     160    1007    1001     160     ---

Note that Na appears to be non-zero in the NIST glass but it is not reported, so the PUBL: value there is my own best attempt to quantify Na in K-412.  But the other elements again are all statistically within zero, with the possible exception of As and Sr. The Sr could be an uncorrected interference from Si Kb'.  Note sure what is going on with As. Has any one ever done a sensitive bulk technique on traces in the NIST K-412, K-411 glasses?

Anyway, something like this might be worth trying if you don't have an amphibole zero blank standard for traces.

Edit by John: just noticed that Mn is also statistically above zero in the K-412 glass. I've remember now that I've seen that  trace Mn for years.  Probably a contaminant.  One thing we could all use is more characterization of the trace elements in our standards...
« Last Edit: July 18, 2018, 02:53:29 pm by Probeman »
The only stupid question is the one not asked!

D.

  • Professor
  • ****
  • Posts: 42
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #8 on: July 19, 2018, 01:56:26 am »
Hi John,

Thank you for the details. I will take a look at it.

FYI: With respect to the NIST 610 glass (it is listed as available on the NIST website), are you familiar with the Jochum et al. 2011 (Geostandards and Geoanalytical Research Vol. 35 no.4 p.397-429) study that evaluated its (and other 61x glasses’) homogeneity for application as a microanalytical standard? They conclude that at the 25 micron spot size (Laser ablation), more trace elements than not (and virtually all lithophiles) are homogeneously distributed.

Cheers,
Deon.

Probeman

  • Emeritus
  • *****
  • Posts: 1914
  • Never sleeps...
    • John Donovan
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #9 on: July 19, 2018, 11:52:13 am »
FYI: With respect to the NIST 610 glass (it is listed as available on the NIST website), are you familiar with the Jochum et al. 2011 (Geostandards and Geoanalytical Research Vol. 35 no.4 p.397-429) study that evaluated its (and other 61x glasses’) homogeneity for application as a microanalytical standard? They conclude that at the 25 micron spot size (Laser ablation), more trace elements than not (and virtually all lithophiles) are homogeneously distributed.

Hi Deon,
Thanks. That is good to know. 

Oh, I mistakenly looked at the "archived" SRMs page.   :-[
john
The only stupid question is the one not asked!

Probeman

  • Emeritus
  • *****
  • Posts: 1914
  • Never sleeps...
    • John Donovan
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #10 on: July 13, 2019, 10:31:39 am »
At the recent QMA 2019 in Minneapolis it just so happened that Mike Jercinovic and I both showed up at "The Beacon" restaurant for breakfast at 7 AM, as we were both staying at "The Graduate". We continued to have breakfast together every morning at 7 AM the rest of the week, and of course the conversation ranged over many subjects while he had his omelet and I had my eggs "over easy".

And of course we talked about trace elements. Let me just say it: Mike "gets" trace element EPMA analysis. For one thing he's a really smart guy and for another, he's thought about trace elements for a long time, and finally he's done a tremendous amount of work on trace elements, in some very difficult materials. In short, he's an expert on the subject. So forgive me, but I'm going to paraphrase Mike on this topic, but I think it's a fairly accurate representation of his feelings on the subject. And he can chime in on where he agrees or disagrees.

Mike said to me, over those eggs, that when we're measuring major elements, there's no question that we need to have accurate matrix corrections and well characterized standards, that ideally are matrix matched to deal with all those nasty details such as peak shifting, beam sensitivity, etc. Simply because the accuracy of major element concentrations are affected by these choices.  Background measurements and background models?  Not so important for characterizing major element concentrations. It's simply a question of P/B.

Trace elements are a totally different beast. When we get down to concentration levels approaching zero, the thing that dominates our trace element accuracy is the accuracy of the background measurement (and possibly spectral interferences, if any are present). Basically the effects of accuracy from our primary standard on our trace elements are "in the noise". That is to say, a few percent error in our primary standard intensities or matrix corrections, is generally going to be significantly smaller that the precision of our trace element measurements.

But accuracy errors in our background characterization are additive, not multiplicative. A 500 PPM accuracy error in our background measurement is a 500 PPM accuracy error in our trace element accuracy. At the 500 PPM level, that is a 100% accuracy error! How accurate are your backgrounds at 0.05 wt% levels?

Of course for minor elements it's a bit of both ends, that is the accuracy of our standards, matrix corrections and background characterization.  Depending of course on the elements, samples and matrix physics.

So, for trace elements at least, lets stop focusing on standards and matrix corrections and place the emphasis on accurate measurements of the background. And how can we accurately determine our background at say, those 0.05 wt% levels? By trying to measure zero!  Ideally a blank standard that matrix matches our unknown, contains a zero concentration of the elements of interest, and of course is measured as an unknown, under the same conditions as our unknowns. As Mike would say "try measuring nothing, because if you can't measure nothing, you can't measure anything!"

But here's another way of looking at it: EPMA sensitivity is limited, compared to many other techniques (SIMS, LA-ICPMS, etc.), to around, let's be generous, say ~1 PPM. But what if we turned that limitation of sensitivity into an advantage in accuracy?  :o

Say we have a standard blank that has less than 1 PPM of the element of interest (as determined by SIMS or LA-ICPMS). As far as the electron microprobe is concerned, that is a big fat zero concentration. And guess what? We can use this standard to determine our trace element accuracy (background at zero concentration) to within the precision of our measurement. If we measure this blank standard (as an unknown) and we get zero wt% +/- 3 PPM, then our trace element accuracy is +/- 3 PPM. And you can take that to the bank.  8)

You might not agree with everything I've said so far, but I think Mike J. does. And if you don't agree, I think that simply means that you haven't thought enough about trace elements. I'm sorry.

Now here is where I'm going to get a little weird and rant a bit. I'm seeing a lot of time and resources spent on synthesizing "doped" trace element standards. But it's really difficult to obtain "doped" trace concentrations that are one, homogeneous, and two, accurately characterized. Assuming one can synthesize something that is perfectly homogeneous (and how do we know that really?), how do we know if it's actually 95 PPM or 105 PPM, or 110 PPM?

But do you know what is perfectly homogeneous, and perfectly accurate (as far as EPMA is concerned)? A standard blank that has less than 1 PPM of the element of interest. If one is measuring Ti in quartz, then any synthetic quartz that has less than 1 PPM of Ti, is perfectly homogeneous and perfectly accurate, to within the precision of our EPMA measurement (e.g., Kartsen Goemann and Sandrin Feig utilize a SpectroSil SiO2 glass with 50 PPB of Ti with excellent results).

So, if you're trying to synthesize doped standards at the 100 PPM, or 200 PPM, or 500 PPM or whatever trace element level, please stop. And instead please synthesize matrix matched standards with *zero* concentrations of the elements of interest. I think we've got synthetic quartz covered, but we could use all sorts of other synthetic standards, for example, synthetic TiO2, ZrSiO4, Fe2SiO4, Mg2SiO4, FeS2, synthetic plag, synthetic k-spar, synthetic sphalerite, etc., etc., etc.

And please produce these materials up to kilogram quantities. Yes, that might require collaborating with laser and optical material companies or other university or national labs. But this way, maybe, just maybe, every damn lab on the planet (including SEM labs!) can have 1 gram of the same standard material and finally, maybe then, we can all be "on the same page" when it comes to standard accuracy!  Then we can retire San Carlos olivine and Amelia albite (fiascos) once and for all!  Frankly it's embarassing what we have for EPMA standards today.

I'll close with an apology to Mike J. in advance, but I just had to get this off my chest! OK, that's Probeman's rant for the week.  Have a great weekend.   :)
« Last Edit: July 13, 2019, 01:11:49 pm by Probeman »
The only stupid question is the one not asked!

Mike Jercinovic

  • Professor
  • ****
  • Posts: 63
    • UMass Geosciences Microprobe-SEM Facility
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #11 on: July 13, 2019, 06:47:29 pm »
Obviously, could not agree more!

Probeman

  • Emeritus
  • *****
  • Posts: 1914
  • Never sleeps...
    • John Donovan
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #12 on: July 14, 2019, 08:18:01 am »
Obviously, could not agree more!

Hi Mike,
Thanks!

So on the subject of matrix matched zero blank standards, I know you've tried various synthetic phosphates, such as GdPO4 (if only to look at the shape of the background near the U and Pb M lines).  But what synthetic material(s) do you think would make an ideal zero blank standard for trace elements in monazites?

I know growing synthetic monazites is problematic. I heard from John Hanchar a while back that all his synthesis attempts produced zoned crystals. But here's the thing. Do we even care what the synthetic monazite composition is, as long as we know that it has zero U and Pb?  It could even have Th in there as long as we utilized a pure Th starting material that contains zero U and Pb, as it would have a zero age!

Of course we'd have to resynthesize it every million year or so...   ;)
The only stupid question is the one not asked!

Mike Jercinovic

  • Professor
  • ****
  • Posts: 63
    • UMass Geosciences Microprobe-SEM Facility
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #13 on: July 14, 2019, 09:22:50 am »
It is not often that a "failed" synthesis turns out to be very useful, but that is the case here.  Our first UMass EPMA post doc, Philippe Goncalves got into synthesizing REE phosphates, consulting with Daniele Cherniak at RPI.  He made some nice ones, but then tried for more monazite-like compositions, so mixtures of major amounts of La, Ce and Nd, then adding Th Y and U also if possible.  Like John Hanchar says, just adding Th produces spectacularly zoned crystals.  Philippe was quite discouraged, but we realized eventually that this was actually very useful as a blank for the overall protocol.  Zero age of course, but the analysis should produce zero Pb, and zero U.  This is such a great test with varying Th because of the interferences of Th on PbMa, UMa and UMb.  You should get zero age, Pb and U for all the different Th zones.  We never pursued more, meaning adding a couple wt.% U or a couple wt.% Y which would also be very useful blanks.  But the Th monazite synthesis experiment is super-useful!
   Maybe we should put together an abstract or something on this, but we have been bashed by reviewers many times about "straw man" arguments and "beating a dead horse".  However, I think this approach is useful and maybe not so obvious.  So, as a rancher back in my native NM might say (and a tip of the old Stetson to you Clay, wherever you are!).. "saddle up your departed ride scarecrow, we're burnin' daylight!"

Probeman

  • Emeritus
  • *****
  • Posts: 1914
  • Never sleeps...
    • John Donovan
Re: Strategies for Improving Accuracy Using Trace Elements Standards
« Reply #14 on: July 14, 2019, 11:54:07 am »
Philippe was quite discouraged, but we realized eventually that this was actually very useful as a blank for the overall protocol.  Zero age of course, but the analysis should produce zero Pb, and zero U.  This is such a great test with varying Th because of the interferences of Th on PbMa, UMa and UMb.  You should get zero age, Pb and U for all the different Th zones. 

Hi Mike,
Exactly. Right on.  This should be a standard used by every lab attempting U-Th-Pb microprobe dating.  To answer the most important question: in this matrix, how accurately can one measure zero!

So how much material was synthesized?

We never pursued more, meaning adding a couple wt.% U or a couple wt.% Y which would also be very useful blanks.  But the Th monazite synthesis experiment is super-useful!

But, but, but again I would argue that adding some U wouldn't be that useful. Because how homogeneously is the U distributed? And how do we know *exactly* what the U concentration is?  We can't.  Though additional Y would be a nice test of the spectral interference correction, as long as the U and Pb contents stay zero. 

Because as we've said above, if the concentrations of U and Pb are *less* than 1 PPM, that's *exactly* a big fat zero as far as electron microprobe trace element accuracy is concerned!   8)  For all of our own sake, please don't get me going again!   :P

Anyway, as you pointed out, the neat thing about this "failed" synthesis is that it does not matter that it's inhomogeneous, nor that the concentrations of the REEs or Th are unknown. One should still always get zero U and Pb no matter where they measure on this material. If they are doing it right that is! 

Can I get some?    :D
« Last Edit: July 14, 2019, 04:46:39 pm by Probeman »
The only stupid question is the one not asked!