Ti-Pt alloy analysis: optimum ZAF?

[Julien]

Hi everyone,

Recently a customer asked me to analyze a bunch of Ti-Pt alloys, each being made of 2 to 4 different phases. Aside of the small size of the particle, I quickly realized that different matrix correction procedure yield pretty different results. The Ti and Pt weight-% content can easily vary by 2-3 wt-% each depending on the correction procedure used (and these variations are not covariant: when Ti increase with a matrix correction, the Pt-content sometimes goes up, sometimes goes down). Here is an example in one phase:

Elemental Weight Percents (average of 5 points, each very similar):
ELEM:        N       O      Ti      Pt   TOTAL
     1    .033    .405  32.423  65.961  98.823   Armstrong/Love Scott (default)
     2   -.527    .203  30.865  73.592 104.133   Conventional Philibert/Duncumb-Reed
     3   -.557    .248  32.523  70.250 102.464   Heinrich/Duncumb-Reed
     4   -.376    .292  31.481  67.185  98.582   Love-Scott I
     5   -.307    .324  31.757  67.929  99.702   Love-Scott II
     6    .150    .298  29.716  71.031 101.195   Packwood Phi(pz) (EPQ-91)
     7    .021    .336  33.631  70.589 104.577   Bastin (original) Phi(pz)
     8   -.597    .251  31.637  68.291  99.582   Bastin PROZA Phi(pz) (EPQ-91)
     9   -.119    .297  31.479  69.211 100.868   Pouchou and Pichoir-Full (Original)
    10   -.132    .312  31.777  69.357 101.314   Pouchou and Pichoir-Simplified (XPP)

This can be problematic as the key for this study is to determine the Ti to Pt (atoms) proportion in each phase. This remind me of the Si-Ir example John Armstrong & Paul Carpenter once presented to me, and in their case the "accurate" value was actually obtain... only when using an Si-Ir standard instead of pure Si and pure Ir standards ("classical" example of the use of CalcZAF here - see also Lehigh School of Microscopy!). Unfortunately in my case, I don't have access to a certified / known composition Pt-Ti alloy, and I now wonder which matrix correction is the "best" / most accurate.

And this is only a small part of the iceberg, as a side-kick of this project would have been to analyze nitrogen and oxygen. Nitrogen is (almost) out of the question here, as the count rate is very low, the expected content is low (most being < 1000 ppm), and N Ka interfere with Ti Ll (=> large error, high correction and very high detection limit around 5000-8000 ppm depending on the counting time). Oxygen is another beast, and I got decent count rate. However, we were surprised by the high O wt-% content. While some phases appear poor in O (as expected: samples were prepared in anoxic environment), other shows up to 1 wt-% O and more (example above show ~0.3%).

Therefore three questions to you:

(a) What ZAF/phi-rho-z correction would you recommend for Ti-Pt alloys, assuming only pure Ti and pure Pt standards are available? Analytical conditions involve Ti Ka and Pt Ma, at 15 keV, 100-200 nA (assumes no problem with PHA / high count rate, each standard yield around or less than 100 cps/nA with the selected setup, and almost half less in the unknown => around 10000 to maximum 20000 cps - should be OK?).

(b) Same as (a), but this time if we further consider the measurement of light element (N and O) at these conditions: which matrix correction to choose (curiously enough, Armstrong yield significantly higher O-content: 0.4% vs. ~0.2-0.3 for all others)?

(c) Without commenting on the (very) bad choice of voltage (15 keV) for light element (*), the high oxygen content appears surprising. Is it possible the samples got heavily oxidized? If so, is this possible that some phases got more O-rich than other (in the same sample!)?

Many thanks in advance, and Happy Thanksgiving!

Julien



(*) with my current instrument / ill (?) BSE, I am loosing all necessary contrast to differentiate the different phases in these Pt-Ti alloys.

[Probeman]

(a) What ZAF/phi-rho-z correction would you recommend for Ti-Pt alloys, assuming only pure Ti and pure Pt standards are available? Analytical conditions involve Ti Ka and Pt Ma, at 15 keV, 100-200 nA (assumes no problem with PHA / high count rate, each standard yield around or less than 100 cps/nA with the selected setup, and almost half less in the unknown => around 10000 to maximum 20000 cps - should be OK?).

Hi Julien,
Well in the absence of a proper secondary standard, looking at all 10 matrix corrections is a good start for evaluating accuracy.

One other thing I can suggest is to perform the analysis using the new Fast Monte-Carlo method as described here:

http://probesoftware.com/smf/index.php?topic=47.msg578#msg578

Because there is a very large atomic number correction in this matrix, Penepma should be able to provide a more accurate matrix correction...  please post these results here, as I would be very interested.

Unfortunately the Ti-Pt binary is not (yet!) one that I have calculated...     I've done about 2/3 of the periodic table but there are still some "holes".  However, I will start that calculation today and it should be done by Monday (110 hours)...

Other than that, because you are attempting to measure trace oxygen (and nitrogen?), you might also want to attempt to measure oxygen (and nitrogen?) in pure end member Ti and Pt as a "blank test".

Oxygen free Pt should be easy to obtain (and maintain!).  Oxygen free Ti will not be so easy to obtain (or maintain!), but you are in luck!  I have a chunk of Ti single crystal grown in iodine vapor at 800C that has around 80 PPM of oxygen bulk!  If you write me at my UofO email I will be reminded to send you a piece.

The problem is keeping the pure Ti surface from oxidizing during polishing and subsequent loading of the sample in the instrument...  not sure what to suggest on that front.   JEOL does have a plasma airlock cleaner as an EPMA instrument option, but I've never seen it "in the wild".

(b) Same as (a), but this time if we further consider the measurement of light element (N and O) at these conditions: which matrix correction to choose (curiously enough, Armstrong yield significantly higher O-content: 0.4% vs. ~0.2-0.3 for all others)?

That is interesting.

The oxygen concentration should be heavily dependent on the MAC table chosen.  Not sure which MAC table is your default but I would try again with the FFAST MAC table in Probe for EPMA as it should be more accurate than the older tabulations.  Please post the same calculation with all 10 matrix corrections but this time using the FFAST MAC table option from the Analytical | ZAF, Phi-Rho-Z, Alpha factor and Calibration Curve Options menu dialog as I would be interested in this also.

Edit by John:  OK I have the Ti-Pt binaries running in Penepma (Penfluor).  It's says 112 hours to go...  I will have to upload the finished calculations on Monday, then re-compile the matrix database and upload everything Monday evening so it should be available for you on Tuesday morning.

[Probeman]

On the problem of polishing a pure Ti standard and preventing subsequent oxidation, I should mention that students here, that prepare oxygen sensitive novel materials, often use a N2 glove box for prep/polishing.

Then perhaps the sample could be coated while still in the N2 glove box using a small coater.  I have a note on my desk to look into what coatings impede oxidation the best, but I haven't had time to pursue this question...

I suspect the answer may lie not only in the coating material and obviously, thickness, but also the coating porosity (deposition method?).

Has anyone looked into this question?

[Julien]

Thanks a bunch Probeman! Wow... what a quick response and reaction! Looking forward to this binary calculation!

BTW, about the aforementioned 10 matrix correction results, I do use FFAST already.

I also did in fact measure oxygen in Pt (not in Ti, though). It is variable in the two Pt standards I have (on in the MAC standard block, another one in my Bence-Albee block - both are "old" and some do not appear very healthy, despite numerous repolishing since I arrived in the lab in 2012). Actually, my initial thoughts were to see the Pt Mz 3rd order line that is around Pt. To my surprise there is a clear O Ka peak that I also observed on a WDS scan (well, probably a composite line of Pt Mz III [expect to be very small... 3rd order and theoretical ~1% the intensity of Pt Ma III] and O Ka)! Close to 0.5 cps/nA even after repolishing twice the standards these past three days (well, without grinding it down, just intensive polishing for >10 min each with 0.2 µm colloidal Al2O3), and recoat it with minimal air exposure, but no luck, still there... (one of the Pt standard improved and "only" yielded 0.3 cps/nA). I start to wonder if indeed a plasma cleaning prior to polishing and prior to loading in the vacuum would be better. Or could this be a contamination during the coating??? I am using the carbon coater from SPI with a classical old rotary pump (=> oil?) and the vacuum reach at the best the low 10-2 mbar.

Here is the plot of O Ka in a Pt standards (initial scan is within each of my 2 standards, other scans are after repolishing once or twice one of these standards). In each cases I used 250 or 200 nA, at 15 keV with defocused beam (20 um) and 3 sec counting time by step of 0.2 mm. O Ka peak is at 109.2 mm. The Pt Mz III should be 1.2 mm on the LEFT of the O Ka (so around 108 mm).



Sorry wrong label on graph... You should read "counts per nA" and not "cps" on the Y-axis; divide the value by 3 to get counts per second per nA...

Now that you raise this oxidation problem, do you suggest to run some kind of blank correction by measuring O in Pt and Ti (or rather correction for contamination)? How can we then assess the "right" contamination level in the unknown? Pt oxidize less likely than Ti, so you would naturally expect a much higher O content in Ti, and my unknowns suggest between 3:1 and 1:1 ratio for Ti:Pt, so there is quite a lot of Ti available to be oxidized... (I have not check the O-content in Ti, but might do this on Monday).

Best and Happy Turkey Day!

Julien

[Probeman]

Now that you raise this oxidation problem, do you suggest to run some kind of blank correction by measuring O in Pt and Ti (or rather correction for contamination)? How can we then assess the "right" contamination level in the unknown? Pt oxidize less likely than Ti, so you would naturally expect a much higher O content in Ti, and my unknowns suggest between 3:1 and 1:1 ratio for Ti:Pt, so there is quite a lot of Ti available to be oxidized... (I have not check the O-content in Ti, but might do this on Monday).

Without looking into it more closely (KLM markers?), it's hard to say, but it appears that you have real oxygen in (or on) your Pt samples. I assume the same oxidation (or adsorbed oxygen/H2O on the surface?), but even more so for pure Ti...  and to make things even worse, there are some Ti L beta emission lines very close to the O Ka line:

For  O ka  WSi60 at  23.9756 angstroms, at an assumed concentration of 0.5 wt.%
  Interference by Ti LB3            at  23.8890 ( 109.655) ( -.39732) =   1560.8%
  Interference by Ti LB4            at  23.8890 ( 109.655) ( -.39732) =    818.1%

So ideally you'd want to perform a quantitative interference correction (a la Probe for EPMA) on your Ti-Pt alloys, but the problem is going to be finding (and maintaining!) pure Ti and Pt standards, not to mention preventing oxidation on your alloy surface.  The JEOL airlock vacuum plasma cleaner might be the only way forward on this front at the moment...

I've been thinking about how one could implement an "in situ" cleaner of some type (H2 plasma?) inside the instrument that would only clean a spot roughly 100 um in diameter (because that is all we need), directly under the beam, but I'm not sure any one is working on such as idea. 

A short wave UV laser or focused UV LED might work for in situ spot cleaning for hydrocarbon removal, but that probably won't help remove oxide layers.

You asked about a type of blank correction for your problem, but I don't think it would work as you would need a matrix match with your unknown alloy for a Ti-Pt standard, which would then simply be a secondary standard.

As difficult as it would be I think we need to find a way to utilize the PFE interference correction but with Ti and Pt standards that are essentially oxygen free (I think I have the Ti standard you need already with 80 PPM of oxygen), but even more important (because we can characterize any bulk oxygen using another method), we need a way to prevent or even better remove the oxidation layer during the acquisition (in situ cleaning).

[Probeman]

Have you quantified the apparent oxygen peak that you show in your Pt wavescan above?  If so, what concentration do you obtain?  Assuming of course that it is oxygen in a bulk matrix...  though it really doesn't matter that much because oxygen x-rays are only emitted from the first 100 nm or so in Pt (8% transmission from 100 nm, 28% from 50 nm).

Pt = Pt1 =  195.09g/mol, Pt 100%
15 keV, 21.45 grams/cm^3
Electron range radius =  .4784372 um
O  ka, at 15 keV, (0.5317 keV edge energy)
X-ray production range radius =  .4766275 um
o  ka absorbed by pt =  11558.87
X-ray transmission fraction of thickness  .1 um (average u/p =  11558.87) =  8.379541E-02
o  ka absorbed by pt =  11558.87
X-ray transmission fraction of thickness  .05 um (average u/p =  11558.87) =  .2894744

[Gseward]

To chip-in on my experiences with the Ti-O issues:

I have to measure O in Ti in some Ti-Al2O3 diffusion bond experiments. Which is obviously fraught with issues. I chose 10kV. Maybe I'll make two passes and  use a lower kV for O in the future.....

I want some way to experiment with correcting the Ti lb interference on O, and Pure Ti always has some O on the surface (and possibly also dissolved), so I ran some models:

I ran a monte carlo simulation of pure Oxygen and a second model of Pure Ti with a 5nm TiO2 surface layer.

Using the predicted intensity of O ka from pure Oxygen and O ka from the layered model I  calculate a  k-ratio for 0 ka. of 0.005.   I used this K-ratio as input to Calczaf. Since the matrix model assumes an homogeneous activation volume, the raw intensity from surface O will be over corrected, as we know. Here is the output:

SAMPLE: 0, TOA: 40, ITERATIONS: 3, Z-BAR: 21.64629

 ELEMENT  ABSCOR  FLUCOR  ZEDCOR  ZAFCOR STP-POW BKS-COR   F(x)u      Ec   Eo/Ec    MACs
   O  ka  6.1812  1.0000   .8178  5.0550   .7408  1.1040   .1360   .5317 18.8076 22449.0

 ELEMENT   K-RAW K-VALUE ELEMWT% OXIDWT% ATOMIC% FORMULA KILOVOL
   O  ka  .00513  .00513   2.592   -----   7.201    .621   10.00
   Ti                     99.990   -----  92.799   8.000
   TOTAL:                102.582   ----- 100.000   8.621


So the intensity from the 5nm layer now looks like 2.6wt% by falsely assuming the O is distributed in the bulk.
Note that the O ka raw intensity has over 600% absorption correction - which gives you a good idea why the wt% can get so high and why a lower kV for O might help!

So what to do? As a first approx. if I make the assumption that the oxide layer is ~ the same on the Pure Ti standard and the Ti metal in the Alumina bond sample (which might be fair, if one polished and coated the samples together), I could use the pure metal as a blank correction.  This would also take into account Ti Lb spectral interference on the O ka line. This would be a reasonable approximation for near pure Ti unknowns perhaps, but I'd like to experiment with an oxygenless Ti material as an interference standard also for comparison. For this I envisage TiB2 TiN or TiC.

Using real probe data:  pure Ti as a standard (99.99)  with no corrections applied for the combination of interference and over-corrected absorption, I measure 5wt% O!!!
The Al2O3 stoichiometry I measure in the Ti-Al2O3 diffusion couples is almost perfect, and Ti as we know CAN dissolve a significant amount of O, which is why (before I though about it) I was inclined to believe my analyses. Note to self: ALWAYS do some calculations in CalcZAF before starting on a new material, regardless of how 'simple' the system might seem.

gareth

[Probeman]

So the intensity from the 5nm layer now looks like 2.6wt% by falsely assuming the O is distributed in the bulk.
Note that the O ka raw intensity has over 600% absorption correction - which gives you a good idea why the wt% can get so high and why a lower kV for O might help!

Good stuff.

I would only comment that (normally) if one wants to *minimize* surface effects (contamination, oxidation, etc.), for a bulk analysis, one should generally use a *higher* keV, as the ratio of x-rays from the surface relative to the x-rays from the bulk will be lower.

Unfortunately, as Gareth points out, because low energy emission lines tend to come *only* from the surface (due to large absorption corrections), you don't have a whole lot of leeway in adjusting the keV.  Here is a plot of K vs. KeV from STRATAGem* for O ka and Ti Ka for 10 nm TiO2 on pure Ti.



Lower keVs do help with sensitivity for low energy emission lines such as oxygen because the overvoltage is getting closer to the "optimum" value of 2-3 times the shell excitation energy, as the beam energy is lowered from typical beam energies.

It would be interesting to perform this measurement (using multiple keVs), on the iodine vapor grown Ti metal I have after it is freshly polished and carbon coated within an N2 glovebox...  or does anyone out there have a JEOL airlock plasma cleaner on their EPMA instrument?

* © Copyright 1993-2016 SAMx

[Gseward]

Some plasma cleaners rely on O radicals for cleaning. I'm guessing one might want Ar in this case!!

[Probeman]

Some plasma cleaners rely on O radicals for cleaning. I'm guessing one might want Ar in this case!!

I'm not an expert in these matters, but my impression is that Ar plasma cleaners work for hydrocarbons and oxides (and more depending on the plasma energy), while O plasma cleaners are more specific to removing only carbon and carbon compounds...?  And may induce oxidation in some materials as you say.

[Probeman]

Edit by John:  OK I have the Ti-Pt binaries running in Penepma (Penfluor).  It's says 112 hours to go...  I will have to upload the finished calculations on Monday, then re-compile the matrix database and upload everything Monday evening so it should be available for you on Tuesday morning.

This morning there is still 30 hours to go, so the Monte-Carlo probably won't be finished and incorporated into the matrix.mdb until Wed or Thursday- sorry.