Too high rutile totals

[Dominik]

I am measuring small rutile inclusions in mostly titanite. I first did with 20 kV, as we are interested in trace elements. The totals were a little high, but almost OK, but we often got mixed analyses from the adjacent titanite. So I switched to 10 kV, less influence by the titanite, but the TiO2 is up between 2-4 wt% (net Ti cps are up 4-6%) compared to the standard, which always gives a fine ~100 wt% TiO2. The rutile is in the µm range, no idea what the problem is. Some effect due to the grain size?
I use 120 nA current (for both 10 and 20 kV), again to get a good signal of trace elements.
I use a JEOL 8530F-Plus and the standard JEOL software.

Thanks, Dominik

[Ben Buse]

Hi Dominik

What standard are you using? What's the count rate for Ti on your spectrometer, is it running into deadtime >60k on either standard or TiO2

[Ben Buse]

If it's worse at 10kV also need to think about are both standard and unknown coated with same thickness of carbon. If a pure metal standard does it have an oxide layer at surface.

[Dominik]

I use a pure rutile as a standard, and then also measure rutile. I am using LIFL and Ti Kb, so I don't have dead time issues. Net counts are around 1650 cps. I thought about coating, but even between grains on the sample, the variation is up to 2-3 wt%. And I use spot mode because of the small grains.

[JonF]

If it's worse at 10kV also need to think about are both standard and unknown coated with same thickness of carbon.

What Ben said. It can make a surprisingly large difference - easily a few percent - and it gets much worse as you go to low overvoltage.
Second easiest thing you can do is to repolish and clean both the sample and standard, and then recoat them both at the same time to ensure the coat thickness is the same.

First easiest thing is to question whether you really need to measure Ti in TiO2...
If you're interested in the trace elements, then the TiO2 can just be added as "Formula by difference" in Calculation Options in Analyze!

Just to check - are you determining oxygen via stoichiometry? If not, the MAC of O in Ti matrix is MASSIVE.

[Probeman]

I am measuring small rutile inclusions in mostly titanite. I first did with 20 kV, as we are interested in trace elements. The totals were a little high, but almost OK, but we often got mixed analyses from the adjacent titanite. So I switched to 10 kV, less influence by the titanite, but the TiO2 is up between 2-4 wt% (net Ti cps are up 4-6%) compared to the standard, which always gives a fine ~100 wt% TiO2. The rutile is in the µm range, no idea what the problem is. Some effect due to the grain size?
I use 120 nA current (for both 10 and 20 kV), again to get a good signal of trace elements.
I use a JEOL 8530F-Plus and the standard JEOL software.

Thanks, Dominik

I would definitely suspect a grain size issue with that much beam current such a small beam.

You might try with the Ka line with the log dead time correction. The log dead time correction should just fine even at high count rates and the Kb line may be less quantitative.

If you run at lower beam currents, do you get better totals?

[Probeman]

If it's worse at 10kV also need to think about are both standard and unknown coated with same thickness of carbon.

What Ben said. It can make a surprisingly large difference - easily a few percent - and it gets much worse as you go to low overvoltage.
Second easiest thing you can do is to repolish and clean both the sample and standard, and then recoat them both at the same time to ensure the coat thickness is the same.

First easiest thing is to question whether you really need to measure Ti in TiO2...
If you're interested in the trace elements, then the TiO2 can just be added as "Formula by difference" in Calculation Options in Analyze!

Just to check - are you determining oxygen via stoichiometry? If not, the MAC of O in Ti matrix is MASSIVE.

I agree with Jon. No need to measure Ti (or O) if you only care about traces. Just do TiO2 by difference, if you're sure it's rutile. 

But I would check for secondary boundary fluorescence from the titanite host phase as it would be easy to excite Ca, Si, etc. Run the Standard app and model it using PENFLUOR/FANAL as a simple boundary just to check for the possibility.

Running it as an inclusion geometry in PENEPMA is more work, but the PENFLUOR should be helpful.

Finally, always check your trace accuracy by running your (99.99%?) TiO2 standard as an unknown using the same conditions as your unknown samples. Then apply a blank correction if necessary.