[dwebb210]

I guess I have to admit defeat.

I work at a metallurgical lab, and the latest puzzler is
a bunch of aluminum samples that need to be analyzed for
strontium content.

Our optical emission spectrophotometer can't detect
strontium, so I have to digest the sample and run it
on a flame AAS.

The aluminum has around 8% silicon.

I don't have metal chip standards with known strontium values.

I do, however, have a 100ppm multielement standard, which
includes strontium.

I know aluminum and silicon both interfere with strontium
analysis, so matching the matrix is important.

I took six flasks and weighed out one gram of chip standard
into each.  I added 0.1, 0.2, 0.5, 1.0, and 5.0 milliliters
of the multielement standard into five of the flasks, and
left one as a "blank".

After digestion, the solutions were diluted to 100mls,
and filtered.

I was essentially doing a "standard addition" test, because
not only did I end up with a set of standards, I was able to
calculate how much strontium was in the metal chip standard.

Or so I thought.

I digested the 6 standards in nitric acid and hydrofluoric.
The reason was because I wanted to save time and avoid
the need to filter the samples if possible.
The use of HF got the silicon and titanium into solution...
but a white precipitate formed which needed to be filtered anyway.

I digested one of the unknown samples the same way.

Realizing I wasn't going to save any time by using HF,
I tried digesting another 6 standards with the Sr spikes,
only this time instead of HF, I used nitric and HCl.
I filtered out the undissolved Si and Ti.

I digested the same unknown sample this way too.

So I have a set of standards with HF, and a sample with HF.
And a set of standards with HCl, and the same sample with HCl.

I ran everything on the AA.

I corrected for the weight of the metal chips,
I corrected for what appeared to be strontium in the chips,
I'm sure I did the calculations right.

I got linear calibration curves for both sets of standards,
although the instrument response was roughly three times
greater for those digested with HF.  
The correlation coefficient for both was >0.999

I expected a difference in instrument response,
and I was happy that I got a linear calibration curve for
both types of dissolution.

HOWEVER, the sample dissolved with HCl calculates to 0.011% Sr,
while the sample dissolved with HF calculates to 0.028% Sr.

Strontium is extremely important for this alloy, and a difference
as large of this is unacceptable.  

Where did I go wrong?

The only thing I haven't done yet is digest another two
samples to verify the values.  There is a remote possibility
the drill hit an inclusion that contained high strontium,
and the drilling made it into one flask but not the other.

Does anyone have any suggestions?

Thanks

Dave

[Borek]

Please post in appropriate forums, this is analytical chemistry question.

Perhaps you are loosing part of strontium with precipitate? Which samples had more precipitate/unsolved material filtered out? These digested in HNO₃/HCl? If so, that could explain weaker signals in these samples.

Just a wild guess.