All Now Mysterious...

Thursday, June 20, 2019

If A Tree Falls In Liberty Park.... (Part II)

Okay, so we’ve established what we want to do: We want to check the rings of a fallen tree for lead to see if the groundwater lead levels have changed significantly over time, specifically as a result of the use of lead additives in gasoline in the mid-20th century. And we have an instrument, the ICP-MS (inductively coupled plasma mass spectrometer) that can detect the concentration of trace metals like lead to something like a few parts per trillion. That’s good.

Mostly.

The thing is, at that level of sensitivity, such an instrument can start picking up traces of lead that come from sources other than the sample. For example, the acid used to process the sample might have tiny (but detectable at this level) impurities of lead. The containers used to process the sample may have lead contamination from prior use. Even dust in the air might contain enough lead to be detectable. This is problematic because it can lead to a false result, showing more lead than is actually there.

First, a quick bit of terminology is in order. In instrumental analytical chemistry, the reading from the instrument that shows the true amount of the substance we care about in the sample is called the ‘signal’.  The reading for the amount of that same substance that comes from sources other than the sample--processing, contamination, the environment, etc.--is referred to as ‘noise’. To be confident of our results, we need to demonstrate that the reading we get from the instrument is primarily signal, with little or no noise--in other words, we want a strong signal-to-noise ratio.

What is a good signal-to-noise ratio? Depending on the test, you really don’t want more than 5% noise. 1%-2% would be even better, of course, but in many cases, you can live with 5%. Knowing that 95% of your reading is due to the sample and nothing else gives you a certain level of confidence, statistically speaking, that what you’re looking at is real. If the noise gets much higher than that, you begin to lose confidence in your results.

So, how do we know how much of the result is signal and how much is noise? We use a chemical blank. For example, I recently prepared three samples of what’s called a Standard Reference Material (or SRM, in this case, NIST SRM 1515, Apple Leaves) for analysis by placing each sample in a clean, dry crucible and baking it in a furnace overnight at temperatures up to 550°C (~1025°F). Once it cools down, I will take the residue and dissolve it in a mixture of hydrochloric acid and ammonia, and then analyze this solution for lead content. Now, as I prepared and processed these three samples, I also placed three empty crucibles into the furnace. I will also add the hydrochloric acid/ammonia mixture to them and then analyze those samples for lead content. In other words, the empty crucibles will undergo the same processing as the real samples. The empty crucibles are the chemical blank; they tell us (at least in theory) how much lead comes from the environment and not from the sample. They tell us the level of noise.

So when the preliminary samples were processed and tested months ago, a chemical blank was made and tested at the same time. And the results were very interesting. The bark of the tree showed extremely high levels of lead. Well, that’s not entirely unexpected. The bark can pick up lead from the air and soil around the tree as well as the groundwater. The interior rings in the tree showed lead concentrations ranging from around 11 parts per billion (remember, 1 ppb is roughly a paper clip in a swimming pool) to around 30 ppb. That’s not a lot. However, the chemical blank showed a lead concentration of around 2 ppb.

2 ppb of noise compared to 11 ppb of signal is around 18%. That’s way too high.

Furthermore, it’s not just the concentration of lead that matters, it’s also the actual amount. Once we identify how much lead is in the sample, we also want to try to isolate it for further testing (specifically isotopic testing, which I’ll tell you about another time). To do that, we need a certain minimum mass of lead to work with. Now, going back to the definition: in a water solution, parts per billion effectively means means micrograms of the substance we care about (again, lead in this case) per liter of water, or nanograms (0.000 000 001 g) per milliliter of water. Since these samples are almost always 10 mL in volume (because the containers can only hold that much), that means that even the 30 ppb sample contained only 300 nanograms (0.000 000 300 g) of lead. We can run isotopic testing on samples that small, but only if they’re very pure--and remember, thanks to the chemical blank, we already have an uncertainty of 20 ng/300 ng, or about 6.7%, in our purity. That won’t give us results that mean anything useful.

So, what do we do about that? If the processing adds that much lead (and uncertainty) to our results, we’re going to need a different way to process the samples.  I’ll start telling you about that in Part III.

0 Comments:

Post a Comment

Subscribe to Post Comments [Atom]



<< Home