I wonder how their Mass Spec testing compares. they have them on sale and $30 per kit seems like a good deal if the data is any good.
We discussed (@taricha, @Rick Mathew ) this question, but I only recall one idea. If the ICP portion of the analytical method is a source of variability, whether you measure light emission or detect ions, the ICP-X results still suffer from variability of the ICP method.
Yep, Kstatefan. I agree with Dan's point here, but I'd go further.
Some of the things that ICP-A gets wrong are things that aren't due to limitations of ICP-OES. They botch salinity sometimes - we ran across several big data sets where they reported all 5 major elements below the values for every other vendors for each element. Randy's pointed out cases where they badly botch Ion balance. They report a ton of values that are less than their uncertainty for "maybe it's really zero". They report results for some elements that are very poorly measured by ICP-OES in general (Rb) and don't tell you how huge their uncertainty is. They sometimes have complete misses (both high and low) for P and I, that other vendors do not have.
Switching to MS vs OES doesn't really address any of those above issues.
Said another way, I have no interest in their ICP-MS results based on looking carefully at what they've done with ICP-OES. Not until they demonstrate they can at least do as well as the rest of the hobby with ICP-OES. If I want an ICP-MS test, it'll be from somebody who's done ICP-OES well.
But what isn't anecdotal evidence, is the fact they [Triton] report out to the hundredth decimal place but all of the values end in .00. The only reason I can see for this is to try to "trick" the customer into thinking the precision is higher than it actually is. I sent them an email asking for clarification on this topic and they stated their system is not designed to deliver results to that level.
I'll defend Triton on this point. I'd say everybody is looking at 1ppb at best (with few exceptions), and all the decimals are mostly decoration.
Triton reporting X.00 for all elements is more transparent than vendors that report numbers after the decimal when they have no such precision. Triton in data sets that I've seen is the stingiest ICP vendor in reporting a number for a trace element. I mean that they report the fewest numbers that "could really be zero" of any vendor. It seems their quality control threshold for deciding if it's really detected or not is slightly higher than others (a good thing).
They are the opposite of ICP-A in this respect. ICP-A will generate a number for almost every element no matter what.
