[Ithakaz]

Hello, I hope I came to the right place, and was hoping someone here might be able to help me out!

We're having an issue at my Lab developing a robust method for analyzing various metals/alloys that we produce for trace metallic impurities via ICP-OES. Its been going on for some time, and I figured it was time to reach out beyond work to see if anyone has any ideas that could help us along.

We produce various precious metals and alloys(Au, Ag, AuSn, AgCu, AuNi etc), and test them for ~44 different metallic impurities. As it stands we have two methods we use, one for gold based alloys and one for basically anything else., but we have a big problem with wavelength selection/interferences and getting results to agree with our more accurate instruments (GDMS). We would also like to eliminate any user subjectivity from our techs.We use Thermo ICAP 6000 series instruments running iTeva software and Our method is roughly as follows.

Digest 0.2 g of our sample in 2 mL of Aqua Regia, and dilute to 10mL with DI water, these are run with a correction factor of 50.

which is run against a calibration of

Blank: 50:50 Aqua Regia:Water
Standard: 5ppm of each element we are analyzing for in 50:50 Aqua regia:water

all using 50 microliters of Yttrium as an internal standard

The method is not bad for analyzing pure metals such as Silver, but our more complicated alloys give us a lot of headaches with false positives and negatives, across various wavelengths.

I'm of the opinion that we need specific methods for each alloy group (ie: AuSn, CuAu, AgCu, AuNiPd), etc) because each will have unique interferences that must be dealt with. Some other ideas we've been looking at

-Digesting more sample to meet the intensity of our standard, and get rid of our 50x correction factor

-Matrix matching our blank and standard to our sample, (so if we are running CuAu, making our blank/standard in a CuAu solution of the same ratio as our sample, instead of just Aqua regia:water

-Utilizing method of standard addition for each sample instead of a normal blank/standard (very time consuming)

-splitting up our standard into multiple standards so the elements don't interfere with each other in the calibration


We've shied away from IECs due to how many elements we are dealing with, but it is still an option.

I know this is slightly vague, and ideally one of our three senior chemists should really be the ones figuring this out, not us lab techs, but I'll answer any questions as best I can. If anyone has had any experience running for trace metal impurities on these instruments, let me know!

[Corribus]

I think it would be more helpful to know the nature of the "interferences". Chemical interferences usually aren't too huge of an issue in ICP-OES, because most elements have at least 2-3 good lines you can use for analysis. The corrective strategy in any case depends on the nature of the interference. Certainly matrix matching is always a good idea, and so is using more sample, although these won't necessarily help with interferences from other species in your sample, because you'll be amplifying the signal from all of them by the same factor. Standard addition is a good technique to use but won't necessarily help either if the interference is coming from the sample; it is more appropriate for samples with high backgrounds or matrix effects (is this what you mean by interference?).

Anyway, if you provide a specific example of the interference you are referring to, more specific solutions may be forthcoming. You also need to keep in mind what your threshold for detection is. Is it possible you are simply operating below the sensitivity of ICP-OES for this application? If so, you may need to move to ICP-MS.

[Ithakaz]

I think it would be more helpful to know the nature of the "interferences". Chemical interferences usually aren't too huge of an issue in ICP-OES, because most elements have at least 2-3 good lines you can use for analysis. The corrective strategy in any case depends on the nature of the interference. Certainly matrix matching is always a good idea, and so is using more sample, although these won't necessarily help with interferences from other species in your sample, because you'll be amplifying the signal from all of them by the same factor. Standard addition is a good technique to use but won't necessarily help either if the interference is coming from the sample; it is more appropriate for samples with high backgrounds or matrix effects (is this what you mean by interference?).

Anyway, if you provide a specific example of the interference you are referring to, more specific solutions may be forthcoming. You also need to keep in mind what your threshold for detection is. Is it possible you are simply operating below the sensitivity of ICP-OES for this application? If so, you may need to move to ICP-MS.

Thank you for getting back to me! With interferences I am mostly speaking about spectral interference, Like, an Arsenic line showing >100ppm when running a sample who has Palladium as a major component, even though our GDMS results show nothing for Arsenic. But also getting results that just don't match up even without interferences involved, such as a CuAu sample showing ~50ppm Fe on the GDMS, and only 4 ppm on the ICP.

I think a lot of it is stemming from the fact that the samples just don't behave the same way as our standards we we analyze them. For example I can calibrate our method with a 5ppm solution of every element we are looking for, and when I run separate solutions made up from our standards, the results are spot on at many different levels, (1ppm, 5ppm, 50ppm, etc). But if I run a sample(who results we can confirmed on a GDMS) on that same calibration, the numbers are way off from what the GDMS says they should be.

We do have a ICP-MS, but the preferred option is to be able to keep these analyses on ICP-OES

[Corribus]

Strictly speaking, a spectral interference refers to a partial (or even full) overlap of your target line by the line from another element. Common elements like iron can cause annoying spectral interferences because they are abundant. The best solution is to choose another line for your target element if one exists of suitable intensity. If one doesn't exist, most good instruments supply line fitting algorithms in their software to correct for overlaps, although I haven't personally used them. I assume the iCAP has something like this (I have an Agilent instrument).

You gave pretty specific concentrations of the interferences. Are you running calibration standards for all the elements and possible interferences simultaneously? Another way to reduce many types of interferences is to adjust the sample inlet parameters (plasma temperature, etc.), but this is less effective when you are trying to measure many elements simultaneously because while one element may improve, others can become worse. If interferences and sensitivity are broadly at issue, you may have better luck if you develop different methods for each element of interest and run the samples multiple times, one time for each element of interest. This can be time consuming but may yield better results. I do not do a lot of broad surveys of unknown samples so I don't have great experience dealing with these types of issues. You may consider reaching out to Thermofisher's applications team - those people are usually pretty helpful if you can pin them down with very specific questions.

Regarding why your results don't match well with GDMS: I couldn't provide much wisdom here because this is not a technique with which I have much familiarity. Ideally two different techniques should give the same results, at least for primary analytes. Unless you are carefully measuring the concentration of your interfering element with a full set of standards and conditions meant to optimize the emission response, it may not be surprising that you get very different concentrations using these two method. If you really want to validate your ICP-OES method, I suggest making up some pure "unknowns" using the target substances (dilute your stock solution of, say, gold to some random concentration) and run them on both instruments using a full range of concentration standards, and see what kind of reproducibility you get. At least this way you know the instruments are behaving properly, and the differences are something related to the samples.

Keep in mind that some substances have peculiar behavior that may affect your results. Gold, for example, is notoriously sticky to glass sample introduction parts and tubing. Each element has its particular challenges and you should do your legwork before attempting analysis. This may also explain differences between the techniques. Total dissolved solids (matrix effects) can also lead to skewed numbers, although ICP-OES is pretty tolerant of this.

Finally, I am not sure what you mean by "calibrate our method with a 5 ppm solution of every element we are looking for". Are you running a full set of calibration for all of these elements? How many elements? Although ICP-OES is frequently advertised as a "simultaneous measurement technique" in product brochures, keep in mind this really only refers to the fact that it can acquire spectral data for a virtually limitless number of elements simultaneously, because the instrument captures spectral data over large spectral windows. This does not necessarily mean that accurate quantitative measurement of a limitless number of elements simultaneously is recommended or even possible. Any given set of conditions will favor some elements at the expense of others. As mentioned above, you may have to settle for narrower experiments tailored to smaller sets of elements. In my lab we routinely analyze 3-4 elements at a time, but even here it can be challenging to thread that needle of getting adequate sensitivity for each element using a single experiment. The sweet spot gets rapidly smaller as the number of target elements increases. This is why analysis of unknown environmental samples is such a challenge! The method development aspect can be brutal.

Hope any of those random musings helps.

[Ithakaz]

MOD Edit: delete monster quote

Thank you again! Its nice to just get an outside perspective, as our chemists here are pretty stuck in their ways, changing methods/techniques is always a struggle.

In response to your questions on what I meant by "calibrate our method with a 5 ppm solution of every element we are looking for", I was referring to the fact that when we calibrate our instrument we use a Blank, that is just Aqua Regia and Water, and a solution that we call simply our "High Standard". Which has 5 ppm of each of 43 different elements in it, prepared from 1000ppm single element solutions. So, to make up a liter of it we use 5 mL Au, 5 mL Ag, 5 mL As, etc. We are then analyzing for all of these elements simultaneously.

So, our calibration is essentially a one point calibration, at 5 ppm, For each element that we are testing for. Does that make it a bit clearer? I think we should break this up into multiple standards, with different element groups, to reduce any interference that may be happening in our calibration itself.

It is generally spectral overlap that is the issue, we've just had trouble narrowing down good wavelengths to use do to the fact that we have to keep 40+ elements in mind at a time.

[wildfyr]

I think ICP-OES is just not equipped to handle 40 elements at a time. That really is a job for ICP-MS. Its design parameters are much better aligned for a job of that complexity.

If its possible to break it down into a few different analyses then perhaps OES will be usable.

I am curious why your two techniques give different results. I think this is a critical tool to look at your issues:

I think we should break this up into multiple standards, with different element groups, to reduce any interference that may be happening in our calibration itself.

[Corribus]

A two point calibration for so many elements? It is no surprise you are having difficulty.

Also you are combining all these elements into a single standard. You have to be careful with this, as you can begin to have matrix interferences because not all elements can be in the same matrix. As an example, we measure gold and silver frequently by ICP-OES, but it is hard to make a single standard that has both gold and silver in it, because of silver's instability in the presence of chloride, and gold standards are usually prepared in dilute HCl. If we have samples that need to be analyzed for both silver and gold, we have to do the analysis separately even though ICP-OES is technically capable of acquiring the spectral data for both elements simultaneously. When you buy multi-element standards from a company like SCP Science, they know all these complex compatibility rules and will tell you up front which elements will work in a single standard. If you are just dumping them all together on your own, this may be a large part of your problem. A simple test would be to measure recovery of your single element standards using your in-home prepared multielement standard (i.e., use your single element standards as samples). See how close you get. This can also tell you which of your elements are the most problematic from a matrix compatibility standpoint.

And use at least a three point calibration!