Non-linear calibration GCMS

[AICMM]

jdlh199,

Is 353 the only ion? If so, one thing I would look at is cutting way back on the EM. You say that the peak is way over the noise so you have plenty of signal. My concern would be that you are approaching saturation although you curve does not have the shape I would expect. Having said that, ppm levels are on the high side for SIM analysis, especially if only looking at one ion.

If you are doing a confirmation ion an easy test would be to calibrate using that (assuming it is less than 353) and see what that looks like.

Best regards.

[jdlh199]

Increasing the transfer line probably isn't the solution, since the analyte peak elutes ~60% of the way into the run. There are two analyte peaks which are diastereoisomers, both are non-linear. The remainder is used to get rid of matrix components that show up in ECD.

It's a standard DB5 at 30mx0.25mmx0.25µm at 1mL/min helium flow rate.

353 is the ion used for quantitation. 181 (PFB) and 311 (derivitized ZDDP with no alkyl chains) are used for confirmation. Both are also non-linear.

I agree ppm levels are high for GCMS, but I wanted to run an extended calibration curve from the lowest level detectable by GCMS up to the highest level that is currently run by ECD to make sure it wasn't just that we were looking at a non-linear concentration range by MS. The MS is non-linear across the entire range. The ECD is linear across the entire detectable range.

More head-scratching....

[jdlh199]

OK, here's yet another (quick) update....

I finally managed to get some time on another instrument. This second instrument (which is the same manufacturer/model) consistently produces a linear calibration line, while it's consistently non-linear on the first.

So to me it looks like a technical issue with the first MS, but this technical issue only affects certain analyses (especially dervitization/methylation methods). I've managed to reproduce the non-linearity with an entirely separate assay for EDTA that uses methylation with BF3. But linearity tests with Pentadecane or OFN as suggested by the manufacturer's PQ testing all work fine giving linear responses...?

Any suggestions? This is driving me nuts...

Both machines are just autotuned and then analysis started. Samples were prepared and aliquoted into two sets of vials and run simultaneously on the two instruments.

[Don_Hilton]

I assume both instruments had the same stationarly phase and type of inlet liner?

If so, the next question is: What about cliping a bit off the front end of the column in the instrument in question and did you try a fresh liner, seal plate and all that in the inlet. (I'm still guessing active sites somewhere.)

[Peter Apps]

Why do you say that the MS is non linear across the entire range ? You only have 3 points at 10 ppm and above, and they are in a straight line (http://i41.tinypic.com/2d0ifcm.jpg). If you want to diagnose the range over which the MS response is non-linear you have to have points spread evenly along the calibration curve, not clustered at the lower left.

If I understand correctly you first dilute your analyte solution, and then you derivatize ? If this is so it is possible that some of the non-linearity is wet chemistry and not instrumental. That you get straight lines with other detectors and instruments can be an artifact of repeatability (have you looked at repeatability ?, this should come before linearity). So, make up a stock solution of analyte, derivatize it, and then dilute it - that way the reaction chemistry is uniform for all the levels of standard.

The issue of serial vs point dilutions has still not been addressed.

Peter

[Don_Hilton]

I think the key here (if I read jdlh199's last post correctly) is that the curve run on a second instrument showed linearity across the range. Which raises the question of what is different between instruments.

The spacing of points shoud not be an issue. The traditional spacing of something like 1, 5, 10, 50, 100... leaves me more comfortable as the appropriate weighting for regression for data with increasing RSD's at lower concentration is 1/x - and, as the lower points are weighted more heavily, I'd rather have more of them to average out noise.

[Peter Apps]

Hi Don

We have not seen the calibration from the other MS, but the curve from the problem calibration is gentle enough that quite minor deviations due to repeatability problems (either in standard preparation or e.g. inlet performance) could turn a gentle curve into what looks like a straight line with the points scattered on either side of it.

I was questioning only the claim that the MS is non-linear throughout the range - by the sophisticated technique of holding a ruler against the screen the three points at 10 ppm and above are in a dead straight line. This could be confirmed (or denied) by running another couple of points in between. The range of the non-linearity is an important diagnostic feature - if it is throughout the range it is not likely to be due (only) to adsorption, which is otherwise the most likely explanation.

Peter

[Don_Hilton]

Peter,

Your concern with three points falling close to a ruler, I understand quite well. I had missed that you wanted to add additional points into the high end of the curve. I agree that this would show linearity or lack thereof at the higher end. I occasionally run my calibration points in replicate, measuring the variability as a function of concentration. Nothing like getting a nice straight line that is as wide as a city street.

I agree that there is a lot we don't see here. My thoughts are that if the results on the other instrument are suitable - as described, let's see what the difference is. If all eight points, generated by injecting the same solutions, are falling along a ruler on the second instrument, we could be headed in the right direction. I don’t want that line of exploration to get lost.

[Peter Apps]

Hi Don

There should be a rule of troubleshooting (to go with don't fix what ain't broke etc) along the lines of "step don't jump" !!

Peter

[jdlh199]

Thanks for the comments.... Let me try and go through the points one by one.

Yes both instruments had the same stationary phase and inlet. I've tried clipping columns, replacing liners, seals etc etc. I'm almost to the point of separating the two GCs from their respective MSs and switching them over to demonstrate the problem is with the MS and not the GC side

I understand that if you take certain points a few may look linear. Bear in mind these are not the only curves/analytes I've run. I've been scratching my head with this for a few weeks now. However, I have stuck to the 1, 2, 5, 10 distribution rather than the 2, 4, 6,8, 10 distribution. Whether I run high or low conc curves, I see the same thing.

I've ruled out the derivitization procedure (I think) by preparing a calibration curve that is diluted then derivitized, and another which is a stock derivitized, then diluted.

This is appearing more and more with different analytes, although it only seems to affect the analytes I actually NEED to work....!!! It's affected the ZDDP assay, an EDTA assay (methylation), a FAME assay and most recently is now appearing with 2-bisethylhexyl-phosphate. All of these are non-linear on the first GCMS and linear on the second. But OFN and Pentadecane is linear on the first GCMS..!!!

I prepared a calibration curve of BEHP in duplicate from 0.1 to 5 µg/mL in DCM. This is prepared from a point dilution of a 10µg/mL working stock solution in DCM, i.e. dilute 0.5mL of the stock with 0.5mL of DCM to get the 5µg/mL, dilute 0.2mL of the stock with 0.8mL DCM to get the 2µg/mL etc etc. One curve was run on the first, another on the second.

Good GCMS: http://i43.tinypic.com/wkh7np.jpg

Bad GCMS: http://i43.tinypic.com/ridjx3.jpg

The bad GCMS is older, which is why I'm wondering if something is old/dying/needs replacing. The second is a newly acquired, but used, instrument, of the same make/model.

Right now I'm method developing on the bad GCMS and running samples on the good one. This is obviously not sustainable.

More comments please....

[AICMM]

jdlh199,

Returning to one of my favorite subjects.... Cut your flow rate down to about 0.7 mL/min and try the curve again. I have seen MS systems (esp. diff. stacks) that are non-linear due to a lack of pumping capacity. This could affect the old more than the new because the old has had the column trimmed enough to not really be a 30 meter and thus the calculated flow is less that the actual flow. I found this issue on the PAH's in environmental, relatively heavy components in that analysis. Should be relatively easy to try.

Best regards.

[jdlh199]

Running right now at 1mL/min and then at 0.7mL/min.

Check back later for an update.

Thanks!

[Don_Hilton]

How do peak shapes match up between the two instruments for the same low level standard? Does one tail any more than the other? Are they integrating the same way.

And a question that is a bit of a jump ahead, but I'll ask: In old instrument vs. new instrument: Besides age, any difference, like length of time since last cleaning? Or one has a newer inert source?

[jdlh199]

Peaks shape is pretty much identical between machines. They are integrating the same way.

The older one has had a major PM in the last few weeks, including having the quad and source pulled and cleaned (since I was trying to rule that out). I've also swapped the inner source between the two MS's and that made no difference.

I suppose I could theoretically pull and swap the entire quads over...

The newer one does reach a lower vacuum level according to the Penning gauge readback (~8.5x10-6) compared the the older one (~1.3x10-5), although the older one is still below the manufacturer's recommended minimum of 3x10-5. The newer one also pumps down a lot quicker, although I put that down to a sexier newer roughing pump?

[Don_Hilton]

You say "pretty much identical." If pushed to say how they differ, what would you say? And, is that difference more pronounced in the low level standards than the high?

Looking at the vacuum readings – without knowing what kind of instrument you are running (if you posted it, I missed it) a couple of things cross my mind. (And this is based on what happens with that kind of change in my instruments) 1) the true column flow differs between instruments – which could be a problem. Or 2) I would be tempted to look for an air leak in the older instrument.

With the instruments that I run, a difference like you describe is a red flag – but that’s the instruments I run.