GC-MS help request: bad accuracy, sensitivity loss, split pe

[kristtorn]

I'm requesting help with troubleshooting recent problems on our Agilent 6890N w/ 5973 quadrupole mass analyzer. Were extracting a highly-lipophilic pesticide from animal tissue, but have also experienced the same problems with standards prepared in solvent. The method has not changed since these problems began. I'm detail the problems and things we've tried this far, but we're pretty stumped at this point so any suggestions are greatly appreciated.

Issues:

-loss of sensitivity. Our LLOQ of 0.3 ng/ml is no longer visible in samples, nor in pure standards. I'll note that we perform regular maintenance on the inlet and source.

-terrible accuracy (precision isn't great but not bad). Our validation attempts have strangely had really great linearity of the cal curve (ascending order, first 10 vials), but the randomized QCs have wildly different RRs throughout the batch. We've run standards at the same concentrations and in the same order and, bizarrely, had the same issue of fantastic cal curve linearity but terrible accuracy of QCs. It's safe to rule out pipetting error for this issue.

-split peaks. Our analyte has split into 2, even 4 peaks sometimes. The internal standard occasionally has the same problem, but these issues only happen for a couple samples of each batch, including the recent "validation" of standards. The analyte abundance didn't decrease in that run. If anything it increased slightly. Compared to last month's performance, however, the abundance of internal standard has vastly decreased.

Things we've tried:

-we perform regular maintenance on the inlet, including putting in a fresh liner, trimming column, replacing gold seal, new ferrule.

-the column has been replaced but is the same kind we've used for the past year without these problems.

-the source has been cleaned. Many times.

-reagent gas: we've tried switching from methane to ammonia, but same problems.

-we're presently cleaning the split line, which is kind of gummed up, though we're operating in splitless.


If anyone has a suggestion I'd love to hear it. It must be a machine problem since standards produce the same issues, but we just can't figure out what. Thank you!

[Steve Reimer]

Are the QC samples run through the same prep procedure as the samples while the standards are made up in fresh solvent? If so could there be a subtle difference in the solvent on injection, e.g. water or polar solvent content, that affects the inlet conditions enough to lose analyte?
You haven't said whether you are doing a splitless injection, direct on column or PTV. With a 0.3 ng/mL low point I doubt you are doing a split injection. If you can't see that level anymore is your calibration still as linear as ever?

[kristtorn]

Yes. The samples, calibration curve, and QCs are extracted in the same manner by spiking the tissue with analyte, extracting with acetonitrile, introducing sorbents, then filtering and evaporating to dryness prior to reconstitution and analysis. The standards, however, were prepared in solvent and simply evaporated prior to reconstitution with toluene.

Though we change out the inlet parts after every batch, this method produces really clean samples with not much apparent buildup in the liner.

Yes, splitless. And the calibration curve has excellent linearity, as good as ever. The curve is run on the first 10 injections, then a toluene injection, then QCs. So the first really inaccurate qc is only 2 injections away from the ULOQ of the curve. That's the part that puzzles me.

[James_Ball]

Yes. The samples, calibration curve, and QCs are extracted in the same manner by spiking the tissue with analyte, extracting with acetonitrile, introducing sorbents, then filtering and evaporating to dryness prior to reconstitution and analysis. The standards, however, were prepared in solvent and simply evaporated prior to reconstitution with toluene.

Though we change out the inlet parts after every batch, this method produces really clean samples with not much apparent buildup in the liner.

Yes, splitless. And the calibration curve has excellent linearity, as good as ever. The curve is run on the first 10 injections, then a toluene injection, then QCs. So the first really inaccurate qc is only 2 injections away from the ULOQ of the curve. That's the part that puzzles me.

Do you see any carryover in the toluene shot after the high calibrator?

If so it could be the needle or inlet has something retaining a little of the compound. It can make a quadratic curve look linear or a linear curve look quadratic but fit very well if the carryover is consistent. I have a similar problem on our ICPMS for Antimony, the carryover gives us a great calibration, but once you flush and run a QC it quants out low. If you place a blank between each calibrator, then the curve look very bad.

That split line will cause problems even if you are injecting in splittless mode. If you think about molecules in high temperature gas phase, those molecules are moving around very fast and will even escape out to the split vent the back into the liner (at least some will) before they make it on the column in the first few seconds after injection. I never thought it would effect it until I found that cleaning the split line stopped breakdown of pentachlorophenol when doing splittless injections.

Another question, what temperature are you running your source?

[Peter Apps]

I have also suffered a progressive and dramatic loss of sensitivity doing isobutane CI in a Varian/Bruker 320 MS. As I write it looks as if the cause was contamination of the ceramic base of the filament assembly with carbon. This conducts away the current that should be going into the source and ionizing stuff. The filament failed the on-board emission current diagnostic test, if your Agilent has something similar it might be worth running it, or even just swapping out the filament. Switching to the unused filament runs the risk that it might also be covered in CI soot.

My favourite cause for erratic peak splitting with a splitless injection is solvent condensing on the column - is your column starting temperature below the solvent boiling point ?

Peter

[kristtorn]

James, we usually don't have carryover problems in the toluene following the uloq (500 ng/mL), or in the tol injections throughout the batch. We do, however, see some in the first tol run following maintenance, so we've been running 2 tols prior to batches to be safe.

Good idea with the split line... it's just been cleaned (it was cruddy), and machine is now pumping down... so I'll be able to give you a better idea of how well that worked in a few hours.

The source is at 150.


Peter, the filament is changed roughly every few months, but the diagnostics are definitely worth looking into.

And EEP! That's a really great point - we're holding the oven at 80 for 1 min, but tol BP is 111. Changing this to 115. Thanks a ton for that one!


I really appreciate all the help so far. I'll post an update as soon we run some standards after these fixes.

[James_Ball]

James, we usually don't have carryover problems in the toluene following the uloq (500 ng/mL), or in the tol injections throughout the batch. We do, however, see some in the first tol run following maintenance, so we've been running 2 tols prior to batches to be safe.

Good idea with the split line... it's just been cleaned (it was cruddy), and machine is now pumping down... so I'll be able to give you a better idea of how well that worked in a few hours.

The source is at 150.


Peter, the filament is changed roughly every few months, but the diagnostics are definitely worth looking into.

And EEP! That's a really great point - we're holding the oven at 80 for 1 min, but tol BP is 111. Changing this to 115. Thanks a ton for that one!


I really appreciate all the help so far. I'll post an update as soon we run some standards after these fixes.

I am not familiar with CI but with EI I run my source at 230c and now with my QQQ I am running it at 300c, seems to keep it cleaner longer. You will see a little more background at higher temps, but if you go to manual tune and scan without tune gas and watch as you raise the source temp you will see what looks like column bleed. It seems the column bleed likes to condense on the source at lower temps. It might be a good idea to run the source bake out built into the Agilent tune menus before each run, just an hour should help.