Riddle me this

[shaun78]

I've got somewhat of a perplexing problem that I am unabel to resolve.

Series of GCs in two different laboratories (mixture of 6890's and 7890's).

Some GCs in some laboratories are able to successfully run a method, while others are not.

There are no problems with blank injections or standard injections. There is never a problem with the first sample injection.

All of the trouble starts after the first sample injection. The problem is a co-eluting peak with the analyte of interest.

I work in an FDA regulated environment and this is a fully validated analytical method that has been in use for over 10 years. We have transferred this method literally 15-20 times and have never observed this problem.

In an effort to troubleshoot I have:

1. Taken column, liner, syringe, septa, and sample preparations from a non-working GC to a working GC and the working GC has no problem.

2. Done the reverse of #1 and the non-working GC has the problem.

3. Have tried different columns, liner, syringes, septa. The analysis works on the good GCs, not on the bad.

4. Replaced inlet, gold seal, FID, split vent copper tubing, and split vent trap on bad GC. Problem presisted, but the interfering peak was about 1/10 of the normal size.

5. In light of #4 suspected contamination of sample contact surfaces within the inlet, so I replaced the reducing nut from the inlet to the column. This caused my problem to regress back to original size.

6. Replaced all carrier gas traps.

7. Replaced FID jet and cleaned ceramics of detector.

As I said, this happens arcoss GCs (6890/7890) and laboratories. I have verified all instrument methods are the same and all relevant parts in the GC are of the same part number.

The question is: what am I missing? What haven't I looked at?

[Peter Apps]

What you are missing is to tell us what it is that you are trying to analyse, and how you are trying to do it.

Assumptions of telepathic powers among forum contributors continue to puzzle me.

Rod's law applies; the more you don't tell us the more we can't help you.

Peter

[Jake]

You've done a lot troubleshooting/maintenance, but it's not clear to me that you thoroughly cleaned the inlet. Maybe I missed it, or you did and didn't list it. The fact that changing the split vent line reduced the contamination 10X suggests a thorough cleanly of the inlet with a wide range of solvents might be beneficial. All your troubleshooting results are consistent with a contaminated inlet. We run primarily FAMEs and find that solvent rinsing the inlet and spit vent line with methylene chloride, followed by methanol and acetone works well. But I have know idea what you're analzying.

As Peter points out, knowing what you're actually analyzing would help.

[shaun78]

Hi Peter and Jake,

Unfortunately I am not able to state exactly what I am analyzing. I've been around these boards enough and helped enough to know the more information that I can give, the better everyone will be.

Having said that, I can shine a little more light, but I can not state exacrly what I am doing. With this being a tried and true validated method with over 10 yers of history, I made the wrong assumption that method specifics were not of significant importance.

As far as the inlet is concerned, I did try cleaning using an approach inline with what you recommended; it did not help either. As a result, I have replaced the entire inlet to no avail. It was the replacing of the entire inlet that temporairly reduced the problem to 1/10 of what it had previously been.

Analyte aside, the analysis is straight forward. Sample is extracted into MeOH and an aliquot is injected on the GC. The extraction is of a multicomponent polymeric material. We are looking at for one of the unreacted / uncrosslinked monomers in the matrix. That is as specific as I am able to get.

Again, initially I did not think the matrix to be of signifigance because some instruments can test the sample without problem, and others can't (and I am talking about the same vial, syringe, liner, septum moved from instrument to instrument). Additionally, the fact that all instrumnets are able to perform one good injection of the sample before problems arise seems to rule out the matrix to me.

This can't be a carry over / incomplete sample elution issue because, again, some instruments don't have a problem.

Anyway, without further ado ...

Instrumental setup:

Injection Temp: 200C
Detection: FID @ 250C, 40 mL/min H2, 440 mL/min O2, 20 mL/min He Makeup.
Oven Gradient: 100C and a fast ramp to 230C (runtime less than 15 min).
8:1 Split injection using standard stright 4 x 78.5 mm liner with deactivated glass wool.
Injection Volume: 2 uL
Column: WAX
Flow: 1.0 mL/min (constant flow)

I realize the injection volume exceeds liner volume by ~2X without even considering the volume the gas takes up. Sure, it's a bad practivce, but systems are able to run this with acceptable precision and without this issue.

I also realize that we would ideally want to increase the injection temperature at least 230C. Again, this can't be root cause given method history / transfers.

Unfortunately, I am not going to be able to get much more specific at this point.

[tkubowicz]

Hello

If your problem is coeluting peak with your analyte it is simple to diagnose.
First thing to check is if peak comes from:

1. GC
2. Sample+injector system

To check this just run blank with no injection (only temperature gradient). If you have peak it means it comes from GC (inlet, column). Then you need to properly do GC cleaning.

If you dont have peak for temperature profile you need to narrow problem down:
-check injector (syringe)
-check solvents used for sample prep
-check vials

Note: I always use retention gap (1-2 m) with columns like WAX. They have tendency to adsorb nasty stuff at the front of column and even with baking out it is tricky to remove it.

Good luck

Regards

Tomasz Kubowicz

[Peter Apps]

The most likely explanation is carry-over caused by injecting too much methanol, whose vapour pushes sample into places it should not be. The reason that some GCs have the problem and others not is probably differences in how fast their electronic flow and pressure controllers respond to the pressure pulse as the sample is injected. To test this you would need to swap the gas controls and inlets between GCs, but in the end you need to develop a method that is robust to inter-instrument differences.

Peter

[shaun78]

Hi Peter,

I've been wondering about this myself.

Internally we have three instruments that can not run this test method (one 6890 and two 7890's); internally we have two GCs that are able to run this test method (both 6890's).

The contract laboratory that we recentlly transferred to has three 7890's, all of which are unable to run the test method.

Internally, people are not too keen on me swapping inlets and pneumatics to prove this point. Is there any other way about it? I've reduced injection volume (2 uL --> 1 uL), but still seem to run into the problem. If I reduce past 1 uL I'm up against a sesnitivity problem.

If I can come up with a definitive way to show that is it, then I am all for it (aside from swapping parts).

This method is going to be in remediation on account of the recent issues. Making the method more robust is certainly on the chopping block.

[Peter Apps]

The simplest approach will be to use a different solvent - the only one that is worse than methanol is water ! Even a higher boiling alcohol will be better.

How about going splitless with a smaller injection volume ?

Peter

[Bigbear]

I'm with Peter, but if you don't want to reduce the injection volume, try a pulsed injection.

[shaun78]

Okay, I have reduced injection volume. 1 uL is borderline unacceptable vapor volume; problem still presists. 0.5 uL is well within acceptable vapor volume; problem still presists. I've also independantly increased / decreased split ratio a long time ago at the initial phase of this investigation; problem still presists.

No matter if I reduce injection volume or increase / decrease split ratios I observe a linear reduction of peak areas. However, I still have a problem with the analysis.

A long long time ago as well I switched to a splitless injection using 0.1 or 0.2 uL. Yes, you guessed correctly; the problem was still present. At the time I thought doing this would point a finger to replacing the split vent trap, which I replaced anyway just to be sure...

[Jake]

For testing purposes I was also going suggest a splitless injection, but you've already done that.

You presumably have a standard. With different amounts on column, does the interfering peak's area change proportionally, or remain constant?

Perhaps posting chromatograms with and without the problem peak would help.

[shaun78]

Hi Jake,

I can't remember if I wrote this already, but this problem is not present in standard analysis; only the sample matrix.

I can try to post chromatograms tomorrow.

Unfortunately, I can't directly answer if this unknown peak is directly proportional to the amount of analyze present. We have a very low specification range (nearly a single order of magnitude). When this peak is present, it would cause a sample otherwise reported as "< LOQ" to be reported as "Fails specification".

I suppose it does not hurt... our spec range is 15-35 ug/g. At the upper end of the specification our analyze of interest has an area of about 15 pA. As you've gathered, it doesn't take a whole lot for us to blow our spec.

Area counts of the interfering peak are routinely around 8-9 pA.

These is always some residual in the sample matrix, it is more of a matter of if it is below our lower specification limit or not.

[Peter Apps]

What extra peaks do you see if you hold the column at the maximum of the programme temperature for about 30 min ?

Peter

[shaun78]

Hi Peter,

I've not held the max temperature for an extended peroid of time myself. However, the contract lab we tried to transfer to has.

You see lots of peaks and this problem goes away.

Now here is the problem. We have fully validated this method for materials A and B. We are testing material A. Granted, there was an R&D aspect of what we are doing, but the R&D portion does not have any impact on the material that we are testing. There have been no actual changes to the manufacturing process.

Additionally, problems with the matrix seem to be less likely as we do have instruments that are able to run this method on the same samples without a problem.

But I suspect you might have something else in mind...

[Peter Apps]

Then your interference may be a ghost peak. Is it a problem to simply extend the run time ?

Peter