Gradual Loss of Signal in Headspace Analysis

[pkenny]

We are using headspace analysis for residual solvents determination. The sample is in a Sesame Oil matrix and we are using 1ml of the sample and 4ml of water in a 20ml headspace vial. We are looking at 7 solvents (Chloroform, Acetone, Ethanol, Hexane, Methanol, Pentane, Pyridine) and have pretty good separation. The problem we are having, though, is during 6 subsequent injections (separate vials) in a sequence, the peak areas of Acetone, Ethanol, Methanol and Pyridine keep decreasing. If we start a new sequence the peak areas are back to where they originally were.

The method run time is 60 minutes and the oven equilibration time if 45 minutes. We have seen this with shaking (automated) and not shaking the sample prior to injection.

The headspace is a Tekmar and the GC is an Agilent 6890.

To me it seems like we have some sort of timing mismatch where the samples are not being treated to an equal amount of time in the oven, so we are looking into that. Any other ideas?

Thanks

[Peter Apps]

Why are you adding water ?

All the problem solvents are water soluble. My guess is that the oil floats on the water and the solvents equilibrate in two directions; into the headspace and into the water. As time goes by more of the water solubles dissolve in the water, lowering the concentration in the oil and subsequently in the gas phase.

Peter

[larkl]

You might check to be sure you don't have some water building up somewhere. Seems odd that the pentane didn't decrease, until you consider that it is not appreciably soluble in water. Could there be a cold spot where the water builds up? Maybe it is slowly vaporized in the carrier and is gone by the time the next sequence runs.

[larkl]

I think Peter has got it. Are you setting up all the HS vials at the same time? You could set them up one at a time, immediately before running each, this would tell you.

[pkenny]

The analyst who developed the method used USP <467> as a starting point - which uses water even for non-water soluble samples.

I immediately thought about the water being the cause also, but figured using the shake option would overcome the problem. With the shake feature it heats the sample vial for 45min @ 80C, then physically shakes it for 2min before beginning sampling for the injection.

The water still bothers me though, especially when you look at the solvents giving us the problem are the most water soluble. I was hoping I was overlooking something so that we wouldn't have to go waterless - the chromatography is actually better with the water (better separation).

[zokitano]

European Pharmacopoeial general method for identification and control of residual solvents, besides water (for water-soluble API) they suggest using dimethylformamide (DMF) for water-insoluble substances (API). If you strictly follow the European Pharmacopoeial method and extrapolate to your oil samples, then it is more practical way to use DMF instead of water as your solvent.

Regards

[Peter Apps]

Try just running the oil, there should not be any need to add anything. Use shake throughout the equilibrium time if the headspace sampler allows you to. The presence of water improving the separation is probably a symptom of adsorptively active silanols somewhere that the water sticks to and deactivates. Paradoxically they were probably caused by water in the first place. Rather optimise the separation by using different columns or temepratrure programmes.

Peter

[krickos]

Try just running the oil, there should not be any need to add anything.
Peter

Peter, I think that would work for identifiation purposes but not quantatively as long static headspace is used. The fact that sample is a sesame oil matrix and alkanes are analysed suggests that a standard addition technique should be used quantatively.

Any external standards with static headspace would most likely over/underestimate the content due to the sample matrix.

So for simplicity reasons I would suggest trying DMF and standard addition.

[Peter Apps]

Hi Krickos

We are jumping ahead a bit here.

I see two ways of calibrating this analyis. First choice would be external standards made up in clean sesame oil (vaccum or gas stripped at elevated temp to make sure that it is really blank), second would be known additions to samples. Which is preferred would depend on number of samples and batch size, and how much sample was available. It is the matix effects that make me reluctant to add anything that does not match the matrix, including DMF.

An advantage of adding DMF is that it might reduce the viscosity somewhat, giving quicker equilibrium, but at 80C sesame oil is very fluid anyway.

Of course a third option for quantitation is multiple injections from a single vial and then fitting the exponential curve to the decreasing peak areas, integate the area unde the whole curve and then compare it to peak areas from liquid splitless (or on-column) injections. It can be very accurate but it is painfully time consuming.

Peter