different results obtained using 6890N v 7890A

[pstaunton]

hi

im currently carrying out an investigation of a method we use in a QC lab. the method is used to quantify glycerine in an eyedrop formulation.

the problem is basically that the 6890N agilent model seems to give high sample bias compared to the 7890A model.

the 6890N was injected with sample and gave results of 1.69%w/v for first injection then 1.47% for second injection, 1.42% third and 1.38% for the fourth. the results show a downward trend from first injection to last injection.

however the same samples ran on the 7890A model gives consistent results at ~1.2% for all the same injections. Why is this so?

I set up both systems with exactly the same parameters and column and both went through the same equilibration before analysis

method - splitless, inlet: 250C, detector:300C.
temp program - initial 100C for 4.5min,
20C/min to 170C (hold 1min)
20C/min to 200C (hold 7.5min)
flow: 12ml/min constant flow
purge flow: 120ml/min after 0.5min, purge off 10min
wash solvents: methanol

sample is made up in methanol - 1g weighed diluted in 50ml of methanol to give conc ~0.02%w/v

column is DB-WAX 30m x 0.53mm id x 1um thickness. guard column 10m.

liner is split liner part number 5190-2294 (glass wool in the middle). I know this is a split liner and the method is not split but unfortunately this was written into the method (don't ask me how or why!)

standard injections seem to stay very consistent (%RSD <2%) but samples do not on the 6890N. is the sample matrix having an effect? can the guard column also have any effects? I noticed that one of the guard columns has a really shiny appearance while the other has the same appearance as the main column.

its probably quite easy to pick holes in this method (something I have no control over) but any ideas as to why response start off very high and start to decrease from injection to injection would be great.

[James_Ball]

When was the last time the split vent lines (copper line from inlet to split valve) were replaced?

Even though it is down stream of the inlet, a contaminated line and inlet fitting can cause such problems. Also has the split vent trap been changed?

With the 6890 being older, it could be more contaminated and therefore less stable.

As for the method, I would think split would be much better than splitless, and will not show the problems mentioned above as much, since there is flow through the line at time of injection instead of it being held with no flow during the splitless phase of the injection. Also that is quite a lot of flow even for a 0.53 column.

[Consumer Products Guy]

I won't comment on the direct answer to the issue of different GC units.

However, I assayed glycerin for over 3 decades in consumer products at levels lower than your sought-for 1% up to 30% using GC but with derivatization, as direct assay for glycerin can be tricky with its polarity due to its three hydroxy groups. I based our assays off this publication "Quantitative determination of glycerin in soap by capillary gas chromatography", JAOCS Volume 64, Issue 9 September 1987, Pages 1356-1357.

We also assayed for stuff like propylene glycol, sorbitol, diethylene glycol using modifications of this technique.

We also used USP assay procedures for impurities in 95% glycerin, those did not use derivatization. Those were trickier.

[w7west]

I suspect there could be a difference in the activities of the two systems towards polar functional groups and there is something in the sample matrix other than glycerin that has active hydrogens. I had this problem when analyzing ethylhexylglycerin in a skin cream and derivatizing with BSTFA as Consumer Products Guy suggested solved my %recovery issues. You could try analyzing derived versions of your solutions on both systems and see if the results start matching up.

[Consumer Products Guy]

I had this problem when analyzing ethylhexylglycerin in a skin cream and derivatizing with BSTFA as Consumer Products Guy suggested solved my %recovery issues. You could try analyzing derived versions of your solutions on both systems and see if the results start matching up.

w7west - I hope the person originally posting this took my advice in January when he/she posted, and just neglected to follow up....