At our lab, we analyse polymeric carbohydrates by hydrolysing the polymers and turning the monomeric sugars into alditol acetates. Solvent is ethylacetate. Carrier is He. Column is Supelco SP-2380, 0.32 ID, 30 m (polar, non bonded phase, T-limit : 275°C). Split ratio is 30, column flow 1 ml/min. Head pressure is about 10 psi. FID settings are classical. Analysis temperature is isothermal(225°C), injection and detection are at 260°C. Injection volume is 1 microL. Solvent peak to component peak ratio is large (1/5.000). Component peaks are rather small.

Reproducebility of 1 calibration sample injected five times is bad for the first peak (arabinose), somewhat better for the second (xylose) and ok for the next three peaks. Areas for the first peaks are larger (up to 50% for the first peak, up to 25% for the second) than the latter peaks. Symmetry for the peaks ranges between 0.8 and 1.2.
This problem is rather new and is observed on two different GC systems and different SP-2380 columns.

Does anybody have any idea what might be the problem ?
Thanks in advance.

What do you consider "bad" reproducibility? Are the initial peaks sitting on a tailing solvent peak? I'm curious why you mention the ratio of solvent to analyte peak sizes; the solvent peak should be completely off scale and thus its area will not be correctly represented and is essentially meaningless here.

It seems like you're looking at the area counts and not the calculated amounts after calibration; exactly what do you mean by "rather small" component peaks? Can you bring up your data system's idea of the peak start, apex, and end points on-screen or a printer/plotter?

-- John

John,

Thanks for the answer. What I call 'bad' is a response factor for 1 component that changes between 1.2 and 1.7 between consecutive runs of the same sample. Peaks are very well resolved and not on the solvent tail. Integration (peak start, apex and end-points) is ok also. Rather small is an apex of 3 pA above baseline.
As you mentioned, solvent peak data is indeed not that important.

Are you collecting enough data points across the peak? For most data systems, 15-30 are required for optimum integration. If the peak apex appears to be correctly located this may not be the problem but it would probably be wise to check.

We sample at 20 Hz, so number of data points should be ok, but while we're at it, we can try an increased rate.

Thanks.

It has been my experience that bad precision in GC is usually related to the injection. This could include any part of the injection: The syringe, the septum, the speed of injection, the mechanics of the septum purge, the characteristics of the inlet liner, the mechanics of splitting (or not splitting, in splitless), and the installation of the column.

Start by considering the septum and syringe disposable items and replace them frequently. Suggested frequency: every 50 samples.

Next make sure that your inlet liner is clean and appropriate to your injection type and speed. I suggest a straight tube with a small, loose wad of silanized glass wool as a good general inlet liner.

Next, make sure that your needle enters the inlet to WELL below any septum purge flow. This is normally the case, but I don't know what instrument you're using...

Of course, be sure the column is installed properly.

If any of these things are in doubt, fix them and try again.

I can't help noticing it is the most volatile components that give the highest variability. Is there a trend, like, high peaks at first then lower, that could suggest the sample itself is changing?

As far as we can tell from our experience, the sample doesn't change over a period of a week or so. So that should not cause a problem.