Chromatography Forum

We're trying to separate a mixture of simple sugars (glucose, galactose, fructose, tagatose) and levulinic acid using the Agilent HiPlex Ca (Duo) column, but have repeatedly noticed very strange behaviour from the levulinic acid.

Essentially, the retention time of the levulinic acid only (not at all the sugars) is affected by the sample concentration. Not in the expected manner, where high concentrations saturate the column, reducing the r.t., but rather higher concentrations consistently increasing the r.t. (from sample concentrations of 0.1 mg/ml --> 10 mg/ml). Rather annoyingly, it sweeps right across the range of the sugar peaks. Any ideas on what might be causing this, and how to fix it?

We use a mobile phase of 5% methanol in water, but have observed the same effect with 10% methanol (not strictly advised with this column but we were trying things out...). Also saw the effect operating at both 0.5 ml/min and 0.35 ml/min. Our column oven is limited to 80C so we can't explore the recommended operating temperature range of 80-90C. For a sense of scale, doubling sample concentration might increase levulinic acid r.t. anywhere from 30 s to 60 s, at a flow rate of 0.35 ml/min.

We are considering adding dilute formic acid to the mobile phase (unrecommended -> manufacturer specs recommend pH 7), or neutralising the levulinic acid with a base like dilute NaOH prior to analysis. Does anyone have experience and advice for trying such techniques with ligand/ion exchange columns?

Reduce the injection volume.

Injection volume is currently 5 uL, which is about as low as we can go while still keeping the peaks above noise at low sample concentrations

Try to use ACN instead of MeOH.

In case somebody else comes across the same issue, in the end what worked was spiking samples with very large amounts of levulinic acid (100 mg/ml), so that it eluted mainly after the carbohydrate peaks. Fructose still sits on the curvature at the base of the peak, but this can be helped by adding spiked blanks to the run, then subtracting their spectra from the sample spectra. The alignment is not perfect so you still see something of an LA peak, but it gives you a much better baseline to integrate from.

We did not try ACN in the mobile phase, maybe that would be a more simple solution for others. We previously found we could not get the required separations for other analytes with ACN.