Optimisation of gradient elution with serially-coupled columns. Part I: Single linear gradients

• Linear gradients are optimised for serially-coupled conventional columns.
• Several time delays are included in the numerical integration for maximal accuracy.
• The effect of the transit through several columns on peak profile is considered.
• Genetic algorithms and Pareto plots were used in the optimisation.
• Gradient optimisation pre-selecting the column combination was the best approach.

A mixture of compounds often cannot be resolved with a single chromatographic column, but the analysis can be successful using columns of different nature, serially combined through zero-dead volume junctions. In previous work (JCA 1281 (2013) 94), we developed an isocratic approach that optimised simultaneously the mobile phase composition, stationary phase nature and column length. In this work, we take the challenge of implementing optimal linear gradients for serial columns to decrease the analysis time for compounds covering a wide polarity range. For this purpose, five ACE columns of different selectivity (three C18 columns of different characteristics, a cyano and a phenyl column) were combined, aimed to resolve a mixture of 15 sulphonamides using acetonitrile–water gradients. A gradient predictive system, based on numerical integration, was built to simulate chromatograms under linear gradient profiles. Two approaches were compared: the optimisation of the combination of columns pre-selecting the gradient profile, developed by De Beer et al. (Anal. Chem. 82 (2010) 1733), and the optimisation of the gradient program after pre-selecting the column combination using isocratic elution, developed for this work. Several refinements concerning the gradient delays along the solute migration and peak half-width modelling were included to improve the realism of the predictions. Pareto plots (expressed as analysis time versus predicted global resolution) assisted in the selection of the best separation conditions. The massive computation time in the gradient optimisation, once the column combination was optimised, was reduced to ca. 3 min by using genetic algorithms.