Description and prediction of retention in normal-phase high-performance liquid chromatography with binary and...

The suitabilities of several earlier reported models for description and prediction of retention in normal-phase systems with mobile phases comprised of two organic solvents-a polar and a non-polar one-were tested on the chromatographic behaviour of phenylurea herbicides and alkyl-, aryl- and nitrophenols as sample solutes with a silica-gel column and 2-propanol, n-heptane and dioxane as mobile phase components. With few exceptions, the data obtained from the best-fit three-parameter retention equation differ from the experimental capacity factors at less than 0.1, or 2%, and for most compounds the fit is better than 1%. Of the two two-parameter equations derived from theoretical models, one failed to describe the retention behaviour. The fit of the other two-parameter equation to the retention data of phenols is slightly inferior than with the three-parameter equation. The differences between the fitted data and the experimental capacity factors of substituted phenylureas were < or = 0.2, or 5%. Three-parameter equations were suggested to describe the retention behaviour in three-component mobile phases either at a constant sum or at a constant ratio of the two stronger solvents in ternary mobile phases. To fit the three-component data at any combination of concentrations of the three solvents with the error comparable to that in binary mobile phases, a nine-parameter equation is necessary. Suitability of these equations to describe the experimental behaviour of substituted phenylureas and phenols was verified in ternary mobile phases with various concentrations of 2-propanol, dioxane and n-heptane. Finally, a method was suggested for the prediction of retention in ternary mobile phases with varying concentration ratios of the two polar solvents from the parameters of best-fit equations in binary mobile phases and a single experimental capacity factor in a ternary mobile phase. For most capacity factors, the error of prediction was lower than 0.2, or 5%. Dried solvents were used to improve the reproducibility of the results and the temperature was controlled to +/-0.1 degree C in all experiments. With these precautions, differences between the original retention volumes and the data from repeated experiments measured after ten months of use of the column in the system tested were less than 0.2 ml for 85% of the values compared.