Abstract

Solvation parameter models relate linearly compound properties with five fundamental solute descriptors (excess molar refraction, dipolarity/polarizability, effective hydrogen-bond acidity and basicity, and McGowan volume). These models are widely used, due to the availability of protocols to obtain the descriptors, good performance, and general applicability. Several approaches to predict retention in reversed-phase liquid chromatography (RPLC) as a function of these descriptors and mobile phase composition are compared, assaying the performance with a set of 146 organic compounds of diverse nature, eluted with acetonitrile and methanol. The approaches are classified in two groups: those that only allow predictions of retention for the mobile phases used to build the models, and those valid at any other mobile phase composition. The first group includes the use of ratios between the regressed coefficients of the solvation models that are assumed to be characteristic for a column/solvent system, and the application of offsets to transfer the retention from a reference mobile phase to any other. Maximal accuracy in predictions corresponded, however, to the approaches in the second group, which were based on models that describe the retention as a function of mobile phase composition (expressed as the solvent volume fraction or a normalised polarity measurement), where the coefficients were made dependent on the solvent descriptors. The study revealed the properties that influence the retention and distinguish the particular behaviour of acetonitrile and methanol in RPLC.

Keywords: Reversed-phase liquid chromatography; General solvation parameter models; Prediction of retention; Mobile phase composition; Acetonitrile; Methanol

Article Outline

1. Introduction

2. Theory

2.1. Method 1: common parameter ratios

2.2. Method 2: mobile phase flag variables

2.3. Method 3: relationships based on φ

2.4. Method 4: separation of polarity contributions

3. Results and discussion

3.1. Probe compounds

3.2. Reference predictions

3.2.1. First reference: individual quadratic relationships

3.2.2. Second reference: common column parameters

3.2.3. Third reference: solvation parameter models developed individually for specific mobile phases

3.3. Prediction capability of the general solvation parameter models

3.3.1. Method 1: common parameter ratios

3.3.2. Method 2: mobile phase flag variables

3.3.3. Method 3: relationships based on φ

3.3.4. Method 4: separation of polarity contributions

4. Conclusions

Acknowledgements

References