The separation of samples that contain more than 15 to 20 analytes (n>15–20) is typically difficult and usually requires gradient elution. We have examined the reversed-phase liquid chromatographic separation of 24 samples with 8≤n≤48 as a function of temperature T and gradient time tG. The required peak capacity was determined for each sample, after selecting T and tG for optimum selectivity and maximum sample resolution. Comparison of these results with estimates of the maximum possible peak capacity in reversed-phase gradient elution was used to quantify the maximum value of n for some required sample resolution (when T and tG have been optimized). These results were also compared with literature studies of similar isocratic separations as a function of ternary-solvent mobile phase composition, where the proportions of methanol (MeOH), tetrahydrofuran (THF) and water were varied simultaneously. This in turn provides information on the relative effectiveness of these two different method development procedures (optimization of T and tG vs. % MeOH and % THF) for changing selectivity and achieving maximum resolution.

By optimizing column temperature T and gradient time tG, complex samples can often be separated by means of reversed-phase high-performance liquid chromatography (RP-LC). Conclusions reached in Part I suggest that the complete separation of such samples will be difficult, however, when more than 15–20 components are present in the sample. An alternative approach is to carry out two separations with different conditions (T, tG) in each run. The combination of results from these two runs then allows the total analysis of the sample, providing that every sample component is adequately resolved in one run or the other. Examples of this approach, carried out by means of computer simulation, are shown here for several samples of varying complexity. Also considered is the ability of a single separation where T and tG are optimized to enable the separation and analysis of one or more individual sample components from complex mixtures (e.g., drugs in animal plasma), including the resolution of isomeric compounds from each other.

Previous studies have shown that four experimental runs, where both temperature T and gradient time tG are varied, can be used for the reliable prediction of separation as a function of these two variables (two-dimensional optimization). Computer simulation (e.g., DryLab) can then be used to predict “optimized” conditions for maximum sample resolution using either isocratic or gradient elution. Samples that contain a large number of components (e.g., n>15–20) present a greater challenge. Resolution for these more complex samples is often quite sensitive to small changes in T or tG, in turn requiring greater accuracy in predictions that result from computer simulation. In the present study of several samples, we have examined computer simulation errors that can arise from inexact expressions for retention time as a function of T, tG or isocratic %B. Resulting conclusions are applicable to both complex and simpler samples, in either one- or two-dimensional optimization. Means to anticipate and minimize the impact of these predictive errors are examined.

Three ion chromatography (IC) retention models, namely the linear solvent strength model (LSSM), empirical end points model (EEPM) and three-point curve fitting using DryLab from LC Resources were evaluated in terms of their ability to predict retention factors for inorganic anions separated on a Dionex AS11 column using electrolytically generated hydroxide eluents. Extensive experimental retention data were gathered for 21 anions (fluoride, acetate, formate, bromate, chloride, nitrite, methanesulfonate, bromide, chlorate, nitrate, iodide, thiocyanate, succinate, sulfate, tartrate, oxalate, tungstate, phthalate, chromate, thiosulfate and phosphate) using hydroxide eluents of varying concentration. Although the purely theoretical LSSM was found to give adequate performance, the EEPM (in which a linear relationship is assumed between the logarithm of retention factor and the logarithm of eluent strength, but the slope is determined empirically) and DryLab performed better, with DryLab giving the best accuracy and precision of the three models. The EEPM and DryLab were also shown to have advantages in terms of their low knowledge requirements and ease of solution. Compared with IC using dual eluent species, the retention behaviour in IC using single eluent species was found to be easier to model by both theoretical and empirical approaches.