HPLC with light scatter detector and chemometric data evaluation for the analysis of cocoa butter and vegetable fats

High performance liquid chromatography equipped with an evaporative light scatter detector was carried out in order to proof the authenticity of cocoa butter. Signals of 17 characteristic triglycerides have been used to develop two chemometric models. PLS was applied for quantitation while neural nets were used for classification. The sample pool was divided in a training set of 18 and a prediction set of 14 samples. The samples included mixtures of several vegetable fats with cocoa butter. A 15 × 4 × 1 feed forward net could be trained and within the prediction set only 2 samples were not correctly assigned. A PLS model with 9 factors was applied and the mean prediction error was found to be 2.5%. The small number of samples was found to be sufficient to show the potential of this data evaluation. Results are expected to improve with a greater data pool.