Chemometrically assisted simultaneous separation of 21 aromatic sulfonatesin ion-interaction RP-HPLC

A method of ion-interaction chromatography (IIR-RP-HPLC) for the simultaneo us separation of 21 polar aromatic sulfonates is developed and optimised by a chemometric treatment. The analytes are: l-naphthalene sulfonic acid (1- NS), 2-naphthalene sulfonic acid, 2-amino-1-naphthalene sulfonic acid (2-A- 1-NS), 5-amino-2-naphthalene sulfonic acid (5-A-2-NS), 8-amino-2-naphthalen e sulfonic acid (8-A-1-NS), 1-amino-5-naphthalene sulfonic acid (1-A-5-NS), 4-amino-1-naphthalene sulfonic acid (4-A-1-NS), 6-hydroxy-2-naphthalene su lfonic acid (6-H-2-NS), 4-hydroxy-1-naphthalene sulfonic acid (4-H-1-NS), 6 -amino-3-hydroxy-2-naphthalene sulfonic acid (6-A-4-H-NS), 6-amino-1-hydrox y-3-naphthalene sulfonic acid (6-A-1-H-3-NS), 3-amino-2,7-naphthalene disul fonic acid (3-A-2,7-NdS), 7-amino-1,3-naphthalene disulfonic acid (7-A1,3-N dS), 2-amino-1,5-naphthalene disulfonic acid (2-A-1,5-NdS), 1-hydroxy-3,6-n aphthalene disulfonic acid (1-H-3,6-NdS), 2-hydroxy-3,6-naphthalene disulfo nic acid (2-H-3,6-NdS), 3-nitrobenzene sulfonic acid (3-NBS). 4-phenol sulf onic acid (4-PS), 4-hydroxy-3-nitrobenzene sulfonic acid (4-H-3-NBS), 1,2-b enzene disulfonic acid (1,2-BdS), 2,6-anthraquinone disulfonic acid (2,6-An tdS). The factors optimised are: (i) the alkyl chain length of the ion-inte raction reagent, (ii) the ion-interaction reagent concentration, (iii) the acetonitrile concentration and (iv) the pH of the mobile phase. An Fraction al Factorial design, together with a star design are used to simultaneously study the effect of the variables and of their interactions. The retention times resulted to be not homogeneously distributed in the variable domain: in particular, for all the analytes considered, the use of nonylamine as t he IIR caused unexpectedly large retention times. In these conditions, the linear regression algorithm failed in building good predictive models. The problem was solved by using the Box-Cox transformation, that considers othe r possible dependencies, besides the linear one, between the response (retention time t(R)) and the variables. In particular, a response 1/tR and 1/root tR permitted to better fit the data and to build reliable predictive models. In turn, the treatment of these models allowed us to locate the experimental conditions to separate, with good resolution, the 21 polar aromatic sulfonates in a total analysis time lower than 37 mi n. (C) 2000 Elsevier Science B.V. All rights reserved.