986 J. Sep. Sci. 2013, 36, 986–991 Monika Michel 1 Luke Chimuka 2,3 Tomasz Kowalkowski 2 Ewa M. Cukrowska 3 Boguslaw Buszewski 2 1 Plant Protection Institute-NRI, Pozna ´ n, Poland 2 Department of Environmental Chemistry and Bioanalytics, Faculty of Chemistry, Nicolaus Copernicus University, Toru ´ n, Poland 3 School of Chemistry, Department of Environmental Analytical Chemistry, University of Witwatersrand, WITS, Johannesburg, South Africa Received February 1, 2012 Revised November 7, 2012 Accepted November 15, 2012 Research Article Prediction of extraction efﬁciency in supported liquid membrane with a stagnant acceptor phase by means of artiﬁcial neural network An artiﬁcial neural network model of supported liquid membrane extraction process with a stagnant acceptor phase is proposed. Triazine herbicides and phenolic compounds were used as model compounds. The model is able to predict the compound extraction efﬁciency within the same family based on the octanol–water partition coefﬁcient, water solubility, molecular mass and ionisation constant of the compound. The network uses the back- propagation algorithm for evaluating the connection strengths representing the correlations between inputs (octanol–water partition coefﬁcients logP, acid dissociation constant pK a , water solubility and molecular weight) and outputs (extraction efﬁciency in dihexyl ether and undecane as organic solvents). The model predicted results in good agreement with the experimental data and the average deviations for all the cases are found to be smaller than ±3%. Moreover, standard statistical methods were applied for exploration of relationships between studied parameters. Keywords: Artiﬁcial neural network / Ionisable organic compounds / Supported liquid membrane DOI 10.1002/jssc.201200105 1 Introduction The supported-liquid membrane technique introduced by Audunsson [1] for sample preparation has demonstrated that it can offer an alternative for various ionisable organic com- pounds in aqueous samples as seen in a number of re- views [2, 3]. Several important parameters such as the pH of the donor and acceptor phases, polarity of the immobilised organic solvent and that of extracted analyte govern the ex- traction process. These parameters need to be optimised for maximum extraction efﬁciency. In previous studies, the in- ﬂuence of the acceptor pH [4], temperature [5, 6] and the hydrophobicity [7] of a compound on the trapping in the ac- ceptor phase and dissolution into the membrane liquid on the extraction efﬁciency have been investigated, respectively. The trapping is set so that the compounds are ionised completely in the acceptor phase, thus maintaining a concentration gra- dient across the membrane of the diffusing compounds. The donor pH is set so that compounds are able to partition into Correspondence: Professor Boguslaw Buszewski, Department of Environmental Chemistry and Bioanalytics, Faculty of Chemistry, Nicolaus Copernicus University, 7 Gagarin Str., 87-100 Toru´ n, Poland E-mail: bbusz@chem.umk.pl Fax: +4856-611-48-37 Abbreviations: ANN, artiﬁcial neural network; PC, principal component; PCA, principal components analysis the membrane liquid. The membrane liquid is often cho- sen so that it gives high partition coefﬁcient for the analyte but low for interfering matrix, thus giving the required high extraction efﬁciency and selectivity. The membrane liquid should also be insoluble in water to assure the desired stabil- ity. Di-n-hexylether and n-undecane have so far been found to have the above properties reﬂected in their frequent use in many application of the technique [8–16]. Since all these parameters need to be optimised, it is important to reduce the experiments as much as possible especially in simultaneous extraction of ionisable compounds of the same family. Statis- tical techniques are therefore potential candidates in helping to predict the extraction process, thus reducing the number of optimisation experiments. A number of researchers have tried to use statistical tech- niques to study the extraction process in membrane-based extractions. Romero et al. [17] used multivariate optimisa- tion of supported liquid membrane extraction of biogenic amines from wine samples prior to LC determination of dansyl derivatives. Transfer prediction by linear discriminate analysis and soft independent modelling of class analogy was used in the extraction of pesticides by membrane separa- tion by Carabias-Martinez et al. [18]. In this case, molecular weight, boiling point, vapour pressure, octanol–water parti- tion coefﬁcients logP, acid dissociation constant pK a , solubil- ity and density of the compounds were investigated on how they inﬂuence the extraction process. Chakraborty et al. [19] is reported to have studied the applicability of artiﬁcial neural network (ANN) in emulsion liquid membranes. ANN has also C  2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.jss-journal.com