Predictive model development for adsorption of aromatic contaminants by multi-walled carbon nanotubes.

In the present study, Quantitative Structure-Activity Relationship (QSAR) and Linear Solvation Energy Relationship (LSER) techniques were used to develop predictive models for adsorption of organic contaminants by multi-walled carbon nanotubes (MWCNTs). Adsorption data for 29 aromatic compounds from literature (i.e., the training data) including some of the experimental results obtained in our laboratory were used to develop predictive models with multiple linear regression analysis. The generated QSAR (r(2) = 0.88), and LSER (r(2) = 0.83) equations were validated externally using an independent validation data set of 30 aromatic compounds. External validation accuracies indicated the success of parameter selection, data fitting ability, and the prediction strength of the developed models. Finally, the combination of training and validation data were used to obtain a combined LSER equation (r(2) = 0.83) that would be used for predicting adsorption of a wide range of low molecular weight aromatics by MWCNTs. In addition, LSER models at different concentrations were generated, and LSER parameter coefficients were examined to gain insights to the predominant adsorption interactions of low molecular weight aromatics on MWCNTs. The molecular volume term (V) of the LSER model was the most influential descriptor controlling adsorption at all concentrations. At higher equilibrium concentrations, hydrogen bond donating (A) and hydrogen bond accepting (B) terms became significant in the models. The results demonstrate that successful predictive models can be developed for the adsorption of organic compounds by CNTs using QSAR and LSER techniques.