Predicting adsorption of aromatic compounds by carbon nanotubes based on quantitative structure property relationship principles

Abstract

Quantitative structure property relationship (QSPR) models were developed to predict the adsorption of aromatic compounds by carbon nanotubes (CNTs). Five descriptors chosen by combining self-organizing map and stepwise multiple linear regression (MLR) techniques were used to connect the structure of the studied chemicals with their adsorption descriptor (K∞) using linear and nonlinear modeling techniques. Correlation coefficient (R2) of 0.99 and root-mean square error (RMSE) of 0.29 for multilayered perceptron neural network (MLP-NN) model are signs of the superiority of the developed nonlinear model over MLR model with R2 of 0.93 and RMSE of 0.36. The results of cross-validation test showed the reliability of MLP-NN to predict the K∞ values for the aromatic contaminants. Molar volume and hydrogen bond accepting ability were found to be the factors much influencing the adsorption of the compounds. The developed QSPR, as a neural network based model, could be used to predict the adsorption of organic compounds by CNTs.