QSAR studies of the dispersion of SWNTs in different organic solvents

Artificial neural network (ANN) and multiple linear regression (MLR) approaches were successfully applied to construct quantitative structure–activity relationship models of the dispersibility of single-walled carbon nanotubes (SWNTs) in different organic solvents. A subset of the calculated descriptors selected by enhanced replacement method (ERM) was used in the QSPR models development. The predictive abilities of ERM–MLR and ERM–ANN models were determined using a test set of six organic solvents affording predictive correlation coefficients of 0.924 and 0.963, respectively, showing good predictive power of the models obtained. The final models satisfied a set of rigorous validation criteria and performed well in predicting of the external test set. The results obtained in this study, confirm the importance of steric and electrostatic interactions, molecular flexibility, polarizability and hydrogen bonding ability of organic solvents in SWNTs dispersibility. This information could be useful for identification of some key molecular features of solvent molecules and to find the proper solvent for SWNTs dispersion.