DFT-based theoretical prediction of intrinsic viscosity of polymer solutions

A four-descriptor quantitative structure-property relationship model was constructed to predict 65 intrinsic viscosities [η] of polymer solutions. Four quantum chemical descriptors, the traceless quadrupole moment θ_(R), the hydrogen bond or electrostatic attraction descriptor Q_H, the partition function Q_BOT(R) and the frontier orbital descriptor E_HOMO, were calculated with density functional theory at the B3LYP/6-31G(d) level and used to develop the model by multivariate linear regression (MLR) analysis. The model possesses coefficients of determination r² of 0.827 for the training set and 0.808 for the test set, and shows better statistical characteristics than the existing MLR models of intrinsic viscosities [η] of polymer-solvent combinations. Moreover, the four descriptors were used to develop a support vector machine model for [η] that possesses a coefficient of determination r² of 0.911 for the whole data set.