QSPR modelling for intrinsic viscosity in polymer-solvent combinations based on density functional theory

Area of Science:

• Computational Chemistry
• Polymer Science
• Physical Chemistry

Background:

• Intrinsic viscosity is influenced by polymer and solvent properties.
• Accurate prediction of polymer-solvent interactions is crucial for material design.
• Quantitative Structure-Property Relationship (QSPR) models offer a predictive framework.

Purpose of the Study:

• To develop and validate linear and nonlinear QSPR models for polymer-solvent systems.
• To investigate the impact of different polymer structure representations (1-5 monomeric units).
• To compare the predictive performance of genetic algorithms-multiple linear regression (GA-MLR) and support vector machine (SVM) models.

Main Methods:

• Calculation of quantum chemical descriptors (e.g., dipole moment, HOMO/LUMO energies) using DFT.
• Inclusion of topological descriptors to account for polymer and solvent structure.
• Screening of molecular descriptors using GA-MLR.
• Development of QSPR models using GA-MLR and SVM.

Main Results:

• GA-MLR models achieved R² values of 0.78 (training) and 0.83 (prediction).
• SVM models demonstrated higher predictive power with R² values of 0.95 (training) and 0.93 (prediction).
• External validation on a literature dataset confirmed SVM's superior performance (R²=0.90 vs. R²=0.81 for MLR).

Conclusions:

• Established QSPR models exhibit good predictability for polymer-solvent interactions.
• SVM models provide a more accurate prediction of intrinsic viscosity compared to MLR.
• The study highlights the utility of combining quantum chemical and topological descriptors in QSPR modeling.