Elucidation of specific aspects of dielectric constants of conjugated organic compounds: a QSPR approach

Abstract

The characteristic aspects of dielectric constants of π-conjugated compounds are elucidated by a quantitative structure property relationship (QSPR) study. To develop a QSPR model, among 141 collected π-conjugated compounds, a subset of 116 compounds was used as the training set for the model building and the rest was used as the test set for the model validation. Statistical regression models using 396 molecular descriptors were generated based on the genetic function approximation algorithm. The predicted dielectric constants obtained by the best model are highly correlated with the experimental values (squared correlation coefficient R(2) of 0.93 and 0.97 for the training and test sets, respectively), while a previous prediction model for general organic molecules (Sild S, Karelson M (2002) J Chem Inf Comput Sci 42:360-367) is not valid for our collected π-conjugated organic compounds. It has been known that the dielectric constants of organic materials are largely influenced by orientational correlations of the constituent molecules. In general, hydrogen bonding is one of the most important intermolecular interactions affecting orientational correlation. In the case of π-conjugated compounds, however, π-π interaction could be another comparable interaction with the hydrogen bonding.

Predicting the Relative Static Permittivity: a Group Contribution Method Based on Perturbation Theory

“…To identify a suitable set of molecular descriptors, experimental permittivity data are correlated with various topological, geometric, and electronic descriptors . The molecular descriptors can be obtained from quantum chemistry calculations or databases − (e.g., highest occupied molecular orbital, dipole moment, van der Waals surface area, or refractive index). Current QSPR methods − for the permittivity are limited to 25 °C and thus cannot resolve the permittivity’s temperature dependence required in process design.…”

“…The molecular descriptors can be obtained from quantum chemistry calculations or databases − (e.g., highest occupied molecular orbital, dipole moment, van der Waals surface area, or refractive index). Current QSPR methods − for the permittivity are limited to 25 °C and thus cannot resolve the permittivity’s temperature dependence required in process design. In addition, the molecular descriptors might be difficult to obtain in the context of high-throughput screenings or molecular design, particularly if quantum chemistry calculations are necessary.…”

Reaction Kinetic Model Considering the Solvation Effect Based on the FMO Theory and Deep Learning

“…However, their model does not perform well for molecules involving the interaction of π−π conjugation. Lee et al 22 select other electron correlation descriptors based on the genetic function approximation algorithm to improve the model performance regarding the π−π conjugation interaction, while the size of 141 samples limits the extrapolation ability of their model.…”

“…To sum up, if data is sufficient, the data-driven empirical model is a choice worth considering for the prediction of the dielectric constant due to their accurate precision (the determination coefficients R 2 are all larger than 0.9 in Schweitzer and Morris, 20 Sild and Karelson, 21 and Lee et al 22 ) compared with mechanism models and semiempirical models. However, the current works still have two issues regarding their model application ranges and descriptor selections.…”