Prediction of the θ(UCST) of Polymer Solutions: A Quantitative Structure−Property Relationship Study

Abstract

One of the industrially important thermodynamic properties of polymer solutions is the upper critical solution temperature at the limit of infinite chain length of polymers, which is used to realize the usage limits of polymer solutions; this property is denoted as the θ(UCST). In this study, the quantitative structure−property relationship technique (QSPR) was used to correlate the θ(UCST) of polymer solutions. Based on molecular descriptors calculated from the chemical structures of the polymer and the solvent, a nine-parameter multilinear equation was obtained. This correlation can predict the θ(UCST) of 107 polymer solutions with a mean relative error of 6.12% and squared correlation coefficient of 0.912.

UCST-type behavior of poly[oligo(ethylene glycol) methyl ether methacrylate] (POEGMA) in aliphatic alcohols: solvent, co-solvent, molecular weight, and end group dependences

“…But also polymers based on N-acetyl acrylamide and N-acryloyl glycinamide or N-acryloyl glutamineamide have been shown to possess UCSTs in water. [24][25][26] Solubility of polymers in organic solvents is important for industrial processability, and thus q temperatures (ideal chain behavior) are tabulated in handbooks 27 and UCSTs for polymers in organic solvents are being predicted 28 and theoretically explained. 29 Well-tunable thermo-switchable materials with UCSTs in organic solvents (or ionic liquids 30 ) are also the basis for preparing micellar self-assemblies.…”

Deep Learning of Binary Solution Phase Behavior of Polystyrene

“…Our earliest understanding of polymer solution thermodynamics began with the lattice model of Flory and Huggins. − While Flory–Huggins theory can capture the phase behavior quantitatively for a few select polymer–solvent combinations, the earliest versions of this theory are oversimplified due to its underlying assumptions . Thus, additional lattice models and improvements have been derived based on the original Flory–Huggins model. − Other models followed, such as equation of state (EoS), quantity structure–property relationship (QSPR), , and activity coefficient models. , Many of these models require solving multiple complex equations simultaneously that are highly sensitive to the initial guess or require properties of the pure components that are either unknown or difficult to measure. In most cases, only qualitative agreement with experiments can be obtained, and each polymer–solvent system requires its own model parametrization.…”

“…3 Thus, additional lattice models and improvements have been derived based on the original Flory− Huggins model. 7−17 Other models followed, such as equation of state (EoS), 18 quantity structure−property relationship (QSPR), 19,20 and activity coefficient models. 21,22 Many of these models require solving multiple complex equations simultaneously that are highly sensitive to the initial guess or require properties of the pure components that are either unknown or difficult to measure.…”

Computation of Upper Flash Point of Chemical Compounds Using a Chemical Structure-Based Model

“…This algorithm was first presented by Leardi et al 55 The presented GA-MLR algorithm that implements the RQK fitness function is used for the subset variable selection. The latter was first proposed by Todeschini et al 56,57 This method has been well studied by Gharagheizi et al [10][11][12][14][15][16]18,20,22,26,27,33,[36][37][38]40 Before starting the computations, the data set must be divided into two new subdata sets; the first one (training set) for developing the model, and the other one (test set) for evaluation of the model. The most accurate model is found by treating the training set data.…”

“…Once a reliable relation has been obtained, it can be applied for prediction of the same property for other structures not yet measured or even not yet prepared. However, these kinds of methods have certain limitations: (I) the family of compounds used to derive the QSPR (“Training set”) should be similar according to the chemical structure, and (II) realistic predictions can only be made for compounds that are chemically related to those from which the QSPR model was derived; that is, predictions should be of interpolations or short extrapolations. − …”