Prediction of Infinite Dilution Molar Conductivity for Unconventional Ions: A Quantitative Structure–Property Relationship Study

Abstract

The infinite dilution molar conductivity (λB ∞) that represents the interactions between ions and solvent molecules is an important transfer property for the utilization of ionic liquids (ILs) in electrochemical applications. However, employing the quantitative structure–property relationship (QSPR) model to predict the λB ∞ of unconventional ions remains to be explored. In this work, new λB ∞-QSPR models were developed to predict the λB ∞ of ions in aqueous solutions by using multiple linear regression (MLR) and stepwise linear regression (SLR) methods based on the molecular descriptors obtained by COSMO-SAC. A total of 132 cations and 158 anions data points at different temperatures were collected and tested. The results showed that the determination coefficients (R 2) of the QSPR model using MLR for cations and anions were 0.9515 and 0.9411, and the average absolute relative deviations (AARD) were 6.79% and 10.42%, respectively, indicating that the proposed λB ∞-QSPR model was excellent at predicting λB ∞. Moreover, R 2 of the QSPR model using SLR for cations and anions were 0.9450 and 0.9406, and AARD were 7.10% and 10.19%, respectively, implying that the λB ∞-QSPR model with fewer descriptors could also predict λB ∞ satisfactorily. We envisage the established λB ∞-QSPR models provide an available method for obtaining λB ∞ of ions in aqueous solutions.

Predicting the Thermal Conductivity of Ionic Liquids Using a Quantitative Structure–Property Relationship

“…The key to constructing a QSPR model is to choose a set of suitable descriptors that are also unique, which can indicate the structural characteristics of cations and anions. This study used work by Song et al 42 to divide the charge density distribution area of cations and anions into certain intervals S σi (e•nm −2 ) as the descriptors to develop the λ-QSPR model. Table 1 shows the σ profiles of cations divided into 5 intervals, concentrating at −2.0−0.5 e•nm −2 .…”

“…The second is that the fragment area descriptor obtained by COSMO-SAC in the work can distinguish the isomers of substances. In recent studies, QSPR models showed accurate predictions for the infinite dilution molar conductivity and infinite dilution diffusion coefficient of ILs developed based on fragment charge density distributions obtained using the COSMO-SAC model. , The reason for selecting the COSMO-SAC model is that it can generate the charge density distribution’s profile, namely, σ profiles that are called molecular fingerprints, which are identical to the human fingerprints and are unique . Moreover, the fragment charge density distributions separated by various intervals are unique.…”

Prediction of the Aqueous Solubility of Compounds Based on Light Gradient Boosting Machines with Molecular Fingerprints and the Cuckoo Search Algorithm

“… 11 − 13 In QSPR, the quantitative relationship among the physicochemical properties, biological properties, and molecular structures of compounds is explored with various statistical methods and mathematical models. 14 , 15 Usually, the molecular descriptors were selected as inputs of the models. 16 In previous studies, the main methods used in the QSPR model include multivariable linear regression (MLR), artificial neural network (ANN), Gaussian process (GP), and support vector machine (SVM).…”

“…To replace traditional mechanistic models, many researchers turned to the quantitative structure–property relationship (QSPR) model. − In QSPR, the quantitative relationship among the physicochemical properties, biological properties, and molecular structures of compounds is explored with various statistical methods and mathematical models. , Usually, the molecular descriptors were selected as inputs of the models . In previous studies, the main methods used in the QSPR model include multivariable linear regression (MLR), artificial neural network (ANN), Gaussian process (GP), and support vector machine (SVM). , However, the complex correlation between molecular descriptors and high-dimensional nonlinear data required for dissolution prediction poses great difficulties in traditional machine learning methods …”

Predicting the Self-Diffusion Coefficient of Liquids Based on Backpropagation Artificial Neural Network: A Quantitative Structure–Property Relationship Study

“…The COSMO-SAC model can be used to obtain the charge density distribution (σ-profile) and cavity volume ( V COSMO ). The charge density distribution area of molecules in a specific interval ( S σ i ) obtained from the σ-profile can be used as a structural descriptor reflecting the interactions, and the obtained V COSMO can be used as a structural descriptor reflecting the size of the molecular volume. ,, Consequently, a unique set of descriptors for the molecular structure can be obtained directly from the molecular structure.…”