Development of a group contribution method for estimating surface tension of ionic liquids over a wide range of temperatures

Extensive Databases and Group Contribution QSPRs of Ionic Liquid Properties. 3: Surface Tension

“…In another paper, the same group adopted an artificial neural network (ANN) combined with gravitational search algorithm (GSA) to represent the dependence of σ on T and GCs. This way more sophisticated data treatment protocol resulted in an improved AARE of 1.18% obtained for the database slightly extended compared with the previous work on the linear model (2307 experimental data points, 229 datasets, 162 ILs). However, some doubts about the statistical significance of the model may arise if one notes that the number of adjusted parameters (ANN’s weights and biases) is relatively high compared to the number of data points used to fit them.…”

“…Alternatively, one can try to adopt a great variety of computational tools readily available nowadays and employ them to “subjugate” the data we already have and to perform computer-aided molecular design (CAMD) of the ILs not synthesized yet, but hopefully exhibiting desired values of the properties of interest. In the literature, a significant amount of effort has been devoted to modeling σ as well, mostly by employing empirical correlations with other properties, or chemical structures encoded by different kinds of molecular descriptors. − In the following paragraph, the most important contributions to this field − are presented and shortly discussed.…”

“…Another category of the methods for estimating σ of ILs comprises the models derived on the basis of general quantitative structure–property relationship (QSPR) scheme, − in many cases the models involving GCs, ,,− which indeed came up very popular since the pioneering work of Gardas and Coutihno . The main difference between QSPRs discussed below and those based on CSP is that they directly incorporate only the information extracted from the chemical structure of constituting ions so that there is no need to “preprocess” the structure (e.g., into critical parameters) to get an input variable for the model.…”

“…The calculation is, however, quite complicated, also since the authors did not provide any examples; besides, the parameter table shown in ref displays some significant errors so that one should contact the authors before applying the model using only the data found in the paper. In 2017, Lazzús and co-workers presented two models employing the GC methodology. , The first model incorporated a multiple linear regression of temperature-dependent σ data (2286 data points in 226 datasets for 154 ILs) performed by means of a genetic algorithm (GA). Distinct correlations were developed for six common chemical families of ILs (imidazolium, ammonium, phosphonium, pyridinium, pyrrolidinium, and piperidinium).…”

Corresponding-States Model for the Correlation and Prediction of the Surface Tension of Silanes

“…On the other hand, the most common semiempirical correlations used are those based on the corresponding-states principle, the chemical or physical group contributions, or the quantitative structure–property relationships. ,− Thus, models based on the corresponding-state principle show a possible universal behavior of different kinds of fluids. ,− Others are multiple-parameter correlations such as the van der Waals type model, the Weibull type model, and some others . The group contribution method has been used to study the properties of surface tension of alkanes, hydrocarbons, and ionic liquids, − etc. Recently, artificial neural networks have been successfully developed for the calculation of surface tension of different kinds of fluids. ,, In the case of pure silanes, Di Nicola et al have proposed a correlation for the surface tension as a function of liquid viscosity data.…”