Correlating and predicting low pressure solubility of gases in [bmim][BF4] by neural network molecular modeling

In this communication a modeling method based on neural network technique and molecular properties has been proposed to model solubility of carbon dioxide, carbon monoxide, argon, oxygen, nitrogen, methane and ethane in 1-butyl-3-methylimidazolium tetrafluoroborate. Molecular weight and acentric factor (i.e. sphericity of molecule) are two network inputs which indicate the structure of gas molecule. Absolute temperature and pressure are two other inputs which exhibit macroscopic condition of studied system. Low deviations during training, validating and testing stages confirmed that the model is reliable within the studied range. Also, the proposed method is able to provide reliable gas solubility estimations based on available solubility data of other gases. This unique capability, which confirms superiority of applied method over traditional methods, enables researchers and engineers to provide acceptable gas solubility estimations without performing long time experiments.

► Reviewing major modeling methods for correlating solubility of gases in ionic liquids.
► Application of neural network molecular modeling for correlating low pressure solubility of gases in [bmim][BF4].
► Comparison of proposed method to other modeling methods.
► Applying the proposed method for predicting solubility of gases based on experimental data points of other gases.