A molecular structure based model for predicting optimal salinity of anionic surfactants

• 20 sets of optimal salinity were acquired for the surfactant/Decane/brine systems.
• A QSPR method was developed to estimate the optimal salinity of the microemulsion system.
• Accuracy of the models was shown by very good statistical validation parameters and the applicability domain.
• The most significant molecular properties affecting optimal salinity are introduced.
• The model outperforms the constitutional descriptors presented in previous empirical correlations.

Optimal salinity (Sopt) is an important property that influences the relative phase volume and solubilization parameters in microemulsion flooding. Selecting an appropriate surfactant for chemical flooding with high Sopt requires characterizing lots of different chemicals and a sophisticated interpretation procedure. The main objective of this paper is to apply quantitative structure property relationship (QSPR) technique to model the Sopt of 20 different surfactant molecules. The surfactant molecules are sorted into two different groups according to their experimental conditions. Geometrical optimization of surfactants was performed at RM1 level. Then, many structural and quantum chemical descriptors were calculated using different computer software programs. Using variable selection of the genetic algorithm (GA-MLR), two descriptors were introduced as independent variables. The squared correlation coefficient (R2) and standard deviation (s) calculated for the selected model were 0.940 and 0.643 for molecular group-A; and 0.984 and 0.445 for molecular group-B, respectively. The results demonstrate high estimation accuracy and strong generalization capacity of the models. The descriptors in both models are related to polarizability and ionization potential of the molecules, thus they are conceptually related to the changes in Sopt. The performance of QSPR was further compared with some commonly used constitutional descriptors in previous literature reports. By comparing the results, one can conclude that the estimation of Sopt can be improved significantly by using QSPR compared with other correlations.

Figure optionsDownload as PowerPoint slide