COSMO-RS prediction of hydrocarbons EACNs calibrated on the n-alkanes reference scale
Abstract
Surfactant–Oil–Water (SOW) systems are commonly described within the Hydrophilic–Lipophilic Deviation (HLD) framework, in which the Equivalent Alkane Carbon Number (EACN) quantifies oil hydrophobicity. The EACN values are usually determined from phase-behaviour studies (fish diagrams, temperature or salinity scans), which are experimentally demanding. Existing predictive approaches (QSPR, Machine-Learning) are all explicitly calibrated against experimental EACN data. This work proposes a physics-based strategy to predict the EACN of structurally diverse hydrocarbons (linear, branched, cyclic, saturated, unsaturated, aromatic) exclusively from their molecular structures, without using any experimental EACN values as input. The method relies on COSMO-RS calculations of the natural logarithm of the activity coefficient ln γ of the oil in the C₁₀E₄ surfactant, which is used as a descriptor of oil penetration into the interfacial surfactant film. For the n-alkane reference series, a linear ACN - ln γ relationship is first established, providing a purely thermodynamic calibration between Alkane Carbon Number and activity coefficient. The EACNs of more complex hydrocarbons are then estimated as the ACN of the n-alkane exhibiting the same calculated ln γ. The predictive power of this model is assessed on a chemically diverse set of 78 hydrocarbons, ranging from C₆ to C₃₀ and including alkanes, alkenes, alkynes and aromatic oils for which reliable experimental EACN values are available. A strong linear correlation is obtained between predicted and experimental EACN (R² = 0.97, RMSE = 1.2), indicating that COSMO-RS-based predictions provide a reliable alternative to experimentally determined EACN values.
Related
- SigmaFormer : Augmenting transformer encoders with COSMO sigma profiles for pure component property prediction
- Solubility prediction for ill-defined polymers: A critical comparison of Hansen parameters and COSMO-RS backfitting
- Molecular Property Prediction with Well-Calibrated Uncertainties via Graph Neural Networks and Conformal Quantile Regression
- Deoxygenative Dialkylation of Vinylic Epoxides to Chiral Alkanes by Double SN2’
- Regio- and Enantioselective Functionalization of Gaseous Alkanes via Nickel-Catalyzed Alkene Dicarbofunctionalization.
BibTeX
@article{aubry2026cosmors,
title = {{COSMO-RS prediction of hydrocarbons EACNs calibrated on the n-alkanes reference scale}},
author = {Jean-Marie Aubry and Théophile Gaudin and Andreas Klamt},
journal = {Journal of Molecular Liquids},
year = {2026},
month = jun,
doi = {10.1016/j.molliq.2026.129498},
url = {https://hal.science/hal-05734337v1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - COSMO-RS prediction of hydrocarbons EACNs calibrated on the n-alkanes reference scale
AU - Jean-Marie Aubry
AU - Théophile Gaudin
AU - Andreas Klamt
PY - 2026
DA - 2026/06/01
JO - Journal of Molecular Liquids
DO - 10.1016/j.molliq.2026.129498
UR - https://hal.science/hal-05734337v1
AB - Surfactant–Oil–Water (SOW) systems are commonly described within the Hydrophilic–Lipophilic Deviation (HLD) framework, in which the Equivalent Alkane Carbon Number (EACN) quantifies oil hydrophobicity. The EACN values are usually determined from phase-behaviour studies (fish diagrams, temperature or salinity scans), which are experimentally demanding. Existing predictive approaches (QSPR, Machine-Learning) are all explicitly calibrated against experimental EACN data. This work proposes a physics-based strategy to predict the EACN of structurally diverse hydrocarbons (linear, branched, cyclic, saturated, unsaturated, aromatic) exclusively from their molecular structures, without using any experimental EACN values as input. The method relies on COSMO-RS calculations of the natural logarithm of the activity coefficient ln γ of the oil in the C₁₀E₄ surfactant, which is used as a descriptor of oil penetration into the interfacial surfactant film. For the n-alkane reference series, a linear ACN - ln γ relationship is first established, providing a purely thermodynamic calibration between Alkane Carbon Number and activity coefficient. The EACNs of more complex hydrocarbons are then estimated as the ACN of the n-alkane exhibiting the same calculated ln γ. The predictive power of this model is assessed on a chemically diverse set of 78 hydrocarbons, ranging from C₆ to C₃₀ and including alkanes, alkenes, alkynes and aromatic oils for which reliable experimental EACN values are available. A strong linear correlation is obtained between predicted and experimental EACN (R² = 0.97, RMSE = 1.2), indicating that COSMO-RS-based predictions provide a reliable alternative to experimentally determined EACN values.
SN - 0167-7322
ER -