COSMO-SAC Variants

In this section, we describe the different COSMO-SAC variants currently available in JCOSMO.
Over the years, several versions of COSMO-SAC have been developed to improve the description of intermolecular interactions, hydrogen bonding, etc. Each variant introduces changes to specific components of the model — such as the sigma-profile generation, the segment interaction energy expression, or the combinatorial term.

JCOSMO provides an implementation of multiple COSMO-SAC variants so users can explore their relative performance and applicability depending on the target property or chemical system. The variant can be selected via the combobox in the top left corner of JCOSMO user interface, allowing users to benchmark and validate different formulations under consistent computational settings.

Note: Compatibility between sigma profiles and the chosen COSMO-SAC variant is essential. Always ensure that sigma-profiles were generated using the corresponding assumptions (e.g., level of theory, basis set, cavity construction, etc.).

CS25

The LVPP-modified COSMO-SAC 2025 model represents our latest COSMO-SAC parametrization, referred to simply as CS25.

Sigma-Profiles

CS25 is based on sigma-profiles calculated using NWChem as the quantum chemical package, in contrast to previous parametrizations that relied on GAMESS. The level of theory employed is B3LYP/def2-SVPD. Molecular geometries were first optimized in the gas phase, followed by a single-point calculation with the COSMO solvation model to generate the sigma-profiles.

The def2-SVPD basis set was chosen as a balance between accuracy and computational cost, offering improved polarization treatment and broader elemental coverage compared to the TZVP basis set used previously.
For larger molecules, the def2-SVP basis set was adopted to reduce computational demand, with negligible loss of accuracy in the resulting sigma-profiles.

Multiple Hydrogen Bonds

In this parametrization, different types of hydrogen bonds (HBs) are explicitly considered. Two types of HB donors are defined: one corresponding to water molecules and another representing hydrogens bonded to electronegative atoms such as N, O, F, Cl, Br, or I. In the user interface, all these donor types are collectively referred to as alcohol donors.

Only the surface area fraction with a charge density exceeding the defined HB cutoff value is assumed to be capable of participating in HBs.

For HB acceptors, several categories are available: ketone, ether, amine, and F+. Some special cases apply to nitrogen-containing groups:

  • When nitrogen is bonded to a single atom (e.g., in nitriles), it is treated as a ketone-type acceptor.
  • When bonded to two atoms (e.g., in pyridine), it is treated as an ether-type acceptor.

Dispersion Contribution

The original COSMO-SAC model typically disregarded dispersion interactions, assuming that their effects would cancel out in excess properties. In CS25, a dedicated dispersion contribution is included for each pair of atoms. This is achieved by storing, along with the surface charge density (sigma-profile), the atom type associated with each surface segment.

Although the resulting dispersion term is generally small, it provides a subtle yet meaningful correction that improves the description of nearly athermal mixtures, systems dominated by weak interactions, and particularly fluorinated/hydrocarbon mixtures, where dispersion effects play a more significant role.

Combinatorial contribution

the Flory-Huggins (FH) equation is used for the combinatorial contribution: where is the volume fraction, is the molecular volume of compound i obtained from the COSMO calculations, and is the mole fraction of component i.

This is in contrast to the typical Staverman-Guggenheim (SG) term with a normalized area12. The reason for using FH is because the potential inconsistencies with the SG formula3. Thus, in all recent COSMO-SAC calculations of our group we have assumed a simple FH combinatorial contribution. Although this can produce slightly worse results for linear hydrocarbon mixtures.

COSMO-SAC-HB2

The LVPP-modified COSMO-SAC model with multiple hydrogen-bond (HB) energy types, improving the description of systems where different donor and acceptor strengths play a significant role. Three parametrizations are available, depending on the quantum chemistry package and level of theory used to generate the sigma-profiles:

  • COSMO-SAC-HB2 (GAMESS) 4: based on sigma-profiles computed with GAMESS using the HF/TZVP level of theory.
  • COSMO-SAC-HB2 (FINE) 5: based on TURBOMOLE calculations employing a fine-grid marching tetrahedron cavity and the BP/TZVPD level of theory.
  • COSMO-SAC-HB2 (BP-TZVP) 5: based on TURBOMOLE calculations using the BP/TZVP level of theory.

All variants use FH as the combinatorial contribution by default, and none includes a dispersive contribution.

  1. Shiang-Tai Lin and Stanley I. Sandler. A Priori Phase Equilibrium Prediction from a Segment Contribution Solvation Model. Ind. & Eng. Chem. Res., 41(5):899–913, 2002. doi:10.1021/ie001047w. ↩

  2. Rafael de P. Soares. The Combinatorial Term for COSMO-Based Activity Coefficient Models. Industrial & Engineering Chemistry Research, 50(5):3060–3063, mar 2011. doi:10.1021/ie102087p. ↩

  3. Gerard J P Krooshof and Gijsbertus de With. Gibbs probability entropy and its implication to combinatorial entropy models. Fluid Ph. Equilib., 584:114146, 2024. doi:https://doi.org/10.1016/j.fluid.2024.114146. ↩

  4. F. Ferrarini, G. B. Flores, A. R. Muniz, and R. P. Soares. An open and extensible sigma-profile database for cosmo-based models. AIChE J., 64(9):3443–3455, 2018. doi:10.1002/aic.16194. ↩

  5. Edgar T. de Souza Jr., Murilo L. Alcantara, Paula B. Staudt, João A. P. Coutinho, and Rafael de P. Soares. Development of a COSMO-SAC parametrization with advanced QM method TZVPD-FINE. Industrial & Engineering Chemistry Research, 64(29):14700–14711, 2025. URL: https://doi.org/10.1021/acs.iecr.5c01146, doi:10.1021/acs.iecr.5c01146. ↩↩