A comprehensive approach to incorporating intermolecular dispersion into the openCOSMO-RS model. Part 2: Atomic polarizabilities
Daria Grigorash, Simon Müller, Esther Heid, Frank Neese, Dimitrios Liakos, Christoph Riplinger, Miquel García-Ratés, Patrice Paricaud, Erling H. Stenby, Irina Smirnova, Wei Yan
TL;DR
The paper addresses improving openCOSMO-RS by incorporating a dispersion term derived from atomic polarizabilities, and evaluates how polarizability processing (e.g., scaling/combining) affects segment-segment dispersion energies, with a focus on halocarbon systems. The approach is validated on a data-rich benchmark (>50,000 data points) and shows improved accuracy with fewer adjustable parameters compared to Part I. The results establish atomic polarizability as a valuable descriptor for refining dispersion interactions in predictive thermodynamics and demonstrate the method's robustness across diverse data types, supported by open-source implementation. The work enhances the reliability of openCOSMO-RS for complex mixtures and broad application in chemical and biochemical engineering.
Abstract
openCOSMO-RS is an open-source predictive thermodynamic model that can be applied to a broad range of systems in various chemical and biochemical engineering domains. This study focuses on improving openCOSMO-RS by introducing a new dispersion term based on atomic polarizabilities. We evaluate different methods for processing polarizability data, including scaling and combining it to compute segment-segment dispersion interaction energies, with a focus on halocarbon systems. The results demonstrate that the modified model outperforms our previous method developed in the first part of this work (Grigorash et al., 2024) , while at the same time requiring fewer adjustable parameters. The approach was applied to a broad dataset of over 50,000 data points, consistently increasing the accuracy across a variety of data types. These findings suggest that atomic polarizability is a valuable descriptor for refining dispersion interactions in predictive thermodynamic models.