Gibbs–Helmholtz graph neural network: capturing the temperature dependency of activity coefficients at infinite dilution

Abstract

The accurate prediction of physicochemical properties of chemical compounds in mixtures (such as the activity coefficient at infinite dilution γij∞) is essential for developing novel and more sustainable chemical processes....

Gibbs–Helmholtz Graph Neural Network for the Prediction of Activity Coefficients of Polymer Solutions at Infinite Dilution

“…It is interesting to note that, during interpolation, the GH-GNN (pss) model achieves a similar MAE when predicting lnΩ ij ∞ of polymer solutions as the one that the original GH-GNN model achieves when predicting small-size systems (i.e., around 0.13). This phenomenon may be attributed to the inherent limitations posed by the experimental uncertainty of logarithmic activity coefficients, which is estimated to lie within the range of 0.1–0.2 .…”

“…In this work, we have used the same graph definition from Battaglia et al and the same node- and edge-level features as in our previous work . Similarly, we constructed one graph for the solvent and one graph for the solute (now a polymer).…”

Gibbs–Duhem-informed neural networks for binary activity coefficient prediction

“…ref. 10 and 14, and considering mixtures with more than two components. Since our investigations are based on activity coefficients obtained from COSMO-RS by ref.…”

“…, extrapolating to new molecules not used in training of the ML models at all, which was successfully demonstrated by ref. 14 and 18, are herein not considered and could be investigated in future work.…”

“…This includes both pure component and mixture properties such as solvation free energies, 1 liquid densities 2 and viscosities, 3 vapor pressures, 2,4 solubilities, 5 and fuel ignition indicators 6 A particular focus has recently been placed on using ML for predicting activity coefficients of mixtures due to their high relevance for chemical separation processes. Here, activity coefficients at infinite dilution, 7–9 varying temperature, 10–15 and varying compositions, 16–18 while considering a wide spectrum of molecules, have been targeted with ML, consistently outperforming well-established models such as UNIFAC 19 and COSMO-RS. 20,21 Given the high accuracy achieved, ML will therefore play an increasingly important role in activity coefficient prediction.…”