Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Graph Neural Networks for Temperature-Dependent Activity Coefficient Prediction of Solutes in Ionic Liquids

Jan G. Rittig, Karim Ben Hicham, Artur M. Schweidtmann, Manuel Dahmen, Alexander Mitsos

TL;DR

The paper introduces a GNN that predicts temperature-dependent infinite-dilution activity coefficients $\gamma^{\infty}$ of solutes in ionic liquids by processing three molecular graphs (cation, anion, solute) into a single IL–solvent fingerprint, augmented with temperature. It demonstrates competitive performance against a state-of-the-art matrix completion baseline and shows strong generalization to ILs and solutes not seen during training via ensemble learning. The findings indicate that GNNs can extend AC predictions to IL solutions with unseen components and varying temperatures, enabling rapid exploration in computer-aided IL design. This approach broadens the applicability of GNNs in thermodynamic property prediction beyond traditional solvents and paves the way for more flexible, data-efficient modeling in ionic-liquid systems.

Abstract

Ionic liquids (ILs) are important solvents for sustainable processes and predicting activity coefficients (ACs) of solutes in ILs is needed. Recently, matrix completion methods (MCMs), transformers, and graph neural networks (GNNs) have shown high accuracy in predicting ACs of binary mixtures, superior to well-established models, e.g., COSMO-RS and UNIFAC. GNNs are particularly promising here as they learn a molecular graph-to-property relationship without pretraining, typically required for transformers, and are, unlike MCMs, applicable to molecules not included in training. For ILs, however, GNN applications are currently missing. Herein, we present a GNN to predict temperature-dependent infinite dilution ACs of solutes in ILs. We train the GNN on a database including more than 40,000 AC values and compare it to a state-of-the-art MCM. The GNN and MCM achieve similar high prediction performance, with the GNN additionally enabling high-quality predictions for ACs of solutions that contain ILs and solutes not considered during training.

Graph Neural Networks for Temperature-Dependent Activity Coefficient Prediction of Solutes in Ionic Liquids

TL;DR

The paper introduces a GNN that predicts temperature-dependent infinite-dilution activity coefficients of solutes in ionic liquids by processing three molecular graphs (cation, anion, solute) into a single IL–solvent fingerprint, augmented with temperature. It demonstrates competitive performance against a state-of-the-art matrix completion baseline and shows strong generalization to ILs and solutes not seen during training via ensemble learning. The findings indicate that GNNs can extend AC predictions to IL solutions with unseen components and varying temperatures, enabling rapid exploration in computer-aided IL design. This approach broadens the applicability of GNNs in thermodynamic property prediction beyond traditional solvents and paves the way for more flexible, data-efficient modeling in ionic-liquid systems.

Abstract

Ionic liquids (ILs) are important solvents for sustainable processes and predicting activity coefficients (ACs) of solutes in ILs is needed. Recently, matrix completion methods (MCMs), transformers, and graph neural networks (GNNs) have shown high accuracy in predicting ACs of binary mixtures, superior to well-established models, e.g., COSMO-RS and UNIFAC. GNNs are particularly promising here as they learn a molecular graph-to-property relationship without pretraining, typically required for transformers, and are, unlike MCMs, applicable to molecules not included in training. For ILs, however, GNN applications are currently missing. Herein, we present a GNN to predict temperature-dependent infinite dilution ACs of solutes in ILs. We train the GNN on a database including more than 40,000 AC values and compare it to a state-of-the-art MCM. The GNN and MCM achieve similar high prediction performance, with the GNN additionally enabling high-quality predictions for ACs of solutions that contain ILs and solutes not considered during training.
Paper Structure (13 sections, 3 equations, 4 figures, 5 tables)

This paper contains 13 sections, 3 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Data set
  3. Methods & Modeling
  4. Molecular graph
  5. Graph neural networks
  6. Graph neural network for activity coefficient prediction
  7. Matrix completion method baseline
  8. Ensemble learning
  9. Implementation & Hyperparameters
  10. Results & Discussion
  11. Prediction of new IL-solute combinations
  12. Generalization to new IL and solute molecules
  13. Conclusion

Figures (4)

  • Figure 1: Graph neural network model for prediction of the temperature-dependent infinite dilution activity coefficient of solutes in ionic liquids.
  • Figure 2: GNN model ensembling parity plot for test set with new IL-solute combinations. Red lines indicate $\pm$ 0.5 error range.
  • Figure 3: Absolute percentage error (APE) of GNN ensemble and MCM ensemble for predicting $\gamma^{\infty}$ of IL solutions in the test set categorized by solute families. For visualization, 13 outliers for the GNN ensemble and 12 outliers for the MCM ensemble with MAPE higher than 100 % are not shown.
  • Figure 4: GNN model ensembling parity plot for generalization test set. Red lines indicate $\pm$ 0.5 error range.