H2S Solubility in Ionic Liquids Predicted by a Hybrid Neural Network

A physics-informed hybrid artificial neural network integrates the van’t Hoff equation to predict hydrogen sulfide solubility in ionic liquids. The model achieved a coefficient of determination (R²) exceeding 0.9987 on a conventional test set and maintained an R² of 0.9925 in a generalization test with unseen cation-anion combinations. The dataset, shared by Jiaping Zhou on figshare in June 2026, supports the rational design of ionic liquids for gas capture.

Use Cases

• Training physics-informed machine learning models for chemical property prediction based on the described hybrid ANN architecture.
• Benchmarking the generalization performance of data-driven models against unseen cation-anion combinations.
• Analyzing the influence of structural features like imidazolium cations and alkyl chain substituents on gas solubility.
• Designing next-generation ionic liquid solvents for hydrogen sulfide capture based on model-derived insights.

Strengths

• The hybrid model demonstrated high predictive accuracy with an R² exceeding 0.9987.
• Model maintained robust performance (R² 0.9925) in a rigorous generalization test where a purely data-driven model's R² dropped to 0.8215.
• Interpretability analysis using SHAP provides actionable insights into feature importance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is small in scale at 681.6 KB, indicating a limited scope.

Provenance

Source

figshare

Collection Method

Likely contains experimental or simulated data on H2S solubility in ionic liquids used to train and validate the described machine learning model.

Freshness

Last updated 2026-06-04 06:07:38; freshness should be verified.