Physics‐Guided 3D Solute–Solvent Interaction Framework for infinite‐dilution activity coefficient prediction
Abstract Accurate prediction of temperature‐dependent infinite‐dilution activity coefficients (IDACs) is essential for solvent screening, extractive separation, and thermodynamic process design. On the benchmark dataset containing 39,840 samples, 695 solvents, 559 solutes, and 13,222 unique solute–solvent pairs over 249.84‐555.60 K, PGSSI achieves an MAE of 0.2043 and an of 0.9536 on the fixed test set. PGSSI shows robust interpolation/extrapolation, 3D‐aware gains on isomeric systems, strong performance on H‐bond‐rich and nonaqueous subsets, effective trend capture in water‐containing systems, and contact‐level interpretability for multi‐criteria, temperature‐dependent solvent screening.
Authors
- Liang Yang (ORCID: https://orcid.org/0000-0001-6291-4359)
- Xinchen Kang (ORCID: https://orcid.org/0000-0003-0593-0840)
- Wei Zhang (ORCID: https://orcid.org/0000-0002-0109-5334)
- Lang Jiang (ORCID: https://orcid.org/0000-0003-3624-1618)
- S.W. Yuan (ORCID: https://orcid.org/0000-0003-1874-9792)
- Jinlin Ye (ORCID: https://orcid.org/0009-0006-7353-7249)
- Yu Wang (ORCID: https://orcid.org/0000-0001-7278-8608)
- Hongchang Zhang (ORCID: https://orcid.org/0000-0002-4798-4364)
- Junhao Wang
- Deming Hu (ORCID: https://orcid.org/0009-0008-7349-8758)
- Kai Li (ORCID: https://orcid.org/0009-0000-3190-0317)
- Xin Yang (ORCID: https://orcid.org/0009-0002-8075-0647)
Institutions
- Hebei University of Technology (CN)
- University of Chinese Academy of Sciences (CN)
Publication Details
- Journal
- AIChE Journal
- Published
- 2026-09-08
- DOI
- https://doi.org/10.1002/aic.70597
- Primary Topic
- Process Optimization and Integration
- Type
- article
- Field-Weighted Citation Impact
- 0.00