Improved vapor pressure predictions using group contribution-assisted graph convolutional neural networks (GC<sup>2</sup>NN)

Krüger, Matteo; Galeazzo, Tommaso; Eremets, Ivan; Schmidt, Bertil; Pöschl, Ulrich; Shiraiwa, Manabu; Berkemeier, Thomas

doi:https://doi.org/10.5194/gmd-18-7357-2025

Articles | Volume 18, issue 20

https://doi.org/10.5194/gmd-18-7357-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-18-7357-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 18, issue 20

Model description paper

|

15 Oct 2025

Model description paper |

| 15 Oct 2025

Improved vapor pressure predictions using group contribution-assisted graph convolutional neural networks (GC²NN)

Matteo Krüger, Tommaso Galeazzo, Ivan Eremets, Bertil Schmidt, Ulrich Pöschl, Manabu Shiraiwa, and Thomas Berkemeier

Supplement

https://doi.org/10.5194/gmd-18-7357-2025-supplement

Data sets

Code and data for 'Improved vapor pressure predictions using group contribution-assisted graph convolutional neural networks (GC2NN)' Matteo Krueger and Thomas Berkemeier https://doi.org/10.17617/3.GIKHJL

Model code and software

Code and data for 'Improved vapor pressure predictions using group contribution-assisted graph convolutional neural networks (GC2NN)' Matteo Krueger and Thomas Berkemeier https://doi.org/10.17617/3.GIKHJL

Short summary

This work uses machine learning to predict saturation vapor pressures of atmospherically-relevant organic compounds, crucial for partitioning of secondary organic aerosol (SOA). We introduce a new method using graph convolutional neural networks, in which molecular graphs enable the model to capture molecular connectivity better than with non-structural embeddings. The method shows strong agreement with experimentally determined vapor pressures, and outperforms existing estimation methods.