Artificial neural networks for the prediction of solvation energies based on experimental and computational data

Jiyoung Yang, Matthias J. Knape, Oliver Burkert, Virginia Mazzini, Alexander Jung, Vincent S.J. Craig, Ramón Alain Miranda-Quintana, Erich Bluhmki, Jens Smiatek*

*Corresponding author for this work

    Research output: Contribution to journalArticlepeer-review

    15 Citations (Scopus)

    Abstract

    The knowledge of thermodynamic properties for novel electrolyte formulations is of fundamental interest for industrial applications as well as academic research. Herewith, we present an artificial neural networks (ANN) approach for the prediction of solvation energies and entropies for distinct ion pairs in various protic and aprotic solvents. The considered feed-forward ANN is trained either by experimental data or computational results from conceptual density functional theory calculations. The proposed concept of mapping computed values to experimental data lowers the amount of time-consuming and costly experiments and helps to overcome certain limitations. Our findings reveal high correlation coefficients between predicted and experimental values which demonstrate the validity of our approach.

    Original languageEnglish
    Pages (from-to)24359-24364
    Number of pages6
    JournalPhysical Chemistry Chemical Physics
    Volume22
    Issue number42
    DOIs
    Publication statusPublished - 14 Nov 2020

    Fingerprint

    Dive into the research topics of 'Artificial neural networks for the prediction of solvation energies based on experimental and computational data'. Together they form a unique fingerprint.

    Cite this