Machine Learning of Dynamic Electron Correlation Energies from Topological Atoms

James L McDonagh, Arnaldo F Silva, Mark Vincent, Paul Popelier

    Research output: Contribution to journalArticlepeer-review


    We present an innovative method for predicting the dynamic electron correlation energy of an atom or a bond in a molecule utilizing topological atoms. Our approach uses the machine learning method Kriging (Gaussian Process Regression with a non-zero mean function) to predict these dynamic electron correlation energy contributions. The true energy values are calculated by partitioning the MP2 two-particle density-matrix via the Interacting Quantum Atoms (IQA) procedure. To our knowledge, this is the first time such energies have been predicted by a machine learning technique. We present here three important proof-of-concept cases: the water monomer, the water dimer and the van der Waals complex H2…He. These cases represent the final step towards the design of a full IQA potential for molecular simulation. This final piece will enable us consider situations in which dispersion is the dominant inter-molecular interaction. The results from these examples suggest a new method by which dispersion potentials for molecular simulation can be generated.
    Original languageEnglish
    JournalJournal of Chemical Theory and Computation
    Early online date6 Dec 2017
    Publication statusPublished - 2017


    • Electron correlation
    • Machine learning
    • FFLUX
    • Force field
    • Interacting Quantum Atoms (IQA)

    Research Beacons, Institutes and Platforms

    • Manchester Institute of Biotechnology


    Dive into the research topics of 'Machine Learning of Dynamic Electron Correlation Energies from Topological Atoms'. Together they form a unique fingerprint.

    Cite this