Learning Intermolecular Electronic Coupling from a Molecular-Orbital-Pair-based Descriptor

09 October 2023, Version 1
This content is a preprint and has not undergone peer review at the time of posting.


In this work, we proposed a graphic molecular-orbital-pair-based descriptor to predict intermolecular electronic couplings (ECs). We transform the numeric grid points data of the molecular orbitals (MOs) involved in the charge or energy transfer process by reducing the thousands of grid points to a few points based on the lobes of MO. The simplified representation consists of two parts, the integral value over the space and the coordinates of centroid of each lobe. To express the intermolecular interaction for a donor-acceptor pair, each simplified MO representation from the donor is paired to the one from the acceptor. Then, the obtained MO-pair representation is the descriptor for learning the intermolecular ECs, including hole transfer (HT), electron transfer (ET) and Dexter energy transfer (DET) coupling, with a multi-layer perceptron (MLP) model. The accuracy of our model reached at a promising level of 0.1~5 meV comparing with the quantum chemistry calculations. The data size dependence tests shows that our model has exhibited good ability on low-shot learning. Therefore, the graphic MO-pair-based descriptor has been proved to be capable to characterize MO interaction features.


Eletronic coupling
Molecular orbital
Machine learning

Supplementary materials

Supplementary Information: Learning Intermolecular Electronic Coupling from a Molecular-Orbital-Pair-based Descriptor
Figures and tables for ET couplings


Comments are not moderated before they are posted, but they can be removed by the site moderators if they are found to be in contravention of our Commenting Policy [opens in a new tab] - please read this policy before you post. Comments should be used for scholarly discussion of the content in question. You can find more information about how to use the commenting feature here [opens in a new tab] .
This site is protected by reCAPTCHA and the Google Privacy Policy [opens in a new tab] and Terms of Service [opens in a new tab] apply.