Accurate quantum chemistry in the condensed phase.


Developing practical new theoretical models for describing non-covalent interactions and predicting the structures and properties of molecular crystals.

Software developed by the Beran Group

HMBI: Hybrid Many-Body Interaction

  • Github link:

    https://github.com/gberan/HMBI

  • Key Literature References:

    "Approximating quantum many-body intermolecular interactions in molecular clusters using classical polarizable force fields." Gregory Beran. J. Chem. Phys. 130, 164115 (2009).

    "Predicting organic crystal lattice energies with chemical accuracy." Gregory Beran and Kaushik Nanda. J. Phys. Chem. Lett. 1, 3480-3487 (2010).

  • Description:

    The HMBI code is used to perform fragment-based calculations in molecular crystals. It can perform energy, gradient, hessian, and NMR chemical shielding calculations. It relies on external software packages such as Q-Chem, Gaussian, Molpro, or PSI4 for the electronic structure and Tinker for the polarizable force field.

MP2D: Dispersion-corrected MP2

  • Github link:

    https://github.com/Chandemonium/MP2D

  • Literature Reference:

    "Accurate non-covalent interactions via dispersion-corrected second-order Moller-Plesset perturbation theory" J. Rezac, C. Greenwell, and G. Beran. J. Chem. Theory Comput. 14, 4711-4721 (2018).

  • Description:

    MP2D addresses the well-known problems of MP2 for describing van der Waals dispersion interactions. Like the closely related MP2C model, this simple, inexpensive correction subtracts out the erroneous dispersion present in MP2 and replaces it with a better version. Because it is based on Grimme's D3 dispersion correction, the dispersion correction takes negligible effort to compute, it can be applied to inter- and intramolecular interactions, and it is differentiable for use in geometry optimizations.

    The MP2D code here reads standard XYZ format geometries and returns the dispersion-correction energy and gradient, which can then be paired with MP2 results from other quantum chemistry software packages.

Δ-ML for NMR Chemical Shieldings

  • Github link:

    https://github.com/pablo-unzueta/dml-nmr

  • Literature Reference:

    "Predicting Density Functional Theory-Quality Nuclear Magnetic Resonance Chemical Shifts via Δ-Machine Learning" P. Unzueta, C. Greenwell, and G. Beran. J. Chem. Theory Comput. 17, 826-840 (2021). DOI: 10.1021/acs.jctc.0c00979

  • Description:

    Training data and neural network implementation of our Δ-ML for predicting small-molecule chemical shieldings.

Data Sets

3B-69 Data Set: Three-body interactions in trimers