Advances in Ab Initio Modeling of the Many-Body Effects of Dispersion Interactions in Functional Organic Materials

Abstract

Accurate treatment of the long-range electron correlation energy, including dispersion interactions, is essential for describing the structure, dynamics, and function of a wide variety of systems. Among the most accurate models for including dispersion into density functional theory (DFT) is the range-separated many-body dispersion (MBD) method [A. Ambrosetti et al., J. Chem. Phys. 2014, 140, 18A508], in which the long-range correlation energy is computed from a model system of coupled quantum harmonic oscillators. In this work, we seek to extend the applicability of the MBD model by developing the analytical gradients necessary to compute MBD corrections to ionic forces, unit-cell stresses, phonon modes, and self-consistent updates to the Kohn-Sham potential. We include all implicit coordinate dependencies arising from charge density partitioning, as we find that neglecting these terms leads to unacceptably large relative errors in the MBD forces. Such errors would impact the predictive nature of ab initio molecular dynamics simulations employing MBD. We develop a new efficient implementation of the MBD correlation energy and forces within the Quantum ESPRESSO software package and rigorously test its numerical stability and convergence properties for condensed phase simulations. Additionally, we re-parameterize the MBD model for use with a wide variety of generalized gradient approximation exchange-correlation functionals. We demonstrate the efficiency and accuracy of these MBD gradient corrections for optimizations of isolated dispersively bound molecular systems, as well as representative condensed phase systems including adsorbed hydrocarbons, layered materials, and hydrogen-bonded crystals. Where highly accurate reference geometries are available, we find the DFT+MBD method significantly improves the predicted structures of these systems and consistently outperforms popular pairwise-additive DFT-D dispersion corrections. Though significant work remains in the benchmarking and testing of these contributions to the MBD model, we are optimistic that these methodological developments will enable many exciting discoveries of beyond-pairwise dispersive effects in organic materials.