MBD+C: how to incorporate metallic character into atom-based dispersion energy schemes

Abstract: The dispersion component of the van der Waals (vdW) interaction in low-dimensional metals is known to exhibit anomalous "Type-C non-additivity" [Int. J. Quantum Chem. 114, 1157 (2014)]. This causes dispersion energy behavior, at asymptotically large separations, that is missed by popular atom-based schemes for dispersion energy calculations. For example, the dispersion interaction energy between parallel metallic nanotubes at separation $D$ falls off aymptotically as approximately $D^{-2}$, whereas current atom-based schemes predict $D^{-5}$ asymptotically. To date it has not been clear whether current atom-based theories also give the dispersion interaction inaccurately at smaller separations for low-dimensional metals. Here we introduce a new theory that we term "MBD+C" . It permits inclusion of Type C effects efficiently within atom-based dispersion energy schemes such as Many Body Dispersion (MBD) and Universal MBD (uMBD). This allows us to investigate asymptotic, intermediate and near-contact regimes with equal accuracy. (The large contact energy of intimate metallic bonding is not primarily governed by dispersion energy and is described well by semi-local density functional theory.) Here we apply a simplified version,"nn-MBD+C", of our new theory to calculate the dispersion interaction for three low-dimensional metallic systems: parallel metallic chains of gold atoms, parallel Li-doped graphene sheets; and parallel (4,4) armchair carbon nanotubes. In addition to giving the correct asymptotic behavior, the new theory seamlessly gives the dispersion energy down to near-contact geometry, where it is similar to MBD but can give up to 15% more dispersion energy than current MBD schemes, in the systems studied so far. This percentage increases with separation until nn-MBD+C dominates MBD at asymptotic separations.