Reilly Anthony M, Tkatchenko Alexandre
Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4-6, 14195 Berlin, Germany.
J Phys Chem Lett. 2013 Mar 21;4(6):1028-33. doi: 10.1021/jz400226x. Epub 2013 Mar 15.
The near endless possibilities for assembling molecular materials has long posed a difficult challenge for theory. All crystal-structure prediction methods acknowledge the crucial contribution of van der Waals or dispersion interactions, but few go beyond a pairwise additive description of dispersion, ignoring its many-body nature. Here we use two databases to show how a many-body approach to dispersion can seamlessly model both solid and gas-phase interactions within the coveted "chemical accuracy" benchmark, while the underlying pairwise approach fails for solid-state interactions due to the absence of many-body polarization and energy contributions. Our results show that recently developed methods that treat the truly collective nature of dispersion interactions are able to reach the accuracy required for predicting molecular materials, when coupled with nonempirical density functionals.
长期以来，组装分子材料几乎无穷无尽的可能性一直给理论带来难题。所有晶体结构预测方法都承认范德华力或色散相互作用的关键作用，但很少有方法超越色散的成对加和描述，而忽略其多体性质。在此，我们利用两个数据库来展示，多体色散方法如何能够在令人垂涎的“化学精度”基准内无缝模拟固相和气相相互作用，而由于缺乏多体极化和能量贡献，基础的成对方法在固态相互作用中失效。我们的结果表明，当与非经验密度泛函相结合时，最近开发的处理色散相互作用真正集体性质的方法能够达到预测分子材料所需的精度。
