Feyersinger Florian, Hartmann Peter E, Hoja Johannes, Reinholdt Peter, Libisch Florian, Kongsted Jacob, Puschnig Peter, Boese A Daniel
Department of Chemistry, University of Graz, Heinrichstraße 28/IV, 8010 Graz, Austria.
Department of Physics, University of Graz, 8010 Graz, Austria.
J Phys Chem A. 2024 Oct 24;128(42):9275-9286. doi: 10.1021/acs.jpca.4c02851. Epub 2024 Oct 15.
Due to the large number of interactions, evaluating interaction energies for large or periodic systems results in time-consuming calculations. Prime examples are liquids, adsorbates, and molecular crystals. Thus, there is a high demand for a cheap but still accurate method to determine interaction energies and gradients. One approach to counteract the computational cost is to fragment a large cluster into smaller subsystems, sometimes called many-body expansion, with the fragments being molecules or parts thereof. These subsystems can then be embedded into larger entities, representing the bigger system. In this work, we test several subsystem approaches and explore their limits and behaviors, determined by calculations of trimer interaction energies. The methods presented here encompass mechanical embedding, point charges, polarizable embedding, polarizable density embedding, and density embedding. We evaluate nonembedded fragmentation, QM/MM (quantum mechanics/molecular mechanics), and QM/QM (quantum mechanics/quantum mechanics) embedding theories. Finally, we make use of symmetry-adapted perturbation theory utilizing density functional theory for the monomers to interpret the results. Depending on the strength of the interaction, different embedding methods and schemes prove favorable to accurately describe a system. The embedding approaches presented here are able to decrease the interaction energy errors with respect to full system calculations by a factor of up to 20 in comparison to simple/unembedded approaches, leading to errors below 0.1 kJ/mol.
由于存在大量相互作用,对于大型或周期性系统评估相互作用能会导致计算耗时。典型例子包括液体、吸附质和分子晶体。因此,迫切需要一种既便宜又准确的方法来确定相互作用能和梯度。一种应对计算成本的方法是将大的团簇分割成较小的子系统,有时称为多体展开,这些子系统可以是分子或其部分。然后可以将这些子系统嵌入到更大的实体中,以代表更大的系统。在这项工作中,我们测试了几种子系统方法,并通过三聚体相互作用能的计算来探索它们的极限和行为。这里介绍的方法包括机械嵌入、点电荷、可极化嵌入、可极化密度嵌入和密度嵌入。我们评估非嵌入碎片化、量子力学/分子力学(QM/MM)和量子力学/量子力学(QM/QM)嵌入理论。最后,我们利用基于单体密度泛函理论的对称适配微扰理论来解释结果。根据相互作用的强度,不同的嵌入方法和方案被证明有利于准确描述一个系统。与简单/非嵌入方法相比,这里介绍的嵌入方法能够将相对于全系统计算的相互作用能误差降低多达20倍,从而使误差低于0.1 kJ/mol。