Saint-Martin Humberto, Hernández-Cobos Jorge, Ortega-Blake Iván
Centro de Ciencias Físicas, Universidad Nacional Autónoma de México, Apartado Postal 48-3, 62251 Cuernavaca, Morelos, Mexico.
J Chem Phys. 2005 Jun 8;122(22):224509. doi: 10.1063/1.1931567.
Up to now it has not been possible to neatly assess whether a deficient performance of a model is due to poor parametrization of the force field or the lack of inclusion of enough molecular properties. This work compares several molecular models in the framework of the same force field, which was designed to include many-body nonadditive effects: (a) a polarizable and flexible molecule with constraints that account for the quantal nature of the vibration [B. Hess, H. Saint-Martin, and H. J. C. Berendsen, J. Chem. Phys. 116, 9602 (2002), H. Saint-Martin, B. Hess, and H. J. C. Berendsen, J. Chem. Phys. 120, 11133 (2004)], (b) a polarizable and classically flexible molecule [H. Saint-Martin, J. Hernandez-Cobos, M. I. Bernal-Uruchurtu, I. Ortega-Blake, and H. J. C. Berendsen, J. Chem. Phys. 113, 10899 (2000)], (c) a polarizable and rigid molecule, and finally (d) a nonpolarizable and rigid molecule. The goal is to determine how significant the different molecular properties are. The results indicate that all factors--nonadditivity, polarizability, and intramolecular flexibility--are important. Still, approximations can be made in order to diminish the computational cost of the simulations with a small decrease in the accuracy of the predictions, provided that those approximations are counterbalanced by the proper inclusion of an effective molecular property, that is, an average molecular geometry or an average dipole. Hence instead of building an effective force field by parametrizing it in order to reproduce the properties of a specific phase, a building approach is proposed that is based on adequately restricting the molecular flexibility and/or polarizability of a model potential fitted to unimolecular properties, pair interactions, and many-body nonadditive contributions. In this manner, the same parental model can be used to simulate the same substance under a wide range of thermodynamic conditions. An additional advantage of this approach is that, as the force field improves by the quality of the molecular calculations, all levels of modeling can be improved.
到目前为止，还无法准确评估模型的缺陷表现是由于力场的参数化不佳还是缺乏足够分子性质的纳入。这项工作在同一力场框架内比较了几种分子模型，该力场旨在包含多体非加和效应：(a) 一个具有可极化且灵活的分子，并带有考虑振动量子性质的约束条件 [B. 赫斯、H. 圣马丁和 H. J. C. 贝伦德森，《化学物理杂志》116, 9602 (2002)，H. 圣马丁、B. 赫斯和 H. J. C. 贝伦德森，《化学物理杂志》120, 11133 (2004)]，(b) 一个可极化且经典灵活的分子 [H. 圣马丁、J. 埃尔南德斯 - 科博斯、M. I. 贝尔纳尔 - 乌鲁丘尔图、I. 奥尔特加 - 布莱克和 H. J. C. 贝伦德森，《化学物理杂志》113, 10899 (2000)]，(c) 一个可极化且刚性的分子，最后 (d) 一个不可极化且刚性的分子。目标是确定不同分子性质的重要程度。结果表明所有因素——非加和性、极化率和分子内灵活性——都很重要。不过，可以进行近似处理，以便在预测准确性略有下降的情况下降低模拟的计算成本，前提是这些近似通过适当纳入有效分子性质来平衡，即平均分子几何结构或平均偶极矩。因此，不是通过对力场进行参数化以重现特定相的性质来构建有效力场，而是提出一种构建方法，该方法基于充分限制适合单分子性质、对相互作用和多体非加和贡献的模型势的分子灵活性和/或极化率。通过这种方式，相同的母模型可用于在广泛的热力学条件下模拟同一种物质。这种方法的另一个优点是，随着力场因分子计算质量的提高而改进，所有建模水平都可以得到提升。
