Department of Chemistry, Durham University, Lower Mountjoy, Stockton Road, Durham, DH1 3LE, UK.
Phys Chem Chem Phys. 2019 Jan 23;21(4):1912-1927. doi: 10.1039/c8cp05889j.
The performance of three methods for developing new coarse-grained models for molecular simulation is critically assessed. Two bottom-up approaches are employed: iterative Boltzmann inversion (IBI) and the multiscale coarse-graining method (MS-CG), using an atomistic n-octane-benzene reference system. Results are compared to a top-down coarse-graining approach employing the SAFT-γ Mie equation of state. The performance of each methodology is assessed against the twin criteria of local structure prediction and accurate free energy representation. In addition, the transferability of the generated potentials is compared across state points. We examine the extent to which the IBI methodology can be improved by using a multi-reference approach (MS-IBI), and demonstrate how a pressure correction can be employed to improve the results for the MS-CG approach. Additionally, we look at the effect of including angle-terms in the SAFT-γ Mie model. Finally, we discuss in detail the strengths and weaknesses of each method and suggest possible ways forward for coarse-graining, which may eventually address the problems of structure prediction, thermodynamic consistency and improved transferability within a single model.
三种用于开发分子模拟新粗粒模型的方法的性能受到严格评估。采用两种自下而上的方法：迭代玻尔兹曼反演（IBI）和多尺度粗粒化方法（MS-CG），使用原子 n-辛烷-苯参考体系。结果与采用 SAFT-γ Mie 状态方程的自上而下的粗粒化方法进行比较。每种方法的性能都根据局部结构预测和准确自由能表示这两个双重标准进行评估。此外，还比较了生成的势能在状态点之间的迁移能力。我们研究了通过使用多参考方法（MS-IBI）改进 IBI 方法的程度，并演示了如何使用压力校正来改进 MS-CG 方法的结果。此外，我们还研究了在 SAFT-γ Mie 模型中包含角度项的效果。最后，我们详细讨论了每种方法的优缺点，并为粗粒化提出了可能的前进方向，这最终可能会在单个模型中解决结构预测、热力学一致性和改进迁移能力的问题。
