Müller Julian, Hartke Bernd
Institute for Physical Chemistry, University of Kiel , Olshausenstrasse 40, 24098 Kiel, Germany.
J Chem Theory Comput. 2016 Aug 9;12(8):3913-25. doi: 10.1021/acs.jctc.6b00461. Epub 2016 Jul 26.
Mechanochemistry, in particular in the form of single-molecule atomic force microscopy experiments, is difficult to model theoretically, for two reasons: Covalent bond breaking is not captured accurately by single-determinant, single-reference quantum chemistry methods, and experimental times of milliseconds or longer are hard to simulate with any approach. Reactive force fields have the potential to alleviate both problems, as demonstrated in this work: Using nondeterministic global parameter optimization by evolutionary algorithms, we have fitted a reaxFF force field to high-level multireference ab initio data for disulfides. The resulting force field can be used to reliably model large, multifunctional mechanochemistry units with disulfide bonds as designed breaking points. Explorative calculations show that a significant part of the time scale gap between AFM experiments and dynamical simulations can be bridged with this approach.
机械化学，尤其是单分子原子力显微镜实验形式的机械化学，在理论上难以建模，原因有二：单行列式、单参考量子化学方法无法准确捕捉共价键断裂，且任何方法都难以模拟毫秒或更长时间的实验。如本工作所示，反应力场有潜力缓解这两个问题：通过进化算法进行非确定性全局参数优化，我们已将一个反应力场（reaxFF）拟合到二硫化物的高水平多参考从头算数据。所得力场可用于可靠地模拟以二硫键为设计断裂点的大型多功能机械化学单元。探索性计算表明，这种方法可以弥合原子力显微镜实验与动力学模拟之间时间尺度差距的很大一部分。
