DERIVATION OF INTERMOLECULAR INTERACTION POTENTIAL FUNCTIONS BASED ON THE CAMBRIDDGE STRUCTURAL DATABASE

由于缺乏合适的分子间作用力模型，原子模拟的分子间势函数的推导遇到了困难。我们在这里提出了一种全新的、非经验性的方法来解决这个问题，它使用剑桥结构数据库作为有机分子之间分子间相互作用的信息来源。该方法利用结构碎片三维几何的逆概率密度分布将结构信息转换为能量函数。我们以C=O...H-O酮/醇氢键为例进行了验证，并对其三维能量面进行了建模，使用的CSD 5.16版本包含了19万多个有机化合物的晶体结构。提取了高质量的氢键结构片段(631)，其结构参数的分布符合已知的氢键势能函数。我们发现改良的利平科特-SCH…的组合C=O…H-O型氢键的第一个三维势函数集令人满意地重现了著名的三维能量面，沿着羰基上非键轨道的方向具有浅极值和双极小值。将这一过程扩展到其他类型和形式的分子间相互作用，应该提供一组潜在的函数和一组适当的参数，对现实模拟有用，特别是对包括蛋白质、糖和DNA在内的生物分子的动态行为。众所周知，到目前为止，在空间项中最重要但最不被理解的是分子间相互作用，但我们的方法应该给出这个问题的现实解决方案。我们最终将获得的力场的雄心勃勃的应用之一将是仅根据化学式先验确定晶体结构。较少

Derivation of intermolecular potential functions for the atomistic simulation has met difficulties due the lack of appropriate models for intermolecular forces. We propose here an entirely new and non-empirical approach to this problem, which uses Cambridge Structural Database as the source of information on the intermolecular interactions among organic molecules. The method uses inverse probability density distribution of the 3D geometry of pertinent structural fragments to convert structural information into energy functions. We tested the idea by using C=O...H-O ketone/alcohol hydrogen bonding as a test case and modeled its 3D energy surface.Version 5.16 of CSD was used which contains more than 190,000 crystal structures of organic compounds. Hydrogen bonding structural fragments (631) with high quality were extracted and the distributions of their structural parameters were fit the known potential functions of hydrogen bonding. We found that a combination of modified Lippincott-Sch … More roder potential for bond stretch, single Gaussian for out-of-plane deformation and a linear combination of two Gaussian functions for in-plane deformation fits best.The first 3D potential function set for the C=O...H-O type hydrogen bond satisfactorily reproduced the well-known 3D energy surface having shallow and double minima along the direction of non-bonded orbitals on a carbonyl group. Extension of this procedure for other types and forms of intermolecular interactions should provide a set of potential functions and a set of appropriate parameters useful for realistic simulations, especially for the dynamic behaviors of bio-molecules including proteins, sugars and DNA. It is well-recognized that by far the most important but the least understood among the steric terms is the intermolecular interactions, but our approach should give a realistic solution to this problem.One of the ambitious applications of the force field that we will eventually obtain will be a priori determination of crystal structure from chemical formula alone. Less