Bosque Ramón, Sales Joaquim
Departament de Química Inorgànica, Universitat de Barcelona, Martí i Franquès, 1, 08028-Barcelona, Spain.
J Chem Inf Comput Sci. 2002 Sep-Oct;42(5):1154-63. doi: 10.1021/ci025528x.
From the experimental polarizability values, alpha, of a large set of solvents containing 426 compounds with very different chemical characteristics, an additive model for the estimation of the polarizability is proposed. The derived average atomic polarizability of 10 elements, C, H, O, N, S, P, F, Cl, Br, and I, allows the calculation of the molecular polarizability of solvents from their chemical composition alone, without any other structural consideration. The average errors are 2.31% and 1.93% for the working set of 340 solvents and the prediction set of 86 solvents, respectively. Semiempirical quantum methods, such as, AM1, PM3, and MNDO, gave errors of about 35%. The density functional theory (DFT) calculations give better results than the semiempirical ones but poorer results than those obtained by the additive approach.
基于包含426种具有非常不同化学特性的化合物的大量溶剂的实验极化率值α，提出了一种用于估算极化率的加和模型。推导出的10种元素（碳、氢、氧、氮、硫、磷、氟、氯、溴和碘）的平均原子极化率，使得仅根据溶剂的化学组成就能计算其分子极化率，而无需任何其他结构方面的考虑。对于340种溶剂的工作集和86种溶剂的预测集，平均误差分别为2.31%和1.93%。半经验量子方法，如AM1、PM3和MNDO，给出的误差约为35%。密度泛函理论（DFT）计算的结果比半经验方法的结果更好，但比通过加和方法获得的结果更差。
