量子化学参数研究聚苯乙烯类玻璃化温度的定量构效关系  

QSPR study on glass transition temperatures of polystyrenes by using Quantum - Chemistry descriptors

作　　者：刘万强[1] 王学业[1] 李新芳 龙清平[1] 文小红[1] 李建军[2] 

机构地区：[1]湘潭大学化学学院,湖南湘潭411105 [2]江苏出入境检验检疫局,江苏南京210005

出　　处：《计算机与应用化学》2005年第9期753-758,共6页Computers and Applied Chemistry

基　　金：湖南省自然科学基金项目(02JJY2019)湖南省中青年科技基金项目(01JZY2099).

摘　　要：用密度泛函理论在6-31G(D)基组上优化了67种聚苯乙烯类结构单元,得到了其单元的结构参数,探讨了这些参数与聚苯乙烯类玻璃化转变温度(Tg)的关系,计算表明,影响聚苯乙烯类Tg的因素有苯环上取代基长度,取代基的分支数,取代基位置,原子的最大负电荷数,氢原子的最大正电荷数,主链碳原子之间的电子布居数,分子偶极矩以及分子的内能等参数。用逐步回归,岭回归和人工神经网络方法分别建立了这些参数与Tg的定量关系,并对这3种方法建立的定量模型进行了比较,3种方法的计算结果与实验值的相关系数分别为0.958 3,0.949 1,0.985 3,标准偏差分别为18.44,19.18,10.41,人工神经网络的拟合值和预测值均优于前2种线性回归方法。The structural units of sixty-seven polystyrenes have been optimized and their quantum-chemical descriptors have been obtained by DFT/6-31G （D） method. The calculated results indicate that the length of substituent, branch numbers of substituents, position of substituent, the most negative net charge of atom, the most positive net charge of hydrogen atom, thermal energy and dipole moment are main factors affecting glass transition temperature. The quantitative relationship have been studied between these descriptors and Ts by using stepwise, ridge regression and BP-ANN methods, The related coefficients between the fitted results and experimental data of Ts for three methods are 0.958 3, 0. 949 1, 0. 985 3, standard deviations are 18.44, 19. 18, 10.41 respectively, the artificial neural networks has advantage over other two kinds of linear regression models in the fitting value and the predicted value.

关 键 词：聚苯乙烯 玻璃化温度 量子化学参数 逐步回归 岭回归 人工神经网络 

分 类 号：O6-04[理学—化学] O631