Prediction of Thermal Decomposition Temperature of Polymers Bull. Korean Chem. Soc. 2008, Vol. 29, No. 10 2009 Prediction of Thermal Decomposition Temperature of Polymers Using QSPR Methods Davood Ajloo, * Ali Sharifian, and Hossein Behniafar School of Chemistry, Damghan University of Basic Science, Damghan, Iran. * E-mail: ajloo@dubs.ac.ir Received February 26, 2008 The relationship between thermal decomposition temperature and structure of a new data set of eighty monomers of different polymers were studied by multiple linear regression (MLR). The stepwise method was used in order to variable selection. The best descriptors were selected from over 1400 descriptors including; topological, geometrical, electronic and hybrid descriptors. The effect of number of descriptors on the correlation coefficient (R) and F-ratio were considered. Two models were suggested, one model having four descriptors (R 2 = 0.894, Q 2 cv = 0.900, F = 172.1) and other model involving 13 descriptors (R 2 = 0.956, Q 2 cv = 0.956, F = 125.4). Key Words : Thermal decomposition temperature, Geometrical descriptors, Semi-empirical, Cross validation, Polymer Introduction Nowadays, before synthesizing a new material, database searching can provide information on related reactions to assist in designing viable pathways to synthesize it. In addi- tion, computational chemistry can help to determine whether these pathways are favored from the thermodynamic point of view. Quantitative structure-property relationships (QSPR) studies can help to predict the properties of the compound of interest. The concept of QSPR is to transform searches for compounds with desired properties using chemical institu- tion and experience into a mathematically quantified and computerized form. Once a correlation between structure and property found, any number of compounds, including those not yet synthesized, can be readily screened on computer in order to select structures with the properties desired. Thus, QSPR approach conserves resources and accelerates the process of development of new molecules for use as any purpose. The different properties of polymer such as thermal stabi- lity, viscosity and dielectric constant, have been studied by QSPR. 1-7 Thermal stability of a material is stability against degradation upon exposure to elevated temperatures in an inert environment. Polymers are often exposed to high temperatures during processing and/or use. Thus, thermal stability is among the most important properties of polymers for a wide range of applications. 8 Thermal stability can be expressed with the temperature of ten percent of decom- position T d,10% , which is defined as temperature at which 10% loss of weight during pyrolysis was recorded by TGA at a heating rate of 10 °C/min. It can be also described with the molar thermal decom- position function Y d,1/2 that correlates T d,1/2 , with an equation: Y d,1/2 = T d,1/2 × M (M is the molecular weight per repeating unit). T d,1/2 is defined same T d,10% , only for half decomposi- tion. Although there are numerous techniques for measuring thermal stability, it is difficulty to interpret the data and/or to compare data obtained in different laboratories or under different test condition. 8,9 Some researchers have found that Y d,1/2 values for polymers can be estimated on the basis of quantitative structure–property relationship models. 10-13 One of the best-known examples of the group additive approach is that of van Krevelen. 10 This method provides a rapid and computationally inexpensive approach to the estimation of Y d,1/2 value, but is purely empirical approach and limited to systems composed only of functional groups that have been previously investigated. Furthermore, the group contributions method is only approximate since this approach fails to account for the presence of neighboring groups or conformational influences. Bicerano extended the group additive concept for the prediction of Y d,1/2 , based on 21 topological and constitutional descriptors. 8 There are too many descriptors involved in the models though the predic- tions are good accuracy. In addition, Sun et al. obtained two kinds of calculated models on polymeric Y d,1/2 by artificial neural networks for linear chain polymers. 11 After this work, Sun et al. further built two models on polymeric Y d,1/2 for the same polymeric data set by fuzzy set theory. 12 The group average method is used to calculate the descriptors for two models, and the connectivity indexes method is used for them. On the other hand, the quantum chemical descriptors used in QSPR models encode information about the electronic structure of the molecule and thus implicitly account for the cooperative effects between functional groups, charge redis- tribution, and possible hydrogen bonding in the polymer. 14 The two quantum chemical descriptors, obtained from the monomers of vinyl polymers were used to predict the molar thermal decomposition function Y d,1/2 . A more physically meaningful quantitative structure-property relationship (QSPR) model obtained from the training set with multiple linear stepwise regression analysis. 7 The goal is to find one equation that is function of a small number of structure-based molecular descriptors that accu- rately predicts the experimental property. We must balance