A New Group Contribution Method for Estimating Critical Properties of

For the estimation of properties of pure compounds, simple group contribution methods (Joback and Reid,1 Lyderson,2Ambrose,3 Klincewicz and Reid,4 and...
0 downloads 0 Views 46KB Size
Ind. Eng. Chem. Res. 2001, 40, 6245-6250

6245

A New Group Contribution Method for Estimating Critical Properties of Organic Compounds Xu Wen* and Yang Qiang School of Chemical Engineering, Tianjin University, Tianjin 300072, China

A new division method of molecular structural units involving group adjacent atom pairs has been developed. Expressions for estimating critical properties Tc, Pc, and Vc have been proposed, with the numerical values of relative structural unit parameters presented. The average percent deviations of estimation of the above three physical properties are 0.73, 2.47, and 1.41, respectively. Introduction In the study of a chemical, biochemical, or environmental system, knowledge of a variety of physical properties of the compounds that undergo transformation is required. However, it is not always possible to find experimental values of properties for the compounds of interest in the literature, nor is it practical to measure the properties as the need arises. Therefore, methods for the estimation of properties are profusely employed. For the estimation of properties of pure compounds, simple group contribution methods (Joback and Reid,1 Lyderson,2 Ambrose,3 Klincewicz and Reid,4 and Lyman et al.5) have been widely used. These methods have the advantage of supplying quick estimates without requiring substantial computational resources. However, many of them are of questionable accuracy and utility and are unable to distinguish among isomers because of their oversimplification. To overcome the above limitations, complex group contribution methods (Marrero-Morejon and PardilloFontdevila6 and Constantinou and Gani7) have been reported in the literature. New problems emerged along with the increment of regression accuracy of these methods. Model parameters of the group method are obtained from fitting property data of a great many substances, which behave like a statistical feature. Only if the number of substances in linear data regression is much more than that of parameters in the model does the group method have a function of extrapolation predicting. Compared with the number of substances in regression, the more the parameter number in the model, the poorer is the predicting function of the model is. If the number of model parameters is more than that of the substances, the value of the model parameter solved will not be unique. For simple group methods, only a single functional group should be taken as the independent molecular structural unit. There are not more than 39 groups for various organic compounds in these methods. An addition of two or more adjacent simple groups is the same as the independent molecular structural unit for complex group methods. When only interaction between two adjacent simple groups is taken into account, the number of parameters in the property * To whom correspondence should be addressed. E-mail: [email protected].

correlation will be 223. The method proposed by Marrero-Morejon and Pardillo-Fontdevila6 was the group interaction contribution (GIC). Among their estimation of critical properties, the number of critical volume data in regression was only 289, which was not much larger than 223. The method proposed by Constantinou and Gani7 was the interaction of two or more adjacent simple groups; the parameter number in their model was still more than 223. For the Constantinou and Gani method, the estimation was performed at two levels: the basic level only used the contribution from the first-order groups, while the second level increased the consideration of the second-order groups. Their 78 first-order groups were not sufficient to describe the molecules of some common compounds, and under certain circumstances, the same molecule may be described in different ways because of overcomplication of their method. The purpose of this study is to develop a new group contribution method to limit the number of model parameters with higher accuracy. Division of Molecular Structural Units Group Containing Carbon and Adjacent Atom Pair. Almost all organic compounds comprise groups containing carbon more than any other group. In each group there are one or more atoms connecting with other groups through their free bonds and being adjacent atoms of other groups. Because some groups of different type contain the same atom connecting with the same adjacent group, if a group and adjacent atom pair are taken as the independent molecular structural unit, the sum of the structural units of molecules will be much less than that of adjacent group and group pairs. Taking group -CH3 as an example, the number of groups adjacent to it is 28 in the GIC method, while the number of atoms adjacent to it is only 11. To reduce the sum of the structural units further, groups containing carbon and adjacent atom pairs are selected as independent structural units. Group Structural Unit. To illustrate clearly, physical property F is expressed as follows:

F)a+

(0) (0) ∑k nk∆(0) k + ∑(ni∆i + ∑nij∆ij ) i i

(1)

where a is the correlation constant, subscripts k, i, and ij represent a group not containing carbon, a group containing carbon, and a group i-adjacent atom j pair,

10.1021/ie010374g CCC: $20.00 © 2001 American Chemical Society Published on Web 11/27/2001

6246

Ind. Eng. Chem. Res., Vol. 40, No. 26, 2001

Table 1. Group-Adjacent Atom Pair Structural Unit Parameters -H

>C
C
CC