Journal of the Chilean Chemical Society
On-line version ISSN 0717-9707
J. Chil. Chem. Soc. vol.51 no.4 Concepción Dec. 2006
J. Chil. Chi. Soc., 51, N°.4 (2006), p.1034-1035
QSPR STUDY OF CORROSION INHIBITORS. IMIDAZOLINES
LEONEL VERA*, MIGUEL GUZMÁN Y PEDRO ORTEGA-LUONI
Departamento de Cs.Químicas y farmacéuticas, Facultad de Ciencias, Universidad Católica del Norte, Avda. Angamos 0610, Casilla 1280, Antofagasta. Chile.
A mathematical study of the QSAR/QSPR type is presented in relating the capacity of substituted imidazolines to the inhibition of steel corrosion. For this, we used structural parameters (quantumand topological).Experimental values from the literature are in agreement with theoretical values, with a low probability of random correlation.
Keywords:QSAR, QSPR, Cross validation, Multiple Regression Models, randomization test.
A corrosion inhibitor is a chemical substance that when applied in small quantities to a corrosive medium, reduces the rate ofcorrosion in metals and or metal alloys. A molecular grouping of 17 imidazolines, (IMDZ), has been reported  as being a potentially good family of inhibitors of steel corrosion; these experimental data were used in this study with the objective of finding a mathematical model of the QSAR/QSPR type, which is able to relate the inhibitory efficiency of the imidazolines with structural parameters (quantum, and/or topological) which can be theoretically calculated, with the ultimate aim ofobtaining a method for the molecular design of corrosion inhibitors.
The computational process for obtaining a multilinear regression model relating inhibitory efficiency of the imidazolines and the six major structural parameters encountered for these molecules consists of the following steps:
1.- Drawing of structures, gross optimization by molecular mechanics and obtaining of atomic coordinates for input to Mopac. The HyperChem versión 4.5 was used.
2.- Calculation of the molecular quantum parameters, to be tested in the search for a QSAR model. A Mopac version 6.0 program was used , following the AM1 method.
3.- Calculation of the molecular topological parameters to be tested in the search for the QSAR model. The program used was MolconnZ versión 3.1.
4.- Testing of the preceding parameters in the search for optimal models.The necessary programs were coded in FORTRAN (Lahey/Fujitsu LF95 PRO v5.7).
RESULTS AND CONCLUSIONS
The numbering and structures of the molecules in this study are shown in Table I. The last column of the table presents the experimental value of the property studied ( Picorr= -log icorr) for each of theimidazolines.
The definition of each of the six best descriptors found, in terms of predictive capacity is given in the following. The first four descriptors X1 (Molecular electronegativity (eV));X2, (Heat of Formation, kcal/mol);X3 (Total Electronic Energy , eV);X4(Charge in N1, uac) are of a quantum nature . The remaining two descriptors, X5(Topological index of connectivity1cp) y X6 (Topological index of form1ka) are ofthe topological type .
The values calculated for each of the six descriptors for the 17 molecules studied are presented in Table II. On the extreme right side of this Table we compare the experimentally calculated values (Picorr (*)) using the model given below each. The average error in the molecular family was about 1%.
Figure 1 allows visual appreciation of the correlation between the experimental values and those calculated for the property modeled.
On the extreme right side of this Table we compare the experimentally calculated values (Picorr (*)) using the model given below each. The average error in the molecular family was about 1%.
Finally, we recognize that given the size of the population of the molecular group studied, the number of descriptors used to obtain a model such as that described is higher than that which would be considered ideal.. Nevertheless, and observing Figure 2, it can be concluded that the probability of obtaining good results with this model by purely random means is extremely low.
This Figure was constructed by randomly disordering the vector of the properties such that each line of descriptors characteristic of a particular molecule would remain aligned -with a very high probability- with a value of the property Picorrrepresenting the other molecule.This was done for the entire molecular group, searching for the best regression model, iterating the process 10,000 times. A plot of r2 versus r2cv was made for each of these models, using random values for the experimental property. The graph shows the distribution obtained. The largest point depicted in the extreme upper right represents the condition in which each line of descriptors is aligned with the true value of the experimental property reported.
The authors thank the General Research and Postgraduate Office of the Universidad Católica del Norte for supporting this research.
1. P. Dupin, A de Savignac, A. Lattes, Werstoffe und Korrosion, 1982, 33, 203-206. [ Links ]
2. R. Karelson, V. Lovanov,Chem. Rev.1996, 96, 1027-1043. [ Links ]
3. L. B. Kier, L. H. Hall, "Molecular Structure Description: The ElectrotopologicalState", Academic Press, New York,1999. [ Links ]