Prediction of thermal stability of organic peroxides for REACH
Résumé
REACH requires the evaluation of the physico-chemical properties of a large number of substances in order to allow their use before 2018. Taking into account the number of substances and properties, the timing, the economic costs, the feasibility at the R&D level and the risks for the manipulator, in particular for the characterization of the dangerous physico-chemical properties (explosibility, flammability), the experimental measurement of all the data was not realistic. Thus, the development of alternative predictive methods for the evaluation of the properties of substances was recommended in the framework of REACH. In this context, the French PREDIMOL (molecular modeling prediction of physico-chemical properties of products) project funded by ANR (National Research Agency) and started in November 2010 for 3 years aimed to demonstrate that molecular modeling was a credible alternative method to experimental characterization to predict a large range of physico-chemical properties of organic peroxides and amines required by EU-REACH’s regulation (annexes VII and IX). In this work, the first predictive and validated QSPR models related to the thermal stability of organic peroxides were obtained after the acquisition of a robust experimental database consisting of 38 data using calorimetry (heat and temperature of decomposition). Multi-linear models (MLR) using different types of descriptors (constitutional, topological and quantum chemical) were derived respecting the OECD guidelines for regulatory acceptability of QSPRs. They were tested by a series of internal and external validation tests and good performances in terms of fitting, robustness and predictive power were obtained compared to previous work. Their applicability domains were defined and analyzed. Considering mechanistic interpretation, the descriptors used in these models were linked to the peroxide bond whose breaking represents the main decomposition mechanism of organic peroxides. Being rigorously validated, the predictive data issued from these models could be used for registration of chemicals in REACH.