(153b) Rational Prediction of Flash Points Based On COSMO-RS

The knowledge of the flash points of a chemical compound or a mixture is crucial for a save handling. Hence it is required within many regulatory and registration procedures. Given the importance of the flash point it is surprising to see the considerable spread of experimental data, which demonstrates the experimental difficulties of reproducible flash point measurements.

As a result a major interest in flash point prediction has grown, resulting in a number of publications with QSAR prediction methods for flash points. The drawback of these methods is that they either explicitly are class specific and hence restricted to compounds out of special classes, or they include descriptors which have nothing to do with the property itself, with the effect that such predictions will fail if they are used out of the space of the fit data set. In this paper we present a novel prediction method, which is based on a vapour pressure predictions with the widely validated and predictive COSMO-RS method, combined with a reasonable estimate of radical reactivity. In essence this model is based on the rational assumption that the flash point is mainly defined by reaching a critical radical reactivity per volume, i.e. by a product of vapour pressure and molecular reactivity.

On the available, very noisy data sets our COSMOflash model yields about the same statistical accuracy as the previously published QSAR approaches, but due to is very small number of adjusted parameters and due to ist rational construction it should have a higher intrinsic accuracy and it should be much more predictive for new compounds.