Model dependence of the activation energy derived from nonisothermal kinetic data

This work intends to explain mathematically the model dependence of the activation energy, E-a, derived from fitting nonisothermal kinetic data to a kinetic model. Artificial data following single reaction mechanisms, both isothermal and nonisothermal, were generated to fit exactly the following simulated kinetic models: first-order reaction, fourth-order Avrami-Erofeyev process, and one-dimensional diffusion. To simulate more closely experimental data, random errors, corresponding to +/-0.1% of the maximum conversion value, were embedded in the data by adding a random number bounded by +/-0.001 to each data point. For isothermal data, any kinetic model leads to the correct value of E-a. However, for nonisothermal data, the calculated E-a deviates from the correct value by an amount, DeltaE(a), that depends strongly on the kinetic model to which the data are fit. In addition, the apparent frequency factor depends slightly on the kinetic model for isothermal data, but depends strongly on the model for nonisothermal data. The results highlight the severe limitations of fitting nonisothermal data to kinetic models.