Template burning inside TS-1 and Fe-MFI molecular sieves: an in situ XRPD study

The high X-ray flux available at the European Synchrotron Radiation Facility (ESRF), combined with the use of a suitably designed area detector setup, allowed us to follow in real time the structural changes occurring during the template burning processes inside TS-1 and Fe-silicalite MFI zeolites with a X-ray powder diffraction technique (XRPD). Rietveld analysis of the XRPD patterns collected in the 350-1000 K interval, integrated each 15 K, yields to the determination of the template overall occupancy factor versus T with an accuracy comparable with that obtained by thermogravimetric measurements, routinely employed for this purpose. The evolution of the structural parameters (V, a, b, c, site occupancy factor of the template molecule) vs T has been obtained. These data allow us to have, for the first time, a complete view of the structural rearrangements induced by the template burning process on the zeolitic framework. The differences caused by the different heteroatom inserted in the MFI lattice (Ti or Fe) are discussed. For both TS-1 and Fe-MFI, the kinetics of the reaction were investigated, to obtain the activation energy of the calcinations process employing the nonisothermal data according to the theory recently proposed by Kennedy and Clark Thermochim. Acta, 1997, 307, 27-35]. For TS-1 only, the time-resolved template burning experiment has been repeated in isothermal conditions at four different temperatures, to obtain the activation energy from isothermal data, according to the standard procedure. Comparison between Arrhenius plots obtained from isothermal and nonisothermal data demonstrates that the Kennedy and Clark method can be also applied to complex materials such as the MFI zeolites. This approach, when applied to time-resolved XRPD studies, is much less time consuming (requesting, in principle, one single nonisothermal run) with respect to the classic approach, which requests at least three isothermal runs. Moreover, it allows a remarkably lower associated error (151 +/- 11 versus 146 +/- 30 kJ mol(-)(1)) due to the much higher number of experimental points employed to perform the linear fit.