The measurement and modelling of the intracrystalline diffusion of cyclohexane in ZSM-5 using zero length column chromatography

Abstract

Zeolites are important solid acid catalysts largely because of their microporous nature giving them strong shape selective properties. This shape selectivity may be improved by depositing an inert (e.g. silaceous) layer on the surface of the catalyst crystals, which inertises non-shape selective surface acid sites and increases diffusional transport resistances. This thesis presents an investigation, using the zero length column (ZLC) method, into changes in the diffusional properties of cyclohexane in ZSM-5 that has had tetraethoxysilane deposited on the surface in the liquid phase. Theoretical analyses of the ZLC technique and its use in a detailed study of cyclohexane in unmodified ZSM-5 were performed in order to prove the reliability of the technique and measure baseline behaviour of the selected system. The ZLC method is a chromatographic technique for the measurement of diffusion coefficients in porous sorbents. Originally developed for gaseous hydrocarbon / zeolite powder systems, it has been experimentally extended to the measurement of liquid sorbate systems, macroporous zeolite pellets and resins and the measurement of self-diffusion through tracer exchange. The technique is robust against the intrusion of external heat and mass transfer effects by the use of relatively high flow rates and small catalyst samples. Analysis of the desorption curves is simple: the slope of the linear (on semilogarthmic axes) long time region gives the diffusional time constant. This so-called 'long time' analysis gives an accurate result for the diffusional time constant. The standard analysis is based on the carrier fluid phase being well mixed throughout the sorbent bed, which is reasonable for a physically short column (i.e. with a small Peclet number) and that the sorbent particles are monosized with a linear adsorption isotherm. It is also assumed that sorbate accumulation in the fluid phase is negligible. Theoretical studies, most backed by experimental data, have shown that most deviations from these assumptions do not preclude the accurate estimation of the diffusional time constant by the 'long time' method. By deriving an alternative model with the fluid phase showing plug flow (infinite Peclet number) behaviour, it was shown that the behaviour of the system becomes independent of the flow pattern in the column for the characteristic dimensionless group L > 15, irrespective of the physical length of or (axial) flow pattern in the column. This behaviour was verified through use of a third model incorporating finite axial dispersion. This gives a 'zero length' criterion, by which it may be ensured that systems conform to the 'zero length' assumptions of the standard (mixed flow) model. To investigate the effect of a distribution of sorbent particle sizes (the standard model assumes a monosized sample) on the desorption curve and measured (using the standard model) parameters, a model was derived which incorporates a log-mean distribution of crystal radii. It was found that, for this case, the normally linear (on semilogarithmic axes) tail of the desorption curve became convex to the time axis. This is the expected result, since the slope of that region indicates the diffusional time constant (D/R2). A varying value of R would thus clearly lead to a varying slope and thus curvature. Analysing such a curve with the standard ZLC model would lead to an underprediction of D/R2 and an overprediction of L. The deviation, as expected, becomes more severe with an increasing distribution width. The size distribution model was applied to experimental data, but the fitted distribution width was found to be a function of L, indicating that the model could not describe the data consistently. This is possibly a result of the chosen distribution function not being representative of the sorbent sample. If an experimental size distribution were available, it was shown through a simple mathematical analysis that an adequate description of the desorption curve could be obtained by simply summing the responses of a number of discrete size fractions, weighted by volume fraction. Finally, numerical analysis indicated that, in the absence of more detailed information, systematic error in the diffusion coefficient caused by analysing a sorbent sample with a size distribution by means of the standard model could be reduced by calculating the mean radius of the particles as the ratio of the total volume of the sample to the total surface area. For the first part of the experimental study, the sorption rate of cyclohexane in three samples of ZSM-5 of different size and morphology was measured. It was found that the diffusion coefficients obtained using the standard analysis method agreed well with one another and with some literature data. The consistency between the large and small crystals' diffusion coefficients is contradictory to many results in the literature. This was attributed to the fact that most small crystals mentioned in the literature are of an industrial origin and might have been hydrothermally treated, whereas the small particles used in this work were laboratory synthesised with no such post-synthesis modification. While the individual experimental desorption curves appeared to be well described by the simple model, it was found that the adsorption constant appeared to be a function of purge flow rate. More specifically, L was not directly proportional to flow rate as required by the model, consistent with the expected deviations caused by a crystal size distribution. There were also differences caused by changes in the sorbate concentration during the equilibration stage of the experiment, particularly at low temperatures. This was the result of the presence of a Langmuir shaped adsorption isotherm. Finally, the effect on the diffusion behaviour of the cyclohexane / ZSM-5 system of depositing a silaceous layer on the external surface of the zeolite was investigated. Models were derived and analysed in order to determine whether the mechanisms of pore mouth narrowing (modelled by using a surface barrier approach) and pore mouth blockage (modelled as a decrease in area available for diffusive flux) could be differentiated. The former case should yield reduced values of diffusion and adsorption constants, while the latter should yield a reduced diffusion coefficient and an unchanged adsorption constant, relative to the unmodified catalyst. Experiments with samples of ZSM-5 modified by tetraethoxysilane in the liquid phase indicated that there was a significant (up to an order of magnitude, depending on the silanisation procedure) decrease in diffusivity. Although there was scatter in the measured adsorption constants, it appeared that these remained constant for all samples. It would thus appear that, at least for the sorbents used in this work, the change in transport rate due to silanisation was due to the pore mouth blockage mechanism.