resumo
The growing interest on inorganic membrane applications is due to their potential in new industrial and research fields, and as alternative processes to more conventional operations. In particular, titanosilicate membranes offer important benefits over the classical zeolite ones, since they may be synthesised without organic templates, to avoid the subsequent calcination usually responsible for irreversible defects, exhibit novel possibilities of isomorphous framework substitution, allowing for fine-tuning catalytic and adsorption properties, and are able to separate mixtures based on differences in affinity and sieving effect. The main objectives of this work were: i) the dynamic characterization of the zeolite type membranes synthesised in our Associated Laboratory CICECO, by carrying out permeation experiments with pure gases and mixtures; ii) the development and validation of new models for the multicomponent mass transport across porous membranes by Maxwell-Stefan approach, taking into account the specific mechanisms found, particularly the surface diffusion contribution; and iii) modelling the experimental points measured, as well as data compiled from literature. In order to carry out the permeation essays, an experimental set-up has been designed, assembled, and tested. For pure gases, the main targets were the measurement of permeances at constant temperature, by varying the transmembrane pressure drop op ( ΔP ), and permeances at programmed temperature, by fixing ΔP . After that, the ideal selectivites were computed. With respect to mixtures, the determination of the real selectivities demanded the mole fractions in permeate and retentate. In the whole, three distinct supports (stainless-steel and α-alumina) and nineteen membranes of AM-3, ETS-10, ZSM-5, and zeolite 4A have been studied, employing H2, He, N2, CO2, and O2. The first exploratory evaluation of membrane quality was accomplished by permeating nitrogen at room temperature. Accordingly, permeances higher than 10−6 mol/m2s.Pa pointed out rough defects, inducing us to perform additional crystallizations over the first layers. This procedure has been implemented with eight membranes. The more detailed experimental programme was carried out with five membranes. Membranes with decreasing permeance-temperature ( Π −T ) curves indicate typically transport by Knudsen and viscous flows, i.e. meso and macrodefects. For instance, the AM-3 membrane number 3 exhibited such behaviour with pure H2, He, N2, and CO2. The Knudsen contribution was confirmed by the linear trend found between permeances and the inverse of the square root of the molecular weights. The viscous mechanism was also identified, since permeances were inversely proportional to gas viscosity or, attending to Chapman-Enskog based equations, directly proportional to 2 0.5 k d M (where k d is the kinetic diameter and M the molecular weight). A dissimilar permeation behaviour observed in this work involved AM-3 membrane number 5. The registered permeances at programmed temperature were approximately constant for N2, CO2 and O2, while for H2 increases significantly. In conjunction they evidence the simultaneous incidence of macro, meso and intercrystalline microdefects. The activated gaseous transport across micropores compensates the lowering impact of meso and macropores. In contrast to N2, CO2 and O2, the small hydrogen diameter makes it possible to permeate through the intracrystalline micropores, which superimposes an additional transfer mechanism responsible for such increasing. With respect to surface diffusion, the CO2/ZSM-5 system may be taken as a paradigmatic example. Once this zeolite adsorbs CO2 the permeances decrease with increasing ΔP , because the surface loading concentrations increase non-linearly tending to the saturation. The contrasting results obtained for nitrogen emphasizes even more this mechanism, since N2 does not adsorb and exhibited constant permeances. Globally, the calculated ideal selectivites ( α* ) ranged from ca. 1 to 4.2. This parameter was also utilized to discriminate the best membranes, since low values denote the non-selective viscous flow typical of macrodefects. For instance, H2/CO2 in AM-3 membrane No 3 presented α* = 3.6 − 4.2 for 40– 120ºC, while in AM-3 No 5 originated α* = 2.6 − 3.1. These results corroborated previous findings that AM-3-5 is better than AM-3-3. Some essays were carried out with membranes saturated with water to increase selectivity: the measurements clearly showed an initial enhancement followed by a consistent reduction with rising temperature, due to the removal of the water molecules responsible for the blockage of some pores. With respect to real selectivities of hydrogen-containing mixtures, more experiments must be performed, and the analytical quantification of hydrogen should be improved. Concerning modelling, new expressions for Maxwell-Stefan thermodynamic factors were derived for pure and multicomponent Nitta, Langmuir-Freundlich, and Toth isotherms, being tested with equilibrium and permeation data from literature. (It is worth noting that only pure and binary classical and dual site Langmuir equations are available). The validation procedure adopted was very stringent: i) the multicomponent isotherms were predicted from pure gas data; ii) the diffusion parameters of pure components were fitted to single permeation data; iii) then, the Maxwell-Stefan crossed diffusivities were estimated by the Vignes relation; finally, iv) the new equations were tested using these parameters, and were able to estimate successfully binary fluxes. In parallel to the main focus of gas permeation, a new Maxwell-Stefan based model has been also derived for ion-exchange in microporous materials. The model was validated with data for Hg2+ and Cd2+ removal from aqueous solution using ETS-4. Its predicting ability has been also analyzed, being possible to conclude it performs very successfully. In effect, good predictions were accomplished with parameters optimized from independent sets of data. Such performance may be attributed to the sound physical principles of Maxwell-Stefan theory.
autores
Patrícia Ferreira Lito
nossos autores
orientadores
Carlos Manuel Silva
Grupos