Você está na página 1de 111

Technische Universitt Mnchen

Department Chemie Lehrstuhl fr Technische Chemie II Prof. Johannes A. Lercher

Diffusion and Separation Characteristics of Binary Hydrocarbon-Mixtures in MFI

Diffusions- und Trenneigenschaften von Kohlenwasserstoffmischungen in MFI

Master Thesis ----------------------------------------------------------------Vorgelegt von Bsc. (TUM) Robin Kolvenbach

Betreuer: Prof. Dr. J. A. Lercher

Garching, April 2010

Wer die erhabene Weisheit der Mathematik tadelt, nhrt sich von Verwirrung Leonardo da Vinci

Acknowledgement

Acknowledgement There are a lot of people without whom this thesis had not been written and whom I owe my deepest gratitude. First of all Prof. Lercher for giving me the opportunity to work in his group and trusting me with the most challenging issue for my work. Then I would like to thank PD Dr. Jentys for guiding me through this thesis and also over the last years during my study at Technische Universitt Mnchen. I would like to address my deepest gratitude to Oliver Gobin who is an archetype for every scientist. I owe my gratitude to him for his guidance and excellent advice throughout this thesis. The intense discussions we had showed me how to perform scientific work in a proper way. I would like to thank the whole group for their support during this thesis. I have to address my special thanks to Xaver Hecht and the whole technical team without whom this thesis would have been impossible. Of course I would like to thank my wonderful parents and my family, brothers and my sister for supporting me my whole live.

Abstract

Abstract The diffusion and separation characteristic of binary hydrocarbon mixtures through zeolitic wafers were studied with a Wicke Kallenbach cell. Therefore, the transport of a benzene/p-xylene mixture and a p-xylene/m-xylene mixture was examined in ZSM5 powder pressed to wafers of 50, 100 and 150 mg. This report analyzes the transport mechanism and tries to falsify two models, the statistically dense membrane and the classical bed adsorber. Therefore several parameters were varied: different measurement and activation temperatures as well as compacting pressures were examined. It was found that the model of the classical bed adsorber suits the experimental results. These were mainly an equilibrium separation factor of 1 and the form of the transmission curves which were both independent on the change of experimental conditions. Considering the transport mechanism the diffusion through the mesopores was found to be the rate determining step once the system was under stationary conditions. The time until stationary conditions were reached as well as the transmission time were clearly influenced by the capacity of the adsorber. The capacity was determined by the temperature during the experiment and activation as well as the weight of the wafer. The steady state diffusion coefficients were 10-6 m2/g obtained by the Wicke Kallenbach measurements, which are much faster than the diffusion coefficients of 10-15 m2/g of the parent MFI as measured by pressure modulation frequency response experiments and uptake rate measurements. As a consequence the transport mechanism in the Wicke Kallenbach wafer can be assumed to be Knudsen diffusion through the mesopores. The pore size of the mesopores was confirmed either by calculations from the Wicke Kallenbach experiments applying the theory of Knudsen diffusion or nitrogen physisorption isotherms using DFT analysis. A pore size of 6 to 15 nm was found in which the transport through the 6 nm pores seem to be the determining step of the rate of diffusion.

Zusammenfassung

Zusammenfassung Die Diffusion und die Trennung einer Mischung bestehend aus zwei Kohlenwasserstoffen wurde in einer Wicke Kallenbach Zelle untersucht. Hierbei wurden Gemische aus Benzol/p-Xylol und m-Xylon/p-Xylol an einem Wafer aus ZSM5 untersucht. Es wurden Pellets zu 50, 100, 150 mg durch Pressen des Pulvers hergestellt. Dieser Bericht analysiert die Transportmechanismen innerhalb des Wafers und falsifiziert zwei Modelle, zum einen die Idee einer statistisch dichten Membran und den klassischen Bettadsorber. Hierbei wurden verschiedene Parameter variiert, unter anderem die Experiment- und Aktivierungstemperatur als auch den Pressdruck whrend der Herstellung des Pellets. Es wurde festgestellt, dass das Modell des Bettadsorbers in der Lage ist, die experimentellen Ergebnisse hinreichend gut zu beschreiben. Diese waren vornehmlich ein Trennfaktor von 1 und die Form der Durchbruchskurve, welche charakteristisch fr Adsorber ist. Der Transport durch den Wafer unter stationren Bedingungen wurde durch die Diffusion in den Mesoporen determiniert. Hierbei hngt die Zeit bis zum Erreichen des Gleichgewichts genauso wie die Durchbruchzeit von der Kapazitt des Adsorbers ab. Diese wiederum wurde beeinflusst von der Experiment- und Aktivierungstemperatur als auch von der Dicke des Wafers. Es wurden Diffusionskoeffizienten von 10-6 m2/g in den Wicke Kallenbach Messungen gefunden welche im Vergleich zu 10-15 m2/g fr das

korrespondierende Pulver deutlich grer waren. Die Diffusionskoeffizienten des Pulvers wurden mittels pressure modulation Frequenz Antwort Methode und Uptake Rate Messungen bestimmt. Anhand der gemessenen

Diffusionskoeffizienten kann angenommen werden, dass die Diffusion in den Poren einer Knudsen Diffusion in den Mesoporen entspricht. Die Porengre wurde daraufhin durch zwei Methoden bestimmt. Zum einen durch eine DFT Analyse einer Stickstoffphysisorptionsisotherme und zum anderen durch Berechnung aus den experimentellen Wicke Kallenbach Daten unter Benutzung der Theorie der Knudsen Diffusion. Es wurde eine Porengre von 6 bis 15 nm gefunden wobei der Transport in den 6 nm Poren den gesamten Diffusionstransport determiniert.

Abbreviations

Abbreviations A Aper poke Aper successful poke amplitude of FR experiment surface area needed statiscally for one poke [m2] surface area needed statiscally for a poke leading to sorption of the adsorbent into the pore system [m2] AS Aw a surface species Cross section of the wafer [m2] parameter corresponding to the amount of adsorbate in the pores cB and on the surface cA B BP b amount of gas present in the sheet adsorbate inside the channel system parameter corresponding to the amount of adsorbate in the pores cB and on the surface cA C Ct C c c0 co c cA ce c(1,2), feed normalized gas phase concentration time dependent concentration [mol/l] concentration at infinite time [mol/l] concentration of diffusing component [mol/m3] initial adsorbed phase concentration [mol/m3] equilibrium adsorbed phase concentration [mol/m3] steady state adsorbed phase concentration [mol/m3] concentration in the adsorbed phase [mol/m3] equilibrium concentration inside the particle [mol/m3] adsorbate concentration in the feed [mol/m3]

Abbreviations

c(1,2), measurement

adsorbate concentration in the permeate during the measurement [mol/m3]

cp cs D(c) D0 Deff DK,eff DLS dp EA,D F FHe Fmolecules

effective heat capacity of the adsorbent sample [J/kgK] surface concentration [mol/m2] concentration dependent diffusion coefficient pre exponential factor of the Arrhenius ansatz diffusion coefficient in the pore system [m2/s] effective Knudsen diffusion coefficient [m2/s] dynamic light scattering pore diameter [m] activation energy of diffusion [kJ/mol] volumetric flow rate [m3/s] volumetric flow rate of the sweep gas [m3/s] Number of molecules transmitting through wafer per second [molecules/s]

H h J K

heat of adsorption [J] external heat transfer coefficient [W/m2K] diffusive molar flow rate [mol/m2s] parameter proportional the slope of the adsorption isotherm

KH k ka L

Henry constant Boltzmann constant [J/mol] rate constant of Adsorption [1/s] characteristic length of diffusion [m]

Abbreviations

mean value of characteristic length of diffusion [m] M MFC m(t) m0 m N molar weight [kg/mol] mass flow controller mass dependent on the time starting mass end mass statiscal number of successful pokes during the

transmission of the adsorbate through the wafer NA n P P0 Pe Avogadro constant [1/mol] number of parallel transport processes partial pressure [Pa] standard pressure (1 bar) partial pressure of the diffusion component in the gas phase at equilibrium condition [Pa] p pB Q Q0 pressure amplitude of FR experiment pressure amplitude of blak of FR experiment dimensionless concentration in the adsorbed phase equilibrium value of the average dimensionless adsorbed phase concentration R r ra S SAd gas constant [J/molK] radius of the Wicke Kallenbach pellet [m] rate of adsorption [mol/s] separation factor adsorption cross section [m2/mol]

Abbreviations

Sw SBET Sext SEM T T0 TI TEM t tHasley

cross section of the adsorbent bed BET surface area [m2/g] external surface area [m2/g] Scanning electron microscopy temperature [K] experimental temperature (FR) [K] initial and final steady state temperature [K] transmission electron microscopy time [s] thickness of a monolayer of physisorbed nitrogen

according to the theory of Halsey [nm] UHV VFR Ve,FR Vma Vme Vmi vmicro,max Vmol Wsat X XRD x ultra high vacuum volume of the FR system [m3] equilibrium volume (FR) [m3] macropore volume [cm3/g] mesopore volume [cm3/g] micropore volume [cm3/g] maximum micropore volume [cm3/g] molar volume [m3/mol] maximum amount of adsorbed gas = W0 [mol/m2] adsorbate in the gas phase X-Ray diffraction space coordinate

Abbreviations

xAdsorbate

molar ratio of the adsorbate either in the feed or the permeate

y Zw

length coordinate [m] Number of hits of a molecule with the wall [1/m2s]

Abbreviations

Greek Symbols the dimensionless parameter describing the heat transport the dimensionless parameter describing the heat transport characteristic function of FR experiment mean value of the characteristic function of FR experiments p B B F porosity factor phase lag of the concentration inside the sheet relative amplitude rate constant of the surface barrier dimensionless radial coordinate phase of the FR experiment phase of the blank FR experiment effective density of the adsorbent sample [kg/m3] density of adsorbent bed surface coverage dimensionless time tortuosity factor amplitude of the perturbation frequency of perturbation [1/s]

Table of Contents

Table of Contents

Acknowledgement ............................................................................................. A Abstract ............................................................................................................. B Zusammenfassung............................................................................................. C Abbreviations ..................................................................................................... D Greek Symbols.....................................................................................................I 1. 2. Introduction................................................................................................. 1 Theoretical Section ..................................................................................... 4 2.1 Zeolites ..................................................................................................... 4 2.2 Wicke Kallenbach Experiments ................................................................ 8 2.3. Frequency Response experiments ........................................................ 10 2.4. Uptake rate measurements ................................................................... 14 3. Experimental ............................................................................................. 19 3.1. Material .................................................................................................. 19 3.2. Wicke Kallenbach experiments ............................................................. 21 3.3.1 Sample preparation ......................................................................... 21 3.2.2 Experimental setup .......................................................................... 22 3.3. Frequency response experiments ......................................................... 25 3.3.1. Sample preparation ........................................................................ 25 3.3.2. Experimental Setup......................................................................... 26 3.4. Uptake rate measurements ................................................................... 27 3.4.1 Sample preparation ......................................................................... 27 3.4.2. Experimental Setup......................................................................... 27 4. Results ...................................................................................................... 29 4.1 Wicke-Kallenbach experiments .............................................................. 29

Table of Contents

4.2. Frequency Response experiments ........................................................ 41 4.3. Uptake Rate measurements .................................................................. 43 4.4. Nitrogen physisorption isotherms.......................................................... 47 5. 6. Discussion ................................................................................................ 49 Conclusion ................................................................................................ 69

Literature .......................................................................................................... 71 Appendix

Introduction

1. Introduction The understanding of diffusion processes in porous media is crucial for the development of new and more effective catalysts and molecular sieves. Especially the diffusion in micropores became a very important field as this is often the rate controlling step in a complex reaction network. Moreover it is possible to increase the selectivity of a process by shape selective catalysts [1,2]. Most of the time these are zeolites or related materials because the pore structure can be tuned to favor the desired product. The difference in the diffusivity of two components due to the pore diameter of a porous substrate can also be used for separation by size exclusion. Sakai et al. showed the separation of p-xylene from a ternary mixture of the xylene isomers with the help of a MFI-type zeolite membrane [3]. Another approach to enhance the separation of hydrocarbons by zeolites was recently published by our group. The enhanced separation is caused by a surface modification with a silica layer of 3 nm thickness [4,5]. This over-layer leads to a variation of the sticking probability on the modified surface which speeds up selectively the sorption rate in the porous over-layer and has in combination with the intrinsic size exclusion of the MFI micropores the consequence of an enhancement of the sorption rate. To be able to determine the intracrystalline diffusivity many different methods were developed in the last decades among these uptake rate measurements [6,7,8,9] as for instance zero length column [10,11], different NMR techniques for example the widely applied pulsed field gradient method [12,13,14,15], frequency response experiments [16,17,18,19] and rapid scan IR-

spectroscopy[4,5,19]. Furthermore microscopic IR techniques offer the possibility to monitor the diffusion process in a sample [20,21]. Another widely used method is the measurement in a Wicke Kallenbach cell. This experiment offers the possibility to determine the diffusive flow through a micro-, meso- or macroporous wafer. Originally Wicke and Kallenbach invented the method for monitoring the diffusivity of CO2 in active coal [22]. Since then the method was used for several different types of experiments either in the classical form or modified to fit the requirements.

Introduction

Sun et al. applied the method to measure intracrystalline diffusivity of fast transport processes such as methane in Silicalite 1 [23]: they mounted large silicalite crystals (100x100x300 m) at the center of an aluminum disk where both sides of the crystals were exposed by using an epoxy resin. They preformed the measurement with methane, ethane, propane and butane. The experiments showed excellent results, which were in agreement with previously reported intracrystalline diffusivities determined by frequency response techniques. Many other authors used zeolite membranes grown on a support, most of the time alumina to investigate intrapore transport processes. Keizer et al. demonstrated for example the possibility to separate mixtures of molecules with small and large kinetic diameters [24]. N-hexane was chosen as small and 2,2-dimethylbutane as large molecule. Separation factors larger than 600 were obtained in a temperature range between 298 and 473 K. Tuchlenski et al. quantified the amount of surface diffusion of non adsorbing gases in porous glass with a pore radius of 3.5 nm by applying a modified Wicke-Kallenbach cell [25]. They obtained excellent results in describing the diffusion of a non-adsorbing gas through the membrane by applying the dusty gas model (DGM). Similar experiments were carried out by Arnst et al. [26]. They examined the transport of ternary mixtures made of hydrogen, helium, nitrogen or argon in mesoporous alumina wafers. They also tried to model the transport processe with the DGM and in addition with the mean pore transport mode model. Unfortunately they found out that it is impossible to model this process as the molecules influence each other in the gas during the transmission through the wafer. The aim of this report is the examination and understanding of transport phenomena of binary mixtures consisting of strong adsorbing substances during the transmission through a pressed wafer of ZSM5 in a Wicke Kallenbach cell. The experiments should also give more insight in the separation behavior of hydrocarbons of such wafers. They should answer the question if these wafers act like a membrane or analogue to a bed adsorber. The experimental part consists of Wicke Kallenbach experiments applying pressed wafers of ZSM5

Introduction

and binary mixtures of benzene and p-xylene as well as p-xylene and m-xylene as adsorbates to see whether it is possible to separate these compounds. To investigate the differences between the powder and the pressed wafer, pressure modulation frequency response and uptake rate measurements were performed. These techniques allow the determination of the time constant of the diffusion and the examination of the transport mechanism by fitting the experimental data to theoretical models. The combination of all these different techniques should answer the question which diffusion mechanism occurs in a pressed Wicke Kallenbach wafer and shows if it is possible to separate hydrocarbons with a pressed zeolitic wafer.

Theoretical Section

2. Theoretical Section 2.1 Zeolites Zeolites are microporous tecto-silicates which are nowadays widely used in industrial processes [27]. Because every zeolite is a potential catalyst, adsorber or cation exchanger they are applied in several processes [28] e.g. methanol to gasoline [29] or selective catalytic reduction of nitric oxide by ammonia [30]. One of the most important operation areas is the petrochemical industry, in particular fluid catalytic cracking [31], isomerization [32,33] and alkylation [34,35]. The name Zeolite was used for the first time in 1756 by the mineralogist A.F. Cronsted who observed bubbles while heating the mineral Stilbite. The origin of the name zeolite are the two greek words zeon(to boil) and lithos(stone) [36]. Nowadays over 180 different zeolite structures are known among them 40 natural and 140 synthetic topologies. They are identified by a three letter code defined by the IUPAC (International Union of Pure and Applied Chemistry) [37] and the IZA (International Zeolite Association) [38] e.g. FAU = Faujasite.

Figure 1:Primary and secondary building units (PBU, SBU) of zeolitic frameworks.

Zeolites consist of SiO44- and AlO45- tetrahedrons as primary constitution which built up a three dimensional secondary structure by linking through the oxygen atoms (Figure 1). These secondary building units (SBU) can be separated in 20

Theoretical Section

different unique polyhedrons. The combination of the primary and secondary units leads to huge and complex structural diversity with three dimensional channel- and pore - systems, cages and super cages [38]. Due to the regular structure of the material the pore size distribution is very sharp. Figure 2 shows the pore size distribution of several different adsorbents in comparism. The pores have a diameter below 1.3 nm which means they are in the range of kinetic diameters of molecules. This property gives the opportunity to perform size exclusive as well as shape selective reactions. Additionally selective adsorption of a molecule from a reaction mixture becomes possible because molecules larger than the pore openings cannot enter in contrast to smaller ones.

Figure 2: Pore size distribution of zeolites in comparism to the standard adsorbents silica gel and active cole [39].

The isomorphous exchange of Si4+ by Al3+ opens out into a negative charge which has to be compensated either by an inorganic or organic cation consequently an alkali, earth alkali cation, quaternary ammonium ion or proton [40] (Figure 3). The general elemental formal for zeolites is Mx/m[(AlO2)x(SiO2)y]wH2O where m is the charge of the cation and w the number of water atoms is the cage.

Theoretical Section

Thereby the Si/Al ratio is defined by the ratio of y/x and has to be greater than 1 due to the Lwenstein rule [41]. This rule indicates that Al-O-Al linkages are forbidden in zeolite frameworks. Si/Al ratios from 1 to infinity which describes pure silica zeolites are common.

Figure 3: Structural formulas showing the effect of isomorphic exchange of silicon by aluminum and the compensation by either an inorganic or organic cation M+ (a) or proton (b) [42].

The aluminum content is one of the characteristics determining the reactivity of a zeolite by the density of negative charge in the framework. This has enormous consequence for the density and strength of the Brnstedt acid sites, the ion exchange capability, the thermal stability and hydrophilic properties [43]. This dependence provides the opportunity to tune the reaction properties of a zeolite by altering the degree of aluminum in the framework. Summarizing the physical and chemical properties [44] of zeolites it may be adhered that they have a high surface area, a molecular dimension of the pores, a high adsorption capacity, high ion-exchange capability, high thermal and hydrothermal stability, the possibility of modulation the electric properties of the active sites and the possibility for pre-activating the molecules in the pores by strong electric fields and molecular confinement. Although the microporousity is one of the main advantages of zeolites it is a disadvantage at the same time. The reason for this is the slow diffusion in the micropores which leads to limitations of the rate of chemical reactions. To solve this problem intense activities are ongoing over the last years.

Theoretical Section

There are mainly two ways to overcome this problem. One is to make nano sized particles [45] and the other one is to synthesize mesoporous materials like MCM-41 [46]. The synthesis of zeolites is usually performed under hydrothermal conditions using reactive gels in alkaline media at 80 200 C. These reactive gels are produced from hydrated aluminum and silicon species e.g. hydroxides. Metalpowder or salts dissolved in sodium hydroxide solution usually serve as aluminum source. The silicon source which is normally the colloidal dioxide is combined with the aluminum containing solutions under hydrothermal conditions [47].

Figure 4: Two different method of ZSM5 synthesis [ 48].

A structure directing agent (SDA) is added to the reaction mixture to control zeolite growth in means of pore size and network dimensions. Usually tertiary amines serve as SDA e.g. a tertiary-propyl-ammonium salts in the case of ZSM5. The pH of the reaction mixture is adjusted between 9 and 13. In this high temperature and pressure conditions a reactive gel is formed. Inside this gel zeolitic nuclei occurs which then starts the crystal growing. Depending on the synthesis method zeolite crystals or polycrystalline zeolite material is formed. These two different methods are described in Figure 4.

Theoretical Section

During the last 50 years many other preparation methods were developed. They were extensively reviewed before [49,50,51,52] and will not be discussed here. 2.2 Wicke Kallenbach Experiments This method for investigating the diffusivity in porous media was developed by Wicke and Kallenbach in 1941 [22]. In principle two countercurrent gas flows graze the sample wafer in a 90 angle. One of the flows is loaded with probe substances and the other one contains of pure sweep gas (Figure 5). The sample is placed into the cell as a wafer which is either produced by compacting of powder or as a grown membrane. The diffusion through the wafer is driven by the concentration difference between both flows. The pressure on both sides of the cell has to be the same to avoid pressure driven transport through the sample. Therefore, the volumetric flow rate of the loaded and the unloaded stream has to be the same.

Figure 5: Flow scheme of the Wicke Kallenbach cell showing the fluxes going in and out of the cell and indication where the diffusive transport takes place.

If the experiment is performed with two or more substances it is possible to determine the ability of the sample to separate the adsorbates by comparing the feed concentration to the concentration in the permeate. Hence it is possible to calculate a separation factor by: (2.2.1) If a separation is observed during the experiment the separation factor has to differ from 1. Furthermore it is possible to calculate the diffusion coefficient of

Theoretical Section

the adsorbates inside the pellet from the steady state molar ratios of these experiments by (2.2.2) (2.2.2) In this equation L is the characteristic length of diffusion, r is the radius of the pellet, FHe is the volumetric flow rate of the sweep gas and x Adsorbate is the molar ratio in the feed and in the permeate at steady state conditions, respectively. In principle three different kinds of diffusion have to be considered in porous media. In the case of molecular diffusion the mean free path of a molecule is shorter than the pore diameter. As a result the diffusion process can be described analogue to the free vacuity by the first fickian law by substituting the diffusion coefficient by an effective diffusion coefficient (Deff). The latter consists of an additional porosity factor (p) which describes the ratio of pore openings compared to the total surface area as well as a tortuosity-factor (F) specifying the differences of the pore geometry from the ideally assumed cylinder [53]. (2.2.3) If the pore size is small or the gas pressure is low the mean free path is longer than the pore diameter. Consequently the molecules hit the pore walls much more often than other molecules. In this case the so called Knudsen diffusion is observed. The conditions for this kind of diffusion are summarized in the following table [53]:
Table 1: Depence of the occurring of Knudsen diffusion on the partial pressure of the adsorbate and the pore size

dp [nm] p [105 Pa]

<1000 0.1

<100 1

<1 10

<2 50

The diffusive flow in the Knudsen theory can mathematically be expressed by equation (2.2.4) [53]. (2.2.4)

Theoretical Section

10

In here dp is the pore diameter, R is the gas constant, T is the temperature, M is the molar mass and c the concentration of the diffusing component. Analogue to the fickian law the effective diffusion coefficient can be described by equation (2.2.5) [53]. (2.2.5) The third kind diffusive transport is the diffusion in pores with a diameter below 1 nm ocurring in zeolites. In this case the pore radius is in the range of the minimum kinetic diameter of the diffusing molecule. Therefore the diffusion coefficients found in zeolites are much lower than the molecular or Knudsen coefficients. The mathematical description is the first fickian law for a system in a steady state as shown in (2.2.3). For non steady state situations the second fickian law is used and calculated separately for each specific case. The solutions for all of these constraints can be found in Crank et al. [54]. The dependence of the effective diffusion coefficient on the pore diameter is shown in Figure 6.

Figure 6: Dependence of the effective diffusion coefficient on the pore size of porous adsorbents.

2.3. Frequency Response experiments The frequency response method offers the opportunity to determine transport pathways and their kinetics in a batch adsorber [60]. In this method the volume of the reaction vessel is altered periodically by 1 % while tracking the pressure simultaneously. By changing the volume the adsorption/desorption

Theoretical Section

11

equilibrium is disturbed and tries to equilibrate again. This equilibration can be followed by tracking the pressure (Figure 7).

Figure 7: Simulated data of a frequency response experiment with a square wave volume perturbation showing the signal of the magnet which describes the volume (blue) and the corresponding pressure trend (red) against the time.

The experimental data is then compared to the solution of theoretical models in order to find the best match. By matching the experimental data to the models the mechanism of the transport process and its kinetic parameters can be obtained. The model equations used for the curve fitting can be found by solving the mass balance in a closed volume as described previously by Yasuda et al. [16]. They obtained two characteristic functions characterizing the system: (2.3.1) (2.3.2) (2.3.3) Where p is the pressure amplitude of the experiment, pB is the corresponding pressure amplitude in a blank experiment, is the phase of the experiment, B is the phase of the blank measurement and is the characteristic function. Herein n is the number of parallel transport processes. Furthermore K is a parameter proportional the slope of the adsorption isotherm, the temperature (T), the volume of the system (VFR) and the gas constant (R). Blank experiments are needed to exclude the non-idealities of the apparatus itself from the outcome of the experiment [60]. The characteristic functions () are solutions of the second fickian law using the right boundary conditions. In this work two different transport mechanisms are examined: the diffusion in an

Theoretical Section

12

infinite sheet and an infinite sheet with surface resistance. The mass balance of an infinite sheet without surface barrier is defined as [17]: (2.3.4) where P is the pressure in the system, V is volume, R is the gas constant, T 0 is the experimental temperature and B is the amount of gas present in the sheet. Assuming the volume is changed sinusoidal we can formulate the volume as: (2.3.5) herein Ve,FR is the equilibrium volume, is the amplitude of the perturbation, is the frequency of perturbation and t is the time. The vapor pressure of the diffusing substance and the concentration in the pores can generally be described by: (2.3.6) (2.3.7) where Pe is the partial pressure of the diffusion component in the gas phase at equilibrium condition, ce is the equilibrium concentration inside the particle, p (p/2) and are the relative amplitudes, and are the phase lags both depending on the frequency . The diffusion inside the pores is described by the second fickian law: (2.3.8) Herein c is the concentration in the pores, D(c) is the concentration dependent diffusion coefficient and x is the space coordinate. Since the perturbations are defined to be small ficks law eases to: (2.3.9) In this work we consider the case of one dimensional diffusion in sheet with plane boundaries at x=0 and x=L. The boundary conditions define the concentration at the boundaries to be proportional to the sinusoidal changing vapor pressure in the gas phase [55], [56]: (2.3.10) (2.3.11) As a consequence the integral of the amount of gas present inside the sheet, B, is given by: (2.3.12)

Theoretical Section

13

Where Be is the equilibrium amount if gas present inside the sheet. The amplitude A is expressed by [56]: (2.3.13) (2.3.14) where KH is Henrys law constant, p is the relative amplitude of the pressure, is the frequency, L is the length of the sheet and D is the diffusion coefficient. Furthermore the coefficient in equation (2.3.12) is defined by: (2.3.15) Substituting (2.3.12) into the material balance (2.3.4) leads to the two equations (2.3.1) and (2.3.2) where the in and out-of-phase functions are given by: (2.3.16) (2.3.17) The other mechanism we want consider is surface controlled diffusion. Therefore we have to add another equilibrium upstream to the one we considered before describing the surface adsorption. This leads to the following reaction network [57]: X AS B P where X is the adsorbate in the gas phase, A S is the surface species and BP is the adsorbate inside the channel system. The corresponding material balance is defined by: (2.3.18) The solution using the previously described terms for P, V and B result in these in- and out of phase functions [16]: (2.3.19) (2.3.20) where is the rate constant of the surface barrier and a and b are parameters corresponding to the amount of adsorbate in the pores cB and on the surface cA [60]:

Theoretical Section

14

(2.3.21) (2.3.22) Additionally the definition of the parameter K of the characteristic functions (2.3.1) and (2.3.2) changed. In the case of a surface barrier it is defined by [57]: (2.3.23) Because most of the zeolites do not have a uniform particle size a normal distribution is introduced which modifies the characteristic functions [60]: (2.3.24) Herein and describing the mean values of the characteristic function and

the thickness of the sheet, respectively. 2.4. Uptake rate measurements To determine the kinetics of adsorption processes in porous materials a step change of the adsorbate pressure is performed and the change in mass of the sample is tracked. The altering of the pressure surroundings causes a shift in the surface concentration of the adsorbate species which can be described by the diffusional time constant (D/r2) [7]. The time constant and thereby the diffusivity can be obtained by matching the experimental data to an appropriate model describing the diffusion process. The method can be used for either macro-, meso- or microporous diffusion. A general mathematical model is applied developed by Ruthven et al. [58] for systems controlled by intracrystalline diffusion. It has the following

assumptions: 1) 2) The sample consists of a ensemble of spherical particles Intracrystalline diffusion is the only significant resistance to mass transfer. Therefore the sorbate concentration at the surface of the particles is always in equilibrium with the adsorbate in the surrounding fluid phase. 3) 4) The diffusivity is constant and the equilibrium relationships are linear. The experiment is performed under isothermal conditions.

Theoretical Section

15

According to these approximations the system can be described by the following differential equations. Based on the first fickian law the first one defines the behavior of the dimensionless adsorbed phase concentration (Q) dependent on the dimensionless time (): (2.4.1) (2.4.2) (2.4.3) in here is the dimensionless radial coordinate, cA is the concentration in the adsorbed phase, c0 is the initial adsorbed phase concentration and c is the steady state adsorbed phase concentration. The second and third equation describe the average of the adsorbed phase concentration over the particle (2.4.5) and its time dependence (2.4.6): (2.4.5) (2.4.6) where Q is the dimensionless concentration in the adsorbed phase, is the dimensionless radial coordinate, H is the heat of adsorption, is the dimensionless time as defined in (2.4.3), is the effective density of the adsorbent sample, cp is the effective heat capacity of the adsorbent sample, c0 is the initial adsorbed phase concentration, c is the final steady state adsorbed phase concentration, T is the temperature, h is the external heat transfer coefficient, TI is the initial and final steady state temperature, D is the Diffusion coefficient and L is the radius of the particles

(= characteristic length of diffusion). Obviously the heat transfer has to be taken into account even if it is disregarded later due to the isothermal conditions. As a result the concentration of the adsorbed phase at the particle surface can be defined by (2.4.7): (2.4.7) where cs is the surface concentration and co the equilibrium adsorbed phase concentration. This set of equations is solved with the initial boundary conditions (2.4.8) and (2.4.9):

Theoretical Section

16

(2.4.8) (2.4.9) in here the first initial boundary condition indicates that the starting adsorbate of the experiment is zero. The second one defines that the concentration of the adsorbed phase does not change in the center of the particles. The solution taking the boundary conditions into account can be obtained by Laplace transformation and is shown in (2.4.10) and (2.4.11): (2.4.10)

(2.4.11) where qn is given by the roots of: (2.4.12) thereby Q0 is defines the equilibrium value of the average dimensionless adsorbed phase concentration, and are the dimensionless parameters describing the heat transport and are formulated by (2.4.13) and (2.4.14) (2.4.13) (2.4.14) Considering the assumption of isothermic conditions ( and 0) the solution for the average dimensionless adsorbate concentration reduces to (2.4.15): (2.4.15) This equation is then used to perform curve fitting with the fit parameter . By knowing the length of diffusion the diffusion coefficient can be obtained.

2.5. Analysis of physisorption isotherms The mostly applied method to analyze nitrogen physisorption isotherms was developed by Brunauer, Emmett and Teller in 1938 [59]. It offers the possibility to determine the surface area of a solid sample. In principle it is an extension of the classical Langmuir isotherm, which is only valid for monolayer adsorption, to multilayer adsorption.

Theoretical Section

17

The following assumption corresponds to this theory: 1) gas molecules adsorb in an infinite number of layers 2) each layer is treated independently which means that it has no interaction with other layers 3) each layer can be treated with the theory of Langmuir adsorption This leads to the BET equation: (2.5.1) with VAd equals the total amount of adsorbate adsorbed on the sample, P is the partial pressure of the adsorbate, P0 is the standard pressure and Vm is the volume of a monolayer. 2.5.1 is the linearized form of this equation. Based on this the monolayer coverage can be obtained from the slope and the intercept of the resulting graph shown in 2.5.2: (2.5.2) Furthermore it is possible to determine the surface area (2.5.3) if the adsorption cross section of the adsorbing molecule is known. (2.5.3) In here Vm is the volume of a monolayer, Vmol is the molar volume of the adsorbing gas, NA is the Avogadro constant, SAd is the adsorption cross section of the adsorbing molecule and M is the molecular weight of the adsorbing molecule. Additionally it is possible to analyze the isotherm concerning the micro- and mesopore volume. A the possible way is the t-plot method: the experimentally obtained isotherm is compared to a standard isotherm by plotting them against each other at the same P/P0 points. The standard isotherm is thereby represented by the thickness of adsorbed layers. In this thesis the method of Halsey [70] is used to produce the standard isotherm. The corresponding equation is: (2.5.4)

Theoretical Section

18

The obtained values are then plotted against the experimentally observed data. Linear regions of the resulting plot correspond to regions where the standard and the experimental isotherm have the same physical properties. These regions can be described mathematically to determine the micro- and mesopore volume. The first linear region with a positive slope is defined as the point closely after the micropores are completely filled and before the mesopore condensation begins. It can be described by: (2.5.5) In here vmicro,max defines the volume of the micropores, kt is a coefficient depending on the layer thickness, the units of vads and Sext. For standard units k is defined by 1/(4.3532xtHasley) with tHasley representing the thickness of a monolayer in the case of nitrogen 0.354 nm. According to equation 2.5.5 the intercept of the first linear region defines the micropore volume whereas the external surface area can be obtained from the slope of the graph. The second linear region with a positive slope describes the region shortly after the mesopore condensation. Consequently the intercept of the second linear region identifies the addition of the micro- and the mesopores. The last step is to determine the mesopore volume. To do so the value of the combination micro- and mesopore volume is subtracted by the previously obtained micropore volume.

Experimental

19

3. Experimental 3.1. Material Polycrystalline H-ZSM5 samples with a particle size < 100 nm and a Si/Al ration of 45 [60] were provided by Sdchemie AG. The nitrogen physisorption isotherms were obtained using a PMI automated sorptometer at liquid nitrogen temperature (77 K) after outgassing under vacuum for at least 6 hours. The specific and the external surface area, the macro-, meso- and micropore volume as well as the total pore volume (Table 2) were determined by nitrogen physisorption (Figure 8) and applying the Brunauer-Emmett-Teller (BET) theory in the range from 0.03 to 0.10. Between 0.1 and 0.3 mbar the plot was not linear with a negative intercept (BET C constant).
200 180 160
150

(a)

Amount Adsorbed (cm STP g )

Amount Adsorbed (cm STP g )

125

Vmi + Vme (b)

-1

140 120 100 80 60 40 20 0 1E-5

(a)

(b)

-1

100

75

Vmi
50

1E-4

1E-3

0.01

0.1

0.0

0.5

1.0

1.5

2.0

Relative Pressure (-)

Figure 8: Nitrogen physisorption isotherms of the parent H-ZSM5 material (a) [5].

Figure 9: s plot of the parent (a) material. Two linear regions can be found at s=0.5 corresponding to the micropore volume and at s= 1.2 belonging to the sum of micro- and mesopore volume [4].

SBET [m2/g] Parent 423

Vmi [cm3/g] 0.12

Vme [cm3/g] 0.04

Vma [cm3/g] 0.22

Sext [m2/g] 65

Vtot [cm3/g] 0.38

Table 2: [60,5]: Results of the nitrogen physisorption experiment for the parent ZSM5 sample. SBET is the total surface area determined by applying the BET theory, S ext is the external surface area and Vmi, Vme, Vma and Vtot are the micro-, meso and micropore volume obtained by using the s comparative plot

Additionally the Langmuir surface area was determined by performing a Langmuir plot in the relative pressure range up to 0.10 mbar. The macro-,

Experimental meso- and micropore volumes were obtained by using a s comparative plot (Figure 9) [61] with nonporous hydroxylated silica [62] as reference adsorbent. The particle size distribution was determined by dynamic light scattering (DLS) and scanning electron microscopy (SEM). DLS measurements were performed on a Malvern Zetasizer Nano ZS. Prior to the measurement the sample was dispersed in water with help of an ultrasonic bath where it was kept for several hours. The measurements were repeated at different concentrations several times. SEM images were produced on a REM JEOL5900 LV microscope operating at 25 kV with a resolution of 5 nm and a nominal magnification of 3.0 x 106. The samples were used without further purification. As shown in Figure 10 the particles contain of larger agglomerates of smaller particles with a diameter of 0.36 0.17 m.

20

Figure 10: Scanning electron microscopy of the H-ZSM5 parent material [60].

The size of the primary particles was obtained by transmission electron microscopy (TEM) and X Ray diffraction (XRD) methods. The TEM images were recorded on a JEOL-2011. The powdered samples were suspended in an ethanol solution and dried on a copper-carbon grid prior to the measurement. The powder diffraction patterns were measured on a Phillips X`pert Pro XRD instrument operating at the Cu-K radiation at 40 kV with a Nickel filter to remove the Cu-K. The samples were used without pretreatment. Powder patterns were produced using a spinner system with a inch divergence slit between two angles of 5 an 70. Step size was 0.017 with a scan time of 115s per step. Both methods show that the primary particle size is below 100 nm (Figure 11, Figure 12).

Experimental

21

Figure 11: TEM images of the parent H-ZSM5 sample[4].

Figure 12: XRD of the parent H-ZSM5 material showing the characteristic reflexes of ZSM5 [60].

The XRD of the parent material shows sharp reflexes at the characteristic positions of ZSM5 (Figure 12). Benzene and p-xylene (> 99.5 % GC standard) were supplied by Sigma Aldrich and were used without further purification. 3.2. Wicke Kallenbach experiments 3.3.1 Sample preparation All powdered samples were compacted into a stainless steel (b) ring at 20 kN for 2 minutes using the pressing instruments shown in Figure 13. The sample was then subsequently built into the Wicke Kallenbach cell and was flushed with helium for 30 minutes. Afterwards the sample was activated at 723 K for 1 hour in a helium flow of 50 ml/min for each of the two flows and an incremental heating rate of 10 K/min. The experiments were performed at a constant gas flow of 50 ml/min for each of the two countercurrent flows. Thereby the loading with benzene was

Experimental

22

performed with a volumetric flow rate of 2 ml/min and p-xylene was saturated with 18 ml/min. Assuming 100% saturation of the streams this leads to a ratio of 1.33/1 bezene/p-xylene in the feed.

Figure 13: CAD model of the pressing instruments used to produce wafers for the Wicke Kallenbach cell containing of a plunger (a), the ring inserted in the cell (b) and the stamp pad (c). The whole instrument consists of stainless steel.

To determine the temperature dependence of the separation, measurements at 130C (403.15 K), 100C (373.15 K) and 70C (343.15 K) were executed. In addition the sensitivity of the wafer thickness was examined in testing wafers of 150, 100 and 50 mg. Also the influence of the compacting pressure and the activation temperature was tested. 3.2.2 Experimental setup The cell shown in Figure 14 is made of 6 pieces, two endplates (a,f) and two rings, one holding the sample wafer (d) and the other holding the screws (c) and two 2 mm thick graphite sealings (b,e). The flow rate is controlled by 4 mass flow controllers (MFC) Bronkhorst EL-flow select, two of them with a range of 200 ml/min used for the sweep gas and the other two with a range of 20 ml/min leading to the saturators. These MFCs are computer controlled by Flow DDE V4.41 and Flow View V1.09 (Bronkhorst). The temperature of the cell is regulated via an Eurotherm 2404 PID controller. The loading of the gas stream with benzene and p - xylene is performed via two saturators, cooled down to 15C in a thermostat (Huber K12-cc-NR) with a 1:1 glycol/water mixture as cooling agent to avoid condensation of the adsorbate inside the system. Figure 15 shows a CAD Model of the saturators

Experimental

23

used. These saturators consist of two nearly identical parts which are connected.

Figure 14: CAD Model of the Wicke Kallenbach cell containing two end plates (a,f), two sealings made out of graphite (b,e), an aluminum ring holding the screws (c) and the inner stainless steel ring clamping the sample (d).

Figure 15: CAD Model of the saturator used to load the gas stream with adsorbates benzene and p Xylene.

Experimental

24

The base body comprises a half inch t - piece and tube which is welded up (a,b) and serves as a flask for the adsorbate. The gas flow enters the saturator at point (c). The following 2 way valve controls the direction of the flow. It can either enter the tube filled with adsorbate (a) or go through the bypass (e) back into the system. The bypass is needed to clean the system before and after the measurement. To save the system from floating with adsorbate in the case of low-pressure inside the system an empty second tube (b) is connected to the first flask. After passing this station the stream gets back into the system through another 2 way valve (d) which connects the bypass (b) and the saturator. An image of the total setup is shown in Figure 16.

Figure 16: Setup for the Wicke Kallenbach experiments containing the Wicke Kallenbach cell, the mass flow controllers (MFC) the saturators and the electronic for controlling the temperature and the flows.

The permeate stream is connected to a vacuum system via a heated capillary and differential pumped inlet system. The UHV is produced by a turbo-

Experimental

25

molecular pump TMH 071-003 (Pfeiffer Vacuum). This allows pressures down to 10-9 mbar. During the measurements the vacuum is 10-5 mbar. The permeate is analyzed by a mass-spectrometer (WR 13302, Hiden Analytical) in the MID mode controlled by MASSoft Professional (Hiden Analytical) computer software. 3.3. Frequency response experiments 3.3.1. Sample preparation Two different kinds of samples, specifically powdered and pressed samples, were used in this experiment. In the case of the powdered samples 35 mg were carefully dispersed on glass-wool at the bottom of a quartz tube to have isolated particles and thereby avoid bed effects. Furthermore two types of pressed wafers were studied: Wicke Kallenbachwafers of 100 mg and IR wafers 6 mg weight. Both were carefully broken into smaller pieces to fit in the quartz tube sample holder. Glass-wool was placed at the bottom of the sample holder to have the same conditions as in the experiments with powdered samples. In both cases the glass-tube was connected to a UHV setup, placed into an oven and pumped to 10-7 mbar. The samples were activated at 723K for 1 hour with a temperature ramp of 10 K/min at ultra high vacuum conditions to remove the adsorbed water. The adsorbate was added with a partial pressure of 0.31 mbar at a temperature of 373 K until the adsorption equilibrium was fully established. During the experiment the volume of the setup was changed periodically in a frequency range of 5 to 10-3 Hz and a square waved volume perturbation 1% in amplitude. The pressure response was fourier transformed by obtaining the amplitude and the phase lag. The resulting characteristic functions were fitted with the model equations described in 2.3. by a nonlinear parameter fitting using a CMA evolutional strategy in matlab [63]. The objective minimized to find the best consistency was the root mean square error normalized to the variance of the data. To be sure that the best parameters were found which is equal to the

Experimental

26

global minimum the calculation was performed three times with different starting parameters.

3.3.2. Experimental Setup An overview of the setup for the frequency response experiments is given in Figure 17. The setup is basically a UHV unit equipped with volume perturbation part which consists of two magnetically driven plates sealed with UHV bellows. The UHV is produced by a turbo-molecular pump TMH 071-003 (Pfeiffer Vacuum). This allows pressures down to 10-8 mbar. To enable the turbo pump to start a zeolite trap is build into the system to produce vacuum of about 10-2 mbar.

Figure 17: Image of the frequency response setup showing the modulation unit, the vacuum chamber, the sample cell and the dosing system.

The pressure is recorded online via a Baratron pressure transducer (MKS 16A11TCC). The system is controlled via HP VEE based software. The

Experimental

27

adsorbing substrate is added by a separately pumped dosing system with an all metal regulating valve UDV 040 (Pfeiffer Vacuum) to put the gas phase corresponding to the liquid adsorbate into the system. 3.4. Uptake rate measurements 3.4.1 Sample preparation A Wicke Kallenbach wafer of 100 mg was pressed as described in 2.1. and broken afterwards into small pieces of 6 to 10 mg. One of these pieces was subsequently filled into the quartz crucible which was then placed into the balance. The balance was connected to an UHV system and pumped 10-7 mbar. The sample was activated at 723 K in UHV conditions to remove the adsorbed water and cooled down to the experiment temperature of 373 K. The adsorbate was added in a pressure jump with partial pressures of 0.1, 1, 2 and 10 mbar and was kept constant during the whole experiment while the sample mass was monitored by a Seteram TG-DSC 111. The mass of the sample was normalized afterwards by: (3.4.1) where m0 is the mass at the beginning of the experiment and m is the mass at the end. This normalized data was then fitted with theoretical models described in 2.4. by nonlinear parameter fitting using a CMA evolutional strategy in matlab [63]. The procedure was identical to the one described in section 3.3. 3.4.2. Experimental Setup The main item of the setup is a Seteram TG-DSC 111 microbalance (Figure 18) equipped with quartz insertions and crucibles being heat insensitive. The balance is connected to a UHV unit with a turbo-pump TMH 071-003 (Pfeiffer Vacuum) which is able to produce a vacuum of 10-8 mbar. Additionally an oil pump is connected to the system to produce on the one hand the fore-vacuum of 10-3 mbar crucial for a turbo-pump system and on the other hand pumping the inlet system for the adsorbates.

Experimental

28

The inlet system contains of a inch connection to fit the round bottomed flask filled with adsorbate to the system. This fitting is linked to an all metal regulating valve UDV 040 (Pfeiffer Vacuum) connecting the inlet part to the vacuum system. The whole system is operated by a Seteram controller which is connected to a computer. A HP VEE based software is used to control the system and to track all relevant data which is the mass, the pressure and the heat signal.

Figure 18: System used for uptake rate measurements consisting of a microbalance, vacuum system with a mass spectrometer, a dosing system and a controller.

Results

29

4. Results 4.1 Wicke-Kallenbach experiments The diffusion and separation of benzene and p-xylene in H-ZSM5 have been studied using the Wicke-Kallenbach method. In general the shape of the benzene concentration curve was exponential with a clear overshooting. The concentration gradient of p-xylene showed also an exponential form but without the over-swinging behavior. Thereby the maximum of the overswinging of the benzene curve occured at the time the p-xylene uptake started. This behavior was independent on the temperature during experiment and activation, the thickness of the wafer and the compacting pressure. The ratio between the equilibrium concentration of benzene and p-xylene had a constant value around 1.43 for every performed experiment and hence similar to the feed. This equals a separation factor of 1 valid for each experiment. The uptake time of benzene and p-xylene differed by more than an hour. An example of the concentration graph is shown in Figure 19 for 100 C, a wafer of 100 mg and 50 ml/min volume flow. In here the p-xylene concentration was multiplied by the ratio of benzene and p-xylene (1.43).
10
-4

10

-5

10
c [mol/l]

-6

Benzene p-Xylene

10

-7

10

-8

10

-9

10

-10

10

20

30

40

time [min]

50

60

70

80

90

100

Figure 19: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 100C, 100mg wafer weight and 50 ml/min volume flow.

Results

30

Because the uptake time for benzene and p-xylene differed as shown in Figure 19 the separation factor was clearly depending on the time. Figure 20 shows this time dependence. It points out that the separation factor was 1 in the beginning, started increasing very fast after 5 minutes and reached a sharp maximum of 104 after 20 minutes corresponding to the maximum of the benzene breakthrough curve. After 100 minutes the graph had a constant value of 1 due to the equilibrium concentration of benzene and p-xylene.
10
6

10
Seperation Factor [-]

10

10

10

-2

50

100

150

time [min]

200

250

300

350

Figure 20: Time dependence of the separation factor of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 at 100C, 100mg wafer weight, a mixture of benzene/p-xylene and 50 ml/min volume flow.

To study whether the uptake and thus the separation factor was influenced by the temperature, experiments at 130, 100 and 70C were carried out. Figure 21 displays the trend of the concentration of benzene at these three temperatures. It points out that the time to the breakthrough of benzene varied between 7 minutes for 130 C and 18 minutes for 70 C. It shows as well that the benzene uptake curve kept its exponential form thus the over swinging got stronger with increasing temperature. The breakthrough time of p-xylene was also temperature dependent changing from 15 minutes for 130C to 37 minutes in the case of 70C as shown in Figure 22.

Results

31

10

-5

c [mol/l]

130 C 100 C 70 C 10
-6

10

-7

10

20

30

40

time [min]

50

60

70

80

90

Figure 21:Concnetration vs. time diagram of benzene breakthrough curves at 130C (black), 100C (red) and 70C (blue) in a Wicke Kallenbach experiment with a pressed H-ZSM5-wafer of 100mg, a mixture of benzene/p-xylene and a volume flow of 50ml/min.
10
-4

10

-5

10
c [mol/l]

-6

130 C 100 C 70 C

10

-7

10

-8

10

-9

50

time [min]

100

150

Figure 22: Concentration vs. time diagram of p-xylene breakthrough curves at 130C (black), 100C (red) and 70C (blue) in a Wicke Kallenbach experiment with a pressed H-ZSM5-wafer of 100mg, a mixture of benzene/p-xylene and a volume flow of 50ml/min.

The breakthrough time as well as the time to the equilibrium concentration decreased with increasing temperature. As in the case of benzene the overall form of the uptake curve stayed exponential but the graph got steeper the higher the temperature was. Similar time dependence was found in the separation factors. Figure 23 shows that the maximum shifted from 15 minutes towards 37 minutes with decreasing

Results

32

temperature. But independent of the temperature the constant value 1 was reached.

10

130C 100C 70C

Seperation Factor [-]

10

10

10

-2

20

40

time [min]

60

80

100

120

Figure 23: Seperation factor vs. time diagram at different temperatures of 130 (blue), 100 (red) and 70C (black) in a Wicke Kallenbach experiment with a pressed H-ZSM5-wafer, a mixture of benzene/p-xylene, a weight of 100 mg and a volume flow of 50ml/min.

To examine whether the amount of adsorbent or the thickness of the wafer had an influence on the uptake or the separation factor pellets with weights of 50, 100 and 150 mg were used at constant temperature of 100C keeping the other experimental conditions untouched. The obtained graph for benzene is shown in Figure 24. It can be seen that the transmission time got longer the more material was used for the wafer and hence with increasing thickness of the wafer. Considering the form of the graphs it can be seen that curves were stretched increasing the thickness of the wafer keeping the overall exponential form with an over-swinging before reaching the equilibrium concentration. Consequently the chart got less steep in exponential region and had a longer over-swinging period the higher the weight of the wafer was. Furthermore the equilibrium concentration got smaller with increasing thickness of the wafer.

Results

33

10

-5

50mg 100mg 150 mg

c [mol/l]

10

-6

10

-7

10

-8

50

100

150

time [min]

200

250

300

350

Figure 24: Concentration vs. time diagram of benzene breakthrough curves with a wafer weight of 50mg (black), 100mg (red) and 150mg (blue) in a Wicke Kallenbach experiment with a pressed H-ZSM5-wafer, a mixture of benzene/p-xylene, a temperature of 100C and a volume flow of 50ml/min

10

-5

50 mg 100 mg 150 mg

c [mol/l]

10

-6

10

-7

10

-8

50

100

150

time [min]

200

250

300

350

Figure 25: Concentration vs. time diagram of p-xylene breakthrough curves with a wafer weight of 50mg (black), 100mg (red) and 150mg (blue) in a Wicke Kallenbach experiment with a pressed H-ZSM5-wafer, a mixture of benzene/p-xylene, a temperature of 100C and a volume flow of 50ml/min

Similar to the benzene curves the graph of p-xylene (Figure 25) got less steep with increasing weight of the wafer. As in all other measurements before an over-swinging is not observed. As in the case of benzene the uptake time got longer with more material used for making the wafer. Looking at the form of the breakthrough curves it is

Results

34

obviously exponential for all of the three weights but was also stretched for thicker wafers. The equilibrium concentration was smaller by the same percentage than in the case of benzene. As a result the separation factor stayed 1 independent on the thickness of the wafer. Remarkable is the time dependence of the separation factors: similar to the uptake curves of benzene and p-xylene the maximum separation factor shifted to longer time periods with increasing thickness of the wafer (Figure 26).

10

150mg 100mg 50mg 10


Seperation Factor [-]
2

10

10

-2

50

time [min]

100

150

Figure 26: Seperation factor vs. time diagram of wafers with weights of 150 (blue), 100 (red) and 50 mg (black) in a Wicke Kallenbach experiment with a pressed H-ZSM5-wafer, a mixture of benzene/p-xylene, a temperature of 100C and a volume flow of 50ml/min

In another experiment the influence of the activation temperature on the separation was examined. In order to do so the activation was performed at 150 C instead of 450 C as in the previous experiments. As shown in Figure 27 and Figure 28 the overall form of the transmission curves was independent on activation temperature. Differences could be observed in the breakthrough time and the total uptake time. The breakthrough time was shortened by 6 minutes from 12 to 6 minutes for benzene and by 9 minutes from 26 to 17 minutes in the case of p-xylene. The breakthrough curves got more flat in both cases which results in a less overshooting in the case of benzene and longer uptake times in both cases.

Results

35

10

-4

10

-5

150C 450C

c [mol/l]

10

-6

10

-7

10

-8

50

100

150

time [min]

200

250

300

350

400

Figure 27: Concentration vs. time diagram of benzene breakthrough curves with a activation temperature of 150 (black), 450C (red) in a Wicke Kallenbach experiment with a pressed HZSM5-wafer of 100 mg, a mixture of benzene/p-xylene, a temperature of 100C and a volume flow of 50ml/min.
10
-4

10

-5

150C 450C

c [mol/l]

10

-6

10

-7

10

-8

50

100

150

time [min]

200

250

300

350

400

Figure 28: Concentration vs. time diagram of p-xylene breakthrough curves with a activation temperature of 150 (black), 450C (red) in a Wicke Kallenbach experiment with a pressed HZSM5-wafer of 100 mg, a mixture of benzene/p-xylene, a temperature of 100C and a volume flow of 50ml/min.

But even if the uptake time got longer a constant separation factor is reached earlier specifically after 80 instead of 120 minutes (Figure 29). The constant separation factor equals the previous experimental results with a value of 1.

Results

36

10

150C 450C

10
Seperation Factor [-]

10

10

-2

10

-4

20

40

60

80

time [min]

100

120

140

160

180

200

Figure 29: Seperation factor vs. time diagram with a activation temperature of 150 (black), 450C (red) in a Wicke Kallenbach experiment with a pressed H-ZSM5-wafer of 100 mg, a mixture of benzene/p-xylene, a temperature of 100C and a volume flow of 50ml/min.

To check, whether the pressure of the compacting had an influence on the result of the experiment a wafer of 100 mg were pressed with minimum force of 10 kN. The curves shown in Figure 30 and Figure 31 revealed the same behavior as the previously described with the over- swinging in the benzene transmission curve but in a less extend as previously observed and the pure exponential curve of the p-xylene graph.

10

-5

10 kN 20 kN

c [mol/l]

10

-6

10

-7

10

-8

50

100

time [min]

150

200

Figure 30: Concentration vs. time diagram of benzene breakthrough curves with a compacting pressure of 10kN (black), 20kN (red) in a Wicke Kallenbach experiment with a pressed HZSM5-wafer of 100 mg, a mixture of benzene/p-xylene, a temperature of 100C and a volume flow of 50ml/min.

Results

37

10 kN 20 kN 10
-5

c [mol/l]

10

-6

10

-7

10

-8

50

100

time [min]

150

200

250

300

Figure 31: Concentration vs. time diagram of p-xylene breakthrough curves with a compacting pressure of 10kN (black), 20kN (red) in a Wicke Kallenbach experiment with a pressed HZSM5-wafer of 100 mg, a mixture of benzene/p-xylene, a temperature of 100C and a volume flow of 50ml/min.

Differences were seen in the breakthrough times. They were 7 instead of 12 and 19 instead of 26 minutes for benzene and p-xylene, respectively. Even if the breakthrough time differed the overall uptake times stayed the same which were 200 for benzene and 240 minutes in the case of p-xylene. The separation factor was identical to the former experiments 1. The time dependence of the separation factor is shown in Figure 32.
10
3

10kN 20kN

10
Seperation Factor [-]

10

10

10

-1

20

40

60

80

time [min]

100

120

140

160

180

200

Figure 32: Seperation vs. time diagrams with a compacting pressure of 10kN (black), 20kN (red) in a Wicke Kallenbach experiment with a pressed H-ZSM5-wafer of 100 mg, a mixture of benzene/p-xylene, a temperature of 100C and a volume flow of 50ml/min.

Results

38

The form stayed untouched of the compacting even the height of the maximum is precisely the same with values around 700. The point of the maximum is shifted from 20 to 15 minutes for 10 kN corresponding to the faster uptake for the less compacted sample. In addition the desorption (Figure 33) of benzene and p-xylene from the WickeKallenbach-wafer of 100 mg was evaluated at 100C. Therefore, a normal uptake procedure was performed as described before. After the concentrations of the adsorbates reached a constant level the adsorbed species were desorbed with unloaded sweep gas streams of 50 ml/min each until both concentrations equaled zero.
4 x 10 3.5 3 2.5
c [mol/l]
-5

Benzene p-Xylene

2 1.5 1 0.5 0 0 200 400 600 800 1000 1200 1400 1600 1800 2000

time [minutes]

Figure 33: Desorption-curve of benzene (blue) and p-xylene (red) at 100C until 3090 min then the desorption was performed with a heating of 1C per minute until 450C and afterwards cooled down stepwise to 100 C again.

At that point an activation program with a heating rate of 1C per minute was started with a final temperature of 450C. This temperature was then hold for 1 hour and subsequently cooled down stepwise to 100C. In here the desorption-curve of benzene was much steeper than the one of pxylene. It reached zero after 37 minutes and showed no more desorption even if the temperature was raised. In contrast to that the p-xylene desorption curve equals zero after 177 minutes and showed a clear maximum in desorption after the temperature was increased corresponding to a temperature of 176C.

Results

39

An identical Wicke Kallenbach experiment using a standard 100 mg wafer as described previously was also performed with m-xylene and p-xylene as single components due to experimental constraints. The result is shown in Figure 34.

10

-5

p-Xylene m-Xylene

c [mol/l]

10

-6

10

-7

50

100

150

time [min]

200

250

300

350

Figure 34: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and consecutively p-xylene (black) and m-xylene (red), 100C, 100mg wafer weight and 50 ml/min volume flow.

As m- and p-xylene have the same mass distribution in a mass spectrometer it was not possible to determine the separation in one experiment. Hence the transmission of the single components was examined. The form of both curves was identically exponential with a large slope in the starting region. Differences could be found in transmission and the uptake times. Thereby m-xylene showed a breakthrough time of 10 minutes whereas p-xylene needed 40 minutes to transmit. In contrast the uptake time was much longer for m-xylene than for p-xylene specifically 350 and 100 minutes for mxylene and p-xylene , respectively.

Results

40

Table 3: Experimental results of the Wicke-Kallenbach experiments

Sample

Experiment Number

Wafer Weight [mg] 100 100 100 150 50 100 100 100

Compacting pressure [kN] 20 20 20 20 20 20 10 20 Uptake Time p-Xylene [min] 185 210 250 240 450 300 370 100

Experimental Activation Temperature Temperature [K] [K] 373 353 403 373 373 373 373 373 723 723 723 723 723 423 723 723

Breakthrough Time Benzene [min] 12 19 7 23 3 7 7

Breakthrough Time m-xylene [min]

Breakthrough Time pXylene [min] 27 41 17 47 9 16 17 40 Separation factor

H-ZSM5 H-ZSM5 H-ZSM5 H-ZSM5 H-ZSM5 H-ZSM5 H-ZSM5 H-ZSM5

1 2 3 4 5 6 7 8

10 Equilibrium Concentration m-Benzene [mmol/l]x102 Equilibrium Concentration p-xylene [mmol/l]x102 2.49 2.76 2.4 1.6 2.96 2.22 2.73 1.66

Experiment Uptake Number Time Benzene, [min] 1 2 3 4 5 6 7 8 270 290 270 295 250 350 400

Uptake Time m-Xylene [min]

Equilibrium Concentration Benzene [mmol/l]x102 3.6 3.7 3.55 2.51 5.4 3.53 3.75

350

0.81

1.01 0.94 1.03 1.14 1.28 1.11 0.91 1.23

Results

41

The most important factor, the separation factor, was 1.23 in this experiment taking the ratio of feed concentration into account which was 2.52 p-xylene/mxylene. All experiments are summarized in Table 3. 4.2. Frequency Response experiments To understand which transport mechanism is rate determining in the Wicke Kallenbach experiment three different frequency response experiments were carried out. The first measurement was performed with a powdered sample, the next with very thin wafers which were of a weight between 6 and 10 mg and lastly with a wafer of the same type as used in the Wicke Kallenbach cell with a weight of 100 mg. The wafers were broken into smaller pieces before starting the measurement to fit into the quartz sample holder. The characteristic functionns obtained from the measurements of the powder and the thin wafer are compared in Figure 35 and Figure 36. The functions of the Wicke Kallenbach wafer are not shown here because the time constant (L2/D) was too large and therefore no suitable data was obtained. The out of phase function (Figure 35) was nearly identical for both samples showing a maximum at 0.02 Hz. This maximum corresponds to a time constant of 50.25s using the method described by Gobin et al. [60] which defines the reciprocal value of the maximum of the out-of-phase function to be the time constant. In addition the diffusion coefficients were calculated by simply dividing L2 by the time constant. Assuming a particle size of 0.4m obtained by SEM the diffusion coefficients were 2.15x10-15 and 2.36x10-15 m2/s for the thin wafer and the powdered sample, respectively. Curve fitting was applied employing the model equations described in 2.3. using on the one hand a simple pore diffusion model in an infinite plane sheet and on the other a more sophisticated model assuming a surface resistance control. Both models were used with and without a particle size distribution to see the impact of a non-uniform size distribution. The fits for the surface resistance model are also shown in Figure 35 and Figure 36. The obtained transport parameters for all theoretical models and both samples are summarized in Table 4 and Table 5.

Results

K = 0.25392

42

0.1 0.08 0.06


out-of-phase characteristic function [-]

powder fit powder fit thin Wafer thin Wafer

0.04 0.02 0 -0.02 -0.04 -0.06 -0.08 -3 10 10


-2

frequency [Hz]

10

-1

10

10

Figure 35: Out of phase function of the ZSM5 sample in powdered state (black) and pressed to a thin wafer (red) including the fit (lines) with a surface resistance model in a non-uniform sheet.
0.35 0.3
in-of-phase characteristic function [-]

powder fit powder thin wafer fit thin wafer

0.25 0.2 0.15 0.1 0.05 0 -3 10

10

-2

frequency [Hz]

10

-1

10

10

Figure 36: In of phase function of the ZSM5 sample in powdered state (black) and pressed to a thin wafer (red) including the fit (lines) with a surface resistance model in a non-uniform sheet.

Both samples showed nearly the same parameters independent on the assumed model. The time constant was in the range of 67 to 75 seconds, the corresponding diffusion coefficient was between 2.16 and 2.59 x 10 -15 m2/s. K showed values between 0.24 and 0.27. All of the models had a apparently good fitting probability confirmed by the normalized root mean square errors which were in an acceptable range of 0.40 to 0.52. Thereby the model with

Results

43

non-uniform particle sizes and a surface resistance had the best correlation with the experimental data.
Table 4: Transport parameters for a powdered sample of ZSM5 obtained by curve fitting assuming theoretical models describing a uniform infinite sheet, a uniform infinite sheet with surface resistance control a non-uniform infinite sheet and a non-uniform infinite sheet with surface resistance control

Model

L2/D [s]

1015 K

NRMS-error

[m2/s] Uniform infinite sheet 70.48 2.27 2.27 0.24 0.24 0.52 0.52

Uniform infinite sheet with surface 70.36 resistance Non-uniform infinite sheet 61.74

2.59 2.38

0.25 0.25

0.45 0.44

Non-uniform infinite sheet with 67.24 surface resistance

Table 5: Transport parameters for a thin wafer of ZSM5 obtained by curve fitting assuming theoretical models describing a uniform infinite sheet, a uniform infinite sheet with surface resistance control a non-uniform infinite sheet and a non-uniform infinite sheet with surface resistance control

Model

L2/D [s]

1015 K

NRMS-error

[m2/s] Uniform infinite sheet 68.40 2.34 2.34 0.26 0.26 0.43 0.44

Uniform infinite sheet with surface 68.32 resistance Non-uniform infinite sheet 73.20

2.19 2.16

0.27 0.27

0.40 0.40

Non-uniform infinite sheet with 74.18 surface resistance

4.3. Uptake Rate measurements The gravimetric uptake curves of an 8 mg piece of a 100 mg Wicke Kallenbach wafer having a thickness of 1 mm were obtained at adsorbate partial pressures of 0.1, 1, 2 and 10 mbar and 373 K. All of the curves showed an exponential uptake with a large time constant (L2/D) between 3.58x102 and 5.77x103 seconds obtained by curve fitting with the theoretical model described in 2.4.

Results

44

1.4 1.2 1
normalized concentration [-]

0.8 0.6 0.4 0.2 0 0

500

time [s]

1000

1500

Figure 37: Uptake curve of benzene in an H-ZSM5 Wicke Kallenbach wafer of 100 mg at a partial pressure of 1 mbar and 373K. The figure shows the experimental data (black) and the result of the curve fitting (red) applying an infinite plane sheet model of 1mm thickness.

1 0.8
normalized mass [-]

0.6 0.4 0.2 0 0

0.1 mbar 1 mbar 2 mbar 10 mbar

200

400

600

time [s]

800

1000

1200

1400

1600

Figure 38: Fitted uptake curves of benzene in an H-ZSM5 Wicke Kallenbach wafer of 100 mg at 373K at partial pressures of 0.1mbar (black), 1mbar (red), 2mbar (blue) and 10 mbar (green).

Figure 37 shows an example of such an uptake curve for the case of 1 mbar having a clear exponential behavior. It also shows the good quality of the fit and therefore of the applied model. Figure 38 summarizes the uptake measurements by showing the results of the curve fitting. Thereby the graphs got steeper with increasing partial pressure of the adsorbate which is equal to a decreasing time constant with the loading. This indicates a clear dependence of the diffusion coefficient on the concentration which can be calculated by

Results

45

dividing L2 by the time constant. Thereby half of wafer thickness which is 0.5 mm equals L. The obtained parameters are summarized in Table 6.
Table 6: Summary of the parameters obtained curve fitting of the experimental data of a HZSM5 Wicke Kallenbach wafer of 100mg applying a theoretical model describing an isothermal infinite plane sheet of the thickness 1mm

Partial pressure [mbar] 0.1 1 2 10

L2/D x 102 [s] 57.7 18.3 7.75 3.58

D x 1010 [m2/s] 0.43 1.37 3.23 6.98

NRMS-error 0.30 0.26 0.73 0.64

To see how the diffusion coefficients of benzene in a ZSM5 Wicke Kallenbach wafer depend on the partial pressure they were plotted with the partial pressure on the x-axes and the diffusion coefficient on the y-axes. The outcome is shown in Figure 39. A linear dependence is obtained.
8,00E+01 7,00E+01 6,00E+01 5,00E+01 4,00E+01 3,00E+01 2,00E+01 1,00E+01 0,00E+00 0 2 4 6 partial pressure [mbar] 8 10 12

Figure 39: Dependence of the diffusion coefficient on the partial pressure of Benzene in a Wicke Kallenbach wafer of 100mg.

To determine differences between the Wicke Kallenbach wafer and the powdered sample the uptake rate of a powdered sample was also measured. In order to do so 4.3 mg of the powder was dispersed on glass wool analogue to the pressure frequency response experiments to avoid bed effects. The adsorbate benzene was dosed into the system in steps of 0.1, 1, 2 and 10 mbar by simultaneously tracking the mass.

Diffusion coefficient x 1011[m2/s]

Results

46

1 0.8
normalized mass [-]

0.6 0.4 0.2 0 0

0.1 mbar 1 mbar 2 mbar 10 mbar

50

time [s]

100

150

200

Figure 40:: Fitted uptake curves of benzene in H-ZSM5 powder at 373K at partial pressures of 0.1mbar (black), 1mbar (red), 2mbar (blue) and 10 mbar (green).

Data processing was performed identical to the other experiment by non-linear parameter fitting using a CMA evolutional strategy in matlab which resulted the fitting shown Figure 40 and Table 7.
Table 7: Summary of the parameters obtained curve fitting of the experimental data of a powdered H-ZSM5 applying a theoretical model describing isothermal infinite plane sheet of the thickness 0.4 m

Partial pressure [mbar] 0.1 1 2 10

L2/D x 102 [s] 2.22 2.7 3.4 6.14

D x 1016 [m2/s] 7.21 5.93 4.71 2.61

NRMS-error 0.29 0.25 0.36 0.28

The characteristic length of diffusion is defined here as the particle size of 0.4 m. In contrast to the wafer the diffusion coefficient decreased with increasing partial pressure which can also be seen in Figure 41. A linear decrease of the diffusion coefficient with the pressure was obtained. This is most likely due to non idealities of the experiment. Especially it is impossible to perform the uptake of partial in a step function. Consequently the bigger the change of partial pressure is the longer the uptake takes which means that it is further away from the ideal step function. This means that the adsorption

Results

47

process of benzene in this MFI powder sample is too fast to get a reliable result from the gravimetric uptake.
8,00E+00 7,00E+00 6,00E+00 5,00E+00 4,00E+00 3,00E+00 2,00E+00 1,00E+00 0,00E+00 0 2 4 6 partial pressure [mbar] 8 10 12

4.4. Nitrogen physisorption isotherms The procedure of the nitrogen physisorption measurement of the Wicke Kallenbach wafer was identical to the one described in the experimental section (3.1). The outcome compared to the isotherm of the parent material is shown in Figure 42 and Figure 43.
350 300

Diffusion coefficient x 1016[m2/s]

Figure 41: Trend of the diffusion coefficient of H-ZSM5 powder obtained by uptake rate measurements.

Vads[cm3/g]

250 200 150 100 50 0 0,00E+00 2,00E-01 4,00E-01 6,00E-01 8,00E-01 1,00E+00 1,20E+00

P/P0 [-]
Figure 42:Nitrogen physisorption isotherms of H-ZSM5 parent powder material (black) and the same material pressed to a wafer of 100 mg (red).

Results

48

350 300

Vads[cm3/g]

250 200 150 100 50 0 2,00E-01 3,00E-01 4,00E-01 5,00E-01 6,00E-01 7,00E-01 8,00E-01 9,00E-01

P/P0 [-]
Figure 43: Nitrogen physisorption isotherms of H-ZSM5 parent powder material (black) and the same material pressed to a wafer of 100 mg (red) in the range of 0.2 to 0.9 p/p0.

Both curves are typical type 4 isotherms. A remarkable change was observed after pressing the powder to a wafer. The pressure region from the starting point until 0.2 was nearly identical for either the powder or the wafer. At 0.2 the hysteresis starts to open in the case of the wafer which occurred at 0.8 in the case of the powder. Moreover the total size of the loop obtained was much bigger in the case of wafer than for the powder. The data was then analyzed applying the BET formula resulting in a BET surface area of 4 46 m2/g.

Discussion

49

5. Discussion In order to understand how the ZSM5 wafer behaves during a Wicke Kallenbach experiment two theories were postulated describing the wafer. On the one hand a statistically dense membrane and on the other hand the classical bed adsorber. Statistically dense membrane represents in this context a mesoporous layered wafer in which the diffusing molecule has to pass at least one particle during the diffusion process (Figure 44).

Figure 44: Principle of a statistically dense membrane.

This means the whole transport mechanism is controlled by the diffusion through the particle which defines the kinetic of the pore diffusion as rate determining. This model assumes a steady state during the diffusion; hence the system is not in equilibrium at any time. The other possible model is a flow adsorber showing transmissions curves as described extensively before [53]. This model assumes an adsorption front going through the particle bed until the adsorbent is saturated.

Figure 45: Principle of a classical bed adsorber [72].

In contrast to the dense membrane this model is based on the assumption of thermodynamic control. Consequently it has an adsorption region (Figure 45)

Discussion

50

where the system is not in equilibrium but the overall rate is controlled by the strength of adsorption. It assumes an established adsorption/desorption equilibrium in the wafer after the breakthrough of the adsorbates is finished. The first remarkable aspects in the obtained data were the breakthrough times. The overall trend was that the breakthrough of benzene was faster than the one of p-xylene. This behavior can be explained with both models. In the case of the dense membrane the faster diffusing component would penetrate faster. This assumption would be in good accordance with Gobin et al. [60] who found intra crystalline diffusion coefficients of 5.28x10 -13 and 2.46x10-13 m2/s for benzene and p-xylene, respectively. But on the other hand this breakthrough time distribution is also typical for flow adsorbers. Based on the theory the weaker adsorbing substance penetrates first. This matches the found breakthrough times which are in good agreement with Mukti et al. who determined adsorption enthalpies of benzene and pxylene of -51 kJ/mol and 94 kJ/mol, respectively [64]. The next interesting result was the temperature dependence. A clear decrease of the transmission time with increasing temperature was found which would also be typical for both models. This can be explained for the membrane with an increase of the diffusion coefficient with increasing temperature as described in the literature [60]. The higher diffusion coefficient leads to a faster transport through the pore system resulting in a decreased breakthrough time at higher temperatures. The explanation for the adsorber model is the decreasing capacity with the increasing temperature as indicated by the isotherms. They show a decrease of loading at increasing temperature. As a result less substance can be adsorbed at the adsorbent which have to lead also to a decreased transmission time. Furthermore an influence on the wafer thickness was found. An increase of the breakthrough time with increasing thickness was observed. This is in agreement with both models. The explanation for the model of a statistically dense membrane is simple, the thicker the wafer is the more particles have to be passed statistically.

Discussion

51

On the other hand this observation agrees with the adsorber model. Increasing the wafer thickness has the same effect as decreasing the temperature: it results in a higher adsorption capacity. Based on this theory a higher capacity leads to a longer breakthrough time [65]. The transmission times clearly decreased when the activation temperature was decreased to 423K because the water was not removed completely from the zeolite. This is inconsistent with the membrane model because a higher loading makes it harder to diffuse through a particle. As a result the breakthrough time should had been increased or at least stayed constant and not decreased as it is shown in Figure 27 and Figure 28. Considering the adsorber model a higher water loading lowers the capacity because less adsorption places are reachable for the diffusing substance. As a consequence shorter transmission times are expected. Thus, the experimental data also agreed with the postulate of a bed adsorber analogue. After discussing the breakthrough times the form of the transmission curve has to be considered. In the experiments an overshooting in the benzene graph and a pure exponential uptake of p-xylene was observed in every experiment with H-ZSM5 as adsorbent independent on the temperature, wafer thickness, activation temperature or compacting pressure. This type of breakthrough curves is typical for bed adsorbers [53]. The overshooting arises due to the competitive adsorption of benzene and pxylene. In this case benzene adsorbs faster than p-xylene but the adsorption is weaker. Consequently benzene is displaced by p-xylene which then occupies the adsorption centers and by this leads to an accumulation of benzene in the non-equilibrium zone. Once the equilibrium in the adsorber is established the concentrations of benzene and p-xylene are constant. It follows that the concentration of benzene shortly after the breakthrough is higher than the steady state concentration which can be seen as an overshooting of the graph. As discussed before the weaker adsorbing substance broke through first and showed an overshooting. In contrast to this the adsorbate with the higher adsorption enthalpy, in this case p-xylene [64], had a longer breakthrough time and showed no overshooting at all. Therefore, the form of the transmission curve can be explained by the model of a classical adsorber.

Discussion

52

To allow the comparison of the experimental data with the postulate of the dense membrane the experimental data was tried to fit numerically. A solution of the second fickian law derived by Crank et al. [54] for the diffusion through a membrane was used: (5.1) where C is the concentration at infinite time, D is the diffusion coefficient which was previously determined by Gobin et al. [60] and L is the characteristic length of diffusion Unfortunately it was impossible to fit the experimental data with a sufficient goodness of fit. This means that it is not possible to describe the experimentally obtained data with the model of a membrane. This strongly suggests that the model of a statistically dense membrane is not describing this particular Wicke Kallenbach experiments in a proper way. The most important parameter for this process is the separation factor because the aim of these wafers was the separation of benzene and p-xylene. Unfortunately the separation factor in the experiments equaled independent on any changes 1. This result can`t be explained with the dense membrane model. If the transport process would had been controlled by pore diffusion a separation factor higher than 1 had to be found because according to Gobin et al. [60] the diffusion of benzene is by an order of 2 faster than the diffusion of p-xylene inside the ZSM5 pore system. As the model assumes a kinetic control the permeation flows of benzene and p-xylene must had been different from each other which would had led to a separation. According to Sun et al. [23] it is possible to calculate the concentration in the permeate at steady state conditions by: (5.2) where AW is the cross section of the wafer, D is the intracrystalline diffusion coefficient, KH is the Henry constant and L is the thickness of the wafer. The concentration of p-xylene in steady can be calculated from the ratio of the pxylene/benzene concentration. This equals taking V>> AxDxK/L into account: (5.3)

Discussion

53

According to 5.3 the separation factor can be mathematically described by Dp-xylene/Dbenzene for the model of a membrane. Assuming this model it is possible to calculate the theoretically expected separation factors according to the previously reported diffusion coefficients [60]. The results are shown in Table 8.
Table 8: Theoretically expected separation factors of the Wicke Kallenbach experiment assuming a statiscally dense membrane applying the diffusion coefficients reported by Gobin et al. [60]

T [C]

Diffusion coefficient Diffusion coefficient Benzene x 10-15 p-xylene x 10-15 [m2/s] [m2/s] 3.96 7.34 12.9 0.97 2.48 6.02

Seperation factor [-] 4.08 2.96 2.14

70 100 130

Also a temperature dependence on the permeation flux [3,66,67] is expected as the diffusion coefficient varies with the temperature which would had influenced separation factor as well. But the results show that neither the expected temperature dependence nor a separation factor at all was observed in the experiments. Furthermore the separation factor showed a clear dependence on the time which means that it reached a maximum after 10 to 50 minutes depending on the temperature, the wafer thickness, the compacting pressure and the activation temperature. The observed behavior fits perfectly the model of an adsorber. As expected the position of the maximum corresponded to the capacity of the adsorber. The time until the maximum was reached increased with increasing capacity produced by a lower temperature during experiment and activation and a higher thickness of the wafer. The maximum of the separation factor occurred at the same time as the maximum of the over swinging of the benzene concentration curve. It was identical with the breakthrough time of p-xylene. To determine the amount of adsorbed species on the pellet a desorption experiment was carried out. In order to do so adsorption until equilibrium conditions was achieved, afterwards the adsorbate stream was stopped and the adsorbate was desorbed with the help of sweep gas (50 ml/min). The amount of adsorbed species was then determined by integration with respect

Discussion

54

to t of the resulted data. The obtained values were then multiplied by 2 because the desorption occurred in both directions of the cell but the amount was only tracked at one of the streams. Hence, the concentration of adsorbed species inside the wafer equals the integration with respect to t of the tracked concentration signal multiplied by 2. The integration resulted in a concentration of 5.86 x 10-4 and 1.33 x 10-3 mol/l for benzene and p-xylene. This led to a ratio of benzene/p-xylene of 0.44 which means that the concentration of p-xylene was more than two times higher than the concentration of benzene. This result is in good accordance with the adsorption enthalpies measured by Mukti et al. [64]. It is another hint for the adsorber theory because this model predicts that the process is

thermodynamically controlled. By this means the concentration ratio inside the sheet must be equal to the ratio calculated from the isotherms which depend on the adsorption enthalpies. To confirm the theory of the adsorber the number of hits was calculated which the molecules have with the pore wall during one transmission through the wafer. This is important because it has to be shown that it is possible to establish an adsorption/desorption equilibrium during the transmission time. If it is impossible to build up an equilibrated state the whole adsorber theory is no longer supportable. The base of the calculation were the sticking probabilities obtained by Reitmeier et al. [4] having values of 2.10x10-7 and 2.18x10-7 for benzene and pxylene. This parameter describes the probability that a poke between the adsorbate and the outer surface of a porous material leads to a pore entering. The starting point of the calculation is the overall number of wall collisions depending on the partial pressure of adsorbate and the temperature derived from the kinetical gas theory [68]: (5.4) where P is the partial pressure of the adsorbate in the feed, m is the mass of one molecule, k is the Boltzmann constant and T is the temperature [K]. To be able to calculate how many hits a molecule has with the walls of the mesopores during the transmission through the wafer, the flow of molecules per second entering the sample has to be calculated:

Discussion

55

(5.5) In here p defines the partial pressure of the adsorbate in the feed, R in the gas constant, T is the temperature, F(He) is the flow of the sweep gas and NA is the Avogadro constant. With this data it is possible to determine the surface area of the adsorbent needed for one poke of the adsorbate: (5.6) In the next step the surface area needed for poke leading to the sorption of the adsorbate is identified by taking the sticking coefficient into account: (5.7) As the external surface area of the wafer is known the successful hits during the transmission of the adsorbate through the wafer could be obtained by: (5.8) The results for temperatures of 343, 373 and 403 K and the adsorbates benzene and p-xylene are summarized in Table 9.
Table 9: Results of the calculation considering the number of pokes leading to a pore entering

during the transmission of benzene and p-xylene through the Wicke Kallenbach wafer

Adsorbate

T [K]

PFeed [Pa]

Zw x 10-24 [1/(m2s)]

Fmolec.x10-16 [Molec./s]

Aper poke x109 [m2]

Aper suc. poke x102 [m2] 1.75 1.78 1.82 1.88 1.83 1.86

Nx102 [-]

Benzene Benzene Benzene p-Xylene p-Xylene p-Xylene

343 313,2 5.04 373 313,2 4.84 403 313,2 4.65 343 234 373 234 403 234 3.23 3.10 2.98

1.86 1.81 1.78 1.32 1.24 1.21

3.68 3.75 3.82 4.10 4.00 4.06

3.71 3.64 3.57 3.46 3.54 3.49

The number of successful pokes is in the range of 350 per molecule during the transmission through the wafer. This means that the establishing of an adsorption/desorption equilibrium as postulated in the adsorber theory is in good accordance with the previous determined sticking coefficients. Consequently the theory of an adsorption zone is also valid. Furthermore it is possible to estimate the length of the zone which is not in equilibrium known as adsorption zone because it equals the length between

Discussion

56

two successful pokes. This value can simply be evaluated by taking the reciprocal value of N and multiplying it with the total length of diffusion. This leads to a length of the adsorption zone of 2.9x10-3 mm. The final evidence for the assumed bed adsorber analogue gave the comparison of the uptake curves of m-xylene and p-xylene which had to be measured in two consecutive experiments due to experimental constraints. The result is shown in Figure 34. Obviously m-xylene transmitted after 10 minutes where as breakthrough of p-xylene took 40 minutes. The form did not change compared to the experiments of benzene and pxylene and showed both the same exponential form as pxylene in the previous measurements. This is due to the consecutive measurement: the overshooting as previously seen in benzene is only observed if two adsorbates are present and the adsorption enthalpies differ clearly [53]. The separation factor calculated as described in 2.2.1 had a value of 1.23 which equals 1 taking the error of the experiment into account. This is a totally different behavior compared to the previously reported separation of p-xylene from a mixture of o-, m- and p-xylene with the help of an MFI membrane [3]. Sakai et al. demonstrated that the differences in permeability due to the larger minimum kinetic diameters of m-xylene in comparism to p-xylene lead to a separation factor of 10 p-xylene/m-xylene at 373 K. This is significantly different to our experiment. This experiment shows that the pressed Wicke Kallenbach wafer did not act similar to a membrane and by this means the theory of a statically dense membrane is not valid for this system. In contrast to the theory a bed adsorber describes the experimental data very good. The transmission times can be understood by the accessible adsorption centers. M-xylene can only reach the outer surface having a BET surface area of 65 m2/g. The sorption into the pores of the zeolite to adsorb at the adsorption places inside the pore system is very slow due to the minimum kinetic diameter. This process could be seen in the slow uptake in the late time region of the uptake curve. The other adsorbate p-xylene was able to access the pore system to adsorb there which led to a transmission time of 40 minutes. Comparing this to the observed BET total surface area of 446 m2/g of the wafer it can be concluded

Discussion

57

that the uptake time has a clear dependence on the accessible adsorbtion centers. This is in good agreement with the postulated bed adsorber. The fact that no separation occurred was due to the rate determining step which was the Knudsen diffusion through the mesopores after the adsorption/desorption equilibrium is established. As m-xylene and p-xylene have the same mass the Knudsen diffusion coefficients have to be identical (2.2.5) as the speed of this transport type is independent on the kinetic diameter. All these facts summarized in Table 10 show that the model of an adsorber describes the wafer of the Wicke Kallenbach experiment bes
Table 10: Summary of the experimental facts and check of consistency to the adsorber and the statistically dense membrane model

Experimental Fact/Model breakthrough time temperature dependence on breakthrough time

Adsorber

Membrane

consistent consistent of consistent consistent

benzene/p-xylene influence of wafer thickness on breakthrough time of consistent inconsistent benzene/p-xylene dependence of the activation temperature on the consistent inconsistent

breakthrough time of benzene/p-xylene form of the uptake curves consistent -

equilibrium/ steady state separation factor = 1 benzene/p- consistent inconsistent xylene equilibrium/ steady state separation factor = 1 p-xylene/m- consistent inconsistent xylene missing temperature dependence on the separation factor consistent inconsistent of benzene/p-xylene time dependence of the separation factor of benzene/p- consistent xylene sticking probalities no activation energy of the diffusion consistent consistent consistent inconsistent

Discussion

58

Table 11: Summary of the results of the calculatng the diffusion coefficients of benzene and p-xylene in a Wicke Kallenbach wafer consisting of HZSM5. Furthermore the pore diameters calculated from these diffusion coefficients assuming Knudsen diffusion are reviewed

Sample

Wafer Weight [mg] 100 100 100 150 50 100 100

Compacting Experimental Activation pressure Temperature Temperature [kN] [K] [K] 20 20 20 20 20 20 10 373 343 403 373 373 373 373 723 723 723 723 723 423 723

Diffusion coefficient Benzene x 10-7 [m2/s] 6.51 6.74 6.39 6.56 4.88 6.32 6.61

Diffusion coefficient p-xylene x 10-7 [m2/s] 6.39 7.20 6.18 5.77 3.60 5.62 7.16

pore diameter calculated from DBenzene x 10-9 [m] 6.14 6.53 6.11 6.18 4.61 5.96 6.24

pore diameter calculated from Dp-xylene x 10-9 [m] 7.02 8.14 6.54 6.35 3.96 6.18 7.88

H-ZSM5 H-ZSM5 H-ZSM5 H-ZSM5 H-ZSM5 H-ZSM5 H-ZSM5

Discussion

59

This fact leads to a characteristic length of diffusion equally to the thickness of the wafer, in contrast to the radius of on particle as in the case of the statistically dense membrane model. Consequently it was the possible to calculate the diffusion coefficients by equation (2.2.2) from the equilibrium molar ratios: (5.10) where L is characteristic length, r is the radius of the wafer, FHe is the flow rate of the sweep gas helium and xAdsorbate is the molar ratio of the adsorbate in the feed and the permeate at equilibrium conditions, respectively. From the observed diffusion coefficients which were in the range of 10 -6 m2/s we could assume Knudsen diffusion. As a consequence it was possible to calculate the radius of the pores by transposing (2.2.5). The transposed equation is expressed by: (5.11) where R is the gas constant, T is the temperature, M is the molar mass. The results of the calculation are shown in Table 11. The assumed Knudsen diffusion is confirmed by the calculated pore diameters which are in the range of 6 to 8 nm. Pores of this size were defined as mesopores [69] by the IUPAC (International Union of Pure and Applied Chemistry). This means that the diffusion in pressed wafers was limited by the transport in mesopores of 6 to 8 nm size.

Figure 46: SEM image of the surface of a Wicke Kallenbach wafer.

Discussion

60

This result fits to the impression of the SEM image which shows separated particles (Figure 46) It indicates as well that the wafer had pores of larger diameters than the calculated ones of 6 to 8 nm but these pores were obviously not rate determining. Thus, we can assume that the calculated pore diameters are the smallest which occurred in the Wicke Kallenbach pellets as they determined the transport rate. The assumption of Knudsen diffusion determining the transport through the wafer is also in good agreement with the physisorption isotherm of nitrogen.

0,5 0,45 0,4 y = 0,0095x + 0,3789

vads [cm3/g STP]

0,35 0,3 0,25 0,2 0,15 0,1 0,05 0 0 1 2 3 4 5 6 7 8 9 y = 0,048x + 0,147

tHasley [nm]
Figure 47: t-plot produced with the method of Hasley for the H-ZSM5 Wicke Kallenbach wafer.

As previously described the isotherm had a much larger loop even at low pressures of 0.2 P/P0 which corresponds to mesopores formed during the pressing procedure. The BET surface area did not change from the powder to the pressed disk and stayed at 420 to 450 m2/g. The total volume of the micro- and mesopores was determined using a t-plot (Figure 47) produced with the method of Halsey [70]. The results are summarized in Table 12. It resulted in a total micropore volume of 0.147 cm3/g and a mesopore volume of 0.232 cm3/g. A clear enhancement of the mesopore volume was obtained compared to the pore volume of the parent material [5] which was 0.147 cm3/g and 0.132 cm3/g micro and mesopore volume. The

Discussion

61

micropore volume stayed untouched which is the expected characteristic as the micropores built up by the zeolite framework. This is not influenced by the pressing of the powder and consequently the micropore volume had to stay the same. To have a closer look on the mesopores a DFT analysis of the physisorption isotherm was performed to obtain the pore diameter of mesopores shown in Figure 48.
Table 12: Results of the analysis if physisorption isotherms of H-ZSM5 in the powdered and the pressed form applying the BET equation [59] and the t-plot method [70]

Material powder 100 mg Wafer

SBET [m2/g] 420 446

VMicro [cm3/g] 0.147 0.147

VMeso [cm3/g] 0.132 0.232

The DFT analysis (Figure 48) showed an increase of mesopores with a maximum of the distribution at a pore diameter of 17 nm. This is in good agreement with the pore diameters calculated from the diffusion

measurements which were 6 to 8 nm. The huge pores indicated by the SEM image are not found in the physisorption isotherm as they are larger than 65 nm.
2,50E-03

2,00E-03

dV [cm3//g]

1,50E-03

1,00E-03

5,00E-04

0,00E+00 50 150 250 350 450 550 650

pore size []
Figure 48: pore size distribution of the parent H-ZSM5 material (black) and the Wicke Kallenbach wafer (red) obtained by DFT calculation.

Discussion

62

The assumption of Knudsen diffusion could also be confirmed by looking at the activation energy which was determined by an Arrhenius ansatz describing the diffusion coefficient by [66]: (5.12) (5.13) In here D is the diffusion coefficient, EA,D is the activation energy of diffusion, R is the gas constant and T is the temperature. According to 5.13 it was possible to obtain the activation energy of the diffusion of benzene and p-xylene for the rate determining step inside the sheet by plotting 1/T against ln (D) (Figure 49).
-14,1 0,0024 benzene p-xylene -14,2 0,0029

ln(D) [-]
-14,3

-14,4

1/T [1/K]

Figure 49: Determination of the activation energy of benzene (black) and p-xylene (red) by an Arrhenius ansatz.

The calculation resulted in activation energies of -1.03 and -2.96 kJ/mol for benzene and p-xylene, respectively. Taking the error of the system including the inlet system and the mass spectrometer into account the activation energy of diffusion was in the range of zero. Afterwards the ideal activation energy of the Knudsen diffusion in the examined system was determined by calculation the diffusion coefficients with: (5.14) The mean value of the obtained pore diameters from Table 11 was used for the calculation. The results are shown in Table 13, Table 14 and Figure 50.

Discussion

63

Table 13: Calculated ideal Knudsen diffusion coefficients of benzene

T [K] 343 373 403

D x 10-7 [m2/s] 6.38 6.65 6.91

Table 14: Calculated ideal Knudsen diffusion coefficients of p-xylene

T [K] 343 373 403


-14,1 0,0023 -14,2

D x 10-7 [m2/s] 5.47 5.70 5.93

0,0028 benzene p-xylene

ln(D) [-] -14,3

-14,4

-14,5

1/T [1/K]

Figure 50: Theoretical evaluation of the activation energy of benzene (black) and p-xylene (red) from the calculated Knudsen diffusion coefficients by a Arrhenius ansatz.

The results imply activation energies of 1.54 kJ/mol for both substances which shows that the Knudsen diffusion has nearly no activation energy in this system. Thus, the obtained activation energies of the measurements are in good agreement with the theory and are consequently another hint for the existence of Knudsen diffusion. This fact is also inconsistent with the previously described theory of a dense membrane because the apparent diffusion of benzene and p-xylene in the used ZSM5 sample has activation energies according to Gobin et al. [60] of 23 and 35 kJ/mol, respectively. Additionally curve fitting was preformed with the transmission curves of the Wicke Kallenbach experiment. The aim of this method is to fit kinetic

Discussion

64

parameters

from

the

transmission

curves

of

the

Wicke

Kallenbach

experiments. In order to do so a kinetic model was assumed and tried to fit the experimental data. The data was fitted with a set of differential equations describing an adsorber with an irreversible Langmuir adsorption. The breakthrough curve was in this case specified by two differential equations, one for the gaseous and one for the solid phase. These formulas define the concentration dependent on time of the adsorbate where the solid phase concentration is replaced by the surface coverage (). Hence, the rate of adsorption can be formulated by (6) [53]. (5.15) In here is the normalized surface coverage [71], ka is the reaction rate coefficient and c is the concentration of the adsorbate in the gas phase. The mass balance of the fluid phase describes the concentration of the adsorbate dependent on the time [72] (5.16) These set of differential equations can be solved analytically according to Bohart et. al [73]. The solution is shown in (5.17): (5.17) where C is the normalized concentration in the gas phase calculated by c/c0, ka denotes the reaction rate coefficient, c0 is the concentration in the feed, t is the reaction time, WSat is maximum amount of adsorbed gas, B is the density of the bed, S is the section of the adsorbent bed and F is the gas flow rate. Nonlinear parameter fitting was performed using a genetic algorithm implemented in matlab. The fitting parameters were F and ka. The objective minimized was the normalized root mean square error divided by t 3 to get a better fit in initial region of the curves. One example of the obtained fits is shown in Figure 51.

Discussion

65

1.2 1 0.8
normalized concentration [-]

0.6 0.4 0.2 0 -0.2 0

fit Benzene measurement Benzene fit p-Xylene measurement p-xylene

0.2

0.4

0.6

0.8

t [s]

1.2

1.4

1.6

1.8 x 10

2
4

Figure 51: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 100C, 100mg wafer weight and 50 ml/min volume flow and the corresponding fit for benzene (black) and p-xylene (blue).

0,09 0,08 0,07 0,06 ka [1/s] 0,05 0,04 0,03 0,02 0,01 0 60 70 80 90 100 T [C] Figure 52: Trend of the parameter k with the temperature. 110 120 130 140

Due to the very primitive model and the in comparison to that very complex adsorption behavior of the system the fits were poor. Nevertheless it was possible to find trends for the strength of adsorption and the diffusion coefficient. Hereby ka served as the parameter describing the limiting factor of the adsorption on particle level which is according to Gobin et al. [60] the pore entering step and F is proportional to the diffusion coefficient as the transport through the pellet is a fully diffusive.

Discussion

66

The only trend observed was the dependence on the temperature. This can be seen in the benzene fits. In here both parameters, k and F increased with increasing temperature which was the expected behavior as activated transport process getting faster with increasing temperature. These trends are shown in Figure 52 and Figure 53. But even if these parameters show a trend in the temperature it is very difficult to interpret the outcome with respect to the experiment. In conclusion the curve fitting shows that the used model is much too trivial to be able to describe the complex behavior of the Wicke Kallenbach system we looked at. Nevertheless it was shown that the fitting gave the same information as the experiment itself. This can be a good starting point for applying a more sophisticated model and try to understand the mechanism inside the pellet by mathematical modeling.
5,00E-11 4,50E-11 4,00E-11 3,50E-11 F [m3/s] 3,00E-11 2,50E-11 2,00E-11 1,50E-11 1,00E-11 5,00E-12 0,00E+00 60 70 80 90 100 T [C] Figure 53: Trend of the parameter k with the temperature. 110 120 130 140

Looking at the frequency response experiments is as well interesting. It was found that the diffusion coefficient of the powdered sample was about 2.5x10-15 m2/s for benzene in H-ZSM5 which is much slower than the measured diffusion coefficients in particles with a larger diameter. As intensely discussed by Gobin et al. [60] the rate determining step is in the case of such small particles no longer the intracrystalline diffusion but the surface adsorption and the pore entering step. Interestingly the time constant did not change after pressing the particles to thin wafers of 6 to 10 mg and diameters of 20 mm. It proves that the

Discussion

67

characteristic length of diffusion equals the one for a powder and a thin wafer which means that the rate controlling transport step is in both cases the same. This denotes that the diffusion in a thin wafer can be described in the same way as for a powdered sample. The behavior changed dramatically if the powder was pressed to Wicke Kallenbach pellets of 1 mm thickness. The characteristic functions for this experiment were not interpretable due to a very large time constants as can be seen in Figure 54. The maximum of the out-of-phase function was below 10-3 Hz which means that they are not in the range of the frequency response experiment. Therefore, the uptake rate measurements were performed because this method is able to determine large time constants.
0.14 0.12 0.1
characteristic functions [-]

fit-v1 in fit-v1 out

0.08 0.06 0.04 0.02 0 -0.02 -0.04 -3 10


-2 -1 0 1

10

Frequency [Hz]

10

10

10

Figure 54: In of phase (blue) and out of phase (red) functions obtained by a pressure frequency response experiment with a Wicke Kallenbach wafer of 100mg and benzene as adsorbate.

With this method time constants of 350 to 5700 s were found which are impossible to obtain with the frequency response method. The characteristic length of diffusion differed for the Wicke Kallenbach wafer as discussed before which was half of the wafer thickness (0.5 mm), in contrast to the particles (0.4 m). This led to diffusion coefficients of 10-10 m2/s. Taking into account the values obtained by the Wicke Kallenbach experiments (10-6 m2/s ) valid for the mesopores and the diffusivities for the powdered sample (10-15 m2/s) it can be assume that the adsorption in a Wicke Kallenbach wafer is controlled by a mixture of both processes. But it can be seen from the

Discussion

68

Wicke Kallenbach experiment that the flow through the wafer once the adsorption desorption equilibrium is reached is only controlled by the transport through the mesopores. Applying the powder to the uptake rate measurement of benzene time constants of 220 to 614 s were found. Surprisingly the time constant increased with increasing benzene partial pressure which means that the diffusion became slower with increasing vapor pressure. This behavior showed clearly the non idealities of the system. It can be concluded that the dosing of adsorbate into the system had the biggest influence on the result. This means that it is impossible to perform a step change of the adsorbate concentration with the used setup. Therefore, the diffusion of benzene in H-ZSM5 powder is too rapid for uptake rate measurements. This is in good agreement with the pressure modulation experiments and thereby with the previously reported diffusivities [60].

Conclusion

69

6. Conclusion The transport processes of strong adsorbing substances on pressed wafers in a Wicke Kallenbach cell were investigated. In order to do so the system HZSM5 with adsorbates benzene and p-xylene was examined. Experimental results confirmed that the wafer could be described as a bed adsorber analogue. The results leading to this conclusion were in particular the clear dependences of the transmission time, the uptake time and the form of the breakthrough curve on the temperature, the wafer thickness, the activation temperature and the compacting pressure. The wafer was a mesoporous accumulation of ZSM5 powder with the same BET surface area as the non-compacted sample of 430 m2/g. During the pressing procedure mesopores of 6 to 15 nm diameter were formed which was confirmed by DFT calculations of the nitrogen physisorption isotherm and calculations from the Wicke Kallenbach experiment. At equilibrium conditions, which means after saturation of the wafer the transport process was limited to the rate of diffusion through the mesopores. Thus the separation factor is 1 independent of external parameters like temperature, activation temperature, wafer thickness and compacting pressure for molecules having a similar molar mass. This could be seen in the experiments applying either p-xylene benzene mixtures or a blend of p-xylene and m-xylene. Especially the measurement with m-xylene and p-xylene varified that the wafer acted like a bed adsorber. The previously reported assumption, that the pressed wafer acts like a grown membrane and could separate pxylene from a mixture of xylenes as described in the literature applying a MFI membrane [3], proved to be wrong. So unfortunately it is impossible to separate hydrocarbons with a pressed membrane of MFI particles. The change in rate determining transport steps from the powder to the wafer was also approved by the combination of pressure modulation frequency response and uptake rate measurements. They show an increase of time constants of two orders of magnitude by compacting the powder to a wafer which is due to a increase of the charactistic length of diffusion from 0.4 m to 1 mm. The diffusion coefficient previously reported by Gobin et al. [60] for the

Conclusion

70

parent wafer was confirmed whereas the uptake rate measurements showed an increase in the diffusion coefficient by 5 orders of magnitude approving that the uptake in the wafer is determined by a combination of Knudsen diffusion and the pore entrance step. Nevertheless the Wicke Kallenbach experiments showed that once the adsorption desorption equilibrium is established the Knudsen diffusion controls the transmission rate.

Literature

71

Literature

[1] [2]

Ryoo, R.; Joo, S.H.; Jun, S., J. Phys. Chem. B, 1999, 103, 7743 Mriaudeau, P.; Tuan,Vu.A.; Sapaly, G., Nghiem, Vu..T.; Naccache, C., Catal. Today,

1999, 49, 285 [3] 297 [4] 15355 [5] 533 [6] [7] Ruthven, D.M.; Eic, M.; Richard, E., Zeolites, 1991, 11, 647 Ruthven, D.M., Principles of adsorption and adsorption processes, New York, Reitmeier, S.J.; Gobin, O.C.; Jentys, A.; Lercher, J.A.; Angew. Chem. Int. Ed. 2009, 48, Reitmeier, S.J.; Gobin, O.C.; Jentys, A.; Lercher, J.A., J. Phys. Chem. C, 2009, 113, Sakai, H.; Tomita, T.; Takahashi, T., Separation and Purification Technology, 2001, 25,

Chichester, Brisbane, Toronto, Singapore, John Wiley & Sons, 1984 [8] [9] [10] Cavalcante, C.L.; Ruthven, D.M.; Ind. Eng. Chem. Res., 1995, 34, 185 Song, L.; Rees, L.V.C., Microporous and Mesoporous Materials, 2000, 35-36, 301 Brandani, F., Deveplopment and application oft he zero length column (ZLC) technique

for measuring adsorption equilibria, The University of Maine, 2002 [11] [12] [13] [14] [15] Brandani, S.; Xu, Z.; Ruthven, D.M., Microporous Materials, 1996, 7, 323 Stejskal, E.O.; Tanner, J.E., J. Chem. Phys., 1965, 42, 288 Stejskal, E.O.; Tanner, J.E., J. Chem. Phys., 1968, 49, 1768 Jobic, H.; Bee, M., Krger, J.; Balzer, C.; Julbe, A., Adsorption, 1995, 1, 197 Jobic, H.; Bee, M.; Krger, J.; Vartapetian, R.S.; Balzer, C.; Julbe, A.; J. Membr. Sc.,

1995, 108, 71 [16] [17] [18] [19] Yasuda, Y., Heterogeneous Chemistry Reviews, 1994, 1, 103 Yasuda, Y., J. Phys. Chem. 1982, 86, 1913 Petkovska, M.; Do, D.D., Chemical Engineering Science, 1998, 53, 3081 Gobin, O.C.; Reitmeier, S.J.; Jentys, A.; Lercher, J.A., Microporous and Mesoporous

Materials, 2009, 125, 3 [20] Karger, J.; Kortunov, P.; Vasenkov, S.; Heinke, L.; Shah, D.R.; Rakocy; R.A.; Traa, Y.;

Weitkamp, J., Ang. Chem. Int. Ed., 2006, 46, 7846 [21] Chmelik, C.; Hibbe, F.; Tzoulaki, D.; Heinke, L.; Caro, J.; Li, J.; Krger, J., Microporous

and Mesoporous Materials, 2010, 3, 340 [22] [23] [24] Wicke, E.; Kallenbach, R.; Kolloid-Zeitschrift, 1941, 97, 135 Sun, M.S.; Talu, O.; Shah, D.B., AlChe Journal, 1996, 42, 3001 Keizer, K.; Burggraaf, A.J.; Vroon, Z.A.E.P.; Verweij; H., Journal of Membrane Science,

1998, 147, 159

Literature

72

[25]

Tuchlenski, A.; Uchytil, P.; Seidel-Morgenstern, A., Journal of Membrane Science,

1998, 140, 165 [26] [27] Arnst, D.; Schneider, P., The Chemical Engineering Journal, 1995, 57, 91 A.F. Hollemann, Lehrbuch der anorganischen Chemie / Hollemann-Wiberg, Begr. Von

A.F. Wiberg, Fortgef. von E. Wiberg - 101., verb. und stark erweiterte Auflage / von N. Wiberg Berlin ; New York ; de Gruyter, 1995 [28] J. Cejka, H. Van Bekkum, Zeolites And Ordered Mesoporous Materials: Progress And

Prospects, Volume 157: The 1st Feza School On Zeolites, Prague, Czech Republic, August 2021, 2005 [29] 86. [30] [31] Long R.Q., Yang R.T., J.Am.Chem.Soc., 1999, 121, 5595 Degnan, T.F., Chitnis, G.K., Schipper, p.H., Microporous and Mesoporous Materials Meisel, S. L.; McCullogh, J. P.; Lechthaler, C. H.; Weisz, P. B., Chem. Tech. 1976, 6,

2000, 35-36, 245 [32] [33] [34] [35] [36] Butler A.C., Nicolaides C.P., Catal. Today, 1993, 18, 443 Young, L.B., Butter, S.A., Kaeding, W.W., J. Catal., 1982, 76, 418 Yashima, T., Ahmad, H., Yamazaki, K., Katsuta, M., Hara, N., J. Catal. 1970, 16, 273 Stcker, M., Mostad, H. Rrvik, T., Catal. Letters, 1994, 28, 203 A.F.Cronstedt: Ron och beskrifning om en oberkant brg art, som kallas zeolites. In:

Kongl. Vetenskaps Acad. Handl. Stockholm, 1756, S. 120123 [37] [38] Compendium of Chemical Terminology IUPAC Research Triangle Park NC, 1997 C. Baerlocher, W. M. Meier, D. Olson, Atlas of Zeolite Framework Types, 5 ed.,

Elsevier, Amsterdam, 2001. [39] [40] [41] [42] [43] J. Weitkamp, Zeolites and catalysis, Solid State Ionics, 2000, 131, 175 Davis, E.M.; Lobo, R.F., Chem. Mater., 1992, 4, 756 W. Lowenstein, Am. Minearalogist, 1954, 39, 92. Dr. rer. nat. Stephan Reitmaier Zampieri, A., Development of MFI-type Zeolite Coatings on SiSiC Ceramic Monoliths

for Catalytic Applications, Erlangen, 2007 [44] [45] [46] Corma, A., J. Catal., 2003, 216, 298 Schmidt, I.; Madsen, C.; Jacobsen, C.J.H., Inorg. Chem., 2000, 39, 2279 Beck, J.S.; Vartuli, J.C.; Roth, W.J.; Leonowicz, M.E.; Kresge, C.T.; Schmitt, K.D.; Chu,

C.T-W. ; Olson, D.H.; Sheppard, E.W.; McCullen, S.B.; Higgins, J.B.; Schlenker, J.L., J. Am. Chem. Soc., 1992, 114, 10835 [47] [48] [49] [50] Cundy, S.C.; Cox, P.A., Chem. Rev. 2003, 103, 663 Derouane, E.G.; Detremmerie, S.; Gabelica, Z.; Blom, N., Appl. Catal. 1981, 1, 201 Barrer, R. M. Chem. Brit. 1966, 380. Barrer, R. M. In Molecular Sieves; Society of Chemical Industry: London, 1968; p 39.

Literature

73

[51] [52] [53]

Breck, D. W. Zeolite Molecular Sieves; Wiley: New York, 1974. Weller, M. T.; Dann, S. E. Curr. Opin. Solid State Mater. Sci. 1998, 3, 137. Baerns, M.; Behr, A.; Brehm, A.; Gmehling, J.; Hofmann, H.; Onken, U.; Renken, A.,

Technische Chemie, Weinheim, Wiley-VCH Verlag GmbH&Co. KGaA, 2006 [54] [55] [56] [57] [58] [59] [60] 20435 [61] [62] Kruk, M.; Jaroniec, M.; Choma, J., J. Carbon 1998, 36, 1447 Gregg, S. J.; W., S. K. S. Adsorption Surface and Porosity, 2 nd ed., Academic press Crank, J., The mathematics of diffusion, Oxford, Clarendon Press, 1975 Carnovale, F., Ph.D. Thesis, La Trobe University, 1980 Evnochides, S.K.; Henley, E.J., J. Polym. Sci. A-2, 1970, 8, 1987 Yasuda, Y., Bull. Chem. Soc. Jpn. 1991, 64, 954 Ruthven, D.M.; Lee, L.K.; Yucel, H., AlChe Journal, 1980, 26,16 Brunauer, S.; Emmett, P.H.; Teller, E., J. Am. Chem. Soc., 1938, 60, 309 Gobin, O. C.; Reitmeier, S. J.; Jentys, A.; Lercher, J. A., J. Phys. Chem. C, 2009, 113,

inc.: New York, 1982 [63] Hansen, N. The CMA evolution strategy: a comparing review. In Towards a new

evolutionary computation. Advances on estimation of distribution algorithms; Lozano, J. A., Larranaga, P., Inza, I., Bengoetxea, E., Eds.; Springer: New York, 2006; p 75. [64] [65] Mukti, R.R.; Jentys, A.; Lercher, J.A., J. Phys. Chem. C, 2007, 111, 3973 Vauck, W.R.A.; Mller, H.A., Grundoperationen Chemischer Verfahrenstechnik,

Stuttgart. Dt. Verlag fr Grundstoffindustrie, 2000 [66] van de Graaf, J.M.; Kapteijn, F.; Moulijn, J.A., Microporous and Mesoporous Materials,

2000, 35-36, 267 [67] [68] [69] [70] [71] [72] van de Graaf, J.M.; Kapteijn, F.; Moulijn, J.A., J. Membr. Sc., 1998, 144, 87 Atkins, P.W., Physical Chemistry, Fifth edition, Oxford University Press, Oxford, 1994 Compendium of Chemical Terminology IUPAC Research Triangle Park NC, 1997 Kruk, M. Jaroniec, J. Choma, Carbon 1998, 36, 1447 Lercher J.A., Katalyse und Reaktionstechnik, 2010 Joly, A.; Volpert, V.; Perrard, A., Dynamic Adsorption with FEMLAB, Modelling

breakthrough curves of gasous pollutants through activated carbon beds, Excerpt from the Proceedingsof the COMSOL Multiphysics Users Conference 2005 Paris [73] Bohart, G.S.; Adams, E.Q., J. Amer. Chem. Soc., 1920, 42, 523

Appendix A: Results

A. Results A.1 Wicke Kallenbach experiments

A.1.1Variation of the temperature


-4

10

10

-5

10
c [mol/l]

-6

Benzene p-Xylene

10

-7

10

-8

10

-9

10

-10

50

100

150

time [min]

200

250

300

350

Figure I: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 70C, 100mg wafer weight and 50 ml/min volume flow.
10
-4

10

-5

10
c [mol/l]

-6

Benzene p-Xylene

10

-7

10

-8

10

-9

10

-10

50

100

150

time [min]

200

250

300

350

Figure II: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 100C, 100mg wafer weight and 50 ml/min volume flow.

Appendix A: Results

II

10

-4

10

-5

Benzene p-Xylene

10
c [mol/l]

-6

10

-7

10

-8

10

-9

50

100

time [min]

150

200

250

300

Figure III: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 130C, 100mg wafer weight and 50 ml/min volume flow.

10

10
Seperation Factor [-]

10

10

10

-1

50

100

150

time [min]

200

250

300

350

Figure IV: Time dependence of the separation factor of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 at 70C, 100mg wafer weight, a mixture of benzene/p-xylene and 50 ml/min volume flow.

Appendix A: Results

III

10

10
Seperation Factor [-]

10

10

10

-2

50

100

150

time [min]

200

250

300

350

Figure V: Time dependence of the separation factor of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 at 100C, 100mg wafer weight, a mixture of benzene/p-xylene and 50 ml/min volume flow.
10
3

10
Seperation Factor 130 C H-ZSM5

10

10

10

-1

10

-2

50

100

150

time [min]

200

250

300

350

Figure VI: Time dependence of the separation factor of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 at 130C, 100mg wafer weight, a mixture of benzene/p-xylene and 50 ml/min volume flow.

Appendix A: Results

IV

A.1.2 Variation of the wafer thickness

10

-5

Benzene p-Xylene

c [mol/l]

10

-6

10

-7

10

-8

100

200

300

time [min]

400

500

600

700

800

Figure VII: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 100C, 50mg wafer weight and 50 ml/min volume flow.
10
-4

10

-5

Benzene p-Xylene
c [mol/l]

10

-6

10

-7

10

-8

100

200

300

time [min]

400

500

600

700

Figure VIII: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 100C, 150mg wafer weight and 50 ml/min volume flow.

Appendix A: Results

10

10

10
Seperation Factor [-]

10

10

10

-1

10

-2

100

200

300

time [min]

400

500

600

700

800

Figure IX: Time dependence of the separation factor of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 at 100C, 50mg wafer weight, a mixture of benzene/p-xylene and 50 ml/min volume flow.
10
4

10

10
Seperation Factor [-]

10

10

10

-1

10

-2

100

200

time [min]

300

400

500

600

Figure X: Time dependence of the separation factor of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 at 100C, 150mg wafer weight, a mixture of benzene/p-xylene and 50 ml/min volume flow.

Appendix A: Results

VI

A.1.3 Variation of the activation temperature


-4

10

10

-5

Benzene p-Xylene

c [mol/l]

10

-6

10

-7

10

-8

50

100

150

time [min]

200

250

300

350

400

Figure XI: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 100C, 100mg wafer weight and 50 ml/min volume flow, activated at 423 K.
10
4

150C 450C

10
Seperation Factor [-]

10

10

-2

10

-4

20

40

60

80

time [min]

100

120

140

160

180

200

Figure XII: Time dependence of the separation factor of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 at 100C, 100mg wafer weight, a mixture of benzene/p-xylene and 50 ml/min volume flow, activated at 423 K (black) and 723 K (red).

Appendix A: Results

VII

A.1.4 Influence of the compacting pressure

10

-5

Benzene p-Xylene
c [mol/l]

10

-6

10

-7

10

-8

50

100

time [min]

150

200

250

300

Figure XIII: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed

wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 100C, 100mg wafer weight and 50 ml/min volume flow, compacted with a pressure of 10 kN.

10

10kN 20kN

10
Seperation Factor [-]

10

10

10

-1

20

40

60

80

time [min]

100

120

140

160

180

200

Figure XIV: Time dependence of the separation factor of the Wicke Kallenbach experiment using

a pressed wafer of ZSM5 at 100C, 100mg wafer weight, a mixture of benzene/p-xylene and 50 ml/min volume flow, compacted with pressure of 10 kN (black) and 20 kN (red)

Appendix A: Results

VIII

A.2

Frequency Response experiments

A.2.1 H-ZSM5 powder

0.3 0.25 0.2


characteristic functions [-]

in-of-phase out-of-phase fit in-of-phase fit out-of-phase

0.15 0.1 0.05 0 -0.05 -0.1 -3 10


-2 -1 0 1

10

Frequency [Hz]

10

10

10

Figure XV: In- (black) and out-of-phase (red) functions of the batch FR experiment of H-ZSM5 powder. The fit was produced by applying a non-uniform infinite sheet model.
0.3 0.25 0.2
characteristic functions [-]

in-of-phase out-of-phase fit in-of-phase fit out-of-phase

0.15 0.1 0.05 0 -0.05 -0.1 -3 10


-2 -1 0 1

10

Frequency [Hz]

10

10

10

Figure XVI: In- (black) and out-of-phase (red) functions of the batch FR experiment of H-ZSM5

powder. The fit was produced by applying a non-uniform infinite sheet model with surface resistance.

Appendix A: Results

IX

A.2.2 H-ZSM5 thin wafer


0.3 0.25 0.2
characteristic functions [-]

in-of-phase out-of-phase fit in-of-phase fit out-of-phase

0.15 0.1 0.05 0 -0.05 -0.1 -3 10


-2 -1 0 1

10

Frequency [Hz]

10

10

10

Figure XVII: In- (black) and out-of-phase (red) functions of the batch FR experiment of an HZSM5 Wicke Kallenbach wafer. The fit was produced by applying a non-uniform infinite sheet model.

0.3 0.25 0.2


characteristic functions [-]

in-of-phase out-of-phase fit in-of-phase fit out-of-phase

0.15 0.1 0.05 0 -0.05 -0.1 -3 10


-2 -1 0 1

10

Frequency [Hz]

10

10

10

Figure XVIII: In- (black) and out-of-phase (red) functions of the batch FR experiment of an HZSM5 Wicke Kallenbach wafer. The fit was produced by applying a non-uniform infinite sheet model with surface resistance.

Appendix A: Results

A.3 Uptake Rate measurements A.3.1 H-ZSM5 powder


1.4 1.2 1
normalized mass [-]

experiment fit

0.8 0.6 0.4 0.2 0 0

50

100

time [s]

150

200

250

Figure XIX: Uptake curve of benzene in H-ZSM5 powder at a partial pressure of 0.1 mbar and

373K. The figure shows the experimental data (black) and the result of the curve fitting (red) applying an infinite plane sheet model of 1mm thickness.
1.4 1.2 1 experiment fit

normalized mass [-]

0.8 0.6 0.4 0.2 0 0

20

40

60

80

time [s]

100

120

140

160

180

200

Figure XX: Uptake curve of benzene in H-ZSM5 powder at a partial pressure of 1 mbar and 373K. The figure shows the experimental data (black) and the result of the curve fitting (red) applying an infinite plane sheet model of 1mm thickness.

Appendix A: Results

XI

1.4 1.2 1
normalized mass [-]

experiment fit

0.8 0.6 0.4 0.2 0 0

50

100

time [s]

150

200

250

Figure XXI: Uptake curve of benzene in H-ZSM5 powder at a partial pressure of 2 mbar and 373K. The figure shows the experimental data (black) and the result of the curve fitting (red) applying an infinite plane sheet model of 1mm thickness.
1.4 1.2 1
normalized mass [-]

experiment fit

0.8 0.6 0.4 0.2 0 0

100

200

300

400

time [s]

500

600

700

800

900

Figure XXII: Uptake curve of benzene in H-ZSM5 powder at a partial pressure of 10 mbar and

373K. The figure shows the experimental data (black) and the result of the curve fitting (red) applying an infinite plane sheet model of 1mm thickness.

Appendix A: Results

XII

A.3.2 H-ZSM5 Wicke Kallenbach wafer


1.4 1.2 1
normalized mass [-]

0.8 0.6 0.4 0.2 0 0

experiment fit

200

400

600

800

time [s]

1000

1200

1400

1600

1800

Figure XXIII: Uptake curve of benzene in an H-ZSM5 Wicke Kallenbach wafer of 100 mg at a partial pressure of 0.1 mbar and 373K. The figure shows the experimental data (black) and the result of the curve fitting (red) applying an infinite plane sheet model of 1mm thickness.
1.4 1.2 1
normalized mass [-]

experiment fit

0.8 0.6 0.4 0.2 0 0

200

400

600

time [s]

800

1000

1200

Figure XXIV: Uptake curve of benzene in an H-ZSM5 Wicke Kallenbach wafer of 100 mg at a partial pressure of 1 mbar and 373K. The figure shows the experimental data (black) and the result of the curve fitting (red) applying an infinite plane sheet model of 1mm thickness.

Appendix A: Results

XIII

1.4 1.2 1
normalized mass [-]

0.8 0.6 0.4 0.2 0 0 experiment fit

200

400

600

time [s]

800

1000

1200

1400

Figure XXV: Uptake curve of benzene in an H-ZSM5 Wicke Kallenbach wafer of 100 mg at a partial pressure of 2 mbar and 373K. The figure shows the experimental data (black) and the result of the curve fitting (red) applying an infinite plane sheet model of 1mm thickness.
1.4 1.2 1
normalized mass [-]

0.8 0.6 0.4 0.2 0 0 experiment fit

200

400

600

800

1000
time [s]

1200

1400

1600

1800

2000

Figure XXVI: Uptake curve of benzene in an H-ZSM5 Wicke Kallenbach wafer of 100 mg at a partial pressure of 10 mbar and 373K. The figure shows the experimental data (black) and the result of the curve fitting (red) applying an infinite plane sheet model of 1mm thickness.

Appendix A: Results

XIV

A.4

Nitrogen physisorption isotherms

A.4.1 H-ZSM5 powder


300 250 200 150 100 50 0 0,00E+00 2,00E-01 4,00E-01 6,00E-01 8,00E-01 1,00E+00 1,20E+00

Vads[cm3/g]

p/p0
Figure XXVII: Nitrogen physisorption isotherm of ZSM-5 parent powder material.

0,45 0,4 0,35

y = 0,0304x + 0,2787 R = 0,9908

vads [cm3/g STP]

0,3 0,25 0,2 0,15 0,1 0,05 0 0 1 2 3 4 5 y = 0,0542x + 0,1471 R = 0,9979

tHasley [nm]
Figure XXVIII: t-plot produced with the method of Hasley for H-ZSM5 powder.

Appendix A: Results

XV

9,00E-04 8,00E-04

(p/p0)/Vads*(1-p/p0) [g/cm3]

7,00E-04 6,00E-04 5,00E-04 4,00E-04 3,00E-04 2,00E-04 1,00E-04 0,00E+00

y = 0,0104x + 3E-06 R = 0,9999

0,00E+001,00E-022,00E-023,00E-024,00E-025,00E-026,00E-027,00E-028,00E-029,00E-02

p/p0 [-]

Figure XXIX: Linearization of the nitrogen physisorption isotherm in the low pressure range according to the BET theory of an H-ZSM5 powder sample.

A.4.2 H-ZSM5 Wicke Kallenbach wafer


350 300 250

Vads[cm3/g]

200 150 100 50 0 0,00E+00 2,00E-01 4,00E-01 6,00E-01 8,00E-01 1,00E+00 1,20E+00

p/p0 [-]
Figure XXX: Nitrogen physisorption isotherm of a H-ZSM5 Wicke Kallenbach wafer.

Appendix A: Results

XVI

0,5 0,45 0,4

y = 0,0095x + 0,3789 R = 0,5502

vads [cm3/g STP]

0,35 0,3 0,25 0,2 0,15 0,1 0,05 0 0 1 2 3 4 5 6 7 8 9 y = 0,048x + 0,147 R = 0,9966

tHasley [nm]
Figure XXXI: t-plot produced with the method of Hasley for a H-ZSM5 Wicke Kallenbach wafer.

0,0009 0,0008

(p/p0)/Vads*(1-p/p0) [g/cm3]

y = 0,0098x + 5E-08 R = 1

0,0007 0,0006 0,0005 0,0004 0,0003 0,0002 0,0001 0 0,00E+001,00E-022,00E-023,00E-024,00E-025,00E-026,00E-027,00E-028,00E-029,00E-02

p/p0 [-]

Figure XXXII: Linearization of the nitrogen physisorption isotherm in the low pressure range according to the BET theory of an H-ZSM5 Wicke Kallenbach wafer.

Appendix A: Results

XVII

A.5

Matlab source code

A.5.1 Processing of Wicke Kallenbach experimental data


function f_Trennfaktor_gegen_Zeit_und_TAbhaengigkeit [ filename dirpath ] = uigetfile ( '130'); [ MZ1 ] = textread ([ dirpath filename ]); t_MZ1 = MZ1 ( :, 1 ); Start1 = 2.17 .* ( 10 ^ 6 ); m1 = find ( t_MZ1 == Start1 ); t_MZ1 = MZ1 ( m1 : end, 1 ); I_MZB1 = MZ1 ( m1 : end, 3 ); I_MZX1 = MZ1 ( m1 : end, 4 ); c_MZB1 = 6936.1 .* I_MZB1; c_MZX1 = 10173 .* I_MZX1; c_MZX1 = c_MZX1 .* 1.44; S1 = c_MZB1 ./ ( c_MZX1 ); t_MZ1 = t_MZ1 - t_MZ1 ( 1 ); t_MZ1 = t_MZ1 ./ 60000; figure( 1 ) plot ( t_MZ1, S1, 'r+' ) xlabel ( 'time [min]' ), ylabel ( 'Seperation Factor 130 title ( 'Trend of Seperation Factor' ) figure( 2 ) semilogy ( t_MZ1, c_MZB1, 'ks', t_MZ1, c_MZX1, 'r+') xlabel ( 'time [min]' ), ylabel ( 'c [mol/l]' ), title ( ZSM5 130 C' ),legend ( 'Benzene', 'p-Xylene' ) [ filename dirpath ] = uigetfile ( '100'); [ MZ2 ] = textread ([ dirpath filename ]); t_MZ2 = MZ2 ( :, 1 ); Start2 = 2.93 .* ( 10 ^ 6 ); m2 = find ( t_MZ2 == Start2 ); t_MZ2 = MZ2 ( m2 : end, 1 ); I_MZB2 = MZ2 ( m2 : end, 3 ); I_MZX2 = MZ2 ( m2 : end, 4 ); c_MZB2 = 6936.1 .* I_MZB2; c_MZX2 = 10173 .* I_MZX2; c_MZX2 = c_MZX2 .* 1.44; S2 = c_MZB2 ./ ( c_MZX2 ); t_MZ2 = t_MZ2 - t_MZ2 ( 1 ); t_MZ2 = t_MZ2 ./ 60000; figure( 3 ) plot ( t_MZ2, S2, 'r+' ) xlabel ( 'time [min]' ), ylabel ( 'Seperation Factor 100 title ( 'Trend of Seperation Factor' ) figure( 4 ) semilogy ( t_MZ2, c_MZB2, 'ks', t_MZ2, c_MZX2, 'r+') xlabel ( 'time [min]' ), ylabel ( 'c [mol/l]' ), title ( ZSM5 100 C' ),legend ( 'Benzene', 'p-Xylene' ) [ filename dirpath ] = uigetfile ( '70'); [ MZ3 ] = textread ([ dirpath filename ]); t_MZ3 = MZ3 ( :, 1 ); Start3 = 2.26 .* ( 10 ^ 6 ); m3 = find ( t_MZ3 == Start3 ); t_MZ3 = MZ3 ( m3 : end, 1 ); I_MZB3 = MZ3 ( m3 : end, 3 ); I_MZX3 = MZ3 ( m3 : end, 4 ); c_MZB3 = 6936.1 .* I_MZB3; c_MZX3 = 10173 .* I_MZX3; c_MZX3 = c_MZX3 .* 1.44; S3 = c_MZB3 ./ ( c_MZX3 );

C H-ZSM5' ),

'Zeolite H-

C HZSM-5' ),

'Zeolite H-

Appendix A: Results
t_MZ3 = t_MZ3 - t_MZ3 ( 1 ); t_MZ3 = t_MZ3 ./ 60000; figure( 5 ) plot ( t_MZ3, S3, 'r+' ) xlabel ( 'time [min]' ), ylabel ( 'Seperation Factor 70 C HZSM-5' ), title ( 'Trend of Seperation Factor' ) figure( 6 ) semilogy ( t_MZ3, c_MZB3, 'ks', t_MZ3, c_MZX3, 'r+') figure( 7 ) xlabel ( 'time [min]' ), ylabel ( 'c [mol/l]' ), title ( 'Zeolite HZSM5 70 C' ),legend ( 'Benzene', 'p-Xylene' ) semilogy ( t_MZ1, c_MZB1, 'kx', t_MZ2, c_MZB2, 'ko',t_MZ3, c_MZB3, 'ks' ) xlabel ( 'time [min]' ), ylabel ( 'c [mol/l]' ), title ( 'Zeolite HZSM5 T - bhngigkeit Benzol' ), legend ( '130 C', '100 C', '70 C' ) figure( 8 ) semilogy ( t_MZ1, c_MZX1, 'rx', t_MZ2, c_MZX2, 'ro',t_MZ3, c_MZX3, 'rs' ) xlabel ( 'time [min]' ), ylabel ( 'c [mol/l]' ), title ( 'Zeolite HZSM5 T - bhngigkeit p-Xylol' ), legend ( '130 C', '100 C', '70 C' )

XVIII

Appendix B: Simulation

XIX

B. B.1

Simulation

Curve fitting of Wicke Kallenbach experiments

1.2 1 0.8
normalized concentration [-]

0.6 0.4 0.2 0 -0.2 0 fit Benzene measurement Benzene fit p-Xylene measurement Benzene

0.5

time [s]

1.5

2.5 4 x 10

Figure XXXIII: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 70C, 100mg wafer weight and 50 ml/min volume flow and the corresponding fit for benzene (black) and pxylene (blue).
1.2 1 0.8
normalized concentration [-]

0.6 0.4 0.2 0 -0.2 0

fit Benzene measurement Benzene fit p-Xylene measurement p-xylene

0.2

0.4

0.6

0.8

t [s]

1.2

1.4

1.6

1.8 x 10

2
4

Figure XXXIV: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 100C, 100mg wafer weight and 50 ml/min volume flow and the corresponding fit for benzene (black) and pxylene (blue).

Appendix B: Simulation

XX

1.2 1 0.8
normalized concentration [-]

0.6 0.4 0.2 0 -0.2 0

fit Benzene measurement Benzene fit p-Xylene measurement Benzene

0.2

0.4

0.6

0.8

time [s]

1.2

1.4

1.6

1.8 x 10

2
4

Figure XXXV: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 130C, 100mg wafer weight and 50 ml/min volume flow and the corresponding fit for benzene (black) and pxylene (blue).

1.2 1 0.8
normalized concentration [-]

0.6 0.4 0.2 0 -0.2 0 fit Benzene measurement Benzene fit p-Xylene measurement p-xylene

0.5

1.5

time [s]

2.5

3.5 x 10

4
4

Figure XXXVI: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 100C, 50mg wafer weight and 50 ml/min volume flow and the corresponding fit for benzene (black) and pxylene (blue).

Appendix B: Simulation

XXI

1.2 1 0.8
normalized concentration[-]

0.6 0.4 0.2 0 -0.2 0 fit Benzene measurement Benzene fit p-Xylene measurement p-xylene

0.5

1.5

time[min]

2.5

3.5

4.5 4 x 10

Figure XXXVII: Concentration vs. time diagram of the Wicke Kallenbach experiment using a pressed wafer of ZSM5 and a mixture of benzene (black) and p-xylene (red), 100C, 150mg wafer weight and 50 ml/min volume flow and the corresponding fit for benzene (black) and pxylene (blue).

Table I: Parameters obtained by non-linear curve fitting of the transmission curves of the Wicke Kallenbach experiments

Benzene T [K] 343 373 403 373 373 Weight [g] 100 100 100 50 150 k [1/s] 0.029 0.053 0.082 0.047 0.300 F [m3/s] 2.14x10-11 3.08x10-11 4.50x10-11 3.00x10-11 1.71x10-11

p-Xylene k [1/s] 0.013 0.012 0.014 0.014 0.010 F [m3/s] 1.1x10-11 9.94x10-12 2.00x10-11 2.45x10-11 7.54x10-12

Appendix B: Simulation

XXII

B.2 Matlab source code of Wicke Kallenbach curve fitting B.2.1 Main program
function [] = f_Wicke_Kallenbach_curve_fitting ( start ) % [ [ t In the first part the data is read and converted filename dirpath ] = uigetfile ( 'Xylol' ); Xylol ] = textread ([ dirpath filename ]); = Xylol ( : , 1 );

% Then the normalized concentration vs. time diagram is calculated from the % experimental data Start1 = start; n = find ( t == Start1 ); t = Xylol ( n : end, 1 ); I_Xylol = Xylol ( n : end, 3 ); I_Benzol = Xylol ( n : end, 4 ); C_Xylol = I_Xylol ./ max ( I_Xylol ); C_Benzol = I_Benzol ./ max ( I_Benzol ); t = t - t ( 1 ); t = t ./ 1000; % Afterwards the curve fitting is performed applying a genetic algorithm [ kB FB ] = f_Wicke_Kallenbach_ga_fit_breakthrough_curve_Benzene( t, C_Benzol ) [ kX FX ] = f_Wicke_Kallenbach_ga_fit_breakthrough_curve_xylene( t, C_Xylol ) %In the last step the solutions are plotted together with the experimental %data Wsat = 0.0744; rhoB = 1.44 * 10 .^ 3; S = 4.18 * ( 10 .^ -8 ); x = 10 ^ -3; c0B = 0.13; c0X = 9.77 .* ( 10 .^ -2 ); for n = 1 : 1 : length ( kB ) XB = (kB ( n ) TB = kB ( n ) .* Wsat .* rhoB .* S ./ ( FB ( n ))) .* x; .* c0B .* t;

CB ( :, n ) = exp ( TB - (0.9999 .* TB )) ./ ( exp ( TB - (0.9999 .* TB )) + exp ( XB - (0.9999 .* TB )) - exp ( - 0.9999 .* TB )); end for m = 1 : 1 : length ( kX ) XX = (kX ( m ) TX = kX ( m ) .* Wsat .* rhoB .* S ./ ( FX ( m ))) .* x; .* c0X .* t;

Appendix B: Simulation
CX ( :, m ) = exp ( TX - (0.9999 .* TX )) ./ ( exp ( TX - (0.9999 .* TX )) + exp ( XX - (0.9999 .* TX )) - exp ( - 0.9999 .* TX )); end for n = 1 : 1 : length ( kB ) for m = 1 : 1 : length ( kX ) x = kB ( n ); y = FB ( n ); z = kX ( m ); u = FX ( m ); i = 3 + ( n - 1 ) .* length ( kX ) + m; figure ( i ) plot ( t, CB ( :, n ), 'k+', t, C_Benzol, 'r-', t, CX ( :, m ), 'b*', t, C_Xylol, 'go' ) legend ( 'fit Benzene', 'measurement Benzene', 'fit p-Xylene', 'measurement p-xylene' ) text( 20000, 1,... ['kB = ',num2str(x)],... 'HorizontalAlignment','center') text( 20000, 0.75,... ['FB = ',num2str(y)],... 'HorizontalAlignment','center') text( 20000, 0.5,... ['kX = ',num2str(z)],... 'HorizontalAlignment','center') text( 20000, 1,... ['FX = ',num2str(u)],... 'HorizontalAlignment','center') end end

XXIII

B.2.2 Genetic algorithm


% This script implements the Simple Genetic Algorithm described % in the examples section of the GA Toolbox manual. % % Author: Andrew Chipperfield % History: 23-Mar-94 file created % % tested under MATLAB v6 by Alex Shenfield (22-Jan-03) function [ kB FB ] = f_Wicke_Kallenbach_ga_fit_breakthrough_curve_Benzene( t, CB_fit ) %imax = 3; GGAP = .9; created NVAR1 = 1; NVAR2 = 1; NIND = 40; MAXGEN = 3000; PRECI = 20; % Generation gap, how many new individuals are % Number of variables % Number of individuals % Number of generations % Precision of binary representation

% Build field descriptor FieldD1 = [rep([PRECI],[1, NVAR1]); rep([10^-5;1],[1, NVAR1]);... rep([1; 0; 1 ;1], [1, NVAR1])];

Appendix B: Simulation
FieldD2 = [rep([PRECI],[1, NVAR2]); rep([0.1 * 10^-10; 10^-5],[1, NVAR2]);... rep([1; 0; 1 ;1], [1, NVAR2])]; % Initialise population Chrom1 = crtbp(NIND, NVAR1*PRECI); Chrom2 = crtbp(NIND, NVAR2*PRECI); % Reset counters Best = NaN*ones(MAXGEN,1); % best in current population gen = 0; % generational counter % Evaluate initial population x = [bs2rv(Chrom1,FieldD1) bs2rv(Chrom2,FieldD2)]; [ObjV] = f_Wicke_Kallenbach_simulate_breakthrough_curve_Benzene_fit ( x, t, CB_fit ); ObjV = ObjV'; % Track best individual and display convergence Best(gen+1) = min(ObjV); figure(2) plot(log10(Best),'ro');xlabel('generation'); ylabel('log10(f(x))'); text(0.5,0.95,['Best = ', num2str(Best(gen+1))],'Units','normalized'); drawnow; % Generational loop while gen < MAXGEN, % Assign fitness-value to entire population FitnV = ranking(ObjV); % Select individuals for breeding SelCh1 = select('sus', Chrom1, FitnV, GGAP); SelCh2 = select('sus', Chrom2, FitnV, GGAP); % Recombine selected individuals (crossover) SelCh1 = recombin('xovsp',SelCh1,0.7); SelCh2 = recombin('xovsp',SelCh2,0.7); % Perform mutation on offspring SelCh1 = mut(SelCh1); SelCh2 = mut(SelCh2); % Evaluate offspring, call objective function x = [bs2rv(SelCh1,FieldD1) bs2rv(SelCh2,FieldD2)]; [ObjVSel] = f_Wicke_Kallenbach_simulate_breakthrough_curve_Benzene_fit ( x, t, CB_fit ); ObjVSel = ObjVSel'; % Reinsert offspring into current population [Chrom1 ObjV]=reins(Chrom1,SelCh1,1,1,ObjV,ObjVSel); [Chrom2 ObjV]=reins(Chrom2,SelCh2,1,1,ObjV,ObjVSel); % Increment generational counter gen = gen+1; % Update display and record current best individual Best(gen+1) = min(ObjV); figure(2)

XXIV

Appendix B: Simulation
plot(log10(Best),'ro'); xlabel('generation'); ylabel('log10(f(x))'); text(0.5,0.95,['Best = ', num2str(Best(gen+1))],'Units','normalized'); drawnow; end I = find(ObjV == (min(ObjV))); P1 = bs2rv(Chrom1,FieldD1); kB = P1 ( I ); P2 = bs2rv(Chrom2,FieldD2); FB = P2 ( I ); % End of GA

XXV

B.2.3 Fitness function


function [ObjV] = f_Wicke_Kallenbach_simulate_breakthrough_curve_Benzene_fit ( x, t, CB_fit ) ka = x ( : , 1 ); Wsat = 0.0744; rhoB = 1.44 * 10 .^ 3; S = 4.18 * ( 10 .^ -8 ); FB = x ( : , 2 ); x = 10 ^ -3; c0B = 0.13; for n = 1 : 1 : length (ka) XB = (ka ( n ) .* Wsat .* rhoB .* S ./ ( FB ( n ))) .* x; TB = ka ( n ) .* c0B .* t; CB = exp ( TB - ( TB .* 0.9999 )) ./ ( exp ( TB - ( TB .* 0.9999 )) + exp ( XB - ( TB .* 0.9999 )) - exp( - TB .* 0.9999 )); x = (( CB - CB_fit ) .^ 2 ); y = ( x ( 2 : end ) .* t ( 2 : end ) .^ 3 ) ./ FitnV ( n ) = sum( y ); end ObjV = FitnV;

( 2 );

Você também pode gostar