diffusion of alkanes in mfi-type zeolites · diffusion of alkanes in mfi-type zeolites citation for...

Diffusion of alkanes in MFI-type zeolites

Citation for published version (APA):Koriabkina, A. O. (2003). Diffusion of alkanes in MFI-type zeolites. Eindhoven: Technische UniversiteitEindhoven. https://doi.org/10.6100/IR565185

DOI:10.6100/IR565185

Document status and date:Published: 01/01/2003

Document Version:Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can beimportant differences between the submitted version and the official published version of record. Peopleinterested in the research are advised to contact the author for the final version of the publication, or visit theDOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and pagenumbers.Link to publication

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, pleasefollow below link for the End User Agreement:www.tue.nl/taverne

Take down policyIf you believe that this document breaches copyright please contact us at:[email protected] details and we will investigate your claim.

Download date: 29. Jun. 2020

https://doi.org/10.6100/IR565185

https://doi.org/10.6100/IR565185

https://research.tue.nl/en/publications/diffusion-of-alkanes-in-mfitype-zeolites(901a59a6-f8f8-463c-b3bc-6dc51cc2d284).html

Diffusion of alkanesin MFI-type zeolites

Diffusion of alkanesin MFI-type zeolites

P R O E F S C H R I F T

ter verkrijging van

de graad van doctor aan de

Technische Universiteit Eindhoven, op gezag van de

Rector Magnificus prof.dr. R.A. van Santen, voor een

commissie aangewezen door het College voor Promoties

in het openbaar te verdedigen

op donderdag 26 juni 2003 om 13.00 uur

door

Alina Olegovna Koriabkina

geboren te Novosibirsk, Rusland

Dit proefschrift is goedgekeurd door de promotoren:

prof.dr. R.A. van Santenenprof.dr. B. Smit

Copromotor:dr.ir. A.M. de Jong

CIP-DATA LIBRARY TECHNISCHE UNIVERSITEIT EINDHOVEN

Koriabkina, Alina O.

Diffusion of alkanes in MFI-type zeolites / by Alina O. Koriabkina. -Eindhoven: Technische Universiteit Eindhoven, 2003.Proefschrift. - ISBN 90-386-3054-9NUR 913Trefwoorden: heterogene katalyse ; zeolieten / poreuze materialen ;diffusie / adsorptie / radiochemie ; positron-emissie / alkanen / fysisch-chemische modelleringSubject headings: heterogeneous catalysis ; zeolites / porous materials ;diffusion / adsorption / radiochemistry ; positron-emission / alkanes /physicochemical modeling

c© 2003 by Alina O. Koriabkina

The work described in this thesis has been carried out at the Schuit Institute ofCatalysis (part NIOK, the Netherlands School for Catalysis Research), EindhovenUniversity of Technology, The Netherlands.

To my parents

Contents

1 Introduction 1

1.1 MFI-type zeolites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Methods of investigation . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.2.1 Theoretical methods . . . . . . . . . . . . . . . . . . . . . . . 5

1.2.2 Experimental methods . . . . . . . . . . . . . . . . . . . . . . 6

1.3 Theoretical models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.3.1 Jump model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.3.2 Continuum thermodynamics model . . . . . . . . . . . . . . . 11

1.3.3 Maxwell-Stefan model . . . . . . . . . . . . . . . . . . . . . . 11

1.4 Scope of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2 Experimental method 17

2.1 Positron Emission Profiling . . . . . . . . . . . . . . . . . . . . . . . 17

2.2 Experimental procedures . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.2.1 Zeolite samples . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2.3 Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2.4 Direct measurement of the zeolite loading . . . . . . . . . . . . . . . . 30

3 Binary mixture diffusion 35

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.2 Experimental section . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

3.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.3.1 Adsorption of Single Components . . . . . . . . . . . . . . . . 37

3.3.2 Binary Adsorption . . . . . . . . . . . . . . . . . . . . . . . . 39

3.3.3 Multi-component Diffusion . . . . . . . . . . . . . . . . . . . . 41

3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4 Influence of the acid sites on diffusion 47

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.2 Experimental section . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

4.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 48

4.3.1 Adsorption and diffusion of binary mixtures in HZSM-5 andsilicalite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.3.2 Adsorptive and diffusive properties of single components inHZSM-5 and silicalite . . . . . . . . . . . . . . . . . . . . . . . 54

4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

viii Contents

5 Diffusion of 3-methylpentane in silicalite 635.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635.2 Experimental section . . . . . . . . . . . . . . . . . . . . . . . . . . . 645.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 66

5.3.1 Influence of the concentration on the self-diffusivity . . . . . . 665.3.2 Activation energy of diffusion . . . . . . . . . . . . . . . . . . 695.3.3 Concentration dependence of the self-diffusivity . . . . . . . . 73

5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

6 Factors enhancing diffusion of n-hexane 816.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 816.2 Experimental section . . . . . . . . . . . . . . . . . . . . . . . . . . . 826.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 82

6.3.1 Influence of the concentration on the self-diffusivity . . . . . . 826.3.2 Activation energy of diffusion . . . . . . . . . . . . . . . . . . 89

6.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

7 Linear alkanes and their mixtures in silicalite 957.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 957.2 Experimental section . . . . . . . . . . . . . . . . . . . . . . . . . . . 967.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 97

7.3.1 Influence of the concentration on the self-diffusivity . . . . . . 977.3.2 n-Pentane/n-hexane mixtures . . . . . . . . . . . . . . . . . . 103

7.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

Summary 109

Samenvatting 113

Acknowledgements 117

List of publications 119

Curriculum Vitae 121

1 Introduction

Zeolites are defined as crystalline inorganic polymers based on a framework of XO4

tetrahedra linked to each other by sharing of oxygen ions, where X may be trivalent(e.g. Al, B, Ga, etc.), tetravalent (e.g. Si, Ge, etc.) or pentavalent (P), [1]. Theprotons that are required to maintain electrical neutrality in the structure, providean acidic function of the zeolite. The crystal structure of a zeolite is determined bya specific order in which a network of tetrahedral units are linked together. The sizeof the zeolite pore openings is determined by the number of oxygen atoms or XO4

units required to form a pore, the shape of the pore and the nature and amount ofthe cations that are present in or at the mouth of the pore. Over the past 50 yearsa lot of new synthetic zeolites have been discovered. Last edition of ”Atlas of zeoliteframework types” contains 136 entries, [2].

Zeolites are used for ion exchange applications as detergents, for adsorption (dry-ing, separation of hydrocarbons) and as catalysts. The most of zeolitic catalysts areused in oil refining and gas-conversion processes, such as the conversion and upgradingof the various oil fractions into transportation fuels, methanol-to-gasoline, conversionof syngas, light paraffins and olefins into gasoline and gasoil, [3]. Huge interest inzeolites is caused by their unique properties that have provided them a vast niche inthe industry:

• high thermal and hydrothermal stability in the industrial environment, espe-cially highly siliceous zeolites.

• presence of ion exchangeability allows the formation of highly dispersed cat-alytically active sites, such as highly acidic sites.

• presence of micropores allow a high concentration of reactant molecules.

• uniform pores, with the openings of similar dimensions as hydrocarbons.

Because the pores of the zeolites are similar to many organic molecules of practicalinterest, it became possible to design novel catalysts, which control the progress ofthe reactants and products through them via selecting the molecules by their size andshape, so called shape selective catalysts. There are three types of shape selectivity:reactant, product and transition state selectivity. Reactant selectivity occurs whenthe feedstock contains two classes of molecules, one of which is too large to passthrough the pores of the zeolite. Product shape selectivity occurs when some of thereaction products can not escape from the zeolite due to their size or shape. Thenthese molecules can undergo a secondary reaction to form molecules that are ableto leave the catalyst. Transition state selectivity is observed when both the reactantand product molecules fit well within the pores of the zeolite, while the reaction

2 Chapter 1: Introduction

Figure 1-1. Schematic drawing of MFI-type zeolite structure.

intermediates are larger than them and are spatially constrained either by their sizeor orientation.

Discovery of ZSM-5 and other medium pore zeolites significantly extended indus-trial application of shape selective catalysis. Transition state selectivity is one of themost important properties of these zeolites, which makes them significantly betterthan other catalysts in a number of important petrochemical processes. These zeo-lites can selectively convert linear, branched alkanes and aromatics with moleculardimensions less than 6 A. Moreover, a new type of diffusion, so called, micropore(or configurational) diffusion, started with these medium pore materials. The therm”configurational diffusion” was introduced by Weisz in 1973, [4]. It refers to the masstransport in the pores of molecular dimension, which is strongly influenced by inter-actions between diffusing molecules and zeolite walls.

1.1 MFI-type zeolites

MFI-type zeolites are the 10-membered oxygen ring (10 oxygen atoms are connectedto form a pore) systems and consist of bidirectional intersecting straight and zigzagchannels connected via intersections. In Fig. 1-1 a schematic presentation of the zeoliteis shown.

The size of straight and zigzag channel openings is 5.3 A×5.6 A and 5.1× 5.5 A,respectively, [2]. The unit cell of the zeolite (size of 20.06 A×19.8 A×13.36 A) contains

Section 1.1: MFI-type zeolites 3

Figure 1-2. The channel system of MFI-type zeolite, showing the channels in the crystal.

2 straight channels and 4 zigzag channels with 4 intersections between them. Thelength of the straight sections is equal to 19.8 A, while zigzag channels are of 6.65 A insize (see Fig. 1-2) and the diameter of the intersections is 5.4 A, [5].

Silicalite is an all-silica zeolite, which means that it consists of SiO4 units. HZSM-5zeolite has an iso-structural framework, in which some of the Si atoms are substitutedwith Al. It possesses acidic properties because of the protons that provide electricalneutrality of the zeolite. These two materials belong to the most applied syntheticzeolites. They are widely used in the petroleum and petrochemical industries as cata-lysts for fluid catalytic cracking (FCC), xylene isomerization, conversion of methanolto gasoline, and as selective sorbents, [6]. Unlike other zeolites, these materials havepores of uniform dimension and do not have any large supercages or bottlenecks,which is a significant factor for the shape selective catalyst as well as for their verylow coke forming abilities as acidic catalysts.

One-dimensional mordenite is a commercially applied catalyst for the hydroiso-merization of C5 − C6 alkanes to increase the gasoline octane number, but HZSM-5zeolite suffers much less from coking due to its structure so it has also been tested inthis reaction, [7]. The disadvantage is that HZSM-5 zeolite shows higher selectivityto monobranched alkanes with lower octane number compared to dibranched alkanesformed in mordenite.

Silicalite provides numerous possibilities for the hydrocarbon separation and itis the most studied membrane material, [8]. Recently, a lot of mainly theoreticalstudies have been conducted on the application of silicalite in the selective separationof mixtures of C5 − C6 alkane isomers, [9, 10, 11, 12], which are the products of thehydroisomerization process, mentioned above. The base for the separation possibilitiesis in the specific adsorption and related micropore diffusion properties of silicalite


towards alkanes.Apparently, commercial interest in silicalite and HZSM-5 zeolites caused a vast

number of scientific investigations of diffusive and adsorptive properties of C5− andC6−alkanes in these materials. The shape selective and molecular sieve propertiesdisplayed by MFI-type zeolites are directly related to the difference in the diffusiveproperties of the hydrocarbons in the pores. In order to be able to control the reactionor separation processes the influence of the process conditions on the diffusion shouldbe known. Obviously, temperature, concentration of the molecules and interactionwith the acid sites are important factors. In separation or reaction processes there areat least two types of molecules present in the zeolite pores, which affect the diffusionand adsorption of each other. Therefore, the research of properties of hydrocarbonsin these materials is focused on the following vital topics:

• concentration dependence of the diffusivity and its origins.

• diffusion of mixtures of alkanes.

• processes in the presence of acid sites.

• temperature dependence of the diffusion coefficient.

Below various methods developed to study zeolites will be discussed.

1.2 Methods of investigation

Modern knowledge on the diffusion and adsorption in zeolites is a result of boththeoretical and experimental studies. Different types of diffusion coefficients are pro-vided using different techniques. Transport or Ficks diffusivity is determined undernon-equilibrium conditions when there is a gradient of molecular concentration in thezeolite. From theoretical methods and some experimental methods, the self-diffusivityis obtained, which is a diffusion coefficient determined under equilibrium conditions.

Various experimental techniques have been developed to measure the diffusivityof organic molecules in zeolites. Unfortunately, each method has its own practicallimitations and the results from these different methods are not always consistent.Recently, theoretical simulations of adsorbed molecules in zeolites and modelling ofdiffusion processes became a very popular method of investigation due to the fastdevelopment of computers.

Micropore (configurational) diffusion of the molecules in zeolites occurs in theadsorbed state. Historically, diffusion and adsorption in zeolites have been investi-gated separately. As computers became more powerful, theoretical studies providedclear conclusion that these two processes are related to each other. Just a decade agoexperimental studies showed that in zeolites there are five types of dependencies ofthe self-diffusion coefficient on the pore occupancy, [13]. It encouraged further theo-retical research to provide an explanation for the experimental observations and to

Section 1.2: Methods of investigation 5

make some predictions. Lately a number of computer simulations have been dedi-cated to the study of the diffusion and adsorption of multi-component mixtures ofhydrocarbons. Unfortunately, due to limited experimental possibilities, these resultsstill remain to be checked.

Further, the most common experimental and theoretical approaches to the ofstudy the diffusion in zeolites will be discussed.

1.2.1 Theoretical methods

Often Monte-Carlo and molecular dynamic simulation (MD) methods are applied forthe theoretical investigations of self-diffusion in zeolites.

Monte-Carlo simulations deal with systems that are in equilibrium and there isno explicit time scale involved in the simulations. Successive configurations evolvein dimensionless time. This approach provides an information on the influence ofvarious factors on the behavior of diffusivity. There are two types of Monte-Carloschemes that have been recently developed for the adsorption and diffusion studies:configurational-bias (CBMC), [14], and kinetic Monte-Carlo, respectively. CBMC al-lows to avoid limitations of conventional Monte-Carlo methods. In the usual scheme,large molecules are inserted into the pores of a zeolite, which results in a high prob-ability of overlap with the zeolite atoms. Moreover, this probability significantly in-creases with the alkane chain length. With the CBMC approach, instead of placingthe molecules in zeolite pores at ones, the molecules are grown atom by atom suchthat overlap with the zeolite framework is avoided. The simulations are used for thecalculations of adsorption isotherms in zeolites, moreover they provide informationon the siting of the molecules in the zeolite framework and their conformations. Itwas shown that size and shape of the molecule play an important role, so that somemolecules are mainly located in the channels, others in the intersections and someare randomly distributed in the zeolite, [15]. It causes some unusual behavior in theadsorption isotherms observed experimentally, [16].

In the kinetic Monte-Carlo approach the zeolite lattice is considered to consistof grid sites connected by bonds, where molecules occupy a certain fraction and hopfrom one site to the other. Statistics on the molecular displacement is collected anddiffusivity is calculated from the mean-square displacement according to the Einsteinequation:

〈∆−→r 2(t)〉 = zDst, (1.2.1)

where Ds is a self-diffusion coefficient, 〈∆−→r 2(t)〉 is a mean-square displacement duringthe observation time t, z is a coefficient that is determined by the space dimension.

On the contrary, molecular dynamics (MD) simulations are capable of predictingthe absolute diffusivities in zeolites using an explicit time scale. Molecular dynamicsmodels attempt to predict a priori the diffusivities in zeolites through considerationof fundamental relationships between the energetics and structure of zeolite-sorbatesystem. The approach is to mimic the system as good as possible. The basis for themethod is the numerical solution of Newton’s equation of motion for a given number


of particles in a specified volume with a fixed total energy. This technique requireslarge quantities of computational time. Because of the complexity of the system, mostof the studies have developed models for zeolite and a single sorbate only. Because ofthe requirements on CPU time, the application of this method is limited to shorterchain length molecules. Moreover, MD is not capable of simulating the isotherms ofadsorption for similar reasons, therefore, CBMC simulations are used.

Theoretical investigations are necessary as they give an inside view on the exper-imentally observed ”external” situations.

1.2.2 Experimental methods

There are two types of experimental methods to measure diffusion in zeolites, socalled, micro- and macroscopic techniques. The experimental procedure determinesthe type of diffusion coefficients measured: transport or self-diffusivity. Generally,macroscopic methods determine transport diffusivity. From microscopic methods,usually, the self-diffusivity is obtained, which is a diffusion coefficient under equi-librium conditions. Below, the most commonly applied methods will be discussed.

Microscopic techniques

Microscopic techniques such as Pulsed Field Gradient NMR (PFG-NMR) and Quasi-Elastic Neutron Scattering (QENS) provide information on the self-diffusion coeffi-cient. PFG-NMR is only applied to sorbate molecules containing unpaired nuclearspins, such as hydrocarbons. Self-diffusivity is derived from the Einstein equationfrom measured mean-square displacement of the molecules. The measurement is per-formed in a sealed glass tube containing a zeolite sample in the equilibrium with thesorbate molecules. The sample is placed in a constant magnetic field. The spin-echosignal, perpendicular to the constant magnetic field is applied by radio frequencypulses with the time interval between the pulses shorter than the spin-relaxationtime T2. In addition to these pulses, two successive pulses of a magnetic gradientfield, which increases linearly with the space coordinate, are superimposed in shorttime intervals with a certain duration. The self-diffusivity is related to the ratio ofthe intensity spin-echo signal in the presence and in the absence of the gradient field,the time interval between the pulses and duration of each pulse. PFG NMR, [13], isonly applicable to the measurement of fast diffusing molecules with the diffusivityrange of 10−9 − 10−11m2/s.

Using QENS, [17], the self-diffusion in zeolites is measured from the broadening ofthe elastic scattering peak (broadening in the energy of the scattered neutron beam).The broadening occurs due to the energy transfer between the incident wave anddiffusing molecules and depends on the diffusion of the molecules. This allows tocalculate self-diffusivity. The method covers the range of the diffusivities similar tothat of PFG NMR.

There is another microscopic method, Interference microscopy, which providesthe information on the transport diffusion coefficient. The method is based on the


time-and-space-resolved measurement of the change in optical density of the zeolitecrystal during adosrption/desorption processes, [18]. Another microscopic method,single crystal IR method, [19], has been recently applied for study of diffusion ofhydrocarbons. It is based on the measurement of evolution of concentration profilesduring desorption by monitoring the change in the IR absorbance. It allows to dis-tinguish the transport processes in different directions in the zeolite crystal.

The main advantage of PFG-NMR and QENS is that these techniques measureself-diffusion coefficient from the molecular mobility directly, and they can be com-pared with theoretical simulations. Even though it is possible to investigate multi-component mixture diffusion, the disadvantage of these techniques is that the mea-surements can be performed at temperatures lower compared to the real life processesand are limited to fast molecules.

Macroscopic techniques

Macroscopic methods are divided by the type of the experimental conditions intosteady-state and transient methods.

Steady-state methods include the reaction kinetics, Wicke-Kallenbach permeationmethods and a novel steady-state single crystal membrane technique (SCM). Thereaction kinetics method allows to measure the zeolitic diffusivity indirectly in acatalytic system. By performing the reaction experiments on the zeolite crystals ofdifferent size, one can measure the reaction kinetics in the kinetic regime and inthe intracrystalline diffusion regime. From the relationship between the effectivenessfactor η, which is a degree of using the catalyst surface and the Thiele modulusφ, which is a parameter related to the effective diffusivity, one can estimate thediffusivity. The obtained diffusion coefficient is an effective parameter and it can notbe directly related to the self-diffusivity.

In Wicke-Kallenbach permeation, [20], the steady-state flux is measured acrossthe zeolite membrane between two gas chambers with different sorbate concentra-tions until the rate of permeation across the membrane will reach the steady-statedue to adsorption by the zeolite. The transport diffusivity is calculated from the re-lationship between the quantity of the component passed through the membrane, Q,during the time, t, and the diffusivity. The method is suitable for single componentstudies as well as for the mixture separation. Large zeolite crystals are required toensure an intracrystalline diffusion as a dominant process. In the transient membranepermeation technique, [34], the flux of the adsorbate through the zeolite membranewith the constant pressure gradient applied.

A newly developed SCM technique measures the diffusive flux through a single-crystal membrane under steady-state conditions, [21]. It should be noted that thismethod provides a directional transport diffusivity. When the pores of the investi-gated zeolite are anisotropic, the results of this method can not be directly comparedto other macroscopic methods that provide the average diffusivity. The importantadvantage of SCM is that it has a very wide range of applicability, from 10−6 to10−15m2/s.


Transient techniques are more popular for studying diffusion in zeolites. They in-clude uptake methods (gravimetric, volumetric), chromatographic, such as FrequencyResponse (FR), zero-length column (ZLC), and Membrane permeation techniques.

Uptake methods are based on the monitoring of the rate of increase or decreaseof sample weight (gravimetric technique) or the uptake rate with a sensitive pressuretransducer (volumetric technique) after making a step change in the external gasphase concentration or pressure. The transport diffusivity is calculated from the so-lution of the equations describing the mass transport, which results in the change ofthe sample weight. The diffusivity range that can be measured with this equipment islimited to 10−13−10−20m2/s. Recently a new uptake method, tapered element oscillat-ing microbalance (TEOM, [22, 23]), has been applied for the diffusion and adsorptionmeasurements of hydrocarbons in zeolites. This method is based on the inertial forcesto measure the mass change in the zeolite sample by changing the frequency of theoscillating tube holding the sample. The corrected diffusivities of 10−12 − 10−15m2/sdetermined from this technique are in a good agreement with those determined bychromatographic methods, that will be described below. The disadvantage of the up-take methods is that they are not capable of studying diffusion in mixtures. This isbecause it is not possible to distinguish different components during the experiment.

In the ZLC method the zeolite sample is equilibrated at very low sorbate partialpressure, then the sample is desorbed by a purge gas and the sorbate concentrationin the purge gas is followed chromatographically, [24]. Diffusivities are obtained byfitting the chromatographic response from the zeolite column with an appropriatemathematical model of the mass transfer. This technique provides the transport dif-fusion coefficient with a possible measured range similar to that of uptake methods.

FR allows to obtain the transient uptake diffusivity via monitoring small and fastchanges in the sorbate pressure, which initiates the adsorption/desorption process inthe zeolite sample recorded by a sensitive pressure transducer. The method is ableto distinguish diffusion processes in different zeolite channels, but it also requiresan appropriate mathematical description in order to obtain the transport diffusioncoefficients. Large zeolite crystals are needed to provide the intracrystalline diffusionlimitations, and an possibility to observe different responses from the diffusion indifferent channels. Shen and Rees observed two diffusion processes for C4−C6 alkanesin silicalite, that correspond to the straight and zigzag channels, respectively, [25].

Recently, a new chromatographic method has been developed, which uses a Mul-titrack apparatus, [26]. The method is based on the measuring of the spreading ofan injected very small pulse of studied molecules in the column containing a zeolitesample under ultra-high vacuum conditions. The pulse response is measured using ahigh speed sampling mass spectrometer. The advantage of this approach is that itallows simultaneous measurement of diffusion and the rate constants of adsorptionand desorption. Moreover, the obtained diffusion coefficient can be considered as theself-diffusion coefficient. This is because the concentration of the molecules is verysmall and there is no carrier gas present, so both do not influence the diffusion. Thedisadvantage of such approach is that it only allows to study the influence of the tem-


perature on the diffusion. With this method, as well as with other chromatographictechniques (FR, ZLC) diffusion in multi-component mixtures in zeolites can not beinvestigated.

The common disadvantage of macroscopic methods is that a direct measurementof the diffusion coefficient is not possible. To extract the diffusion coefficient, a modeldescribing the mass transfer is used. In general, the diffusivity is determined from themacroscopic data using the solution of the second Fick’s low:

∂c

∂t= D

∂2c

∂x2, (1.2.2)

where c is a concentration, D is a diffusion coefficient. Eq. (1.2.2) assumes that masstransport occurs via diffusion and can be adopted for any geometry of the zeolitecrystal. When there are other processes involved in the mass transport, they shouldbe included in this equation. In some cases a numerical solution of the model isrequired. Fitting the model equation to the experimental data will provide a diffusioncoefficient. A similar approach is used in this thesis to extract the diffusion coefficientand will be discussed later.

Another problem with macroscopic techniques is that in order to interpret theresults on the microscopic scale and to compare it with theoretical data, transportdiffusivity has to be converted into the value comparable to the self-diffusivity. Astransport diffusivity is measured in the presence of the concentration gradient inzeolite, a correction on the concentration is made. The obtained diffusivity is calledthe corrected or jump diffusivity. This diffusion coefficient is treated as one that isrelated to the diffusion process on the microscopic scale. Darken’s relation has beenapplied to obtain the corrected diffusivity:

Dt(θ) = Dc∂ln p

∂ln c. (1.2.3)

Here θ is the zeolite pore occupancy, c is the adsorbed concentration at partial pres-sure p, Dt(θ), Dc are transport and corrected diffusivities, respectively. Even thoughcorrected diffusivity is quite often directly compared with the self-diffusion coefficient,it is not really correct unless Dc is determined at θ = 0. A relationship similar tothe Darken’s one has been proposed, [27], to link self-diffusion Ds to the transportdiffusivity:

Dt(θ) =1

1− γ(c)pDs

∂ln p

∂ln c. (1.2.4)

In Eq. (1.2.4), γ(c) is some factor of the zeolite occupancy. Physical meaning of theequation is that at high loadings a hindrance of the molecules decreases the jumprate. At low partial pressure Eq. (1.2.4) becomes equal to Darkens relation. Recently,Krishna and Paschek proved that corrected and self-diffusion coefficients are relatedby the following expression, [28]:

Ds(θ) =Dc

1 + θ. (1.2.5)


As it was mentioned earlier, there are four major directions in the research of thediffusion of hydrocarbons in MFI-zeolite, such as influence of the temperature, con-centration of the adsorbed species, study of multi-component mixtures and influenceof the presence of acid sites in the zeolite. In principle all the methods described aboveare suitable for studying the temperature dependence and influence of the interac-tions with the Brønsted sites and zeolite loading on the diffusion, while only a fewmethods, including microscopic methods PFG NMR and QENS have been applied forthe measurement of the diffusion in binary mixtures. Usually, macroscopic techniquesare not capable to measure diffusion coefficients of the components in the mixturessince they are not able to distinguish the molecules. Many of the discussed methodsare not capable to determine the zeolite loading simultaneously with the diffusivity,thus these method require additional equipment to measure the concentration of themolecules in zeolite pores.

Positron Emission Profiling (PEP) based on the use of radio-labelled moleculesis a unique macroscopic method, which allows to measure self-diffusion of alkanes inzeolites in situ. The greatest advantage of the method compared to the others is thepossibility to study multi-component mixtures under reaction conditions, which is notpossible with microscopic methods such as PFG NMR or QENS. Often, possibilities ofthese techniques for studying mixtures are limited to the small molecules, or in somecases, mixtures of isomers can not be investigated. All these limitations are avoidedwith the PEP technique, which makes the method really exclusive. Moreover, tracer-exchange experiments (TEX-PEP) allow simultaneous measurement of diffusion andadsorption of the components in multi-component mixtures. Thus, this techniqueopens a new opportunity to investigate the most vital issues on zeolitic diffusion.

1.3 Theoretical models

To be able to provide an explanation for the experimental observations from themicroscopic point of view or to perform computer simulations a theoretical model isrequired to describe diffusion in zeolite pores. There are several theoretical modelsto describe adsorption and diffusion in zeolites, the most common are: jump model,continuum thermodynamics model and Maxwell-Stefan models.

1.3.1 Jump model

In the jump model the zeolite crystals are presented as a network of interconnectedpores. Diffusion occurs via a sequence of jumps from one adsorption site to the other.Adsorption site is the place within the crystal where the potential energy of themolecule is minimal. Neighboring sites are separated by high energy barriers. Locationof the adsorption sites and direction of the jumps are related to the zeolite topologyand the interactions between the molecules. The most popular methods to simulatethis model use Monte-Carlo methods. The self-diffusion coefficient is derived fromthe mean-square displacement of the molecules using the Einstein equation.

Section 1.3: Theoretical models 11

A lot of diffusion simulations have been performed using this model for variouszeolites to study the dependence of the diffusion coefficient on the pore occupancy.Using this model, the influence of various factors, such as presence of the acid sites,attractive and repulsive interactions between the particles on the diffusivity undervarious zeolite occupancies have been performed, [29]. Introduction of such inter-actions in the diffusion model might provide an explanation of different types ofconcentration dependencies of the self-diffusion in zeolites that have been measuredusing PFG-NMR technique, [13].

There is also an analytical way to calculate self-diffusivity in terms of a jumpmodel using the mean-field theory. The disadvantage of the mean-field approach isthat it does not take into account correlation effects. At the pore occupancies θ > 0,a molecule has a tendency to return to the site it has visited before, so that successivesteps are correlated, therefore, concentration at various sites of the pore lattice arein general different and mutually correlated as it was shown for silicalite, [30]. Mean-field approximation for the jump model assumes that the occupancy of all sites ofthe same type is the same. In order to estimate the diffusivity one has to know themean jump length a and the mean residence time τ spent on the adsorption site. Forzeolite NaY it was shown that diffusivity can be calculated as following, [31]:

Ds =1

6· a2

τ. (1.3.1)

Of course, one should make some estimates of the residence time and mean jumplength for each type of zeolite.

1.3.2 Continuum thermodynamics model

According to the continuum thermodynamics model the molecular jump probabilityfrom one adsorption site to the other is described by transition state theory. Passingof the molecules between the adsorption sites is considered as a rate process with anactivated transition state. The migration of the molecules is caused by a sequence ofrandom discrete jumps between the neighboring sites with a minimum energy neededfor each jump. The intrazeolite space should be represented by an energy profile, sothat molecules pass through a less favorable transition state when they are movingfrom one adsorption site to another. In order to calculate the diffusivity one has toknow the information on changes in degrees of freedom between the molecules in theadsorbed phase and those in the transition state. Using this model, studies on theconcentration dependence of the diffusion have been performed, [32].

1.3.3 Maxwell-Stefan model

Maxwell-Stefan theory applied for diffusion is capable of predicting an informationon diffusion in multi-component mixtures and concentration dependencies, [33]. Itis a macroscopic method based on the fact that the driving force for the diffusion


is the gradient of the chemical potential of the adsorbed species. This approach al-lows to predict the character of diffusion on the macroscopic scale and provides acorrected diffusion coefficient or jump diffusivity (see Section 1.2). Jump diffusivitycan be related to the self-diffusion via Eq. (1.2.5). This model is very useful for theexperimental studies, especially those obtaining the corrected diffusivities, [34]. Thedisadvantage of the model is that it does not take into account the zeolite topology.

Existence of these models helps to explain or predict experimental observations. Insome cases only one model can be applied due to the technical limitations of the com-puter simulations. For other zeolite/adsorbate systems there are several theoreticalpredictions available, which makes it interesting to compare them with the experi-mental results. A combination of the experimental and theoretical approaches to thestudy of adsorption and diffusion in zeolites may provide a complete understandingabout these processes.

1.4 Scope of this thesis

This thesis focuses on the experimental study of self-diffusion and adsorption of C5−C6 alkanes in MFI-type zeolites, i.e. silicalite and HZSM-5. The principles of thetechnique Positron Emission Profiling (PEP), which has been applied are describedin Chapter 2 together with the mathematical tools needed for the interpretation ofthe experimental data. An introduction with a short overview of the literature dataon the particular topic is given in each chapter.

PEP enabled the measurement of the diffusion of C6 isomers in their mixturesin zeolite silicalite, which is a vital topic of many theoretical researches, [11]. InChapter 3, the influence of the mixture composition on the component’s diffusivitiesand adsorptive properties for 2-methylpentane/n-hexane mixtures is discussed. Acomparison with the results obtained from Monte-Carlo simulations on the adsorptionof single components and their mixtures is involved in the result interpretation. Forthe mixtures of the alkanes it is shown that siting of the molecules in a silicaliteframework plays a crucial role in their behavior.

As mentioned in Section 1.1, HZSM-5 zeolite is used as a catalyst for hydrocarbonconversion processes, therefore diffusion and adsorption of mixtures of alkanes wasstudied in this zeolite and compared with non-acidic silicalite. The results of thisinvestigation will be discussed in Chapter 4.

Another important issue in zeolitic diffusion is its dependence on the zeolite poreoccupancy. A number of theoretical studies have been performed and some theoriesand predictions came out as a result, [30, 32]. In Chapters 5 and 6, experimentalobservations on the concentration dependence of monobranched and linear hexanes,respectively, is described. Chapter 5 deals with 3-methylpentane diffusion in silicalite,and a check of various theoretical predictions on the concentration dependence of theself-diffusion is made.

Chapter 6 is dedicated to the n-hexane diffusion in both HZSM-5 and silicalite atvarious pore occupancies. The behavior of linear alkane is different from that of iso-

Section 1.4: Scope of this thesis 13

hexane due to the different siting of the molecules. For n-hexane it causes additionalinteractions discussed in this Chapter that influence molecular motion.

Chapter 7 consists of two parts: one discusses the study of C5 n-alkane diffusionin silicalite, while the second part is dedicated to the diffusion of mixtures of n-pentane/n-hexane in silicalite with the various ratios between the components. Inthe first part, the influence of the zeolite pore occupancy was studied at varioustemperatures, and the reasons for the observed behavior are discussed.

References

[1] Liebau, F. Zeolites 3 (1983) 191.

[2] Baerlocher, Ch.; Meier, W.M.; Olson, D.H. Atlas of zeolite framework types, Fifthedition, Elsevier, 2001.

[3] Sie, S.T. Past, Present and future role of microporous Catalysts in the PetroliumIndustry, in: Advanced Zeolite Science and Application Stud. Surf. Sci. 85 (1994)578.

[4] Weisz, P.B. Chem.Tech. 3 (1973) 498.

[5] Kokotailo, G.T.; Meier, W.M. in R.P. Townsend, ed., Properties and Applicationsof Zeolites, Special Publ. 33, 1979, p.133.

[6] Naber, J.E.; de Jong, K.P.; Stork, W.H.J.; Kuipers, H.P.C.E.; Post, M.F.M.Stud. Surf. Sci. 84 (1994) 2197-2219.

[7] de Gauw, F.J.M.M.; van Grondelle, J.; van Santen, R.A. J. of Catal. 206(2)(2002) 295-304

[8] Karger, J.; Ruthven, D.M. Diffusion in zeolites and other microporous solids,John Wiley & Sons, New York, 1992.

[9] Funke, H.H.; Argo, A.M.; Falconer, J.L.; Noble, R.D. Ind. Eng. Chem. Res. 36(1997) 137.

[10] Krishna, R.; Smit, B.; Vlugt, T.J.H. J. Phys.Chem. A 102 (1998) 7727-7730.

[11] Schenk, M.; Vidal, S.L.; Vlugt, T.J.H.; Smit, B.; Krishna, R. Langmuir 17 (2001)1558-1570.

[12] Calero, S.; Smit, B.; Krishna, R. J. of Catal. 202 (2001) 395-401.

[13] Karger, J. and Pfeifer, H. Zeolites 7 (1987) 90-107.

[14] Smit, B., Siepman, JI. Science 264 (1994) 1118.

[15] Smit, B.; Maesen, L.M. Nature 374 (1995) 42.

[16] Lohse, U.; Fahlke, B. Chem.Tech. 35 (1983) 350.

[17] Jobic, H. J. Molec. Catal.A: Chem. 158 (2000) 135.

[18] Schemert, U.; Karger, J.; Weitkamp, J. Microporous Mesoporous Mater. 32(1999) 101.

16 REFERENCES

[19] Lin, Y.S.; Yamamoto, N.; Choi, Y.; Yamaguchi, T.; Okubo, T.,; Nakao, S.I.Microporous Mesoporous Mater. 38 (2000) 207.

[20] van de Graaf, J.M.; Kapteijn, F.; Moulijn, J. Microporous Mesoporous Mater.35-36 (2000) 267-281.

[21] Talu, O.; Sun, M.S.; Shah, D.B. AIChE Journal 44(3) (1998) 681.

[22] Zhu, W.; van de Graaf, J.M.; van den Broeke, L.J.P.; Kapteijn, F.; Moulijn, J.A.Ind. Eng. Chem. Res. 37 (1998) 1934.

[23] Zhu, W.; Kapteijn, F.; Moulijn, J.A. Microporous Mesoporous Mater. 47 (2001)157.

[24] Eic, M., Ruthven, D.M.. Zeolites 8 (1988) 40.

[25] Song, L.; Rees, L.V.C. Proc. 12th Int.Zeolites Conf., 1999, pp.67-74.

[26] Nijhuis, T.A.; van den Broeke, L.J.P.; van de Graaf, J.M.; Kapteijn, F.; Mak-kee, M.; Moulijn, J.A. Chem. Eng. Sci. 52(19) (1997) 3401.

[27] Karger, J.; Bulow, M. Chem. Eng. Sci. 30 (1975) 893.

[28] Paschek, D.; Krishna, R. Chem. Phys. Lett. 333 (2001) 278-284.

[29] Tsikoyiannis, J.G.; Wei, J. Chem. Eng. Sci. 46 (1991) 233.

[30] Coppens, M.-O.; Bell, A.T.; Chakraborty, A.K. Chem. Eng. Sci. 53 (1998) 2053.

[31] Auerbach, S.M.; J. Phys. Chem. B 106 (1997) 7810.

[32] Chen, Y.D.; Yang, R.T. AIChE Journal 1991 (37(10)) 1579-1582.

[33] Krishna, R.; Paschek, D. Phys. Chem. Chem. Phys. 4 (2002) 1891.

[34] Krishna, R. Trans IChemE 101 (1997) 10121.

2 Experimental method

This chapter is dedicated to the positron emission profiling (PEP) technique, whichhas been applied for the in situ study of diffusion and adsorption of hydrocarbons inMFI-zeolites. The method is based on the imaging of radio-labelled molecules insidethe packed bed reactor. It requires the production of radio-labelled alkanes, which isdescribed here. PEP experiments can be performed in tracer-exchange (TEX-PEP)and pulse modes. TEX-PEP experiments are very helpful in the study of multi-component mixtures, while the pulse experiments provide information on the diffu-sion under zero zeolite loading. Both experimental procedures are suitable for themeasurement of the self-diffusion coefficient under various pore occupancies of thezeolite. Modelling of the obtained images of radio-labelled molecules inside the reac-tor is needed in order to extract the information on the diffusion and adsorption. Abrief introduction into the model used for the interpretation of the experimental datais provided in this chapter.

2.1 Positron Emission Profiling

Positron Emission Profiling (PEP) is a non-invasive in situ experimental techniquebased on the use of the radioactive isotopic labelling. It is applied for studying ad-sorption, mass transfer, [1, 2, 3, 4] and heterogeneous reactions, [5, 6].

By definition isotopes are the chemical elements that have the same atomic num-ber but different atomic weight. Radioactive nuclei are unstable, hence undergo a nu-clear decay, which results in another chemical element and nuclear fragments. Thosefragments can be α-particles (helium nuclei), β+- or β− particles (positron and elec-tron, respectively) and high energy γ-photons. All of them can be detected. Isotopeshave identical chemical properties, therefore, introducing the radio-labelled analogof studied molecules in the investigated system will not influence its behavior butwill allow to ”visualize” it. Isotopic labelling has been successfully applied in variouskinds of research including medicine, biology, chemistry, physics, etc. In the presentwork use of radio-labelled molecules allows one to visualize the mass transport insidethe zeolite pores.

The idea of the PEP technique comes from the positron emission tomography(PET). The concept of both techniques consists of addition of a very small amountof positron emitting isotopes to the investigated system. The course of the isotopethrough the system is followed by a specially designed detector. A major advantageof the positron emitters used in this study is their short half-life time. This meansthat only very small quantities of labelled material are required. When it is injectedin the system no disruption of the process occurs. PET was developed over 20 yearsago. Now it is a diagnostic technique widely used in medicine, which provides 3D-

18 Chapter 2: Experimental method

Figure 2-1. Schematic drawing of the annihilation event.

images of the distribution of radio-labelled molecules within living organs. Moreover,it provides the rate information on the biochemical or metabolical processes. Jonkerset al., [7], tried to apply a commercial PET-camera to the study of chemical reactions,however, only a one-dimensional image could be obtained with insufficient spacial andtime resolutions. Therefore, a necessity to improve the method for the study of thechemical systems has emerged and a one-dimensional PEP technique with higherresolutions has been developed, [8].

The main advantage of PEP compared to other applications of isotopes for the re-action studies is that it provides space-and-time resolved information on the locationof radioactive molecules, i.e. imaging. Even though it is a quite powerful technique,it is unable to directly identify the molecules inside the reactor (outlet reaction mix-ture composition can be measured with chromatography), which is important for thereaction study. Therefore, modelling is involved in the data interpretation. In thisthesis the behavior of alkanes and their mixtures in zeolites has been studied in theabsence of any reaction. The molecules have been identified before being introducedin the reactor. Nevertheless, for the quantitative analysis, modelling of the processand fitting to the experimental data are required.

To study the diffusive and adsorptive properties of hydrocarbons in zeolites, cor-responding positron emitting 11C-labelled molecules are used. The nuclear decay of11C is quite fast with the half-life time of 20.4 min, therefore an on-line productionof this isotope is required alongside with the production of radio-labelled alkane foreach experiment. The procedure and the PEP experiments will be described below,but first, the main detection principle of the PEP technique will be provided.

Section 2.1: Positron Emission Profiling 19

Figure 2-2. Photon detectors: a) schematic drawing; b) photo of the PEP detectors.

Detection principals

Imaging of the mass transfer inside the reactor using radio-labelled molecules is basedon the following principle. The radioactive 11C-nucleus decays with the half-life timeof 20.4 min emitting a positron, which is the antiparticle of an electron, and a neutrinoν:

11C −→ β+ +11 B + ν. (2.1.1)

The positron annihilates with electrons from the surrounding matter, yielding a pair ofγ-photons with the energy of 511 keV each emitted under angle of 1800 (Fig. 2-1). Thisproperty of the emitted photons underlies the PEP technique: via coincident detectionof the photons, the position of the annihilation event, and therefore a location of theradioactive molecules can be reconstructed. For this purpose a specially designedsystem of the photon detectors (called PEP detector) is used.

The PEP detector, [8], consists of two detector banks. Each detector bank consistsof 9 detectors made of Bismuth-Germanium oxide scintillation crystals coupled tophotomultipliers and capable of detection of the γ-photons. They are mounted aboveand below the reactor on equal distance from the cylindrical axis of the reactor asshown in Fig. 2-2. 511 keV γ-Photons produced from the annihilation are able topenetrate the reactor walls and travel a significant distance, therefore it is possibleto detect these outside the reactor. Since the photons are emitted in exactly oppositedirections as schematically shown in Fig. 2-1 the position of the decay is determinedby coincident detection of a pair of photons along the cylindrical axis of the reactor by


Figure 2-3. Scheme of the Positron Emission Profiling setup.

two elements, one from each array. The detection is considered to be coincident if itoccurs within 50 ns. The position of the decay is reconstructed from the intersectionof the line which connects these two elements and the cylindrical axis of the reactor.The activity is measured at 17 equidistant positions with the distance between themof 3 mm. The data is collected every 1 s. In this way a space and time resolvedconcentration profile of the injected radio-labelled alkanes can be obtained with atime resolution of 1 s and a spatial resolution of 3 mm. The reactor containing thezeolite sample is placed so that the middle of the zeolite packed bed is in the middleof the PEP detector. If the reactor bed length is equal to 30 mm, then 10 detectionpositions are involved in the imaging of the molecules inside the reactor bed, whilethe rest of the positions monitor radioactive molecules in the gas phase in the reactorspace not filled with the zeolite.

Production of 11C-labelled alkanes

The isotope 11C is produced via irradiation of a nitrogen target with 12 MeV protonsfrom the 30 MeV AVF Cyclotron of Eindhoven University of Technology accordingto the following nuclear reaction:

14N + p −→11 C + α, (2.1.2)

here p is a proton. Due to the oxygen impurities in the target, the produced isotopeis converted into 11CO by radiolysis. Further, it is transferred to the setup for theradio-labelled hydrocarbon production via a homologation reaction on a vanadium-promoted Ru/SiO2 catalyst, [9]: 11C-labelled C6−alkanes are synthesized using non-labelled 1-pentene, while 11C-labelled C5−alkanes are produced using non-labelled

Section 2.2: Experimental procedures 21

Figure 2-4. TEX-PEP profiles measured at 7 positions, n-hexane in silicalite, T=413 K.

1-butene, respectively. As a result, a range of C1 − C6 and C1 − C5 alkanes in eachcase is produced.

After the separation of the reaction products using a chromatographic column,the hydrocarbon of interest (n-hexane, 2-methyl-, 3-methylpentane or n-pentane) istransferred to the PEP-setup, schematically shown in Fig. 2-3. There it is collectedinto the syringe in order to be used in the PEP-experiments, which can be performedin the tracer-exchange (TEX-PEP) or pulse modes.

2.2 Experimental procedures

TEX-PEP experiments

The experimental setup is schematically shown in Fig. 2-3. In the tracer-exchangePEP experiments (TEX-PEP) 11C-labelled hydrocarbons from a syringe are contin-uously introduced using the pump in a flow of non-labelled molecules with prede-termined composition, passing through the zeolite-packed bed reactor (PEP reactorin Fig. 2-3). The flow of labelled molecules is significantly lower than the total flow,thus there is no disturbance of the equilibrium between the non-labelled alkane andthe zeolite. The exchange process between 11C-labelled and non-labelled molecules ismonitored using the PEP detector described earlier. When the equilibrium betweenradio-labelled molecules in the gas and adsorbed phase has been reached a constantsignal is received from the detector. At that moment, the flow of radio-labelled hy-drocarbons is switched off by stopping the syringe pump. After that the re-exchangeprocess takes place. The concentration of the non-labelled hydrocarbons is kept con-stant during the experiment, thus, steady-state conditions are fulfilled.


Figure 2-5. Pulse experiment: concentration profiles for 7 detection positions, 3-methylpentane

in silicalite, T=443 K.

The flow composition of non-labelled hydrocarbons in hydrogen as a carrier gas isset using a Bronkhorst CEM system (marked as ”mixer” in Fig. 2-3). The liquid alka-nes (n-hexane, 3-, 2-methylpentane or n-pentane) used in the experiments have beenobtained from Aldrich (99.9 % purity) and are used as received. The binary mixturewas produced using a controlled evaporator and mixer (CEM) mixing unit consist-ing of two branches each of which was equipped with a mass flow controller (MFC).By the use of these controllers for two different alkane branches, the composition ofthe mixture can be set by varying the ratio between the gas flow of the componentswith the constant total flow. The outlet mixture composition and the absence of thereaction can be checked using two gas-chromatographs, GC (NaI) and GC (FlameIgnition Detector (FID)) in Fig. 2-3. Radio-labelled components can be detected withGC (NaI), while non-labelled molecules can be analyzed using GC (FID).

An example of the measured concentration profile of radio-labelled molecules inthe reactor is shown in Fig. 2-4. Each plot in this Figure corresponds to the detectionposition which corresponds to the certain distance from the beginning of reactor bed.At time 0 s, when the injection of radio-labelled is started the number of counts isincreasing at every positions inside the reactor due to exchange with non-labelledmolecules. The exchange is complete in the whole reactor within approximately 100 sin this experiment. Thereafter, the signal received by the PEP detector is constantand equal between almost all the positions. For the first plot (TEX-PEP profile), lowernumber of counts (number of radio-labelled hydrocarbon molecules) at the equilib-rium is measured. This is because the beginning of the zeolite packed bed does notcoincide with the beginning of the detector bank. As discussed earlier (Section 2.1),the spacial resolution is 3 mm. Apparently, the beginning of the zeolite bed corre-

Section 2.2: Experimental procedures 23

sponds to the middle of the fifth detection position. Therefore, only half of what isobserved from the other position is measured. Another feature of the concentrationprofiles is the time delay between them, which is caused by the adsorptive proper-ties of the injected component. The first profile, which corresponds to the beginningof the zeolite packed bed has reached the equilibrium earlier, while the 7th curvewas the latest. After the flow of radio-labelled molecules is stopped at approximatetime of 500 s (Fig. 2-4), due to re-exchange with non-labelled molecules the num-ber of counts detected at each position is decreasing up to some background valueof approximately 50 counts. Then, the re-exchange is considered to be finished, andthe measurement is stopped. The time of the decay of each curve is determined bythe diffusion of the component. The time delay between the curves is determined byits adsorptive properties. The re-exchange part of the TEX-PEP profiles is used forthe quantitative data interpretation, [10]. From single experiment the loading andself-diffusion coefficient of the investigated component in zeolite are obtained.

TEX-PEP experiments have been implemented only recently, [10], but they openeda unique opportunity to study diffusion and adsorption processes simultaneously inmulti-component mixtures of alkanes in zeolites. The method allows to introduce aradio-labelled counterpart of one of the mixture components and monitor its behav-ior in the mixture under steady-state conditions. Thus one can measure in situ theloading and self-diffusivity of the given hydrocarbon. In the present study, in orderto measure the diffusion of the components in binary mixture, two experiments haveto be performed. In one experiment a radio-labelled alkane-1 is used in order to mon-itor its behavior in the mixture. Then the experiment was repeated with labelledcomponent-2 under similar conditions.

Pulse experiments

Pulse experiments are very useful when the information on the diffusive propertiesof the component under zero loading conditions are of the interest. Unfortunately,this method does not allow to study multi-component mixtures. Nevertheless, themeasurement of the diffusivity of a single component under various partial pressureconditions is possible. This approach has been used for the investigation of the con-centration dependence of the self-diffusion coefficient in zeolites.

During the experiments a constant flow of non-labelled hydrocarbon in a carriergas (hydrogen) with predetermined composition is fed in a plug flow reactor con-taining the zeolite sample. The partial pressure of the alkane was set as describedearlier. Under zero partial pressure conditions, only hydrogen as a carrier gas was fedinto the reactor. In order to measure the diffusivity of the alkane, a pulse contain-ing approximately 10−15 moles of the 11C-labelled hydrocarbon of interest is injectedas a pulse of 2 s in the flow passing through the reactor. The development of thepulse (change of the radio-labelled concentration profile in time) is monitored using aPEP detector. As the concentration of the non-labelled component does not changeduring the experiment, and the quantity of the injected radio-labelled molecules isvery small, the diffusion coefficient estimated from the modelling of the data is the


Figure 2-6. SEM picture of HZSM-5 sample.

self-diffusion coefficient.Typical output of the pulse experiment is shown in Fig. 2-5. Each plot shows the

changing concentration of radio-labelled alkane at several detection positions in time.The maximum number of counts is observed for the first profile, which correspondsto the beginning of the zeolite bed. Further, a broadening of the concentration profileoccurs when the pulse moves along the reactor bed, due to adsorption and diffusionof the molecules in zeolite. Therefore, the maximum of each plot is smaller comparedto the previous profile. Delay between the profiles is related to the adsorption. Here,it is quite small. For stronger adsorbed molecules, a delay would be larger. The asym-metric shape of the profiles is observed in Fig. 2-5. This indicates that intracrystallinediffusion is a rate-limiting process compared to the adsorption. How fast is the decayof each profile is determined by the diffusivity.

2.2.1 Zeolite samples

The measurements have been performed on large crystals of silicalite-1 and HZSM-5zeolites. Use of large zeolite crystals simplifies significantly the modelling and exper-imental procedure because it allows to avoid pelletizing of the crystals, thus there isno macropore diffusion and intracrystalline diffusion is the dominating process.

The sample of silicalite-1 has been kindly supplied by Shell Research and Tech-nology Center, Amsterdam. Scanning electron microscopy (SEM) showed that it con-sisted of regular coffin-shaped crystals with an average size of 150µm × 50µm ×30µm.

HZSM-5 has been kindly provided by Dr. L.Gora from Delft University. The

Section 2.3: Modelling 25

average crystal size was determined with scanning electron microscopy and turnedout to be 160µm × 25µm × 25µm. Fig. 2-6 shows the SEM picture of the sample.One can see that the crystals have quite unform size and shape without very ob-vious shape defects. The Si/Al ratio was 40 in the initial gel composition, yieldingthe concentration of the Brønsted sites [H+]= 7.5×10−6 mol/g ± 2.5×10−6 mol/g.This concentration has been determined by temperature-programmed isopropylaminedecomposition, [11]. This is possible since it is known that one molecule of isopropy-lamine is decomposed per Brønsted site. Thus, the acid site concentration is equal tothe amount of isopropylamine decomposed per gram of zeolite. The bed porosity wasdetermined to be ε=0.44.

Samples of approximately 300 mg have been used in the experiments. This pro-vided a zeolite bed length of approximately 25-30 mm. Prior to experiments, zeoliteshave been activated in a hydrogen stream for 2 hours at 723 K in order to removeadsorbed water and any other adsorbed impurities.

2.3 Modelling

In order to interpret the data provided by the PEP experiments, an appropriatemathematical model has to be used to describe the processes in the zeolite reactorbed. A detailed description of the model and more information about its applicationcan be found in work of Schuring et al., [4]. Below a brief description of the modellingof the experiments is provided.

General description of the model

The model has to describe several stages of the mass transport: inside the crystals andbetween them, in the bed. The use of non-pelletized zeolite crystals described aboveallows to exclude a mass transport in macro-pores1. The transport of molecules occursvia convection and axial diffusion in the bed, adsorption/desorption at the crystalsurface and diffusion inside the pores of the crystals.

The model is used to analyze three types of experiments: pulse and TEX-PEP ex-periments with single component, and TEX-PEP experiments with binary mixtures.Diffusion coefficients can be estimated from modelling all three types of experiments.Both types of single component experiments performed under similar conditions pro-duce results consistent with each other. This confirms the validity of the approach.Additionally, from TEX-PEP experiments (both single component and binary mix-tures) zeolite loading can be obtained. When studying a binary mixture in TEX-PEP experiment only one component is detected during the experiments. Therefore,a model for single-component can be adopted to describe this case as well. The ob-tained parameters describing the different processes in the bed are then the effectivevalues for the transport of this component in the mixture.

1If crystals were pelletized, as is usually the case for small crystals, diffusion in the macro-poreswould be the third stage of the mass-transport process.


Model for the mass transport in the zeolite packed bed

Let us start by describing the transport process in the bed. Mass transfer throughthe bed, resulting from diffusion, convection and mass transfer to the zeolite crystals,can be described by the following equation:

∂cgas

∂t= Dax

∂2cgas

∂z2− νint

∂cgas

∂z+

3(1− ε)

RcεNc. (2.3.1)

In this equation z is the coordinate along the cylindrical axis of the reactor, cgas

is a gas phase concentration in the catalyst bed. Dax is the axial diffusion coeffi-cient, calculated from the molecular diffusion coefficient of the component. Rc is acharacteristic crystal size, νint is the interstitial velocity, which is defined as

νint =νsup

εz

, (2.3.2)

where νsup is the gas flow velocity in the empty reactor, εz-porosity of the bed, ε isthe zeolite crystal porosity. Rc, Dax, νint and ε are fixed parameters, which can bedetermined or estimated. In Eq. (2.3.1) the first term in the right-hand-side describesaxial diffusion in the bed. The second term describes the flow of the gas through thebed. Finally, the third term describes the flow to the crystals, where Nc stands forthe mass flux through the boundary of the zeolite crystal. It is determined by therate-limiting step for the adsorption/desorption at the crystal boundary. Diffusionthrough the laminar fluid film surrounding the particles is neglected, because thisprocess is much faster than diffusion inside the zeolite crystal, [4]. Adsorption anddesorption at the outer surface of the zeolite crystal are fast processes comparedto the diffusion inside the pores. Therefore, diffusion is a rate-limiting step for thedesorption.

Boundary conditions between the zeolite bed and the crystal

A the crystal boundary the flow to the surface must be equal to the desorptionrate at the crystal boundary. The model of Nijhuis et al., [13], explicitly accountsfor adsorption-desorption at the crystal boundary, assuming Langmuir adsorptionkinetics. As the PEP experiments are conducted under steady-state conditions, thismechanism can be replaced by a simple first-order adsorption-desorption process:

Nc = kdcx(Rc, z, t)− kacgas(z, t). (2.3.3)

Here x is a coordinate in the crystal, ka and kd are the adsorption and desorptionconstants, respectively. On the other hand, the flow is equal to the diffusive flow fromthe bulk of the crystal,

Nc = −Dc∂cx

∂x

∣∣∣∣boundary

(2.3.4)

where cx is an adsorbed concentration inside the crystal and Dc is the intracrystallinediffusivity.


Moreover, the adsorption equilibrium can be assumed at the crystal boundary,which has been shown to be a reasonable assumption, [4], when adsorption-desorptionis fast compared to the diffusion in the crystal. Thus, the boundary equation at thecrystal surface is expressed by adsorption equilibrium condition:

cx(Rc, z, t) = Ka × cgas(z, t), (2.3.5)

where Ka = ka/kd is the adsorption equilibrium constant.

Model for the mass transport in the crystals

Let us now describe the model for the mass transport process inside the crystals.A single crystal has a coffin-like shape, as shown in Fig. 2-6. Moreover, in a singlecrystal diffusion is anisotropic in general, [14]. It is not enough, however, to considera processes in a single crystal. In the experimental situation a very large sample ofrandomly oriented crystals is present. The experimentally observed transport processis an average over orientation of crystals. This suggests to use an approximationin order to simplify the modelling: the zeolite crystals can be assumed to have aspherical shape and micro-pore diffusion can be assumed to be an isotropic process.This approximation is commonly made in the literature, [12]. It has been shown to bequite reasonable, [12], when the radius of the spherical crystal is chosen to match theexternal surface to volume ratio of the real ones. In our case this gives Rc = 25µmfor silicalite and Rc = 20µm for HZSM-5 crystals.

In the approximation discussed above the transport inside the zeolite crystals isdescribed by the following equation:

∂cx

∂t= Dc

(∂2cx

∂x2+

2

x

∂cx

∂x

), (2.3.6)

where x is the radial coordinate in the crystal.Intracrystalline diffusivity, Dc, is the average value of diffusivities in all the direc-

tions. In general, Dc is concentration dependent. However, during the single experi-ment the concentration of the non-labelled molecules is constant. In particular, in thepulse experiments under zero zeolite loading, the amount of the labelled moleculesinjected is negligibly small. Therefore, Dc can thus be regarded as constant during asingle experiment.

Boundary conditions for the crystal

Eq. (2.3.6) receives two boundary conditions. At the center of the particle the bound-ary condition is fixed by the symmetry:

∂cx

∂x

∣∣∣∣x=0

= 0. (2.3.7)

The boundary condition at the surface has been discussed already in Eq. (2.3.5).


Initial conditions for the model

Finally, Eqs. (2.3.1) and (2.3.6) require the initial conditions. Initial condition is alsowhat distinguishes application of the model to pulse and TEX-PEP experiments.At the beginning of the tracer re-exchange process (TEX-PEP) the system is inequilibrium. The initial conditions corresponding to this case are

cgas(z, 0) = c0,

cx(x, z, 0) = Ka × c0, (2.3.8)

where c0 is the injected initial tracer concentration. In contrast, during the pulseexperiments the zeolite reactor bed does not contain any radio-labelled moleculesbefore the pulse of labelled molecules is injected. This provides the following initialconditions for the bed and for the crystal:

cgas(z, 0) = 0, cx(x, 0) = 0. (2.3.9)

Pulse of radio-labelled molecules injected in the flow of non-labelled molecules isapproximated by a function f(t), which is a Gaussian function here. Thus, additionalinitial boundary condition is provided by the concentration of radioactive moleculesjust in front of the zeolite bed cz(−z, 0):

cz(−z, 0) = f(t)× c0, (2.3.10)

Solution of the model

The partial differential equations (PDE) described above are solved using the numer-ical method of lines, [15]. This procedure has been described in detail in Noordhoeket al., [16]. In short, the method consists of converting of PDEs into a set of ordinarydifferential equations (ODE) that can be easily solved. It is done as following. Thetime, t, is used as a continuous variable, while the spacial variables are discretized.There are two sets of grid points for the zeolite bed (z) and the crystal (x). Thus, or-dinary time-differential equations for the concentration at each grid point are derived.Solving them yields values for the concentration at each bed and crystal gridpoint.The PEP detector measures the total concentration of labelled molecules in a certainsection of the catalyst bed. Therefore, the gas phase in the bed and the phase inthe zeolite crystal both contribute to the measured PEP profiles. Thus, solutions forthose phases have to be converted into the PEP profile. A volume averaging has tobe applied to simulate the detector’s response. First, the volume-averaging for thecrystal situated at z coordinate in the bed is made:

〈cx(z, t)〉 =3

R3c

Rc∫0

cx(x, z, t)x2dx. (2.3.11)

The total concentration at the z position in the zeolite bed is calculated as a sum ofthe concentration in the crystal and in the gas phase at this position:

ctotal(z, t) = εcgas(z, t) + (1− ε)〈cx(z, t)〉. (2.3.12)


Figure 2-7. Experimental data fitted by the model: n-pentane in silicalite, T=373 K.

Use of the model

An example of a fit the measured concentration profiles for TEX-PEP and pulseexperiments is shown in Fig. 2-7 and Fig. 2-8, respectively. From these figures onecan see that the model is able to explain the observed concentration profiles quitewell. Fitting the modelled concentration profiles to the measured ones yields a valueof the unknown parameters such as the adsorption constant Ka and intracrystallinediffusivity Dc. To be able to extract reliable data from the experiments, the parame-ters of interest must have a significant influence on the shape of the exchange curves.This has been shown to be the case for Ka and Dc in the work of Schuring, [4].Both parameters appear to influence the measured concentration profiles. Adsorp-tion constant, Ka, affects the time scale at which the change in concentration travelsthrough the bed without influencing the actual shape of the concentration profile.This effect is most pronounced in the case of TEX-PEP experiments, Fig. 2-7. Onthis figure the curves appear to be simply shifted with respect to each other, and theshift is related to adsorption constant, Ka. In the case of pulse experiments, Fig. 2-8,the effect manifests itself in a more complicated way: curves are not simply shifted.The diffusion constant, in contrast, mainly influences the shape of the concentrationprofile and causes longer tails with decreasing diffusivity. To summarize, analysis ofthe shape of the concentration profiles provides the information on both the diffusionconstant and the adsorption constant.

Loading

In order to study the concentration dependence of the diffusion coefficient it is im-portant to be able to measure the zeolite loadings. PEP-experiments themselves pro-


Figure 2-8. Example fit of a pulse of 3-methylpentane in silicalite, T=443 K.

vide such information. Adsorption constant Ka determined from the model fit of theTEX-PEP concentration profiles allows, in principle, to calculate the zeolite loading.Moreover, for multi-component mixtures one can determine the zeolite loading foreach component separately. The loading of the zeolite is given in [mmol/g] by

cads = Kap

ρRT, (2.3.13)

where p is the partial pressure of the alkane, ρ is the zeolite density, R is the gasconstant and T is the temperature.

2.4 Direct measurement of the zeolite loading

As was already mentioned above the knowledge of the zeolite loadings is importantfor the studies of the diffusion concentration dependence. Here we describe an inde-pendent way to measure loadings not relying on the model fits. This method can beapplied for single components only.

The experiments to determine the component loading were performed using Balz-ers Quadrupole mass-spectrometer system QMG-420. The zeolite sample was exposedto a hydrocarbon/hydrogen stream until an equilibrium between the gas and adsorbedphase was reached, which was monitored using the mass-spectrometer. The systemwas considered to be in equilibrium when the signals of the hydrogen and alkanemasses were constant. After equilibration, the flow of the hydrocarbon was switchedoff, resulting in the desorption of the adsorbed species. The total amount of desorbedalkane was then calculated from the integration of a calibrated signal of hydrocarbon’smass over desorption time. The loading at a given partial pressure and temperature is

Section 2.4: Direct measurement of the zeolite loading 31

determined as the amount of the desorbed hydrocarbon per gram of the zeolite. Thezeolite pore occupancy θ was calculated as a ratio between the loading under givenconditions and the maximum loading. The loading has been taken as a saturationone when further increase in the partial pressure of the hydrocarbon did not resultin the increase of the adsorbed concentration.

Some nomenclature

• t [s] - time;

• z [m] - coordinate along the cylindrical axis of the reactor;

• x [m] - radial coordinate in the zeolite crystal;

• Rc [m] - radius of zeolite crystal;

• cgas [mol/m3] - gas phase concentration in the bed;

• cx [mol/m3] - concentration in the crystal;

• Dax [m2/s] - axial diffusion coefficient;

• Dc [m2/s] - intracrystalline diffusivity;

• νint [m/s] - interstitial velocity;

• νsup [m/s] - gas flow velocity in the empty reactor;

• εz [-] - porosity of the bed;

• ε [-] - zeolite crystal porosity;

• Nc [mol/m2·s] - mass flux through the boundary of the zeolite crystal;

• Ka [-] - adsorption equilibrium constant;

• ka [m/s] - adsorption constant;

• kd [m/s] - desorption constant;

• c0 [mol/m3] - injected initial tracer concentration;

• θ [-] - zeolite pore occupancy;

• cads [mol/g] - zeolite loading;

• p [Pa] - partial pressure of the alkane;

• ρ [g/m3] - zeolite density;

• R [Pa·m3/mol· K] - gas constant;

• T [K] - temperature.

References

[1] Anderson, B.G.; de Gauw, F.J.; Noordhoek, N.J.; van IJzendoorn, L.J.; vanSanten, R.A.; de Voigt, M.J.A. Ind. Eng. Chem. Res. 37 (1998) 815.

[2] Anderson, B.G.; van Santen, R. A.; van IJzendoorn, L.J. Appl. Catal. A 160(1997) 125.

[3] Anderson, B.G.; van Santen, R.A.; de Jong, A.M. Top. Catal. 8 (1999) 125.

[4] Shuring, D.; Koriabkina, A.O.; de Jong, A.M.; Smit, B.; van Santen, R.A. J.Phys. Chem. B 105(32) (2001) 7690-7698.

[5] Sobzcyk, D.P.; de Jong, A.M.; Hensen, E.J.M.; van Santen, R.A. submitted toTopics in Catal. (2003) .

[6] van der Linde, S.C.Application of Positron Emission Profiling in catalysis, 1999,Ph.D.Thesis, Delft: Technische Universiteit Delft.

[7] Jonkers, G.; Vonkeman, K.A.; van der Waal, S.W.A.; van Santen, R.A. Nature355 (1992) 63.

[8] Mangnus, A.V.G.; van IJzendoorn, L.J.; de Goeij, J.J.M.; Cunningham, R.H.;van Santen, R.A.; de Voigt, M.J.A. Nucl. Instr. and Meth. 99 (1995) 649.

[9] Cunningham, R.H.; Mangnus, A.V.G.; van Grondelle, J.; van Santen, R.A. J.Molec. Catal.A: Chem. 107 (1996) 153.

[10] Schumacher, R.R.; Anderson, B.G.; Noordhoek, N.J.; de Gauw, F.J.M.M.; deJong, A.M.; de Voigt, M.J.A.; van Santen, R.A. Microporous Mesoporous Mater.35-36 (2000) 315.

[11] Juskellis, M.V.; Slanga, J.P.; Roberrie, T.G.; Peters, A.W. J. of Catal. 138 (1992)391-394.


[13] Nijhuis, T.A.; van den Broeke, L.J.P.; Linders, M.J.G.; van de Graaf, J.M.;Kapteijn, F.; Makkee, M.; Moulijn, J.A. Chem. Eng. Sci. 54 (1999) 4423.

[14] Hong, U.; Karger, J.; Kramer, R.; Pfeifer, H.; Seiffert, G.; Muller, U.;Unger, K.K.; Luck, H.B.; Ito, T. Zeolites 11 (1991) 816.

[15] Schiesser, W.E. The numerical method of lines: integration of partial differentialequations; Academic Press: San Diego, 1991.

34 REFERENCES

[16] Noordhoek, N.J.; van IJzendoorn, L.J.; Anderson, B.G.; de Gauw, F.J.; vanSanten, R.A.; de Voigt, M.J. Ind. Eng. Chem. Res. 37 (1998) 825.

3 Binary mixture diffusion

In this chapter, the results on the diffusive and adsorptive properties of n-hexane/2-methylpentane mixtures in zeolite silicalite measured with TEX-PEP are presented.A slight preference for the adsorption of n-hexane was found because it is entropicallymore favorable to adsorb these molecules as they have no preferential siting in thezeolite pores. The diffusivity of the slowly moving iso-hexane decreases with its con-centration in the zeolite pores. The mobility of the linear alkane strongly decreaseswith increasing 2-methylpentane fraction in the gas phase. Sudden drop in the dif-fusion coefficient of n-hexane is observed at a loading of 2-methylpentane equal toapproximately 2.75 molecules per unit cell. This is because the branched alkanes arepreferentially sited in the intersections between the straight and zigzag channels ofsilicalite and therefore effectively block the zeolite pore network. These results showthat the siting and adsorptive properties of the components as well as the structureof the zeolite network play an important role in the behavior of multi-componentmixtures in zeolites.

3.1 Introduction

Zeolites have found a vast application in the petroleum industry as catalysts in themost important processes of gasoline production and as molecular sieves in separationprocesses, [1, 2]. A lot of studies have been dedicated to the diffusive and adsorptiveproperties of molecules inside these materials, [3], as they might play a crucial role inthese processes. Most of them are focused on the diffusion and adsorption of singlecomponents. However, in the catalytic or separation process, at least two, or evenmore, species are present inside the zeolite. Mainly, this is caused by technical limi-tations of most conventional methods for studying mass transfer in zeolites, like thegravimetric and volumetric methods, because they are unable to distinguish differenttypes of molecules in zeolite pores. Moreover, it is often assumed that diffusion inmulti-component systems can be predicted from single-component data.

However, recently the number of mainly theoretical investigations on multi-com-ponent mixtures is increasing. Most of these studies are limited to small molecules,like methane/xenon, [4, 5], ethane/ethene, [6], and n-butane/methane, [7]. Most ex-perimental data are obtained with techniques such as NMR and QENS and at rel-atively low temperatures. There are few studies dealing with larger hydrocarbons,such as xylene/benzene mixtures, [8], and cyclic, branched, and linear hydrocarbonsin silicalite membranes, [9]. Recently, Masuda et al. performed the experiments onn-heptane/n-octane and ortho- and meta-xylene mixtures at elevated temperaturesusing a desorption rate process (DRP) method, [10]. The authors did not observe theinfluence of the fast component on the character of diffusion of the slow component,

36 Chapter 3: Binary mixture diffusion

while the diffusivity of the fast component decreases monotonically with the increas-ing fraction of slow components. This is in line with results obtained for the mixturesof smaller components.

In this chapter, TEX-PEP technique is being used to study the adsorption andself-diffusion of n-hexane/2-methylpentane mixtures in zeolite silicalite. These sys-tems were chosen because of their practical applications, the availability of other(theoretical) studies, and the peculiar adsorption behavior observed for the singlecomponents, [15]. With total pressure kept constant, the adsorptive and diffusivebehaviors of n-hexane and 2-methylpentane have been studied as a function of then-hexane/2-methylpentane ratio in the gas phase.

3.2 Experimental section

Experiments have been performed using a PEP setup in the tracer-exchange modeas described in details in Chapter 2.

During the TEX-PEP experiments, a constant flow of non-labelled hydrocarbonsin a hydrogen carrier stream was fed into the reactor. The total flow rate of hydro-carbons and carrier gas was set to a value of 80.2 mL/min. The total hydrocarbonpressure was fixed at 6.6 kPa, and the experiments were conducted at a temperatureof 433 K. At different mixture ratios, the tracer re-exchange of both n-hexane and2-methylpentane was recorded.

For a tracer-exchange experiment, a quantity of labelled molecules of either n-hexane or 2-methylpentane was continuously injected using a syringe pump intothe feed stream containing non-labelled hydrocarbons. For each mixture composi-tion, two experiments have been performed, using radio-labelled n-hexane and 2-methylpentane, respectively, in order to follow the behavior of each mixture compo-nent. In each case, the tracer-exchange and tracer-reexchange processes were moni-tored using the PEP detector (see Chapter 2).

For the experiments, a sample of silicalite-1 described earlier (Section 2.4) has beenused. The length of the zeolite bed was 3 cm. Prior to being used in experiments, thezeolite sample was activated for at least 2 h at 673 K in a hydrogen stream.

A model described in Section 2.3 was used to fit the experimental data in order toextract the information on the self-diffusion coefficients and zeolite loadings. Load-ings of the components at various mixture compositions have been calculated usingEq. 2.3.13.

To measure a zeolite loading at 6.6 kPa, 433 K with n-hexane and 2-methylpentane,respectively, independent experiments have been performed using a mass-spectrometer,as described in Chapter 2. The obtained loadings have been compared with those de-termined from TEX-PEP experiments.

Section 3.3: Results and discussion 37

Figure 3-1. Experimental data: TEX-PEP profiles for labelled 2-methylpentane at 3 detection

positions, 2-methylpentane in 50:50 gas mixture with n-hexane in silicalite, 6.6 kPa, T=433 K.

3.3 Results and discussion

Self-diffusivities and loadings of both 2-methylpentane and n-hexane have been mea-sured in their mixtures varying the ratio between the components. Fig. 3-1 and Fig. 3-2 show the tracer-reexchange curves at different positions along the reactor for 2-methylpentane and n-hexane, respectively, in a 1:1 mixture. Modelling translates themeasurement error to the following error on the fitted parameters: the diffusivities andadsorption constants (see Chapter 2) were extracted from the fitting with an error ofless than 10%. A lower signal-to-noise ratio was observed in case of 2-methylpentane(see Fig. 3-1), which resulted in a slightly lower accuracy (still the error was lessthan 15%). It was caused by a lower, compared to n-hexane adsorbed concentrationof 2-methylpentane in zeolite. The plots also show that the transport properties ofn-hexane and 2-methylpentane in this mixture are different. The re-exchange of thebranched molecule is slower than that of the linear component, indicating a corre-sponding behavior of diffusion for each component. The larger separation in time ofthe exchange curves for the linear alkane indicates that this component has a largeradsorption constant.

Below the diffusive and adsorptive characteristics derived from concentration pro-files for both alkanes, single and in their mixtures, will be discussed.

3.3.1 Adsorption of Single Components

From the fitted adsorption equilibrium constants, the loadings (in moles per gramof zeolite) for n-hexane and 2-methylpentane were calculated using Eq. (2.3.13) asdescribed in Chapter 2. The data for the pure components under the experimental


Figure 3-2. Experimental data: TEX-PEP profiles for labelled n-hexane at 4 detection positions,

n-hexane in 50:50 gas mixture with 2-methylpentane in silicalite, 6.6 kPa, T=433 K.

Adsorbent loading, mmol/g loading, mol./u.c. loading, mmol/g (MS)

n-hexane 0.64±0.03 3.6 0.59±0.032-methylpentane 0.59±0.03 3.4 0.53±0.03

Table 3-1. Loadings of single components in silicalite at 433 K, 6.6 kPa.

conditions are shown in Table 3-1.The slightly lower loading of the branched compared to the linear alkane un-

der equal conditions observed here is in accordance with other studies, [16]. In-deed, Zhu et al., [17], measured higher loading for n-hexane in silicalite comparedto 2-methylpentane even under a significantly higher partial pressure of iso-alkane.At 408 K, a loading of 2.36 molecules per unit cell was measured for n-hexanein silicalite at 0.47 kPa, while for 2-methylpentane the loading reached only 2.28molecules per unit cell at partial pressure of 0.73 kPa, [17]. Loadings for both n-hexane and 2-methylpentane determined from the independent experiments using amass-spectrometer (MS) are in agreement with those provided by TEX-PEP (seeTable 3-1).

Theoretically calculated values of the heat of adsorption for n-hexane and 2-methylpentane are 70 kJ/mol and 65 kJ/mol, respectively, [18, 19], which is in agree-ment with the average values determined by Zhu et al., [20]. As the heats of ad-sorption of these alkanes are very close the difference in adsorption is also causedby an entropic effect. Indeed, the conformations of the bulkier branched alkanes aremuch more restricted in the narrow pores of the zeolite. For the branched isomer insilicalite-1 there is a large difference in the adsorption entropy between the molec-


Figure 3-3. Loadings of mixture components in silicalite as a function of 2-methylpentane frac-

tion in the gas phase, total partial pressure 6.6 kPa, 433 K.

ular locations in the intersections and in the channels as shown by Zhu et al., [20].Therefore, the adsorption of 2-methylpentane from the gas phase leads to a higherreduction in entropy compared to adsorption of n-hexane. This makes it entropicallyless favorable to adsorb the branched isomer, [16]. Thus, higher partial pressure of2-methylpentane compared to linear hexane is required in order to reach the sameloading.

3.3.2 Binary Adsorption

In Figure 3-3, the adsorption of both n-hexane and 2-methylpentane from their gasmixtures are shown as a function of the gas phase mixture composition. Obviously,the n-hexane loading monotonically decreases while the 2-methylpentane loading in-creases with an increasing partial pressure of the branched alkane. The total hydrocar-bon loading (Fig. 3-3) varies only little and slightly decreases at high 2-methylpentanefraction in the gas phase due to the reasons discussed above. Small deviations in ad-sorbed concentrations for both alkanes from a linear dependence on the mixture com-position as shown in Fig. 3-3, indicate a slight preferential adsorption for n-hexanecompared to its monobranched isomer.

In the literature, different types of adsorptive behavior of binary mixtures havebeen reported. For the adsorption of a mixture of para- and ortho- xylene in variousY-type zeolites, it was found that are adsorbed independently and the character ofadsorption for each component in mixtures is similar to that of a single component,[22]. On the contrary, in zeolite 13X from ethane and ethylene gas mixtures, [23], theethylene was preferentially adsorbed at low ethane partial pressures. As a result, anonlinear dependence of the loading of each component on the mixture composition


Figure 3-4. Loadings of mixture components in silicalite as a function of 2-methylpentane frac-

tion in the gas phase, experiment (total hydrocarbon partial pressure 6.6 kPa, 433 K) and CBMC

simulations (total hydrocarbon partial pressure 7.8 kPa, 433 K).

was observed. A similar picture was observed by Heuchel et al., [4], for CF4 and CH4

in silicalite.

The observations of this study (Fig. 3-3) are supported by the study on the resultsof CBMC simulations performed by Vlugt et al., [16], and by Calero et al., [21], onadsorptive behavior of linear and branched alkanes and their mixtures. The simula-tions were performed with a fixed mixture ratio at lower temperatures (300 K and362 K). It was shown, that at a total loading of approximately 4 molecules per unitcell, the loading of the branched alkanes reaches a maximum value. At lower loadings,both components are adsorbed independently, while at higher loadings the branchedalkanes is squeezed out by the linear alkanes. Vlugt et al., [16], showed that thisbehavior of the component is related to the siting of the molecules in silicalite poresystem. It was found that the n-hexane is adsorbed anywhere in the silicalite pores,while 2-methylpentane molecules are located at the intersections between the straightand zigzag channels. As a result of that, n-hexane has a higher packing efficiency.

Apparently, under the conditions used in the present study, the loading of thecomponents were quite high, so that the 2-methylpentane is squeezed out. As a re-sult of the higher packing efficiency of n-hexane, a preferential adsorption for thiscomponent is observed (Fig. 3-3). The linear alkane is replaced only at higher 2-methylpentane fractions, and a nonlinear dependence on the mixture composition isobserved.

Comparison between our experimental results and those calculated from CBMCsimulations, [24], performed for 2-methylpentane/n-hexane mixtures under conditionsclose to those in this study is shown in Fig. 3-4. The loadings of the single components


Figure 3-5. Self-diffusivities of mixture components in silicalite as a function of 2-methylpentane

fraction in the gas phase, total hydrocarbon pressure 6.6 kPa, 433 K.

provided by CBMC simulations are slightly higher, especially for the branched alkane,than the values obtained from TEX-PEP measurements. This is apparently due tothe higher partial pressure of hydrocarbons applied in the CBMC simulations. Theslight decrease of the total loading with the branched alkane fraction in the gasphase predicted from the simulations is in agreement with the TEX-PEP results.However, a slight preferential adsorption for the branched alkane is deduced from thesimulations, while PEP measurements provide lower values for the 2-methylpentaneloadings. This disagreement with our results and the tendency shown by Vlugt et al.,[16], can probably be attributed to imperfections in the model parameters used forthe CBMC simulations.

3.3.3 Multi-component Diffusion

Figure 3-5 shows the self-diffusion coefficients obtained from the TEX-PEP exper-iments for both alkanes as a function of the gas phase mixture composition. As itmight be expected, the self-diffusivity of pure n-hexane is an order of magnitudehigher than that of the 2-methylpentane. Indeed, a minimum kinetic diameter of n-hexane is smaller than that of iso-hexane (4.30 A and 5.00 A, respectively, [26]), whichapparently makes a diffusion of 2-methylpentane more spatially restricted comparedto n-hexane.

For both components, a decrease in mobility is observed with an increasing frac-tion of the branched alkane in the gas phase. Analogous behavior was found forCH4/CF4 mixtures, where the self-diffusivity of both components decreased as theloading of slow tetrafluoromethane increased, [25]. In our study, diffusion trend for2-methylpentane is similar to that one would expect for a single-component system,


Figure 3-6. Self-diffusivities of mixture components in silicalite as a function of 2-methylpentane

loading, total hydrocarbon pressure 6.6 kPa, 433 K.

[27, 28], whereby the self-diffusivity monotonically decreases with increasing loadingof this component. It is also in agreement with other studies, [5, 10, 30]. In all thesesystems, it was found that the diffusivity of the slow component is not noticeablyaffected by the presence of the fast component.

In our study, the diffusivity of the faster n-hexane slowly monotonously decreaseswith 2-methylpentane fraction in the gas phase up to 0.75. At this mixture compo-sition a sudden drop in the linear alkane diffusion is observed. n-Hexane diffusivitybecomes quite close to the diffusivity of the branched alkane. Even though it wasshown for silicalite-1 that diffusivity should decrease with the pore occupancy, [29],this decrease in mobility must result from hindrance by its branched isomer in thesystem, because the total mixture loading is approximately constant (Fig. 3-3).

Diffusion in zeolites is consider to proceed via sequence of activated jumps fromone site to the other. A jump is successful if the neighboring site, to which the moleculeattempts to jump is empty. As it was discussed in Section 3.3.2, iso-hexane moleculesare preferentially adsorbed in the channel intersections that connect straight andzigzag channels. When the amount of slowly moving molecules (2-methylpentane)increases, they block the connection between the channels. Then, the number ofsuccessful jumps of the fast component (n-hexane) should be determined by therate at which an empty site is created by a jump of the slow component. Thus, athigh loadings of 2-methylpentane, the self-diffusivity of n-hexane becomes stronglydetermined by the self-diffusion rate of its branched isomer.

In Fig. 3-6, self-diffusion coefficients of both components are shown versus 2-methylpentane loading. One can see that the sudden drop in the n-hexane diffusivityoccurs at a 2-methylpentane loading of approximately 2.75 molecules per unit cell.

Section 3.4: Conclusions 43

Since the unit cell of silicalite contains 4 channel intersections, 2.75 molecules of iso-hexane per unit cell is probably enough to block the pore network. When all theintersections are occupied by the slowly diffusing branched alkanes, the entire poresystem will be blocked. As a consequence, diffusion of the fast n-hexane will be deter-mined by diffusion the slow component. Indeed, when the loading of 2-methylpentaneincreases upon further, n-hexane diffusivity continues to decrease (see Fig. 3-6).

The blockage of the pore network was also observed in a system of methane andbenzene in zeolite NaY, [31], and silicalite, [32]. In NaY, the benzene molecules blockthe windows of the supercages for the faster methane. Forste et al., [32], showed thatthe decrease of methane diffusivity was also caused by blocking of the channel inter-sections by benzene in zeolite silicalite. For the methane/xenon mixtures in silicalite,both components are preferentially sited in the interiors of the (straight and zigzag)channels, causing the blocking by the slow components to be less dramatic, [5].

For the n-butane/methane, [30], and methane/tetrafluoromethane, [25], systemsin silicalite, a decrease in the diffusivity of both mixture components was observedwith increase of the loading of slow n-butane, and an increase in methane loading,respectively. In the first case, methane shows a preferential adsorption for the inter-sections, while n-butane is approximately equally adsorbed in the straight and zigzagchannels, [18]. In mixtures of CH4/CF4 in silicalite, CF4 adsorbs preferentially in thestraight channels while methane adsorbs in zigzag channels. The decrease in the eachcomponent diffusion is probably caused by the decreasing with the loading probabil-ity for the molecule to jump to the free neighboring adsorption site. An accelerateddrop in the diffusivity was indeed observed but only at much higher loadings. Ma-suda et al., [10], studied diffusion of the components in mixtures of n-heptane andn-octane in silicalite. These alkanes do not have any preference for the particular ad-sorption site, therefore fast component has to diffuse with the same speed as the slowone, because all possible adsorption sites for n-heptane can be randomly occupied byn-octane.

Thus, multi-component diffusion is strongly related to the zeolite topology andadsorptive properties of the components, which might also be affected by the pres-ence of another adsorbate. As the diffusion of the component might be affected bythe presence of another type of the molecules, its diffusion coefficient in mixtureswill depend on the mixtures composition and can not be predicted from the singlecomponent behavior.

3.4 Conclusions

In this work, TEX-PEP has been successfully applied for determining the adsorp-tive and diffusive properties of an n-hexane/2-methylpentane mixture at elevatedtemperatures (433 K).

The system shows a slight preference for the adsorption of n-hexane over 2-methylpentane. This is caused by the higher packing efficiency of linear alkane be-cause it can be sited in the pore system of silicalite, while the branched alkane is


preferentially adsorbed in the intersection between straight and zigzag channels. Thisconfirms to the conclusions deduced from the CBMC simulations performed by Vlugtet al., [16], at lower temperatures and a constant mixture ratio. CBMC simulationsperformed at similar conditions as in this work show a reasonable agreement, althougha slightly different behavior is predicted by these calculations.

The self-diffusion of the fast n-hexane molecules is strongly influenced by thepresence of the slowly diffusing 2-methylpentane. Moreover, in this study at the 2-methylpentane fraction in the gas phase approximately equal to 0.75, a drastic de-crease in the diffusivity of n-hexane occurs. We believe, this is caused by the blockingof the channel intersections by the slowly moving branched alkanes, as this suddendrop takes place at a 2-methylpentane loading of approximately 3 molecules per unitcell, which corresponds to the situation when approximately 3 out of 4 channel in-tersections are occupied by iso-hexane. This indicates that the adsorptive propertiesof the different components, together with the topology of the zeolite pores, playan important role in the character of adsorption and diffusion of multi-componentmixtures in zeolites.

References

[1] Sie, S.T. Past, Present and future role of microporous Catalysts in the PetroliumIndustry, in: Advanced Zeolite Science and Application Stud. Surf. Sci. 85 (1994)578.

[2] Weitkamp, J. Hydrocarcking and Hydrotreating, in: J.W. Ward (Ed.), ACS Sym-posium Series 20, Am.Chem. Soc., 1975, 1-27.


[4] Heuchel, M.; Snurr, R.Q.; Buss, E.F. Langmuir 13 (1997) 6795.

[5] Jost, S.; Bar, N.-K.; Fritzsche, S.; Haberlandt, R.; Karger, J. J. Phys. Chem. B102 (1998) 6375.

[6] Gladden, L.F.; Sousa-Goncalves, J.A.; Alexander, P. J. Phys. Chem. B 101(1997) 10121.

[7] Gergidis, L.N.; Theodorou, D.N.; Jobic, H. J. Phys. Chem. B 104 (2000) 5541.

[8] Niessen, W.; Karge, H.G. Microporous Mater. 1 (1993) 1.

[9] Funke, H.H.; Argo, A.M.; Falconer, J.L.; Noble, R.D. Ind. Eng. Chem. Res. 36(1997) 137.

[10] Masuda, T.; Fujikata, Y.; Ikeda, H.; Hashimoto, K. Microporous MesoporousMater. 38 (2000) 323.

[11] Anderson, B.G.; van Santen, R. A.; van IJzendoorn, L.J. Appl. Catal. A 160(1997) 125.

[12] Anderson, B.G.; van Santen, R.A.; de Jong, A.M. Top. Catal. 8 (1999) 125.

[13] Anderson, B.G.; de Gauw, F.J.; Noordhoek, N.J.; van IJzendoorn, L.J.; vanSanten, R.A.; de Voigt, M.J.A. Ind. Eng. Chem. Res. 37 (1998) 815.

[14] Schumacher, R.R.; Anderson, B.G.; Noordhoek, N.J.; de Gauw, F.J.M.M.; deJong, A.M.; de Voigt, M.J.A.; van Santen, R.A. Microporous Mesoporous Mater.35-36 (2000) 315.

[15] Smit, B.; Maesen, L.M. Nature 374 (1995) 42.

[16] Vlugt, T.J.H.; Krishna, R.; Smit, B. J. Phys. Chem. B 103 (1999) 1102.

46 REFERENCES

[17] Zhu, W.; Kapteijn, F.; Moulijn, J.A. Microporous Mesoporous Mater. 47 (2001)157-171.

[18] June, R.L.; Bell, A.T.; Theodorou, D.N. J. Phys. Chem. B 96 (1992) 1051.

[19] June, R.L.; Bell, A.T.; Theodorou, D.N. J. Phys. Chem. B 94 (1990) 1508.

[20] Zhu, W.; Kapteijn, F.; van der Linden, B.; Moulijn, J.A. Phys. Chem. Chem.Phys. 3 (2001) 1755-1761.

[21] Calero, S.; Smit, B.; Krishna, R. J. of Catal. 202 (2001) 395-401.

[22] Cottier, V.; Bellat, J.-P.; Simonot-Grange, M.-H.; Methivier, A. J. Phys. Chem.B 101 (1997) 4798.

[23] Buffham, B.A.; Mason, G.; Heslop, M.J. Ind. Eng. Chem. Res. 38 (1999) 1114.

[24] Schuring, D.; Koriabkina, A.O.; de Jong, A.M.; Smit, B.; van Santen, R.A. J.Phys. Chem. B 105 (2001) 7690-7698.

[25] Snurr, R.Q.; Karger, J. J. Phys. Chem. B 101 (1997) 6469.

[26] Breck, D.W.Zeolite Molecular Sieves, John Wiley:New York, 1974.

[27] Schumacher, R.R.; Karge, H.G. J. Phys. Chem. B 103 (1999) 1477.

[28] Schuring, D.; Jansen, A.P.J.; van Santen, R.A. J. Phys. Chem. B 104 (2000)941.

[29] Coppens, M.-O.; Bell, A.T.; Chakraborty, A.K. Chem. Eng. Sci. 53 (1998) 2053.

[30] Gergidis, L. N.; Theodorou, D.N. J. Phys. Chem. B 103 (1999) 3380.

[31] Nivarthi, S.S.; Davis, H.T.; McCormick, A.V. Chem. Eng. Sci. 50 (1995) 3217.

[32] Forste, C.; Germanus, A.; Karger, J.; Pfeifer, H.; Caro, J.; Pilz, W.; Zikanova, A.J. Chem. Soc. Faraday Trans. 83 (1987) 2301.

4 Influence of the acid sites on diffusionof hexanes and their mixtures withinMFI-zeolites.

The self-diffusivities and component loadings of mixtures of n-hexane and 2-methyl-pentane in silicalite-1 and H-ZSM-5 have been measured at 433 K as a function ofthe n-hexane/2-methylpentane ratio in the gas phase (at constant total hydrocarbonpressure of 6.6 kPa) using the PEP technique. Strong preferential adsorption of thelinear alkane over the branched one has been observed in HZSM-5, while in silicaliteonly a slightly higher adsorption of n-hexane was observed. The self-diffusivities ofboth components in the mixtures decrease with increasing 2-methylpentane contentand have been found to be approximately two times lower in H-ZSM-5 than those insilicalite. When the concentration of the branched hexane in the adsorbed phase be-comes close to approximately 3 molecules per unit cell, the influence of the acid siteson the diffusion of linear hexane diminishes compared to the influence of the blockageof the zeolite network with the slowly moving 2-methylpentane. The apparent activa-tion energies of the diffusion for n-hexane and 2-methylpentane in both zeolites havebeen measured at temperatures between 393 K and 533 K under similar conditions.For n-hexane, these values were found to be very close in silicalite and ZSM-5: 18± 2kJ/mol and 22± 2 kJ/mol, respectively. For 2-methylpentane the activation energieswere found to be very high: 66± 6 kJ/mol and 72± 3 kJ/mol, respectively.

4.1 Introduction

A number of experimental and theoretical studies has been conducted on the adsorp-tion and diffusion of single components within zeolites, [1], and it has been acceptedthat the behavior of a component in multi-component mixtures cannot be predictedfrom the single component behavior, [2, 3, 4].

Several experimental and theoretical studies have been dedicated to the investi-gation of binary mixtures of hydrocarbons, [2, 4, 5], in silicalite-1, which is all-silicaMFI-type zeolite without acid sites. However, the diffusion under reaction conditionstakes place inside the catalyst containing Brønsted sites, which also might affectmolecular transport of the reactant mixture components. So far, no studies on multi-component mixtures have been made to establish a possible influence of the acid siteson the diffusion of binary mixture components.

Nevertheless, there are some studies, which compare the diffusivities of single com-ponents in silicalite and HZSM-5, [6, 7, 8]. Zikanova et al., [6], using a constant volumemethod observed that at temperatures between 303 and 423 K diffusion of benzene

48 Chapter 4: Influence of the acid sites on diffusion

in HZSM-5 is twice as low as in silicalite. These results are in a good agreement withthose obtained by FR method, [8]: at temperatures of 375-415 K the diffusivity ofbenzene was 2-3 times lower in the presence of acid sites than in silicalite. Later,Masuda et al., [7], studied the diffusion of aromatics (benzene, toluene, para-xylene)in MFI-type zeolites with different amount of acid sites at 423-723 K. They concludedthat at low temperatures, the diffusivities decrease with increasing amount of acidsites and the activation energy of diffusion is close to the adsorption enthalpy value.

Here, we use the PEP technique. We studied the adsorption and diffusion of n-hexane and 2-methylpentane and their mixtures in two MFI-type zeolites: silicaliteand HZSM-5. n-Hexane and 2-methylpentane represent the reactant/product systemfor the hydroisomerization reaction of n-hexane. This reaction is catalyzed by thenoble metal (Pt) loaded zeolite, containing acid sites.

The main purpose of this study was to compare the behavior of linear and iso-hexanes and their binary mixtures in silicalite and HZSM-5. The experiments havebeen performed at 393 K and 533 K, the absence of any reaction in HZSM-5 has beenchecked by chromatographic analysis of the outlet mixture composition.


The TEX-PEP experiments have been performed using a PEP setup (Chapter 2).During the experiments, a constant flow of non-labelled hydrocarbons in a hydrogenas a carrier gas (80.2 mL/min) was fed into the reactor. The total partial pressure ofhydrocarbon was kept constant at 6.6 kPa, and the experiments were conducted at atemperature of 433 K. At different mixture ratios, the tracer reexchange process forboth n-hexane and 2-methylpentane was recorded in silicalite and HZSM-5 zeolites,respectively. Both zeolite samples are described in Section 2.2.1. For each mixturecomposition, two experiments have been performed, with radio-labelled n-hexane and2-methylpentane, respectively. The tracer-exchange and tracer-reexchange processeswere monitored using the PEP detector (see Chapter 2).

A model described in Section 2.3 was used to interpret the experimental data inorder to extract the information on the self-diffusion coefficients and zeolite loadings.

To measure a maximum zeolite loadings in both zeolites for n-hexane and 2-methylpentane, respectively, independent experiments have been performed usinga mass-spectrometer, as described in Chapter 2. The loading was considered as amaximum one when further increase in the partial pressure of alkane did not resultin the increase of the loading.


In this section the results on the binary mixtures concerning adsorption and diffusionin silicalite and HZSM-5 are presented. Comparison between these zeolites providesinformation about the effect of the component interactions with the acid sites.


Figure 4-1. Loadings of mixture components in both MFI-type zeolites as a function of 2-

methylpentane fraction in the gas phase, total hydrocarbon pressure 6.6 kPa, 433 K.

The single component diffusion and adsorption in these two zeolites is being dis-cussed in order to determine the influence of the Brønsted sites on n-hexane and2-methylpentane separately.

4.3.1 Adsorption and diffusion of binary mixtures in HZSM-5 and silicalite

Diffusivities and loadings of n-hexane and 2-methylpentane in their mixtures withdifferent ratios between the components have been measured in silicalite and HZSM-5 at 433 K.

Adsorption

The loadings of the components in silicalite and HZSM-5 are shown in Fig. 4-1. Theadsorbed concentration of n-hexane in HZSM-5 is higher than in silicalite. This resultshould be expected from the presence of the acid sites, since the enthalpy of n-hexaneadsorption in HZSM-5 (-82 kJ/mol) was reported to be higher than in silicalite (-72kJ/mol), [10]. For iso-hexane these values are 6 kJ/mol lower.

F. Eder, who studied adsorption of 2-methylpentane and n-hexane in MFI-typezeolites, [10], found that in HZSM-5 at high loadings, a complex of 2 molecules ofn-hexane with the hydroxyl group is formed, while iso-alkane molecules are unableto form such complex of two molecules with the acid site. Indeed, in our study, 2-methylpentane loadings in mixtures in HZSM-5 are very close to those in silicalite asone can see in Fig. 4-1.


Zeolite n-hexane 2-methylpentane

6.6 kPa cmax 6.6 kPa cmax

HZSM-5 0.75 mmol/g 1.1± 0.2 mmol/g 0.62 mmol/g 0.74± 0.04 mmol/gsilicalite 0.63 mmol/g 1.2± 0.2 mmol/g 0.59 mmol/g 0.75± 0.02 mmol/g

Table 4-1. Loadings of single components in silicalite and HZSM-5 at 433K.

The loading of n-hexane in mixtures is somewhat higher than it might be asif it were proportional to its partial pressure (Fig. 4-1). On the contrary, the 2-methylpentane loading is lower. This indicates stronger preferential adsorption ofn-hexane over iso-hexane in their mixtures in HZSM-5 than in silicalite.

In the earlier experimental and CBMC simulation study, [2, 11], of n-hexane/iso-hexane mixtures in silicalite a slight preferential adsorption of linear alkane over thebranched one has been shown. As a possible explanation the molecular siting hasbeen proposed. It is considered that n-hexane is probably situated everywhere insidesilicalite micropores, whereas for iso-hexane it is more suitable to be located in thechannel intersection because of entropy reasons, [11]. Consequently, 2-methylpentaneis being pushed out from silicalite by n-hexane. In fact, in HZSM-5 the situation forn-hexane appears to be even more favorable, because of stronger interactions with theacid sites since at high loadings 2 molecules of n-hexane can interact with the acidsite, [10]. As a result, the linear alkane has a higher packing efficiency (Table 4-1).

Table 4-1 shows the adsorbed concentrations of pure components. At 6.6 kPa, insilicalite the amount of n-hexane is just slightly higher than the one of iso-hexane,while in HZSM-5 due to the interactions with the acid sites the linear alkane isobviously adsorbed better than 2-methylpentane.

The maximum loadings of each component have been measured by the aid ofa mass-spectrometer (see Section 2.4). The loading was considered as a maximumone when further increase in the partial pressure of alkane did not result in theincrease of the loading. The values found for the alkanes in both zeolites (Table 4-1)confirm our assumption. n-Hexane, indeed, has a sorption capacity approximatelyequal to 7 molecules per unit cell, which is in a fair agreement with other studies,where the values of 7-8 molecules per unit cell are measured, [10, 12, 13, 14]. For2-methylpentane, maximum loading is close to 4 molecules per unit cell, which isequal to the amount of the channel intersections per unit cell of MFI-type zeolites.As it was mentioned before, the intersections of the straight and zigzag channels arethe most sterically and energetically suitable positions for iso-alkanes to reside. Toobtain loadings higher than 4 molecules per unit cell, the iso-alkanes must find theenergetically less favorable locations, most probably in the channels. This requiresvery high pressures, [11].

The experimental study of single n-butane and iso-butane adsorption with thevolumetric method in silicalite and HZSM-5 leads to similar conclusions, [15]. Valyonet al., [15], found that under identical conditions the n-butane loading was 1.5-2.0


Figure 4-2. Self-diffusivities of mixture components in both MFI-type zeolites as a function of

2-methylpentane fraction in the gas phase, total hydrocarbon pressure 6.6 kPa, 433 K.

times higher than that of iso-butane in the temperature range of 273-413 K. ForHZSM-5 the complete saturation with butanes was reached at lower pressures becauseof the stronger interactions with the acid sites. For n-butane the maximum loadingwas found to be equal to approximately 8 molecules per unit cell, while for iso-butanethe inflection in the isotherm was observed at adsorption of 4 molecules per unit cell.

Therefore, we conclude that in HZSM-5 zeolite interaction with the acid sitesresults in the preferential adsorption of linear hexane over the 2-methylpentane intheir mixtures because of the n-hexane higher packing efficiency.

Diffusion

Fig. 4-2 shows the self-diffusivities of the components in these zeolites as a function ofthe 2-methylpentane fraction in the gas phase. The self-diffusivities of both hexaneslinearly decrease with increasing content of the branched hexane in the gas phase inboth MFI-type zeolites (Fig. 4-2 and Fig. 4-3).

In HZSM-5, the self-diffusion of alkanes is approximately two times slower thanin silicalite. Obviously, the presence of the acid sites is affecting molecular transportvia stronger interactions with the n-hexane molecules. A similar effect of Brønstedsites on the single component diffusion of aromatics was observed in MFI-type zeo-lites with different concentration of acid sites, [6, 7, 8]. The FR technique providedsimilar results for n-butane and iso-butane diffusion in silicalite and HZSM-5, [15]:the diffusivity of both components was approximately twice as low in the zeolite con-taining acid sites, while the diffusion of iso-butane was significantly slower comparedto n-butane.

In the present study of binary mixtures the self-diffusivity of the fast compo-


Figure 4-3. Self-diffusivities of 2-methylpentane in mixtures with n-hexane in both MFI-type

zeolites as a function of 2-methylpentane fraction in the gas phase (enlarged picture from

Fig. 4-2).

nent (n-hexane) in HZSM-5 is influenced by two factors: the presence of slow 2-methylpentane molecules and by the interaction with acid sites. As long as the con-centration of the branched hexane does not exceed a critical value the effect of theBrønsted sites is dominating and the diffusion of n-hexane in mixtures is two timeslower in HZSM-5 than in the zeolite without acid sites.

A behavior similar to that of n-hexane is observed for 2-methylpentane (Fig. 4-3).Indeed, with increasing loading of 2-methylpentane its self-diffusivity decreases inboth zeolites. As well as for n-hexane, in HZSM-5 the diffusivity of 2-methylpentaneis approximately two times lower than in silicalite at any fraction of 2-methylpentane.Therefore, the presence of the Brønsted sites noticeably decreases the diffusivities ofboth hexanes, probably, because of stronger interactions between the diffusants andthe acid sites, which causes the increased holding time at the adsorption site.

As soon as the concentration of 2-methylpentane in the gas phase reaches a certainvalue (about 0.83 of the total concentration under the conditions of our study), theinfluence of the acid sites on the n-hexane diffusivity is not dominant anymore incomparison with the pore occupation with slow moving 2-methylpentane. Fig. 4-4shows the dependence of the diffusivities of both components versus the concentrationof adsorbed 2-methylpentane in terms of molecules per unit cell. The diffusivitiesof n-hexane in silicalite and HZSM-5 become equal when the concentration of 2-methylpentane reaches approximately 2.75 molecules per unit cell. In the case of2-methylpentane itself, the self-diffusivity in silicalite becomes very close to the valuein HZSM-5 at the same loading (Fig. 4-4).

In Chapter 3 the behavior of binary mixtures of linear and monobranched hex-


Figure 4-4. Self-diffusivities of mixture components in both MFI-type zeolites as a function of

2-methylpentane loading, 433 K.

anes in silicalite have been discussed. It was shown that the zeolite pore networkblockage occurs when the concentration of the branched alkane (2-methylpentane)reaches approximately 2.7 molecules per unit cell. From CBMC simulations, [11], itis known that monobranched molecules such as 2-methylpentane or 3-methylpentane,due to their structure, prefer to occupy the intersections between straight and zigzagchannels in MFI-type zeolites. On the other hand, from the crystallographic zeolitestructure, [16], it is known that the unit cell of MFI-type zeolites has 4 intersections.Therefore, the diffusivity of n-hexane, which prefers to stay in straight and zigzagchannels, sharply decreases when more than half of the intersections are occupied byslow 2-methylpentane.

From Fig. 4-4, it is clear that when the 2-methylpentane loading increases fur-ther (up to 3 molecules per unit cell), the n-hexane diffusivity in silicalite becomesclose to that in HZSM-5. Probably, this is related to the influence of adsorbed 2-methylpentane, which diffusivity is ten times lower, on n-hexane is stronger than theinfluence of the acid sites that causes the only two-times decrease in the diffusiv-ity. Indeed, in silicalite n-hexane molecules move very fast compared to the ones inHZSM-5, but as soon as 2-methylpentane molecules appear the diffusion of linearalkane slows down. At the moment when the majority of the intersections is blockedby iso-hexane, n-hexane molecules are trapped inside zigzag or straight channels. Asa result, n-hexane molecules cannot freely diffuse between the channels. Since thediffusion of 2-methylpentane is significantly slower, the diffusivity of n-hexane in sil-icalite drops till the same value as in HZSM-5. A further decrease in the diffusivitieswith the increasing iso-hexane fraction in both zeolites is caused by the pore networkblockage with 2-methylpentane.


Our observations are confirmed by a recent experimental study of diffusion insilicalite of binary mixtures of alkanes (n-heptane, n-octane) and aromatics (ortho-and meta-xylenes), [4]. Masuda et al. found the same effect of a slower component todecrease the diffusion of the fast component via pore connection blockage, [7]. Theyhave also measured that the diffusion of the slow component is not affected by thepresence of the fast component, which is in agreement with studies of methane/xenonmixtures, [17]. This is in agreement with our study (Fig. 4-2).

So, from the experiments performed on the binary mixtures of 2-methylpentane/n-hexane in HZSM-5 and silicalite, we can conclude that in HZSM-5 both componentsexperience the influence of two factors: interactions with the acid sites, and the con-centration of 2-methylpentane. Interactions with the acid sites causes preferentialadsorption of n-hexane over 2-methylpentane and a decrease in the diffusion of bothcomponents by a factor of two. Increased loading of slow 2-methylpentane diminishesthe influence of the acid sites in HZSM-5, and the diffusion of n-hexane becomesequal in silicalite and HZSM-5 and sharply decreases almost until the value of 2-methylpentane.

4.3.2 Adsorptive and diffusive properties of single compo-nents in HZSM-5 and silicalite

From the experiments with mixtures in both zeolites we concluded that there is aninfluence of the interaction with the acid sites on both components. Therefore, thesingle component experiments have been performed in order to investigate possibledifferences in the diffusion parameters such as the activation energy for both compo-nents. Activated diffusion is described by the Arrhenius-type equation:

D = D0 · exp(− Ea

RT

), (4.3.1)

where D0 is called pre-exponential factor. Ea is the activation energy of diffusion, R isthe gas constant and T the temperature. The pre-exponential factor D0 is related to ajump frequency between adsorption sites in the zeolite lattice, while the exponentialexpresses the chance that the molecules are able to overcome the free energy barrierEa between these sites.

The loadings of n-hexane and 2-methylpentane in HZSM-5 and silicalite also havebeen measured at temperatures between 373 and 533 K with an interval of 20 degrees.The hydrocarbon pressure was taken identical to that in the binary mixture experi-ments (6.6 kPa). The values of the apparent activation energies have been obtained.

n-Hexane

Fig. 4-5 shows the loadings of n-hexane and 2-methylpentane in both zeolites. Undersimilar conditions, the adsorbed concentration of n-hexane is higher than that of 2-methylpentane, especially at lower temperatures. As discussed in Section 4.3.1, for


250 300 350 400 450 500 550

1

2

3

4

5

6

7

C ads [m

ol./u.

c.]

T [K]

n-hexane in HZSM-5 n-hexane in silicalite 2-methylpentane in HZSM-5 2-methylpentane in silicalite

Figure 4-5. Loadings of hexanes measured at various temperatures in silicalite and HZSM-5.

n-hexane molecules the interactions with the Brønsted sites results in higher enthalpyof adsorption, [10]. This leads, under similar conditions, to a higher loadings of n-hexane in HZSM-5 than in silicalite.

From the temperature dependence of the diffusivity of n-hexane in both zeolites,the apparent activation energy has been deduced and the results are shown in Table 4-2. Corresponding Arrhenius plots are shown in Fig. 4-6. The data obtained in thiswork are compared with values found in literature.

The results show that the values of the activation energy measured with theTEX-PEP are very close for both zeolites. Because of the accuracy of the TEX-PEP technique it is difficult to establish a certain difference between H-ZSM-5 andsilicalite (Table 4-2). The experiments performed with the mixtures at 433 K (see Sec-tion 4.3.1), assure us that the difference between the activation energies of n-hexanediffusion in HZSM-5 and silicalite provides a two times decrease in the diffusivity ofn-hexane at 433 K in the presence of the acid sites.

Valyon et al., [15], have measured the activation energy of diffusion for n-butanein silicalite and HZSM-5 with FR method. They found values of 10.7 and 13.1 kJ/molrespectively, which resulted in approximately a double decrease in the n-butane diffu-sivity in the presence of the Brønsted sites. This confirms our observations that thereis an effect of the interactions with the acid sites on the activation energy of alkanediffusion in MFI-type zeolites, but the contribution is not very significant.

Our data are in a good agreement with the values provided by other techniques(Table 4-2). The activation energy of n-hexane diffusion in HZSM-5 equal to 24kJ/mol has been obtained by the FR method, [18]. The constant volume methodprovided the value of 20 kJ/mol, [19]. So, Ea = 22± 2 kJ/mol measured in this workoverlaps very well with these values. The value of the apparent activation energy for


0.0020 0.0025 0.0030

-26

-25

-24

-23ln

D

1/T [1/K]

HZSM-5 (experiment) silicalite (experiment) silicalite (linear fit) HZSM-5 (linear fit)

Figure 4-6. Arrhenius plots for diffusivities of n-hexane in silicalite and HZSM-5, 6.6 kPa.

Zeolite Ea, kJ/mol

TEX-PEP literature

silicalite 18.5± 1.5 16− 19

HZSM-5 22.0± 2.0 20− 24

Table 4-2. Apparent activation energy of diffusion for n-hexane in MFI-type zeolites, comparison

with literature.

silicalite, provided by TEX-PEP method, is confirmed by other methods, such as theZLC-method and the Square wave method, [20, 21], (Table 4-2). The discrepanciesoccur with the techniques such as membrane permeation and TEOM, that provideda somewhat higher values of the activation energy of n-hexane: 34.7 kJ/mol, [22],and 38 kJ/mol, [23], respectively. In the membrane permeation technique very highloadings up to 8 molecules per unit cell were used, which can explain the discrepancywith other techniques. The value provided by Zhu et al., [23], is the activation energydeduced from the corrected diffusivities. The diffusivities were measured at differentconditions (partial pressures) and the loadings were also up to 5.25 molecules perunit cell.

One would expect an increase in the activation energy of diffusion in the presenceof the acid sites. Because of the interactions of alkanes with the Brønsted sites themolecules have to overcome a certain energy barrier to detach from the acid site andhop to the next one. A comparison between the data obtained by different techniquesfor silicalite and HZSM-5 agrees with this presumption, but the difference in thevalues is not very significant to be established with an absolute assurance with the


0.0020 0.0022 0.0024 0.0026-32

-30

-28

-26

-24

lnD

1/T [1/K]

HZSM-5 (experiment) HZSM-5 (linear fit) silicalite (experiment) silicalite (linear fit)

T=393-513K

Figure 4-7. Arrhenius plots for diffusivities of 2-methylpentane in silicalite and HZSM-5, 6.6

kPa.

Zeolite Ea, kJ/mol

TEX-PEP literature

silicalite 66± 6 50, [23]; 46, [24]HZSM-5 72± 3 24, [26]; 36, [25]

Table 4-3. Activation energy of diffusion for 2-methylpentane, comparison with literature.

aid of the PEP technique.

2-methylpentane

Arrhenius plots for the diffusion of 2-methylpentane in silicalite and HZSM-5 arepresented in Fig 4-7. In silicalite, the diffusivities have been measured at 393-513 Kand in HZSM-5 at 413-533 K, respectively.

It turned out that, at elevated temperatures, the signal-to-noise ratio for thePEP-detectors is too low, because the adsorbed concentration of 2-methylpentaneis low, especially on silicalite. Thus, we were not able to measure the diffusivityand the loading of 2-methylpentane in silicalite at 533 K. For the same reason theexperimental error was slightly higher than 10%.

The apparent activation energies of diffusion found for 2-methylpentane in sili-calite and HZSM-5, both are significantly higher than those for n-hexane. This resultsin an order of magnitude difference in the diffusion coefficients for these alkanes. Thisin agreement with the earlier studies of linear and mono- and di-branched alkanes inMFI-type zeolites, [24, 25]. The commonly accepted explanation for the high activa-


Zeolite D(m2/s), T = 423K

Grav., [24] Grav., [25] TAP, [26] PEP

silicalite 2 · 10−12 1.1 · 10−12

HZSM-5 9 · 10−13 1 · 10−12 4 · 10−13

Table 4-4. Diffusion coefficient for 2-methylpentane in MFI-type zeolites, comparison with lit-

erature.

tion energies of diffusion for the iso-alkanes in MFI-type zeolites is that due to itshigher critical diameter, iso-alkanes experience a steric hindrance to their diffusion.

Similar to n-hexane, we find the best estimated value for the apparent activationenergy of 2-methylpentane diffusion in HZSM-5 to be higher than in silicalite (Table 4-3). This is in agreement with the findings of the FR technique for iso-butane diffusionin MFI-type zeolites, [15]. The activation energy of iso-butane was measured to be 1.2kJ/mol higher in HZSM-5. PEP does not allow us to be that precise in this case, butin combination with the data on mixture experiments (Section 4.3.1) it is possibleto claim that the alkane interactions with the Brønsted sites result in the increase inthe activation energy of diffusion.

The values of the activation energy for 2-methylpentane in both zeolites, measuredhere are significantly higher than those measured by other techniques (Table 4-3).TEOM, [23], and Gravimetric measurements, [24], provide an activation energy forbranched hexane diffusion in silicalite of 50 and 46 kJ/mol, respectively, which iseven higher than for HZSM-5 measured by Xiao and Wei with the same method (36kJ/mol, [25]) and by Keipert and Baerns with a transient technique (Ea = 24 kJ/mol,[26]).

The discrepancies between the values of the activation energies provided by dif-ferent authors can be attributed to the different alkane partial pressures used in theexperiments. There are some theoretical and experimental studies which indicate asignificant concentration dependence of diffusion in zeolites, [24, 25, 27, 28, 29]. Cop-pens et al., [29], have shown for ZSM-5 zeolite with Monte-Carlo simulations that thediffusivity can drop by a factor of ten when the occupancy is close to the saturation.In this work we performed our experiments under hydrocarbon partial pressure of 6.6kPa, which is higher than pressure conditions used in the TEOM, gravimetric andvolumetric measurements.

Fig. 4-5 shows the loadings of 2-methylpentane and n-hexane in both zeolites inthe temperature interval used to determine the activation energy. The loading of 2-methylpentane in our experiments reaches 0.5-3.5 molecules per unit cell with theloadings in HZSM-5 slightly higher than in silicalite at the same temperatures. Themaximum loadings of 2-methylpentane in silicalite and HZSM-5 were measured tobe 0.75 mmol/g, which is equal to approximately 4.2 molecules per unit cell (seeTable 4-1). At the partial pressure of 6.6 kPa, the loadings of the zeolites were upto 80% of the saturation, which is higher compared to the experimental conditions


of other techniques. Indeed, in the transient experiments with the TAP reactor, [26],a pulse of a very small amount of molecules is made on the zero-occupancy zeolite,so the influence of the loading on diffusion can be excluded. It is also importantto note that the experimental conditions (partial pressure) used in the gravimetricexperiments, [24, 25], to determine the diffusion parameters such as activation energyare not specified. Therefore, it is actually not very correct to compare these values(Table 4-3), obtained under different conditions. Nevertheless, diffusion coefficientsprovided by those techniques are in a fair agreement with the values measured here.The data are shown in Table 4-4.

Unfortunately, it is not possible to make a definite conclusion from the literaturedata about the relation between the influence of the acid sites and the activationenergy of 2-methylpentane diffusion. Our experimental data indicate that in HZSM-5the best estimated value of the diffusion activation energy for 2-methylpentane isslightly higher than in silicalite, but because of the higher experimental error for2-methylpentane experiments in silicalite, we cannot unambiguously conclude this.

4.4 Conclusions

The diffusion and adsorption of mixtures of 2-methylpentane and n-hexane as well asthat of single components in silicalite and HZSM-5 have been studied with TEX-PEP.Comparison between these zeolites leads to the following conclusions:

1. The interaction with the acid sites causes a decrease in the self-diffusivities ofboth alkanes. The apparent activation energy of diffusion for both alkanes was foundto be very close in silicalite and HZSM-5 at 6.6 kPa.

2. In HZSM-5 the diffusivity of n-hexane in the mixtures with 2-methylpentaneis influenced by two factors: a) interaction with the acid sites which decreases thediffusivity; b) the presence of 2-methylpentane, which has a ten-times lower diffusivity.As soon as the loading of iso-hexane exceeds approximately 2.7 molecules per unitcell, the effect of the Brønsted sites on the diffusion becomes negligible, so that thediffusivities in silicalite and HZSM-5 are equal and decrease further with increasing2-methylpentane content.

3. Interaction with the acid sites causes a preferential adsorption of the linearhexane over the branched one from their mixtures. In HZSM-5, the loadings of n-hexane were higher compared to those in silicalite due to the interactions with theacid sites. On the contrary, 2-methylpentane loadings in mixtures in silicalite andHZSM-5 are very close.

References


[2] Shuring, D.; Koriabkina,A.O.; de Jong, A.M.; Smit, B.; van Santen, R.A. J.Phys. Chem. B B,105(32) (2001) 7690-7698.

[3] Krishna, R.; Vlught, T.J.H.; Smit, B. Chem. Eng. Sci. 54 (1999) 1751-1757.

[4] Masuda, T.; Fujikata, Y.; Ikeda, H.; Hashimoto, K. Microporous MesoporousMater. 38 (2000) 323-332.

[5] Paschek, D.; Krishna, R. Phys. Chem. Chem. Phys. 3 (2001) 3185-3191.

[6] Zikanova, A.; Bulow,M.; Schlodder, H. Zeolites 7 (1987) 115-118.

[7] T. Masuda, Y. Fujikata, T. Nishida, K. Hashimoto, Microporous MesoporousMater. 23 (1998) 157-167.

[8] Shen, D.; Rees, L.V.C. Zeolites 11 (1991) 666-671.

[9] Cunningham, R.H.; Mangnus, A.V.G.; van Grondelle, J.; van Santen, R.A. J.Molec. Catal.A: Chem. 107 (1996) 153-158.

[10] Eder, F. Thermodynamics and siting of alkane sorption in molecular sieves, 1996,Ph.D.Thesis, Twente: University of Twente.

[11] Vlugt, T.J.; Krishna, R.; Smit, B. J. Phys. Chem. B B,103 (1999) 1102-1118.

[12] Wu, P.; Debebe, A.; Ma, Y. Zeolites 3 (1983) 118.

[13] Anderson, J.; Foger, K.; Mole, T.; Rajadhyaksha, R.; Sanders, J. J. of Catal. 58(1979) 114.


[15] Valyon, J.; Onyestyk, Gy.; Rees, L.V.C. in proc. of 2nd Pac. Basin. Conference,2000, 482-486.

[16] Baerlocher, Ch.; Meier, W.M.; Olson, D.H. Atlas of zeolite framework types, 2001,Elsevier.

[17] Jost, S.; Bar, N.-K.; Fritzshe, S.; Haberlandt, R.; Karger,J. in proc. of 12thInternational Zeolite Conference, 1999, 149-152.

62 REFERENCES

[18] Bulow, M.; Schodder, H.; Rees, L.V.C.; Caro, J.; Richards, R. in proc. of 7thInternational Zeolite Conference, 1986,Tokyo.

[19] Hashimoto, K.; Masuda, T.; Murakami, N. in Zeolite Chemistry and Catalysis,eds P.A. Jacobs et al., 1991, 477-484.

[20] van den Begin, N.; Rees, L.V.C. Caro, J.; Bulow, M. Zeolites 9 (1989) 287.

[21] Eic, M.; Ruthven, D.M. Stud. Surf. Sci. 49 (1989) 897.

[22] Millot, B.; Methivier, A.; Jobic, H.; Moueddeb, H.; Dalmon, J.A. MicroporousMesoporous Mater. 38 (2000) 85-95.

[23] Zhu, W,; Kapteijn, F.; Moulijn, J.A. Microporous Mesoporous Mater. 47 (2001)157-171.

[24] Cavalcante, C.L.; Ruthven, D.M. Ind. Eng. Chem. Res. 34 (1995) 185.

[25] Xiao, J.; Wei, J. Chem. Eng. Sci. 47 (1992) 1143.

[26] Keipert, O.P.; Baerns, M. Chem. Eng. Sci. 53 (1998) 3623.

[27] Chen, N.Y.; Degnan, T.; Smith, C.Molecular transport and reaction in zeolites,1994, John Wiley & Sons.

[28] Coppens, M.-O.; Bell, A.T.; Chakraborty, A.K. Chem. Eng. Sci. 54 (1999) 3455-3463.


5 Diffusion of 3-methylpentane insilicalite: concentration dependence

The influence of the pore occupancy on the intracrystalline diffusion of 3-methyl-pentane in silicalite-1 has been studied at 403-483 K. A monotonous decrease in thediffusivities of alkane was observed with an increasing zeolite loading. The apparentactivation energy of diffusion increased dramatically with increasing partial pressure:from 44 kJ/mol at zero partial pressure to 80 kJ/mol at 4.5 kPa. Analysis of theexperimental data revealed that the concentration dependence of the diffusivity forthis particular system can be approximated by an exponential decay rather than bya linear dependence (1− θ), where θ is a pore occupancy.

5.1 Introduction

For several decades, many studies of the diffusive and adsorptive properties of alka-nes in MFI-zeolites have been carried out, [1]. These materials are of great inter-est in the petroleum industry as catalysts for hydrocarbon conversion processes. 3-Methylpentane is one of the main products of the hydroisomerization of n-hexane,an important reaction to upgrade the gasoline octane number. In Chapter 3 study ofthe diffusion of n-hexane/iso-hexane mixtures in silicalite has been discussed. It hasbeen shown that the diffusion of both alkanes is controlled by the concentration of theslow component. Therefore, the concentration dependence of the diffusion coefficient,especially for bulky iso-hexane is an important issue. Silicalite was used as a modelof the active zeolite matrix. This all-silica zeolite does not contain acid sites, whichexcludes any possible influence on the diffusion process, [2].

In the literature, there are five different phenomena mentioned, that describe theconcentration dependence of self-diffusion in the zeolites, [3]: 1) the diffusivity de-creases monotonically with occupancy; 2) the diffusivity is constant up to mediumpore filling and at high loadings a decrease in the diffusivity can be observed; 3) at lowconcentration the diffusivity monotonically increases and then has a constant valuewith further increase in the loading; 4) the diffusivity has a maximum at certain porefilling; 5)the diffusion coefficient monotonically increases with the concentration. Usu-ally, the concentration dependence of type-1 is observed for saturated hydrocarbonsin silicalite.

FR method, [4, 5], has been used to measure the decrease in the diffusivity forn-hexane in silicalite with loading, which is in agreement with the results obtained byMicke et al. from the sorption uptake kinetics, [6]. For smaller alkanes like methane,ethane and propane in silicalite NMR has shown the diffusivities to reduce with theconcentration, [7]. Kinetic Monte-Carlo simulations performed by Coppens et al.,

64 Chapter 5: Diffusion of 3-methylpentane in silicalite

[8], have also shown that in MFI-type zeolites, the diffusivity strongly depends onthe loading and decreases several orders of magnitude as the occupancy of silicaliteincreases. Similar behavior was found for the diffusion of linear and branched alkanesin various zeolites using MD simulations, [9].

In the past years, several theories have been developed trying to derive an ex-pression, which would be able to predict and describe the concentration dependenceof the diffusion in zeolites. In these theories, diffusion is assumed to proceed via thesequence of the activated jumps between the adsorption sites. One of the first predic-tions has been made in 1960 using mean-field theory, [10], according to which a lineardecrease with the pore occupancy can be observed. The diffusivity would be propor-tional to the fraction of unoccupied adsorption sites 1− θ, reflecting the probabilitythat a neighbor adsorption site is free for the molecule to jump. Later, Monte-Carlosimulations revealed deviations from this linear dependence, [8, 11, 12, 13], show-ing a stronger decrease in the diffusivity with pore occupancy for MFI-type zeolites.In these studies the topology of this particular zeolite has been taken into account.Correlation effects were found to be more significant for the molecular diffusion inthe poorly connected zeolite matrixes such as silicalite, so that the diffusion coef-ficient decreased faster with the zeolite loading, [8]. A unified model based on therandom walk/hopping mechanism has been offered by Chen and Yang, [14], capableto describe both decreasing and increasing concentration dependencies.

In the present experimental study of the concentration influence on self-diffusion of3-methylpentane in silicalite, we have tried to establish this dependence and analyzeit with existing models. Factors influencing the apparent activation energy of diffusionhave been also studied.


The PEP experiments has been performed in the pulse mode for this study in or-der to obtain the diffusivity under zero loading. Silicalite sample described in Sec-tion 2.2.1 has been used. During the experiments, a constant flow of non-labelled3-methylpentane in hydrogen was fed in a plug flow reactor containing the zeolitesample. The partial pressure of the alkane was set by varying the ratio between thealkane gas flow and hydrogen flow (with the total flow kept constant and equal to150 Nml/min) using a Bronkhorst CEM system. Under zero partial pressure con-ditions, only the hydrogen as a carrier gas was fed into the reactor with the flowrate of 150 Nml/min. In order to measure the diffusivity of the alkane, a pulse con-taining approximately 10−15 moles of the 11C-labelled 3-methylpentane is injected inthe flow passing through the reactor. The development of the pulse (change of theradio-labelled concentration profile in time) is monitored using PEP detectors. Thebed length of the reactor was 30 mm, so 10 detection positions were involved. Thediffusivities of 3-methylpentane have been measured at temperatures from 403 to 483K at a hydrocarbon partial pressures from 0 to 4.5 kPa with a constant total flowrate of 150 Nml/min.

Section 5.2: Experimental section 65

Figure 5-1. Example fit of a pulse of labelled 3-methylpentane in silicalite, T=443 K.

Each plot on Fig. 5-1 shows changing concentration of radio-labelled alkane atthe certain detection position in time and the corresponding modelled concentrationprofile. Model adjusted for the pulse experiments as described in Chapter 2 has beenapplied for the fitting of the experimental data. Modelling translates the measurementerror to the following error on the fitted parameters. The diffusivities were estimatedfrom the fitting with the error of approximately 10%.

Isotherms of adsorption of 3-methylpentane in silicalite

In order to perform some of the PEP experiments under isosteric conditions, thepartial pressure of the component should be known to obtain the needed loading.Therefore, isotherms of adsorption are required. The isotherms of adsorption for 3-methylpentane in silicalite have been measured using a Balzers Quadrupole mass-spectrometer system QMG-420 at 403 - 483 K. The method is given in Section 2.4.The alkane loading at a given partial pressure and temperature is determined asthe amount of the desorbed hydrocarbon per gram of the zeolite. The adsorptionisotherms have been fitted using a Langmuir curve (Fig. 5-2). Although the dual-siteLangmuir model is applied to describe the isotherm of adsorption for iso-hexanes,[15], we used a simple Langmuir model because the partial pressure conditions of thisstudy (0-4.5 kPa) were significantly lower compared to those required to observe aninflection in the isotherm.

Fitting the data provided the equilibrium constants of adsorption. The maximumloading was found to be Cmax = 0.75± 0.05 mmol/g. This value is close to approxi-mately 4 molecules per unit cell, which is predicted by CBMC simulations, [15], as theloading for the inflection to occur. This is in a good agreement with Zhu et al., [16],and Cavalcante et al., [17]. The authors obtained 0.69 and 0.7 mmol/g as maximum


Figure 5-2. Isotherm of 3-methylpentane adsorption in silicalite at 403 K.

loading for 3-methylpentane in silicalite using TEOM and a gravimetric techniques,respectively.

From the temperature dependence of the fitted adsorption constant (Fig. 5-3) theenthalpy of adsorption has been determined to be 65 ± 5 kJ/mol. The heat of ad-sorption obtained experimentally by different techniques are of the same value: 66.4(inverse chromatography technique, [18]), 60 (wall-coated capillary chromatographicmethod, [19]) and 62.7 kJ/mol (gravimetric sorption uptake method, [17]). Theoret-ical calculations from June et al., [20], provided the value of 63 kJ/mol, which alsosupports the experimental data of this study. Thus, the adsorption isotherm datameasured in this work agree with those from other techniques and will be furtherused for the diffusion data interpretation.


5.3.1 Influence of the concentration on the self-diffusivity

Concentration dependence of the self-diffusivity

The self-diffusivity of 3-methylpentane in silicalite as a function of loading is shownin Fig. 5-4. For clarity sake, it is shown in both logarithmic and linear scales. Theself-diffusivity monotonically decreases with the loading, which corresponds to thetype-1 concentration dependence, [3], mentioned above. Quite an opposite effect ofthe pore occupancy was observed by the Inverse Chromatography technique, [18], for3-methylpentane as well as for iso-pentane and 2,2-dimethylbutane. Most probably,this was due to the small size (2µm) of silicalite crystals used in the experiments, sothe process was not controlled by the intracrystalline diffusion.


Figure 5-3. Arrhenius plot of fitted Langmuir adsorption constants for 3-methylpentane in sili-

calite.

The decrease in the diffusivity with the pore occupancy was observed for other hy-drocarbons in zeolites and predicted from some theoretical simulations. As has beenmentioned above, there are five types of concentration dependencies of the diffusionand it is mainly determined by the zeolite-adsorbate and adsorbate-adsorbate inter-actions, [3]. Dynamic Monte-Carlo simulations and mean-field studies performed byCoppens et al., [2], recovered all five types of diffusional behavior and it was shownthat in silicalite due to the absence of interactions with acid sites a monotonous de-crease of the self-diffusion coefficient can be observed. This is in agreement with ourobservations as well as with the results of MD simulations performed by Schuringet al., [9], that showed a decrease in the diffusivity for n-butane in silicalite.

In general, the diffusion mechanism is considered to be a sequence of activatedjumps from one adsorption site to another. Activated diffusion is described by Eq. 4.3.1.The pre-exponential factor D0 is related to a jump frequency between adsorptionsites in the zeolite lattice, [21], while the exponential expresses the chance that themolecules are able to overcome the free energy barrier Ea between these sites.

Observed decrease of the diffusion coefficient with the pore occupancy might becaused by the concentration dependence of the activation energy of diffusion or by de-crease of the jump frequency. According to the MD simulations, [9], the potential en-ergy of the alkane-alkane interaction increases monotonically with the loading, whilethe interaction between the alkane and zeolite lattice remains approximately con-stant. This means that the diffusivity in silicalite decreases because molecules hindereach other at high pore occupancies and the activation energy does not change withthe occupancy. Mean-field theory predicted a very general behavior for the moleculesin zeolites: a molecule can only hop to a neighbor site if this site is vacant, the averageprobability of this is (1−θ), where θ is the pore occupancy. Thus, a linear decrease in


Figure 5-4. Concentration dependence of the self-diffusivity of 3-methylpentane in silicalite (top

plot - logarithmic scale; bottom plot - linear scale).

the diffusion with the loading can be expected: D(θ) = D(0)·(1−θ), where D(θ), D(0)are the self-diffusivities at θ and zero pore occupancies, respectively. Lately, a numberof theoretical studies showed that a negative deviation from this linearity can be ex-pected. One of the main arguments is that the molecular displacements are correlatedto each other and zeolite topology has to be taken into account, [8]. Indeed, Coppenset al., [8], showed that the diffusivity in MFI-type zeolites decreases even strongerwith the occupancy because of the low lattice connectivity. The best approximationdescribing the concentration dependence of the diffusion coefficient calculated fromdynamic Monte-Carlo simulations would be an exponential decay. Earlier, Theodorouet al., [11], modelled diffusion in ZSM-5 zeolite as a random walk process in a two-dimensional pore network and a deviation from the linear dependence similar to theone revealed by Coppens et al., [8], was also observed. Tsikoyiannis and Wei, [12],used Monte-Carlo simulations to calculate the dependence of the diffusivity on the


occupancy and they found for the cubic and square lattices that the dependence isdescribed by a function that lies somewhere between (1− θ) and (1− θ)2.

The experimentally obtained concentration dependence of the diffusivity of 3-me-thylpentane in silicalite shown in Fig. 5-4 can not unambiguously be described aslinear, and the deviation discussed above is not likely to be observed. This can bedue to the limited experimental accuracy. At lower temperatures, a bit slower thanlinear decrease in the diffusivity with the pore occupancy was observed, while athigher temperature (483 K) it could be described as linear. Thus, from Fig. 5-4 itis difficult to draw clear conclusion about the function most suitably describing theexperimental data.

Nevertheless, it is important to establish wether the activation energy of diffusionis increasing with the pore occupancy of the zeolite thus causing the observed decreasein the diffusion.

5.3.2 Activation energy of diffusion

Influence of the loading

In this section the factors influencing the activation energy of diffusion, such as par-tial pressure and zeolite pore occupancy, will be discussed. In some particular cases,activation energy was influenced by the zeolite loading. Indeed, MD simulations per-formed by Schuring et al., [9], for n-butane in mordenite demonstrated that withan increased pore loading the potential energies of alkane-lattice and alkane-alkaneinteractions change, and an increase in the activation energy of diffusion is observed.A double increase in the activation energy of diffusion occurred for n-butane in mor-denite from the MD simulations, when the pore occupancy reached 0.8 from themaximum loading, [9].

In order to investigate the influence of the silicalite pore occupancy, the activationenergies of diffusion measured at two significantly different degrees of pore filling havebeen compared. Zero and approximately 50% pore occupancies were used. To measurethe diffusivities at a fixed loading, partial pressures were chosen as to ensure that atall temperatures it was fixed at approximately 0.35 mmol/g. The value of the partialpressure of 3-methylpentane corresponding to the particular loading was obtainedfrom the isotherms of adsorption (see Section 5.2).

In Fig. 5-5 the Arrhenius plots obtained for both pore occupancies are shown.The activation energies of diffusion determined from the slope of the Arrhenius plotat high and zero loadings are equal to 46± 6 kJ/mol and 44± 4 kJ/mol, respectively,meaning that for 3-methylpentane in silicalite the activation energy does not dependon the pore occupancy. Heink et al., [22], using PFG NMR observed a similar picturefor C3−C6 n-alkanes in silicalite. For these alkanes the diffusivities decrease with theloading, while the activation energies were found to be independent.

The Arrhenius plots in Fig. 5-5 corresponding to different loadings (0 and 0.35mmol/g, respectively) are parallel to each other, so that the pre-exponential factor D0

is smaller for high pore occupancy than that for empty zeolite (smaller value of the


Figure 5-5. Arrhenius plots for the self-diffusivity of 3-methylpentane in silicalite at zero and

constant occupancy of approximately 0.47 from the maximum loading.

intercept (ln D0) corresponds to a higher loading). This trend is also in agreement withthe experiments performed for C1−C6 n-alkanes in silicalite using PFG NMR, [22]. Forthose alkanes, the activation energy did not change, while the pre-exponential factordecreased as the zeolite loading increased. Apparently, the concentration dependenceof the diffusion coefficient is related to the change in the jump frequency but not tothe activation energy. Therefore, the pre-exponential factor (see Eq. (4.3.1)) dependson the pore occupancy, which probably means that with increased occupation of theadsorption sites the jump frequency decreases because the probability of the successfuljump decreases.

Comparison with literature data

The influence of the concentration on the diffusivity makes the comparison of thediffusional parameters with the literature more difficult, because the experimentalconditions might be the reason for the discrepancies. The most common expressionused to extract self-diffusion coefficients from the transport diffusivities obtainedfrom macroscopic measurement of uptake and permeation rates, or from the analysisof frequency responses, is the Darken equation (see Eq 1.2.3). Values of correcteddiffusivity are used to compare with the self-diffusivities, which is not really correctas was shown by Maginn et al., [23], and recently by Paschek and Krishna, [24]. Inthese studies, the discrepancies between the self-diffusion coefficient and the correcteddiffusivity at non-zero pore occupancies have been demonstrated. Linear dependenceof the corrected diffusivity on the pore occupancy (1 − θ) and stronger decrease forself-diffusion for methane and 2-methylhexane in silicalite were observed using kineticMonte-Carlo simulations performed by Paschek and Krishna, [24]. Therefore, when


Method D × 10−13, m2/s conditions

PEP 3.1 423 K, 4.5 kPaPEP 2.9 403 K, 1.3 kPa

Membrane Permeation, [25], 2.47 423 K, 0.1 barInverse Chromatography, [18], 0.007 423 K, 0

TEOM, [26], 4.20-4.71 408 K, 0.94-6.35 kPa

Table 5-1. Diffusion coefficient for 3-methylpentane in silicalite obtained by various methods at

various partial pressure conditions.

Method Ea, kJ/mol

PEP 44± 3Membrane Permeation, [25], 50

Gravimetry, [27], 58Wall-coated capillary chromatography, [19], 29

Inverse Chromatography, [18], 56TEOM, [26], 52.4

Table 5-2. Apparent activation energies of 3-methylpentane diffusion in silicalite obtained by

various methods.

comparing the diffusivities obtained by different methods, one should keep in mindthat both corrected and self-diffusivities are concentration dependent and are notnecessarily close to each other unless measured at zero loading.

The self-diffusivity of 3-methylpentane in silicalite measured here have been com-pared with those from other methods determined under close conditions. The dataare shown in Table 5-1. Apparently, diffusivities measured with PEP technique are inreasonable agreement with those from other studies, except for the self-diffusion co-efficient determined by Inverse Chromatography technique, [18], that is significantlylower compare to others. The reason for this disagreement is not clear.

The values of the activation energy of 3-methylpentane diffusion in silicalite mea-sured by different authors are shown in Table 5-2. The activation energy obtainedfrom the corrected diffusivities by Millot et al. using the Membrane permeation tech-nique, [25], is equal to 50 kJ/mol, which is quite close to the value of the activationenergy measured here. Deviation from the linear dependence was observed for thelogarithm of the diffusivity vs reciprocal temperature, which was attributed by theauthors to the strong variation of diffusion coefficient with the coverage.

One can find in the literature activation energy values essentially different fromeach other. Gravimetric measurements on large crystals give the corrected activationenergy equal to 58 kJ/mol, [27], partial pressure conditions were not specified. Two-times lower activation energy (29 kJ/mol) was found by the wall-coated capillarychromatographic method, [19]. Inverse Chromatography experiments provided the


activation energy of diffusion at zero occupancy of 56 kJ/mol, [18], which is slightlyhigher than measured in this study. Modelling of the data used in the Inverse Chro-matography method implies intracrystalline diffusion limitations, which probably wasnot the case, since the size of silicalite crystals (2 µm) used in the study, [18], was notbig enough. Probably, due to the absence of the intracrystalline diffusion limitationsthe increase in the diffusivity with the loading was observed as well as the deviationof the data from the Arrhenius dependence. Zhu et al., [26], calculated 52.4 kJ/molas the value for the activation energy of 3-methylpentane diffusion from the cor-rected diffusivities measured by TEOM on large silicalite crystals (length of 120µm).In this study, [26], the activation energy was obtained from the diffusivities mea-sured at different partial pressure and pore occupancy conditions. From TEOM data,[26], we have chosen the diffusivities measured at similar loadings (approximately 3.8mol/u.c.) at different temperatures to estimate so-called true activation energy, thatis determined under constant loading to exclude its influence. The value obtainedturned out to agree very well with the activation energy measured in this study:46± 3 kJ/mol and 44± 3 kJ/mol (zero loading), from TEOM and PEP, respectively.

Therefore, due to the concentration dependence of the diffusivity, the ”true” ac-tivation energy of diffusion can only be measured at zero pore occupancy or underconstant loading conditions in order to eliminate a possible influence of the experi-mental conditions.

Influence of the partial pressure

In the macroscopic experiments the activation energy is usually measured under con-stant partial pressure of the component. This has also been done in this study withinthe temperature range from 403 to 483 K.

As the pressure increases from 0 kPa to 4.5 kPa (Fig. 5-6), an unusual effect ofthe partial pressure of 3-methylpentane on the apparent activation energy of diffusionwas observed: a significant rise from 44 kJ/mol to 80 kJ/mol occurred. The values ofthe apparent activation energy measured at different partial pressures of alkane areshown in Table 5-3.

Significantly lower activation energies (50 kJ/mol) compared to the values re-ported here for 4.5 kPa, were measured at 10 and at 6 kPa of feed pressure with theMembrane permeation technique, [25]. The diffusivities measures with this techniquewere almost equal at these partial pressures, [25], which means that there was no con-centration effect observed. Probably, this is due to loadings that were high enough sothat the diffusivity did not change very much.

We believe, the observed effect of the partial pressure on the apparent activationenergy (Fig. 5-6) is a consequence of the concentration dependence of the diffusivity.As the concentration inside zeolite pores under given partial pressure depends on thetemperature, the combined effect of temperature and loading dependence is measured.With increasing temperature, the loading decreases. According to the observed effectof the loading on the diffusion coefficient (Fig. 5-4), in addition to the increasedmobility of the molecules due to the higher temperature, diffusivity also increases


Figure 5-6. Influence of the partial pressure on the 3-methylpentane apparent activation energy

of diffusion in silicalite.

Pressure, kPa Ea, kJ/mol

0 44± 31.33 65± 42.00 61± 33.70 70± 54.50 80± 6

Table 5-3. Apparent activation energies of 3-methylpentane diffusion in silicalite at different

partial pressures.

due to the decreased loading, which is caused by the increase in D0, related to thejump frequency. As a result, the apparent activation energy can be much higherthen the real activation energy, and this value depends on the hydrocarbon partialpressure, at which the measurements are performed.

5.3.3 Concentration dependence of the self-diffusivity

It was experimentally established that for 3-methylpentane in silicalite, the depen-dence of the apparent activation energy of diffusion on the partial pressure is causedby the concentration dependence of pre-exponential factor D0(θ). Therefore, fromthe diffusivities at given loadings and the true activation energy we can estimatethe concentration dependence of the pre-exponential factor, D0(θ). It is derived fromEq. (4.3.1):


D0(θ) = D(θ) · exp(

Ea

RT

), (5.3.1)

where D0(θ) is pre-exponential factor that corresponds to the diffusivity D(θ) at poreoccupancy θ and temperature T , Ea is activation energy measured at zero or constantloading.

Several predictions have been made on the concentration dependence:

• Mean-field theory predicted a linear decrease in the diffusivity proportional to(1−θ), which is the probability factor for the molecule to jump into the neighboradsorption site, [10]. It does not take into account zeolite topology or specificinteractions between the molecule and zeolite:

D(θ) = D(0) · (1− θ), (5.3.2)

with the diffusivity at zero pore occupancy D(0):

D(0) = D∞ · exp(− Ea

RT

). (5.3.3)

In this expression, D∞ is a diffusion coefficient at infinite temperature in the ze-olite with zero loading. Therefore, the pre-exponential factor D0(θ) is expressedfrom Eqs. (5.3.1), (5.3.2) and (5.3.3) as follows:

D0(θ) = D∞ · (1− θ). (5.3.4)

• Kinetic Monte-Carlo simulations predict deviations from the linear dependencefor MFI-type zeolite due to its poorly connected lattice, [8]. The average connec-tivity of silicalite is 2.67 if the molecule can jump to any adsorption site: zigzag,straight channels or the intersections. In this particular case of 3-methylpentanein silicalite, the connectivity is 4, because according to Vlugt et al., [15], chan-nel intersections are the most favorable locations for monobranched hexanes.Only very high pressure is required for the molecules to adsorb in the channels,[15]. For this connectivity the deviation from the linear dependence is expectedto be very small, [8]. Simulated dependence can be approximated in general byan exponential decay:

D(θ) = D(0) · exp

(−θ

b

); (5.3.5)

D0(θ) = D∞ · exp

(−θ

b

). (5.3.6)

b is a fitted coefficient that depends on the connectivity for the molecule/zeolitesystem. The diffusivity at zero coverage depends on dimension of the latticeonly, because there are no interactions between the molecules.


Figure 5-7. Pre-exponential factor D0(θ) for 3-methylpentane self-diffusion in silicalite vs. zeolite

loading. The data fitted by theoretically predicted dependencies.

• From calculations based on transition state theory a general formulae was de-rived to describe the concentration dependence of surface and zeolitic diffusions,[14]:

D(θ)

D(0)=

1− θ + 12λθ(2− θ) + H(1− λ) · (1− λ)1

2λθ2

(1− θ + 12λθ)2

; (5.3.7)

D0(θ) = D∞1− θ + 1

2λθ(2− θ) + H(1− λ) · (1− λ)1

2λθ2

(1− θ + 12λθ)2

. (5.3.8)

In this expression, λ is a blockage parameter and is equal to the ratio betweenthe rate constant of return migration due to blockage by another molecule andthe rate constant of forward migration. λ is equal to zero for surface diffusionand λ > 0 for zeolitic diffusion. Proximity of the sizes of adsorbate and zeolitepore should result in a large values of λ. H(1 − λ) is the Heaviside step func-tion. Eq. (5.3.7) was claimed to describe all five possible types of concentrationdependence of diffusion in zeolites discussed in the work of Karger and Pfeifer,[3].

Fig. 5-7 shows the attempt to fit the pre-exponential factor calculated from theexperimental data according to Eq. (5.3.1) with the Eqs. (5.3.4), (5.3.6) and (5.3.8).In Table 5-4 one can find the characteristics of fitting: χ2 chi square value of fit andCoefficient of Determination (R2). The minimum value of χ2 and higher value of R2

correspond to a better fit.


Function χ2 R2

Mean-field 1.6558 · 10−15 0.85388Monte-Carlo 1.07623 · 10−15 0.91233

Transition state theory 1.7439 · 10−15 0.8078

Table 5-4. Fitting characteristics for the functions predicted by theoretical calculations.

Apparently, the stretched exponential approximation predicted by Kinetic Monte-Carlo simulations, [8], seems to be a slightly better fit for the experimental data: firstorder decay with the fitted coefficient b = 3.239. Even though Eq. (5.3.7) failed todescribe the data for 3-methylpentane in silicalite, the fitted value of λ = 10.02 is inline with the values found by Chen and Yang, [14], for diffusion of thriethylamine in13X zeolite and for benzene in ZSM-5, that are equal to 10.2 and 2.11, respectively.

According to Coppens et al., [8], the difference between mean-field theory andMonte-Carlo simulations should be insignificant for the investigated system, whichis indeed observed here. A stronger decrease in the self-diffusion with the occupancyis caused by the higher probability of the return jump, because the probability fora previously visited site to be vacant is higher than that for any other site. On thecontrary, mean-field theory predicts that the probability to hop to any of the neighborsites is equal and zeolite topology is not considered. This is a main reason for thecontradictions in these two theories. According to kinetic Monte-Carlo simulations,the probability of the successful jump depends also on the connectivity between theadsorption sites. For the intersections the connectivity factor is higher, which makesthe deviations from the mean-field theory dependence smaller. Indeed, one can seefrom the fit (Fig. 5-7) that both functions are quite close to each other. Unfortunately,there is no possibility to compare obtained coefficient b with the literature, since theexponential dependence is just an approximation of the simulated effect of the zeoliteloading on the diffusion coefficients. Nevertheless, we tried to compare the approxi-mated dependencies with those for the diamond lattice that also has a dimension of 3and connectivity of 4, which is similar to those of 3-methylpentane in silicalite. In thediamond lattice, the deviations from a linear dependence are very insignificant, [8],and the fitting provide a coefficient b = 2.753. The more essential are the correlationeffects, so that the diffusivity decreases faster with the concentration, the smallershould be the value of b. Apparently, for the studied case, the deviations from thelinear dependence of the diffusion coefficient on the pore occupancy are within theexperimental accuracy. Only based on the data analysis, one may conclude that aslight deviation from the linear dependence is observed.


5.4 Conclusions

Using the PEP technique, the concentration dependence of 3-methylpentane in sili-calite has been studied. The self-diffusion coefficient can be best described as monoto-nously decreasing as the pore occupancy increases, which corresponds to type-1 of theconcentration dependence according to Karger and Pfeifer, [3]. Since the activationenergy of diffusion of the alkane turned out to be independent on the pore occupancy,the jump frequency related pre-exponential factor D0 is concentration dependent. Formolecular diffusion in silicalite, theoretical studies predict a linear decrease of the dif-fusion coefficient D(θ) = D(0) · (1 − θ) (mean-field theory) and negative deviationfrom linear behavior, which can be approximated by an exponential decay (Eq. 5.3.5)that is an approximation of Monte-Carlo simulations. Analysis of the experimentaldata showed that the exponential is a slightly better fit for the experimental data.Moreover, the theoretically predicted deviations from the linear dependence are alsovery small for this system. Since 3-methylpentane molecules most probably jump tothe channel intersections only, the connectivity of the system is 4, which leads to asmaller deviation from the linear dependence.

As a consequence of the concentration dependence of the pre-exponential factorD0(θ), the apparent activation energy increases with partial pressure. Usually, inthe macroscopic experiments the measurement of the activation energy is performedunder fixed partial pressure conditions. If the conditions are different from those usedin other studies it can be the reason for the discrepancies in the apparent activationenergy and diffusivities obtained by different authors. Values of activation energy ofdiffusion should be compared only measured at zero pore filling or under fixed loadingconditions.

References




[4] Bulow, M.; Schodder, H.; Rees, L.V.C.; Caro, J.; Richards, R. in proc. of 7thInternational Zeolite Conference, 1986,Tokyo.

[5] van den Begin, N.; Rees, L.V.C. Caro, J.; Bulow, M. Zeolites 9 (1989) 287.

[6] Micke, A.; Bulow, M.; Kocirik, M.; Struve, P. J. Phys. Chem. B 98 (1994) 12337-12344.

[7] Caro, J.; Bulow, M.; Schimer, W.; Karger, J.; Heink, W.; Pfeifer, J. J. Chem.Soc. Faraday Trans. 81 (1985) 2541.


[9] Schuring, D.; Jansen, A.P.J.; van Santen, R.A. J. Phys. Chem. B 2000 (104)941-948.

[10] Jost, J. Diffusion in solids, liquids, gases, Academic Press, New York, 1960.

[11] Theodorou, W.; Wei, J. J. Molec. Catal.A: Chem. 83 (1983) 205-224.

[12] Tsikoyiannis, J.G.; Wei, J. J. Chem. Eng. Sci. 46 (1991) 233.

[13] Trout, B.L.; Chakraborty, A.K.; Bell, A.T. Chem. Eng. Sci. 1997 (52(14)) 2265-2276.

[14] Chen, Y.D. and Yang, R.T. AIChE Journal 1991 (37(10)) 1579-1582.

[15] Vlugt, T.J.; Krishna, R.; Smit, B. J. Phys. Chem. B 103 (1999) 1102-1118.


[17] Cavalcante, C.L.; Ruthven, D.M. Ind. Eng. Chem. Res. 34 (1995) 177-184.

[18] Jolimaitre, E.; Tayakout, M. Jallut, C.; Ragil, K. Ind. Eng. Chem. Res. 40 (2001)914-926.

80 REFERENCES

[19] Jama, .M.A.; Delmas, M.P.F.; Ruthven, D.M. Zeolites 18 (1997) 200-204.

[20] June, R.L; Bell, A.T.; Theodorou, D.N. J. Phys. Chem. B 94 (1990) 1508-1516.

[21] Chen, N.Y.; Degnan, T.; Smith, C. Molecular transport and reaction in zeolites,1994, John Wiley & Sons.

[22] Heink, W.; Karger, J.; and Pfeifer, H.; Datema, K.P.; Nowak, A.K. J. Chem.Soc. Faraday Trans. 1992 (88) 3505-3509

[23] Maginn, E.J.; Bell, A.T.; Theodorou, D.N. J. Phys. Chem. B 97 (1993) 4173-4181.



[26] Zhu, W.; Kaptein, F.; Moulijn, J.A. Microporous Mesoporous Mater. 47 (2001)157-171.

[27] Cavalcante, C.L.; Ruthven, D.M. Ind. Eng. Chem. Res. 34 (1995) 185-191.

6 Factors enhancing diffusion ofn-hexane in MFI-zeolites

The concentration dependence of the self-diffusivity of n-hexane in large crystals ofsilicalite and HZSM-5 zeolites has been investigated with PEP in the temperaturerange of 393-483 K. A monotonous increase in the self-diffusivity was observed withincreasing alkane loading for both zeolites up to 4 molecules per unit cell. The diffusioncoefficient in HZSM-5 was approximately two times lower compared to silicalite dueto the interaction of n-hexane with the Brønsted acid sites. The increase in thediffusivities with the loading is ascribed to the repulsive interactions between themolecules adsorbed in the channel intersections and the adjacent straight channel.The apparent activation energy of self-diffusion was measured to be independent ofthe partial pressure of alkane, while the jump frequency increased. In HZSM-5 zeolitethe apparent activation energy decreased with the partial pressure due to strongerinteractions with the acid sites at low pressures.

6.1 Introduction

It is a well established phenomenon that diffusion in zeolites depends on the con-centration of the adsorbate. There are five types of the experimentally observed con-centration dependencies of self-diffusion in the zeolites have been described in theliterature, [1]. Usually, diffusion of n-hexane in silicalite is found to decrease with thezeolite pore occupancy.

Indeed, using FR method,[2, 3], a decrease in the diffusivity for n-hexane in sil-icalite with loading was observed. The diffusivities of smaller alkanes like methane,ethane and propane in silicalite also reduce with increasing concentration as shownby Caro et al., [4]. A number of theoretical studies have been devoted to this issue,[5, 6, 7, 8]. Kinetic Monte-Carlo simulations, [5], have shown that in MFI-type ze-olites, the diffusivity strongly depends on the loading and the zeolite topology. Forsilicalite, the diffusivity decreases with loading and the dependence can be approxi-mated by a stretched exponential, [5]. A similar behavior was found for the diffusionof linear and branched alkanes in various zeolites using MD simulations, [9].

In silicalite, which consists of straight and zigzag channels, the diffusion coefficientobtained from experimental observations or deduced from computer simulations is theaverage value of the contribution of the diffusivities in both channels, which in theirturn depend on the pore occupancy. Recently, some theoretical, [10], and experimen-tal, [11, 12], studies indicate different diffusional behavior of hydrocarbon moleculesin straight and sinusoidal channels of silicalite-1. Using the FR method, which isable to distinguish diffusion in different channels, Song and Rees, [11], observed an

82 Chapter 6: Factors enhancing diffusion of n-hexane

increase in n-hexane diffusivity in straight channels of silicalite with the total concen-tration. On the other hand, an average n-hexane diffusivity determined from NMRexperiments, [13], under similar conditions decreased with concentration. The detailsof the diffusion mechanism are still not completely understood. In HZSM-5 zeolite,the diffusivity is also influenced by the acid sites as well as geometrical peculiaritiesof the zeolite framework. In the presence of protons, as revealed by Monte-Carlo sim-ulations, [14], the diffusivity might exhibit all five types of behavior mentioned above.Concentration, location and strength of the acid sites are the factors that determinethis dependence of the diffusivity, [14]. In this study, the diffusion of n-hexane insilicalite-1 and HZSM-5 has been investigated with TEX-PEP (see Chapter 2). Theinfluence of the concentration on the diffusion and the activation energy of diffusionhave been investigated for silicalite and HZSM-5 zeolites.


The diffusivities of n-hexane in silicalite have been measured at temperatures from 393to 513 K at various partial pressures from 0.9 to 6.6 kPa with a constant total flow rateof 150 Nml/min. For HZSM-5, experiments have been performed at 393 K at variouspartial pressures from 0.9 to 6.6 kPa with a constant total flow rate of 150 Nml/min.n-Hexane loadings have been calculated by the method as described in Section 2.4.In order to measure the diffusivity of the alkane, 11C-labelled n-hexane was used inthe TEX-PEP experiments as described in Chapter 2. In order to interpret the dataprovided by the PEP experiments, an appropriate mathematical model is used todescribe the transport through the zeolite reactor bed. Description of the model isprovided in Section 2.3. Studied samples of silicalite and HZSM-5 are discussed inSection 2.2.1. Fitting the modelled concentration profiles to the measured ones yieldsthe value of the self-diffusivity D and the information on the zeolite loading. Anexample of such a fit is shown in Fig. 6-1. One can see that the model is capableto describe the experimental data quite well. The diffusivity determined from theexperiments is the self-diffusion coefficient because it measured under steady-stateconditions. Modelling translates the measurement error to the following error onthe fitted parameters. The values of the fitted parameters, self-diffusivity D andadsorption constant Ka are determined with the error of less than 10%.



Self-diffusion of n-hexane in silicalite

Fig. 6-2 shows the effect of concentration on the self-diffusion coefficients at varioustemperatures. At temperatures lower than 513 K, the diffusivity increases as a func-tion of zeolite loading up to approximately 4 molecules per unit cell. At 513 K, the


Figure 6-1. Experimental data (solid curves) fitted by the model (dotted curves): n-hexane in

silicalite, concentration profiles at 7 detection positions, 6.6 kPa, T=413 K.

diffusion coefficient remains the same at approximately 0.1 and 1.2 molecules per unitcell of the zeolite loading. Higher loadings could not be achieved at this temperaturedue to technical limitations. At the lowest temperature applied in this study (393K), the diffusivity does not increase further when the concentration of n-hexane insilicalite exceeds 4 molecules per unit cell.

Apparently, the behavior of n-hexane observed here is different from what has beenmeasured or calculated in the works of other authors. Using PFG NMR Heink et al.,[13], measured a decrease in the diffusivity of n-hexane in silicalite at loadings up toone molecule per unit cell and higher than 4 molecules per unit cell. Millot et al.,[15], did not observe any change in the diffusion with the Membrane permeationmethod when the partial pressure increased from 8 kPa to 13 kPa. Under conditionssimilar to our study, Van Den Begin et al., [3], have also observed a decrease inthe n-hexane diffusivity by a single-step frequency response technique. Theoreticalsimulations generally agree with these experimental observations. Dynamic Monte-Carlo simulations of Coppens et al., [5], predict a decrease in the diffusivity with poreoccupancy which can be approximated by a stretched exponential. This behavior isrelated to the probability for the molecule to jump to the neighboring position thatis free which decreases with the pore occupancy. This is in agreement with the MDsimulations of Schuring et al., [9], for n-butane. Thus, our results are in apparentcontradiction with other studies.

However, an increase in the diffusion coefficient with the pore loading of the zeo-lite measured here (Fig. 6-2), is in agreement with the data of Song and Rees, [11],describing diffusion in the straight channels of silicalite. The FR technique has beenapplied to distinguish the diffusion in straight and zigzag channels of MFI-zeolites.


Figure 6-2. Self-diffusivity of n-hexane in silicalite at various temperatures and loadings.

The authors proposed that two different diffusion processes occur in these channelsand molecular exchange is possible. This has been confirmed by comparison with dif-fusion of hydrocarbons in silicalite-2, which only consists of the intersecting straightchannels. In this zeolite only a single diffusion process was observed, [16]. The dif-fusivity of n-hexane in zigzag channels of silicalite-1 was found to be an order ofmagnitude lower than that in the straight channels, [16]. This is in line with MD sim-ulations of Hernandez and Catlow, [17], that indicate that diffusion in the straightchannels is several times faster than that in zigzag channels even at loadings of 4molecules per unit cell. Song and Rees proposed that with increased concentrationn-hexane molecules get immobilized in zigzag channels and the diffusivity via thestraight channels increases. In our study, due to the random orientation of the zeo-lite crystals along the reactor, we obtain the average diffusion coefficient. Therefore,we can not attribute the diffusional behavior of n-hexane found in our study to thestraight channel diffusion only. The diffusion of the molecules in the zeolites is de-termined by the adsorbate-adsorbate and adsorbate-zeolite interactions. Theoreticalinvestigations performed by Schuring et al., [9], show that for n-butane molecules theadsorbate-adsorbate interactions increase with the silicalite pore occupancy, while theadsorbate-zeolite potential energy remains approximately the same. Evidently, as thezeolite pores get more crowded the collisions between the molecules increase, whichcauses diffusivity to decrease in the absence of any other interactions. Approximatelya decade ago, Tsikoyiannis and Wei showed that under the presence of the repulsiveforces between the neighboring molecules, uptake diffusivity proceed via maximum asthe pore occupancy increases, [7]. The stronger the repulsions, the more pronouncedmaximum is observed. However, it is difficult to compare the uptake diffusivity tothe self-diffusion coefficient measured here. Recent kinetic Monte-Carlo and CBMC


simulations of Paschek and Krishna, [18], provided indications for the presence ofthe repulsive interactions between molecules adsorbed at adjacent intersection andstraight channel adsorption sites for iso-butane self-diffusion in silicalite. Probably,diffusion in zigzag channels is not affected by the interactions with the molecules inthe adjacent intersections because it is too slow compared to diffusion in the straightchannels. These repulsive interactions have been shown to have a noticeable effect onthe mobility of the molecules, especially at low temperatures, so that the diffusivityincreases, [18]. We suspect that the increase in the diffusivity of n-hexane with loadingobserved here might be related to the presence of similar repulsive interactions.

In fact, Monte-Carlo simulations performed by Smit and Maessen, [19], indicatedthat up to the loadings of 4 molecules per unit cell, n-hexane molecules are randomlydistributed in the straight, zigzag channels and the intersections between them. Asthe loading increases, there are more molecules adsorbed in the intersections andadjacent straight channels, which results in an increase of the repulsive interactionscausing an increase in n-hexane mobility.

Kinetic Monte-Carlo simulations for methane and iso-butane diffusion in silicalite,[20], have shown an interesting behavior for both hydrocarbons: a decrease of thediffusivity at loadings up to 4 molecules per unit cell, followed by an increase inthe self-diffusion coefficient at higher loadings due to repulsive interactions. Theseobservations do not contradict our experimental findings for n-hexane in silicalite. Itis known that at low partial pressures, both iso-butane and methane molecules occupythe intersections only, [21]. Thus, there are no repulsive interactions between themolecules because they are separated from each other by straight or zigzag channels.In this way, the diffusivity decreases due to lower probability for the molecule tojump to a neighboring adsorption site (the channel intersection). A similar behaviorwas observed in our study of 3-methylpentane diffusion in silicalite, discussed inChapter 5. The concentration dependence of the self-diffusion coefficient for the iso-hexane could be approximated by a stretched exponential at the loadings up to 4molecules per unit cell, which is in agreement with the simulations of Krishna andPaschek for methane and iso-butane, [20].

At loadings higher than 4 molecules per unit cell (at high partial pressures), iso-butane as well as methane molecules are pushed into the straight and zigzag channels.This results in the repulsion between the molecules located in the intersections and inadjacent straight channels causing an increase in the diffusivity, [20]. It is importantto note that the length of iso-butane molecules is smaller than the length of n-hexane.Therefore, one expects a stronger repulsion between n-hexane molecules than for iso-butane because the distance between the molecules in the intersections and adjacentstraight channel will be shorter.

Since the adsorption behavior of n-hexane molecules in silicalite is different fromthat of iso-butane, [21], we expect a concentration dependence of the diffusion co-efficient to be different from that of iso-butane. At low partial pressures, the diffu-sivity will increase with loading due to increased repulsive interactions between themolecules in the straight channels and the intersections. At high partial pressures,


when the loading reaches 4 molecules per unit cell, all the molecules are in the zigzagchannels, [19]. This is because the length of n-hexane molecule is commensuratewith the length of the zigzag channels, so these molecules ”freeze” there. When themolecules occupy the sinusoidal channels, the intersections are free, so the straightchannels can get completely filled with the n-hexane molecules upon further increaseof the pressure. It means that there is no additional repulsive interaction betweenthe molecules, and a decrease in the diffusivity is expected because a pore occupancyeffect, [5], mentioned above will play a dominating role.

Indeed, as shown in Fig. 6-2 at a temperature of 393 K, at the partial pressure of6.6 kPa the loading was approximately 4 molecules per unit cell and the diffusion co-efficient does not increase anymore. According to the calculated adsorption isothermof n-hexane, [19], a n-hexane pressure of 3 to 7 kPa is needed for commensuratefreezing at 398 K, then, all the molecules are in zigzag channels. So, we assume thatthe observed behavior of n-hexane diffusion at this loading is caused by the freezingof the molecules in the zigzag channels of silicalite, which leads to the absence of therepulsive interactions because molecules are separated by the channel intersections.At 513 K, the diffusivity did not change either (Fig. 6-2). Krishna and Paschek, [20],pointed out that repulsive interactions will not have a noticeable impact on molecularmobility at elevated temperatures. This might be one of the reasons why we did notobserve any change in the diffusion at 513 K. On the other hand, the loading mighthave been too low (Fig. 6-2) to have a significant amount of the n-hexane moleculesin the intersections and adjacent channels for repulsion to occur.

Our results are supported by the FR measurement performed by Song and Rees,[11]. n-Hexane diffusivity was found to increase with total loading in the straightchannels at relatively low temperature (336 K), while at higher temperatures dif-fusivities decreased or did not change with loading. Probably, the reason is thatat lower temperature, the repulsive interactions are more pronounced compared tohigh temperatures, and that the loading was high enough to induce repulsion. In ourstudy, higher n-hexane pressures were applied (see Fig. 6-2) than in the work of Songand Rees, [11]. Therefore, we observed an increase in the diffusion coefficient even athigher temperatures, except for 513 K. As the repulsion occurs between the moleculesin the intersections and straight channels, it has more impact on the diffusion in thesechannels as measured by Song and Rees, [11]. With TEX-PEP the average diffusioncoefficient is obtained, but as we mentioned earlier, diffusion in zigzag channels isan order of magnitude lower, so it does not have a significant impact on the averagevalue.

Summarizing, one can distinguish the following regimes of n-hexane adsorptionand diffusion in silicalite:

• at very low loadings, n-hexane molecules are randomly distributed through outthe zeolite pore network, the probability of finding the molecules adsorbed in theintersections and in the adjacent straight channels is small. Therefore, diffusionprobably does not change or slightly decreases with the total pore occupancybecause pore occupancy and repulsion effects are very weak.


• At intermediate loadings (up to 4 molecules per unit cell), there are moremolecules in the intersections and straight channels, so repulsion between themolecules has more effect on the mobility than blockage, causing the diffusivityto increase.

• At the highest loadings, when the zigzag channels are completely filled with then-hexane molecules, increase in the pore occupancy leads to a decrease in thediffusivity. Probably because the straight channels get filled first, and then theintersections, or simply due to a prevailing effect of the pore blockage.

In short, the concentration effects on the self-diffusivity are determined by the bal-ance between intermolecular repulsive interactions, which increase diffusion rates withincreasing concentration, and the probability to find a free adjacent site, which de-creases with increasing concentration.

Our interpretation of the peculiarities of n-hexane diffusion and adsorption insilicalite is in line with the results from PFG NMR, [13], where a decrease in thediffusivity is found at loadings up to 1 molecule per unit cell and higher than 4molecules per unit cell. The temperatures of the experiments were higher than in thepresent study, which might have diminished the effect of the repulsive interactions.

As to the apparent contradictions with other theoretical and experimental studiesmentioned in the beginning of this section, we believe, this can be explained. Firstof all, some theoretical studies, [5, 6, 7, 8, 9, 10], of the concentration dependence ofself-diffusion coefficients did not take into account a molecular repulsion. This wasalready pointed out by Krishna and Paschek, [20]. These studies thus focused on thecorrelation effects and a decreased molecular mobility was observed due to increasedprobability of blocking adjacent sites with pore occupancy.

Regarding the experimental work, we note that the concentration dependencewill be strongly influenced by the conditions. In the experimental study of n-hexanediffusion in silicalite performed by Millot et al. using the membrane permeation tech-nique, [15], no change in the diffusivity was measured under 5 to 13 kPa at 303 to 473K. We suppose, the partial pressure was high enough for the commensurate freezingof n-hexane molecules in zigzag channels of silicalite to occur or perhaps, the poreoccupancy effect increased so that there was a balance between it and effect of therepulsive interactions. These observations are in agreement with our data at 393 K,3.3 and 6.6 kPa (Fig. 6-2). The data of Square Wave, [3], and FR, [2], showed onlya very slight decrease of the corrected diffusivity of n-hexane in silicalite, which doesnot correspond to our observations. Recently, Krishna showed that corrected and self-diffusion coefficients can not be directly compared unless they have been determinedat the loading close to 0, [22]. Therefore, we believe, it might be the reason for thediscrepancies.

Comparison with the literature

In the Table 6-1 data on the diffusion coefficient of n-hexane in silicalite measuredby various macro- and microscopic and simulation techniques are presented. One ob-



TEX-PEP 0.4-1.9 393 K, 2.8-4.5 mol./u.c.FTIR, [23] 0.0011 353 K,-TEOM, [24] 0.14-0.125 388 K, 3.27-2.5 mol/u.c.QENS, [25] 69 373 K

FR, [2] 0.7-0.4 373 K, 1-4 mol./u.c.Square Wave, [3] 0.09-0.07 379 K, 1-3 mol./u.c.

Membrane Permeation, [15] 1 373 K, 0.08-0.13 barNMR, [13] 15 330 K, 4 mol./u.c.MD, [26] 100 330 K, 4 mol./u.c.

Table 6-1. Diffusion coefficients of n-hexane in silicalite obtained by various methods.

serves that our TEX-PEP values are in a good agreement with the values provided byFR [2], Membrane Permeation, [15], and TEOM, [24], techniques. An order of mag-nitude lower values were measured by Square Wave, [3], and FTIR, [23], methods,which might be due to a lower temperature of those experiments. The diffusion coeffi-cients measured by microscopic techniques such as NMR, [13], QENS, [25], and thosecalculated from MD simulations, [26], are traditionally significantly higher than thosederived from macroscopic methods. The authors of the macroscopic ultra-high vac-uum Multitrack technique, [27], claim that one of the reasons for lower values of thediffusivities measured by macroscopic methods is the diffusion resistance caused bythe presence of a carrier gas. This, probably, can explain the observed discrepancies.

Diffusion of n-hexane in HZSM-5

We have also studied the influence of the n-hexane concentration on the diffusion inHZSM-5 zeolite in order to determine the effect of Brønsted sites. The measurementshave been performed at n-hexane partial pressures from 0.9 kPa to 6.6 kPa at 393K. The obtained self-diffusion coefficients have been compared to those in silicalitein Fig. 6-3.

The self-diffusivity in HZSM-5 exhibits the same concentration dependence asfound for silicalite. However, at similar loadings, the diffusivity is approximatelytwo times lower in the presence of acid sites. Thus, next to the pore occupancyeffect and repulsive interactions between the molecules sited in the straight channelsand adjacent intersections, in HZSM-5 zeolite, n-hexane molecules experience theinteractions with the acid sites, that causes the twofold decrease in the diffusivity.Obviously, the presence of the acid sites increases the residence time for the moleculesat the adsorption site via stronger interactions. A twofold decrease in the diffusivityin HZSM-5 compared to silicalite was also observed for benzene, [28, 29], and formixtures of n-hexane/2-methylpentane in silicalite and HZSM-5 (see Chapter 4).

On the other hand, we note that PFG NMR experiments, [13], did not detect any


Figure 6-3. Influence of the concentration on the self-diffusivity of n-hexane in silicalite and

HZSM-5 at 393 K.

influence of the acid sites on the mean diffusivity. The reason for this disagreementwith the macroscopic observations mentioned above is not really clear.

6.3.2 Activation energy of diffusion

In general, activated diffusion is described by an Arrhenius-type equation (Eq. 4.3.1).The pre-exponential factor D0 is related to the jump frequency between adsorptionsites in the zeolite lattice, [30]. The jump frequency might depend on the concentra-tion of the adsorbate (as has been shown in Chapter 5) and on the interactions withthe acid sites. In the study discussed in Chapter 4, we tried to establish the influenceof the acid sites on the apparent activation energy of diffusion but no definite conclu-sion was made. In this study we have measured the apparent activation energies ofdiffusion for n-hexane in silicalite and HZSM-5 under different partial pressure condi-tions in order to establish which parameter (activation energy or the pre-exponentialfactor) is affected by the concentration and the interaction with the acid sites.

The apparent activation energy of n-hexane in silicalite was measured at 0.9, 3.3and 6.6 kPa, in the temperature range of 393-513 K. For HZSM-5, the experimentshave been performed at 0.9 and 6.6 kPa. The results are shown in Fig. 6-4 and Fig. 6-5for silicalite and HZSM-5, respectively.

For silicalite, the apparent activation energy does not change with the partialpressure and is equal to approximately 19± 3 kJ/mol, which is in a good agreementwith the values of provided by other techniques as shown in Chapter 4. This valueis also in agreement with the value of the activation energy of n-hexane diffusionthrough straight channels of silicalite determined from the FR experiments (21.7± 2kJ/mol), [11]. This probably indicates that the diffusion in the straight channels has


Figure 6-4. Arrhenius plots for the self-diffusivity of n-hexane in silicalite at various partial

pressures.

Zeolite Pressure, kPa Ea, kJ/mol D0, m2/s

silicalite 0.9 18± 3 1.3× 10−9

3.3 20± 4 4.7× 10−9

6.6 19± 2 1.2× 10−8

HZSM-5 6.6 21± 2 7.7× 10−9

0.9 32± 6 9.1× 10−8

Table 6-2. Diffusion parameters for n-hexane in silicalite and HZSM-5 at different partial pres-

sures.

a significant contribution to the overall diffusion process.

From Fig. 6-4 one observes that the activation energy does not change as a functionof partial pressure. In case of 3-methylpentane, increase in the apparent activationenergy of diffusion was observed (see Chapter 5, Fig. 5-6). There, diffusion decreaseswith the pore occupancy. However, for n-hexane the opposite behavior is observedbecause of the repulsive interactions. One might suppose that the effect of the poreoccupancy on the slope of the Arrhenius plot is compensated by the increase in thediffusivity due to the repulsion (see Fig. 6-4). Moreover, the intercept of the Arrheniusplots, which relates to the jump frequency, increases with partial pressure (Table 6-2).Thus, we assume, that the repulsive interactions have an effect on the jump frequency.From the PFG NMR experiments, [13], the activation energy was also found to beindependent on the loading, while the pre-exponential factor D0(θ) decreased withthe concentration at low and high loadings (> 4 mol./u.c.), which is in agreementwith our hypothesis on the n-hexane diffusion in silicalite.


Figure 6-5. Influence of the partial pressure on the n-hexane apparent activation energy of

diffusion in HZSM-5 zeolite.

At low loadings, the jump frequency decreases because zeolite pores become morecrowded so that the average probability that the neighboring adsorption place is freefor the molecule to jump decreases, and the molecule should stay longer at its currentlocation. At certain loading, the repulsive interactions between the molecules in theintersections and adjacent straight channels increase the mobility of the molecules,so that the jump frequency increases. At loadings higher than 4 molecules per unitcell, concentration effect starts to dominate due to redistribution of the molecules,and jump frequency decreases as observed by PFG NMR, [13].

For HZSM-5 zeolite, the situation appears to be more complicated as the apparentactivation energy of diffusion decreases with the partial pressure. At high partialpressure (6.6 kPa) the activation energies in silicalite and HZSM-5 are very close, andthe pre-exponential factor for the diffusion in the presence of the acid sites is lowerthan that in silicalite. At lower partial pressure (0.9 kPa), the activation energy andthe jump frequency in HZSM-5 are both higher than in silicalite (Table 6-2). Probably,when the concentration inside acidic zeolite is not very high, the interactions of n-hexane molecules with the acid sites have a strong influence on the overall diffusionprocess. On the other hand, under similar partial pressure, loading of HZSM-5 ishigher than of silicalite (Fig. 6-3), therefore, pre-exponential factor D0 in HZSM-5 ishigher than in silicalite. Anyway, at 0.9 kPa the repulsion is not enough to overcomethe interactions with acid sites, and the apparent activation energy is high. As thepartial pressure increases, the repulsive interactions start to dominate. Therefore,the apparent activation energy of n-hexane diffusion measured at high pressure isequal to the activation energy of diffusion in silicalite. Fact, that the jump frequencyD0 in HZSM-5 is decreasing with the pressure, might be explained by the influenceof the pore blockage at high loadings, meaning that it might decrease the jump


frequency, while the influence of the repulsion could result in the increase of theapparent activation energy.

6.4 Conclusions

This study revealed some interesting features of the n-hexane diffusion in MFI-typezeolites that are probably caused by the repulsive interactions between the moleculeslocated in the intersections of the straight and zigzag channels and those sited in theadjacent straight channels. Indications for this kind of interactions have been reportedby Paschek and Krishna, [18], for iso-butane in silicalite as a result of Monte-Carlosimulations.

We found that the diffusivity increases with the concentration at loadings upto 4 molecules per unit cell. At these loadings, n-hexane molecules are randomlydistributed in the zeolite matrix, so that the amount of the molecules in the intersec-tions and adjacent straight channels is increasing with loading, therefore the repulsionincreases. This leads to an increase in the apparent pre-exponential factor (jump fre-quency). We have also observed that at high temperatures, the repulsion between themolecules does not have a noticeable effect on the diffusion as predicted by Krishnaand Paschek, [20]. Although repulsions between molecules located in the intersectionsand adjacent zigzag channels may also occur, they do not have a significant influenceon the average diffusivity measured here, because diffusivity in zigzag channels is anorder of magnitude lower compared to that in the straight channels, [11].

At higher loadings, the diffusivity did not change because of the re-arrangementof n-hexane molecules in the channels due to a possible commensurate freezing ofthe molecules in zigzag channels. This can diminish repulsive interactions betweenthe molecules. The pore occupancy effect, which causes a decrease in the molecularmotion, started to dominate.

In HZSM-5, at low partial pressures, the activation energy of diffusion in HZSM-5was measured to be higher than at high pressures, probably due to the interaction withthe acid sites, which dominate over concentration dependence and repulsive forces.In acidic zeolite, the interaction with the acid sites causes an approximately twofolddecrease in the jump frequency (pre-exponential factor) compared to silicalite. At highpressures, the repulsive interactions start to have a stronger effect on the diffusion,so the apparent activation energy becomes very close to the activation energy insilicalite.

References


[2] Bulow, M.; Schodder, H.; Rees, L.V.C.; Caro, J.; Richards, R. in proc. of 7thInternational Zeolite Conference, 1986, Tokyo.

[3] van den Begin, N.; Rees, L.V.C.; Caro, J.; Bulow, M. Zeolites 9 (1989) 287.

[4] Caro, J.; Bulow, M.; Schimer, W.; Karger, J.; Heink, W.; Pfeifer, J. J. Chem.Soc. Faraday Trans. 81 (1985) 2541.






[10] Raj, N.; Sastre, G.; Catlow, C.R. J. Phys. Chem. B 103 (1999) 11007-11015.

[11] Song, L.; Rees, L.V.C. J. Chem. Soc. Faraday Trans. 93(4) (1997) 649-657.

[12] Talu, O.; Sun, M.S.; Shah, D.B. AIChE Journal 44(3) (1998) 681-694.




[16] Rees, L.V.C. and Shen,D Stud. Surf. Sci. 87 (1994) 563-571.

[17] Hernandez, E. and Catlow, C.R.A. Proc.R.Soc.Lond A 1995 (448) 143-160

[18] Paschek, D.; Krishna, R. Chem. Phys. Lett. 2001 (342) 148-154

[19] Smit, B.; Maesen, T. Nature 1995 (374) 42-44

94 REFERENCES

[20] Krishna, R.; Paschek, D. Chem. Ing. J. 2002 (85) 7-15

[21] Vlugt, T.J.H.; Krishna, R.; Smit, B. J. Phys. Chem. B B,1999 (103(7)) 1102-1118


[23] Lin, Y.S.; Yamamoto, N.; Choi, Y.; Yamaguchi, T.; Okubi, T.; Nakao, S.-I. Mi-croporous Mesoporous Mater. 38 (2000) 207-220.

[24] Zhu, W.; Kaptein, F.; Moulijn, J.A. Microporous Mesoporous Mater. 47 (2001)157-171.

[25] Jobic, H.; Bee, M.; Caro, J. Proc.Int.Conf. on ZeoZites, 1992, Montreal, Canada,p.121

[26] June, R.L; Bell, A.T.; Theodorou, D.N. J. Phys. Chem. B 96 (1992) 1052.

[27] Nijhuis, T.A.; van den Broeke, L.J.P.; ven de Graaf, J.M.; Kapteijn, F.; Mak-kee, M.; Moulijn, J.A. Chem. Eng. Sci. 52(19) (1997) 3401-3404.

[28] Shen, D.; Rees, L.V.C. Zeolites 11 (1991) 666-671.

[29] Zikanova, A.; Bulow,M.; Schlodder, H. Zeolites 7 (1987) 115-118.

[30] Chen, N.Y.; Degnan, T.; Smith, C. Molecular transport and reaction in zeolites,1994, John Wiley & Sons.

7 Diffusion of linear alkanes and theirmixtures in silicalite

The influence of the zeolite pore occupancy on the diffusivity of n-pentane in sili-calite, has been investigated using PEP in the temperature range of 373-473 K. Athigh temperatures, diffusion was found to be independent on the zeolite loading, whileat 373 and 393 K, a slight increase in the diffusivity was observed with increasingalkane loading. Most probably, this behavior was caused by the repulsive interactionsbetween the molecules adsorbed in the channel intersections and the adjacent straightchannels. These interactions resulted in a decrease of the apparent activation energyat high partial pressure of n-pentane. A study of n-pentane and n-hexane mixtureswith various ratios between the components has been performed. The preferential ad-sorption of n-hexane over n-pentane was observed. Diffusion of both alkanes increasedwith the fraction of n-hexane in the gas phase. Probably, the mobility of n-pentanewas affected by the repulsive interactions with the n-hexane molecules in the adjacentadsorption sites.

7.1 Introduction

MFI-type zeolites are widely used in the petroleum industry as molecular sieves inseparation processes and as catalysts for hydrocarbon conversion processes. Amongstothers, hydroisomerization of a mixture of C5−/C6− alkanes is one of the most im-portant reactions in the modern oil refinery. This explains the large interest in theadsorptive and diffusive properties of hydrocarbons in the zeolites. However, so farno experimental studies have been performed on the diffusion of mixtures of thesetwo linear alkanes.

Self-diffusion of a single component in the zeolites might exhibit various concen-tration dependencies, [1], which are mainly determined by the zeolite topology, sitingof the molecules, adsorbate-adsorbate interactions and adsorbate-zeolite interactions.The situation is even more complicated for mixtures of components. Evidently, thepresence of another type of molecules might affect the diffusion of both mixture com-ponents, [2]. This has been shown in numerous computer simulations, [3, 4, 5, 6, 7],and some experimental studies, [2, 8, 9, 10], for binary mixtures of linear and branchedalkanes or aromatics. Indeed, in silicalite, the diffusivities of both components in 2-methylpentane/n-hexane mixtures decrease due to the blockage of the zeolite porenetwork by the branched alkane, [2]. This is because a bulky alkane is slower and it ispreferentially located in the channel intersections, blocking the connection betweenthe channels and forcing the faster n-hexane molecules to stay longer at their currentlocations.

96 Chapter 7: Linear alkanes and their mixtures in silicalite

Meanwhile, for the single n-hexane, an increase of the diffusion with the loading(up to 4 molecules per unit cell) in silicalite-1, was measured in our previous study (seeChapter 6). It was ascribed to dominant adsorbate-adsorbate interactions, that arebelieved to be caused by repulsion between the molecules located in the intersectionsand adjacent straight channels. The relevance of such interactions has been pointedout by Paschek and Krishna, [11], for iso-butane in silicalite by the Monte-Carlosimulations. For n-pentane, one would expect a behavior similar to that of n-hexanein silicalite. It would be particularly interesting to study the mixture of homologmolecules.

A study of binary diffusion in mixtures of linear alkanes (n-heptane/n-octane) inhighly siliceous MFI-type zeolite using a desorption under reduced pressure (DRP)method has been performed by Masuda et al., [12], using a mass-spectrometer. Thismethod allowed to measure a decrease in the transport diffusivity of the fast compo-nent (n-heptane) in mixtures with the slow n-octane. It was measured that the rateof desorption of n-heptane decreased in the presence of n-octane, while desorption ofslow n-octane remained unchanged.

In this study, the diffusion of n-pentane and n-pentane/n-hexane mixtures insilicalite-1 has been investigated with TEX-PEP. The diffusivities of both n-pentaneand n-hexane have been measured in their mixtures with different ratios between thecomponents. For n-pentane, the influence of the concentration on the diffusion hasbeen studied.


Experiments have been performed in the TEX-PEP mode using a PEP setup. Theinformation on the studied silicalite sample is provided in Chapter 2. The radio-labelled n-pentane and n-hexane were produced from 11CO and non-labelled 1-buteneand 1-pentene, respectively, as described in Chapter 2. The production of 11C-labelledn-pentane for the first time was developed in this study.

In order to measure the diffusion of the components in the mixture in silicalite,two sets of experiments have been performed. In one set a labelled n-pentane wasused in order to monitor a behavior of this hydrocarbon in the mixture, then underthe same conditions, the experiments were repeated with the labelled n-hexane.

Interpretation of the experimental data using modelling provided the estimationfor the loading of silicalite with both components and their diffusion coefficients.Loadings of the components at various mixture compositions have been calculatedusing Eq. 2.3.13. Modelling translates the measurement error to the following erroron the fitted parameters. For n-hexane, the relative error of the determination ofthe self-diffusion coefficient from the fitting normally does not exceed 10%. In thepresent study, at high temperatures the accuracy in the estimation of diffusivitieswas lower, because of the lower signal-to-noise ratio in the experimental data. Itwas caused by lower adsorption of n-pentane compared to n-hexane, therefore, alower signal from radio-labelled n-pentane molecules was detected. An example of


Figure 7-1. Experimental data fitted by the model: n-pentane in silicalite, concentration profiles

at 6 detection positions, 6.6 kPa, T=413 K.

n-pentane experimental data fitted by the model is shown in Fig. 7-1.The diffusivities of n-pentane in silicalite have been measured at temperatures

from 373 to 473 K at various partial pressures from 0.1 to 13 kPa with a total flowrate of 80 to 200 Nml/min. Experiments on binary mixtures have been performed at433 K varying the ratio between the components and keeping the total pressure ofhydrocarbons fixed at 6.6 kPa. n-Pentane and n-hexane loadings have been calculatedas described in Section 2.4.



Diffusion of n-pentane in silicalite

In Fig. 7-2, the influence of the alkane concentration on the self-diffusivities at vari-ous temperatures is shown. At 473 and 453 K the loading is quite low (less than 1.5molecules per unit cell) making it difficult to determine its influence on the diffusion,while at 433 K, the loading increases from 0.5 to approximately 3.5 molecules perunit cell. Thus, we conclude that in this temperature range (433-473 K), the diffu-sivity does not change with the loading within the experimental accuracy. At lowtemperatures, 373 and 393 K, the diffusion coefficient of n-pentane increases with itsconcentration in the zeolite pores. This is in line with the results of a previous studyof n-hexane diffusion in silicalite reported in Chapter 6, where a strong increase ofn-hexane diffusivity at loadings from 0.1 to 4 molecules per unit cell was indeed ob-served. This increase was attributed to the repulsive interactions between molecules


Figure 7-2. Self-diffusivity of n-pentane in silicalite-1 at various temperatures and loadings.

located in the channel intersections and the adjacent straight channels. Evidence forthese interactions was provided by Paschek and Krishna for iso-butane in silicalite[11]. We assume, that the concentration dependence of n-pentane diffusivity measuredhere (Fig. 7-2) is due to the same effects as outlined for n-hexane and iso-butane.

As n-pentane molecules are randomly distributed throughout the zeolite matrix,without any preference for a specific site (straight, zigzag channels or the intersec-tions between them), [14], more molecules are located in the straight channels andintersections with increasing zeolite loading. This leads to an increase in the repulsiveinteractions mentioned above, which increases the mobility of n-pentane molecules.

One would expect weaker repulsions for n-pentane compared to the longer n-hexane molecules, especially at high temperatures. Indeed, as only one molecule fitsinto the channel or an intersection, the distance between the n-pentane molecules ad-sorbed in the straight channels and the intersections is larger than that for n-hexanemolecules. Repulsive forces decrease when the distance between the molecules in-creases, which means a weaker repulsion for n-pentane. At high temperatures, theseinteractions are even weaker as noticed by Paschek and Krishna, [11]. It is indeed ob-served in our experiments: diffusivity of n-pentane does not increase with the loadingat high temperatures 433-493 K (Fig. 7-2).

Besides repulsive interactions between the molecules, there are other factors thatinfluence the diffusion of n-pentane in silicalite. Theoretical calculations, [15, 16, 17,18, 19], showed a decrease in the diffusivity with the loading when repulsive inter-actions are not taken into account. In this case, pore occupancy plays a significantrole. Indeed, as the loading increases, the probability for the molecule to jump to theneighboring adsorption site decreases since the chance that the site is already occu-pied increases, so the molecule has to stay at its current location longer. It causes


the decrease of the jump frequency, which leads to a slower diffusion. According tothe mean-field theory, diffusivity should be proportional to the fraction of the unoc-cupied sites, [20]. For silicalite, self-diffusivity was shown to decrease slightly fasterwith the occupancy due to correlation effects, [15]. Krishna and Paschek, [21], deter-mined a jump diffusion at a given occupancy as a jump diffusion at zero occupancyproportional to the vacancy factor and to the repulsion factor, which on its turn isalso concentration dependent. Therefore, we assume that self-diffusion of n-pentanein silicalite is simultaneously affected by the occupancy of the zeolite, which causesa decrease in the diffusion, and by repulsive interactions between the molecules, thatcause an increase in their mobility. These two effects compete with each other. Ex-perimental conditions determine the domination of one of them. For example, up toa certain partial pressure, molecules such as branched alkanes are preferentially ad-sorbed in the channel intersections, [11, 14, 22]. Under these conditions, there is norepulsion between the molecules, and diffusion decreases with the loading. As soonas the molecules start to appear in the straight channels, diffusion increases, [11].

To our opinion, the fact, that at high temperatures (433-473 K), diffusion of n-pentane does not change with the concentration (Fig.7-2), indicates that the poreoccupancy effect is compensated by the repulsive interactions between the molecules.At 433 K, the concentration of n-pentane in the pores varied from 0.5 to 3.5 mol./u.c.,which is close to 50% pore occupancy. At these pore occupancies in the absence ofany other interactions a decrease in the diffusivity should be observed as shown byCoppens et al., [15]. Thus, we may conclude that there are repulsive interactions thatcompensate this decrease of the molecular mobility. On the other hand, at high tem-peratures diffusion is quite fast, and the repulsive interactions are weaker. Thereforethese interactions do not dominate over the pore occupancy effect and the diffusioncoefficient does not change. At low loadings, both, the occupancy and repulsion effectsare too insignificant, so the diffusivity does not change either.

At lower temperatures (373-393 K), diffusion becomes faster at higher loadings(Fig.7-2), probably, due to the dominance of the repulsive interactions, which arestronger at these temperatures.

In the work of Heink et al., [23], a decrease in the diffusivity of n-pentane insilicalite at loadings up to 0.75 molecules per unit cell and higher than 4 moleculesper unit cell was measured with PFG NMR. Despite an apparent disagreement, theseresults can be explained within our hypothesis. We assume, that in their experimentsat low concentrations the pore occupancy effect was dominating, so the jump fre-quency was measured to decrease with the loading [23], which resulted in the slowerdiffusion. Here, we did not observe any change in the diffusivity in a similar loadingrange perhaps, due to the lower experimental accuracy. On the one hand, at sucha low loadings, one should not expect a noticeable decrease in the diffusion due tothe pore occupancy, [15]. However, at high loadings (4-12 molecules per unit cell), adecrease of the n-pentane self-diffusion was measured, [23, 24]. Probably, the occu-pancy effect suppresses the repulsion between the molecules. On the other hand, it isknown that maximum sorption capacity of silicalite for n-pentane is approximately 8



TEX-PEP 0.7-1.05 373 K, 2.9-7.1 mol./u.c.Membrane technique [26] 0.24 334 K,-

ZLC, [27] 0.2 334 K, -FR, [28] 40 303 K, 7 mol./u.c.

PFG NMR, [23] 40 330 K, 4 mol./u.c.PFG NMR, [24] 20-4 334 K, 4-8 mol./u.c.

SCM, [29] 1 300 K, 0.5 kPaMD, [16] 98.5 333 K, 4 mol./u.c.

Table 7-1. Diffusivities of n-pentane in silicalite obtained by different methods.

molecules per unit cell, meaning that at the loadings higher than 8 mol./u.c. it wasnot the micropore diffusion that was measured.

The interpretation of the result presented in Fig. 7-2 is supported by the diffusionalmeasurements of light alkanes in silicalite-1 performed by van de Graaf et al., [25],using a Membrane permeation technique. At 303 K, an increase of the intracrystallinediffusivity was observed for C1−C3 alkanes. Moreover, for shorter alkanes, there was asmaller increase detected. At higher temperatures, the diffusivity did not change withthe concentration, which is in line with our results. For n-propane, the diffusivity had amaximum at pore occupancy of approximately 0.8. Apparently, up to that occupancydiffusivity increase due to the stronger influence of the repulsive interactions, thanat very high loadings, a pore occupancy effect starts to be dominating and diffusioncoefficient decreases.

Thus, we assume that there are the following features of n-pentane diffusion insilicalite. At a very low loadings, n-pentane diffusion probably does not change orslightly decreases with the total pore occupancy because of the weak pore occupancyeffect. At intermediate loadings, there are more molecules in the intersections andstraight channels. This leads to an increase of repulsions between the molecules.This would increase the diffusion coefficient, but this increase is compensated bya decreasing jump frequency due to the increased loading. At high loadings, therepulsion factor overcomes the occupancy influence, so the diffusion increases. Ateven higher loadings, the decrease in the mobility or perhaps, or stabilization shouldoccur because of increasing pore occupancy effect. Note, that repulsive interactionsdo not have a noticeable impact on the molecular mobility at high temperatures, sothe diffusivity does not change with the occupancy, or even might start to decreasedue to the pore blockage effect.

Comparison with the literature

Table 7-1 shows a comparison between the diffusion coefficients of n-pentane in sil-icalite measured by various macro- and microscopic techniques as well as deduced


from MD simulations. As the values for the diffusion coefficient of n-pentane in sil-icalite measured under conditions similar to this study have not been found in theliteratures, this Table provides the data obtained under somewhat different condi-tions.

TEX-PEP values are in a good agreement with the values for the diffusivitiesprovided by macroscopic techniques such as the Membrane technique, [26], ZLC,[27], and the SCM technique, [29]. The diffusivity provided by another macroscopicmethod, FR, is an order of magnitude higher, [28], compared to our self-diffusivity. Itcan probably be explained by a significantly lower temperature and a high loading,so the repulsive interactions were very strong under these conditions.

Usually, diffusivities provided by microscopic techniques (PFG NMR) or computersimulations are significantly higher than those from macroscopic methods, which isindeed the case here. The authors of a novel macroscopic ultra-high vacuum techniqueMultitrack, [30], claimed that one of the reasons for lower values of the diffusivitiesmeasured by macroscopic methods is the diffusion resistance caused by the presenceof a carrier gas.

Comparison with the data found in the literature shows that the diffusivitiesprovided by the TEX-PEP method are reasonable and in a good agreement with thevalues supplied by other macroscopic techniques.

Activation energy of n-pentane diffusion in silicalite

We have tried to establish wether there is any influence of the repulsive interactionsdiscussed above on the apparent activation energy. In the theoretical study of iso-butane diffusion in silicalite, [11], there was a change in the apparent activation energyobserved, so that at lower temperatures, the apparent activation energy was lowerthan at elevated temperatures. This is probably caused by the repulsive interactions.

The activation energy of n-pentane in silicalite was measured at 0.5, 1.0, 3.3 and9.0 kPa, in the temperature range of 393-473 K. The results are shown in Table 7-2. Apparently, the accuracy at low partial pressures was not very high, which isrelated to the lower accuracy in the determination of the self-diffusivity as explainedin Section 7.3.1. Nevertheless, obtained values are in agreement with the literaturedata, which vary from 8.3 kJ/mol (PFG NMR, [23]) to 21 kJ/mol (FR, [28]).

Evidently, the apparent activation energy does change with the partial pressure(Table 7-2). We assume, that the change is related to the competing influence of twofactors: pore occupancy and repulsive interactions. Dominance of the factor dependson the temperature and the concentration of the n-pentane molecules in silicalitepores. Silicalite loadings with n-pentane in these experiments are shown in Fig. 7-3.

In Chapter 5 that discusses the concentration dependence of 3-methylpentanediffusion in silicalite we have shown that the apparent activation energy increases withthe partial pressure due to the pore occupancy influence on the pre-exponential factor(jump frequency) D0. It is because under fixed partial pressure conditions, as one cansee from Fig. 7-3, the loading of the zeolite is different at different temperatures.


Figure 7-3. n-Pentane loadings at various temperatures and partial pressures in silicalite.

Pressure, kPa Ea, kJ/mol

0.5 18± 41.0 27± 43.3 21± 39.0 13± 2

Table 7-2. Apparent activation energy of diffusion for n-pentane in silicalite at various partial

pressures.

On the other hand, the apparent activation energy of n-pentane diffusion in silicalitemight increase due to the repulsive interactions.

Thus, at low partial pressure (0.5 kPa), the apparent activation energy is lowercompared to higher pressures, because both pore occupancy and repulsive interactionfactors compensate each other, so that the diffusivity is independent on the loading(Fig. 7-2), and the apparent activation energy is probably, close to the ”true” one (atzero pore occupancy). At higher partial pressures, the pore occupancy might startto play a slightly more significant role, so the activation energy slightly increases.At high partial pressure (9.0 kPa), the apparent activation energy has the lowestvalue, which is probably related to the increase in the diffusivity due to dominatingrepulsive interactions (Fig. 7-2).

Therefore, for n-pentane an influence of pore occupancy and repulsive interactionson the apparent activation energy is probably observed.


Figure 7-4. Loadings of mixture components in silicalite as a function of n-hexane fraction in

the gas phase. Total hydrocarbon pressure 6.6 kPa, 433 K.

7.3.2 n-Pentane/n-hexane mixtures

As discussed in Section 7.3.1, diffusion of both n-hexane and n-pentane in silicaliteare influenced by similar factors. We have conducted a study of binary mixtures ofthese two linear alkanes in order to establish wether the diffusivity of the componentswill be affected by the presence of another alkane.

The experiments have been performed at a temperature of 433 K. The totalhydrocarbon pressure was kept constant at 6.6 kPa by varying the ratio betweenn-hexane and n-pentane in the gas phase.

Adsorption

The loadings of both components in the mixtures have been measured. In Fig. 7-4, itis shown as a function of n-hexane fraction in the gas phase. Obviously, n-hexane ispreferentially adsorbed over n-pentane, which is an expected result. In the equimolargas mixture, the n-hexane loading was approximately 3 times higher compared to thatof n-pentane (0.39 and 0.13 mmol/g, respectively). Indeed, the heat of adsorptionfor n-hexane in silicalite is larger than that of n-pentane, 71 and 42 kJ/mol, [32],respectively. Since both alkanes are likely to be situated everywhere within the zeoliteframework, [14], under these conditions, the reason for the better adsorption of n-hexane is probably in the stronger adsorbent-zeolite interactions due to the longerchain length in comparison with n-pentane. Indeed, the loading of pure n-hexane isapproximately 0.63 mmol/g, while for n-pentane it is 0.45 mmol/g (Fig. 7-4).


Figure 7-5. Self-diffusivities of mixture components in silicalite at various gas mixture compo-

sitions. Total hydrocarbon pressure 6.6 kPa, 433K. Gas mixture composition is determined by

the ration between the components in the gas phase.

Diffusion

Fig. 7-6 shows the diffusivities of n-pentane and n-hexane as a function of gas mixturecomposition. Loadings of both components depend on their gas fraction (Fig. 7-4).In Fig. 7-6, the influence of the loading of n-hexane on the diffusion coefficients ofthe mixture components is shown. Here, the total hydrocarbon pressure is constant(6.6 kPa), so that when the concentration of n-hexane increases, the concentrationof n-pentane decreases. At 0 mol./u.c. of n-hexane, there is only n-pentane in sili-calite. Concentration of pure n-hexane under these conditions (433 K, 6.6 kPa) is 3.6mol./u.c.

First of all, quite an unusual fact is observed: the diffusion coefficient of puren-pentane in silicalite at 433 K, 6.6 kPa is lower then that of pure n-hexane. InFig. 7-2, one can see that at 433 K, n-pentane diffusion does not change with theloading for the reasons discussed earlier. It means, that the diffusion coefficient isclose to that at zero loading. In the case of-hexane, a drastic increase of the diffusivitywith the loading is observed (see Chapter 6, Fig. 6-2) because of stronger repulsiveinteractions between n-hexane molecules compared to n-pentane. Therefore, at theloading of approximately 3.6 molecules per unit cell, the diffusion coefficient is severaltimes higher than ”true” diffusivity. It is clear from Fig. 7-6, that at low n-hexaneconcentrations, its diffusion is slower than of n-pentane. Thus, this strange behaviorof the single components is explained by the conditions of the experiment.

The diffusivity of n-pentane increases with n-hexane loading (Fig.7-6), while thediffusivity of single n-pentane is independent on the concentration at this temperature(Fig. 7-2, 433 K). Repulsion between n-pentane should be weaker than between n-


Figure 7-6. Influence of n-hexane loading on self-diffusivities of mixture components in silicalite,

total hydrocarbon pressure 6.6 kPa, 433K. Gas mixture composition is established by the ration

between the components in the gas phase. The gas phase mixture composition determines the

loading of both components (see Fig. 7-4).

hexane, since the length of n-hexane molecules is larger than that of n-pentane, sothe distance between C6−alkanes is shorter than that for n-pentane, which makes then-hexane’s interactions stronger. In mixtures, n-pentane and n-hexane molecules arerandomly distributed in the zeolite, so the interaction occurs between n-pentane/n-hexane molecules too. In this situation, the distance between n-hexane/n-pentane isshorter than that of n-pentane/n-pentane molecules. Thus, with increasing n-hexaneloading n-pentane molecules have more interactions with n-hexane molecules, leadingto an increasing diffusivity.

From Fig. 7-6 it is clear that n-hexane diffusion is slower compared to its diffusionas a single component. Apparently, it happens for a similar reason as the mobility ofn-pentane increases: repulsive interactions with n-pentane molecules are weaker thanwith n-hexane molecules. This also indicates that in the absence of such interactionsn-hexane is slower than n-pentane, which is a reasonable behavior to expect. Asthe concentration of n-hexane in the silicalite pores increases further, diffusion ofn-hexane slightly increases due to the increased repulsion.

These results are from the first sight somewhat different from those measured byMasuda et al., [12], for n-heptane/n-octane mixtures in silicalite. A decrease in thediffusivity of fast n-heptane was observed as the loading of slow n-octane increased,while diffusivity of C8−alkane did not change in the presence of fast alkane. First ofall, it might be explained by the experimental conditions, where the temperature washigher (448-498 K) and total pressure significantly lower (13 Pa), so one would expectthe pore occupancy to be quite low. Under these conditions repulsive interactions


negligibly small. Since the molecules are randomly distributed, slow alkane can causethe fast one to move with the same rate.

Thus, in mixtures, diffusion of both components more or less is affected by thepresence of each other. Mainly, the difference with the single component diffusionis caused by the additional stronger or weaker repulsive interactions with the otheralkane.

7.4 Conclusions

TEX-PEP allowed us to study a concentration dependence of self-diffusion of n-pentane and its diffusion in mixtures with n-hexane in silicalite.

Diffusion of n-pentane was found to be independent on the loading at high tem-peratures, while at lower temperatures, a slow increase in the self-diffusion coefficientwas observed. We believe, that there are some repulsive interactions between themolecules located in the intersections of the straight and zigzag channels and thosesited in the adjacent straight channels, that cause an increase in the mobility of themolecules, [11]. However, n-pentane is also affected by the pore occupancy, whichcauses a decrease in the diffusivity. These two factors are competing. At high tem-peratures, they compensate each other, because the repulsions are weaker. As a resultdiffusion is independent on concentration. At low temperatures, repulsive interactionsare stronger, and diffusivity increases. It also has an effect on the apparent activationenergy.

In mixtures with n-hexane, of which the diffusivity is significantly affected by themolecular repulsion, n-pentane molecules are accelerated by stronger repulsion withn-hexane. Thus, at high loading of h-hexane, the mobility of n-pentane moleculesbecomes very close to that of n-hexane. On the contrary, n-hexane diffuses slowerin mixtures with n-pentane compared to the single component under the same con-ditions. With increasing fraction of n-pentane repulsive interaction become weaker,and the diffusivity of n-hexane becomes even slower than that of n-pentane. Thus, inthe absence of strong repulsive interactions n-hexane diffusivity is slower comparedto n-pentane because n-hexane is a bulkier molecule. At high n-hexane loadings inmixtures, the diffusion is not influenced by the presence of n-pentane.

References


[2] Schuring, D.; Koriabkina, A.O.; de Jong, A.M.; Smit, B.; van Santen, R.A. J.Phys. Chem. B 105(32) (2001) 7690-7698.

[3] Krishna, R.; Smit, B.; Vlugt, T.J.H. J. Phys.Chem. A 102(40) (1998) 7727-7730.

[4] Paschek, D.; Krishna, R. Phys. Chem. Chem. Phys. 3 (2001) 3185-3191.

[5] Krishna, R. Chem. Ing. J. 84 (2001) 207-214.

[6] Krishna, R.; Chem. Phys. Lett. 355 (2002) 483-489.

[7] Krishna, R.; Paschek, D. Phys. Chem. Chem. Phys. 4 (2001) 1891-1898.

[8] Masuda, T.; Fujikata, Y.; Ikeda, H., Hashimoto, K. Microporous MesoporousMater. 38 (2000) 323-332.

[9] Niessen, W.; Karge, H.G; Microporous Mater. 1 (1993) 1.

[10] Funke, H.H.; Argo, A.M; Falconer, J.L.; Noble, R.D. Ind. Eng. Chem. Res. 36(1997) 137.


[12] Masuda, T.; Fujikata, Y.; Ikeda, H.; Hashimoto, K. Microporous MesoporousMater. 2000 (38) 323-332.

[13] Schumacher, R.R.; Anderson, B.G.; Noordhoek, N.J.; de Gauw, F.J.M.M.; deJong, A.M.; de Voigt, M.J.A.; van Santen, R.A. Microporous Mesoporous Mater.35-36 (2000) 315-326.

[14] Smit, B.; Maesen, T. Nature 1995 (374) 42-44





108 REFERENCES


[20] Jost, J. Diffusion in solids, liquids, gases, Academic Press, New York, 1960.

[21] Krishna, R.; Paschek, D. Chem. Ing. J. 2002 (85) 7-15.

[22] Vlugt, T.J.H.; Krishna, R.; Smit, B. J. Phys. Chem. B B,1999 (103(7)) 1102-1118


[24] Datema, K.P; den Ouden, C.J.J.; Ylstra, W.D.; Kuipers, H.P.C.E.; Post M.F.M.;Karger,J. J. Chem. Soc. Faraday Trans. 87 (1991) 1935.

[25] van de Graaf, J.M.; Kapteijn, F.; Moulijn, J. Microporous Mesoporous Mater.35-36 (2000) 267-281.

[26] Hayhurst, D.T.; Paravar, A. Zeolites 8 (1988) 27.

[27] Eic, M.; Ruthven, D.M. Zeolites:Facts, Figures, Future, ed. Jacobs, P.A. and vanSanten, R.A. 1989, Elsevier, Amsterdam, pp.897-906.

[28] Song, L.; Rees, L.V.C. Proc. 12th Int.Zeolites Conf., 1999, pp.67-74.

[29] Talu, O.; Sun, M.S.; Shah, D.B. AIChE Journal 44(3) (1998) 681-694.

[30] Nijhuis, T.A.; van den Broeke, L.J.P.; van de Graaf, J.M.; Kapteijn, F.; Mak-kee, M.; Moulijn, J.A. Chem. Eng. Sci. 52(19) (1997) 3401-3404.

[31] Chen, N.Y.; Degnan, T.; Smith, C.Molecular transport and reaction in zeolites,1994, John Wiley & Sons.

[32] Sun, M.S.; Talu, O.; Shah, D.B.; J. Phys. Chem. B 100 (1996) 17276-17280.

Summary

Diffusion of alkanes in MFI-type zeolites

Zeolites are widely used in the petrochemical industry as catalysts and selectivesorbents. Due to their application, the adsorptive and diffusive properties of hydro-carbons in these materials have attracted significant interest. In order to enhanceor model a catalytic or separation process a thorough understanding of adsorptionand diffusion of the reactants and reaction products is required. This is because themolecules participating in the process have to adsorb and diffuse inside the zeolitepores towards the active sites or to diffuse out of the zeolite crystal. These processesmight seriously affect the behavior of the catalyst or sorbent. In this work MFI-typezeolites have been studied, which are one of the most industrially applied zeoliticmaterials. The zeolites consist of the intersecting straight and zigzag channels with 4intersections per unit cell. The diameter of the pores is close to the kinetic diameterof many hydrocarbons.

The present research have been focused on the diffusion of single linear andmonobranched alkanes and behavior of those compounds in their mixtures. To explorethose topics the PEP technique has been applied. This is a powerful technique foran in situ investigation of the adsorptive and diffusive properties of various alkanesin zeolites. PEP is based on the use of radio-labelled molecules. This is a uniquemacroscopic method, which allows to measure self-diffusion of alkanes in zeolites.The greatest advantage of the method compared to the others is the possibility tostudy multi-component mixtures under reaction conditions. Often, possibilities ofthese techniques for studying mixtures are limited to the small molecules, or in somecases, mixtures of isomers can not be investigated. All these limitations are avoidedwith the PEP technique, which makes the method really exclusive.

In Chapter 3 an investigation of diffusion and adsorption of binary mixtures oflinear (n-hexane) and branched (2-methylpentane) alkanes in silicalite have beendiscussed. It turned out that not only the size but the siting of the molecules in theparticular zeolite plays an important role in the behavior of the mixture components.A slight preference for the adsorption of n-hexane over 2-methylpentane was observedbecause of the higher packing efficiency of linear alkane. Indeed, it can be sitedanywhere in the pore system of silicalite, while the branched alkane is preferentiallyadsorbed in the intersection between straight and zigzag channels. This is also reflectedin the character of diffusion for both components. The diffusion of the fast n-hexanemolecules is strongly influenced by the presence of the slowly diffusing 2-methylpen-tane. It was observed that when the 2-methylpentane loading has reached approxi-

110 Summary

mately 2.75 molecules per unit cell, a drastic decrease in the diffusivity of n-hexaneoccurs. Most probably, this is caused by the blocking of the channel intersectionsby the slowly moving branched alkanes, as this sudden drop takes place at a 2-methylpentane loading, which corresponds to the situation when approximately 3 outof 4 channel intersections are occupied by iso-hexane. This indicates that the sitingof the different components, together with the topology of the zeolite pores, playan important role in the character of adsorption and diffusion of multi-componentmixtures in zeolites.

In Chapter 4 the influence of the interactions of the mixture components with theacid sites has been studied. A comparison between the behavior of binary mixturesdiscussed above in silicalite and acidic HZSM-5 have led to the following conclusions.First of all, indeed, the interaction with the acid sites causes a decrease in thediffusivities of both alkanes, while at the apparent activation energy of diffusion forboth alkanes were found to be very close in silicalite and HZSM-5 measured at 6.6 kPa.A significant preferential adsorption of the linear hexane over the 2-methylpentanewas observed in HZSM-5 zeolite. Probably it is caused by the ability of n-hexaneto form a bimolecular complex with one acid site and its higher packing efficiency.On the contrary, 2-methylpentane loadings in mixtures in silicalite and HZSM-5 arevery close. In HZSM-5 zeolite the diffusivity of linear alkane in the mixtures withbranched alkane is influenced by two factors: a) interaction with the acid sites, whichdecreases the diffusivity by approximately a factor of two; b) the presence of 2-methylpentane, which has a ten-times lower diffusivity. At low loadings of branchedalkane, the interactions with the acid sites is prevailing. As soon as the loading of iso-hexane exceeds approximately 2.7 molecules per unit cell, the effect of the Brønstedsites on the diffusion becomes negligible compared to the blockage of the pore networkconnection by the branched alkane.

Apparently, the diffusivity is influenced by the concentration of the moleculesinside zeolite pores. The concentration dependence of the diffusion coefficients ofsingle branched and linear alkanes has been studied. In Chapter 5 diffusion of 3-methylpentane in silicalite has been discussed. It was found that diffusivity of thisalkane decreases with the pore occupancy. The dependence can be best approximatedby an exponential decay. It was shown that the pre-exponential factor D0, which isrelated to the jump frequency is concentration dependent, while the activation energyof diffusion does not change with the loading. Since 3-methylpentane molecules mostprobably jump to the channel intersections, the diffusion rate is mainly determinedby the availability of the intersection to which a molecule attempts to jump. As aconsequence of the concentration dependence of the pre-exponential factor D0, theapparent activation energy increases with partial pressure. Usually, in the macroscopicexperiments the measurement of the activation energy is performed under fixed partialpressure conditions. If the conditions are different from those used in other studiesit can be the reason for the discrepancies in the apparent activation energy anddiffusivities obtained by different authors. Values of activation energy of diffusionshould be compared only measured at zero pore filling or under fixed loading conditions.

Summary 111

For linear C5− and C6−alkanes the dependence on the concentration turnedout to be quite different from that of the branched alkane. For n-hexane, whichis discussed in Chapter 6 the diffusivity increases with the concentration at loadingsup to 4 molecules per unit cell. We believe that this behavior is caused by additionalrepulsive interactions between the molecules situated in the straight channels and theintersections. Since n-hexane molecules are randomly distributed in the zeolite matrix,the amount of the molecules in the intersections and adjacent straight channels isincreasing with loading, therefore the repulsive interactions increase. This leads to theincrease in the diffusivity. Even though there is always the effect of the pore occupancyas discussed in Chapter 5, in this case the repulsion plays a dominating role. It was alsoobserved that at high temperatures, the repulsive interactions between the moleculesdo not have a noticeable effect on the diffusion. Although repulsion between themolecules located in the intersections and adjacent zigzag channels may also occur,this does not have a significant influence on the average diffusivity measured here,because diffusivity in zigzag channels is an order of magnitude lower compared tothat in the straight channels. At higher loadings, the diffusivity does not changebecause of the re-arrangement of n-hexane molecules in the channels due to a possiblecommensurate freezing of the molecules in zigzag channels, so that the repulsiveinteractions between the molecules are diminished and the pore occupancy effect,which causes a decrease in the molecular motion, starts to dominate. In HZSM-5,this diffusion trend remains, while the diffusion coefficient is two times lower. Webelieve that in acidic zeolite, the interaction with the acid sites causes a an increasein the apparent activation energy, but it is only observed at low partial pressures,otherwise the repulsion effect is dominating.

For n-pentane, the effect of the repulsive interactions on the molecular mobilitywas found to be less dramatic compared to n-hexane. It is discussed in Chapter 7.Diffusion of n-pentane was found to be independent on the loading at high tempe-ratures, while at lower temperatures, a slow increase in the self-diffusion coefficientwas observed. We believe, that there are repulsive interactions between the moleculessimilar to n-hexane that cause an increase in the mobility of the molecules. Theyare weaker compared to n-hexane because the molecules are shorter than n-hexane,so that when they are adsorbed in the adjacent sites the distance between themis bigger than between n-hexane molecules and the repulsions are weaker. At hightemperatures, the pore occupancy effect and the repulsion effect compensate eachother, because the repulsions are weaker. As a result diffusion is independent onconcentration. At low temperatures, repulsive interactions are stronger, and diffusivityincreases. It also has an effect on the apparent activation energy.

In Chapter 7 the behavior of mixtures of n-pentane and n-hexane are discussed.We observed that n-pentane molecules accelerate in the presence of n-hexane. This isbelieved to be caused by stronger repulsive interactions with n-hexane than betweenn-pentane molecules along. Thus, at high loading of h-hexane, the mobility of n-pentane molecules becomes very close to that of n-hexane. On the contrary, n-hexane diffuses slower in mixtures with n-pentane compared to the single component

112 Summary

under the same conditions. With increasing fraction of n-pentane repulsive interactionbecome weaker, and the diffusivity of n-hexane becomes even slower than that of n-pentane. Thus, in the absence of strong repulsive interactions n-hexane diffusivityis slower compared to n-pentane because n-hexane is a bulkier molecule. At highn-hexane loadings in mixtures, the diffusion is not influenced by the presence of n-pentane.

Therefore, diffusion of the alkane molecules in MFI-type zeolites is determinedby several factors, internal and external. Siting of the molecules plays a crucial roleas it might induce additional interactions between the molecules, such as repulsiveinteractions, when the molecules situated too close to each other. In mixture diffusiona pore blockage by one of the components might occur. Zeolite-adsorbate interactionsalso play an important role in the diffusion, such as interaction with the acid sites.Concentration of the molecules is another important external factor because it deter-mines the significance of all above mentioned factors on the diffusion process.

Samenvatting

Diffusie van alkanen in MFI-type zeolieten

Zeolieten worden op grote schaal toegepast in de petrochemische industrie alskatalysatoren en selektieve adsorptie middelen. De adsorptie en diffusie eigenschappenvan koolwaterstoffen in deze materialen hebben als gevolg van hun toepassing veelaandacht gekregen. Gedegen kennis van adsorptie en diffusie van reactanten en reactie-producten is noodzakelijk om katalytische en scheidingsprocessen to modeleren en teoptimaliseren. De reden hiervoor is dat moleculen die bij deze processen betrokken zijnin de zeoliet porien moeten adsorberen en naar de actieve plaatsen moeten diffunderenen ook weer uit het zeoliet kristal. Dit kan grote invloed hebben op het gedrag vande katalysator of sorbent. In dit onderzoek zijn MFI-type zeolieten bestudeerd, eenvan de in de industrie meest toegepaste typen. De zeolieten bevatten elkaar kruisenderechte en zigzag kanalen met vier kruispunten per eenheidscel. De porie-diameter lichtdicht bij de kinetische diameter van veel koolwaterstof moleculen.

Dit onderzoek heeft zich gericht op de diffusie van lineaire en enkel-vertaktealkanen en hun gedrag in mengsels van deze componenten. Om deze onderwerpente onderzoeken is de positron emmissie profilering (PEP) techniek toegepast. Dit iseen krachtige techniek voor in situ onderzoek van adsorptie en diffusie eienschappenvan alkanen in zeolieten. PEP is gebaseerd op het gebruik van radioactief gelabeldemoleculen. Het is een unieke macroscopische mehode, die het mogelijk maakt dezelfdiffusie van alkanen in zeolieten te meten. Het grootste voordeel van deze methodet.o.v. andere is de mogelijkheid om meer-component mengsels onder reactie conditieste bestuderen. Vaak zijn de mogelijkheden van technieken om mengsels te bestuderenbeperkt tot kleine moleculen en in sommige gevallen kunnen mengsel van isomerenniet onderzocht worden. Al deze beperkingen worden vermeden met de PEP techniek,wat deze methode exclusief maakt.

In Hoofdstuk 3 wordt het onderzoek van de diffusie en adsorptie in silicalietvan lineaire (n-hexaan) en vertakte (2-methylpentaan) alkanen in binaire mengselsbesproken. Het blijkt dat niet alleen de grootte maar ook de situering van de moleculenin de betreffende zeoliet een belangrijke rol speelt in het gedrag van de mengsels vande componenten. Een lichte voorkeur voor de adsorptie van n-hexaan ten opzichtevan 2-methylpentaan is waargenomen, wat te wijten is aan een hogere pakkings-efficientie van het lineaire alkaan. Dit kan overal in het poriesysteem van silicalietworden geplaatst terwijl het vertakte alkaan bij voorkeur op de kruispunten vanlineaire en zigzag kanalen wordt geadsorbeerd. Dit komt ook tot uitdrukking in hetkarakter van de diffusie van beide componenten. De diffusie voor de snelle n-hexaan

114 Samenvatting

moleculen wordt sterk beinvloed door aanwezigheid van de langzaam diffunderende2-methylpentaan. Een drastische afname van de diffusie van n-hexaan werd waar-genomen als de 2-methylpentaan belading ongeveer 2.75 moleculen per eenheidscelheeft bereikt. Hoogst waarschijnlijk wordt dit veroorzaakt door het blokkeren vande kanaal-kruispunten door de langzaam bewegende vertakte alkanen, omdat deplotselinge afname optreedt bij een 2-methylpentaan belading, die overeenkomt metde situatie waarbij ongeveer 3 van de 4 kanaal-kruispunten zijn bezet met iso-hexaan.Dit geeft aan dat de situering van de verschillende componenten tezamen met detopologie van de zeoliet-porien een belangrijke rol speelt in het adsorptie en diffusiekarakter van multi-component mengsels in zeolieten.

In Hoofdstuk 4 is de invloed van de interactie van de componenten in het mengselmet de zure plaatsen in de zeoliet bestudeerd. Een vergelijking tussen het gedragvan binaire mengsels zoals hierboven beschreven in silicaliet en zure HZSM-5, leiddentot de volgende conclusies. Allereerst veroorzaakt de interactie met de zure plaatseninderdaad een afname in de diffusie voor beide alkanen, terwijl de schijnbare active-ringsenergie voor diffusie voor beide alkanen erg dicht bij elkaar blijkt te liggen insilicaliet en HZSM-5, gemeten bij 6.6 kPa. Een significante preferentiele adsorptievoor het lineaire hexaan t.o.v. 2-methylpentaan werd waargenomen in HZSM-5.Waarschijnlijk wordt dit veroorzaakt door het vermogen van n-hexaan om een bimo-leculair complex te vormen met een zure site en door zijn hogere pakkings-efficientie.Daar tegenover staat dat de belading van 2-methylpentaan in mengsels in silicalieten HZSM-5 vrijwel gelijk is. In HZSM-5 wordt de diffusie van het lineaire alkaan inmengsels met het vertakte alkaan beinvloed door twee factoren: a) interactie metde zure sites, waardoor de diffusie met ongeveer een factor twee afneemt; b) deaanwezigheid van 2-methylpentaan, waarvan de diffusie tien keer zo langzaam is. Bijlage beladingen van vertakte alkanen heeft de interactie met de zure sites de overhand.Zo gauw als de belading van iso-hexaan meer dan circa 2.7 moleculen per eenheidscelbedraagt, wordt het effect van de Bronsted sites op de diffusie verwaarloosbaar invergelijking met het blokkeren van het porie-netwerk door de vertakte alkanen.

Blijkbaar wordt de diffusie beinvloed door de concentratie van de moleculen inde zeoliet porien. De concentratie afhankelijkheid van de diffusie coefficienten vande enkel-vertakte en lineaire alkanen is onderzocht. In Hoofdstuk 5 is de diffusievan 3-methylpentaan in silicaliet besproken. Er werd gevonden dat de diffusie vandit alkaan afneemt met de bezettingsgraad van de porien. De afhankelijkheid kan hetbest worden benaderd door een exponentiele afname. Er werd aangetoond dat de pre-exponentiele factor D0, gerelateerd aan de sprongfrequentie, concentratie afhankelijkis terwijl de activeringsenergie voor diffusie niet veranderd met de belading. Omdat 3-methylpentaan moleculen hoogst waarschijnlijk naar de kruisingen tussen de kanalenspringen, wordt de diffusiesnelheid voornamelijk bepaald door de beschikbaarheid vande kruising waar het molecule naar toe wil springen. Als gevolg van de concentratieafhankelijkheid van de pre-exponentiele factor D0 neemt de schijnbare activerings-energie toe met de partiele druk. In macroscopische experimenten wordt het metenvan de activeringsenergie gewoonlijk uitgevoerd bij een vaste partiele druk. Als deze

Samenvatting 115

omstandigheden anders zijn dan in andere studies kan dat de reden zijn voor dediscrepanties in de schijnbare activeringsenergie en diffusiesnelheden verkregen doorverschillende auteurs.Waarden voor de activeringsenergie voor diffusie kunnen alleenvergeleken worden als deze gemeten zijn bij een belading die aan nul nadert of onderomstandigheden met een vaste belading.

Voor lineaire C5− en C6−alkanen bleek de afhankelijkheid van de concentratiebehoorlijk anders dan voor de vertakte alkanen. Voor n-hexaan, besproken in Hoofd-stuk 6, neemt de diffusie toe met de concentratie bij beladingen tot 4 moleculen pereenheidscel. We nemen aan dat dit gedrag wordt veroorzaakt door extra repulsieveinteracties tussen de moleculen die zich bevinden in de rechte kanalen en de kruispun-ten. Omdat n-hexaan moleculen random verdeeld zijn in de zeoliet matrix, neemt hetaantal moleculen dat zich in de kruispunten en aanliggende rechte kanalen bevindttoe met de belading, waardoor de repulsieve interacties toenemen. Dit leidt tot eentoename in de diffusie. Zelfs al is het effect van de porie-bezetting, zoals besprokenin Hoofdstuk 5, altijd aanwezig, in dit geval speelt de repulsie een dominante rol. Erwerd ook waargenomen dat bij hoge temperaturen die repulsieve interacties tussende moleculen geen merkbaar effect hebben op de diffusie. Hoewel afstoting tussenmoleculen die zich in de kruispunten en aangrenzende zigzag kanalen bevinden ookkan voorkomen, heeft dit geen significante invloed op de hier gemeten gemiddeldediffusiesnelheden, omdat de diffusie in zigzag kanalen een orde van grootte lanzameris dan die in de rechte kanalen. Bij hogere beladingen verandert de diffusie niett.g.v. een herrangschikking van de n-hexaan moleculen in de kanalen veroorzaaktdoor een commensuraat bevriezen van de moleculen in de zigzag kanalen, zodat derepulsieve interacties tuusen de moleculen teniet worden gedaan en het effect vanporie-bezetting, wat een afname veroorzaakt in de moleculaire beweging, begint tedomineren. In HZSM-5 blijft de trend in de diffusie behouden, terwijl de diffusie-coefficient twee keer zo laag is. We veronderstellen dat in de zure zeoliet de interactiemet de zure sites een toename in de schijnbare activeringsenergie veroorzaakt. Ditwerd echter alleen bij lage partiaal drukken waargenomen, anders overheerst het effectvan de repulsie.

Voor n-pentaan is het effect van de repulsieve interacties op de moleculaire mo-biliteit minder dramatisch in vergelijking met n-hexaan. Dit wordt besproken inHoofdstuk 7. De diffusie van n-pentaan bleek onafhankelijk van de belading bijhoge temperaturen, terwijl bij lage temperaturen een langzame toename van de zelf-diffusie coefficient werd waargenomen. We veronderstellen dat repulsieve interactieseen vergelijkbare rol spelen als bij n-hexaan, welke een toename in de moleculairemobiliteit veroorzaken. Zij zijn zwakker in vergelijking met n-hexaan omdat de mole-culen korter zijn, zodat wanneer deze geadsorbeerd zijn op aangrenzende plaatsende afstand tussen hen groter is dan tussen n-hexaan moleculen en de repulsieveinteracties dus zwakker zijn. Bij hoge temperatuur compenseren het effect van deporie-bezetting en van de repulsie elkaar, omdat de afstoting zwakker is. Hierdoorwordt de diffusie onafhankelijk van de concentratie. Dit heeft ook een effect op deschijnbare activeringsenergie.

116 Samenvatting

In Hoofdstuk 7 wordt het gedrag van mengsels n-pentaan en n-hexaan beschreven.We namen waar dat n-pentaan moleculen versnellen in de aanwezigheid van n-hexaan.Er wordt verondersteld dat de oorzaak ligt in sterkere repulsieve interacties met n-hexaan dan tussen n-pentaan moleculen onderling. Hierdoor wordt bij hoge n-hexaanbelading, de mobiliteit van n-pentaan moleculen ongeveer gelijk aan die van n-hexaan.Daar tegenover staat dat de diffusie van n-hexaan langzamer wordt in mengselsmet n-pentaan in verhouding tot de pure component onder gelijke omstandigheden.Met een toenemende fractie van n-pentaan, wordt de repulsieve interactie zwakkeren de diffusiesnelheid van n-hexaan zelfs langzamer dan die van n-pentaan. In deafwezigheid van repulsieve interacties wordt de diffusie van n-hexaan dus langzamer invergelijking met n-pentaan, omdat n-hexaan een omvangrijker molecule is. Bij hoge n-hexaan beladingen in mengsels wordt de diffusie niet beinvloed door de aanwezigheidvan n-pentaan.

De diffusie van alkaan moleculen in MFI-type zeolieten wordt dus bepaald door eenaantal factoren: intern en extern. De lokatie van de moleculen speelt een cruciale rol,omdat dit additionele interacties tussen de moleculen kan induceren, zoals repulsieveinteracties als de moleculen te dicht bij elkaar zitten. Bij diffusie in mengsels kan hetblokkeren van de porien door een van de componenten optreden. Zeoliet-adsorbaatinteracties spelen ook een belangrijke rol bij de diffusie, zoals interacties met dezure sites. De concentratie van de moleculen is een andere belangrijke externe factor,omdat dit de belangrijkheid van alle voorgenoemde factoren voor het diffusieprocesbepaald.

Acknowledgements

First of all, I would like to thank my promotor Prof.dr. R.A. van Santen for givingme an opportunity to work in his research group. Rutger, thank you very much formany interesting discussions and your continuous support.

I would like to thank Dr.ir. Arthur de Jong for being my co-promotor. Arthur,thanks a lot for your support and help during all these years, with discussions,experiments and thesis. Thanks to you I know how to start up the cyclotron!

Dr.ir. E.J.M. Hensen is acknowledged for the fruitful discussions on the last twochapters of this thesis. Emiel, I am very grateful for your help and enthusiasm. Thankyou very much for your constructive criticism and suggestions.

I would like to thank the members of the reading committee of this thesis:Prof.dr Freek Kapteijn, Prof.dr. Berend Smit and Prof.dr. Martien J.A. de Voigtfor the time they spent reading my thesis and for their comments and suggestionsthat helped to improve the thesis. I would also like to express my gratitude toProf.dr. R. Krishna and Prof.dr. B. Smit for the discussions we had on the results ofthe last two Chapters.

I am very grateful to Dr.ir. D. Schuring for creating and allowing me to usehis beautiful SimanPEP programm for the interpretation of the PEP experiments.Danny, thank you very much for the discussions we had.

I want to thank Joop van Grondelle for the technical support of the experimentalwork. Dear Joop, as we would say in Russia, you have “golden hands”. I was verylucky to work with you.

I would also like to thank Erik, Frits, Jan, Rinus for the technical support incyclotron. I appreciate very much help with the PEP experiments provided by Emielvan Kimmenade and Martijn Kersbulck.

Dr. Frank de Gauw, Dr. Bruce Anderson, Dr. Sing, Nick Lousberg, Marco Hendrix,and Dr. Pieter Magusin thank you all very much for sharing your knowledge on howto perform reaction kinetics experiments, how to make large mordenite crystals, SEMand XRD analysis of them and NMR study.

I thank all current and former members of the department for creating a welcomingand friendly atmosphere. Especially I would like to thank members of the MetalCatalysis group. First of all, my officemates Mayela and Darek. Thank you verymuch for your friendship and support during all these years. Darek, thanks a lot foryour patience in answering my numerous questions and your help in adaptation tothe Dutch life. My best friend Mayela, thank you very much for your friendship andcompanionship in different kinds of sports we tried (I do not want to list them dueto a limited space). I have learnt a lot about Mexican people and culture. I hope, oneday I will visit your great country. I also want to thank another Mexican, Maru, forbeing our jogging coach and a very good friend.

118 Acknowledgements

I was very lucky to have Christophe and Zhu as my colleagues. Christophe, thanksa lot for sharing your knowledge on how to grow zeolite crystals.

I would like to thank Ine, Ingrid and Joyce for their professionalism and helpduring all these years.

I want to thank another very important University organization: Student SportCenter of Eindhoven. I thank all the stuff for the work they do, fantastic facilitiesand for friendly atmosphere they created there.

Finally, I want to thank all my family for their help and support.

List of publications

• Adsorption and diffusion of alkanes and their mixtures in silicalite studied withpositron emission profiling technique,A.O. Koriabkina, D. Schuring, A.M. de Jong, J. van Grondelle, R.A. van Santen,Stud. Surf. Sci. 135 (2001) 3129-3136.

• Adsorption and diffusion of n-hexane/2-methylpentane mixtures in zeolite Sili-calite: experiments and modelling,D. Schuring, A.O. Koriabkina, A.M. de Jong, B. Smit, R.A. van Santen,J. Phys. Chem. B 105(32) (2001) 7690-7698.

• Influence of the acid sites on diffusion of hexanes and their mixtures withinMFI-zeolites,A.O. Koriabkina, A.M. de Jong, J. van Grondelle, D. Schuring, R.A. van Santen,J. Phys. Chem. B 106(37) (2002) 9559-9566.

• Diffusion of 3-methylpentane in silicalite-1: concentration dependence,A.O. Koriabkina, A.M. de Jong, D. Schuring, R.A. van Santen,J. Phys. Chem. B (2002) submitted.

• The application of non-hydrothermally prepared stevensites as support for hy-drodesulfurization catalyst,M. Sychev, R. Prihodko, A. Koriabkina, E.J.M. Hensen, J.A.R. van Veen,R.A. van SantenStud. Surf. Sci. 143 (2002) 257-265.

• Factors enhancing diffusion of n-hexane in MFI-zeolites,A.O. Koriabkina, A.M. de Jong, E.J.M. Hensen, R.A. van Santen,Phys. Chem. Chem. Phys. submitted (2003) .

• Diffusion of linear alkanes and their mixtures in silicalite,A.O. Koriabkina, A.M. de Jong, E.J.M. Hensen, R.A. van Santen,Phys. Chem. Chem. Phys. submitted (2003) .

Curriculum Vitae

Alina Koriabkina was born on June 30, 1976 in Akademgorodok (Novosibirsk), Russia.In 1993 she graduated School N -◦163 with distinction in Akademgorodok (Novosibirsk).The same year she started her study at Chemistry and Ecology department atNovosibirsk State University. In June, 1997 she received her Bachelor degree. InSeptember 1997 she started to work on her Master thesis entitled “Study of selectiveammonia oxidation to nitrous oxide on Mn-Bi/α-Al2O3 catalyst” at Boreskov Instituteof Catalysis (Novosibirsk) in the group of prof.dr. A.S. Noskov. In June, 1999 shereceived her Master diploma in Chemistry and Catalysis.

Since September, 1999 she was employed as AIO at Eindhoven University ofTechnology, Schuit Institute of Catalysis in the group of Inorganic Chemistry andCatalysis lead by Prof.dr. R.A. van Santen. She worked on the experimental studyof adsorption and diffusion of various alkanes in zeolites using PEP technique. Themain results of the work are described in this thesis.

diffusion of alkanes in mfi-type zeolites · diffusion of alkanes in mfi-type zeolites citation for...

Documents