The potential of zeolite membranes in. hydroisomerization processes. Maikel L. Maloncy

Transcription

1 The potential of zeolite membranes in hydroisomerization processes Maikel L. Maloncy

2 ISBN: Cover: cross section of a zeolite membrane (top), a gasoline pump (middle) and a tree (bottom) symbolizing the potential of membranes in the production of high quality, environmentally friendly gasoline components. Cover by Dr. M.S. Hamdy.

3 The potential of zeolite membranes in hydroisomerization processes Proefschrift ter verkrijging van de graad van doctor aan de Technische Universiteit Delft, op gezag van de Rector Magnificus prof. dr. ir. J.T. Fokkema, voorzitter van het College voor Promoties, in het openbaar te verdedigen op dinsdag 25 april 2006 om 15:30 uur door Maikel Laurence MALONCY Mestre em Engenharia Química, Universidade Federal de São Carlos, Brazilië geboren te Paramaribo, Suriname

4 Dit proefschrift is goedgekeurd door de promotoren: Prof. dr. J.A. Moulijn Prof. dr. J.C. Jansen Samenstelling promotiecommissie: Rector Magnificus Prof. dr. J.A. Moulijn Prof. dr. J.C. Jansen Prof. dr. ir. G. Baron Prof. dr. ir. H. van Bekkum Prof. dr. D. Cardoso Prof. dr. D. Cazorla Amorós Dr. L. Gora voorzitter Technische Universiteit Delft, Nederland, promotor Universiteit van Stellenbosch, Zuid Afrika, promotor Vrije Universiteit Brussel, België Technische Universiteit Delft, Nederland Federal University of São Carlos, Brazilië University of Alicante, Spanje Polish Academy of Sciences, Polen, adviseur Reservelid Prof. dr. F. Kapteijn Technische Universiteit Delft, Nederland

5 ! ""

6

7 Contents Chapter 1. General introduction and thesis outline 1. General 1 2. Zeolites 2 3. Zeolite membranes 5 4. Zeolite membrane reactors 8 5. Hydroisomerization and its position in fuel supplies Outline of the thesis 11 References 11 Chapter 2. The performance of silicalite-1 membranes in the separation of linear and branched alkanes Abstract Introduction Experimental Results and discussion Conclusions 24 References 25 Chapter 3. Preparation of zeolite beta membranes and their pervaporation performance in separating di- from monobranched alkanes Abstract Introduction Experimental Results and discussion Concluding remarks 38 Acknowledgements 39 References 40 Chapter 4. Design of a state of the art C 5 /C 6 hydroisomerization process Abstract Introduction Process description and simulation Results and discussion 47

8 3.1. Reactor Adsorber/Desorber Overall simulation Economics Total investment Operating costs Product based cost (cash) factor Concluding remarks 57 References 58 Chapter 5. Hydroisomerization of hexane within a reactor composed of a tubular silicalite-1 membrane packed with Pt-loaded chlorided alumina catalyst Abstract Introduction Experimental Results and discussion Conclusions 67 Acknowledgement 67 References 67 Chapter 6. Hydroisomerization of heptane: mechanistic aspects and industrial challenges Abstract Introduction Synthesis and reaction mechanisms Reaction intermediates Formation of carbenium ions Reaction of carbenium ions Mechanisms of acid-catalyzed isomerization Monomolecular mechanism Bimolecular mechanism Relation between isomerization and cracking Recapitulation of reactions via protonated cyclopropane Catalysts Monofunctional acid catalysts Bifunctional acid catalysts Bifunctional zeolite-supported catalysts Influence of the zeolite structure on isomerization

9 reactions Effect of temperature, metal loading and acidity The oil industry Recent developments and trends Appraisal of the heptane isomerized product Challenges in the heptane hydroisomerization process Concluding remarks 96 Acknowledgement 96 References 96 Chapter 7. Technical and economical evaluation of a zeolite membrane based heptane hydroisomerization process Abstract Introduction Process concept Process description and simulation Results and discussion Reactor design Membrane design Overall simulation Economics Concluding remarks 115 Acknowledgments 116 Notation 116 References 117 Summary 121 Samenvatting 125 Acknowledgements 129 Publications 133 Curriculum vitae 135

10

11 1 General introduction and thesis outline 1. General The oil we are burning in two centuries took hundreds of millions of year to form from primeval plants. The average light vehicle burns 100 x the car weight in prehistoric plants in the form of gasoline. Only 12.5 % of the fuel energy reaches the wheels, 7% is used in accelerating the car and less than 1% moves the driver [1]. The draconically inefficient use of our energy sources in car transportation is alerting the community in other areas as well. Gradually, a mind setting turns to focus on higher efficiencies. A typical example of the size of changes that should be made is given in the Innovative Roadmap Separation Technology, 2004 [2]. The authors recommend to reduce energy consumption in separation processes up to 75%. Such numbers cannot be realised by improving state-of-the-art systems but only by developing revolutionary, new technologies, based on new materials. The content of this thesis pertains to a specific case in energy efficiency, the separation of linear and branched alkanes using membranes as an alternative for distillation columns and adsorption towers. Production processes of branched alkanes applying membranes are suggested. The membranes used for the study were zeolite-based materials as they have extremely interesting properties for application in this area [3]. In the following paragraphs the zeolite properties and the zeolite membrane synthesis and performance in general are given.

12 Chapter 1 2. Zeolites Zeolites are crystalline, microporous, aluminotectosilicates. The tectosilicate configuration is based on so-called corner-connected tetrahedral units, see Figure 1. (a) (b) Figure 1. (a) tetrahedral unit comprising formally 1x silicon and 4 x 1/2 oxygen. (b) 2 tetrahedral units sharing 1 oxygen atom. The total charge of a Si unit is zero. In the case of an Al-unit the charge is 1. The corner-connected tetrahedral units result in a microporous well ordered topology, Figure 2, with remarkable specifications, see intermezzo 1. The model presented in Figure 2 for the MFI type zeolite shows the so-called straight channels while the so-called sinusoidal channels are running in the plane of the picture, and intersecting the straight channels, see also Figure 3b. Each pore direction has a unique pore diameter, providing an opportunity to separate mixtures of small molecules. Figure 2. Typical presentation of the topology of the zeolite type MFI which can be found for other zeolites as well in the Atlas of Zeolite Structures [4]. Only the connections of the Si and Al centers is given. The oxygen atoms are omitted for reasons of clarity. 2

13 General introduction and thesis outline Intermezzo 1 Remarkable specifications of zeolites are the following: The specific pore wall area can be about m 2 /g. The external area, depending on the particle size can be 5% up to 50% of the pore wall area. Each atom of the framework is part of the pore wall. Minor changes such as a defect or a modification in the framework can change the adsorption or catalytic properties. Silicalite-1 is an all silica lattice and therefore hydrophobic, see also Table 2. However, the material can show uptake of water. This can only be explained by defects in the framework generating hydrophilic centers. Even at a Si/Al ratio of Brønsted acid cracking catalytic activity can be still present and measurable in cracking of e.g. n-hexane. The pore volume can occupy half of the crystal volume. The relative density can be between 1.4 and 1.9 Zeolites have ferro-elastic properties, implying that upon external stimuli the pore aperture can deform and adsorb unexpected molecule sizes. Molecules of naphthalene with a kinetic diameter of 0.72 x 0.38 nm were adsorbed in the pores of MFI, which are 0.51 x 0.55 nm. The passage of the guest molecule was realized through change of the pore shape [5]. Of the more than 150 types of zeolite each type differs in a particular property. In Table 1 the mnemonic code of the structure (topology) of frequently used zeolite types is given, together with the specific properties, which result(ed) in various applications. Table 1. Mnemonic code, specific properties of zeolites as well as (potential) applications. Code a Type Si/Al Pore configuration (nm) Channel system Application b LTA NaA D ion exchange, drying, separation GIS NaP x D ion exchange 0.28x0.48 FAU HY CaX D 3-D catalysis catalysis, separation MOR HMor 6 0.7x D catalysis FER HFer x D catalysis 0.35x0.48 MFI HZSM x D catalysis, separation 0.55x0.51 BEA HBeta x D catalysis 0.64x0.76 a [3], b [7] 3

14 Chapter 1 As zeolites are crystalline materials the pore size distribution is unique per zeolite type and the pore configuration clearly identified in the crystal form, see Figure 3. The topology of the framework as well as the pore configuration was well established for most of the zeolite types based on single crystal analysis applying x-ray diffraction techniques. In the case the zeolite contains Al ions in the framework the lattice is balanced by cations. Most favorable for catalysis applications is a proton, resulting in Brønsted catalytic sites [7]. Together with the size and the uniformity of the pores the zeolites unit two properties in one material; the molecular shape selectivity and the Brønsted acidity. Next to these attractive properties a third function, e.g. Pt or a coordination complex can be introduced in the zeolite pores. Table 2 illustrates the change in properties upon the influence of Al isomorphously substituted for silicon in the framework. (a) (b) Figure 3. (a) Single crystal of MFI type zeolite, dimensions 250x150x80 µm, and (b) the orientation of the pores in the crystal form. 4

15 General introduction and thesis outline Table 2. General properties of zeolites as a function of the Si/Al ratio. (The direction of the arrows indicates an increase in a certain property) Si/Al property Acid sites Acid strength Cations Hydrophobic Hydrophylic Thermal stab. As indicated in Table 2 the number of acid sites is at its maximum at a Si/Al ratio of about Acid strength increases from Si/Al 1.26 up to Si/Al ~ 10 and is further on more or less constant. Interesting for maximum ion exchange capacity is the lowest Si/Al ratio (1.0) while the material is simultaneously hydrophilic. Reducing the Al content provides gradually a more hydrophobic material. Thermal stability is controlled by the Si/Al ratio as well. From Si/Al ratio 1 up till infinity the thermal stability can range from 600 o C till 1100 o C. 3. Zeolite membranes In general membranes are continuous phases that selectively obstruct or retard a component in a mixture. Membranes composed of zeolitic materials are applied, up to ~60 m 2, in separation processes [8]. Zeolite membranes can be divided in two specific modes of performance, viz., the so-called absolute separation and the separation based on affinity with the zeolitic material. In the absolute separation mode one of the components of a mixture is too large in kinetic diameter to enter the pores of the zeolitic phase of the membrane. The absolute separation of a mixture containing two molecules with a difference in kinetic diameter can be done by choosing a zeolite with a pore aperture between the kinetic diameters of the two components in the mixture. The main element determining the zeolite pore aperture is the oxygen ring. The fine tuning of the pore diameter is depending on the type and number of cations involved. Figure 4 depicts an industrial example where zeolites are used to achieve absolute separation. Linear and iso-hexane are separated using zeolite CaA. It is clear, based on the dimensions, that the zeolite phase can provide absolute separation. 5

16 Chapter 1 Figure 4. n/i-hexane separation with zeolite membrane as an example of industrial application. In the column denoted with Dp the kinetic diameters of n- and i-hexane are given together with the pore aperture of zeolite A (actually CaNaA as not all the Na has been exchanged by Ca) Separation based on affinity allows actually both or more types of molecules in a mixture to permeate through the micropores of the membrane. Two modes of component permeation can be identified. One of the mode is surface diffusion in which, e.g. hydrocarbon molecules, interact with different energies with the pore wall through a number of C-H---O interactions. The other mode is the activated gaseous diffusion. In this mode the molecules follow the gas rules thus increasing distance between molecules while displaying lesser interaction with the pore wall. Surface diffusion dominates at low temperature whereas activated gaseous diffusion prevails at elevated temperature. At intermediate temperature both modes occur. The two modes are illustrated in Figure 5 for single component permeation of C 1 C 4 alkanes through a silicalite-1 membrane [9]. Figure 5. Permeances of methane (circles), ethane (squares), propane (triangles) and butane (diamonds) through silicalite-1 as a function of temperature [9]. 6

17 General introduction and thesis outline The separation of mixture components smaller than the pore size occurs at relatively low temperatures where the component permeation is governed by surface diffusion. Thus, the molecules diffuse with different fluxes through the membrane due to the different interactions with the pore wall. At elevated temperatures the activated gaseous diffusion is dominant resulting in a loss in separation selectivity. The flux of permeating components is optimal at a temperature where the diffusion is just still described by surface diffusion and not by activated gaseous diffusion. This is typical for linear hydrocarbons diffusion through zeolites. Almost absolute separation due to affinity with the pore wall can also occur as was demonstrated experimentally with n-butane/hydrogen [9], and olefin/hydrogen mixtures [10]. In both cases the pore aperture was larger than the kinetic diameters of the molecules. However, the interaction between the wall and the hydrocarbons excluded the transport of hydrogen through the pores. Only at elevated temperatures the transport was governed by activated gaseous diffusion thus allowing both molecules i.e. hydrocarbons and hydrogen to permeate. In general the transport through the separation layer is optimal facilitated when the micropores are aligned perpendicular to the support of the zeolite membrane. In contrast to the 3-D pore geometry, the 1-D and 2-D pore geometries should comprise oriented crystals in a perfectly closed layer on a macroporous support. The preparation of such a layer has been a matter of research in the last two decades. To achieve crystal orientation the crystals can either be forced by a natural cause like the aspect ratio of the crystal to precipitate from slurry on the support surface or grow in-situ from the support under particular conditions. In our laboratories one of the conditions studied was phase separation of nutrient and template to grow the crystals from. Another study performed was applying an extremely thin film of the ingredients needed to grow crystals from. A third attempt was extended growth of oriented seed crystals with a template that promoted growth perpendicular to the support surface. In all cases the crystals should grow into a continuous layer. Prerequisites are: 1) sufficient local supply of nutrient and 2) attractive support functions to initiate nucleation and growth along the support surface. A proven concept is the so-called grain growth. In the preparation of the membrane layer the nutrient concentration is chosen as high as possible. In that case the nucleation rate is high and crystal growth as well, both with a preference along the support and not into the solution. Figure 6 shows a 2-D geometry MFI type zeolite originated from a synthesis aiming at oriented crystal formation. A randomly oriented crystalline material of zeolite type NaA with 3-D geometry is illustrated in Figure 7. 7

18 Chapter 1 (a) (b) Figure 6. (a) cross sections at low (upper part) and high magnification (lower part) and (b) top view of grains of oriented crystalline material of MFI type zeolite in a membrane configuration. The 2-D channel configuration is oriented as follows: straight channels perpendicular to the support provide optimal passage. Sinusoidal channels run parallel to the support. (a) (b) Figure 7. (a) cross section and (b) top view of randomly oriented crystallites of zeolite NaA on alumina support. The 3-D geometry provides maximal transport through the membrane. 4. Zeolite membrane reactors Zeolite membrane as part of a membrane reactor should at least meet the requirements regarding selectivity and flux at the reactor temperature. In the case of absolute separation this can be expected, however, in affinity separations surface diffusion might be changed in activated gaseous diffusion as a function of temperature resulting in an unfavorable selectivity. Various concepts regarding zeolite membrane and catalyst phase in one unit are given below. Figure 8 illustrates the so-called parallel passage membrane reactor developed for 8

19 General introduction and thesis outline hydrogenation reaction studies [10]. The membrane reactor consists of a rutile support, an intermediate layer of Pt and a silicalite-1 top layer. The component of interest permeates through the silicalite-1 layer and reacts on the Pt layer. The advantage of using a silicalite-1 layer is that no Si is isomorphously substituted by Al, thus, no cations and/or Brønsted acid sites are present, neither on the extenal surface nor on the internal surface. Side reactions and pore blocking are therefore excluded. Furthermore, by coating the Pt layer with the zeolite no Pt is exposed to the open space and no side reactions will occur. Figure 8. Parallel passage membrane reactor developed for hydrogenation studies [10]. The reactor comprises a support layer of rutile, a porous Pt sputtered coating of 3 nm and a continuous silicalite-1 film of 0.5 µm. The concept of a micro membrane reactor consisting of a catalyst particle covered with a membrane layer [11] is depicted in Figure 9. The advantage of a system containing micro membrane reactors is that no continuous layer is needed and in the case of defects or leakage only one micro reactor fails. The membrane area per unit reactor volume is relatively large. There is simultaneously selectivity on the feed side as well as on the product side. In e.g. hydrogenation reaction it is mandatory that the temperature in the parallel passage as well as in the micro reactor concept must be at least 275 o C as otherwise hydrogen as a reactant is repelled from the catalyst sites by the preferred occupancy of the hydrocarbons. Figure 9. Micro membrane reactor comprising Pt on rutile coated with a silicalite layer [11], allowing selective hydrogenation of linear olefins in the presence of branched olefins. 9

20 Chapter 1 Figure 10. Tubular membrane reactor comprising a macroporous support coated with a zeolite layer and loaded with catalyst [12] for the hydroisomerization of mixtures of alkanes. The tubular membrane reactor concept illustrated in Figure 10 is actually comprising a porous tube of alumina or stainless steel coated with zeolite and containing a catalyst bed [12]. The unit performs at one temperature which can be chosen as such that the performance of the membrane and the catalyst is optimal. 5. Hydroisomerization and its position in fuel supplies The most important isomerization reaction within the context of energy-related catalysis is probably the skeletical isomerization of alkanes [13]. Hydroisomerization is the skeletical isomerization of alkanes in the presence of hydrogen and can be considered the basis of several important processes in the hydrocarbon processing industry. These include the skeletical rearrangement of light alkanes into their branched isomers. The hydroisomerization of mainly C 5 and C 6 hydrocarbons is becoming extremely important in improving the octane quality of the gasoline fraction, since in the course of the awareness of worldwide pollution through exhaust gases of motor cars the use of gasoline additives such as tetraethyllead, benzene and oxygenates, all providing high octane numbers, were reduced or abandoned. C 5 /C 6 hydroisomerization processes use bifunctional catalysts, i.e. Pt supported on H-Mordenite or Pt supported on chlorinated alumina. The reactor effluent consists of a mixture of branched and linear alkanes, which are separated in distillation columns or in adsorption towers packed with zeolite CaA. These units are quite energy intensive or discontinuous. Membranes could be a good alternative for these units. 10

21 General introduction and thesis outline 6. Outline of the thesis The thesis describes studies performed on the synthesis and application of zeolite membranes in the separation of linear and branched light alkanes. This separation is of great importance in hydroisomerisation processes. These processes are also described in the thesis. Different applications of zeolite membranes in these processes are suggested. In chapter 1, a general introduction is given. The importance of an alternative for energy intensive processes is stressed. Zeolitic materials are introduced and the application as membranes in separation and combined separation and reaction processes are described. Chapter 2 and 3 present and discuss experiments in which the synthesis and performances of zeolite membranes in the separation of light alkanes mixture are evaluated. In chapter 2 different silicalite-1 membranes were synthesized and used in the separation of monobranched from linear C 4, C 6 and C 7 alkanes. The use of zeolite beta membranes in separating dibranched from mono-branched hexane is described in chapter 3. In chapter 4 and 5 hydroisomerisation processes of light alkanes are presented. Chapter 4 describes a conceptual design of a state of the art C 5 /C 6 total hydroisomerization process. This design was used as a basis for comparison for a conceptual design of an industrial scale C 5 /C 6 hydroisomerization processes that uses membrane technology to combine reaction and separation in one process unit. In chapter 5 preliminary experimental studies are presented in which zeolite membranes are used to combine separation and reaction functions in one process unit as an alternative for the state of the art C 5 /C 6 hydroisomerization process. In Chapter 6 and 7 the hydroisomerization of heptane is discussed. Chapter 6 gives an overview of the mechanistic aspects, the current state and the potential application of heptane hydroisomerization. A conceptual design of an industrial heptane hydroisomerization process is presented in chapter 7. This process uses zeolite membrane technology for the separation of heptane isomer mixtures. Most of the chapters in this thesis are based on the author s publications. Therefore, each chapter can be read independently and overlaps between chapters can occur. References [1] A.B. Lovins, E.K. Datta, O-E. Bustnes, J.G. Koomey, N.J. Glasgow, Winning the oil endgame, Innovations for Profits, Jobs and Security, B.T. Aranow (ed.), Rocky Mountain Institute, Colorado, USA,

22 Chapter 1 [2] Separation Technology Innovation Roadmap, Report 17 September 2004, Dutch Ministry of Economic Affairs, DG Innovation, The Hague, the Netherlands. [3] H. van Bekkum, E, Flanigen, P.A.Jacobs and J.C.Jansen (eds.) Introduction to Zeolite Science and Practice, second edition, Elsevier, Amsterdam, [4] Atlas of Zeolite Structure Types, (assessed 15/12/2005) [5] H. van Koningsveld, F. Tuinstra, H. van Bekkum, J.C. Jansen, Acta Cryst. B45 (1989), [6] H. Robson, K.P. Lillerud, Verified syntheses of zeolitic materials, 2nd edition, Elsevier, Amsterdam, [7] R.A. Sheldon, H. van Bekkum (eds.), Fine Chemicals through Heterogeneous Catalysis, First edition, Wiley-VCH, Weinheim, [8] Y. Morigami, M. Kondo, J. Abe, H. Kita & K. Okamoto, The first large-scale pervaporation plant using tubular-type module with zeolite NaA membrane, Sep. Purif. Technol. 25 (2001) [9] F. Kapteijn, J.M. van de Graaf, J.A. Moulijn, J. Mol. Catal. A: Chemical 134 (1998) 201. [10] N. van der Puil, E.J. Creyghton, E.C. Rodenburg, S.T. Sie, H. van Bekkum, J.C. Jansen, J. Chem. Soc. Faraday Trans. 92 (1996) [11] N. Nishiyama, K. Ichioka, D-H. Park, Y. Egashira, K. Ueyama, L. Gora, W. Zhu, F. Kapteijn, J.A. Moulijn, Ind. Eng. Chem. Res. 43(5) (2004) [12] M.L. Maloncy, L. Gora, E.E. McLeary, J.C. Jansen, Th. Maschmeyer. Catal. Commun. 5 (6) (2004) 297. [13] S.T. Sie, in: G. Ertl, H. Knözinger, J. Weitkamp (eds.), Handbook of Heterogeneous Catalysis volume 4, VCH Verlagsgesellschaft mbh, Weinheim, 1997, p

23 2 The performance of silicalite-1 membranes in the separation of linear and branched alkanes Abstract Different silicalite-1 membranes were synthesised on seeded and nonseeded Trumem tubular supports. The nonseeded membranes were prepared by the in-situ hydrothermal synthesis using a two-step temperature crystallization procedure. The seeded membranes were prepared by the secondary growth method. Two types of seeded membranes were prepared; membranes prepared with seeds of about 700 nm and membranes with seeds of about 220 nm. The largest thickness of the continuous phase of the zeolite layer was obtained on the nonseeded membranes. The thickness was about 15 µm against 10 µm and 5 µm obtained on the 700 nm and 220 nm seeds membranes, respectively. The performance of the membranes were evaluated in permeation experiments using mixtures of 50/50 wt% n-butane/i-butane, 80/20 wt% n-hexane/2-methylpentane, and 80/20 wt% n-heptane/ 2-methylhexane. The experiments were performed at temperatures between K. Selectivities as high as 50 in favour of the linear alkane were obtained. The hydrocarbon fluxes range from 0.04 x 10-4 mol m -2 s -1 to 12 x 10-3 mol m -2 s -1. There was a decrease of hydrocarbon fluxes with increasing number of carbon atom. The selectivity decreases and the hydrocarbon flux increases when the temperature increases. The nonseeded membrane containing the thickest zeolite layer showed the best separation performance with the highest linear/branched alkane selectivities at reasonable hydrocarbon fluxes. This chapter is based on: M.L. Maloncy, L. Gora, J.A. Moulijn, J.C. Jansen, J. Membr. Sci., submitted

24 Chapter 2 1. Introduction Membranes composed of the MFI type zeolite (silicalite-1, ZSM-5) have been often investigated for catalytic and separation applications. Zeolites with the MFI structure have a high SiO 2 /Al 2 O 3 ratio inducing hydrophobic properties and relatively high thermal and chemical stability. The structural porosity of MFI zeolites consists of channels of about 0.55 nm, which makes this zeolite type interesting for application in size and shape selective chemical and physical processes [1,2]. The synthesis of a good quality zeolite membrane and its reproducibility is a rather difficult task, due to several factors involved. The work of Gora et al. [3] is one of the few publications on reproducibility. The membrane quality can be determined by intercrystalline porosity [4,5], crystal orientation [6], size of the crystal [7], thickness and uniformity of the zeolite layer, and the presence of defects [8]. A good quality membrane needs to fulfil two counteracting requirements; a thin layer, to achieve sufficiently high fluxes, and a defect free layer, to achieve high separation efficiencies [9]. A common method for the synthesis of zeolite membranes is the in-situ hydrothermal synthesis. In this method a support is put in direct contact with the synthesis solution or gel allowing the growth of a zeolite film on the surface of the support under hydrothermal conditions. A competitive crystallization on the support and in the reaction mixture occurs. Another method is the synthesis by secondary growth. Two steps are involved in this method. The surface of a support is first coated with zeolite seed crystals, followed by the growth of the seeds layer to a continuous zeolite film under hydrothermal synthesis conditions [10]. In this work we evaluate the separation performance of silicalite-1 membranes prepared by the in-situ hydrothermal synthesis method and by the secondary growth method. The separation of branched from linear C 4 -C 7 alkanes was studied. This separation is of great importance in the oil industry, due to the higher octane number of the branched hydrocarbons. The product of catalytic hydroisomerization of linear alkanes is a mixture of linear and branched hydrocarbons, and it s a necessity to separate the branched alkanes from the linear ones. Several studies showed that silicalite-1 (pore diameter around 0.55 nm) is a good candidate for this separation (kinetic diameter of the branched hydrocarbons is about nm, and that of the linear isomers is about 0.43 nm) [11-15]. 2. Experimental Trumem tubular supports (Trumem International, LLC) were used in this work. These supports are composed of porous stainless steel coated with TiO 2 [16]. The stainless steel layer has a thickness of 0.25 mm and a mean pore size of nm. The TiO 2 layer has 14

25 The performances of Silicalite-1 membrane a thickness of mm and a mean pore size of 160 nm. The supports (outer diameter 10 mm) were welded at the ends to nonporous metal tubes with a Swagelok connection (12 mm) on one side for mounting in the permeation testing equipment, and the other side of the tube was closed (dead end). The porous part of the support available for permeation had a length of 0.1 m, giving a total membrane surface area of about m 2. The support is shown in Figure 1. Figure 1. Tubular Trumem membrane support. The zeolite layer was synthesised on the Trumem supports according to two different synthesis approaches: an in-situ hydrothermal synthesis and a secondary growth synthesis. In the in-situ hydrothermal synthesis the membranes were prepared by two subsequent crystallizations under different conditions. This synthesis was already described elsewhere [17]. The condition for the first crystallization was 393 K for 114 h. The second crystallization condition was set at 453 K for 17 h. A synthesis mixture with molar composition 100 SiO 2 :59.3 TPABr:63.7 TPAOH:14200 H 2 O was used. From the synthesis mixture a zeolite layer was grown on the TiO 2 side of the support, which was placed vertically in a Teflon-lined autoclave. After the first synthesis the membrane was transferred into a new Teflon-lined autoclave with fresh synthesis solution. The reactants used for the synthesis mixture were: tetrapropylammonium hydroxide (TPAOH) (Chemische Fabriek Zaltbommel CFZ B.V., 25% in water), tetrapropylammonium bromide (TPABr) (CFZ B.V.), tetraethyl orthosilicate (TEOS) (Aldrich) and deionized water. TPAOH, TPABR and deionised water were mixed until TPABr was completely dissolved. TEOS was added and the solution was stirred at room temperature for 6 h. In the secondary growth synthesis two different types of membranes were synthesised; membrane synthesised using seeds of about 700 nm and membranes synthesised on seeds of about 220 nm. The seeds of 700 nm were prepared with a synthesis mixture similar to that reported by Jansen et al. [6], with molar composition SiO 2 :32.68 TPA: OH: H 2 O. The mixture was stirred at room temperature for 1.5 h. Thereafter, the solution was kept at 433 K for 3 h. The molar composition of the synthesis mixture used to prepare the 220 nm seeds was 25 SiO 2 :5 TPAOH:0.1 Na 2 O:1500 H 2 O:100 EtOH (Persson et al. [18]). The synthesis time was 22 h. 15

26 Chapter 2 Helium MFC Retentate P P Permeate Helium Hydrocarbon MFC Silicalite-1 Shell side Tube side Oven Figure 2. Schematic representation of the experimental set-up. (MFC) mass flow controller, (P) pressure transducer. Once the silicalite-1 seeds were formed after the required period they were separated from the mixture, washed repeatedly and put into suspension in deionised water. The seeds were applied on the Trumem supports by dipping the supports into the suspension. The supports were then dried at 373 K and used in the second step of the synthesis. The synthesis mixture used in the second step was similar to that used in the in-situ hydrothermal synthesis approach. The supports were placed vertically in Teflon-lined autoclaves and the synthesis mixture was added. The condition set for crystallization was 453 K for a period of 17 h. This condition was set for the support containing the 700 nm seeds as for the support containing the 220 nm seeds. Each end of the tubular supports, equipped with the Swagelok connection, was wrapped with Teflon tape and plugged with a Teflon tap, so that the zeolite layer could only grow on the TiO 2 side of the support. After crystallisation the membranes were washed with distilled water, dried overnight and calcined at 673 K for 16 h in air with heating and cooling rates of 1 K min -1. The surface and cross-section morphology of the resulting membranes was analysed by scanning electron microscopy (SEM) using a Philips XL20 microscope. Separation experiments were performed in a set-up presented previously [19], and schematized in Figure 2. The set-up comprised of the membrane in a hot air oven, as well as feed and carrier gas supply units. Three sets of experiments were carried out to evaluate the separation of the different linear/branched hydrocarbon mixtures. The first set of experiments was carried out with a 50/50 wt% n-butane/i-butane mixture at temperatures ranging from 303 K to 473 K. The mixture was supplied at a flow rate of 100 ml min -1. Helium with a flow rate of 100 ml min -1 was used as sweep gas. In the second set of experiments a 80/20 wt% n-hexane/2-methylpentane mixture was used as feed supplied at a flow rate of 0.08 ml min

27 The performances of Silicalite-1 membrane The experiments were performed at temperatures between 353 K and 473 K. The final set of experiments was carried out using a feed containing a 80/20 wt% n-heptane/2-methylhexane mixture with a feed flow rate of about 0.07 ml min -1. The separation was studied at temperatures ranging from 393 K to 473 K. Helium was used as carrier and sweep gas at a flow rate of 50 ml min -1 in the second and the last set of experiments. In all the experiments the pressure set on both sides of the membrane was 1 bar. Feed, retentate and permeate were analysed online with a gas chromatograph (FID detector for the hydrocarbons measurements). A GC-column, CP-Sil PONA CB fused silica, WCOT, 100 m x 0.25 mm, 500 nm df, was used to determine the amounts of hydrocarbons quantitatively. The separation performances were evaluated in terms of linear/branched alkane selectivity and the hydrocarbon flux through the membrane. The flux and the selectivity (S n/i ) were calculated as follows: F m Flux = A (1) X n X i S n/i = X n X i Permeate Feed (2) Where F m, A, X n and X i are the component flow through the membrane (mol s -1 ), the membrane area (m 2 ), the mole fraction of the linear alkane and the mole fraction of branched alkane, respectively. 3. Results and discussion In a typical zeolite membrane synthesis nucleation takes place in the bulk of the synthesis solution and on the support. By placing the support vertically in the synthesis solution the formation of the zeolite layer is predominantly due to nucleation on the support surface and the subsequent crystal growth from the nuclei towards a continuous layer. Formation of the zeolite layer by precipitation of crystals formed due to nucleation in the bulk of the solution, which occurs when the support is horizontally placed, is avoided. 17

28 Chapter 2 Silicalite-1 (a) Silicalite-1 (b) Silicalite-1 (c) Figure 3. SEM images of the nonseeded membrane (a) and the membranes prepared with seeds of 700 nm (b) and 220 nm (c). The last image is 90 degrees rotated compared to the first two images, moreover another magnification is used. The different synthesis procedures resulted in membranes in which the supports where fully covered with silicalite-1 crystals. The silicalite-1 layer was firmly bounded to the surface of the titania layer of the supports (see Figure 3). Figure 3a and 3b show clearly the different layers composing the membrane; the zeolite layer on the top, the stainless steel 18

29 The performances of Silicalite-1 membrane layer at the bottom and the titania layer in between. Figure 3c shows only the zeolite layer on top of the titania phase. Although other morphology characteristics can determine a membrane quality the focus on the membrane characterization will be on the zeolite layer thickness. The thickness can be easily deduced from SEM images. The thickness influences significantly the membrane separation performances. A thin membrane can provide sufficiently high fluxes. The thickness of the monolithic part (almost continuous phase) of the silicalite-1 layer indicated by the double arrows in Figure 3 varied from 5 to 15 µm depending on the synthesis procedures followed. The smallest thickness was found for the membranes prepared with the seeds of 220 nm. These membranes had a thickness of about 5 µm, as deduced from Figure 3c. The membranes synthesised with 700 nm seeds, see Figure 3b, had a thickness of around 10 µm while the membranes prepared on the nonseeded membranes had the largest thickness, of about 15 µm, see Figure 3a. Basically, each membrane was prepared by using two synthesis steps. The second synthesis step was similar for the three membranes. The preparation of the three membranes only differs in the first synthesis step. For the nonseeded membrane the first synthesis step consists of a crystallization process at low temperature and a longer crystallisation time aiming at nucleation. For the seeded membranes seeds of 220 nm and 700 nm were applied to the support from which they eventually should grow into a continuous layer in the second synthesis step. As the second synthesis step is similar the thickness of the continuous phase of zeolite layer is mainly influenced by the first synthesis step. The different membrane thicknesses obtained indicate that the use of a low temperature and longer crystallisation time leads to the formation of a thicker continuous layer while the use of seeds reduces the thickness. Moreover, smaller seeds further decrease the thickness. Probably, the different seeds grow with different rate at the same experimental conditions. The small seeds may not be fully crystallised. The growth of the small seeds may be a surface limited reaction. The first synthesis step in the preparation of the nonseeded membrane was performed at low temperature and aimed at the enhancement of nucleation and the coverage of the support with a dense population of small crystals. The subsequent in-situ crystallisation at elevated temperature in the second synthesis step envisaged the acceleration of the crystal growth and the formation of a closed layer. The temperature influences the nucleation of the crystals and the crystal growth. Increasing the temperature increases the crystal growth while the number of nuclei decreases. When the crystal growth and the nucleation processes are separated and performed at different conditions the synthesis is better controlled and can result in high quality membranes. Compared to the secondary growth approach used to prepare the seeded membranes the in-situ method with the two subsequent crystallization temperatures has the disadvantage to be time consuming. The first synthesis step performed 19

30 Chapter 2 at low crystallisation temperature was about 114 h. However, due to the longer crystallisation time and especially the lower temperature in the first synthesis step a denser zeolite layer with the largest thickness was formed. The separation performances of the membranes evaluated in the three sets of experiments are presented in Figures 4 and 5 in terms of the linear/branched alkane selectivities and fluxes of the various hydrocarbons as a function of temperature. In all cases there was an increase in hydrocarbon fluxes with temperature. The fluxes of the linear alkanes were always higher than those of the branched ones. The highest hydrocarbon fluxes were obtained for the membranes prepared with the 220 nm seeds. The branched isomer fluxes were the lowest on the nonseeded membranes. At temperatures below 410 K the lowest linear alkane fluxes were obtained with the membranes prepared with 700 nm seeds. Above this temperature the lowest linear alkane fluxes were observed on the nonseeded membranes. This behaviour could be an indication of a better quality of the nonseeded membrane having less defects in the zeolite layer compared to the 700 nm seeded membrane. In zeolitic pores the linear alkanes, preferentially adsorbed from the mixture, permeate by surface diffusion at low temperatures thereby hindering the branched hydrocarbon permeation. The fluxes of the linear alkanes are thus higher than those of the branched ones. As the temperature increases the hydrocarbon molecules vibrate more vigorously and have reduced interaction with the zeolite pore wall and consequently will diffuse faster through the pores yielding generally higher fluxes. The surface diffusion becomes less dominant while the activated diffusion increases. The adsorption of the linear hydrocarbons reduces and simultaneously an increase in the branched hydrocarbons permeation occurs. The presence of the branched alkanes in the zeolitic pores can obstruct the permeation of the linear hydrocarbon, resulting in a small decrease in the linear hydrocarbon permeation. The combined effects result in a maximum for the linear hydrocarbon flux that is typically for linear hydrocarbons in zeolitic pores. As shown in Figure 4 a maximum linear hydrocarbon flux is observed at a temperature around 393 K for the nonseeded membrane. As the temperature increases the fluxes of the linear molecules initially decrease slightly but subsequently increase with temperature. This behaviour is only observed on the nonseeded membrane and not on the 700 nm seeded membrane. This could be an indication of the presence of nonzeolitic pores (defects) on the 700 nm seeded membrane contributing to the continuous increase of the linear hydrocarbon fluxes. Typically, the linear hydrocarbon fluxes decrease with increasing number of carbon atoms. The fluxes of n-butane are about one order of magnitude higher than those of n-hexane and n-heptane, see Figure 4. The fluxes of n-hexane on their turn are higher than those of n-heptane. Similar behaviour is observed for the branched hydrocarbon fluxes. 20

31 The performances of Silicalite-1 membrane n-c 4 Flux (x 10-3 mol m -2 s -1 ) 12 (a) Temperature ( K ) i-c 4 Flux (x 10-3 mol m -2 s -1 ) 12 (b) Temperature ( K ) n-c 6 Flux (x 10-4 mol m -2 s -1 ) (c) 2MP Flux (x 10-4 mol m -2 s -1 ) (d) Temperature ( K ) Temperature ( K ) n-c 7 Flux (x 10-4 mol m -2 s -1 ) 12 (e) MHx Flux (x 10-4 mol m -2 s -1 ) 3 (f) Temperature ( K ) Temperature ( K ) Figure 4. Hydrocarbon fluxes through the different MFI membranes as a function of temperature. From the n/i-c 4 mixture the n-butane (a) and iso-butane (b) fluxes are shown. From the n/i-c 6 mixture the n-hexane (c) and 2-methylpentane (d) fluxes are given. From the n/i-c 7 mixture the n-heptane (e) and 2-methylhexane (f) fluxes are presented. Symbols: 220 nm seeds membrane (circles), 700 nm seeds membrane (triangles), nonseeded membrane (squares). Note that different flux and temperature scales are used. 21

32 Chapter 2 40 (a) n/i-c 4 Selectivity Temperature ( K ) n-c 6 /2MP Selectivity (b) Temperature ( K ) n-c 7 /2MHx Selectivity 10 (c) Temperature ( K ) Figure 5. Linear/branched alkanes selectivities at different temperatures: (a) n-butane/iso-butane, (b) n-hexane/2-methylpentane, (c) n-heptane/2-methylhexane. Symbols: 220 nm seeds membrane (circles), 700 nm seeds membrane (triangles), nonseeded membrane (squares). Note that different flux and temperature scales are used. The selectivity values obtained for the linear/branched hydrocarbon mixtures on the different membranes are presented in Figure 5 as a function of temperature. For all the membranes the selectivities decrease at elevated temperatures. This is because the selective adsorption of the linear alkanes becomes less dominant when the temperature increases 22

33 The performances of Silicalite-1 membrane resulting in a decrease in selectivity. The contributions by adsorption and diffusion govern the permeation selectivity of alkanes through silicalite-1 membranes. The diffusion differences between individual components dominate the selectivity at higher temperatures. The highest selectivities over the whole temperature ranged studied were observed for the nonseeded membranes. At the lowest temperatures studied using these membranes the selectivities towards the linear alkane for the n-butane/i-butane, n-hexane/2-methylpentane and n-heptane/2-methylhexane mixture were 37 (at 303 K), 48 (at 353 K) and 8 (at 393 K), respectively. Even at the highest temperature of 473 K the nonseeded membrane showed selectivity towards the linear component with selectivity values around 3, while the seeded membranes were not selective at all. The lowest selectivity values were obtained on the 220 nm seeded membrane. The higher linear/branched alkane selectivities and the reasonable alkane fluxes for the nonseeded membrane suggest a better quality membrane compared to the seeded ones. Among the seeded membranes the 700 nm seeded membrane shows better separation performance than the 220 nm seeded membrane. Relating the fluxes and the selectivity with the membrane thickness it seems that a decrease in the zeolite layer thickness results in an increase in the hydrocarbon fluxes through the membrane. This is mainly observed at the high temperatures studied. The selectivity increases with an increase in the zeolite layer thickness, which is observed mainly at low temperatures. Besides the zeolite layer thickness, a defect-free zeolite layer influences the membrane separation performances. High separation efficiency can be obtained on a membrane without defects. The lower separation performances of the seeded membranes, mainly the 220 nm seeded membrane could be due to the existence of defects in the zeolite layer. These defects may form during hydrothermal synthesis or calcinations. The preparation of the seeded membranes by the secondary growth method aimed at sealing the intercrystal voids. Most probably, the crystal growth was not sufficient to close the intercrystalline gaps, resulting in the presence of defects. During calcination defects can be formed due to template removal and the difference in thermal expansion between the support and the zeolite layer. The use of seeds reduces the thickness of the zeolite layer and this can have consequences for the resistance of the zeolite layer integrity to calcinations. The nonseeded membrane is the thickest membrane and can be considered more resistant to calcinations. The 220 nm seeded membrane is the thinnest membrane and has less resistance. The 700 nm seeded membrane has a resistance in between that of the nonseeded and the small seeded membrane. Consequently, the 220 nm membranes may present more defects compared to the other membranes resulting in the lowest separation performance as observed from the separation experiments. It is true that 23

34 Chapter 2 the thickness alone doesn t fully describe the resistance to calcinations, but it can be used as an indication. The best separation performance of the nonseeded membrane is most probably due to the use of two subsequent crystallisation temperatures which not only influence the membrane thickness but also tighten the intercrystalline boundaries, resulting in a dense membrane with less defects. The separation performance of the membranes can be compared with literature data. The selectivity values for the n-butane/i-butane mixture obtained are comparable with those found in literature on silicalite-1 membranes. Values in the range of can be found [20-22]. Literature data on n-hexane/2-methylpentane and n-heptane/2-methylhexane selectivity are scarce, making direct comparison rather difficult. However, our results compare well with those of Flanders et al. [11] who studied the separation of a n-hexane/3-methylpentane mixture and reported similar n-hexane/3-mp selectivity ranging between 50 and 75 at 373 K. Funke et al. [12] also observed similar selectivity behaviour. The selectivity decreases with increasing temperature from 24 (at 326 K) to 1 (at 443 K). The single component fluxes of n- hexane through MFI membranes reported in the literature are around 1 x 10-3 mol m -2 s -1 (at about 373 K), comparable to our results. 4. Conclusions Silicalite-1 membranes were prepared successfully on tubular supports by two different methods. In-situ hydrothermal synthesis with a two-step crystallisation procedure was applied to prepare nonseeded membranes, while seeded membranes with seeds of 700 nm and 220 nm in size were prepared by the secondary growth method. The size of the seeds influences the zeolite layer thickness. The thinnest zeolite layer was observed for the membranes synthesised with the 220 nm seeds. The hydrothermal synthesis approach with a two-step crystallisation procedure provides the membrane with the thickest zeolite layer. This membrane showed the best separation performance with the highest selectivities at reasonable fluxes. Thus, good quality membranes can be obtained by applying the in-situ hydrothermal synthesis with a two-step crystallisation procedure. In separating linear from branched alkanes using silicalite-1 membranes, the linear alkanes permeate faster than the branched ones. At higher temperatures the hydrocarbon fluxes increase while there is a decrease in the linear/branched alkane selectivity. The hydrocarbon fluxes decrease with increase of carbon number. 24

35 The performances of Silicalite-1 membrane References [1] L. Cot, A. Ayral, J. Durand, C. Guizard, N. Hovnanian, A. Julbe, A. Larbot, Solid state Sci. 2 (2000) 313. [2] J. Caro, M. Noack, P. Kölsch, R. Schäfer, Microporous Mesoporous Mater. 38 (2000) 3. [3] L. Gora, N. Nishiyama, J.C. Jansen, F. Kapteijn, V. Teplyakov, Th. Maschmeyer, Sep. Purif. Tech (2001) 223. [4] E.R Geus, H. van Bekkum, W.J.W. Bakker, J.A. Moulijn, Microporous Mesoporous Mater. 1 (1993) 131. [5] J.M. van de Graaf, E. van der Bijl, A. Stol, F. Kapteijn, J.A. Moulijn, Ind. Eng. Chem. Res. 37 (1998) [6] J.C. Jansen, W. Nugroho, H. van Bekkum, in: J.B. Higgens, R. von Ballmoos, M.M.J. Treacy (Eds.), Proceedings of the ninth International Zeolite Conference, Butterworth- Heinemann, Montreal, 1993, 247. [7] Z.A.E.P. Vroon, K. Keizer, M.J. Gilde, H. Verweij, A.J. Burggraaf, J. Membr. Sci. 113 (1996) 293. [8] Z.A.E.P. Vroon, K. Keizer, A.J. Burggraaf, H. Verweij, J. Membr. Sci. 144 (1998) 65 [9] F. Kapteijn, J. van de Graaf, J.A. Moulijn, J. Mol. Catal. A: Chemical 134 (1998) 201. [10] Y.S. Lin, I. Kumakiri, B.N. Nair, H. Alsyouri, Sep. Purif. Methods 31 (2) (2002) 229. [11] C.L. Flanders, V.A. Tuan, R.D. Noble, J.L. Falconer, J. Membr. Sci. 176 (2000) 43. [12] H.H. Funke, A.M. Argo, J.L. Falconer, R.D. Noble, Ind. Eng. Chem. Res. 36 (1997) 137. [13] C.J. Gump, R.D. Noble, J.L. Falconer. Ind. Eng. Chem. Res. 38 (1999) [14] J. Coronas, R.D. Noble, J.L. Falconer. Ind. Eng. Chem. Res. 37 (1998) 166. [15] T. Matsufuji, K. Watanabe, N. Nishiyama, Y. Egashira, M. Matsukata and K. Ueyama. Ind. Eng. Chem. Res. 39 (2000) [16] L. Trusov, Membr. Technol. 128 (2000) 10 [17] L. Gora, J.C. Jansen, J. of Catal. 230 (2005) [18] A.E. Persson, B.J. Schoeman, J. Sterte, J.E. Otterstedt, Zeolites 14 (1994) 557. [19] M.L. Maloncy, L. Gora, E.E. McLeary, J.C. Jansen, Th. Maschmeyer. Catal. Commun. 5 (6) (2004) 297. [20] S. Alfaro, M. Arruebo, J. Coronas, M. Menendez, J. Santamaria, Microporous Mesoporous Mater. 50 (2001) 195. [21] K. Keizer, A.J. Burggraaf, Z.A.E.P. Vroon, H. Verweij, J. Membr. Sci. 147 (1998) 159. [22] J. Hedlund, J. Sterte, M. Anthonis, A.-J. Bons, B. Carstensen, N. Corcoran, D. Cox, H. Deckman, W. de Gijnst, P.-P. de Moor, F. Lai, J. McHenry, W. Mortier, J. Reinoso, J. Peters, Microporous Mesoporous Mater. 52 (2002)

36 Chapter 2 26

37 3 Preparation of zeolite beta membranes and their pervaporation performance in separating di- from monobranched alkanes Abstract Two synthesis methods were used to prepare zeolite beta membranes on betaseeded, commercial Trumem disk supports. The first synthesis method produced an alkali free zeolite beta layer with a Si/Al ratio of 40. A Na-containing zeolite beta layer with a Si/Al ratio of 13 was obtained from the second synthesis method. The thickness of the zeolite beta layers was around 1 µm. XRD analysis confirmed the existence of the zeolite beta phase. Nitrogen permeation experiments indicated permeation governed by surface diffusion. Thus, the membranes prepared had hardly any defects. The performance of the membranes in separating di- from monobranched alkanes was evaluated in pervaporation experiments at 303 K with a 50/50 wt% 2-methylpentane (2MP) and 2,2-dimethylbutane (22DMB) mixture. Only the membranes prepared by the first method were selective. A selectivity of 1.5 in favour of 2MP was obtained which is higher than the Knudsen selectivity. Adsorption energies estimated by simulation of 2MP and 22DMB in zeolite beta were -48 kj/mol and -42 kj/mol, respectively. The higher adsorption value of 2MP and the selectivity towards 2MP suggest preferential adsorption of the mono over the dibranched isomer in the zeolite beta membranes. Membrane post-treatment with trimethylchlorosilane seems to improve membrane performance. A selectivity of 1.8 was obtained on a post-treated membrane. This chapter is based on: M.L. Maloncy, A.W.C. van den Berg, L. Gora, J.C. Jansen, Microporous Mesoporous Mater. 85(1-2) (2005) 96

38 Chapter 3 1. Introduction Zeolite beta is relevant to petrochemical industrial processes such as aromatic alkylation and the hydroisomerization of alkanes. Pt-loaded beta zeolite catalyst shows very high activity and selectivity in the hydroisomerization of C 5 -C 7 alkanes [1,2]. Adsorption experiments using different zeolites [3,4] suggest a relatively good performance of zeolite beta in separating branched C 5 -C 8 isomers. Denayer et al. [3] studied ZSM-5, ZSM-22, mordenite, NaY, NaUSY and beta type zeolites. The authors suggest that zeolite beta and ZSM-22 can be used to separate C 5 -C 8 alkane isomers based on the degree of branching. Huddersman and Klimzcyk [4] studied the separation of 3-methylpentane and 2,3- dimethylbutane over zeolite beta, mordenite, silicalite, EU-1, ZSM-12 and SAPO-5. The beta type zeolite proved to be the most effective separator with stronger adsorption capacities for the monobranched compared to the dibranched isomer. These findings encourage studies on the preparation of zeolite beta membranes for application in hydroisomerization processes. In these processes the membranes could be potentially applied as a separator for branched alkane isomers or even as a catalytic membrane reactor, combining separation and reaction functions. Literature data on zeolite beta membranes are very limited. Furthermore, to our knowledge there is no report on branched alkane mixture separation using this type of membrane. It is difficult to use zeolite beta membranes to separate mono- from dibranched isomers mainly because these molecules have similar diameters ( nm) and they fit into the zeolite beta pores (0.75 nm). Therefore, it is not possible to achieve separation due to absolute molecular sieving. Preferential adsorption of one of the molecules could bring about the separation of monofrom dibranched isomers. In this work, we report the synthesis of zeolite beta membranes on stainless steel disk supports and their performance in separating mono- from dibranched hexane by pervaporation. In literature it is shown that for a given zeolite membrane, fluxes and selectivities for pervaporation and vapour permeation are the same if the feed fugacities are the same [5]. However, vapour permeation is usually not carried out with the feed at the saturated vapour pressure. In general, pervaporation fluxes are higher at the same temperature because of higher feed side coverage [6]. The influence of post-treatment with trimethylchlorosilane on the membrane performance is also reported in this work. The silane reagent can be very effective for the modification of the surface of inorganic materials, altering their adsorption and catalytic properties [7-9]. To support the hypothesis of preferential adsorption of one of the isomers, preliminary simulations were performed to estimate the adsorption energies of the hexane isomers in zeolite beta. 28

39 Preparation of zeolite beta membranes 2. Experimental In this work commercial Trumem disks with a diameter of 25 mm were used as supports. These disks are composed of porous stainless steel coated with TiO 2 [10]. The stainless steel layer has a thickness of 0.25 mm and a mean pore size of 2-5 µm. The TiO 2 layer has a thickness of mm and a mean pore size of 160 nm. Some of the supports were first seeded with beta particles. The size of the seeds was about nm. The seeds were grown from a reaction mixture identical to that of method 2 (vide infra). The support was put into a flat holder with one open end and the other end attached to a vacuum pump. The seeds were placed on the titania side of the support by applying vacuum at the stainless steel side of the holder and by dipping the holder vertically in the zeolite beta seeds suspension. The seeds were than dried in air at 353 K overnight. The membranes were prepared according to two different methods. In the first method a synthesis mixture with molar composition 1.00 SiO 2 : Al 2 O 3 :0.269 (TEA) 2 O:15.5 H 2 O was used followed by crystallisation at 423 K for 96 h. The reactants used for the synthesis were: tetraethylammonium hydroxide (TEAOH) (35 wt%), silica (Degussa Aerosol 200), aluminium nitrate (Al(NO 3 ) 3.9H 2 O) and deionised water. The membranes from the second method were prepared from a mixture with composition 1.00 SiO 2 :0.02 Al 2 O 3 :0.25 (TEA) 2 O:15 H 2 O: Na 2 O:0.02 K 2 O:0.058 HCl followed by crystallisation at 423 K for 40 h. The reactants used for the synthesis were: tetraethylammonium hydroxide (TEAOH) (35 wt%), sodium chloride, potassium chloride, sodium hydroxide, silica (Degussa Aerosol 200), sodium aluminate (56 wt% Al 2 O3, 37 wt% Na 2 O) and deionised water. The preparation of the synthesis mixtures was done by following the procedure of zeolite [Ti,Al]-beta (without use of the Ti-containing reagent) and zeolite Al-beta, both given in [11] for method one and two, respectively. Crystallisation in both methods was done with the support vertically placed inside a Teflon-lined autoclave. A Teflon holder protected the stainless steel side of the support so that the membrane could only grow on the TiO 2 side. After crystallisation the membranes were washed with distilled water, dried overnight and calcined at 673 K for 4 h in N 2 followed by 16 h in air with heating and cooling rates of 0.5 K/min. The initial absence of O 2 and the low heating rates prevent crack formation caused by exothermal burn-off of the template and difference in thermal expansions. On the resulting membranes a second layer was prepared following the same procedure as that of the synthesis of the first layer. Thus, on each support two zeolite layers were synthesised. The zeolite beta membranes prepared are given in Table 1. In order to compare the permeation behaviour of the beta membranes with a smaller pore zeolite membrane, a silicalite-1 membrane (pore size, 0.55 nm) prepared according to a seeded surface synthesis approach (700 nm size seeds) [12] was used in the permeation experiments. 29

40 Chapter 3 Table 1. Beta membranes synthesised. Membrane Beta seeded Method No. of layers BEA1 yes 1 2 BEA2 yes 2 2 BEA3 no 1 2 BEA4 no 2 2 Zeolite beta particles formed after crystallisation were collected from the Teflon bottom, washed, dried and element analysis (Inductively Coupled Plasma Optical Emission Spectroscopy, ICP-OES) was performed using a Perkin Elmer Optima 3000DV. The identification of the existing phases on the collected beta particles and on the membranes was done by X-ray diffraction analysis (XRD) with CoKα radiation (λ = nm). The surface and cross-section morphology of the membranes was analysed by scanning electron microscopy (SEM) using a Philips XL20 microscope. To check the permeation performance and the existence of defects, N 2 permeation experiments according to the batch method [13] were performed on the calcined membranes. In this method, pure N 2 was introduced to the feed side (the zeolite side of the membrane) up to a pressure of 200 kpa. The feed stream was shut down and vacuum was applied at the permeate side. The decrease of pressure on the feed side was monitored with a pressure transducer to determine the flux trough the membrane. The experiments were performed at ambient temperature. As the volume of the compartment at the feed side is known, ideal gas law can estimate the amounts of moles present at a measured pressure. The N 2 permeance over a certain time interval, t (s) was determined as follows: n N 2 permeance = t A P (1) where n, P and A are the amount of N 2 permeated (mol), the pressure difference across the membrane (Pa) and the membrane area exposed to permeation (m 2 ), respectively. The pervaporation experiments were performed in the set-up schematised in Figure 1. The membrane was placed in the membrane holder with the zeolite side facing the sealing ring and was closed tight with the closing piece facing the stainless steel side of the membrane (see Figure 2). In this manner the system was sealed leaving the membrane as the only route for components to permeate through. The diameter of the exposed permeable area was 15 mm. The holder was then placed in the pervaporation apparatus. 30

41 Preparation of zeolite beta membranes Feed solution Stop valves Vent Vacuum Stir bar Membrane in holder PI Pressure indicator Cold traps Figure 1. Schematic view of the pervaporation set-up. Membrane Sealing ring Closing piece (facing stainless steel side of membrane) Membrane holder Figure 2. Membrane holder used in the pervaporation set-up. Pervaporation experiments were performed at 303 K with a 50/50 wt% 2-methylpentane (2MP) / 2,2-dimethylbutane (22DMB) mixture. The mixture was poured on the feed side. The permeate side was kept under vacuum. The permeate was continuously collected in a cold trap cooled with liquid nitrogen and analysed using a gas chromatograph (FID detector for the hydrocarbons measurements). A GC-column, CP-Sil PONA CB fused silica, WCOT, df = 500 nm, 100 m x 0.25 mm was used to determine the amounts of hydrocarbons quantitatively. 10 g of n-heptane was used as solvent in the traps, because of the small amount of components that permeated. In this manner it was easier to collect the mixture and to inject it for composition analysis into the GC. Pure n-hexane was used as internal standard. The amount of permeate can be determined by comparing the peak area for the permeant with that of n-hexane. The pervaporation experiments were performed 31

42 Chapter 3 during a period of 6 h. Before each experiment, the membranes were calcined overnight at 573 K in air with heating and cooling rates of 0.5 K/min to remove adsorbed species. The flux and the selectivity (S 2MP/22DMB ) were calculated from the following equations. n Flux = t A (2) X X S 2MP/22DMB = X X 2MP 22DMB 2MP 22DMB Permeate Feed (3) Where n, t, A, X 2MP and X 22DMB are the permeate amount (mol), the permeation time (s), the membrane area (m 2 ) the mole fraction of 2MP and the mole fraction of 22DMB, respectively. Modification by post-treatment with trimethylchlorosilane was done on the non-seeded beta membrane prepared by method 1 (BEA3). The membrane was placed into 5 ml trimethylchlorosilane and kept there at room temperature for about 30 min. The membrane was then dried and calcined at 673 K for 4 h in N 2 followed by 16 h in air with heating and cooling rates of 0.5 K/min. This membrane was used also in the pervaporation experiments. The adsorption energies of the 2MP and 22DMB in a beta pore channel were estimated by means of interatomic-potential-based calculations. Structural optimisations of both the framework and the adsorbed molecules were performed using the computer code GULP (General Utility Lattice Program) [14] with the RFO [15] optimisation algorithm at a constant pressure of 1 bar. As a starting structure for zeolite BEA, the crystallographic data were taken from the database of the IZA website [16] and sized as a 2x2x1 unit cell (Si 256 O 512 ). This cell was modelled as a periodic box with no symmetry constrains. Since the interaction between zeolite beta and the alkanes is strongly dominated by the framework oxygen atoms (Kiselev model [17]), for simplicity, all other framework atoms were chosen to be silicon. The interactions between the framework atoms were modelled according to the force field of Sanders [18], which is specially designed for reproducing the structure, mechanics and vibrations of zeolitic materials [19-21]. The starting geometries for 2MP and 22DMB were obtained by optimising these structures in the computer code Hyperchem [22] with the AMBER fore field [23]. In GULP, 2MP and 22DMB were modelled using the force fields given in references [24] and [25], respectively. The non-bonding interaction between 32

43 Preparation of zeolite beta membranes the alkanes and the zeolite beta O-atoms is described by a set of Lennard Jones equations [26]. 3. Result and discussion Figure 3 shows the XRD spectra of the membranes BEA1 and BEA2, confirming the existence of the zeolite beta phase on the supports. Peaks from the support (rutile, TiO 2 and iron austenite) were detected as well. Membrane BEA1 had higher beta zeolite peak intensities compared to BEA2, suggesting higher crystallinity. Among the different variables distinguishing the two synthesis methods, the longer crystallization time of method one is most probably accounting for the higher crystallinity of BEA1. XRD analysis of the particles collected from the Teflon bottom also shows higher peak intensities of the zeolite beta prepared by the first method, as well as the existence of zeolite beta as the only zeolite phase. support BEA1 support support support BEA2 support support Θ Figure 3. XRD pattern using CoKα radiation (λ = nm) of the zeolite membranes BEA1 and BEA2. Table 2. Elemental analysis of the zeolite beta particles prepared by the two methods. Si (%) Al (%) Na (%) Si/Al (mol/mol) Method Method

44 Chapter 3 The results of the elemental analysis performed on the collected zeolite beta particles from the Teflon bottom is given in Table 2. The obtained Si/Al ratios are in agreement with the ratios given in [11]. SEM images of the cross sections of the different membranes are shown in Figure 4. The images clearly show the zeolite layers on top of the titania layer of the support. The zeolite and the titania layers are well-intergrown. The membrane surfaces also show wellintergrown zeolite beta crystals. SEM images from the membrane cross sections suggest a zeolite layer thickness around 1 µm for all the membranes. In principle, the use of seeds on a support gives rise to a dense layer when the system is exposed to further synthesis steps. The use of a seeded mixture was found essential for a dense coating [12]. Apparently, the use of seeds in this work does not influence significantly the membrane thickness as deduced from the SEM images. BEA1 BEA2 BEA3 BEA4 Figure 4. SEM images of the different zeolite beta membranes (cross sections). Results from the nitrogen permeation experiments are depicted in Figure 5. The permeance of N 2 was plotted against the pressure difference over the membrane. For all membranes the permeance decreases with increasing pressure difference across the membrane. This behaviour indicates that the permeation of N 2 is mainly governed by surface diffusion, that the permeation is basically through zeolitic pores and that there is an absence of large defects (non-zeolitic pores). 34

45 Preparation of zeolite beta membranes 10 N 2 Permeance (x 10-7 mol m -2 s -1 Pa -1 ) viscous Knudsen / activ. gaseous surface BEA3 BEA4 Silicalite-1 BEA P (x 10 5 Pa ) BEA1 Figure 5. N 2 permeance through the different membranes according to the batch method [13]. (Note: The x-axis of the inset should be based on the feed pressure rather than the pressure difference in the case of viscous flow) The permeance of a component through a membrane could take place according to various transport modes, such as viscous flow, Knudsen diffusion, activated gaseous diffusion and surface diffusion. Nishiyama et al. [13] discuss these transport modes in zeolite membranes. For single component systems viscous flow and Knudsen diffusion occurs in nonzeolitic pores. Viscous flow occurs in the presence of an absolute pressure gradient. The permeance from viscous flow increases with increasing feed pressure. The permeances from Knudsen diffusion and activated gaseous diffusion are independent from the pressure difference across the membrane. The permeance from surface diffusion decreases with increasing pressure difference across the membrane. Surface diffusion and activated gaseous diffusion occur in zeolitic pores. The determined permeance values vary from 0.4 x 10-7 mol m -2 s -1 Pa -1 for BEA1 to 7.7 x 10-7 mol m -2 s -1 Pa -1 for BEA4, see Figure 5. The permeance values are in the same order of magnitude as those reported by Tuan et al. for Al-beta zeolite [27] and boron substituted beta zeolite membranes [28]. To obtain the low N 2 permeation Tuan et al. synthesized Al-beta en B-beta membranes that required five and three synthesis layer of zeolite beta, respectively. Six days were needed for the synthesis of one layer. In the present work only two synthesis layers were required to obtain the low N 2 permeation, because the use of seeds and the titania-coated support also favours the low N 2 permeation. 35

46 Chapter 3 The influence of the different synthesis methods on the N 2 permeation can also be deduced from Figure 5. The membranes synthesized by method 2 have higher permeance values than those from method 1. This could be due to the higher crystallinity, which resulted from method 1 as previously mentioned in the XRD analysis. It is possible that BEA1 and BEA3 have a better intergrown membrane surface of zeolite beta particles than the membranes from method 2. The use of seeds does not influence significantly the N 2 permeance of the membranes synthesized by method 2. The influence of the use of seeds is more pronounced for the membranes synthesized by method 1. The seeded BEA1 had permeances of roughly one order of magnitude lower than the unseeded BEA3. Although the use of seeds didn t seem to influence the membrane thickness it surely affects the N 2 permeation through the membranes synthesized by method 1. The seeds could account for a dense and compact layer of zeolite beta increasing the resistance to N 2 permeation. The BEA1 membrane had permeance values even lower than those of the seeded silicalite-1 membrane, as can be seen in Figure 5. This gives an indication of the good quality of the membrane synthesized, because silicalite-1 membranes have zeolite pores of around 0.55 nm, which are smaller than the zeolite beta pores (0.75 nm). The performance of the silicalite-1 membrane on the permeation of N 2 (decreasing slope) also suggests the absence of large defects. Overall it seems that synthesizing the beta membranes by method 1 and using seeds gives good quality membranes with hardly any defects. The membranes BEA1 and BEA2 were used in pervaporation experiments at 303 K to separate 2MP and 22DMB. The results of the pervaporation experiments are given in Table 3. The membrane prepared by method one, BEA1, showed a better performance in separating 2MP from 22DMB with higher permeation flux of 2MP compared to 22DMB. BEA1 also showed a better performance than the silicalite-1 membrane. Comparison of the BEA1 membrane performance with that of a membrane reported in the scarce literature on branched alkane separation [29] shows a better separation performance of the BEA1 membrane. Funke et al. [29] used a silicalite membrane but were not able to separate a 50/50 wt% 3-methylpentane/22DMB mixture in vapour permeation experiments at 363 K. They obtained selectivity equal to unity. Table 3. Pervaporation experiments at 303 K with a 50/50 wt% 2MP/22DMP mixture. Membrane Fluxes (mol s -1 m -2 ) S 2MP/22DMB 2MP 22DMB BEA x x BEA x x BEA3 post-treated 9.34 x x Silicalite x x

47 Preparation of zeolite beta membranes In principle separation by molecular sieving could be achieved by using a defect-free silicalite-1 membrane, since one of the isomers (the monobranched hexane) has a diameter similar to the pore diameter of silicalite-1 and the other one (the dibranched hexane) a larger diameter. However, this possibility is not straightforward since studies showed that larger molecules could penetrate into zeolite MFI pores [30]. Moreover, one should be very cautious with the values of the isomers diameter, since in literature accurate estimation is lacking. As given in Table 3, both molecules 2MP and 22DMP, permeate with similar fluxes through the silicalite-1 membrane. This suggests that separating 2MP from 22DMB by molecular sieving using silicalite-1 is rather difficult. The fact that the larger pore zeolite BEA1 membrane could separate these isomers suggests that this separation is most probably due to preferential adsorption of the monobranched isomer on zeolite beta. Simulation studies discussed further on showed higher adsorption energy for the monobranched isomer. Another cause for the higher flux of 2MP could be that during its permeation through the zeolite pore it suffers less sterical hindrance than the more bulky 22DMB. To evaluate the influence of post-treatment with a silane coupling reagent, BEA3 was post-treated with trimethylchlorosilane and used in the pervaporation experiments. The result is presented in Table 3. This membrane showed the best separation performance. An explanation for the better performance could be that the silane post-treatment changes the surface in a manner that increases the adsorption of the 2MP. This is in accordance with the higher fluxes of 2MP. During the post-treatment the zeolite external surface could be covered with the products of reactions between the silane reagent and the surface hydroxyl groups. Calcination with O 2 removes the hydrocarbon residue and produces silica coated zeolites [7]. The external surface may become more apolar, thus increasing the interaction between the surface and the apolar 2MP isomer. The pore interior remains unchanged because trimethylchlorosilane is too large to penetrate into the pore. The post-treatment with possible formation of a silica coated surface could also cause a reduction in the zeolite pore openings. In combination with the increased interaction, the reduction in pore size could contribute to the larger difference in fluxes, thus increasing the selectivity. For a complete understanding of the surface modification and the enhanced pervaporation performance further investigation is needed. The selectivies obtained in this work using BEA1 and post-treated BEA3 were higher than the unity value of the Knudsen selectivity. The values are moderate compared to selectivity values found in literature for linear/monobranched or linear/dibranched isomer mixture separation using MFI type membranes. Nevertheless, the results in this work can be regarded as an important progress in the separation of di- from monobranched alkanes using zeolite membranes. 37

48 Chapter 3 To gain more insight in the adsorption of the two isomers in zeolite beta molecular simulations were performed to estimate adsorption energies for 2MP and 22DMB in zeolite beta pores. Figure 6 gives a schematic representation of the adsorbed 2MP and 22DMB in the zeolite beta pore obtained by performing a full optimisation of this system (see experimental section). From the simulations adsorption energies of -48 and -42 kj/mol were predicted for 2MP and 22DMB, respectively. This supports the assumption that the monobranched molecule is stronger and preferentially adsorbed in the zeolite beta pore, which could explain the selectivity towards the monobranched isomer. The larger interaction area most likely causes the preference for 2MP. 22DMB has a quaternary C-atom, which is largely shielded from the pore wall by its surrounding C-atoms, while all C-atoms of 2MP are exposed to the framework O-atoms (see also Figure 6). Furthermore, 2MP is more flexible so it can easier adapt to the shape of the pore. Other researchers have also observed the difference in adsorption energy between monobranched and dibranched hexane experimentally. Huddersman and Klimczyk [31] found a difference between 3-methylpentane and 2,3-dimethylbutane of -5.0 kj/mol, comparable to the calculated value of -6.0 kj/mol of the present work. Denayer et al [3] found a difference of about -7 kj/mol between 2MP and 22DMB. It should be stressed that the energy values of the present work should be seen as a qualitative indication, because they are based on optimisation calculations. Monte Carlo simulations are required in order to obtain more accurate quantitative data. 4. Concluding remarks Two synthesis methods were used to prepare successfully zeolite beta membranes with a zeolite layer thickness of around 1 µm. Membranes with zeolite top layers having Si/Al ratios of 40 and 13 were obtained for the first and second method, respectively. N 2 permeation experiments using the batch method suggest the absence of large defects in the membranes surfaces. The use of beta seeds influences the N 2 permeation performance of the membranes prepared by method 1 more significantly than that of the membranes prepared by method 2. The membranes prepared by method 1 showed a lower N 2 permeation and a better performance in the separation of a 50/50 wt% 2MP/22DMB mixture. Membrane posttreatment with trimethylchlorosilane improves membrane performance. The calculated adsorption energies indicate that the separation of branched hexane isomers on beta membranes is due to the preferential adsorption of the monobranched hexane. As multibranched isomers have higher octane values than monobranched ones the application of beta membrane has potential in the field of gasoline production and can be promising mainly when integrated reaction/separation systems are envisaged. 38

49 Preparation of zeolite beta membranes Figure 6. Schematic representation of an optimised all silica zeolite BEA structure with 2MP (above) and 22DMB (below) inside one of the pores (black sticks are oxygen, grey sticks are silicon). The CH x -groups (x = 0-3) of the alkanes are given as spheres in accordance with the united atom model as employed in the calculations. Acknowledgements Philippe Ungerer and Emeric Bourasseau of the Université de Paris Sud are acknowledged for supplying the parameters for the Lennard Jones interaction between quaternary C-atoms and zeolites. 39

50 Chapter 3 References [1] T. Yashima, Z.B. Wang, A. Kamo, T. Yoneda, T. Komatsu, Catal. Today, 29 (1996) 279. [2] Z.B. Wang, A. Kamo, T. Yoneda, T. Komatsu, T. Yashima, Appl. Catal. A, 159 (1997) 119. [3] J.F. Denayer, W. Souverijns, P.A. Jacobs, J.A. Martens, G.V. Baron, J. Phys. Chem. B, 102 (1998) [4] K. Huddersman, M. Klimczyk, AIChE Journal, 42 (1996) 405. [5] M. Namuro, T. Yamaguchi, S. Nakao, J. Membr. Sci. 144 (1998) 161. [6] T.C. Bowen, R.D. Noble, J.L. Falconer, J. Membr. Sci. 245 (2004) 1. [7] E.F. Vasant, P. Cool, Colloids Surf. A, 179 (2001) 145. [8] N.R.E.N Impens, P.van der Voort, E.F. Vansant, Microporous Mesoporous Mater. 28 (1999) 217. [9] T. Sano, M. Hasegawa, S. Ejiri, Y. Kawakami, H. Yanagishita, Microporous Mater. 5 (1995) 179. [10] L. Trusov, Membr. Technol. 128 (2000) 10. [11] H. Robson, K.P. Lillerud, Verified syntheses of zeolitic materials, 2nd edition, Elsevier, Amsterdam, 2001, pp [12] L. Gora, G. Clet, J.C. Jansen, Th. Maschmeyer, in: A. Galarneau, F. Di Renzo, F. Fajula, J. Védrine (Eds), Zeolites and Mesoporous Materials at the Dawn of the 21st Century, Studies in Surface Science and Catalysis, Vol. 135, Elsevier Science, Amsterdam, 2001, p [13] N. Nishiyama, L. Gora, V. Teplyakov, F. Kapteijn, J.A. Moulijn, Sep. Purif. Technol (2001) 295. [14] J.D. Gale, J. Chem. Soc. Faraday Trans. 93 (1997) 629. [15] A. Banerjee, N. Adams, J. Simons, R. Shepard, J. Phys. Chem. 89 (1985) 52. [16] International Zeolite Association website (assessed June 2004): [17] A.H. Fuchs, A.K. Cheetam, J. Phys. Chem. B, 105 (2001) [18] M.J. Sanders, M. Leslie, C.R. A. Catlow, J. Chem. Soc., Chem. Commun. (1984) [19] N. J. Henson, A. K. Cheetham, J.D. Gale, J. Chem. Mater. 6 (1994) [20] G.D. Price, I.G. Wood, D.E. Akporiaye, in: C.R.A. Catlow, C. R. A. (Ed.), Modelling of Structure and Reactivity in Zeolites, 1st edition, Academic Press Inc., San Diego, 1992, p. 38. [21] D. W. Lewis, G. Sastre, Chem. Commun. (1999) 349. [22] Hyperchem, version 7.0, Hypercube Inc., Canada,

51 Preparation of zeolite beta membranes [23] S.J. Weiner, P.A. Kollman, D.A. Case, U.C. Singh, C. Ghio, G. Alagona, S. Profeta, P. Weiner, J. Am. Chem. Soc. 106 (1984) 765. [24] T. J. H. Vlugt, R. Krishna, B. Smit, J. Phys. Chem. B, 103 (1999) [25] S.K. Nath, R. Khare, J. Chem. Phys. 115 (2001) [26] Ph. Ungerer, B. Tavitian, A. Boutin, Applications of Monte Carlo simulation in oil and gas technology, Technip, Paris, [27] V.A. Tuan, S. Li, J.L. Falconer, R.D. Noble, Chem. Mater. 14 (2002) 489. [28] V.A. Tuan, L.L. Weber, J.L. Falconer, R.D. Noble, Ind. Eng. Chem. Res. 42 (2003) [29] H.H. Funke, A.M. Argo, J.L. Falconer, R.D. Noble, Ind. Eng. Chem. Res. 36 (1997) 137. [30] H. van Koningsveld, J.C. Jansen, Microporous Mater. 6 (1996) 159. [31] K. Huddersman, M. Klimczyk, J. Chem. Soc. Faraday Trans. 92 (1996)

52 Chapter 3 42

53 4 Design of a state of the art C 5 /C 6 hydroisomerization process Abstract A state of the art C 5 /C 6 hydroisomerization process was designed as a base case for comparison with a new process that combines reaction and separation in one process unit (McLeary et al. Stud. Surf. Sci. Catal. 135 (2001) 3273). The design was based on processing continuously 1000 ton/day of feed (RON = 74) on an 8000 h a year operation. A final product with a RON value of 86 is predicted; an increase of 12 points compared to the feed RON. The C 5 yield was 95%. The economic analysis of the process showed a total investment of 20 million euros and an annual operating cost of 70 million euros. The increase in RON from feed to product, the C 5 yield and the total investment of the present design are comparable to literature data on typical hydroisomerization processes. This indicates that the present design can be used as a base case for comparison with the new process. This work was developed in cooperation with P.C. Perez, currently at the Laboratory for Process Equipment, TUDelft

54 Chapter 4 1. Introduction Rearranging the structure of hydrocarbons in order to achieve high octane number in gasoline fraction is becoming increasingly necessary because of environmental regulation. To increase the octane number n-pentane and n-hexane (the main component of the light strain run gasoline fraction) are hydroisomerized into isopentane and the dibranched hexanes [1,2]. A typical example of a hydroisomerization process is the Shell Hysomer process [3], first commercialized in 1970, and currently employed in over 70 plants worldwide. This process uses a bifunctional catalyst composed of a noble metal supported on a zeolite i.e. platinum supported on Mordenite. The catalyst is resistant to water and sulfur compounds up to quite high concentrations, eliminating the need for extra pre-treatment facilities. Different from the other hydroisomerization processes that use chlorinated catalysts the Hysomer process doesn t require continuous addition of a chloride activator, removal of HCl from the effluent streams, and precautions against chloride corrosion [4]. The process operates at a temperature in the range of o C and at pressures between 8 30 bar. Due to thermodynamic equilibrium limitations, complete conversions of linear alkanes into branched ones are not achieved by once-through operation. To obtain conversion to extinction the hydroisomerization process can be completed with a physical separation process that allows isolating and recycling of linear alkanes. An example of such a physical separation process is the IsoSiv process of Union Carbide Corporation [5,6] This process provides efficient branched/linear paraffin separation using a molecular sieve (zeolite CaA) unit instead of distillation columns. The mixture of pentanes and hexanes is a close boiling point mixture, which makes the separation by distillation difficult and energy consuming. The use of Zeolite CaA enables the selective adsorption of linear alkanes over branched ones. The branched alkanes have larger molecular diameter, inhibiting them to enter the zeolite CaA pores. The Total Isomerization Process (TIP) is a combination of the Hysomer and the IsoSiv processes [7]. The TIP, commercialized since 1975 [8], is widely used and is considered a process for virtually complete hydroisomerization of the linear paraffins. In the TIP the application of molecular sieve separation and recycle of linear alkanes not only leads to a higher octane number of the product, but also decreases the effect of hydroisomerization temperature on product quality. Therefore, the disadvantage of zeolites with their higher operating temperatures than catalysts based on chlorinated alumina largely disappears. In the present work the design of a state of the art hydroisomerization process is reported. The focus was put on the reaction section, the separation section and on the 44

55 Design of a state of the art C 5 /C 6 hydroisomerization process process economics. The design served as a base case for comparison for a new process that combines reaction and separation in one process unit. The new process described by McLeary et al. [9] uses membrane technology and aims at better performance in technological and economical aspects than the state of the art process. 2. Process description and simulation A simplified scheme of the process is shown in Figure 1. The design was based mainly on the description reported in the patent of Holcombe [7]. The operation conditions for the reaction were obtained from literature data [3,4,10]. For the separation section information was retrieved from the works of Silva [11-13]. The feed streams going into the process were a make-up hydrogen feed and a hydrocarbon feed. The hydrogen purity in the make-up stream was about 85 mol%. The impurity consisted of light hydrocarbons (C 4 -). The hydrocarbon feed had a RON of 74 and contained essentially a mixture of linear and branched C 5 and C 6 hydrocarbons (about 90 wt%). The component distribution in the hydrocarbon feed was similar to that in a feed stream of a typical hydroisomerzation process [3]. The process feed is normally the result of refinery distillation operations, and thus contains small amounts of heptanes, aromatics and cycloparaffins. The hydrocarbon feed was about 1000 ton/day, which is in the range of industrial hydroisomerization process feed streams ( ton/day). Purge Reactor Hydrogen Adsorption/ Desorption section Feed Product Figure 1. Process flow scheme. 45

56 Chapter 4 The feed combined with the recycle stream is sent to the hydroisomerization reactor where mainly the linear alkanes are catalytically isomerized into branched ones in the presence of hydrogen. The hydrogen/hydrocarbon molar ratio was about 1. The catalyst used was a platinum supported on zeolite type. The temperature chosen for the reactor was 250 C and the pressure 20 bar. The reactor effluent is sent to a separator (flash) where the C 5 and higher hydrocarbons are separated from hydrogen and light hydrocarbons (C 1 up to C 4 ). The flash is operated at 29 o C and 15 bar. The effluent streams from the flash unit are sent to the adsorption/desorption section. In this section the hydrogen rich stream is used as purge gas. The stream containing C 5 and higher hydrocarbons is passed as feed at 250 o C and 20 bar periodically in sequence through a system of four columns, each containing a fixed bed of zeolite CaA as adsorbent and undergoing the following stages: A-1, introduction of the hydrocarbon feed: The adsorption bed contains initially the hydrogen purge gas from the previous stage (D-2). The hydrocarbon feed rich in the linear and branched C 5 and C 6 hydrocarbon enters the adsorption bed and forces the hydrogen purge gas out of the bed. A-2, adsorption of the linear alkanes: At this stage the hydrogen purge gas is already removed from the bed. From the hydrocarbon feed that is now present in the adsorption bed, the linear alkanes are selectively adsorbed. The branched alkanes remain in the bulk phase of the hydrocarbon feed and leave the bed as the final product. D-1, purging of the branched alkanes: The bed is purged with hydrogen, countercurrently with respect to the direction of the A-2 adorption, to remove the remaining branched components. D-2, desorption and purging of the linear alkanes: By continuing the supply of the hydrogen purge gas countercurrently with respect to A-2 adsorption the adsorbed linear alkanes are now desorbed and removed from the bed. This is done until the bed consists mainly of hydrogen purge gas. The linear alkanes are then recycled to the reactor. Hydrogen is used in the adsorber/desorber columns and also in the isomerization reactor. The hydrogen feed used is generally a combination of two or more refinery hydrogen streams. Light hydrocarbons containing from 1 to 4 carbon atoms will appear in the course of operation of the process since these low boiling materials are present in the 46

57 Design of a state of the art C 5 /C 6 hydroisomerization process hydrocarbon feed and are also produced in the catalytic unit. To prevent build-up of these low hydrocarbon materials, part of the hydrogen rich stream coming from the separator is purged before entering the adsorption/desorption section. The final product coming from the adsorber/desorber section should contain mainly branched C 5 and C 6 contributing to an improved RON. The process was simulated and designed using Aspen and Excel. 3. Results and discussion The focus of the result and discussion is mainly on the reactor, the adsorber/desorber units and the economics of the process. 3.1 Reactor A fixed bed reactor was selected because of its ease of design and operability. In the reactor mainly the linear pentanes and hexanes are isomerized into the branched isomers. In Table 1 the conversions of the linear components are given. Hydrocracking was considered for all hydrocarbons higher than C 6 at conversion of about 88 wt% with propane and butane as the reaction product. For cycloparaffins and aromatics respectively hydrogenolysis and hydrogenation reactions were assumed. The conversion data were chosen as such to obtain roughly the same components distribution as that of a typical TIP process. The reaction conditions were assumed at a temperature of 250 o C and a weight hourly space velocity (WHSV) of about 1.3 kg hydrocarbon feed h -1 kg -1 catalyst. The WHSV chosen is comparable to the space velocities in typical hydroismerization processes using zeolitic materials [10,14]. Table 1. Conversions of the linear alkanes. Reactant Product Conversion (wt%) n-pentane: isopentane 44 n-hexane: 2-methylpentane (48%) 56 3-methylpentane (29%) 2,3-dimethylbutane (11%) 2,2-dimethylbutane (12%) Value between parentheses is the selectivity, amount of product formed S = x 100% amount of reactant converted In Aspen a stoichiometric reactor model was used at 250 o C and 20 bar. The total amount of catalyst was estimated from the reactor feed flow and the WHSV. The 47

58 Chapter 4 hydrocarbon feed into the reactor was 2160 ton/day. The estimated amount of catalyst and other reactor specification are given in Table 2. Table 2. Reactor specifications. Reactor type Fixed bed adiabatic Temperature ( o C) 250 Pressure (bar) 20 Pressure drop (bar) 0.25 Length (m) 9.2 Diameter (m) 3.1 Catalyst Pt/zeolite Catalyst amount (ton) 69 Particle diameter (m) (sphere) Catalyst density (ton/m 3 ) 1 (assumed) 3.2 Adsorber/Desorber In designing the adsorber/desorber columns the adsorption of mainly n-pentane and n-hexane was considered since these components are the most important ones to be recycled to the isomerization reactor. Adsorption equilibrium isotherms of n-pentane and n- hexane on zeolite CaA were plot according to the equation of Nitta et al. [15]: 1 θ Keq = (1) p (1 - θ) n Cs θ = Cs (2) max where: Keq n p Cs Cs max the equilibrium constant (1/bar) number of active sites occupied by an adsorbed molecule partial pressure of the adsorbate (bar) concentration of adsorbate in the adsorbed phase of the solid (g/l) maximum concentration of adsorbate in the adsorbed phase (g/l) Figure 2 and 3 show the plots of the isotherms for n-pentane and n-hexane, respectively. The isotherm parameters of n-c 5 and n-c 6 were obtained from the work of Silva et al. [11-13]. The isotherms show a concave to the fluid concentration axis, which 48

59 Design of a state of the art C 5 /C 6 hydroisomerization process characterizes it as a favourable isotherm. This means that the points of high concentration in the adsorption wave move more rapidly than the points of low concentration. Therefore, the adsorption zone becomes narrower as it moves along the bed, termed self sharpening [16,17]. The design of the adsorber/desorber was based on this concept. Considering equilibrium over the whole column the assumption made is that at the breakthrough time (saturation time) the bed is fully saturated. 10 n-c 5 Loading (g /100g adsorbent) n-c 5 partial pressure (bar) Figure 2. n-pentane isotherm at 250 o C using zeolite CaA. 10 n-c 6 Loading (g / 100g adsorbent) n-c 6 partial pressure (bar) Figure 3. n-hexane isotherm at 250 o C using zeolite CaA. 49

60 Chapter 4 Assuming a bed void fraction of 0.5 the total volume of the bed was determined using the following model [16]: Co Vbed = Q t (Co Cs) (3) where: Vbed volume of the adsorbent bed (m 3 ) Q total volumetric feed flow (m 3 /s) t saturation time (s) Co concentration of adsorbate in the fluid phase (g/l) Cs concentration of adsorbate in the adsorbed phase of the solid (g/l) The feed stream into the adsorber/desorber section was 1860 ton per day. The molar compositions of n-c 5 and n-c 6 in the feed stream were 20.3% and 2.6%, respectively. Ideal gas behaviour was considered. Saturation time of 4 min was used [7]. The estimated amount of adsorbent and other column specifications are given in Table 3. The desorption step of the column is performed by purging countercurrently the adsorbed straight chain parrafins in the zeolite bed with the hydrogen rich stream coming from the flash separation vessel. Table 3. Adsorber/desorber column specification. Column bed type Fixed bed Temperature ( o C) 250 Pressure (bar) 20 Length (m) 6.5 Diameter (m) 2.2 Adsorbent Zeolite 5A Adsorbent amount (ton) 14.3 Particle diameter (m) (sphere) Adsorbent density (ton/m 3 ) 0.6 (assumed) To assure a continuous process an adsorption/desorption cycle is needed. Four columns are needed which is a typical figure for the TIP process. The description of the adsorption/desorption steps was already mentioned in the process description and simulation section. The time required to complete one step is half of the breakthrough time [18]. Thus, 2 min were used for one step. As there are four columns, each column will be assured to perform until complete saturation of the zeolite bed within the first two steps of adsorption (A-1 and A-2). The complete adsorption/desorption cycle time is 8 min. Table 4 illustrates 50

61 Design of a state of the art C 5 /C 6 hydroisomerization process the adsorption/desorption sequence. The adsorption/desorption section data is summarized in Table 5. Table 4. Adsorption/desorption sequence. Adsorption/desorption stage 0-2 min 2-4 min 4-6 min 6-8 min Column 1 A-1 A-2 D-1 D-2 Column 2 A-2 D-1 D-2 A-1 Column 3 D-1 D-2 A-1 A-2 Column 4 D-2 A-1 A-2 D-1 Table 5. Adsorber/desorber section summary. Adsorbent Zeolite CaA Number of columns 4 Saturation time (min) 4 Step time (min) 2 Adsorption/desorption cycle time (min) 8 Column L x D ( m) 6.5 x 2.2 Mass of adsorbent per column (ton) 14.3 Total mass of adsorbent (ton) 57.2 Roughly 92% of the n-c 5 and 97% of the n-c 6 present in the adsorber/desorber feed stream were separated and recycled to the reactor. For the simulation performed in Aspen a separator representing the adsorber/desorber section was used. The split fractions used for the different component are given in Table 6. The split fraction was defined as the ratio between the amount of a component i leaving the adsorber/desorber section and the amount of the component i entering the section. The fractions were chosen as such, that a typical product could be obtained. As the zeolite adsorbents can provide absolute separation unity values were chosen for the branched alkanes, aromatics and the cycloparaffins. Table 6. Split fractions of the Separator used in Aspen representing the Adsorber/desorber section. Component Split fraction H C n-c n-c n-c Branched alkanes 1.00 Aromatics and cycloparaffins

62 Chapter Overall simulation The overall simulation of the process illustrated by the simplified block scheme of Figure 4 resulted in the flows and yields given in Table 7 for the main process streams. The yield is defined as the ratio between the stream and the hydrocarbon feed stream (stream <1>). The hydrogen used contains hydrocarbon impurities and that accounts for the slightly higher product yield compared to the hydrocarbon feed. <9> <2> <7> <10> <8> <1> <3> <4> REACTOR SEPARATOR ABSORBER / <5> <6> DESORBER <11> 250 o C 20 bara 29 o C 15 bar 250 o C 20 bar Figure 4. Process block scheme with the main process streams. Table 7. Summary of the main process streams. Stream Flow (ton/day) Yield (ton/ton) <1> Hydrocarbon feed (HC) <2> Make-up H <3> HC make up H <4> Reactor feed <5> Reactor effluent <6> Adsorber/desorber Feed <7> H 2 from separator <8> Adsorber/desorber purge <9> Recycle <10> Purge <11> Product The composition of the final product obtained from the simulation is given in Table 8. The final product has a RON of 86; an improvement of 12 points in relation with the process feed (RON 74). The process has a 95% C 5 yield. 52

63 Design of a state of the art C 5 /C 6 hydroisomerization process The C 5 yield is defined as: mass of C 5 and higher hydrocarbon in product C 5 Yield = 100% mass of hydrocarbon feed (4) The RON of the streams were estimated as follows: ( i i ) (5) RON = x RON where RON i and x i are the RON and weight fraction of component i present in the stream. Table 8. Hydrocarbon feed and product distribution. Components Feed (wt%) Product (wt%) C n-pentane methylbutane ,2-dimethylpropane n-hexane methylpentane methylPentane ,3-dimethylbutane ,2-dimethylbutane C aromatics cycloparaffins RON C5 Yield (%) Economics The economic performance of the process was based on an 8000 hours a year operation, an isomerate product price of 230 euros/ton, and an amount and price of the hydrocarbon feed of 1000 ton/day and 162 euros/ton, respectively [19]. An economic lifetime of ten year was assumed. The total investment, operating cost and the product based cost (cash) factor of the different items are described in the following sections. 53

64 Chapter Total Investment The total investment is the sum of the fixed costs, the working capital and the start up costs. [20]. The Lang factorial method of cost estimation was used in order to estimate the fixed capital as a function of the total purchase equipment cost [21]. C = f PCE (6) f where C f, f, PCE are the fixed capital, the estimation factor and the total purchase equipment cost, respectively. Table 9 shows shows typical factors for estimation of project fixed capital. The total physical plant cost (PPC) is calculated as: PPC = (1f 1..f 9 ). PCE => PPC = PCE (7) The relation between the fixed capital and the PCE can be determined as: C f = (1f 10 f 12 ). PPC = PPC => C f = PCE (8) Table 9. Factors for fixed capital estimation [21]. Total purchase cost: Equipment erection f Piping f Instrumentation f Electrical f Buildings, process f Utilities f Storage f Site development f Ancillary buildings f Subtotal (1f 1.f 9 ) 2.97 Total physical plant cost: Design & Engineering f Contractor's fee f Contingency f Subtotal (1f 10.f 12 )

65 Design of a state of the art C 5 /C 6 hydroisomerization process The PCE was estimated with help of literature data [20-22] and updated using European base indexes. The PCE estimated was 8.5 million euros. Figure 5 shows the contribution of the different items to the PCE. The catalyst and the heat exchangers had a significant impact on the overall cost. These two items are the major cost drivers and it is recommended to optimize catalyst as well as the energy utilization. Reactor vessel 1.1% Adsorbent 5.3% 2.7% Ads/Des columns 0.6% Flash 4.3% Compressor 1.0% Pumps 16.8% Heat exchangers Catalyst 67.2% Figure 5. Distribution of the Purchase Equipment Costs The catalyst and adsorbent were included in the PCE assuming that they are bought once and can be regenerated each two years. For the estimation of the fixed capital the Lang factor of 3.86 was used over the PCE, excluding the catalyst and the adsorbent. Over these two items a factor of 1.17 was used, since not all the factors of Table 9 can be applied on them. The total fixed capital cost was calculated as: C f = (PCE - PCE ) PCE (9) where PCE is the cost of the catalyst and the adsorbent. The estimated total investment is given in Table 10. The estimated value is comparable to that of state of the art hydroisomerization processes [23]. An IFP plant with half of the production capacity is estimated at 12 million euros. A UOP Penex/Molex plant with similar capacity is estimated at 23 million euros. 55

66 Chapter 4 Table 10. Total investment estimation. Million euros Fixed Capital 16.3 Working capital (15% of the fixed capital) 2.4 Start-up costs (8% of the fixed capital) 1.3 Total Investment Operating Costs The operating costs can be divided in direct cost (fixed and variable cost) and indirect cost. In Table 11 the estimation of the operating cost is given based on typical figures [21]. The operating cost per ton of product is 209 euros. Table 11. Operating cost. Million euros/year Direct cost: - Variable (raw materials, utilities, etc) 58 - Fixed (maintenance, operating labour, insurance, etc) 6 Indirect cost (sales expense, general overheads, R&D etc) 6 Annual operating cost Product based cost (cash) factor The product based cost (cash) factor of an item i (X i ) is defined as: total cost or cash of an item i entering or leaving the process X i = total price of the product (10) The product based cost (cash) factor scheme of the process is illustrated in Figure 6. The items entering the process are the costs while those leaving the process provide cash. From the figure it can be deduced that the different items related with the cost do not have a significant financial impact on the process. There is still a financial margin between the cash and the cost although the feed has a factor of 0.7. As heating medium the heat transfer fluid Dowtherm Q [24] was used instead of high pressure steam. The hot oil contains very stable coumpounds (diphenylethane and 56

67 Design of a state of the art C 5 /C 6 hydroisomerization process alkylated aromatics) so it can be recirculated. One of the cost related to the hot oil is the amount of energy supplied to it before entering the process. This energy can be supplied by using fuel. The financial impact of the fuel cost is shown in Figure 6. As the purge gas contains mainly light hydrocarbons a similar price as that of fuel was used. Hot Oil Fuel Cooling Water Electricity Feed Purge Hydro-Isomerization Process H Product Figure 6. Product based cost (cash) factor scheme. 4. Concluding remarks In the present work a design of a state of the art C 5 /C 6 hydroisomerization process was developed. The process was designed to process continuously 1000 ton/day of feed with RON 74, containing mainly C 5 and C 6 compounds. The process consisted of a hydroisomerization reactor and adsorber/desorber units. In the reactor the feed was converted into C 5 and C 6 isomers over a Pt/zeolite catalyst. The adsorber/desorber units were packed with zeolite CaA which separated the final product from the linear alkanes. A product with a RON value of 86 was predicted. This is an increase of 12 units compared to the feed RON. The C 5 yield obtained was 95%. The economics of the process showed a total investment of 20 million euros. An annual operating cost of 70 million euros was estimated (209 euros/ton product). These values are comparable with typical C 5 /C 6 hydroisomerization processes. Thus, the present design can be used as a basis for comparison with the new process described by McLeary et al [9]. In Table 12 a summary is presented of the results of the present design and the new process. 57

68 Chapter 4 Table 12. Summary of the present design and the new process State of the Art New Process [9] Feed RON Final product RON C 5 Yield (%) Total investment (million euros) Annual Operating cost (million euros) References [1] A. Chica, A. Corma, J. Catal. 187 (1999) 167. [2] C.L. Li, L. Shi, G.X. Huang, R.Y. Wang, Chem. Eng. Comm. 121 (1993) 1. [3] H.W. Kouwenhoven, W.C. van Zijll Langhout, Chem. Eng. Progr. 67 (1971) 65. [4] S.T. Sie, in: Handbook of Heterogeneous Catalysis, Vol. 4, eds. G. Ertl, H. Knözinger, J. Weitkamp (VCH Verlags Gesellschaft mbh, Weinheim, 1997) ch. 3, pp [5] G.J. Griesmer, W.F. Avery, M.N.Y. Lee, Hydrocarbon Processing 44 (6) (1965) 147. [6] J.H. Olive, M.F. Symoniak, Oil Gas J. 26 June 1972, pp. 68. [7] Th.C. Holcombe, U.S. Patent , [8] M.F. Symoniak, Hydrocarbon Processing May 1980, pp [9] E.E. McLeary, R.D. Sanderson, C. Luteijn, E.J.W. Buijsse, L. Gora, Th. Maschmeyer, J.C. Jansen, Stud. Surf. Sci. Catal. 135 (2001) [10] A. Hennico, J-P. Cariou, Hydrocarbon Technology International 1990/1991, 68. [11] J.A.C. Silva, F.A. da Silva, A.E. Rodrigues, Stud. Surf. Sci. Catal. 120 (1998) 371. [12] J.A.C. Silva, A.E. Rodrigues, Ind. Eng. Chem. Res. 36 (1997) 493. [13] J.A.C. Silva, A.E. Rodrigues, Ind. Eng. Chem. Res. 36 (1997) [14] A.I. Lugovskoi, S.A. Loginov, V.A. Sysoev, S.A. Makeev, A.N. Shakun, M.L. Fedorova, Chem. Tech. Fuels Oils 36 (5) (2000) 330. [15] T. Nitta, T. Shigetomi, M. Kuro-Oka, T. Katayama, J. Chem. Eng. Japan 17 (1984) 39. [16] Coulson & Richardson s Chemical Engineering Vol. 2, 4th Edition, [17] Wesselingh and Kleizen, Scheidingprocessen, [18] Perry and Green, Perry s Chemical Engineers Handbook, 7th edition, MGH Intl., (1998), section 16 Adsorption and Ion Exchange. [19] Sager, T.C., Reno, M.E., Denny, R.F., UOP Inc., Cost effective isomerisation options for gasoline processing requirements, Hydrocarbon Technology International, [20] Biegler, L.T., Grossmann, I.E. and Westerberg, A.W. Systematic Methods of Chemical Process Design, Pretence Hall,

69 Design of a state of the art C 5 /C 6 hydroisomerization process [21] R.K. Sinnott, Coulson and Richardson Chemical Engineering Volume 6, Chemical Engineering Design, Butterworth-Heinemann, Oxford, [22] DACE Prijzenboekje, 17e editie, mei [23] Special Report Refining Processes 98, Hydrocarbon Processing: International edition, 77, 11 (November 1998) 98. [24] accessed April

70 Chapter 4 60

71 5 Hydroisomerization of hexane within a reactor composed of a tubular silicalite-1 membrane packed with Pt-loaded chlorided alumina catalyst Abstract Experiments with hexane were carried out using a reactor, in which linear molecules were separated from monobranched ones on a silicalite-1 membrane prior to conversion on a Pt-loaded chlorided alumina catalyst bed. The results indicate separation selectivity with a factor higher than 20, high hexane conversion and product selectivity towards dibranched isomers. These results suggest that the system has potential in upgrading low octane value hydroisomerization feed streams. This chapter is based on: M.L. Maloncy, L. Gora, E.E. McLeary, J.C. Jansen, Th. Maschmeyer, Catal. Commun. 5 (2004) 297

72 Chapter 5 1. Introduction Environmental restrictions imposed on gasoline resulted in the removal of lead compounds and a strengthening of the limits for toxic compounds such as aromatics, in particular benzene, olefins and sulfur-containing components, and reformulation in gasoline composition is occurring worldwide. In the European Union the limits placed on gasoline composition for the year 2000 were sulfur levels of 150 ppm, 18% olefins, 1% benzene, 42% aromatics and 2.7% oxygen. For the year 2005 sulfur must be further reduced to 50 ppm and aromatics to 35% [1]. As the octane boosters (Lead, MTBE, ETBE, aromatics) are removed or reduced in the gasoline pool, hydroisomerization of light alkanes is becoming extremely important as an alternative for octane upgrading. Since the branched isomer products formed have high research octane numbers (RON) and burn cleanly they are the only acceptable alternative as octane booster. Hydroisomerization of light alkanes such as C 5 and C 6 is already applied industrially in a process consisting of a reaction and separation section. In zeolite catalyst based hydroisomerization processes high temperature is employed, however these processes are far from optimal. The high temperature pushes the equilibrium to the wrong side, and the setup is fundamentally equilibrium limited. Additionally a pressure swing adsorption separation unit follows the conversion. This means that large volumes of gas are continuously pumped around. Combination of these two sections into one unit could offer new and promising industrial features, densifying the process and reducing operating costs [2]. For the hydroisomerization of hexane the reactor concept as shown in Figure 1 was applied. In this concept a zeolite membrane surrounds a catalyst bed. There is no recycling of unconverted components. However, if needed, two or more reactors could be connected in sequence. The process feed, mainly linear and partly branched molecules, contacts the membrane in a parallel-passage mode. Figure 1. Reactor concept for hexane hydroisomerization. 62

73 Hydroisomerization of hexane within a reactor composed The high RON branched molecules (2,3-dimethylbutane RON-105.0; 2,2-dimethylbutane 91.8; 2-methylpentane 73.4; 3-methylpentane 74.5) do not permeate and leave the system as part of the product. The linear molecules permeate through the membrane pore and are mostly converted into branched products on the catalyst bed. The reaction products together with the unconverted linear components leave the system and form the remaining part of the product. 2. Experimental Hydroisomerization experiments were performed using a 10 cm tubular membrane reactor. The membrane part of the reactor, containing Silicalite-1 as the selective layer, was prepared by using a two-step temperature synthesis (120 O C for 114 hours followed by 180 O C for 17 hours). The development of this membrane is described elsewhere [3]. The membrane area is about 2.98 x 10-3 m 2. The reaction part of the reactor contains a catalyst bed of Ptloaded chlorided alumina catalyst particles, supplied by Akzo Nobel. At low temperature operation, i.e. 120 o C, the formation of the branched products is enhanced making it preferable to use catalysts composed of noble metal supported on chlorided alumina than zeolite-supported ones. The experimental set-up (see Figure 2) consists of the reactor in a hot air oven, feed and carrier gas supply units and analytical parts. He or H 2 Retentate P P Permeate and/or Reaction product He P Silicalite-1 membrane catalyst space Shell side n-c 6 i-c 6 tube side oven Figure 2. Reactor set-up. Two sets of experiments were carried out to examine the performance of the reactor. The first experiment was performed without catalyst and aimed to determine permeation data 63

74 Chapter 5 (selectivities, fluxes). A feed mixture of 80 mol% n-c 6 and 20 mol% 2-MP at a flow rate of 0.08 ml min -1 was supplied via a HPLC pump to the system. Helium was used as a carrier and sweep gas at a flow rate of 50 ml min -1 each. In the second set of experiments the combined separation and catalytic features of the membrane reactor were analysed by filling the membrane tube (volume of 7.46 x 10-6 m 3 ) with 5.20 g of catalyst. The sweeping He gas was replaced with H 2, because H 2 is required for hydroisomerization. The space velocity and H 2 /C 6 molar ratio inside the reactor were 2.39 mmol hexane feed g -1 Catalyst h -1 and 9.8, respectively. An additional experiment was performed with the catalyst in its inactive form to compare permeation behaviour with that of the experiment without catalyst. All experiments were performed at 120 o C. Pressures on both sides of the membrane were equal to 1 bar. Feed, retentate and permeate streams were analysed with an on-line gas chromatograph (FID detector for the hydrocarbons measurements). A combination of two GC-columns, CP-Sil PONA CB, df = 0.5 µm (50 x 0.21 mm and 100 x 0.25) was used in series to determine the amounts of hydrocarbons quantitatively. 3. Results and discussion Figure 3 shows SEM images of the cross-section and surface of the silicalite-1 membrane. The images show a well-intergrown layer without the visible intercrystalline boundaries between the crystals forming the membrane. This indicates that the preparation by the two-step synthesis results in a promising quality of silicalite-1 membrane. Figure 3. SEM images of the silicalite-1 membrane: Cross section (left) and Top view (right). Separation experiments performed at 120 o C (first set of experiments) suggest a relatively good membrane performance with n-c 6 flux and selectivities comparable and even higher than those reported in literature [4,5]. The fluxes (Φ), separation selectivities (α) and molar fractions of the feed components going through the membrane (f) are given in Table 1. 64

75 Hydroisomerization of hexane within a reactor composed Table 1. Flux (Φ i ), separation selectivity (α) and separation fraction (f) at 120 o C. Component i Φ i (mmol m -2 s -1 ) α i/2-mp f n-c MP The experiment with the inactive catalyst produced similar permeance data as given in Table 1, consistent with the permeance through the membrane not being hindered by the physical presence of the catalyst. The reaction products should have little influence on the permeance, since the hydrogen sweeping gas is continuously removing these components. The composition of the feed, permeate (experiment with no reaction) and product stream (experiment with reaction) as well as the product selectivities are given in Table 2. Table 2. C 6 isomers distribution and product selectivities (S i ) at 120 O C, n-c 6 conversion is 73 mol%. Component i Distribution (%) S i (%) Feed Permeate Products n-c MP MP ,3-DMB ,2-DMB Cracked (<C 6 ) A n-hexane conversion of 73 mol% was achieved. The n-hexane conversion (X n-c 6 ) and the product selectivities (S i ) were calculated as follows: ( moles of n-c 6 permeated ) - ( moles of n-c 6 in reaction product stream ) X n-c 6 = x 100% ( moles of n-c permeated ) 6 ( moles of reaction product i formed ) S i = x 100% ( moles of permeated n-c converted ) 6 As can be seen from Table 2 the catalyst is very selective towards isomerization compared to cracking; less than 1 mol% of the converted n-hexane is cracked into smaller components. The relatively high selectivity towards the dibranched components, 36%, is of great importance when aiming at high RON products. The estimated RON numbers of the different streams are given in Table 3. The RON of the streams were estimated as follows: 65

76 Chapter 5 ( i i ) RON = x RON Where, RON i and x i are the RON and fraction of component i present in the stream. Table 3. RON values for different process streams. Stream RON Feed 35 Permeate 25 Retentate 39 Reaction Product (C 6 RON) 67 Reaction product Retentate (C 6 RON) 49 Reaction product Retentate (i-c 6 RON) 78 The isomerised product gave a RON value of about 3 points higher compared to values determined from literature data on n-hexane hydroisomerization using Pt-loaded zeolite beta and mordenite catalyst at similar n-c 6 conversion [6,7]. Improvement in RON of 14 points was achieved when going from feed to the combined stream of reaction product and retentate. Industrial processes show upgrading of around 10 points. In contrast to the feed used in this work, industrial feed streams (RON of around 70) contain substantial amounts of i-c 5 and i-c 6 components. Since the membrane has shown reasonable selectivity and separation, higher retentate RON values can be expected if a representative industrial process feed stream is used. Additionally, the use of two or more of similar membrane reactor systems in series could result in higher octane upgrading. In contrast to current processes that use distillation columns or adsorption systems for separation, the system presented does not have the disadvantages of high energy consumption, pseudo-continuous or cyclic operation that often rely on large volume of a third fluids for desorption and delivery steps. Moreover, permeation of compounds that are strong poison for the catalyst i.e. moisture or sulfur compounds, could be retained by the membrane in the system. However, to explore the inherent advantages of the combined system making it compatible with existing state of the art processes, further optimisation is needed. Important points needing further investigation are: 1) production of thinner membranes that provide higher permeation fractions and selectivities; 2) study on optimal space time velocity resulting in higher conversion and product selectivities towards dibranched isomers. These two points should provide the combined separation and reaction system at its optimal condition. 66

77 Hydroisomerization of hexane within a reactor composed 4. Conclusions Results of n-hexane hydroisomerization experiments performed within a reactor composed of a tubular silicalite-1 membrane packed with Pt-loaded chlorided alumina catalyst indicate the potential of hydroisomerization processes within a combined separation and reaction system. Selectivity with a factor higher than 20, high n-hexane conversion, high selectivity towards dibranched isomers, and RON upgrading capacity of 14 points were obtained. As there is space for optimisation, further development on the application of hydroisomerization processes using these systems will depend on progress in the fabrication of high flux, high selectivity zeolite membranes. Acknowledgement Akzo Nobel in Amsterdam, nowadays Albemarle Catalysts is acknowledged for providing the catalyst. References [1] T.G. Kaufmann, A. Kaldor, G.F. Stuntz, M.C. Kerby, L.L. Ansell, Catal. Today 62 (2000) 77. [2] E.E. McLeary, R.D. Sanderson, C. Luteijn, E.J.W. Buijsse, L. Gora, Th. Maschmeyer, J.C. Jansen, Stud. Surf. Sci. Catal. 135 (2001) [3] Chapter 2 of this thesis. [4] H.H. Funke, A.M. Argo, J.L. Falconer, R.D. Noble, Ind. Eng. Chem. Res. 36 (1997) 137. [5] C.L. Flanders, V.A Tuan, R.D. Noble, J.L. Falconer, J. Membr. Sci. 176 (2000) 43. [6] T. Yashima, Z.B. Wang, A. Kamo, T. Yoneda, T. Komatsu, Catal. Today 29 (1996) 279. [7] K.-J. Chao, H.-C. Wu, L.-J. Leu, Appl. Catal. A 143 (1996)

78 Chapter 5 68

79 6 Hydroisomerization of heptane: mechanistic aspects and industrial challenges Abstract This report presents an overview on heptane hydroisomerization. The reaction mechanisms involved in the synthesis of heptane isomers are discussed, as well as the catalysts employed. The focus in this report is on acid-catalyzed hydroisomerization utilizing zeolite-based catalysts. The need for branched heptane isomers is expected to increase due to more stringent environmental regulations. The first heptane hydroisomerization process will most likely be operational when the bottlenecks in the process development are solved. These include: 1) the selectively of the catalyst to produce multibranched isomers, 2) the separation of mono- from multibranched isomers and 3) controlling hydroisomerization versus hydrocracking. This chapter is based on: M.L. Maloncy, V.F.M. Tjon Soei Len, J.C. Jansen, J.A. Moulijn, to be submitted for publication

80 Chapter 6 1. Introduction Over 760 million tons of petroleum is processed per year in refineries in Europe alone to meet the demand for liquid transportation fuels such as gasoline, diesel and jet fuel [1]. Despite major innovations in the automobile industry over the past decades, cars still have combustion engines that run on gasoline with a minimum octane number of around 88 [2]. Not until engine technology advances to alternative engines, e.g. fuel cells, worldwide consumption of transportation fuels will continue to grow and is expected to remain strong [1-3]. In the past, lead-containing compounds like tertraethyllead, were used in gasoline to enhance the octane number. These octane boosters were banned because of their negative impact on the environment. Environmentally driven regulations are requiring significant improvement in the quality of gasoline in many parts in the world. Due to the concerns of groundwater contamination, high octane oxygenates such as MTBE are under increasing pressure to be eliminated from gasoline. The actual lead phase-out in gasoline together with the future restrictions in the aromatic content and MTBE will have a negative impact on the octane number and the total amount of gasoline produced will be reduced [4]. This even worsens if one considers that the requirement of ultra-low sulfur content could mean removing part of the higher boiling gasoline produced in the FCC unit from the gasoline pool [1]. Taking into account the above-described scenario, together with further limitations in the olefins content in order to decrease ozone problems, it is evident that the hydrocarbon composition of the fuel will shift away from aromatics and olefins to hydrocarbon types such as naphtenes and branched paraffins that burn cleaner [3]. An example of a gasoline pool composition is shown in Table 1. Table 1. Different refinery processes contribution to the gasoline pool composition (vol%), adapted from [4]. Distribution Aromatics in Pool Olefins in Pool FCC Reforming Alkylation 13.3 < 0.1 < 0.1 Light Straight Run < 0.1 Isomerization 4.0 < 0.1 < 0.1 Hydrocracking 2.7 < 0.1 < 0.1 Coker 0.7 < Polymerization 0.3 < Dimerization 0.2 < Butanes Total RON

81 Hydroisomerization of heptane: mechanistic aspects and industrial challenges The olefins and aromatics will have to be substituted by other high-octane components in the near future. Therefore, there is an increasing interest for new octane enhancement processes, particularly for the hydroisomerization of heptane and octane paraffins. Heptane represents an important part of naphta, but has a low Research Octane Number (RON 0). This explains the need to isomerize it into monobranched paraffins (RON 42-65) and preferably into multibranched isomers (RON ). The RON numbers, as well as the abbreviations of heptane isomers used further in the text, are presented in Table 2. Table 2. RON numbers of heptane isomers [5]. Hydrocarbon Abbreviation RON n-heptane nc7 0 2-methylhexane 2MHx methylhexane 3MHx ethylpentane 3EP ,3-dimethylpentane 23DMP ,4-dimethylpentane 24DMP ,2-dimethylpentane 22DMP ,3-dimethylpentane 33DMP ,2,3-trimethylbutane 223TMB The objective of this work is to present an overview of the syntheses of branched heptane isomers. These branched alkanes can be produced through acid-catalysed hydroisomerization of heptane. Although the hydroisomerization of heptane does not occur in the absence of catalysts, the description of the process is divided into a reactions and a catalysts part. In section 2 the reaction mechanisms are discussed. The focus is on acidcatalysed isomerization reaction mechanisms. The relation between hydrocracking and hydroisomerization is also discussed. In section 3 the different catalysts types employed in hydroisomerization are presented with emphasis on zeolite type catalysts. Section 4 includes a brief discussion of the major trends in the oil industry, an appraisal of the branched isomerized product and the main challenges in the heptane hydroisomerization production. Finally concluding remarks are given in section 5. 71

82 Chapter 6 2. Synthesis and reaction mechanisms Acid-catalyzed isomerization of higher alkanes (>C 5 ) occurs via carbenium ions as intermediates [6-10]. The isomerization reaction itself is part of a chain reaction involving chain initiation, carbenium ion rearrangement and chain propagation. In acid-catalyzed isomerizations the reaction intermediate may not exist as free carbenium ion, but the charged intermediates may be complexed with the acid catalyst. Discussions on carbenium ion formation can be found also in [9-15]. For the sake of simplicity, the acid-catalyzed isomerization is sometimes represented as carbenium ions, without showing the complexed bond to the catalyst. In section 2.1 the formation and reactions of the intermediates are discussed. In section 2.2 the reaction mechanisms in the hydroisomerization process of heptane are discussed as well as the isomerized products that can be formed. In section 2.3 the relation between hydroisomerization and hydrocracking is presented and finally in 2.4 an overview of the isomerization reactions of heptane is given Reaction intermediates Formation of carbenium ions Carbenium ions can be formed from alkanes by carbonium ion intermediates, as shown in Figure 1. The transformation of alkenes into carbenium ions is shown in Figure 2. H H H CH H2 CH Figure 1. Formation of carbenium ion from alkanes adapted from [11]. H CH CH 2 Figure 2. Formation of carbenium ion from alkenes by proton addition. 72

83 Hydroisomerization of heptane: mechanistic aspects and industrial challenges Reactions of carbenium ions A. Intramolecular reactions: Hydride and alkyl shifts Important reactions of carbenium ions include isomerizations proceeded by 1,2-hydride and 1,2-alkyl shifts. In the mechanism, it may involve a three-membered (or larger) ring intermediate and a more or less symmetrical transition state involving concerted bond making and breaking [11]. An example of a carbenium reaction involving hydride shift is shown in Figure 3. Alkyl shifts involve the migration of alkyl groups and an example of this reaction is depicted in Figure 4. R CH 2 O Al O H H R' R CH 2 R' R' R C CH CH CH O O O O O O Al Al O O O O Figure 3. Hydride shift in carbenium ion reaction on zeolite. Adapted from [11]. Methyl shift Figure 4. Methyl shift in carbenium ions. Beta-elimination When an adsorbed carbenium ion reacts to decouple the proton to the surface with desorption of an alkene, the reverse reaction of the protonation of an alkene occurs. This type of reaction is called beta-elimination and proceeds as depicted in Figure 5. H Surface O - Surface OH Figure 5. Beta-elimination of a carbenium ion, adapted from [11]. 73

84 Chapter 6 Beta-scission An important reaction of carbenium ions is beta-scission, in which the C-C bond located beta to the charged carbon atom is broken. This cracking reaction [10-12] is shown in Figure 6. beta scission H beta scission with hydride shift Figure 6. Beta-scission of a carbenium ion, adapted from [11]. B. Intermolecular reactions: The intermolecular reactions involve carbenium ions as Lewis acids reacting with alkanes and alkenes. In the reaction with alkanes, the carbenium ion is converted into an alkane and vice versa (hydride shift), as shown in Figure 7. With alkenes dimerization occurs; the alkene is added to the carbenium ion forming a larger carbenium ion. In fact, this is the reverse of the beta-scission reaction shown in Figure 6. H hydride shift CH Figure 7. Reaction of carbenium ion of with an alkane, adapted from [11] Mechanisms of acid-catalyzed isomerization Monomolecular mechanism The simplest way to visualize the skeletal isomerization is by the classical alkyl shift, i.e. a methyl or ethyl shift [10], in which a methyl group is detached from the carbenium ion chain as a methyl ion and reattached at another place. This simple reaction mechanism is contestable. The reason is that the methyl ion is a high-energy species and detachment from the carbenium ion chain involves prohibitively high activation energy, as can be seen from 74

85 Hydroisomerization of heptane: mechanistic aspects and industrial challenges Table 3. Another argument against the classical alkyl shift mechanism is that from an energy point of view, this mechanism would favour the formation of isomers with larger side groups than methyl. The most likely mechanism of skeletal isomerization of the intermediate carbenium ion, involves the rearrangement of the classical secondary carbenium ion into a non-classical carbonium ion, namely a protonated dialkylcyclopropane. The transformation of the classical secondary carbenium ion into the protonated dialkylpropane will not involve a high energy barrier, because of the hybrid resonance structures as shown in Figure 8. The resonance contributes to the stability of the strain in the ring and is demonstrated by the high stability of the cyclopropenylium cation. Table 3. Heats of formation of alkyl carbenium ions [16]. Carbenium ion Heat of formation (kj mol -1 ) Methyl (primary ion) 1080 Ethyl (primary ion) 942 n-propyl (primary ion) 913 n-butyl (primary ion) 883 i-propyl (secondary ion) 812 s-butyl (secondary ion) 795 t-butyl (tertiary ion) 728 R C C R' R C C R' R C C C H C H C H R' R R' Figure 8. Resonance structures of protonated dialkylcyclopropane, adapted from [13]. In the remainder of the text, the protonated cyclopropane (PCP) is drawn as a triangle with dotted lines and a sign indicating the protonation. The evidence for the existence of the PCP mechanism for isomerization is further supported by isomerization experiments done by Weitkamp [9] and by Sie [10,15]. According to the PCP mechanism for isomerization of higher alkanes (>C 6 ), the formation of the 2-methyl isomer will be less likely than that of the 3-methyl isomer, which is experimentally observed. Weitkamp [9] demonstrated that the surprisingly low rate of formation of 2-methyl isomers were consistent with a branching mechanism via protonated cyclopropanes while they could not be explained by a classical mechanism via alkyl and hydride shifts. The location of the breakage of the bonds in the PCP isomerization mechanism depends on which kind of carbenium ion is formed. As can be seen from the heat of formation in Table 3, the stability of carbenium ions decreases in the following order; 75

86 Chapter 6 tertiary>secondary>primary. The possible and forbidden rearrangements are illustrated with the example of the s-butyl cation in Figure 9. The mechanism of acid-catalyzed isomerization via the PCP is shown in Figure 10. carbenium ion b a a H b PCP Figure 9. Possible and forbidden rearrangements of the s-butyl cation, adapted from [10]. linear paraffin hydride abstraction/transfer classical carbenium ion non-classical carbonium ion classical carbenium ion hydride transfer isomerized product Figure 10. Mechanism of acid-catalyzed isomerization of n-heptane, adapted from [6,10,17]. The skeletical C-C bond rearrangements of an acyclic alkylcarbenium ion formally belong to either type A or type B isomerization. Figure 11 illustrates this type of isomerizations. In type A isomerization, the position of a side chain changes, but the number of primary, secondary, tertiary and quaternary C atoms in the molecule are the same. In type B isomerization, the degree of branching changes and consequently the number of primary, secondary, tertiary and quaternary C atoms in the molecule. 76

87 Hydroisomerization of heptane: mechanistic aspects and industrial challenges H 1,2 hydride shift 3 ring closure TYPE A 3 ring opening 1,2 hydride shift H ring closure corner to corner H jump TYPE B 3 ring opening 1,2 hydride shift H Figure 11. Monomolecular mechanism of type A and type B isomerization, illustrated with the formation of the carbenium ion of 3MHx, adapted from [12] Bimolecular mechanism Blomsma and co-workers [18] stated that hydroisomerization of higher alkanes can also occur through a bimolecular mechanism. This involves the addition of an alkene to an adsorbed tertiary alkylcarbenium ion, followed by type A isomerization and beta-scission. After beta-scission, the initial tertiary alkylcarbenium ion is regenerated and can enter a new catalytic cycle. The conventional monomolecular isomerization of heptane to 2-methylhexane occurs via type B rearrangement. The bimolecular isomerization mechanism involves faster alkylcarbenium ion reactions compared to the monomolecular one. The reaction shown in Figure 12 is an alkylation/dealkylation mechanism, with an intermediate skeletal rearrangement. 77

88 Chapter 6 Figure 12. Bimolecular isomerization of nc7 into 2MHx, adapted from [18]. The net result of all reactions of this reaction cycle is the same as a simple skeletal isomerization. Blomsma and co-workers [18] stated the following rules for bimolecular isomerization: 1) bimolecular isomerization can only occur if the double bond in the alkene is allocated between two secondary atoms. Thus, only hept-2-ene, hept-3-ene, 2-methylhept-4-ene, 2-methylhept-3-ene, 3-methylhept-4-ene and 2,2-dimethylhept-3-ene are susceptible to this type of isomerization 2) the position of the branching is determined by the double bond in the alkene, i.e. 2-alkenes are branched at C 2, 3-alkenes at C 3, etc. The branched isomers that can be generated are 2MHx, 3MHx, 23DMP, 24DMP and 223 TMB, as shown in Figure

89 Hydroisomerization of heptane: mechanistic aspects and industrial challenges MS 2MHx MS 24DMP ES 3MHx MS 23DMP ES 23DMP MS 223TMB Figure 13. Branching isomerization reactions of C 7 alkenes via dimerization cracking. MS= methyl shift, ES=ethyl shift, adapted from [18] Relation between isomerization and cracking Cracking and isomerization are interrelated and occur through the same mechanisms. The generally accepted theory of acid-catalyzed cracking of paraffins is a carbenium ion mechanism according to which a classical carbenium ion is formed from a hydrocarbon, which then undergoes β-scission to give a carbenium ion of lower carbon number and an olefin [10,13]. However, from an energy point of view, beta scission is rather unlikely in the case of a secondary ion with a straight carbon chain. Experimentally observed characteristics of acidcatalyzed cracking also strongly argue against this beta scission cracking mechanism [14]. The earlier mentioned non-classical carbonium ion, namely the PCP, is the reaction intermediate in acid-catalyzed cracking as well as in isomerization. Furthermore, cracking and isomerization are both catalyzed by similar acid catalysts. Therefore, cracking may occur alongside isomerization. This is experimentally observed for heptane (Table 4). Table 4. Distribution of isomerization products of nc 5, nc 6 and nc 7 over Pt/HMOR [10]. nc 5 isomerization nc 6 isomerization nc 7 isomerization Temperature ( C) Cracked products (wt%) Branched products (wt%) Unconverted alkane (wt%) Branched / linear (%) Equilibrium approach (%)

90 Chapter 6 The relationship between cracking and isomerization is shown in Figure 14. This mechanism suggests that hydroisomerization and hydrocracking may occur as parallel reactions, in addition to sequential ones, i.e. cracking of pre-isomerized molecules and postisomerization of cracked fragments. linear paraffin hydride abstraction/transfer classical carbenium ion non-classical ion hydride transfer classical carbenium ion isomerized products hydride shift carbenium ion H hydride shift scission hydride transfer cracked products Figure 14. Mechanistic relation between hydroisomerization and hydrocracking, adapted from [10,13,15]. Skeleton isomerization requires more than four carbons, whereas cracking requires at least seven atoms in the carbon chain. This means that pentanes and hexanes are in a unique position; they can easily be isomerized, but not easily cracked (Table 4). Some well-known processes for pentane and hexane isomerization are based on these principles i.e. Shell Hysomer process, Shell/Union Carbide TIP and BP Isomerization Process [10]. The competition for isomerization and cracking in bimolecular pathways is shown in Figure

91 Hydroisomerization of heptane: mechanistic aspects and industrial challenges -H2 ic7 -H2 nc7 H slow Beta-scission fast Beta-scission C3-C6 cracked products 2 x ic7 Figure 15. Competition for isomerization and cracking in bimolecular pathways, [18] Recapitulation of reactions via protonated cyclopropane In view of the above-described mechanisms, isomerization via a PCP intermediate is the most likely mechanism that explains experimentally observed distribution of heptane isomers. An overview of possible isomerization and cracking reactions through the PCP mechanism is presented in Figure

92 Chapter 6 (a) nc7 2MHx 3MHx 3MHx 3MHx 22DMP 2MHx 23DMP 23DMP 24DMP 3MHx 33DMP 23DMP 3EP 24DMP 223TMB (b) 23DMP propane, n-butane 22DMP propane, isobutane 24DMP propane, isobutane Figure 16. Mechanism of (a) Isomerization via PCP intermediates; (b) beta-scission of heptane dibranched isomers, adapted from [19]. 82

93 Hydroisomerization of heptane: mechanistic aspects and industrial challenges 3. Catalysts. In principle, two types of catalysts exist for skeletal isomerization of paraffins via carbenium or carbonium ions as intermediates: monofunctional and bifunctional acidic catalysts [10,11]. The latter combines the acidic function with the hydrogenationdehydrogenation function of a metal. The focus in this paper is on bifunctional zeolite catalysts. However, a brief discussion of monofunctional catalysts in section 3.1 is appropriate. A general discussion of bifunctional catalysts is presented in section 3.2 and in section 3.3 bifunctional zeolite-based catalysts in the heptane hydroisomerization process are discussed in detail Monofunctional acidic catalysts The most prominent representative of this type of monofunctional catalysts is aluminum chloride. This acid is used in combination with HCl to give HAlCl 4, in which the Brønsted acid site is active in the isomerization [10]. The HAlCl 4 system has been extensively applied in older industrial processes. It has the disadvantage that replenishment with fresh acid during operation is required to compensate for acid consumption. From both an economic and environmental point of view, processes based on bifunctional catalysts are much more desirable than those based on monofunctional catalysts. The mechanism of isomerization on monofunctional catalysts is shown in Figure 17. acid site Figure 17. Mechanism of n-heptane conversion on a monofunctional acid catalyst Bifunctional acid catalysts Bifunctional catalysts, consisting of a noble or transition metal loaded on an acidic support are used for hydroisomerization and hydrocracking reactions. In contrast to monofunctional catalysts, bifunctional catalysts have a synergy between the metallic and acid sites, leading to an enhanced catalytic activity and higher yields of products with tertiary and quaternary carbon atoms. The main advantage of using bifunctional catalysts is that the 83

94 Chapter 6 process can be operated under stable operating conditions in the presence of sufficiently high hydrogen pressure [10]. The catalyst is utilized in the presence of hydrogen and low octane n-paraffins are converted into branched paraffins that have an improved octane number [3,18,20]. The reaction occurs through the mechanism where alkanes are dehydrogenated on the metal phase and the alkenes formed are protonated at the Brønsted acid sites yielding alkyl carbenium ions. After C-C bond rearrangement and eventually scission, isomeric or fragmented alkylcarbenium ions desorb as products alkenes that are hydrogenated at the metal phase to yield product alkanes [9,10,13-15,18]. The reactions are illustrated in Figure 18. -H 2 Pt Pt dehydrogenation Pt Pt adsorption on acid site H carbenium formation H desorption isomerization to metal site H 2 Pt Pt hydrogenation Figure 18. Acid catalyzed hydroisomerization of n-heptane on bifunctional catalyst. To improve the isomerization reactions different approaches are made [20]: 1) using solid acidic catalysts such as zeolites, the reactions can be carried out at lower temperatures favouring the formation of isomerization products; 2) using hydride transfer agents, cracking can be significantly reduced; and 3) using the structural properties of i.e. zeolites, the diffusion properties can be altered and hence, the yield of isomerized products can be increased. Examples of bifunctional catalysts include chlorinated alumina, silica-alumina and zeolites loaded with platinum or palladium. Zeolite type catalysts are preferable, because of their thermal and chemical stability as well as their structural properties. The use of chlorinated alumina type catalysts is less desirable because of environmental concerns. 84

95 Hydroisomerization of heptane: mechanistic aspects and industrial challenges 3.3. Bifunctional zeolite-supported catalysts Aluminosilicate zeolites have found extensive use as catalysts in crude-oil refining processes [12]. Zeolites have well-defined channels, making them also suitable to use as molecular sieves. It is generally known that the bridging hydroxy groups, linking a silicon with an aluminium atom in the framework, are responsible for the Brønsted acidity and catalytic activity (Figure 19a). Other theories describe the active site as a bifunctional active site comprising a Brønsted acid and one or more Lewis bases. The Brønsted acid being the bridging hydroxy and the Lewis bases, the oxo ligands of the aluminium that is in interaction with the bridging hydroxy (Figure 19b). The interaction of the oxygen atom of the hydroxy group with aluminium, leads to weakening of the hydrogen-oxygen bond, increases the acidity and enhances the basicity of the oxo ligands of aluminium. H O O O O Si Al Si O O O O O O O O Si H O O Al O O Si O O (a) (b) Figure 19. (a) Brønsted acid site; (b) bridging hydroxy and oxo ligand(s) of aluminum acting as bifunctional Brønsted acid and Lewis base sites [12]. Alkane transformation of linear paraffins on platinum containing H-form zeolite catalysts occurs as a result of bifunctional catalysis, previously discussed in section 3.2. In Figure 20 the dehydrogenation of heptane on a Pt-loaded zeolite is schematized and in Figure 21 the isomerization of dehydrogenated heptane is illustrated. H 3 C H C H C C 4 H 9 H H Pt Pt Pt Pt H 2 Figure 20. Dehydrogenation of heptane on a Pt-loaded zeolite. 85

96 Chapter 6 (a) H H H C H H H 3 C C C C 3 H 7 Si O Al O Si H - H (b) H H H C H H H 3 C C C C 3 H 7 Si O Al O Si H - H Figure 21. Skeletal isomerization on a zeolite: (a) methyl shift in 2MHx; (b) branching of 3 heptene, adapted from [12]. The catalytic activity of a zeolite, i.e. catalytic turnover frequency of a site, depends on many parameters including the zeolite structure type, the crystallographic siting, the concentration and distribution of the sites in the framework and the nature of the specific reaction to be catalysed [12]. The sterical factors in the zeolite controlling molecular reaction and motion are denoted as molecular shape selectivity, which is discussed in section

97 Hydroisomerization of heptane: mechanistic aspects and industrial challenges Influence of the zeolite structure on isomerization reactions The porous structure of catalysts has a strong influence on the selectivities in n- heptane isomerization [21]. This will be explained by comparison of a more open and a more restricted structure catalyst. The highest content of multibranched isomers are obtained on open solids, but the thermodynamically predicted value was never reached due to the fact that cracking of multibranched isomers is initiated quickly. Examples of open and restricted structure catalysts are given in Table 5. Table 5. Examples of open (restricted) zeolites [22]. Zeolite Windows Channel dimensions (nm) Open/restricted structure HBeta 12MR 0.55x0.55 Open 0.64x0.76 HEU-1 10MR 0.41x0.57 large side pockets Restricted HZSM-22 10MR 0.41x0.55 Restricted A. Open structure Zeolites with an open structure, e.g. Y and Beta, produce less cracking during C7 hydroisomerization. The methylhexanes, formed from n-heptane are in equilibrium at low conversions. Consequently they are converted into multibranched isomers. Patrigeon et al. [19,22,23] compared several zeolites in the heptane isomerization process and demonstrated that multibranched formation is consecutive to the formation of monobranched in open structure zeolites. The cracked products, which are clearly secondary products of the reaction, appear at medium conversion and increase strongly at high conversion. For open structure catalysts, 2MHx and 3MHx reach equilibrium at low conversions, but when the contact time is increased, they are converted to multibranched isomers, as shown in Figure 22a and 23a. Chao et al. [24,25] stated that H-Beta should be promising for the isomerization of light n- paraffins such as heptane. The more open structure of H-Beta presumably plays a key role in its higher activity and selectivity. 87

98 Chapter (a) nc7 MB MuB CP MB Yields (w%) MuB 10 CP nc7 conversion (w%) Yields (w%) (b) nc7 MB MuB CP MB CP 10 MuB nc7 conversion (w%) Figure 22. Yields of monobranched (MB), multibranched (MuB) and cracked products (CP) in nc7 conversion on (a) Pt/HBeta at 210 C and (b) Pt/HEU-1 at 220 C [19,22,23]. B. Restricted structure Zeolites with a more restricted structure, e.g. ZSM-22, result in more cracking products. Presumably 2MHx is formed relatively easier due to sterical constraints. Next this molecule can be converted into 3MHx to reach equilibrium or into multibranched isomers. In very restricted zeolites with narrow channels the first branching occurs at the pore mouths, but the second branching to form multibranched isomers is suppressed due to sterical constraints. This results in a high selectivity for monobranched isomers. In restricted catalysts two types of rearrangements for the 2MHx are in competition: 1) the formation of 3MHx and 2) the formation of multibranched isomers. Figure 22b and 23b illustrate this. 88

99 Hydroisomerization of heptane: mechanistic aspects and industrial challenges (a) nc7 2MHx equilibrium 3MHx equilibrium 3EP multibranched isomers (b) nc7 2MHx 3MHx 3EP multibranched isomers Figure 23: Reaction routes of nc7 conversion on (a) open structure and (b) restricted structure catalysts [23]. Theories about the influence of the molecular shape of the catalyst on the selectivity can be classified into [21]: 1) pore mouth catalysis, 2) transition state shape selectivity and 3) product shape selectivity. In the theory of pore mouth catalysis (reactant selectivity, see Figure 24a) the terminal branching occurs entirely at the pore mouths, because the required transition state does not fit inside a too small channel. Since this model postulates that none of the branched paraffins can fully enter too small pores, it has to postulate that subsequent hydroisomerization reactions to form di- and tri-branched paraffins occur at the pore mouth or external surface. The transition shape selectivity suggests that the transition state required for terminal methyl groups is better able to fit inside smaller channels than the transition state for internal methyl groups, as shown in Figure 24c. The product reactant selectivity (see Figure 89

100 Chapter 6 24b) suggests a higher desorption rate for paraffins with terminal methyl groups than for paraffins with internal methyl groups. The former therefore have shorter residence times and are less susceptible to consecutive reactions. Figure 24. Shape selectivity concepts for n-heptane conversion: reactant (top), product shape (middle) and transition state (bottom) selectivity [21]. Table 6. Comparison of molecular sieve dimensions in heptane hydroisomerization [21]. Structure type Max. diameter (nm) Max. monobranched C7 yield (w%) Max. dibranched C7 yield (w%) BEA TON MTT AEL Heptane hydroisomerization studies performed by Maesen et al. [21] suggest that catalysts with narrow channels have lower dibranched isomers yields compared to the wider channel catalysts, see Table 6. The works of different authors [21,26] suggest that in the case of narrow channels, i.e. H-ZSM-22, methylbranching does not occur in the channels but at the pore mouths. 90

101 Hydroisomerization of heptane: mechanistic aspects and industrial challenges Effect of temperature, metal loading and acidity Hydroisomerization of heptane is usually carried out on bifunctional zeolites at temperatures starting from approximately 170 C and upwards. The effect of the reaction temperature on the isomerization yield over a Pt/H-Beta catalyst studied by Wang and coworker [27] is shown in Figure Selectivity Conversion Conv., sel. and yield (C%) Yield Reaction temperature ( 0 C) Figure 25. Effect of reaction temperature on n-heptane isomerization over Pt/HBeta catalyst [27]. The conversion of n-heptane increases with reaction temperature, while the selectivity to isomers decreases. The total yield of all heptane isomers passes through a maximum at 210 o C. This indicates the consumption of isomerized products in consecutive cracking reactions. The main cracking products were propane and butane. Similar results were reported by other authors [19-25,27-29]. Chica et al. [4] and Patrigeon et al. [22] found that the acidity seems to be a less important parameter in C7 isomerization than porosity. Different results were obtained depending on the solid structure. Wang and co-workers [27] found that the higher the content of Pt and the lower the acidity of the zeolite beta, the higher the selectivity to isomers. Based on experimental results with zeolite beta catalysts, Chao et al. [25] stated that the transformation route of n-heptane over Pt/HBeta catalyst could be represented as in Figure

102 Chapter 6 nc 7 H 2 C 7 - H MHx H 2 - H 2MHx, 3MHx (MB) C 3 C 4 H 2 - H C 3 C 4 (CP) DMP - H H 2 22DMP, 23DMP, 24DMP (MuB) C 7 = C 3 C 4 or C 3 C 4 H 2 - H C 3 C 4 (CP) 33DMP or 223TMB H 2 - H 33DMP, 223TMB (MuB) C 14 H 2 - H C 3 C 5 C 6 or C 6 2C 4 C 4 2C 5 (CP) Figure 26. Monomolecular and bimolecular reaction routes of hydrocracking and hydroisomerization of n- heptane via a bifunctional mechanism [25]. Blomsma et. al. [18] presented similar results and stated that heptane conversion on zeolite beta catalysts promoted with Pd occurs through a combination of monomolecular and bimolecular reaction mechanisms. An important feature of the bimolecular mechanism is that it generates in a selective manner isomers. The bimolecular mechanism produces 2MHx, 3MHx, 23DMP and 24DMP in large quantities, suppressing the formation of 22DMP and 3EP. Blomsma et. al. [18] further stated that the contribution of each of the two mechanisms on zeolite beta is deduced from the formation of specific iso-heptanes. It depends on the metal content and the acidity of the zeolite. The monomolecular mechanism predominates at high metal loadings and the bimolecular mechanism is active especially at low metal loadings of the catalyst. 4. The oil industry The heptane isomerized product can be produced as an octane enhancing compound for the gasoline pool. Therefore, it is appropriate to highlight the most important aspects of the oil industry, accounting for the most recent developments and trends, presented in section 4.1. In section 4.2 an appraisal of heptane isomers is presented and section 4.3 summarizes the major challenges encountered in the heptane hydroisomerization process. 92

103 Hydroisomerization of heptane: mechanistic aspects and industrial challenges 4.1. Recent developments and trends The oil industry is a capital-intensive industry. One of the most crucial drivers is technology, which drives: 1) the ability to perform activities more efficiently in terms of cost reduction, 2) the opportunity to extract more from existing oil fields and 3) the ability to develop alternative sources of energy. The high fixed costs and the low oil price were the main drivers behind the most recent consolidation in the oil industry. The industry landscape changed drastically due to some mega mergers, namely the mergers between Exxon and Mobil in 1999 [30], British Petroleum and Amoco in 1998 and Arco in 2000 [31], Petrofina and Elf in 2000 [32] and Chevron and Texaco in 2001 [33]. These mergers enabled the companies to be more vertically integrated. In the perspective of the heptane isomers as a product, the oil industry is best analyzed using a gasoline supply and demand framework. The demand for heptane isomers is directly correlated to the number of combustion engines that require gasoline with a minimum octane number. These types of gasoline can be divided into regular unleaded (RON 91), premium unleaded (RON 95) and super unleaded (RON 98). The shift in expected supply and demand in Europe is shown in Figure 27. Total refined product consumption has increased over the years and it will continue to increase in the future. However, the gasoline demand in the European Union 15 countries (EU15) has been slowly declining since 1992 and projections show continued decline in demand through to The decline in demand of gasoline is mainly the result of the commitment by car manufacturers to meet reduced CO 2 emission levels of 140 grams CO 2 /km in the new car fleet sold in 2008 [1,34]. This commitment is expected to result in car manufacturers offering higher efficiency gasoline engines (using gasoline direct injection) and the encouragement of the use of more efficient diesel engine cars instead of gasoline powered cars [1]. Alternative engine technology, e.g. hybrid vehicles using both internal combustion engines and electricity, are already available from Toyota [35] and Honda [36] in certain markets and prototype fuel cell vehicles are also on the road [2,37]. 93

104 Chapter 6 Figure 27. (a) Gasoline supply for total Europe and (b) gasoline demand for Europe Union 15 countries. Data obtained from [1]. ( ) Regular unleaded RON 91, ( ) Premium unleaded RON 95, ( ) Premium leaded, ( ) Super unleaded RON Appraisal of the heptane isomerized product An octane enhancing heptane isomerized product is not yet sold on the market as part of the gasoline pool. Octane-enhancing aromatics and olefins typically have RON values of The heptane hydroisomerization product could have a RON of Since there is a relation between the RON and price, as shown in Figure 28, this indicates a price range of 94

105 Hydroisomerization of heptane: mechanistic aspects and industrial challenges US dollar/ton. On an average octane ton price of 2.00 US dollar [38] a heptane isomerized product with RON values between 92-99, would have a price in the range of US dollar/ton. This is based on the assumption that supply and demand of gasoline do not change drastically and currency exchange fluctuations are ignored. 260 MTBE 240 super unleaded Price [US$/ton] naphta premium unleaded 180 regular unleaded RON Figure 28. Relation between market price and RON. Data obtained from [1] Challenges in the heptane hydroisomerization process Major oil companies such as Shell and British Petroleum currently employ hydroisomerization processes of C 5 /C 6. Since branched heptane isomers are extremely promising in producing an octane-enhancing product that is less harmful to the environment, research and development on this topic is in full progress. A conceptual process on heptane hydroisomerization was published recently [39,40] and Institut Francais du Petrole already patented a hydroisomerization process for higher alkanes [41]. The first heptane isomerization process will most likely be operational when the bottlenecks in the process development are solved. These include: 1) selectivity of the catalyst to produce high octane multibranched isomers, 2) the separation of mono- from multibranched isomers and 3) controlling hydroisomerization versus hydrocracking. 95

106 Chapter 6 5. Concluding remarks Acid-catalyzed isomerization of heptane can occur through a monomolecular or a bimolecular mechanism. Since the bimolecular mechanism does not account for all reaction products it is concluded that the acid-catalyzed isomerization of heptane occurs mainly through a monomolecular mechanism with carbenium ions and a protonated cyclopropane ion as reaction intermediates. A variety of catalysts can be used in the hydroisomerization of heptane, but it is believed that bifunctional zeolites possess the most promising features to be used in this process. Bifunctional catalysts are preferred above monofunctional catalysts, because of the synergy between the acid sites and the hydrogenation-dehydrogenation step on the metal sites. With zeolites, the acidity can be controlled and the structural properties can also be altered to either promote isomerization or cracking. The highest yield of multibranched isomers was found on the more open structure zeolites such as HBeta, but the thermodynamic equilibrium was never reached due to cracking. In more restricted structures such as HZSM-22, a higher selectivity for monobranched isomers was measured. Experimental evidence showed that acidity seems to be a less important parameter in C 7 isomerization than porosity. The demand for branched heptane isomers is expected to increase due to more stringent environmental regulations. The first heptane isomerization process will most likely be operational when the bottlenecks in the process development are solved. These include: 1) the selectivity of the catalyst to produce multibranched isomers, 2) the separation of monofrom multibranched isomers and 3) controlling hydroisomerization versus hydrocracking. Acknowledgement Dr. U. Hanefeld is gratefully acknowledged for his contribution. References [1] C.H. Birch, R. Ulivieri, Uls gasoline and diesel refining study. Prepared for European Commission Directorate-General Env. 3. Purvin & Gertz Inc. (November 2000). (accessed April 2003) [2] R. Barber, K. Carabell, Motor gasolines - technical review. Chevron Products Company (1996). (accessed April 2003) [3] T.G. Kaufmann, A. Kaldor, G.F. Stuntz, M.C. Kerby, L.L. Ansell, Catal. Today 62 (2000) 77. [4] A. Chica, A. Corma, P.J. Miguel, Catal. Today 65 (2001) 101. [5] W.G. Lovell, Ind. Eng. Chem. Res. 40 (1948)

107 Hydroisomerization of heptane: mechanistic aspects and industrial challenges [6] F.E. Condon, in: Catalysis, Vol. 6, ed. P.H. Emmett (Reinhold, New York, 1958) ch.5, pp [7] D.M. Brouwer, Recl. Trav. Chim. Pays Bas 87 (1968) 210. [8] D.M. Brouwer, J.M. Oelderik, Recl. Trav. Chim. Pays Bas 87 (1968) 721. [9] J. Weitkamp, Ind. Eng. Chem. Res. 21 (1982) 550. [10] S.T. Sie, in: Handbook of Heterogeneous Catalysis, Vol. 4, eds. G. Ertl, H. Knözinger, J. Weitkamp (VCH Verlags Gesellschaft mbh, Weinheim, 1997) ch. 3, pp [11] D.S. Santilli, B.C. Gates, in: Handbook of Heterogeneous Catalysis, Vol. 3, eds. G. Ertl, H. Knözinger, J. Weitkamp (VCH Verlags Gesellschaft mbh, Weinheim, 1997) ch. 5, pp [12] J.A. Martens, P.A. Jacobs, in: Handbook of Heterogeneous Catalysis, Vol. 3, eds. G. Ertl, H. Knözinger, J. Weitkamp (VCH Verlags Gesellschaft mbh, Weinheim, 1997) ch. 5, pp [13] S.T. Sie, Ind. Eng. Chem. Res. 31 (1992) 1881 [14] S.T. Sie, Ind. Eng. Chem. Res. 32 (1993) 397. [15] S.T. Sie, Ind. Eng. Chem. Res. 32 (1993) 403. [16] J.L. Franklin, in: Carbonium Ions, Vol. 1, eds. G.A. Olah, P. von R. Schleyer, (Interscience, New York, 1968) pp. 85. [17] D.M. Brouwer, Recl. Trav. Chim. Pays Bas 87 (1968) [18] E. Blomsma, J.A. Martens, P.A. Jacobs, J. Catal. 159 (1996) 323. [19] P. Raybaud, A. Patrigeon, H. Toulhoat, J. Catal. 197 (2001) 98. [20] G. Kinger, D. Majda, H. Vinek, Appl. Catal. A Gen. 225 (2002) 301. [21] Th.L.M. Maesen, M. Schenk, T.J.H. Vlugt, J.P. de Jonge, B. Smit, J. Catal. 188 (1999) 403. [22] A. Patrigeon, E. Benazzi, Ch. Travers, J.Y. Bernhard, Catal. Today 65 (2001), 149. [23] A. Patrigeon, E. Benazzi, Ch. Travers, J.Y. Bernhard, Stud. Surf. Sci. Catal. 130 (2000) [24] K.J. Chao, H.C. Wu, L.J. Leu, Appl. Catal. A Gen. 143 (1996) 223. [25] K.J. Chao, C.C. Lin, C.H. Lin, H.C. Wu, C.W. Tseng, S.H. Chen, Appl. Catal. A Gen. 203 (2000) 211. [26] E.B. Webb, G.S. Grest, M. Mondello, J. Phys. Chem. B 103 (1999) [27] Z.B. Wang, A. Kamo, T. Yoneda, T. Komatsu, T. Yashima, Appl. Catal. A Gen. 159 (1997) 119. [28] F.Y.A. El-Kady, M.F. Menoufy, H.A. Hassan, Indian J. Technol. 21 (1983) 213. [29] A. Corma, J. Planelles, F. Tomas, J. Catal. 94 (1985) 445. [30] Exxon Mobil Corporation, website (accessed May 2003) [31] British Petroleum, website (accessed May 2003) 97

108 Chapter 6 [32] Totalfinaelf, website (accessed May 2003) [33] ChevronTexaco, website (accessed May 2003) [34] S. Smith, The future of energy markets. BP Statistical Review of World Energy June (accessed May 2003) [35] Toyota Motor Corporation, website (accessed May 2003) [36] Honda Motor Company, website (accessed May 2003) [37] Royal Dutch/Shell, website (accessed May 2003) [38] A.I. Lugovskoi, S.A. Loginov, V.A. Sysoev, S.A. Makeev, A.N. Shakun, M.L. Federova, Chem. Tech. Fuels Oils 36 (2000) 330. [39] M.L. Maloncy, L. Gora, J.C. Jansen, Th. Maschmeyer, Ars Separatoria Acta 2 (2003) 18. [40] M.L. Maloncy, Th. Maschmeyer, J.C. Jansen, Chem. Eng. J. 106 (2005) 187. [41] G. Hotier, O. Clause, S. Jullian, K. Ragil, J.P. Durand, US Patent 6,338,791, January 15,

109 7 Technical and economical evaluation of a zeolite membrane based heptane hydroisomerization process Abstract An industrial scale heptane hydroisomerization process was simulated based on a concept of two reactors and a zeolite membrane. A product stream containing tribranched, and part of the dibranched C 7 isomers with octane number up to 92 is predicted. The economics of the process shows an investment cost of 40 million euros, with the membrane unit as the main cost driver. The technical and economical feasibility of this industrial scale heptane hydroisomerization process depends mainly on further development and performance of zeolite membranes. This chapter is based on: M.L. Maloncy, Th. Maschmeyer, J.C. Jansen, Chem. Eng. J. 106 (2005) 187

110 Chapter 7 1. Introduction Regulations to minimise the adverse impact of automotive fuel combustion on the environment have resulted in the need for changes in the automotive fuel composition worldwide. The challenge faced by refiners is to produce environmentally friendly gasoline with sufficiently high research octane number (RON). In the oil industry C 5 and C 6 paraffins are typically used in hydroisomerization units to obtain high octane number components. Paraffins larger than C 6, such as heptane are usually present in catalytic reforming feed streams and converted into aromatic compounds. Since regulation aims to reduce aromatic components in gasoline an alternative for the use of the higher alkanes is hydroisomerization. Besides the isomerization reaction the separation of high-octane value isomers from lower ones is also of great importance in the hydroisomerization process. There are numerous experimental efforts published on heptane isomerization, however, on heptane isomer separation literature data is scarce. Currently no commercially operating process for the hydroisomerization of heptane exists. In this work a preliminary design of a heptane hydroisomerization process using zeolite membrane is shown, aiming at the production of high octane number heptane isomers. Besides the technical aspects of the process, involving mainly the reaction and the membrane separation sections, the economics of the process is also evaluated. If essential data is missing in the early stage of process evaluation, reasonable assumptions have to be made, which of course have to be confirmed by experiments if the evaluation is favourable. Early feasibility studies, based on literature data, preliminary experiments or reasonable assumptions, are essential to identify the key parameters and to give guidance towards promising technologies [1]. 2. Process concept Noble metal supported zeolite type catalysts are often used for n-heptane hydroisomerization. The selectivity towards heptane isomers using these catalysts depends on various factors such as; acid/metal site ratio [2] zeolite structure and zeolite acid strength [3,4]. If the catalyst is too acidic, cracking of mainly multibranched isomers is enhanced. On the other hand if the catalyst is a too weak acid the isomerization may not proceed appropriately. For the hydroisomerization of n-heptane we propose a concept that uses two reactors, each with a different type of catalyst, and a separation unit composed of a zeolite membrane (see Figure 1). 100

111 Technical and economical evaluation of a zeolite membrane Lights (< C 7 ) Feed Primary Separations multibranched-c 7 rich stream Zeolite Membrane Reactor2 < 0.62 nm Product > 0.62 nm (2,2- & 3,3-DMP; 2,2,3-TMB) linear & mono-c 7 rich stream Reactor1 Heavies (> C 7 ) Figure 1. Heptane hydroisomerization block scheme. The first reactor (Reactor1) contains a strong acidic bifunctional type catalyst and aims to convert sequentially n-heptane to mono- and dibranched isomers. The second reactor (Reactor2) is used to convert preferably 2,4-dimethylpentane into 2,2,3-trimethylbutane. According to the protonated cyclopropane (PCP) mechanism [5] 2,4-dimethylpentane is the main source for the formation of 2,2,3-trimethylbutane, the isomer with the highest RON (see Table 1). The catalyst used must be a moderately to weakly acidic bifunctional type so that cracking of dibranched components is not enhanced while the formation of 2,2,3-TMB is promoted. Examples of catalysts for the first and second reactor are Platinum loaded Hydrogen-Beta zeolite (Pt/H-BEA) type, which can have relatively high conversion and selectivity towards isomerization and less cracking [3], and Nickel loaded on high area amorphous silica-alumina (Ni/ASA) type [9], respectively. Table 1. Kinetic diameter, Boiling point and RON of heptane isomers. b Component Kinetic diameter Boiling point (K) RON (nm) a n-heptane (n-c 7 ) methylhexane (2-MHx) methylhexane (3-MHx) ethylpentane (3-EP) ,3-dimethylpentane (2,3-DMP) ,4-dimethylpentane (2,4-DMP) ,2-dimethylpentane (2,2-DMP) ,3-dimethylpentane (3,3-DMP) ,2,3-trimethylbutane (2,2,3-TMB) a [6,7], b [8] 101

112 Chapter 7 The separation unit is used to separate the final product of the process composed of the dibranched 2,2-DMP, 3,3-DMP and tribranched 2,2,3-TMB, from smaller components, especially 2,4-DMP that is sent to the second reactor. Separation by distillation is rather difficult as can be seen from the boiling point data in Table 1. A zeolite membrane with a channel aperture of approximately 0.55 nm [10] can be applied for this kind of separation. Assuming absolute separation within a perfect zeolite membrane, only the small molecules can permeate through the channel aperture. Thus, the di- and tribranched isomer products with their diameter of 0.62 nm will not enter the channels, while smaller molecules will permeate through the pores. Note that the values of the kinetic diameters given in Table 1 should be viewed as qualitative indications, since the concept of the kinetic diameter is based on simplified spherical representation and rigidity of a molecule or of a framework [11]. 3. Process description and simulation A process simulation was performed based on the proposed concept of two reactors and a separation unit. Figure 1 shows the process block scheme. The process flow scheme has been previously presented [12,13] and is shown in Figure 2. As feed to the process hydrogen was used as well as a simplified industrial naphtha feed with a RON of 57. The hydrocarbon feed contained C 6, C 7 and C 8 linear and branched alkanes. The C 7 components comprised about 40 wt% of the feed. The throughput of the feedstock was assumed 907 metric ton per day, comparable to that of existing C 5 /C 6 isomerization processes (between 600 and 1200 metric ton per day). The distillation train shown in Figure 2 is composed of three distillation columns. The purpose of the first two columns is to separate the C 7 fraction from the heavier and lighter hydrocarbons, while the third column is used to separate the multibranched from the linear and monobranched C 7 components. The recycle stream is sent via a hydrogen separation membrane to the second distillation column. 102

113 Technical and economical evaluation of a zeolite membrane 103

114 Chapter 7 Within an overall refinery process, the lighter and heavier hydrocarbons separated in the distillation columns, could be used in state of the art C 5 /C 6 hydroisomerization process and the reforming process respectively. Experimental data from [3] and [9] were used as a basis to obtain kinetic data. For hydroisomerization a first order reaction rate was assumed. The cracking reactions were assumed to be first order in reactant and zero order in hydrogen. The reaction models used are simple and do not increase simulation complexity. For the membrane simulation a countercurrent pervaporation model was used. By using this model it is assumed that the individual components do not influence each other s fluxes. Figure 3 shows a schematic view of the membrane unit. Permeance data were estimated from experiments performed in our laboratory [14] and also from the experimental work of Flanders et al. [15]. 2,3-DMP and 2,4-DMP were considered the slowest permeating components and the recovery of these species was set greater than 98% at the membrane outlet. Gas out Liquid in Gas in Liquid out Figure 3. Schematic view of the membrane unit. The reactor and membrane operation conditions set for process simulation are specified in Table 2. The process simulation was performed using Aspen and Excel, and different membrane design configurations were calculated. Table 2. Reactor and membrane operation conditions. Reactor 1 Reactor 2 Membrane Reactor/membrane model Plug flow Plug flow Countercurrent Plug flow Adiabatic Adiabatic Temperature (K) Pressure (x 10 5 Pa) at feed side 1 at permeate side H 2 /Hydrocarbon molar ratio

115 Technical and economical evaluation of a zeolite membrane 4. Results and Discussion The focus of the results and discussion presented below will be mainly on the reactors, the product separation membrane and in lesser extend on the hydrogen separation membrane. The economics of the process will be highlighted as well Reactor design Assuming the PCP mechanism for n-c 7 isomerization, the following reactions are mechanistically possible: n-c 7 2-MHx (1) n-c 7 3-MHx (2) 2-MHx 3-MHx (3) 2-MHx 2,2-DMP (4) 2-MHx 2,3-DMP (5) 2-MHx 2,4-DMP (6) 3-MHx 2,3-DMP (7) 3-MHx 3,3-DMP (8) 3-MHx 3-EP (9) 2,2-DMP 2,3-DMP (10) 2,3-DMP 2,4-DMP (11) 2,3-DMP 3,3-DMP (12) 2,4-DMP 2,2,3-TMB (13) For the reactions (1)-(13) kinetic equations can be derived. For a reaction A B this equation holds: k j rj = kj[ A] - [ B] (14) K j K j can be obtained from equilibrium compositions at the studied temperature. First order reaction rates were assumed. In literature first order reaction rate for hydroisomerization of heptane is described in the work of El Kady et al. [16]. Using a plug-flow reactor for process simulation, Arrhenius parameters and activation energy data were used as input: k k e (15) j EA / RT = 0, j In Table 3 the Arrhenius parameters and activation energies of the forward and backward isomerization reactions used in the simulation are given. 105

116 Chapter 7 Table 3. Arrhenius parameters k 0 and activating energies E A for forward (j) and backward (-j) isomerization reactions. Reaction k 0 (x s -1 ) k 0 (x 10 8 s -1 ) E A (kj mol -1 ) (Reactor1) (Reactor2) k 0,j k 0,-j k 0,j k 0,-j E A,j E A,-j (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) Besides isomerization reactions cracking reactions were taken also into account, with mainly propane and isobutane as the cracking products. It is found in literature that multibranched isomers are preferably cracked and that hydroisomerization and cracking occurs in consecutive steps as shown in the reaction scheme below [17]: n-alkane MB MTB CR (16) MB stands for single branched, MTB for multibranched and CR for cracked alkanes. The cracking reactions where assumed to be first order in reactant. The rate of cracking for a component A is given by: r [ ] C = kc A (17) The cracking coefficient k C is assumed to be only a function of the degree of branching: k C, n-c7 < k C, MB < k C, MTB (18) This assumption is valid for the first reactor that uses a stronger acidic catalyst with a relatively high cracking ability. As input for the reactor simulation, the activation energy and the Arrenhius parameters were used: k C = k C0 e E / RT A 106

117 Technical and economical evaluation of a zeolite membrane The activation energy used for the cracking reactions was 175 kj mol -1. This value was estimated from experimental data [3] and roughly assumed to be equal for all cracking species. The Arrhenius parameters of the different isomer groups are given in Table 4 for the first reactor. Because of the moderate acid catalyst used the cracking ability of the second reactor should be lower, using therefore the Arrhenius parameter of 1.56 x s -1 for all reactants, excluding n-c 7 (k C0 = 0 s -1 ). Zero order in hydrogen was assumed in the reactions. Table 4. Arrhenius parameters used for cracking reaction (k C0 ) in the first reactor. Reactant k C0 (x s -1 ) linear n-c monobranched 7.56 dibranched tribranched Simulations using the reaction models within a plug flow reactor consideration resulted in the product distribution of the reactors effluent given in Table 5. The total amount of feed to the first and second reactor was 1813 and 1033 metric ton per day, respectively. The estimated amount of catalyst and the reactors dimensions together with other reactor specifications are given in Table 6. Because of their large scale and ease of design and operability, both reactors used in the process, are fixed bed reactors. The short residence time distribution of these reactors favours a high selectivity. Both reactors are operated adiabatic and in gas phase. Table 5. Reactor product distribution (%wt). Reactor 1 Reactor 2 Feed Product Feed Product H Propane Isobutane n-c MHx MHx ,2DMP ,3DMP ,4DMP ,3DMP EP ,2,3TMB Others

118 Chapter 7 Table 6. Reactor specifications. Reactor 1 Reactor 2 Reactor type Fixed bed Fixed bed Adiabatic Adiabatic Temperature (K) Pressure (x 10 5 Pa) Pressure drop (x 10 5 Pa m -1 ) Length (m) Diameter (m) Catalyst Pt/H-BEA (0.5 wt% Pt) Ni/ASA (5 wt% Ni) Catalyst mass (metric ton) Dp (m) (sphere) (sphere) 4.2. Membrane design A countercurrent membrane pervaporation model is used for the membrane design. In pervaporation, the driving force is the difference in partial pressure on the gas side and the vapour pressure on the liquid side of the membrane. The flux in pervaporation can be calculated using the driving force and the permeance value of the specific component for the specific membrane. By using this model it is assumed that the individual components do not influence each other s fluxes. sat ( γ ) J = x P yp (19) i i i i i i The mass balances can be specified using the countercurrent plug-flow model (Figure 3). Both the composition and the total flow of both streams change and have to be integrated over the length of the membrane. Component balance for the liquid stream: ( ) d F x ( γ ) =Π x P y P a (20) L i sat i i i i i M dz Component balance for the gas stream: ( ) d F y ( γ ) =Π x P yp a (21) G i sat i i i i i M dz 108

119 Technical and economical evaluation of a zeolite membrane And the total mass balance: df dz df sat ( γ ) (22) L G = = Πi xi ipi yp i am dz i Evaluation of the mass balance equations yields the required membrane area (a M ). To evaluate these equations permeance data for the components is required. Vapour permeation experiments performed in our laboratory with C 7 mixtures gave at 353 K permeance values of 4.0 x mol m -2 s -1 Pa -1 for 2,4-DMP and 4.4 x 10-8 mol m -2 s -1 Pa -1 for linear C 7. At a temperature of 393 K the dibranched permeance was 1.6 x 10-9 mol m -2 s -1 Pa -1 and the linear C 7 permeance was 7.7 x 10-8 mol m -2 s -1 Pa -1 [14]. An increase of permeance with increasing temperature was observed for the double branched and the linear heptane. A silicalite-1 membrane with a thickness of about 16 µm was used. Literature data on the separation of heptane isomers is scarce making it difficult to compare our data with literature. However, when comparing with literature data of C 6, the permeances are in the same order of magnitude [15]. Flanders et al. [15] compared pervaporation with vapour permeation and their results showed that the double-branched permeance for pervaporation was an order of magnitude higher than vapour permeation. The authors also observed relative constant permeance behaviour of the dibranched components with increased driving force. If we extend those finding to our work, the permeance of the 2,4-DMP would be in the order of 10-8 applying pervaporation. So, in the equations permeance of 1 x 10-8 mol m -2 s -1 Pa -1 for the dibranched components was used. For the boundary conditions at the inlet, the process streams coming from the distillation column and the hydrogen separation membrane were used (see Figure 2). Mainly multibranched components were considered with 2,3-DMP and 2,4-DMP as the slowest permeating species. The boundary condition set at the outlet was that the recovery of the slowest permeating species should not be greater than 98%. Vapour pressures of 8.0 x 10 5 and 9.8 x 10 5 Pa for 2,3-DMP and 2,4-DMP were used, respectively, at the operation temperature of 458 K. The solution of the equations yields a membrane area of m 2. Meindersma and de Haan [18] in their work on aromatic compounds separation using zeolite membranes estimated membrane areas ranging from m 2 to m 2 depending on the purity requirement. A feed of 300 metric ton per hour was used in their work, which is about six times higher than the amount of feed sent to the membrane unit of the present work. The estimated membrane areas given above are currently not produced commercially for zeolite membrane units. The first large-scale pervaporation plant using zeolite membrane put into industrial operation has a total membrane area of about 60 m 2 [19]. The zeolite that 109

120 Chapter 7 is used is the NaA-type zeolite. The feed flow to the pervaporation plant is 0.48 metric ton per hour with a water content of 10 wt%. Thus, the technical feasibility of the proposed n- heptane process in this work is restricted by further development of zeolite membranes on a larger scale. For the calculation of the dimensions of the membrane unit different geometries can be used to apply membrane technology in industrial practice. The most straightforward solution is the shell and tube configuration (Figure 4a). It is similar to a countercurrent heat exchanger but here mass is exchanged. One of the disadvantages is the large internal volume of the unit (low membrane surface to unit volume ratio). Monolith structures, however, have a very high surface to volume ratio. They consist of a large number of parallel square channels with sizes ranging from 10 to 100 cells per square inch. Research of MFI type membranes on monolith used to separate light alkanes is already in development [20-22]. Although in theory these structures can be applied in countercurrent mode (Figure 4b) it will have practical difficulties to connect all individual channels to the corresponding process streams. A feasible solution is the use of cross-flow monolith structures [23,24] shown in Figure 4c. In this structure cross flow can be established because each layer of channels is rotated by 90. By interconnecting multiple units one can approach countercurrent operation. The dimensions of the shell and tube configuration are calculated as follows (liquid flow inside the tubes): a M n = (23) π dl al π 4 2 = d n (24) 3 π = (25) ag s n d n ( ) V = a a L (26) L G For both monolith configurations the equations are: a M n = (27) β dl 1 = = (28) 2 2 al ag d n ( ) V = a a L (29) L G 110

121 Technical and economical evaluation of a zeolite membrane Figure 4. Different membrane configurations: (a) Shell and tube membrane unit (top view); (b) Countercurrent monolith; (c) Cross-flow monolith. The dark areas indicate liquid flow inside the tube/channels. 111