t 0000019

+\GURJHQVHOHFWLYH6LOLFD

0HPEUDQHVIRU8VHLQ

0HPEUDQH6WHDP5HIRUPLQJ

$ULDQ1LMPHLMHU

Nijmeijer, Arian

Hydrogen-Selective Silica Membranes for Use in Membrane Steam Reforming

Thesis University of Twente, Enschede – With ref. – With summary in Dutch

ISBN 90-36513863

Copyright © 1999 by A. Nijmeijer, The Netherlands

Printing and binding by Printpartners Ipskamp, Enschede

Cover illustration: J.R. Rostrup-Nielsen, “Catalytic Steam Reforming”, Springer Verlag, Berlin (1984).

+<'52*(16(/(&7,9(6,/,&$0(0%5$1(6)2586(,1

0(0%5$1(67($05()250,1*

352()6&+5,)7

WHUYHUNULMJLQJYDQ

GHJUDDGYDQGRFWRUDDQGH8QLYHUVLWHLW7ZHQWH

RSJH]DJYDQGHUHFWRUPDJQLILFXV

SURIGU)$YDQ9XJKW

YROJHQVEHVOXLWYDQKHW&ROOHJHYRRU3URPRWLHV

LQKHWRSHQEDDUWHYHUGHGLJHQ

RSYULMGDJGHFHPEHUWHXXU

GRRU

$ULDQ1LMPHLMHU

JHERUHQRSPDDUW

WH(VSHO

'LWSURHIVFKULIWLVJRHGJHNHXUGGRRUGHSURPRWRU

SURIGULU+9HUZHLM

Here we stand

Like an Adam and an Eve

Waterfalls

The Garden of Eden

Two fools in love

So beautiful and strong

The birds in the trees

Are smiling upon them

From the age of the dinosaurs

Cars have run on gasoline

Where, where have they gone?

Now, it’s nothing but flowers

(Nothing but) flowers, David Byrne, The Talking Heads

The work described in this thesis was financially supported by the EU in the framework of theBrite-Euram program.

&RQWHQWVChapter 1

Introduction

1

Chapter 2

The process technology of (membrane) steam reforming

15

Chapter 3

Colloidal processing of ceramic membrane supports. General introduction

37

Chapter 4

Colloidal processing of ceramic membrane supports. The experimental part

53

Chapter 5

The preparation and properties of hydrothermally stable γ-alumina membranes

69

Chapter 6

Preparation, characterisation and properties of microporous silica membranes

85

Chapter 7

Low temperature CVI modification of γ-alumina membranes

105

Chapter 8,

The thermal dehydrogenation of H2S in a membrane reactor

115

Chapter 9

Evaluation and recommendations

127

&KDSWHU

,QWURGXFWLRQ

1. Concepts for membrane reactors

With the development of new, highly selective ceramic membranes the possibility of using

them in high temperature membrane reactors came into scope. Hence, a large amount of re-

search is currently being done on a world-wide scale on these reactors. An extensive review of

current research on membrane reactors was given in a recent paper by Bredesen [1] and by

Saracco et al. In somewhat older work, Saracco and Specchia [3] and Zaman and Chakma [4]

provided good reviews on the use of membrane reactors as well. Saracco and Specchia fo-

cused on catalytic membrane reactors while Zaman and Chakma paid more attention to mem-

brane synthesis. Armor [5] gave a good review on membrane catalysis and defined the current

problems. In this chapter an overview is given on the general ideas of membrane reactors and

types of reactions under investigation in the world.

Membrane reactors can be used to shift the equilibrium in thermodynamically limited reac-

tions. Several types of membrane reactors are currently under investigation, especially for de-

hydrogenation reactions such as the dehydrogenation of propane to propene [6] or of ethyl-

benzene to styrene [7]. Also the dehydrogenation of H2S has been studied in membrane reac-

tors [8,9].

Reforming reactions have been studied in membrane reactors as well. Most well-known is the

steam-reforming of various hydrocarbons [10-13], especially methane steam-reforming which

is the major source of hydrogen in the world [14]. Some research has been performed on CO2

reforming of methane [15] and also a considerable amount of effort has been put in perform-

ing the water-gas shift reaction in a membrane reactor [16,18].

In all of the above reactions, the selective removal of hydrogen from the reaction zone is the

main role of the membrane, hence, such membranes must be hydrogen selective. Several types

Chapter 12

of membranes can be used and their advantages and disadvantages will be discussed in section

2 of this chapter. In general, problems with the existing membrane concepts are either an in-

sufficient selectivity or permeance. Also the thermochemical stability of membranes is often

insufficient for the process conditions used [17].

Another type of membrane reactors involves the reactors where the reactant is fed in a con-

trolled manner to the reactor via a membrane. Examples are the oxidative coupling of methane

[19-21] and ethylene hydrogenation [22]. Especially in the case of dense perovskite-type

membranes large problems with respect to materials stability are to be expected in the reactive

environments. Also, because of the very high temperatures (900-1000ºC) used in this type of

reactors, the high temperature sealing technology that should be developed involves a large

technological risk.

A novel type of membrane reactor, emerging presently, is the pervaporation reactor. Conven-

tional pervaporation processes only involve separation and most pervaporation set-ups are

used in combination with distillation to break azeotropes or to remove trace impurities from

product streams, but using membranes also products can be removed selectively from the re-

action zone. Next to the polymer membranes, microporous silica membranes are currently un-

der investigation, because they are more resistant to chemicals like Methyl Tertair Butyl Ether

(MTBE) [23-24]. Another application is the use of pervaporation with microporous silica

membranes to remove water from polycondensation reactions [25]. A general representation

of such a reaction is:

R(COOH)x + R’(OH)y o RCOOR’ + H2O

Clearly, in such a reaction the removal of the produced water will lead to an enhanced conver-

sion. Commercially available polymer membranes cannot withstand the severe operation and

cleaning conditions for this process (150-300ºC) and microporous silica membranes again

come into the picture.

2. Hydrogen selective membranes

When considering membrane reactors for dehydrogenation and reforming reactions, three

types of membrane are of most interest: dense palladium or palladium composite membranes,

Introduction 3

silica membranes produced by Chemical Vapour Infiltration (CVI) and silica membranes pro-

duced by sol-gel techniques. These membrane types will now be discussed.

2.1 Palladium membranes

Dense palladium membranes are the most investigated membranes for hydrogen separation.

They are investigated by several research groups as the group of Gryaznov [26-28] and the

group of Uemiya and Kikuchi [29,30].

However, a large problem when pure palladium membranes are used is the phase transition

between the α- and β-form in hydrogen-containing environments. This phase transition can be

suppressed by doping the palladium with other metals. Silver is used most and Pd-alloy mem-

branes with 23-25% silver are produced commercially by Johnson Matthey [31,32]. These

membranes show increased stability and slightly higher hydrogen permeance compared with

the conventional, undoped, palladium membranes. Johnson Matthey claims a hydrogen per-

meance of 1 * 10-6 mol/m2sPa at 500ºC.

Currently, palladium and palladium composite membranes are used only when ultra-pure hy-

drogen is needed and only in the separation step. First, the hydrogen is produced, for example

with methanol steam reforming [32] after which the produced H2/CO2 mixture is fed to the

palladium membrane. The chemical stability of the palladium membranes, however, remains a

large problem. They are very sensitive towards sulphur and chlorine and also the stability to-

wards CO might be problematic. It has been reported that a CO concentration of only

0.2 vol-% gives a significant reduction in hydrogen flux [33,34]. The poisoning effects of

these gases, however, depend largely on the type of alloy [35] and can therefore be limited by

choosing the right type of alloy for a specific gas mixture.

2.2 CVI-silica membranes

Much research has been performed on silica membranes produced by Chemical Vapour Infil-

tration for hydrogen separation purposes. In chapter 7, CVI experiments are described and a

concise literature review is provided as well. Below some highlights will be presented briefly.

In CVI, the pores of a porous medium are plugged by the reaction product of a precursor and

an oxidising agent. For the preparation of silica layers, the precursor is a gaseous or volatile

Chapter 14

silica compound, such as silane [36-38] or Tetra Ethyl Ortho Silicate [39-42] and the oxidis-

ing agent is usually pure oxygen, air or steam. The porous materials used for CVI are com-

monly α-alumina supports [39], γ-alumina membranes [40] or porous Vycor glass [36,37].

In theory CVI membranes are very promising. Especially membrane stability is expected to be

very good, because the separative layer is located inside the support, where it is protected

against mechanical damage and chemical attack. In addition, measured permselectivities of

CVI-silica membranes are very high: for H2:N2 values as high as 3000 have been measured

[36,37].

However, large-scale commercial production of CVI-membranes is hindered by the compli-

cated equipment needed for their synthesis and the associated high costs. Furthermore the

permeance of CVI membranes (1*10-7 mol/m2sPa or lower) are still too low. Maybe these low

permeances can be increased using different precursors or other reaction conditions, but a

large improvement is not to be expected.

2.3 Sol-gel silica membranes

A third type of membrane is the sol-gel microporous silica membrane. This type of membrane

is of major importance in this thesis. Below, a short overview will be provided of state-of-the-

art silica membranes at the start of the project (1995). This has been the starting point from

which the new membranes described in this thesis were developed.

2.3.1 Synthesis

At the start of the project microporous sol-gel silica membranes were under investigation in

various research groups. An extensive literature review is provided in the introduction of the

thesis of De Vos [43]. Common supports for sol-gel silica membranes are mesoporous

γ-alumina membranes [44-47]. These mesoporous membranes are either home made [43-45]

or obtained from commercial sources [46,47].

A standard membrane as prepared by de Lange [45] consisted of a die-pressed α-alumina sup-

port, fired at 1360ºC with a pore diameter of 160 nm on which a γ-alumina membrane was

coated with a home-prepared boehmite sol. The coated γ-alumina layer was calcined at 600ºC,

had a thickness of 7 µm with a pore diameter of ~5 nm. On top of this mesoporous membrane,

Introduction 5

a silica top-layer was coated with a home-prepared polymeric silica sol. The silica layer was

fired at 400ºC and had a pore-diameter of around 3Å. Some defects were however present in

the microporous silica layer resulting in relatively low permselectivities. For example the

H2/CO2 permselectivity was only 5, which is very close to the ideal Knudsen selectivity of 4.7.

An SEM micrograph of a supported silica membrane is provided in Figure 1. Because of the

very small thickness (~30 – 50 nm) of the silica layer, this layer is not visible in the micro-

graph.

2.3.2 Properties

At the start of the project (1995) state-of-the-art microporous silica membranes as prepared by

de Lange [45] and described above had a permselectivity of 43 of hydrogen towards methane

and a hydrogen permeance of 1.6*10-6 mol/m2sPa.

By performing the synthesis of the membranes under cleanroom conditions later, De Vos

[48,49] showed the influence of particle contamination on the integrity of silica membrane

layers. Furthermore, the firing temperature of the silica was increased to 600ºC, which resulted

Figure 1: SEM micrograph of a supported silica membrane.

Chapter 16

in a decrease of the hydroxyl group concentration on the pore-surface in the silica layer and a

significant decrease in the CO2 permeance [50]. Therefore, the H2/CO2 permselectivity im-

proved by more than a factor 10, while the hydrogen permeance remained high (6 * 10-7

mol/m2sPa at 300ºC).

Gas transport properties through silica membranes have not been extensively studied. Espe-

cially the resistance of gas transport of small molecules like H2 through the thin SiO2 layer are

currently such that the resistance in the supporting layers should not be ignored or might even

dominate the transport properties of the final membrane, see for example [50].

In macro- and mesoporous membrane layers the nature of the flow is determined by the rela-

tive magnitude of the mean free path λ of the molecules and the pore size dp. When the mean

free path of the gas molecules is much larger than the pore size, i.e. λ >> dp, collisions of

molecules with the pore walls are predominant and the mass transport takes place by the well-

known selective Knudsen diffusion process. If the pore radius is much larger than the mean

free path of the molecules and a pressure difference over the membrane exists the mass trans-

port takes place by non-selective viscous flow.

Studies with many types of porous media have shown that for the transport of a pure gas the

Knudsen diffusion and viscous flow are additive (Present and DeBethune [52] and references

therein). When more than one type of molecules is present at intermediate pressures there will

also be momentum transfer from the light (fast) molecules to the heavy (slow) ones, which

gives rise to non-selective mass transport. For the description of these combined mechanisms,

sophisticated models have to be used for a proper description of mass transport, such as the

model presented by Present and DeBethune or the Dusty Gas Model (DGM) [53]. In the DGM

the membrane is visualised as a collection of huge dust particles, held motionless in space.

Benes and Verweij provide a thorough theoretical description of the multi-component mass

transport in microporous systems [54]. Lately, some systematic gas transport data has been

obtained for different microporous membranes in our group [50], but more extensive meas-

urements are necessary to get a good insight in the detailed transport properties of the different

types of silica membranes.

A good description of the transport phenomena in the membrane systems studied here may

result in the possibility to develop quality estimators for membrane units. Such quality esti-

mators can be used in process industry to evaluate compositions of permeate and retentate

Introduction 7

streams from the reactors. In this way, it is possible to control product streams by the use of

easy measurable quantities, such as feed and sweep flows, total pressure and temperature. By

doing this, the number of expensive and maintenance consuming gas chromatographs in proc-

ess streams can be reduced.

3. Project description and objectives

The goal of the present study is the development of a high temperature membrane reactor for

steam reforming of natural gas (methane), which occurs by the following reaction:

CH4 + 2H2O o 4H2 + CO2

Actually this is a combination of steam reforming by:

CH4 + H2O o 3H2 + CO

and the shift conversion reaction by:

CO + H2O o CO2 + H2

The removal of H2 from the reaction zone in a membrane reactor under equilibrium condi-

tions, enables three possible changes, or a combination of these changes:

1. By keeping temperature and conversion constant one is able to reduce the catalyst volume.

2. By keeping the catalyst volume and the temperature constant one can increase the conver-

sion of the reaction.

3. By keeping the catalyst volume and the conversion constant one can lower the reaction

temperature. It is mainly this last option that is considered in this thesis.

Improvements suggested above have a large positive environmental effect. When environ-

mental regulations regarding CO2 emissions become stricter, a reactor that utilises one or

more of the above options might even be cost-effective compared with the conventional steam

reformers.

3.1 General objectives

The development of a novel membrane reactor requires considerable effort, so a European

consortium of universities, institutes and industry was formed. The complete consortium con-

sists of seven partners including the University of Twente. The development of a new micro-

Chapter 18

porous silica membrane, stable under steam reforming conditions, is the main objective of the

University of Twente in the consortium. The results of this development are described in this

thesis, while a detailed list of objectives is provided in section 3.2. The other tasks are de-

scribed briefly below.

After preparation of newly developed membranes, high temperature permeation measurements

were performed by VITO in Belgium. Steam treatment and membrane material characterisa-

tion were performed at SINTEF in Norway and reactor testing together with kinetic modelling

of the reactor at IRC in France. The development of a high temperature test module for this

reactor testing was the task of Velterop BV in the Netherlands.

Clearly, for steam reforming at lower temperatures, other catalysts are necessary, so the choice

of a catalyst and the testing of this catalyst under real process conditions are important. This

task has been a cooperation between IRC and Norsk Hydro in Norway.

To get more insight in the economics of the project and the merits and drawbacks of the mem-

brane reactor with respect to a conventional steam-reformer, also a techno-economic evalua-

tion (TEE) was performed. For the use of the produced hydrogen two cases were considered.

The hydrogen can be used for the production of NH3 with ammonia being the feedstock for

fertiliser production (case 1, Norsk Hydro) or for the production of electricity by using the hy-

drogen as fuel for a gas turbine (case 2, KEMA, the Netherlands). Some results of the TEE,

together with some general ideas about reactor engineering and how to operate a membrane

steam reformer are provided in chapter 2 of this thesis.

3.2 Derived objectives for membrane development

The main task for the University of Twente was the development of a hydrogen selective mi-

croporous silica membrane for use at high temperatures. The membrane should be suitable for

use in a membrane reactor for steam reforming of natural gas. The target goals stated in the

project were to develop a membrane with a H2 flux equal to 1 * 10-5 mol/m2sPa with a sepa-

ration factor > 50 with respect to the other gas components like CH4, CO and CO2 for 1000

hours at 600°C in Simulated Ambient Steam Reforming Atmosphere (SASRA). SASRA con-

ditions are: 30 bar total pressure with CH4 : H2O = 1:3. Apart from the stability towards the

high temperature, mainly the stability in steam-containing environments was expected to be a

large problem.

Introduction 9

4. Problem definition

In summary, the main goal of the present work is the development of a hydrothermally stable

microporous silica membrane with prescribed transport properties. Preferably, these steam

stable membranes should have very high permselectivities. Because the permselectivity of a

molecular sieving silica membrane will drop to the Knudsen value of the γ-alumina supporting

membrane when the silica membrane deteriorates under steam reforming conditions, a selec-

tivity of the silica layer higher than the Knudsen selectivity is sufficient. In this way the meas-

urement of the permselectivity is a powerful tool to assess the hydrothermal stability of a sup-

ported microporous membrane.

For the preparation of high-quality membranes, also high quality supports are needed. It was

decided that the project would start with the development of colloidal filtrated flat supports

and centrifugal cast tubular supports, which have a higher degree of homogeneity than con-

ventional die-pressed, tape-cast and extruded supports. The development of these new sup-

ports is described in chapter 4 of this thesis.

5. Thesis outline

The purpose of the thesis is to provide a detailed description of the improvements that have

been made since the project start-up on silica membranes.

In chapter 2, some basic ideas about steam reforming in conventional and membrane reactors

are worked out. In this chapter the operation of conventional steam-reformers is compared

with possible membrane steam-reformers. In this chapter also a techno-economic evaluation

of a membrane reactor compared with the conventional process is provided. The boundary

conditions imposed by process technology and the techno-economic evaluation result in the

formulation of requirements for the development of the membranes, i.e. selectivity, flux, tube

length, operating pressure, etc.

The improvements that have been made in the preparation of molecular sieving silica mem-

branes started with the development of high quality membrane supports, because quality of the

supporting system is of crucial importance for the quality of the final molecular sieving mem-

brane. To this end, the synthesis of the supports was performed by means of colloidal proc-

Chapter 110

essing. A literature review of the basic concepts of the preparation of colloidal suspensions is

provided in chapter 3, whereas the actual preparation is dicussed in chapter 4.

On top of the newly developed supports a steam-stable intermediate layer was coated. The

preparation of these layers is treated in detail in chapter 5. After this, the permselective silica

layer was applied, which should be resistant against high temperature and steam-containing

environments as well. The experimental procedure together with some transport and Ruther-

ford BackScattering (RBS) studies are described in chapter 6.

Apart from the silica membranes prepared by dipcoating, also Chemical Vapour Infiltration

(CVI)-type membranes have been prepared. Chapter 7 is dedicated to this type of membrane.

In chapter 8 a new project has been formulated for the use of membrane reactors for the ther-

mal dehydrogenation of H2S. Compared to the conventional Claus process, the application of

a membrane reactor in the thermal H2S might have some large advantages.

Finally, in chapter 9, conclusions are drawn and suggestions made for further research on

(steam-stable) molecular sieving silica membranes or mesoporous γ-alumina membranes.

Though not all of the project objectives were obtained, progress was made in the synthesis of

micro- and mesoporous membranes. Especially the development of steam stable membranes

may be a large step forward in the development of ceramic membranes.

6. References

1. R. Bredesen, “Key Points in the Development of Catalytic Membrane Reactors” Paper no. A7.0 in Proc.

13thInt. Congr. Chem. Process Eng., August 23-28 1998, Praha, Czech Republic.

2. G. Saracco, H.W.J.P. Neomagus, G.F. Versteeg and W.P.M. van Swaaij, “High-Temperature Membrane

Reactors: Potential and Problems”, Chem. Eng. Sci., 54 1997-2017 (1999).

3. G. Sarracco and V. Specchia, “Catalytic Inorganic Membrane Reactors: Present Experience and Future Op-

portunities”, Catal. Rev. Sci. Eng., 36 [2] 305-84 (1994).

4. J. Zaman and A. Chakma, “Inorganic Membrane Reactors”, J. Membrane Sci., 92 1-28 (1994).

5. J.N. Armor, “Membrane Catalysis: Where is it Now, What Needs to be Done”, Catal. Today, 25 199-207

(1995).

6. Y. Yildirim, E. Gobina and R. Hughes, “An Experimental Evaluation of High-Temperature Composite

Membrane Systems for Propane Dehydrogenation” J. Membrane Sci., 135 107-15 (1997).

Introduction 11

7. J.C.S. Wu, T.E. Gerdes, J.L. Pszczolkowski, R.R. Bhave, P.K.T. Liu and E.S. Martin, “Dehydrogenation of

Ethylbenze to Styrene Using Commercial Ceramic Membranes as Reactors”, Sep. Sci. Tech., 25 [13-15]

1489-510 (1990).

8. T. Kameyama, M. Dokiya, M. Fujihige, H. Yokokawa and K. Fukuda, “Production of Hydrogen from Hy-

drogen Sulfide by Means of Selective Diffusion Membranes”, Int. J. Hydrogen Energy, 8 [1] 5-13 (1983).

9. T. Kameyama, M. Dokiya, M. Fujishige, H. Yokokawa and K. Fukuda, “Possibility for Effective Production

of Hydrogen from Hydrogen Sulfide by Means of a Porous Vycor Glass Membrane”, Ind. Eng. Chem. Fu-

nam., 20 97-99 (1981).

10. S.L. Jorgensen, P.E.H. Nielsen and P. Lehrmann, “Steam Reforming of Methane in a Membrane Reactor”,

Catal. Today, 25 303-7 (1995).

11. S. Uemiya, N. Sato, H. Ando, T. Matsuda and E. Kikuchi, “Steam Reforming of Methane in a Hydrogen

Permable Membrane Reactor", Appl. Catal., 67 223-30 (1991).

12. M. Chai, M. Machida, K. Eguchi and H. Arai, “Promotion of Hydrogen Permeation on Metal-Dispersed

Alumina Membranes and its Application to a Membrane Reactor for Methane Steam Reforming”, Appl.

Catal. A, 110 239-50 (1994).

13. A.M. Adris, S.S.E.H. Elnashaie and R. Hughes, “A Fluidized Bed Membrane Reactor for the Steam Re-

forming of Methane”, Can. J. Chem. Eng., 69 1061-70 (1991).

14. T. Johansen, K.S. Raghuraman and L.A. Hacket, “Trends in Hydrogen Plant Design – Steam Reforming will

Continue to be the Main Source of H2”, Hydrocarbon Processing, [8] 119-27 (1992).

15. E. Kikuchi and Y. Chen, “Low-Temperature Syngas Formation by CO2 Reforming of Methane in a Hydro-

gen Permselective Membrane Reactor”, Stud. Surf. Sci. Catal. 107 547-53 (1997).

16. E. Kikuchi, S. Uemiya, N. Sato, H. Inoue, H. Ando and T. Matsuda, “Membrane Reactor Using Micropo-

rous Glass-Supported Thin Film of Palladium. Application to the Water Gas Shift Reaction”, Chem. Lett.,

489-92 (1989).

17. C.H. Chang, R. Gopalan and Y.S. Lin, “A Comparative Study on thermal and Hydrothermal Stability of

Alumina, Titania and Zirconia Membranes”, J. Membrane Sci., 91 27-45 (1994).

18. S. Uemiya, N. Sato, H. Ando and E. Kikuchi, “The Water Gas Shift Reaction Assisted by a Palladium

Membrane Reactor”, Ind. Eng. Chem. Res. 30 585-89 (1991).

19. J.E. ten Elshof, H.J.M. Bouwmeester and H. Verweij, “Oxidative Coupling of Methane in a Mixed-

Conducting Perovskite Membrane Reactor”, Appl. Catal. A, 130 195-212 (1995).

20. A.M. Ramachandra, Y. Lu, Y.H. Ma, W.R. Moser and A.G. Dixon, “Oxidative Coupling of Methane in Po-

rous Vycor Membrane Reactors”, J. Membrane Sci., 116 253-64 (1996).

21. K. Omata, S. Hashimoto, H. Tominaga and K. Fujimoto, “Oxidative Coupling of Methane using a Mem-

brane Reactor”, Appl. Catal., 52 L1-L4 (1989).

22. A.F.Y. Al-Shammary, I.T. Caga, J.M. Winterbottom, A.Y. Tate and I.R. Harris, “Palladium-Based Diffusion

Membranes as Catalysts in Ethylene Hydrogenation”, J. Chem. Tech. Biotech., 52 571-85 (1991).

23. M. Asaeda, “Preparation of thin Porous Silica Membranes for Separation of Non-Aqueous Organic Solvent

Mixtures by Pervaporation”, Ceram. Trans., 31 411-20 (1993).

Chapter 112

24. M. Asaeda, P. Uchytil, T. Tsuru, T. Yoshioka, M. Ootani and N. Nakamura, “Pervaporation of Metha-

nol/MTBE Mixture by Porous Silica-Zirconia (10%) Membranes”, pp 322-25 in Proc. ICIM5 June 22-28,

Nagoya, Japan (1998).

25. J.W. Bakker, “Application of Ceramic Pervaporation Membranes in Polycondensation Reactions”, pp. 448-

51 in Proc. ICIM5 June 22-28, Nagoya, Japan (1998).

26. V.M. Gryaznov “Hydrogen Permeable Palladium Membrane Catalysts”, Platinum Metals Rev., 30 [2] 68-72

(1986).

27. V.M. Gryaznov, O.S. Serebryannikova, Y.M. Serov, M.M. Ermilova, A.N. Karavanov, A.P. Mischenko and

N.V. Orekhova, “Preparation and Catalysis over Palladium Composite Membranes” Appl. Catal. A, 96 15-

23 (1993).

28. V.M. Gryaznov, “Platinum Metals as Components of Catalyst-Membrane Systems”, Platinum Metals Rev.,

36 [2] 70-79 (1992).

29. S. Uemiya, N. Sato, H. Ando, T. Matsuda and E. Kikuchi, “Promotion of Methane Steam Reforming by Use

of Palladium Membrane”, Sekiyu Gakkaishi, 33 [6] 418-21 (1990).

30. E. Kikuchi, “Palladium/Ceramic Membranes for Selective Hydrogen Permeation and Their Application to

Membrane Reactor”, Catal. Today, 25 333-37 (1995).

31. J.E. Philpott, “Hydrogen Diffusion Technology, Commercial Application of Palladium Membranes”, Plati-

num Metals Rev., 29 [1] 12-16 (1985).

32. J.E. Philpott, “The On-Site Production of Hydrogen, A Mobile Generator for Meteorological and Industrial

Purposes”, Platinum Metals Rev., 20 110-113 (1975).

33. F. Sakamoto, Y. Kinari, F.L. Chen and Y. Sakamoto. “Hydrogen Permeation Through Palladium Alloy

Membranes in Mixtures Gases of 10% Nitrogen and Ammonia in the Hydrogen”, Int J. Hydrogen Energy,

22 [4] 369-75 (1997).

34. F.L. Chen, Y. Kinari, F. Sakamoto, Y. Nakayama and Y. Sakamoto, “Hydrogen Permeation through Palla-

dium-Based Alloy Membranes in Mixtures of 10% Methane and Ethylene in the Hydrogen”, Int. J. Hydro-

gen Energy, 21 [7] 555-61 (1996).

35. H. Yoshida, S. Konishi and Y. Naruse, “Effects of Impurities on Hydrogen Permeability through Palladium

Alloy Membranes at Comparatively High Pressures and Temperatures”, J. Less-Common Metals, 89 429-36

(1983).

36. G.R. Gavalas, C.E. Megiris and S.W. Nam, “Deposition of H2-Permselective SiO2 Films”, Chem. Eng. Sci.,

44 [9] 1829-35 (1989).

37. S.W. Nam and G.R. Gavalas, “Stability of H2-Permselective SiO2 Films Formed by Chemical Vapor Depo-

sition”, AIChE Symp. Series, 85 [268] 68-74 (1989).

38. S. Kitao and M. Asaeda, “Gas Separation Performance of Thin Porous Silica Membrane Prepared by Sol-

Gel and CVD Methods”, Key Eng. Mater., 61 & 62 267-72 (1991).

39. S. Morooka, S. Yan, K. Kusakabe and Y. Akiyama “Formation of Hydrogen-Permselective SiO2 Membrane

in Macropores of α-Alumina Support Tube by Thermal Decomposition of TEOS”, J. Membrane Sci., 101

89-98 (1995).

Introduction 13

40. C.L. Lin, D.L. Flowers and P.K.T. Liu, “Characterization of Ceramic Membranes II. Modified Commercial

Membranes with Pore Size under 40 Å”, J. Membrane Sci., 92 45-58 (1994).

41. S. Yan, H. Maeda, K. Kusakabe, S. Morooka and Y. Akiyama, “Hydrogen-Permselective SiO2 Membrane

Formed in the Pores of Alumina Support Tube by Chemical Vapor Deposition with Tetraethyl Orthosili-

cate”, Ind. Eng. Chem. Res., 33 2096-101 (1994).

42. J.C.S. Wu, D.F. Flowers and P.K.T. Liu, “High-Temperature Separation of Binary Gas Mixtures Using Mi-

croporous Ceramic Membranes”, J. Membrane Sci., 77 85-98 (1993).

43. R.M. de Vos, “High-Selectivity, High-Flux Silica Membranes for Gas Separation”, PhD Thesis, University

of Twente, 1998.

44. R.J.R. Uhlhorn, M.H.B.J. Huis in ‘t Veld, K. Keizer and A.J. Burggraaf, “High Permselectivities of Micro-

porous Silica-Modified γ-Alumina Membranes”, J. Mater. Sci. Lett., 8 1135-38 (1989).

45. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, “Permeation and Separation Studies on

Microporous Sol-Gel Modified Ceramic Membranes”, Microporous Mater., 4 169-86 (1995).

46. C.J. Brinker, T.L. Ward, R. Sehgal, N.K. Raman, S.L. Hietala, D.M. Smith, D.W. Hua and T.J. Headley,

“Ultramicroporous Silica-Based Supported Membranes”, J. Membrane Sci., 77 165-79 (1993).

47. N.K. Raman and C.J. Brinker, “Organic “Template” Approach to Molecular Sieving Silica Membranes”, J.

Membrane Sci., 105 273-79 (1995).

48. R.M. de Vos and H. Verweij, “High Selectivity, High Flux Silica Membranes for Gas Separation”, Science,

279 1710-11 (1998).

49. R.M. de Vos and H. Verweij, “Improved Performance of Silica Membranes for Gas Separation”, J. Mem-

brane Sci., 143 [1] 37-51 (1998).

50. N.E. Benes, A. Nijmeijer and H. Verweij, “Microporous Silica Membranes”, to be published in “Recent

Advances in Gas Separations by Microporous Membranes”, N. Kannellopoulos ed.

51. R.M. de Vos, W.F. Maier and H. Verweij, “Hydrophobic Silica Membranes for Gas Separation”, J. Mem-

brane Sci., 158 277-88 (1999).

52. R.D. Present and A.J. DeBethune, “Separation of a Gas Mixture Flowing Through a Long Tube at Low

Pressure”, Phys. Rev., 75 [7] 1050-55 (1949).

53. E.A. Mason and A.P. Malinauskas, “Gas Transport in Porous Media: The Dusty-Gas Model”, Chem. Eng.

Monographs, 17 1-175 (1983).

54. N.E. Benes and H. Verweij, “Comparison of Macro- and Microscopic Theories Describing Multicomponent

Mass Transport in Microporous Media”, accepted for publication in Langmuir.

&KDSWHU

7KHSURFHVVWHFKQRORJ\RI PHPEUDQHVWHDPUHIRUPLQJ

1. Introduction

Hydrogen is one of the most important industrial chemicals and energy carriers. Today

hydrogen is mostly produced using the steam reforming process [1-4]. In this process the

overall reaction is:

CH4 + 2H2O o CO2 + 4H2

At thermodynamic equilibrium conditions H2 conversion may be far from complete, and

hence a high temperature and high steam to carbon ratios are needed to obtain sufficient

conversion. Normally the process is carried out at 800-900°C and a pressure of 30 - 40 bar,

resulting in a conversion of 90%. In order to obtain the same conversion at a lower

temperature, hydrogen must be removed selectively from the reaction zone during the

process. This can be done in a hydrogen selective membrane reactor. Such a reactor, provided

with a membrane having a separation factor >50 for H2 towards CO2/CO/CH4/H2O would

give the same conversion at 600°C as obtained at 900°C in the conventional process [5].

The use of a membrane reactor in steam reforming has several advantages. Because of the

lower temperature operation, the energy consumption of the process is reduced which results

in lower emission of CO2. The lower temperature also requires less expensive catalyst, tubing

and other reactor materials. Since hydrogen of sufficient purity is produced directly from the

reformer, the downstream shift conversion can be omitted. Moreover, the dimensions of the

CO2 removal and final purification units can be reduced. Hence, significant savings in

equipment costs can be expected.

If the membrane surface area in the reactor is sufficiently high, conversion depends only on

the selectivity of the membrane. In this case, all the natural gas that is not lost by transfer

through the membrane will be converted in the reactor. Because of the large costs of the high

Chapter 216

surface area needed, this option is not of industrial interest. A surface area comparable to the

heat exchange area will be more realistic for industrial purposes. To obtain a conversion of

90% with this membrane surface area, the membrane must have a separation factor >50 and a

H2 permeance of 1 mol/m2s bar under steam reforming conditions [5].

The objective for the project, on which this thesis is based, is the extension of the applicability

of H2 selective microporous silica membranes to higher temperatures and harsh environments

by improvement of the material properties. Compared to the 1994 state-of-the-art, both H2

permeance as well as selectivity towards H2 had to be increased. In the present chapter, the

conventional process is discussed first for the sake of comparison together with several

catalyst issues. The membrane process is analysed after that and a comparison between both

approaches is made.

2. Conventional Process

The present treatment of the conventional process will be based on the process diagram

presented in Figure 1, which represents a steam reformer coupled with an ammonia synthesis

plant [6]. This is one of the two cases, which were considered in the project. The other was

the use of the produced hydrogen as fuel in combined cycle gas turbines. In this chapter, the

steam reforming part will be treated only, but some comments on the ammonia plant will be

made, in view of the composition of the product stream leaving the steam reformer.

2.1 Feed gas purification

The nickel-based reforming catalysts which are commonly used in steam reforming are quite

sensitive to sulphur, halogen and heavy metal poisons. Since these elements may all be found

in natural gas, a feed gas purification section is normally required. Of the mentioned catalyst

poisons, sulphur is by far the most important [6].

In process industry, quite a number of processes are available for the removal of sulphur from

gaseous feedstock. In steam reforming hot or cold zinc oxide beds are generally used for that

purpose. Zinc oxide is not only effective in removing the sulphur compounds but removes

some chlorides as well. H2S reacts irreversibly to the solid ZnS when it is led through the

ZnO-bed. The spent bed must be discarded afterwards. The overall reaction is:

The process technology of (membrane) steam reforming 17

ZnO + H2S o ZnS + H2O

Other organic sulphur compounds that are not easily removed by zinc oxide can be

hydrogenated to H2S first by reacting with the hydrogen over a cobalt or nickel molybdenate

catalyst. A conventional zinc oxide bed as described above can then remove the formed H2S.

If the chloride content of the natural gas feed is too high, a modified alumina catalysts that

can irreversibly absorb the chloride can be used.

Figure 1: Process diagram for natural-gas-based steam reformer with a connected ammonia plant [6].

Chapter 218

2.2 Primary and secondary reforming

In the primary and secondary reformer the following steam reforming reaction takes place:

CH4 + H2O o CO + 3H2

The catalyst for this reaction is normally nickel on a refractory or aluminate support. The

steam reforming reaction is highly endothermic (-∆H0298 = -206 kJ/mol) and high

temperature, low pressure and high steam-carbon ratios (3-4 is commonly used) favour

conversion [1].

Primary reforming

The primary reformer is a process furnace in which fuel is burned with air to provide the heat

of reaction to the catalyst contained within tubes. This area of the furnace is usually referred

to as the radiant section, so named because radiation is the primary mechanism for heat

transfer at the high (600-700°C) temperatures required by the process. Reforming pressures in

the range 3-4 MPa in the reactor provide a reasonable compromise between costs and

downstream recompression requirements. Carbon formation (coking) in the primary reformer

must be prevented (as is discussed further in paragraph 4).

State-of-the-art primary reformer designs differ in the arrangement of tubes and burners, tube

material, and feed gas distribution and reformer gas collector systems. A primary reformer

contains typically between 40 and 400 tubes. The internal tube diameter is in the range 70-

160 mm with a tube thickness of 10-20 mm. The heated length is 6-12 meter depending on the

furnace type. The tubes are made from high alloy nickel chromium steel by centrifugal

casting. In this casting process sections of ca. 6 meter length are produced which are welded

together to the required tube length. The practical limit on the primary reformer exit

temperature is determined by tube metallurgy considerations. One of the numerous possible

configurations is provided in Figure 2.

Secondary reforming

The reforming process is completed in the authothermic secondary reformer, which is a

refractory lined vessel containing a fixed-bed catalyst. The remainder of the endothermic heat

requirement is provided by the combustion of part of the primary reformer effluent directly

with air. This allows much higher process temperatures, of the order of 1000ºC, to be attained

at the secondary reformer exit and consequently low methane slips in the range of 0.2-

The process technology of (membrane) steam reforming 19

0.3 vol-%. The secondary reformer catalyst is similar to that used in the primary reformer.

Because the amount of air added to the secondary reformer is determined by the nitrogen

requirements of the downstream ammonia synthesis, the split between the primary and

secondary reformers obey in a heat balance consistent with equipment design temperatures

[6].

2.3 Shift conversion

Carbon monoxide, which is formed in the steam reforming reaction, deactivates the ammonia

synthesis catalyst and must be removed by means of the exothermic water-gas shift reaction,

which also maximises hydrogen production. To this end, CO is converted first to more easily

removable CO2:

CO + H2O o H2 + CO2

Figure 2: Side-fired primary reformer [6].

Chapter 220

Initially, the bulk of CO is shifted to CO2 in a high temperature shift (HTS) converter

operating at 350-450°C to take advantage of the faster reaction kinetics at those temperatures.

The HTS converter is operated at a temperature much lower than in the secondary reformer to

protect the used catalyst. The gases are cooled and the remaining CO is shifted to CO2 in a

low temperature shift (LTS) converter, operating at about 220ºC to achieve almost complete

CO conversion due to more favourable equilibrium conditions. HTS catalysts consist of

magnetite (Fe3O4) crystals stabilised by chromium oxide. Phosphorus, arsenic acid and

sulphur are poisoning catalyst. The LTS catalyst is normally copper oxide supported by zinc

oxide and alumina. After LTS the product stream contains some CO, only 0.25 to 0.4 vol-%

2.4 Carbon dioxide removal

The effluent gases from the shift converters contain about 17-19 vol-% (dry) carbon dioxide,

which is ultimately reduced to a few ppm by bulk CO2 removal, using an absorber-stripper

configuration. Three configurations are used in industry, illustrated by the examples in the

subsequent paragraphs:

1. Pure reaction

2. Combined reaction with physical adsorption

3. Pure physical adsorption systems

The choice of a specific CO2 removal system depends on the overall ammonia plant design

and process integration. Important considerations include: CO2 slip permitted, CO2 partial

pressure in the synthesis gas, presence of sulphur, process energy demands, investment cost,

availability of solvent, and CO2 recovery requirements.

Alkanolamine process

In this case, carbon dioxide reacts reversibly in the adsorber with aqueous alkaline solutions

to form a carbonate adduct (configuration 1). This adduct decomposes in the stripper upon

heating. In early ammonia plants, an aqueous solution of 15-20 wt % monoethanolamine

(MEA) was always standard for removing CO2. Primary alkanolamine solutions, however,

require a relatively high heat of regeneration so that, nowadays, secondary and tertiary ethanol

amines are mainly used.

Hence, activated tertiary amines such as triethanolamine (TEA) and methyl diethanolamine

(MDEA) have now gained wide acceptance for CO2 removal. These materials require very

The process technology of (membrane) steam reforming 21

low regeneration energy because of the weak CO2-amine interaction energy, and do not form

corrosive compounds.

Activated carbonate process

The activated carbonate process is based on absorption of CO2 by potassium carbonate to give

potassium bicarbonate (configuration 2). When potassium bicarbonate is heated it releases

CO2 while potassium carbonate is formed back again. The original hot carbonate process was

found too corrosive for carbon steel reactor walls. Nowadays, however, improvements in

additives and optimisation of operation have made activated carbonate processes competitive

with state-of-the-art MDEA systems.

Water stripping

A third method is CO2 removal by physical absorption in a (sea)water scrubber (configuration

3). Because of the low costs of (sea)water, large quantities can be used and a stripping section

is not necessary because the water is discarded.

PSA-unit

In modern plants, a Pressure Swing Adsorption (PSA) unit replaces the complete LTS, the

CO2 stripping section and the final purification.

2.5 Final purification

Oxygen-containing compounds (CO, CO2, H2O) contaminate the ammonia synthesis catalyst

and must be removed or converted to inert species before entering the ammonia synthesis.

The presence of CO2 in the synthesis gas may lead to the formation of ammonium carbamate,

which may cause fouling and compressor breakdown due to corrosion. Most ammonia plants

use a methanation process to convert carbon oxides to methane, while cryogenic processes

that are suitable for purification of synthesis gas have been developed as well.

Methanation

The methanation reactions used are the reverse of reforming and shift reactions:

CO + 3H2 o CH4 + H2O

CO2 + 4H2 o CH4 + 2H2O

Chapter 222

The methanator catalyst is nickel, supported by alumina, kaolin or calcium aluminate cement.

After methanation the CO and CO2 content of the treated gas is of the order of a few ppm. A

methanator typically operates in the temperature range of 300-400°C. Methanation reactions

are strongly exothermic and hence the CO and CO2 concentrations at the inlet of the

methanator should be carefully monitored, to avoid thermal runaway.

Dehydration

The use of molecular sieve dryers for removal of the remaining carbon oxides and water in

the synthesis gas to levels of < 1 ppm levels has gained prominence in low-energy-

consumption ammonia plant designs. Instead of molecular sieves so-called knockout drums

(high pressure vessels to remove traces of liquids) can be used as well.

3. Catalyst

The choice of the catalyst is of large influence on the behaviour of the reforming process.

Ni-based catalyst are most common, but recently more advanced catalysts have been

developed as well. As indicated before, one of the advantages of a membrane reactor is that it

can be operated at much lower temperatures but with the consequence that state-of-the-art

catalysts might not be sufficiently active anymore. In this paragraph, an overview is provided

on commonly used catalysts and some of the problems that may be encountered [7].

3.1 Nickel-based catalysts

The process design for steam reforming is based on the minimisation of the costs of hydrogen

production. As catalyst costs are high, their activity and stability play a critical role. Because

of the relatively low surface area of steam reforming catalysts, a high surface coverage of the

active nickel component is required to achieve an acceptable catalytic activity per unit weight

of catalyst. Consequently, the active nickel crystallites are situated close to each other,

however insufficient adherence to the carrier may lead to severe sintering (loss of nickel

surface area) during catalyst pre-treatment or actual operation.

On active Ni-based catalysts, coke formation is apt to occur. The primary site of carbon

formation is the acidic metal-promoted supporting oxide [7]. This catalytically active oxide is

however necessary for the majority of catalytic reactions and is essential for high steam

The process technology of (membrane) steam reforming 23

reforming activity. So the metal oxide site confers activity and imparts unselective carbon-

reforming properties if not correctly moderated. Modification of the highly active, carbon

forming catalyst site is generally accomplished by the introduction of basic species to

partially neutralise the active acid sites. In general, the level of basic moderator is chosen such

that the “super” active sites are neutralised, leaving the medium activity sites unaffected to

obtain the required process activity. Typical basic additives to catalyst formulations are

usually one or more of the metal oxides of sodium, potassium, lithium, cesium, calcium,

barium, strontium, magnesium, lanthanum and cerium.

Doping with alkali elements

Potassium is one of the most common constituents of basic additives that reduce carbon

formation. Potassium salts are highly soluble, however, and mobile at relatively low

temperatures and therefore prone to migration and loss from the catalyst surface. This might

lead to downstream deposits and potential process upsets. Andrew [8] stated, however, that

the presence of an adequate quantity of mobile alkali appears to be the key factor to enable a

supported nickel-reforming catalyst to operate successfully at low steam to carbon ratios. In

addition, it is well documented that potassium has an activity-moderating effect on steam

reforming catalysts [7,8,9], so that more catalyst has to be used. Another disadvantage of the

use of potassium in a membrane steam reformer is the possibility of reaction with the

separative silica layer of the membrane. At the envisaged steam reforming conditions this

may lead to the formation of a crystalline keatite phase [10,11].

Doping with lanthanum

To avoid the problems encountered with potassium, that additive can be replaced by

lanthanum oxide1[12]. Lanthanum oxide is a high melting point oxide with strong basic

properties. It neutralises carbon forming acidic sites and does not suffer from surface

migration or enhanced mobility at the catalyst surface as potassium does under influence of

steam. Contrary to potassium, lanthanum additions have a positive effect on catalyst activity

and it promotes the reduction of nickel as required to obtain sufficient steam reforming

activity. Moreover, no reactions of lanthanum with silica at steam reforming conditions are

known to occur.

1 Product information Dycat international, Mandeville, Louisiana, USA.

Chapter 224

3.2 Non-nickel catalysts

Instead of nickel, other catalytically active metals are used as well. Rhodium and ruthenium,

for example, show an activity that is about ten times higher than that of nickel, platinum and

palladium [6]. The addition of small amounts of copper to the conventional nickel catalyst is

reported to improve the activity of nickel at high temperatures [13].

Complications with in desulphurising heavy feedstocks have also lead to attempts to use non-

metallic catalysts for steam reforming, but their activity is still inferior to that of nickel

catalysts [14,15].

3.3 Catalyst poisoning

Sulphur is the most severe poison for steam reforming catalysts. A detailed study of sulphur

contamination is provided in [7]. On the other hand, sulphur may have a positive effect too,

because it may depresse coke formation on nickel catalysts [16].

A second important poison is As2O3 but its poisoning effect is much less than that of sulphur

[17]. The mechanism of As2O3-poisoning is based on the formation of an alloy with nickel.

The arsenic typically originates from the solutions used in carbon dioxide wash of the catalyst

or is present as an impurity in some zinc oxide sulphur removal beds. Also silica is mentioned

as a pore mouth poison by physically blocking the entrance to the pore system by which the

catalyst activity is decreased [18].

4. Coking and process conditions

As mentioned before, coke formation is apt to occur in the primary reformer, which is highly

undesirable because the catalyst conversion rate is then reduced significantly. A detailed

discussion of the mechanism behind coking and how to avoid coking at process conditions is

provided in [7]. Carbon formation on catalyst materials is discussed in paragraph 3 of this

chapter and a proper choice of catalyst, depending on the feedstock used for reforming, can

solve many coking problems. A good choice of process conditions, however, may also help to

minimise coke formation and if the right catalyst is chosen one can operate a steam reformer

for ten years without extensive coking problems [7].

The process technology of (membrane) steam reforming 25

Carbon may be formed from carbon monoxide and methane by the following reversible

reactions [6,12,18].

2CO o C + CO2 (Boudouard reaction, ∆H0298 = -173 kJ/mol)

CH4 o C + 2H2 (Decomposition of methane, ∆H0298 = 75 kJ/mol)

Depending on the operation conditions three different types of carbon can be formed:

whisker-like carbon, encapsulating carbon and pyrolytic carbon.

Whisker-like carbon is formed by diffusion of carbon through the Ni-crystal. After nucleation,

the whisker grows further with a Ni-crystal on top. This mechanism does not deactivate the

catalyst, but causes breakdown of the catalysts after some time. Whiskers are formed at

temperatures > 450ºC.

Encapsulating carbon consists of carbon polymers, which encapsulate the complete catalyst

particles. This type of carbon is formed at temperatures <500ºC. Encapsulating carbon

formation results in a progressive deactivation of the catalyst.

Pyrolytic carbon is formed by thermal cracking of the hydrocarbon feed. It will encapsulate

the catalyst as well finally resulting in deactivation of the catalyst and an increased pressure

drop over the reactor.

Apart from the catalyst type and its modification, the steam to carbon ratio has the largest

influence on coke formation. Formation may be expected below a certain, critical, steam to

hydrocarbon ration. The critical ratio was found to increase rapidly with temperature and to be

influenced by the type of hydrocarbon and by catalyst [6].

To avoid carbon deposition, the steam-to-carbon ratio is normally kept between 2.5 –3.0, but

processes exist too where a steam to methane ratio is used, as high as 4.0 [6].

The hydrogen content in the gas influences the coke formation rate [7] as well. In a recent

article Hou et al. [19] showed the influence of the removal of hydrogen on coking rates in a

membrane steam reformer using palladium membranes. The need of a minimum

concentration of hydrogen is of special importance when operating a membrane steam

reformer, because it limits the process conditions at which such a reactor can be operated.

A minimum hydrogen concentration is not only required to minimise coke formation, but it is

also important to avoid oxidation of the used catalyst [9,20]. Mostly steam to hydrogen ratios

of approximately 10 are used [20].

Chapter 226

Thirdly, also a minimum hydrogen concentration is required for inhibiting H2S poisoning of

the used catalyst [19]. Poisoning takes place by reactive adsorption of H2S on the nickel of the

catalyst surface:

H2S + Ni = NiS + H2

Of course, this reactive adsorption is favoured by removal of hydrogen from the reaction

zone. When 80% of the hydrogen is removed in the membrane reactor, the H2S tolerance of

the catalyst is about halve the tolerance when no hydrogen is removed from the reaction zone.

A higher degree of sulphur removal from the feed stream should be accomplished when

operating a membrane steam reformer.

5. Membrane reactor process

In this paragraph two of the most appropriate concepts for steam reforming employing a

membrane will be discussed, namely Membrane Steam Reforming (MSR) and Gas Heated

Reforming (GHR) with enriched air. The cases are described below, and are based on a

Techno-Economic Evaluation prepared by KEMA, SINTEF and Norsk Hydro [21].

5.1 Membrane Steam Reforming

The proposed process design for Membrane Steam Reforming (MSR) is shown in Figure 3.

The natural gas feed is depressurised first from 100 bar to 30 bar. The depressurised natural

gas feed is then heated in heat exchanger, passed through a water saturation column and

heated further before mixing with process steam to meet the selected steam to carbon ratio.

The mixed feed stream is heated up to 430ºC and fed to the catalytic membrane steam

reformer. Hydrogen formed by the steam reforming and water-gas-shift reactions is then

selectively removed from the reaction zone through the membrane.

The high-pressure gas (retentate) stream leaves the membrane steam reformer at 625°C and

30 bar, while the H2-rich permeate stream leaves the membrane steam reformer at 555°C and

1.5 bar. Pure nitrogen from an air separation plant is supplied as a sweep gas on the permeate

side of the membrane.

The process technology of (membrane) steam reforming 27

The hydrogen content of the retentate stream is too small to make hydrogen recovery

economically feasible. On the other hand, the heat content of the retentate stream is reused.

The stream is cooled by heat exchange with part of the reformer feedstock and subsequently

used for preheating the water feed of the saturation column.

The permeate (product) stream is split into two streams, providing heat to the natural gas feed

and for producing LP steam (not drawn in Figure 3). After a further temperature decrease

down to 25ºC by using cooling water, the product stream is compressed in three steps from

1.4 bar to 36.7 bar using repeated inter-stage cooling to 25°C and water knockout drums after

each compressor. CO is methanised, as described in paragraph 2.5 and residual water is

removed either by molecular sieves or knockout drums. The ammonia synthesis gas (N2, H2)

is finally compressed to 100 bar before being used for the ammonia synthesis.

5.2 Gas heated reforming using enriched air

Another possible concept for membrane steam reforming is Gas Heated Reforming (GHR). A

flowsheet of this process is provided in Figure 4.

Figure 3: Process layout of a membrane steam reformer [21].

Chapter 228

The natural gas feed is depressurised again from 100 to 30 bar. The heated gas stream is

saturated in a column by counter-current scrubbing with hot water. The saturated gas stream

is heated further, before being mixed with additional steam to obtain the required steam to

carbon ratio of 3.0.

The mixed feed stream is given a final preheating to 430ºC, and fed to the gas-heated

reformer, where the feedstock is partially converted to synthesis gas by conventional

membrane steam reforming (paragraph 5.1). The partially reformed gas leaves the gas-heated

reformer and is fed to the secondary reformer together with enriched air and hence partially

combusted.

The product gas leaves the secondary reformer at a temperature of 885°C and is heat-

exchanged in the primary membrane reformer. After that, the product gas leaving the gas-

heated reformer is utilised for preheating of the natural gas feed, heating of circulating water

in the saturator loop and generation of LP steam at 3 bar. Finally, after a temperature decrease

to 265ºC the gas is fed to a shift converter, after which again methanation takes place and

removal of CO2 and traces of water.

Figure 4: Process layout for a gas heated reformer [21].

The process technology of (membrane) steam reforming 29

5.3 Membrane design

For operation in a steam reformer, membranes must be found with a proper balance between

permeance and selectivity. Ideally, a membrane with both high selectivity and high

permeance is required, but one may expect on forehand that, typically, attempts to maximise

one are compromised by a reduction in the other. State-of-the-art hydrogen-selective

membranes were already discussed in chapter 1 and the reader is referred to that for more

information on suitable membrane types.

The membrane surface area in the reactor has to be optimised with respect to the number of

membrane tubes. There are, however, two important boundary conditions:

• The length of the used tubes. With current technology it is not possible to produce proper

membrane tubes longer than 2 m. Requirements on both membrane quality and

microstructural homogeneity along the tube determine the maximum allowable length. On

the other hand, two or more complete membrane tubes might be sealed head to head to

obtain longer lengths.

• The location of the catalyst. A membrane steam reformer contains two types of tubes, the

reformer tubes and the membrane tubes. The membrane tubes are placed inside the

reformer tubes. There are two possibilities for both the location of the separative layer on

the membrane tube and the location of the catalyst. The separating layer can be located at

the inside or at the outside of the support tube. Membrane tubes with their separative

layer inside are less sensitive towards operational and handling damage.

The catalyst can be placed inside the membrane tubes or between the membrane and the

reformer tubes. Unfortunately, both possibilities have specific disadvantages:

• If the catalyst is placed inside the membrane tubes and also the separative layer is coated

at the inside of the tube, any compounds such as potassiumoxide from the catalyst might

react with the silica separative layer to form keatite. This will destroy the molecular

sieving properties of the silica toplayer, see paragraph 3.1. Additionally there is a risk that

catalyst loading will damage the membrane layer.

• Positioning of the catalyst between the membrane and the reformer tubes can result in a

lower hydrogen flux through the membrane in comparison with the configuration in

which the catalyst is placed near the separative layer at the inside of the tube. This

decrease will be due to gas transport limitations through the support, because the pores of

the support might be partly blocked by molecules from the feed stream. This effect will

Chapter 230

not occur when the feed-stream is at the side of the separative layer, because the

molecules, which may cause blocking, cannot pass the separative layer.

In view of the above-mentioned phenomena, the best and most reliable configuration will be

the one with the catalyst in the annular space between the outer reformer tube and the inner

membrane tube. The separative layer can be located best at the inside of the membrane tube.

This configuration is shown in Figure 5.

5.4 Sealing

The membrane tubes must be sealed, possibly to each other and to the collector plate of the

reactor vessel. The sealing has to be adherent and mechanically strong but also gas-tight and

thermally resistant, up to temperatures of at least 700°C. These high demands make sealing

one of the most important problems in current high temperature membrane technology.

State-of- the-art high temperature sealing materials are based on glass and glass-ceramic [22-

25]. The major disadvantage of such materials in steam reforming environments is their

possible limited resistance against the high temperatures and corrosive environments

occurring in the steam reforming reactor space.

Figure 5: Cross-section of a reformer tube in a membrane steam reformer. The separative layer is located at

the inside of the membrane tube.

The process technology of (membrane) steam reforming 31

Compared to glass, glass-ceramics are mechanically stronger, more resistant to chemical

attack and have a wider range of thermal expansion coefficients. Complex non-linear thermal

expansion characteristics can be achieved, resulting in very close thermal expansion matching

to a variety of metals and alloys, including those with non-linear behaviour [24].

A good sealing not only dictates the properties of the sealing material, but also some

requirements on the tube design. In the ideal case the tubes are perfectly round, which enables

the application of a very thin seal in the annular space between the tube and the collector

plate. This, in its turn, is advantageous, because the chance of seal cracking during heating up

of the reactor is then largely reduced. To obtain tubes with a superior roundness compared to

conventional ones, centrifugal cast tubes have been developed. These are discussed further in

chapter 4.

6. Comparison of the different processes

In this paragraph, a sensitivity analysis is made of the costs of the different processes. It must

be noted in advance, however, that it is very difficult to provide really accurate quantitative

cost estimations, because of the lack of information on, for example, membrane selectivity

and life-time and the costs of supported membranes and sealing. An attempt to provide yet a

quantitative cost analysis has been made in [21].

6.1 Tube length

The costs of a membrane reactor is highly dependent on the number of tubes used. In [21] a

comparison is provided of the (membrane) tube length against the investment. By using

longer tubes, the costs of burners and tube collector plates decrease considerably.

Conventional steam reformers contain tubes with a length of 12-14 meters. According to [21],

a membrane reactor consisting of 12 meter tubes would have investment costs which are 50%

lower than a membrane reactor with 2 meter tubes. For ceramic membranes, however, a

length of 12 meter is not realistic today. Therefore, an option can be to prepare tubular

membranes with a length of 2 m and to seal them together to a length of 12 m. In this case

only the sealing costs increase, but the advantage of a large tube length remains. Sealing two

tubes together is a completely new technique, however, and will need a large amount of

development work.

Chapter 232

The costs of the GHR concept are somewhat less sensitive to the number of tubes, because a

GHR does not contain burners.

6.2 Temperature

The reaction temperature has a significant influence on operating costs. When using a

membrane reactor it might be possible to operate the system at a much lower temperature,

enabling the use of less expensive tubing materials. As a starting point for the project, we

chose 600ºC as reaction temperature. The economic evaluation prepared during the project

[21], however, uses a reaction temperature of 700ºC.

6.3 Membrane selectivity

The selectivity of the used membranes is of large influence as well. A selectivity larger than

500 would significantly reduce the amount of impurities in the permeate stream and thereby

downstream processing of the synthesis gas.

6.4 Permeate pressure

The largest inherent weakness of membrane reactors is the low pressure permeate streams

which should be recompressed for further use. This compression represents 30 to 40% of the

total annual operating costs.

6.5 Comparison of the different concepts

An attempt has been made to compare the different concepts on the basis of a cost estimation.

The results obtained in [21] are summarised in Table 1. They have been calculated with a

simulation spreadsheet, prepared by SINTEF [26]. For reasons of comparison, the results

from [21] have been compared with designs in which 12 meter (sealed) tubes are assumed

together with an operation temperature of 600ºC. Please note that in that case the total costs

are highly dependent on the costs of sealing. Moreover, sealing membrane tubes one to

another might be unrealistic. It is evident, however, from the comparison, that there is a large

uncertainty in cost-estimations for membrane reactors and that with only slight changes in the

input parameters of the cost-model, membrane steam reforming can be cost-effective. When

The process technology of (membrane) steam reforming 33

better membranes with higher permeances and selectivities become available and when

sealing technology is getting more developed and cheaper, membrane steam reforming

remains worthwhile studying. Moreover the possibility to operate membrane steam reformers

at lower temperatures than conventional steam reformers might make them even more cost-

effective when in future environmental regulations become more strict. These environmental

regulations might provide a large drive towards energy effective processes, which favours

membrane techniques. Implementation of membrane steam reforming in process industry in

future largely depends on the above mentioned factors.

7. Conclusions

Replacing conventional steam reformers by membrane steam reformers is an interesting

option providing the possibility of lower operating temperature, thereby creating a more

energy efficient process. Another advantage is the possibility to at least partly omit the

methanation and CO2 removal section. On the other hand, the use of lower temperatures

might involve serious problems with reformer operation. Coking problems might arise and

this effect might even become more outspoken since hydrogen is removed from the reaction

zone. In the worst case a lowered hydrogen concentration might even lead to (partial)

oxidation and sulphur poisoning and thereby extended deactivation of the catalyst.

Cost calculations for a membrane reactor are very cumbersome. Numerous uncertainties and

assumptions have to be made for a large number of parameters. Tube length and sealing costs

are very important and up to now it is not even sure whether a sealing material can be

developed that is able to withstand the severe operation conditions. This makes a proper

Total annual costs Investment costs

Conventional reforming 4.7 42.5

MSR (700ºC, 2 meter) 9.5 49.6

GHR (700ºC, 2 meter) 5.7 45.9

MSR (600ºC, 12 meter) 4.6 45.0

GHR (600ºC, 12 meter) 1.2 41.7

Table 1: Cost-comparison of steam reformers. All costs are given in million Euro.

Chapter 234

estimation of sealing costs and hence total costs very difficult. Besides that, it was found that

the cost estimation is rather sensitive towards assumptions for the design. This led us to the

conclusion that, membrane steam reforming might be cost effective with the GHR concept as

the most interesting option.

And last but not least costs are found to depend largely on different membrane properties and

first of all a suitable membrane has to be developed. The development of such a membrane is

described in the remaining chapters of this thesis.

8. References

1. S.L. Jorgensen, P.E.H. Nielsen and P. Lehrmann, “Steam Reforming of Methane in a membrane Reactor”,

Catal. Today, 25 303-7 (1995).

2. T. Johansen, K.S. Raghuraman and L.A. Hacket, “Trends in Hydrogen Plant Design – Steam Reforming

will Continue to be the Main Source of H2”, Hydrocarbon Processing, [8] 119-27 (1992).

3. F.W. Hohmann, “Improve Steam Reformer Performance”, Hydrocarbon Processing, [3] 71-74 (1996).

4. J.M. Abrardo and V. Khurana, “Hydrogen Technologies to Meet Refiners’ Future Needs”, Hydrocarbon

Processing, [2] 43-49 (1995).

5. Proposal to the EC Framework Programme IV, “Membrane Reactor for Cost-Effective Environmental-

Friendly Hydrogen Production”, Brite Euram, no. BE95 1930 (1995).

6. J.R. Rostrup-Nielsen, “Catalytic Steam Reforming”, Springer Verlag, Berlin (1984).

7. J.R. Rostrup-Nielsen, “Steam Reforming Catalysts”, Danish Technical Press Inc., Copenhagen (1975).

8. S.P.S. Andrew, “Catalysts and Catalytic Processes in the Steam Reforming of Naphta”, Ind. Eng. Chem.

Prod. Res. Develop., 8 321-24 (1969).

9. J.R. Rosrup-Nielsen, “Activity of Nickel Catalysts for Steam Reforming of Hydrocarbons” J. Catal., 31

173-99 (1973).

10. P.P. Keat, “A New Crystalline Silica”, Science, 120 328-30 (1954).

11. J. Shropshire, P.P. Keat and P.A. Vaughan, “The Crystal Structure of Keatite, a New Form of Silica”, Z.

Kristall., 112 409-13 (1959).

12. J.R. Rostrup-Nielsen, T.S. Christensen and I. Dybkjær, “Steam Reforming of Liquid Hydrocarbons”, Stud.

Surf. Sci. Catal., 113 81-95 (1998).

13. J. Barcicki, A. Denis, W. Grzegorizyk, D. Nazimek and T. Borowiecki, “Promotion of Nickel Catalysts for

the Steam Reforming of Methane”, React. Kinet. Catal. Lett., 5 [4] 471-78 (1976).

14. T. Tomita and M. Kitagawa, “Ein Neues Steam Reforming-Verfahren für Hochsiedende

Kohlenwasserstoffe”, Chem. Ing. Tech., 49 [6] 469-75 (1977)

15. T. Tomita, A. Moriya, T. Shinjo, K. Kikuchi and T. Sakamoto, J. Jap. Petrol. Inst., “ The Influence of

Steam on Coking Rates in Steam Reforming”, 23 [2] 69-74 (1980).

The process technology of (membrane) steam reforming 35

16. J.R. Rostrup-Nielsen, “Sulphur-Passivated Nickel Catalysts for Carbon-Free Steam Reforming of

Methane”, J. Catal., 85 31-43 (1984).

17. G.W. Bridger and W. Wyrwas, “Steam Reforming of Liquid Hydrocarbons”, Chem. Process Eng., 48 [11]

101-7 (1967).

18. T.S. Christensen, “Adiabatic Prereforming of Hydrocarbons – an Important Step in Syngas Production”,

Appl. Catal. A, 138 285-309 (1996).

19. K. Hou, M. Fowles and R. Hughes, “Potential Catalyst Deactivation Due to Hydrogen Removal in a

Membrane Reactor Used for Methane Steam Reforming”, Chem. Eng. Sci., 54 3783-91 (1999).

20. J.D. Rankin and J.G. Livingstone, “Catalysts: a Recipe for Longer Life”, Ammonia Plant Saf. 23 203-12

(1981).

21. R. Meijer, D. van der Vlist, F. Janssen, A. Anundskås, T. Pettersen and T. Strøm, “Membrane Reactor for

Cost Effective Environmental-Friendly Production of Hydrogen – Techno-economic Evaluation”, Internal

Report, (1997).

22. I.W. Donald, “Preparation Properties and Chemistry of Glass- and Glass-Ceramic-to-Metal Seals and

Coatings”, J. Mater. Sci., 28 2841-66 (1993).

23. C. Günther, G. Hofer and W. Kleinlein, “The Stability of the Sealing Glass AF45 in H2/H2O and O2/N2

Atmospheres”, Electrochem. Proc., 97 [18] 746-56 (1997).

24. M.A. Ritland, D.W. Ready, R.N. Kleiner and J.D. Sibold “Method for Sealing a Filter”, US Patent,

5.700.373 (1997).

25. F.M. Velterop, “Method of Connecting Ceramic Material to Another Material”, US Patent, 5.139.191

(1992).

26. Sintef Applied Chemistry, “Simulation and Cost Estimation of Membrane Steam Reformers”, Microsoft

Excel spreadsheet (1996).

&KDSWHU

&ROORLGDOSURFHVVLQJRI FHUDPLFPHPEUDQHVXSSRUWV

*HQHUDOLQWURGXFWLRQ

1. Introduction

In the preparation of multi-layered ceramic membranes, the quality of the support is of crucial

importance to the integrity of the membrane layers that are applied in the subsequent prepara-

tion steps. First, the surface roughness and homogeneity of the support will determine the in-

tegrity of these membrane layers, and, second the surface roughness determines the minimal

thickness of the membrane layer for complete surface coverage.

In this work, two support shapes are of particular interest: tubular and flat supports, which are

currently the most used supports in membrane research. Apart from these shapes also ceramic

multi-bore tubes and honeycomb structures are produced for membrane applications and re-

cently α-alumina hollow fibre supports were developed as well [1].

Though the majority of this chapter is related to the production of tubular supports, a few

comments on the preparation of flat supports will be made. Emphasis is put on the stabilisa-

tion of the suspensions used in the preparation of the supports. Experimental procedures are

provided in more detail in chapter 4 of this thesis; the present chapter mainly provides the ba-

sic knowledge for suspension preparation and shaping techniques.

1.1 Flat supports

Flat supports can be produced in various ways. Die pressing (or dry pressing) is most often

used, but also tape casting can be applied. Both methods have disadvantages due to the exten-

sive use of binding agents and other additives. In the case of die pressing, mainly PolyVi-

Chapter 338

nylAlcohol (PVA) is used as a binding agent, while in tape casting very complicated slurries

are used, consisting of plasticisers, binders and anti-foaming agents [1,3]. The use of extensive

amounts of these additives might result in inhomogeneities in the support, which certainly in-

fluence the quality of the subsequent membrane layers.

A rather new technique for the preparation of flat membrane supports is colloidal filtration. In

this technique an aqueous suspension is made of high purity α-alumina particles with a narrow

size distribution using a very low concentration of electrostatic or electrosteric stabiliser. The

suspension is homogenised using ultrasound, which results in the break-up of the large ag-

glomerates in the powder. The suspension is poured out in moulds after which the liquid in the

suspension is removed by vacuum filtration and the particles are stacked in a controlled man-

ner. The resulting cake (green cast) is dried overnight, released from the mould and fired to

the final product. Due to the homogeneous stacking, the product can be shaped and polished

with relative ease to a surface roughness necessary for applying membrane layers.

A more detailed synthesis procedure for the supports used in this study is provided in chap-

ter 4 and information on the strength of these supports as a function of synthesis conditions

can be found in [4]. Some general information about the colloidal filtration method can be

found in [5,6].

1.2 Tubular supports

For research on stability, permeability and separation characteristics of the membranes, flat

membranes are perfectly suitable. For the use in process industry, however, tubular mem-

branes are necessary, because of the ease of sealing tubes into modules compared to stacked

flat plate modules. The conventional way of preparing ceramic tubes is extrusion but a prob-

lem of extruded ceramic tubes may be the inhomogeneity of the surfaces due to inhomogene-

ous distribution of binders and other extrusion additives and stresses induced during process-

ing. Inhomogeneities often lead to an enhanced roughness, which may be detrimental to the

quality of subsequently applied layers. Moreover, the minimal thickness of an applied layer

has to be higher than the surface roughness of the supporting system, which means that if the

surface roughness is reduced, subsequent membrane layers may be thinner, thereby reducing

the resistance to flow.

Colloidal processing of ceramic membrane supports. General introduction 39

For preparing tubular ceramic parts, centrifugal casting of a colloidal suspension can be used.

In this process a suspension of particles with a particle size distribution as narrow as possible

is used to diminish segregation of the prepared suspension. The suspension is obtained by dis-

persing the powder in a suitable liquid by milling or ultrasound, after which it is poured or in-

jected in a mould-tube, which is placed in a horizontal or vertical centrifuge. The mould-tube

is rotated for a certain time at a typical rotational speed of 15.000 to 40.000 rpm and released

from the centrifuge. Afterwards the remaining liquid is poured or sucked out of the mould and

the mould-tube containing the wet green compact is dried in an atmosphere with controlled

temperature and humidity. The compact can be released from the mould after drying and fired

at a suitable temperature.

With centrifugal casting one can prepare homogeneous tubes with a very smooth inside sur-

face without the need of polishing. The importance of a low surface roughness has already

been pointed out.

While the preparation of flat supports by colloidal filtration is rather simple and robust, the

preparation of tubes by centrifugal casting offers more challenges. Especially during mould

release and drying cracking and warping of the green cast often occurs. These problems can be

eliminated by the use of additives in the starting suspension or the use of release agents on the

inner surface of the mould-tube. However, these techniques also result in inhomogeneities in

the final product and its use should therefore be limited.

The stability of the starting suspension should also be controlled carefully. If the starting sus-

pension is too stable, the final sediment will remain fluid-like, so that an actual compact is not

formed and redispersion will occur as soon as rotation ends. On the other hand, a less stable

suspension might give rise to attraction of the particles in the suspension and to flocculation,

which influences the homogeneity and the surface roughness of the final product. Controlled

flocculation, however, might be advantageous for mould release, because in this case the par-

ticles are not completely close-packed which gives rise to some shrinking during drying. This

shrinkage enables easy release of the green compact from the mould-tube.

Most literature on centrifugal (slip) casting deals with the preparation of dense and homoge-

neous flat ceramic parts from colloidal suspensions by centrifuging [7-15]. This method util-

ises centrifugal forces, but also makes use of forces involved in conventional slip casting to

produce higher green densities. The above mentioned articles contain valuable information

Chapter 340

about stability of very different kinds of colloidal suspensions. Another application of cen-

trifugal casting can be found in the production of steel tubes, as described in [16] and [17].

One of the most important articles on the centrifugal casting of ceramic tubes is from the

group of Bachmann [18]. They developed a method to synthesise silica tubes for the produc-

tion of optical telecommunication fibres, consisting of a commercial centrifuge equipped with

an injection tube. Using this technique, it was possible to produce a silica tube, layer by layer.

Because the doping of the silica could be changed during the process, axial varying optical

properties were obtained in the resulting tubes. An identical technique was used in the present

work to prepare multilayer tubular membrane supports. Results are described in section 9 of

chapter 4.

2. Theoretical background

2.1 The DLVO-theory

The DLVO-theory is named after Derjaguin, Landau, Verwey and Overbeek and predicts the

stability of colloidal suspensions by calculating the sum of two interparticle forces, namely the

Van der Waals force (usually attraction) and the electrostatic force (usually repulsion) [19].

The van der Waals force between atoms consists of three different dipole induced forces, the

Keesom interaction, the Debye interaction and the London interaction.

• Keesom interaction occurs when a permanent molecular dipole creates an electric field,

which orients other permanent dipoles in such a way that they will attract each other.

• Debye interaction occurs when a permanent dipole induces a dipole in a polarisable atom

or molecule. The induced dipole is oriented in such a way that attraction occurs.

• London interaction occurs by fluctuations in the electrons in atoms or molecules in such a

way that instantaneous dipoles are formed. This effect leads to attraction between the two

induced dipoles.

The origin of the electrostatic force is the surface charge that solid particles acquire when they

are immersed in a liquid that contains a sufficient amount of ions. Possible charging mecha-

nisms are ionisation, ion adsorption and ion dissolution, which are now discussed.

Colloidal processing of ceramic membrane supports. General introduction 41

Ionisation

Colloidal metal-oxide particles, with hydroxyl groups at their surface, may undergo proton

association or dissociation depending on the pH of the solution. At low pH, a metal-oxide

particle will be charged positively and at high pH negatively. The pH at which the net charge

is zero, is the iso-electric point.

Ion adsorption

A net surface charge can be acquired by the adsorption of ions on the surface of the particle.

Ion adsorption may be positive or negative. Surfaces, which are already charged (e.g. by ioni-

sation), usually show a tendency to adsorb counter-ions. It is possible that counter-ion adsorp-

tion causes a reversal of charge.

Surfaces in contact with aqueous media are more often negatively charged than positively.

This is a consequence of the fact that cations are usually more hydrated than anions and there-

fore have a greater tendency to reside in the bulk of the aqueous medium. The smaller, less

hydrated and more polarising anions have the greater tendency to be specifically adsorbed.

The interaction between the counter-ion and the surface is not only physical; chemical inter-

action may occur as well. Chemical interaction is not only influenced by the valence and size

of the ions, but also by the chemical composition of the surface and the ion. This type of ad-

sorption is called specific adsorption.

Depending on the concentration of the counter-ions a layer can be formed on the surface. If

the surface is positively charged, first a layer of negatively charged counter-ions will adsorb,

present within the inner Helmholtz-plane. On this layer of specifically adsorbed negatively

charged ions, a layer of positively charged ions can again be adsorbed, which are enclosed by

the outer Helmholtz plane, or simply the Helmholtz plane. The total layer of specifically and

non-specifically adsorbed ions is called the Stern layer.

Another layer of counter-ions will also form a diffuse layer surrounding the Stern layer, which

is called the Gouy layer. The Gouy and Stern layers as well as the Helmholtz plane are shown

in Figure 1.

Chapter 342

Ion dissolution

Ionic substances can acquire a surface charge by

unequal dissolution of the ions of which they are

composed. An example is silver iodide particles

in an aqueous suspension, which are in equilib-

rium with a saturated solution. With excess I--

ions, the silver iodide particles are negatively

charged. With an excess of Ag+-ions, the parti-

cles will be positively charged.

The DLVO-theory

In the DLVO-theory, the force between two particles is seen as a summation of the attractive

force imposed by the van der Waals force and the repulsive force imposed by the electrostatic

interaction.

For two spheres with radius Rs at a distance D, the force is given by:

In which, A is the Hamaker constant, N0 the number of ions, k the Boltzmann constant, T the

temperature and κ the Debye length. Z is given by:

In which z is the valence of the ions surrounding the sphere, e the electronic charge and ψ0 the

surface potential.

The behaviour of the interaction energy between two particles as a function of the surface

separation (distance between the surfaces) is plotted in Figure 2.

Figure 1: The different layers due ion adsorption

on a particle.

F F FAR

DN kTZDLVO attr rep

s D= + = − + −.

( )exp12

6420 2 κ

Zze

kT= tanh

ψ0

4

Colloidal processing of ceramic membrane supports. General introduction 43

In several cases, however, the DLVO-theory

has shown to be inadequate, due to the occur-

rence of other inter-particle forces that may be

present in colloidal suspensions. These phe-

nomena are summed up below:

Hydrophobic forces

Hydrophobic forces are long-range attractive

forces between macroscopic, hydrophobic sur-

faces in water. The force is significantly

stronger than the Van der Waals attraction and

can still be measured at surface separations as

large as 70 nm.

Solvation forces

When two surfaces are immersed in a liquid,

the force between them can be greatly affected

by the interaction of the liquid with the sur-

face. In this case the surface may be “solvated”

in a particular way. An isolated surface will

thereby modify the structure of the liquid adja-

cent to it. The nature and thickness of the sol-

vation layer depend on properties of surface

and liquid. The solvation force in water is

called the hydration force.

Capillary forces

It is well known that a vapour is apt to condense in a narrow gap, capillary or between two

solid surfaces in close proximity. Such a condensed meniscus between two particles results in

a strong attractive force between these particles due to surface tension.

Figure 2: Schematic diagram of the variation of

interaction free energy with surface

separation according to DLVO-theory.

The net energy (solid line) is given by

the sum of the double layer repulsion

(-.-.-.) and the Van der Waals attraction

(--------). Two different situations are

depicted, one with a low salt concen-

tration (a) and one with a high salt con-

centration (b) [20].

Chapter 344

Steric repulsion forces

Surfactants or polymers adsorbed on the particle surface are able to keep particles that far ap-

part that the Van der Waals attraction cannot become effective. This phenomenon is called

steric stabilisation.

Gravitational force

Gravitational forces might influence the stability of the suspension especially for larger parti-

cles (larger than 100 µm).

Brownian motion

Brownian motion, which is a result of random collisions between the particles in the suspen-

sion, might influence suspension stability especially for smaller particles (smaller than

100 µm).

2.2 Stabilising a colloidal suspension

Stabilising a colloidal suspension implies that the total interparticle potential decreases with

increasing inter particle distance. The different kinds of stabilisation all use some of the

above-mentioned interparticle forces.

There are mainly six different methods of stabilisation, as summed up by Everett [21], see

Figure 3.

1. Electrostatic stabilisation. Repulsion between the double layers formed by non-adsorbing

counter-ions causes stabilisation. Simple salts can be used which do not have a steric ef-

fect.

2. Electrosteric stabilisation. By using larger molecules with charge-carrying groups electro-

static and steric stabilisation are combined. The thicker the layer that is formed, the larger

the steric part of the stabilisation. Polyelectrolytes with a high number of -COO- groups

are mostly used for this type of stabilisation [23, 24].

3. Steric stabilisation. Particles with large molecules adsorbed on the surface are repelled by

each other because the freedom for chain movement decreases if particles approach (this

Colloidal processing of ceramic membrane supports. General introduction 45

would imply a decrease of entropy). For small enough particles, Brownian motion is then

sufficient to keep the particles suspended for an indefinite time. Note that the polymer

must not adsorb too strong on the particles because if polymer segments stick strongly and

irreversibly to the first area of the surface that they encounter, further polymer adsorption

is hindered and a comparatively poor polymer coverage of the surface may result.

4. Depletion stabilisation. This type of stabilisation is related to steric stabilisation and arises

only when a high concentration of non-adsorbing large molecules is dissolved in the sus-

pension. In this case, there would be an effect arising from the entropic penalty arising

from compressing the free polymer chains between the particles if not all polymer chains

are excluded from the gap between the particles. This effect would result in a repulsive

force between the particles and thereby stabilisation. This last idea is however not gener-

ally accepted and it therefore remains a question whether the depletion stabilisation effect

occurs in practice.

5. Hydration stabilisation. Due to the polarisability of the water molecules, even in a de-

ionised aqueous solution, stabilisation may occur. Positively charged alumina particles,

for example, bind preferentially to the negative oxygen of the water molecule. As a result,

a double layer is formed, similar to ionic electrostatic repulsion. In ionic solutions, the hy-

dration repulsive force occurs simultaneously with the electrostatic force, while the pro-

portion of the two forces depends on the ionic concentration [22].

6. Masking of the Van der Waals forces. By choosing a suspending medium with dielectric

properties as close as possible to those of the solid (particularly at optical and UV fre-

quencies), the van der Waals force is minimised.

Chapter 346

Clearly steric, electrosteric or depletion stabilisation as described above may occur, but attrac-

tion, leading to destabilisation, can also be the result, as can be seen in Figure 4 [20].

Destabilisation by polymer addition

Stabilisation of a suspension by just adding a polymer solution can have a beneficial, but also

a detrimental effect. When a polymer chain is very long, the possibility arises that one polymer

molecule can adsorb to more than one particle, thereby forming a link between the particles in

suspension. This effect is known as bridging flocculation and results in the destabilisation of

the suspension. The use of block copolymers can overcome the problem of bridging floccula-

tion. These block copolymers consist of two different polymer chains connected end to end. If

a block copolymer is dissolved in a liquid that is a good solvent for one of its ends but a poor

solvent for the other, the latter end will have strong tendency to adsorb to the particles, while

the remainder of the molecule extends into the solvent. In this way a coat of non-adsorbing

and non-bridging polymer is attached to the particle, thereby effectively preventing other par-

ticles from approaching.

Figure 3: Methods of stabilising colloidal suspensions [21].

Colloidal processing of ceramic membrane supports. General introduction 47

When using polyelectrolytes even more complicated effects can arise. The addition of large

polyelectrolytes to a suspension of opposite charged particles is likely to lead to bridging with

again one polymer chain attached to two or more particles, in this way forming a “necklace”.

Because of restrictions on the number of possible configurations, non-adsorbing polymers

tend to stay out of a region near the surfaces of the particles, known as the depletion layer. As

two particles approach, the polymers in the solution are repelled from the gap between the sur-

faces of the particles. In effect the polymer concentration in the gap is decreased and is in-

creased in the solution. As a result, an osmotic pressure difference is created which tends to

“push” the particles together. The resulting attractive force is the reason for depletion floccu-

lation. In contrast to this, depletion stabilisation has been mentioned above.

3. Stabilising alumina suspensions

Stable suspensions are necessary to obtain homogeneous casts by colloidal filtration. Different

kinds of stabilisation mechanisms for alumina suspensions are used in literature. Pure electro-

static stabilisation can be obtained using HNO3, while electrosteric stabilisation is obtained

using polyelectrolytes like PolyMethylAcrylicAcid (PMAA) and PolyAcrylicAcid (PAA). In

Figure 4: Different effects of adding polymers [20].

Chapter 348

the following section an overview will be given of several studies on the stabilisation of alu-

mina-containing suspensions with different stabilisation agents.

3.1 Stabilisation with HNO3

Hidber et al. [9] described the colloidal processing of wet-milled α-alumina suspensions by

centrifugal casting from a 80 wt-% suspension at pH 4.3. The mean particle diameter of the

α-alumina powder was 0.3 µm and the BET surface area 8.59 m2/g*. Nitric acid or ammonia

was used for electrostatic stabilisation. Very dense ceramics could be obtained with relative

densities up to 99.9%.

Also Huisman et al. [13] used nitric acid to obtain stable alumina/magnesia† suspensions. The

powder had an average particle diameter of 0.46 µm and a surface area of 9.4 m2/g. The re-

sulting pH was 4 - 4.2. Again high sintered densities up to 99.7% were reached.

α-Alumina/zirconia composites were made by Chang et al. [26]. AKP-50‡ powder was used as

alumina source. This powder has an average particle diameter of 0.2 µm and a surface area of

9.9 m2/g. For the zirconia, TZ-3Y powder§ with a mean particle diameter of 0.30 µm and a

BET surface area of 7.7 m2/g was used. Chang et al. prepared three slurries, a slurry dispersed

by electrostatic forces at pH 4, a flocculated slurry at pH 9 without salt addition and a coagu-

lated (weakly flocculated) slurry at pH 4 with 2M NH4Cl. For the second and third system, pH

adjustments were made after sonification. After centrifugation, the coagulated system gave the

highest green density (58.0%), while the flocculated suspension gave the lowest green density

(42.4%).

The addition of large concentration of electrolytes (the coagulated slurry case) can signifi-

cantly increase the viscosity of an otherwise dispersed slurry. Hereby the mass segregation of

the α-alumina and zirconia parts of the suspension was prevented. The conclusion is therefore

that coagulated slurries with large salt additions are beneficial for preparing homogeneous

dense ceramic composites.

* Reynolds Aluminium Co, Arkansas, USA.† Ceralox, Tuscon, Arizona, USA.‡ AKP series, Sumitomo, Tokyo, Japan.§ Tosoh, Tokyo, Japan.

Colloidal processing of ceramic membrane supports. General introduction 49

Roeder et al. [14] used dilute mixtures (8 vol-% solids) for the preparation of dense alumina

composites with CeO2-ZrO2 and Al2O3-platelets. As main alumina powder (the matrix),

AKP-30 was used with an average particle diameter of 0.4 µm and a surface area of 6.5 m2/g.

The resulting specimens were sintered at 1600°C. No further densification behaviour was

mentioned.

In this case segregation behaviour during consolidation is even more pronounced than in the

work of Chang et al. [26] In the work of Roeder [14] prevention of segregation is described in

terms of particle drafting during consolidation. In the case of particle drafting one particle

“carries” one or more other particles by which a joint effect in particle movement is created

during the consolidation step.

3.2 Stabilisation with polyelectrolytes

In two articles, J. Cesarano III et al. [23, 24] described the stability of aqueous α-alumina sus-

pensions stabilised with PMAA and PAA polyelectrolytes. The powders used were AKP-20

with a mean particle diameter of 0.52 µm and a surface area of 4.5 m2/g and AKP-30 as de-

scribed above. The electrolytes used were the Na-salt of PMAA with an average molecular

weight of 15000 g/mol and PAA of various molecular weights (1800, 5000 and 50 000 g/mol).

The structural formulas of the polymers are shown in Figure 5.

Chapter 350

The effect of the pH on suspension

stability was measured and is

shown in a so-called stability map,

see Figure 6.

Using polyelectrolytes, powder

loadings as high as 60 vol-% were

obtained in stable suspensions,

which is extremely high. A possi-

ble drawback of the use of poly-

electrolytes, however is the inher-

ent pollution with sodium when

using the sodium salt of PMAA.

For obtaining dense ceramics this

seems to be no problem, because

99% dense AKP-30 based ceram-

ics were obtained after sintering at

1350°C. However it might be bet-

ter to use the ammonium salt of

PMAA, NH4+ PMAA (here called APMA). This component can be burned out completely

during sintering and thereby any contamination of the sintered compact avoided.

Steinlage et al. [25] indeed used electrosteric stabilisation with APMA for α-alumina suspen-

sions to prepare dense tube- and gear-shaped ceramic components by centrifugal slipcasting.

The aqueous suspensions contained 20-33 vol-% solids and 8 vol-% APMA** and were

brought at pH 9.5 with NH4OH after which the slurries were treated with ultrasound.

3.3 Conclusions

All in all, the results from the research mentioned above show that it is possible to obtain good

compacts by suspension processing using a centrifugal force. However, centrifugation has

been limited to dense systems up to now and for preparing porous systems challenges arise as

** Darvan C, R.T. Vanderbilt Co. Inc., Norwalk, CT, USA.

Figure 5: Structural formulas of PMAA and PAA [23].

Figure 6: Stability map of PMAA/alumina suspension as a

function of the pH for 20 vol-% AKP-30 [23].

Colloidal processing of ceramic membrane supports. General introduction 51

discussed in chapter 4 of this thesis, which treats the synthesis of flat and tubular membrane

supports using colloidal techniques.

4. References

1. J. Smid, C.G. Avci, V. Günay, R.A. Terpstra and J.P.G.M. van Eijk, “Preparation and Characterisation of

Microporous Ceramic Hollow Fibre Membrane” J. Membrane Sci., 112 85-90 (1996).

2. R. Moreno, “The Role of Slip Additives in Tape-Casting Technology: Part I - Solvents and Dispersants”,

Am Ceram. Soc. Bull., 71 [10] 1521-31 (1992).

3. R. Moreno, “The Role of Slip Additives in Tape-Casting Technology: Part II - Binders and Plasticizers”, Am

Ceram. Soc. Bull., 71 [11] 1647-57 (1992).

4. P.M. Biesheuvel and H. Verweij, “Ceramic Membrane Supports, Permeability, Tensile Strength and Stress”,

J. Membrane Sci. 156 141-52 (1999).

5. F.F. Lange and K.T. Miller, “Pressure Filtration, “Consolidation Kinetics and Mechanics”, Am. Ceram. Soc.

Bull., 66 1498-504 (1987).

6. F.F. Lange, “Powder Processing Science and Technology for Increased Reliability”, J. Am. Ceram. Soc., 72

3-15 (1989).

7. W. Huisman, T. Graule and L.J. Gauckler, “Centrifugal Slip Casting of Zirconia (TZP)”, J. Europ. Ceram.

Soc., 13 33-39 (1994).

8. G. Steinlage, R. Roeder, K. Trumble, K. Bowman, S. Li and M. McElfresh, “Preferred Orientation of

BSCCO via Centrifugal Slip Casting”, J. Mater. Res., 9 [4] 833-36 (1994).

9. P. Hidber, F. Baader, Th. Graule and L.J. Gauckler, “Sintering of Wet-Milled Centrifugal Cast Alumina”, J.

Europ. Ceram. Soc., 13 211-19 (1994).

10. F.F. Lange, “Forming a Ceramic by Flocculation and Centrifugal Casting”, United States Patent, 4.624.808

(1986).

11. A. Yamakawa, Y. Doj, M. Miyake, “Formation of a Ceramic Composite by Centrifugal Casting”, United

States Patent, 5.262.366 (1993).

12. J.S. Moya, A.J. Sánchez-Herencia, J. Requena, R. Moreno, “Functionally Gradient Ceramics by Sequential

Slip Casting”, Mater. Lett. 14 333-35 (1992).

13. W. Huisman, T. Graule and L.J. Gauckler, “Alumina of High Reliability by Centrifugal Casting”, J. Europ.

Ceram. Soc., 15 811-21 (1995).

14. R.K. Roeder, G.A.Steinlage, K.P. Trumble and K.J. Bowman, “Preventing Segregation during centrifugal

Consolidation of particle suspensions: Particle Drafting”, J. Am. Ceram. Soc., 78 [9] 2367-73 (1995).

15. W. Huisman, T. Graule and L.J. Gauckler, “High Quality Ceramics by Centrifugal Slip Casting”, Third

Euro-Ceramics, V.1 537-42 (1993), P. Durán ed.

16. A. Royer, “Horizontal Centrifugation: A Technique of Foundry Well Adapted to the Processing of High

Reliability Pieces”, J. Mater. Shaping Techn. 5 197-209 (1988).

Chapter 352

17. L. Northcott and V. Dickin, “The Influence of Centrifugal Casting (Horizontal Axis) upon the Structure and

Properties of Metals”, J. Inst. Metals, 70 301-23 (1944).

18. P.K. Bachmann, P. Geittner, E. Krafczyk, H. Lydtin and G. Romanowski, “Shape Forming of Synthetic Sil-

ica Tubes by Layerwise Centrifugal Particle Deposition”, Ceram. Bull., 68 [10] 1826-31 (1996).

19. D.J. Shaw, “Introduction to Colloid and Surface Chemistry”, 3rd edition, Butterworths, London, 1980.

20. R.G. Horn, “Particle Interactions in Suspensions”, pp 58-101 in: Ceramic Processing, R.A. Terpstra,

P.P.A.C. Pex and A.H. de Vries (eds.), (1995).

21. D.H. Everett, “Basic Principles of Colloid Science”, 2nd ed., John Wiley & Sons, Inc., New York (1995).

22. P. Somasundaran, B. Markovic, S. Krishnakumar, and X. Yu, “Colloid Systems and Interfaces – Stability of

Dispersions Through Polymer and Surfactant Interaction”, Ch. 14 in “Handbook of Surface and Colloid

Chemistry”, K.S. Birdi (ed.), CRC Press, Boca Raton, Florida (1997).

23. J. Cesarano III and I.A. Aksay, “Processing of Highly Concentrated Aqueous α-Alumina Suspension Stabi-

lized with Polyelectrolytes”, J. Am. Ceram. Soc., 71 [12] 1062-67 (1988).

24. J. Cesarano III, I.A. Aksay and A. Bleier, “Stability of Aqeous α-Al 2O3 Suspensions with Poly(methacrylic

acid) Polyelectrolyte”, J. Am. Ceram. Soc., 71 [4] 250-55 (1988).

25. G.A. Steinlage, R.K. Roeder, K.P. Trumble and K.J. Bowman, “Centrifugal Slip Casting of Components”,

Am. Ceram. Soc. Bull., 75 [5] 92-94 (1996).

26. J.C. Chang and B.V. Velamakanni, F.F. Lange and D.S. Pearson, “Centrifugal Consolidation of Al2O3 and

Al2O3/ZrO2 Composite Slurries vs Interparticle Potentials: Particle Packing and Mass Segregation”, J. Am.

Ceram. Soc., 74 [9] 2201-4 (1991).

&KDSWHU

&ROORLGDOSURFHVVLQJRI FHUDPLFPHPEUDQHVXSSRUWV

7KHH[SHULPHQWDOSDUW

This chapter is split in two parts. The first part will briefly treat the preparation of flat ceramic

membrane supports by colloidal processing. In our laboratory, these supports are used to study

stability and gas separation properties of microporous silica membranes because they are easy

to prepare and demand less complex testing equipment.

The second part of this chapter will treat in detail the, more cumbersome, preparation of high

quality tubular membrane supports. This geometry is necessary for upscaling to process in-

dustry. Therefore some research has been performed on the production of high quality tubular

membrane supports. This part has been published in a concise form [1].

3DUW7KHSUHSDUDWLRQRI IODWVXSSRUWV

1. Introduction

As already mentioned in chapter 3, several methods exist to produce flat membrane supports,

as die pressing and tape casting. Relatively new is colloidal filtration that is discussed in this

section. For the preparation of α-alumina supports, a colloidal suspension is obtained by dis-

persing the suitable powder in an aqueous solution using ultrasound. The resulting suspension

is poured out in moulds and the solution is separated from the powder particles using a pres-

sure difference as driving force. The resulting cake is dried, released from the mould and sin-

tered at the desired temperature. The advantage of colloidal filtration is the ease of preparation

and the high degree of homogeneity in the resulting cast, which makes them excellent sup-

ports for molecular sieving silica membranes.

Chapter 454

2. The experimental part

The starting α-Al 2O3 powders were AKP-30 and AKP-15* with a mean particle diameter of

0.40 and 0.62 µm and a BET surface of 6.2 m2/g and 3.5 m2/g respectively. Both powders

have narrow particle size distributions of (1.5 wt-%<0.25 µm + 95 wt-%<1 µm) and (1.5

wt-%<0.27 µm + 89 wt-%<1 µm), respectively and a chemical purity of >99.99% as stated by

the producer. Because of the above-mentioned properties, the powders are excellent starting

materials for the preparation of membrane supports.

A 50 wt-% suspension is obtained by dispersing the α-alumina powder either in a 0.02M ni-

tric acid solution for the AKP-30 powder or a 0.02M nitric acid solution, mixed with Poly Vi-

nyl Alcohol [PVA†] (5 g/l) for the AKP-15 powder and using ultrasonic treatment‡ for

15 minutes. From the resulting suspension, the liquid is removed by filtration using water-jet

evacuation. Polyester filters§, consisting of a biological mixture of cellulose nitrate and cellu-

lose acetate, with a pore diameter of 0.8 µm are used. The resulting filter cake (cast) is dried

overnight at ambient temperature and fired at 1100°C [AKP-30] or 1150°C [AKP-15] for 1

hour with a heating/cooling rate of 2°C/min.

After firing, the supports are machined to the required dimensions and polished until a shiny

surface is obtained. Before use the supports are cleaned in ethanol by ultrasonic treatment to

remove any loose particles in the pores caused by machining of the samples. After cleaning

the supports are fired at 800°C to remove any ethanol left in the pores and to release stresses

induced by machining and polishing. Typical dimensions are a diameter of 39.0 mm and a

thickness of 2.0 mm.

3. Properties

The porosity of the resulting flat supports was measured after firing of AKP-30 and AKP-15

at respectively 1100ºC and 1150ºC with the Archimedes method by immersion in mercury.

* Sumitomo Chemical Company, Ltd, Japan.† E. Merck, Darmstadt, Germany.‡ Model 2520 Sonifier, Branson Ultrasonics Corporation, Danbury, USA.§ ME 27, Schleicher & Schuell, Dassel, Germany.

Colloidal processing of ceramic membrane supports. The experimental part 55

The sintered compacts had a porosity of 32%. No difference in porosity was found between

the AKP-30 and AKP-15 supports. Their pore-size distributions, measured by mercury poro-

simetry*, are given in Figure 1. The mean pore-radii were found to be 40 and 80 nm respec-

tively.

Hydrogen permeances were measured following the dead-end permeation method and showed

to be 8*10-7 mol/m2sPa at 500ºC for AKP-30 supports and 2*10-6 mol/m2sPa for the AKP-15

supports. No pressure testing was performed, but the tensile strength was measured in a four

point bending set-up [2]. Results are that the flat ceramic supports are strong enough to with-

stand typical pressures used in lab-scale testing of the resulting membranes (i.e. at least up to

5 bars pressure difference).

4. Conclusions

Colloidal filtration showed to be a very convenient way of preparing high quality flat mem-

brane supports. These supports are extremely homogeneous as can be seen from the very

* Series 200, Carlo Erba, Milan, Italy.

Pore-size distribution flat supports

0

10

20

30

40

50

60

70

80

90

100

0 20 40 60 80 100 120 140

radius (nm)

pore

vol

ume

(cc/

g)

AKP-15

AKP-30

Figure 1: Pore size distribution of AKP-30 and AKP-15 supports made by colloidal filtration.

Chapter 456

sharp pore size distribution. As already stated in chapter 3 the homogeneity of the supporting

system is very important because it will result in a very narrow pore-size distribution. This is

necessary for a good control of the dip coating process, because of the capillary forces in-

volved in this process. During dip coating a broad size distribution might influence the for-

mation of the layer and thereby the integrity of the final membrane. Moreover, a homogeneous

support will exhibit a very low surface roughness after polishing, so that very thin layers can

be coated on the support.

The above mentioned advantages make the supports very suitable for the preparation of flat

microporous silica membranes for lab-scale tests. However, due to the almost perfect particle

packing, the hydrogen permeance may be too low for application in process industry. For sta-

bility testing, on the other hand, the permeance of the membranes is of a far lower importance

than the selectivity of the layer under investigation. More information about stability testing

can be found in chapter 5 and 6 for the γ-alumina and the silica layer respectively.

3DUW7KHSUHSDUDWLRQRI WXEXODUPHPEUDQHVXSSRUWV

5. Introduction on centrifugal casting

Porous α-Al 2O3 tubes are frequently used as support for inorganic membranes. The conven-

tional way of producing such tubes is by extrusion or isostatic pressing followed by sintering.

These techniques are fully accepted for the production of dense ceramic tubes, but may be less

suitable for the production of porous membrane supports. Especially the occurrence of un-

roundness, inhomogeneities and a considerable surface roughness may impose problems in

both cases. For the application of defect-poor meso- and micro-porous membrane layers for

gas separation [1,4] a very smooth inner surface together with a narrow pore-size distribution

of the membrane support tube is needed [5].

Ceramic tubes can also be prepared by centrifugal casting (CC) of colloidal particles [6-8]. In

this process, a powder is dispersed in a liquid with a stabilising agent, followed by rotating for

some time in a cylindrical mould around its axis. The resulting cast is dried, released from the

mould and slightly sintered. If particles are used with a narrow size distribution and a low de-

gree of agglomeration one may expect the formation of a nearly random close packed (RCP)

green compact [9]. This requires the use of a proper colloidal stabiliser at a concentration such

Colloidal processing of ceramic membrane supports. The experimental part 57

that the particles stay well-dispersed in the liquid but form a coherent rigid structure in the

compact. Examples of possible stabilisers are nitric acid [10-12] or polyacrylate-based prod-

ucts [8, 13, 14]. If the concentration of stabiliser is too low the particles will already flock in

the liquid and form a low-density compact that will exhibit a rough surface. At higher stabi-

liser concentrations the dispersion may become too stable so the compact remains fluid-like

[15] and redispersion might occur as soon as the rotation stops.

At optimum conditions the compact shape will closely follow the cylindrical mould shape

which can be made with roundness close to perfection. In addition the surface roughness of

the inside surface of the compact can be expected to be of the order of the particle size. Sin-

tering mainly serves to obtain sufficient strength by the formation of necks without significant

grain growth and shrinkage. In this section process optimisation of porous membrane support

tubes by CC of high-quality α-Al 2O3 particles is described.

6. Experimental

The starting α-Al 2O3 powders were the same as for the preparation of flat supports, AKP-30

and AKP-15. To obtain tubes with 2 mm wall thickness and ~20 mm diameter, 120 gram of

powder was mixed with different amounts of APMA (Ammonium PolyMethAcrylate aqueous

solution, Darvan C*) and distilled water. The mixture of water and APMA, 120 ml in total,

was brought on pH = 9.5 by adding (~1.5 ml) concentrated ammonia†. The resulting suspen-

sion was ultrasonically‡ treated for 15 minutes using a frequency of 20 kHz and a transducer

output power of 100 W. This suspension was used to prepare tubes with three different

lengths: short, 6&10 cm, tubes in a home-built apparatus, using steel moulds and long tubes

(16 cm) in a commercial centrifuge§ using Delrin** moulds. The inner diameter of the tubes

was ~20 mm diameter. Before pouring the suspension into the moulds, the moulds were

coated at the inside with a solution of Vaseline†† in petroleum ether (boiling range 40-60°C)

* R.T. Vanderbilt Company, Inc., Norwalk, USA.† E. Merck, Darmstadt, Germany.‡ Model 250 Sonifier, Branson Ultrasonics Corporation, Danbury, USA.§ CEPA, GLE, Carl Padberg GmbH, Lahr, Germany.** Du Pont de Nemours, Dordrecht, The Netherlands.†† Elida Fabergé, Bodegraven, The Netherlands.

Chapter 458

to obtain easy mould release. The tubes were centrifuged for 20 minutes at 20.000 rpm and the

remaining liquid was poured out of the moulds afterwards. The green tubes were horizontally

dried inside the moulds in a climate chamber* for two days at 30°C and 60% relative humid-

ity. After drying the green tubes were removed from the moulds and sintered horizontally on a

flat support at 1150°C for 1 hour with a heating/cooling rate of 1°C/min.

To study the influence of the amount of APMA on the drying and sintering behaviour of the

AKP-30 tubes a series with different APMA concentrations in the suspension was made.

* Heraus Vötch, Ballingen, Germany.

Figure 2: AKP-30 tubes made by centrifugal casting: 1,3 with sinter warping and cracking defects ([APMA]

= 417 kg/m3) and 2,4 without visible processing defect ([APMA] = 167 kg/m3).

Colloidal processing of ceramic membrane supports. The experimental part 59

7. Results

The suspension mixtures of

120 gram powder and 120 ml stabi-

lising liquid were sufficient for two

tubes with a length of 16 cm (and a

wall thickness of 2 mm). The wall

thickness of the supports could be

varied from 1 to 2 mm at least, de-

pending on solid concentration. It

was found that porous tubes, visually

free of processing defects could be

obtained only with a certain, opti-

mum, APMA concentration,

[APMA], in the suspension. With

[APMA] below the optimum, drying

cracks were observed after drying. At [APMA] higher than the optimum, typical defects were

obtained such as surface corrugation and excessive warping and cracking during sintering.

Examples are given in Figure 2.

7.1 Influence of binder concentration

The results of the study on the influence of the APMA concentration on the quality of AKP-30

tubes after sintering are summarised in Table 1. It was found that [APMA] = 167 kg/m3 (ad-

dition of 20 ml APMA) gave optimal results. The AKP-30 results enabled us to use a less ex-

tended optimisation procedure for AKP-15, resulting in an optimum of [APMA] = 83 kg/m3

(10 ml APMA).

[APMA]

(kg/m3)

Observations

0 No suspension possible

8 Low green strength; green tube difficult to release

42 Better green strength; some surface roughness

83 Good green strength; some surface roughness

100 Ibid

167 No visible processing defects

250 Ibid

292 Reasonable quality; some sinter-cracking

333 Some sinter-cracks; some surface roughness

417 Considerable cracks after sintering, warping and

surface roughness

Table 1: Influence of liquid phase [APMA] on the quality

of sintered porous AKP-30 tubes.

Chapter 460

7.2 Properties

The porosity of AKP-30 (AKP-15) tubes,

made with optimum [APMA] was 42.5%

(43.2%)* after firing at 500°C, measured with

the Archimedes method by immersion in mer-

cury. The sintered compacts had a porosity of

34.8% (34.5%). Their pore-size distributions,

measured by mercury porosimetry† are given

in Figure 3. The mean pore radius was found

to be 60 (92) nm.

* The numbers between parentheses refer to AKP-15 tubes.† Series 200, Carlo Erba, Milan, Italy.

Pore size distribution tubular supports

0

20

40

60

80

100

120

140

0 20 40 60 80 100 120 140

radius (nm)

pore

vol

ume

(cc/

g)

AKP-30AKP-15

Figure 3: Pore-size distribution of AKP-30 and AKP-15 tubes made by CC at optimum conditions.

Figure 4: Roundness diagram of an AKP-30

centrifuged tube.

Colloidal processing of ceramic membrane supports. The experimental part 61

The surface roughness* of the tubes was found to be ~0.25 µm for the inside and ~0.9 µm for

the outside. The mean unroundness† (deviation from a perfect circle) was ~0.025 mm, based

on a 100 point measurement; the unroundness

diagram is shown in Figure 4. In this diagram

the drawn line around the concentrical circles

gives the deviation from a perfectly round ob-

ject. Figure 4 shows a slight elliptic deforma-

tion, possibly caused by “gravitational stress”

during sintering. For comparison: the unround-

ness of a typical extruded α-Al 2O3 tube was

measured to be 0.16 mm as shown Figure 5.

The surface roughness of this tube was ~6 µm.

8. Discussion and conclusion

With the CC technique excellent tubular membrane supports can be prepared with a very low

surface roughness (0.25 µm). The roundness of tubes was found to be 6× better than that of a

typical extruded tube. The roundness can possibly be improved further if special attention is

paid to mould roundness and drying and/or sintering is done vertically or in a rotating set-up.

This roundness is very important for application in reactors. If the tubes are glass-soldered in

ring-shaped machined flanges, a good roundness may result in a minimal and evenly solder-

filled space between the tube and the flange. This, in turn, will generally result in a better

sealing process, quality and stability. If deformable (graphite) gaskets are used in a removable

flange, a large unroundness will result in a radially inhomogeneous stress distribution in the

tubes near the sealing, increasing the risk of brittle fracture.

The wall-thickness of 2 mm provides the CC tubes with sufficient mechanical strength to

withstand gas and liquid pressures that are common in membrane technology. The measured

pore diameters of 120 and 184 nm are well in the range used for mesoporous membrane

preparation and it is expected that γ-alumina layers can be applied on the supports by conven-

* Mitutoyo Surftest III, Mitutoyo Mfg C0., Ltd., Tokyo, Japan.† MC 850, Carl Zeiss, Oberkochen, Germany.

Figure 5: Roundness diagram for a typical ex-

truded α-Al 2O3 tube.

Chapter 462

tional dip-coating and without the need of further intermediate layers [4]. The minimum

membrane thickness that can be obtained defect-free can be expected to be of the order of the

support roughness. This leads to the conclusion that the thickness of membranes on CC tubes

can be 24× less than those on extruded tubes. This, in turn, may result in a large flow increase

for gases and liquids.

The best tube quality was obtained with an optimum APMA concentration that is proportional

to the specific surface area of the powders. In the present experiments (with α-Al 2O3 powders)

the optimum ratio between [APMA] and specific surface area was found to be ~0.03 kg2/m5.

An [APMA] of 8 kg/m3 only, showed to be sufficient for electrosteric stabilisation and a

rather stable suspension but also resulted in some drying cracks and roughness on the inside

tube surface. This is likely to be caused by the fact that the suspension is partly flocked, lead-

ing to a poor particle packing that densifies significantly during drying. In addition it was

found that the green strength was insufficient at low [APMA]. This can be ascribed to poor

particle packing too, but it is more likely that APMA acts as a polymeric binder. At optimum

[APMA], tubes can be prepared with sufficient green handling strength, which exhibit no sur-

face roughness or cracking during drying or sintering. With higher [APMA], the green state

shows no visual defects, but significant warping and cracking is obtained during sintering.

This observation can be explained best by the presence of internal stresses in the green state

caused by green state handling, or thermal processing. These stresses neither relax nor lead to

cracks in the green state because of the combined effect of particle packing, close to RCP and

a significant amount of interparticle bonding.

The occurrence of drying cracks is quite familiar in ceramic technology. On the other hand,

warping and sintering cracks due to internal stresses is less frequently reported. However it

may be expected, that both phenomena are actually quite common in many practical ceramic

processes. In such processes, compact homogeneity is often lower than in the present case so

that the identification of processing defects as mentioned, may become obscured. Less dense

areas in the green state cause (local) drying cracks; more dense areas cause (local) sintering

stress. The observations made can be generalised for all ceramic suspension processing since

they are not limited to shapes made with APMA and may be reproduced, for instance, when

HNO3 together with PVA is used as a stabiliser.

Colloidal processing of ceramic membrane supports. The experimental part 63

9. Centrifugal injection casting

Next, experiments are discussed using the so-called Centrifugal Injection Casting (CIC) tech-

nique [16]. Here, the same centrifuge as for conventional centrifugal casting is used, but the

mould is not filled with the suspension before the experiment, but the suspension is sprayed

into the mould during centrifuging. This method was developed by Bachmann et al. [6] for the

production of silica tubes for optical glass fibres. In this method it is possible to change the

suspension composition during spraying which results in property changes of the produced

tubular part over the cast radius. For membrane support applications, this might be a favour-

able option, because one can start with a very coarse porous outside layer and end up with a

layer with smaller pores at the inside of the tube. In this way one can combine very high fluxes

(coarse porous layer) with a very smooth surface (inside layer) which is a suitable support for

coating high quality membrane layers.

9.1 Experimental

Two types of tubes have been prepared, single powder tubes (AKP-30 tubes) and tubes pre-

pared with two different powders (AKP-30 and CR-1*). This CR-1 powder has a broad parti-

cle size distribution 5% < 0.6 µm and 95% < 3 µm with a mean particle diameter of 1.5 µm

and a BET specific surface area of 4 m2/g as stated by the producer.

For the single powder AKP-30 tube the following recipe was used. AKP-30 and 0.02M HNO3

were mixed in a 1:1 weight ratio. After mixing, the suspension was treated with 175 Watt ul-

trasound for 15 minutes. The obtained suspension was filtered through a 200 µm steel filter

and under magnetic stirring more 0.02 M nitric acid is added until a final weight composition

of AKP-30:HNO3 = 1:5. Because the suspension is not very stable, stirring had to be contin-

ued until the end of the preparation, including the injection process.

Injection is started in an empty mould when the final rotational speed of 20.000 rpm is

reached. The movement of the injection tube is controlled by a commercial stepmotor sys-

tem†, this to ensure a smooth and reproducible movement of the injector tube. A typical axial

* Baikolox CR-1, Baikowski Chimie, Annecy, France.† Festo BV, Delft, The Netherlands.

Chapter 464

speed of the injector is 3.4 cm/s, while a peristaltic pump* is used to inject the suspension in

the mould. Injection can be terminated when the desired amount of suspension has been in-

jected. With the powder loading of the suspension one can easily calculate the thickness of the

formed compact. Drying and sintering is essentially the same as described in section 6 for the

conventional casted (CC) tubes.

With the same method also tubes consisting of a CR-1 outside layer and an AKP-30 inside

layer were prepared. The CR-1 suspension was produced in essentially the same way as de-

scribed above for the AKP-30 suspension. Injection also took place in the same way as the

single powder AKP-30 tube. The process started with injecting the CR-1 suspension and af-

terwards the AKP-30 suspension was injected. An SEM† photograph of the resulting tube is

shown in Figure 6.

* Masterflex, model 7523-01, Barnant, Barrington, UK.† Jeol, JSM 5800, Tokyo, Japan.

Figure 6: SEM micrograph of CIC-tube consisting of an outer CR-1 part (light, striped) and an inner

AKP-30 part (grey, even).

Colloidal processing of ceramic membrane supports. The experimental part 65

Because the CR-1 powder has a rather broad particle size distribution, also some segregation

in the subsequent sprayed layers can be found. This segregation is nicely shown in Figure 7,

which is a close-up of the CR-1 (outside) part of the SEM picture shown in Figure 6.

9.2 Results

AKP-30 tubes made by CIC had a porosity of 28%, which is somewhat lower than the con-

ventional cast AKP-30 tubes. This is most probably due to the way of stabilising the suspen-

sion. In the case of conventional casting the suspension is stabilised by a relatively large

amount of polymeric stabiliser and the suspension is most likely to be partly flocked. This

flocculation will cause some extra porosity in the green compact. This porosity is not removed

during the firing process and a more porous support will result. In the case of nitric acid stabi-

lisation the suspension might be less flocked. The resulting green and sintered density of the

compacts shall therefore be higher. The surface roughness of CIC tubes is somewhat larger

(by optical examination) than the roughness of the CC tubes. This is most probably due to the

Figure 7: Close-up of the CR-1 part of the combined AKP-30/CR-1 tube.

Chapter 466

spraying action of the injection tube. This spraying causes some turbulence during centrifug-

ing which disturbs the formation of a smooth inside surface of the tube in production.

9.3 Conclusions on Centrifugal Injection Casting

With CIC, it is possible to produce tubes without visual defects, but when using only a single

powder, the conventional centrifugal casted tubes showed not only a higher porosity, but also

a somewhat smoother inside surface. However, the largest advantage of using the CIC tech-

nique is the possibility of creating multilayer tubes in just one production step. One is able to

build up tubes which have a very coarse outside, ensuring high permeability and a smooth in-

side surface, ensuring the right conditions for coating high quality membrane layers. Moreover

with the CIC technique one has the possibility to apply layers with different properties, like

catalytic active layers. So, depending on the application of the tubes, one can choose either to

use the CC technique to obtain a very homogeneous, rather porous single powder tube or to

use the CIC technique to obtain graded tubes or layered tubes with special properties of a

somewhat lower quality.

10. Perspectives

It may be questioned whether the high degree of perfection of the CC tubes justifies the higher

costs in mass-scale production and the limitation to circular shapes when compared to extru-

sion processes. Extrusion processes are cheap, continuous and enable more complex shapes

such as multi-bore tubes. On the other hand, the CC technique allows a radially varying con-

centration and morphology of composition and particle morphology. The use of a suspension

that consists of largely different sizes of particles may automatically result in specific radial

variations that can be predicted quantitatively on basis of the method described in [17]. How-

ever, these suspensions might be more difficult to stabilise and in general there are less de-

grees of freedom for varying properties of the produced tubes.

With the CIC technique it is possible to inject small amounts of suspension layerwise so that

all thinkable radial distributions can be realised. Suspensions with large particles and suspen-

sions with smaller particles can be stabilised seperately, which will be easier than stabilising a

suspension with a large variation in particle size. Multilayer membranes and solid fuel cell

Colloidal processing of ceramic membrane supports. The experimental part 67

structures can be made in this way with radial variations of transport properties and catalytic

activity to obtain innovative solutions for dedicated reactor and separation problems.

11. Acknowledgement

The author wishes to express his thanks to Dr. P. Geittner (Philips Forschungslabor Aachen)

for the fruitful discussions on centrifugal (injection) casting.

12. References

1. A. Nijmeijer, C. Huiskes, N.G.M. Sibelt, H. Kruidhof and H. Verweij, “Centrifugal Casting of Tubular

Membrane Supports”, Am. Ceram. Soc. Bull. 77 [4] 95-98 (1998).

2. P.M. Biesheuvel, H. Verweij, “Design of Ceramic Membrane Supports: Permeability, Tensile Strength and

Stress”, J. Membrane Sci. 156 [1] 141-52 (1999).

3. R.J.R. Uhlhorn, M.H.B.J. Huis in ‘t Veld, K. Keizer and A.J. Burggraaf, “Synthesis of Ceramic Membranes.

Part 1. Synthesis of Non-Supported and Supported Gamma-Alumina Membranes without Defects”, J. Ma-

ter. Sci., 27, 527-37 (1992).

4. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, “Formation and Characterization of Sup-

ported Microporous Ceramic Membranes by Sol-Gel Modification Techniques”, J. Membrane Sci., 99, 57-

75 (1995).

5. R.M. de Vos and H. Verweij, “Improved Performance of Silica Membranes for Gas Separation”, J. Mem-

brane Sci., 143 [1] 37-51 (1998).

6. P.K. Bachmann, P. Geittner, E. Krafczyk, H. Lydtin and G. Romanowski, “Shape Forming of Synthetic Sil-

ica Tubes by Layerwise Centrifugal Deposition”, Am. Ceram. Soc. Bull., 68 [10] 1826-31 (1989).

7. P.K Bachmann, P. Geittner, H. Lydtin, G. Romanowski and M. Thelen, “Preparation of Quartz Tubes by

Centrfugational Deposition of Silica Particles”, pp 449-52 in: Proc. 14th. Europ. Conf. Optical Commun.

Vol.1 Inst. Electr. Eng., London, Sept 11-15 (1988).

8. G.A. Steinlage, R.K. Roeder, K.P. Trumble and K.J. Bowman, “Centrifugal Slipcasting of Components”,

Am. Ceram. Soc. Bull., 75 [5] 92-94 (1996).

9. E.A. Barringer and H.K. Bowen, “Formation, Packing, and Sintering of Monodisperse TiO2 Powders”, J.

Am. Ceram. Soc., 62 [12] C199-C201 (1982).

10. W. Huisman, T. Graule and L.J. Gauckler, “Alumina of High Reliability by Centrifugal Casting”, J. Europ.

Ceram. Soc., 15 811-21 (1995).

11. P. Hidber, F. Baader, T. Graule and L.J. Gauckler, “Sintering of Wet-Milled Centrifugal Cast Alumina”, J.

Europ. Ceram. Soc., 13 211-19 (1994).

Chapter 468

12. J.C. Chang, B.V. Velamakanni, F.F. Lange and D.S. Pearson, “Centrifugal Consolidation of Al2O3 and

Al2O3/ZrO2 Composite Slurries vs Interparticle Potentials: Particle Packing and Mass Segregation”, J. Am.

Ceram. Soc., 74 [9] 2201-4 (1991).

13. J. Cesarano III, I.A. Aksay and A. Bleier, “Stability of Aqueous α-Alumina Suspensions with

Poly(methacrylic acid) Polyelectrolyte”, J. Am. Ceram. Soc., 71 [4] 250-55 (1988).

14. J. Cesarano III, I.A. Aksay and A. Bleier, “Processing of Highly Concentrated Aqueous α-Alumina Suspen-

sion Stabilized with Polyelectrolytes”, J. Am. Ceram. Soc., 71 [12] 1062-67 (1988).

15. C.P. Cameron and R. Raj, “Better Sintering through Green-State Deformation Processing”, J. Am. Ceram.

Soc., 73 [7] 2032-37 (1990).

16. C. Huiskes, A. Nijmeijer, H. Kruidhof and H. Verweij, “Porous Ceramic Supports by Centrifugal Injection

Casting”, to be submitted to J. Europ. Ceram. Soc.

17. P.M. Biesheuvel, A. Nijmeijer and H. Verweij, “A Theory of Batchwise Centrifugal Casting”, AIChE J., 44

[8] 1914-22 (1998).

&KDSWHU

7KHSUHSDUDWLRQDQGSURSHUWLHVRI

K\GURWKHUPDOO\VWDEOHγDOXPLQDPHPEUDQHV

1. Introduction

There is a growing interest in the application of inorganic membranes in high temperature gas

separation and membrane reactors. Microporous amorphous silica membranes are of particu-

lar interest for H2-separation in processes like steam reforming, the water-gas shift reaction,

dehydrogenation and coal gasification [1-3]. Selective removal of hydrogen in numerous

equilibrium-restricted processes (e.g. dehydrogenation processes) may lead to a significant

increase in conversion and yield, provided membrane permeance and selectivity are high. In

addition, by selectively removing hydrogen, strongly endothermic equilibrium-limited proc-

esses may be operated at temperatures that are significantly lower than those in conventional

reactors without loss in conversion (e.g., steam-reforming of methane).

In most of the above-mentioned processes, high operation temperatures are necessary while

the reaction atmospheres usually contain considerable amounts of steam because water is one

of the reactants, or because water is added to reduce coke formation. Also in food processing

and medical applications, steam is often used for sterilisation. Thus, in many applications the

membranes must be sufficiently stable in environments of both increased temperature and

containing steam. In this work, this is called “hydrothermal” stability.

Somewhat surprisingly, however, only a very limited amount of literature is available on hy-

drothermal stability of even the most commonly applied mesoporous membrane type, namely

γ-alumina membranes on α-Al 2O3 supports. These mesoporous γ-alumina membranes are the

common supports for the microporous silica membranes to be used in membrane steam re-

formers. In the investigations that finally led to the present study, delamination of the

γ-alumina membrane from the α-Al 2O3 supports in hot steam was found to be a major compli-

Chapter 570

cation. In a few studies [4,5] a rather large change in pore size under hydrothermal conditions

has been reported without mentioning delamination from the support. At high temperatures

changes in pore-size of the γ-alumina membrane can be related to sintering and conversion of

γ-alumina to other transition alumina forms (with increasing temperature δ-alumina and

θ-alumina) and α-alumina as described by Wefers and Misra [6]. Most authors conclude the

conversion from one low temperature phase to another, for example γ-alumina to δ-alumina

from X-ray diffraction data. These alumina forms and their transitions, however, are not well

defined and it may be difficult to distinguish between the different types. Most reliable data

are therefore available for the overall change of the transition aluminas (γ, δ and θ) to

α-alumina, which shows distinct peaks in the X-ray diffraction spectrum.

A well-known method to improve the thermal, and to a certain degree the hydrothermal sta-

bility of γ-alumina is the use of dopants. Several authors [7-16] made detailed studies of the

effects of lanthanum doping, which generally increases pore stability and impedes the phase

transition to α-alumina. A large number of doping elements was investigated by Vereshchagin

et al. [17], and, contrary to other reports, they found that lanthanum enhanced the conversion

of transition aluminas to α-Al 2O3. On the other hand the same authors reported that elements

like scandium, cerium, calcium and strontium retarded conversion. The discrepancies in the

literature made us to investigate, besides lanthanum, the effects of gadolinium and calcium on

pore stability under hydrothermal conditions.

In refractory industry, phosphate bonding is a well-known method for bonding ceramic pow-

ders together [18-21] in either a reactive or non-reactive manner. Phosphate bonding may

yield strong compacts at temperatures far below the normal sintering temperature of the pow-

ders. Much research has been conducted on the use of MonoAluminumPhosphate (MAP,

Al(H2PO4)3) for bonding alumina powders. In addition to MAP, other phosphor-containing

compounds like orthophosphoric acid (H3PO4) can be used as well. In the work described

here, phosphate bonding with MAP precursor solutions has been used to anchor the γ-alumina

membranes to the α-Al 2O3 support to inhibit delamination in the interface. The purpose of the

present paper is to describe the procedures for and effects on hydrothermal stability of La-

doping and phosphate bonding of γ-alumina membranes on α-Al 2O3 supports.

The preparation and properties of hydrothermally stable γ-alumina membranes 71

2. Experimental

All stability tests were performed on γ-alumina membranes coated on flat α-Al2O3 supports.

These supports were prepared as follows:

A 50 wt-% suspension of α-alumina powder (AKP-30)* in a 0.02M nitric acid solution was

treated with ultrasound† for 15’ in a specially designed beaker (a so-called Glass Rosett cell)

using a frequency of 20 kHz and a transducer output power of 100W. The resulting suspen-

sion was filtered on polyester‡ with 0.8 µm pore size. The resulting filter cake (cast) was dried

overnight at ambient temperature and sintered at 1100°C for 1 hour using a heating/cooling

rate of 2°C/min.

After firing, the supports were machined to the required dimensions of Ø 39 mm, 2 mm thick-

ness and polished until a shiny surface was obtained. Before use the supports were cleaned in

ethanol by ultrasonic treatment to remove any debris that remained in the pores after machin-

ing. After cleaning the supports were fired at 800°C to remove ethanol left in the pores. The

typical thickness of a final flat membrane support was 2 mm. The sintered compacts had a po-

rosity of 32%.

A MAP layer was coated on the supports according to the following procedure: A commercial

50 wt-% MAP solution§ was diluted either 10 or 20 times, further indicated as MAP10 and

MAP20, respectively. The shiny surface of a flat support was brought in contact with this so-

lution for 3 seconds, after which it was dried. Next to this pre-treatment, the supports were

coated under class 100 clean room conditions with either pure or doped 0.5M boehmite sols.

Sols were prepared by reacting 0.5 mole of aluminium-tri-sec-butoxide** (ATSB) with

70 moles of double-distilled water of 90ºC [22]. The ATSB was added drop-wise under a ni-

trogen flow to avoid premature hydrolysis. The temperature of the reaction mixture should at

least be 80ºC to avoid the formation of any Bayerite (Al(OH)3) [23]. After the addition of

ATSB, the mixture was kept at 90ºC for at least one hour to evaporate off the butanol formed.

* Sumitomo, Tokyo, Japan.† Model 2520 Sonifier, Branson Ultrasonics Corporation, Danbury, USA.‡ ME 27, Schleicher & Schuell, Dassel, Germany.§ Alfa, Johnson Matthey GmbH, Karslruhe, Germany.** Acros, 97% purity, Geel, Belgium.

Chapter 572

The mixture was subsequently cooled down to ~60ºC and peptised with 1M HNO3* at a pH of

about 2.5. During the synthesis, the sol was stirred vigorously. The peptised mixture was re-

fluxed for 20 hours at 90ºC, resulting in a very stable 0.5 molar boehmite sol with a clear

white/blue appearance.

Doping of this sol was performed by thorough mixing with the appropriate amount of a 0.3M

metal nitrate solution. The mixing was done directly before coating to avoid possible ageing

effects that have been reported in the literature, for example by Lin and Burggraaf [11]. No

such ageing studies were, however, performed in the present work.

A dip-coating solution was obtained by diluting 30 ml of (doped) boehmite sol with 20 ml of

a solution of 30 g PVA/l in 0.05M HNO3. Both solutions were filtered before use through a

0.8 µm cellulose acetate filter† to remove any large particles from the solutions. The shiny

surface of a polished α-Al 2O3 support was then brought in contact with dip-coating sol for

3 seconds, using a dedicated dip-coating apparatus‡ and class 100 cleanroom conditions. After

dip-coating, the membranes were dried and fired in air at temperatures between 600ºC and

1000ºC for three hours with a heating/cooling rate of 1ºC/min. Unsupported bulk membrane

material was obtained for specific surface area and XRD-characterisation by drying the dip-

coating solutions and subsequently firing the dried gels at the above-mentioned conditions. To

detect any possible reaction of MAP with La-doped γ-alumina, powder XRD samples were

prepared by mixing doped γ-alumina that was fired at 1000ºC and pure MAP that was fired at

300ºC in 50 wt-% ratio. The mixture was subsequently fired at 1000ºC for 3 hours.

The pore-size of the resulting mesoporous membranes was determined by permporometry.

This method and the home-built equipment that we used for the measurements have been de-

scribed in detail elsewhere [24]. During the measurements the following processes took place:

1. By capillary condensation the mesopores of the membranes became filled with cyclohex-

ane.

2. Starting with all pores filled, the cyclohexane partial pressure above the membrane was

reduced in intervals, gradually emptying the pores.

3. At each cyclohexane vapour pressure level, the oxygen permeance through the open pores

of the membrane was measured.

* E. Merck, Darmstadt, Germany.† FP030/50, Green rim, Schleicher & Schuell GmbH, Dassel, Germany.‡ Velterop BV, Delden, The Netherlands.

The preparation and properties of hydrothermally stable γ-alumina membranes 73

The measurements were performed at 20ºC and the oxygen concentration was determined

with a gas chromatograph*. From the oxygen permeance data as a function of cyclohexane

partial pressure the pore-size distribution was calculated with the Kelvin equation [25].

After the pore-size was established, the membranes were treated in a steel reactor with a

Simulated Ambient Steam Reforming Atmosphere (SASRA) for 100 hrs at 600°C with

H2O/CH4 = 3/1 (by volume) at 2.5 MPa total pressure. Heating and cooling was performed in

an argon atmosphere at the same total pressure at a rate of 1°C/min. In a few experiments a

pure steam treatment was carried out at 0.2 MPa total pressure at 150ºC or 300ºC in the same

manner as for SASRA treatment. A pure CO2 treatment was done likewise, but at 500ºC at

1.2 MPa pressure.

The surface area of the membrane materials was measured before and after SASRA treatment

by a single point BET instrument with a TC-detector. Three samples were run in parallel and

the amount of adsorbed nitrogen on the sample surface was measured and used for calculating

the surface area. The phase composition was characterised by XRD†. After SASRA treatment

the adherence of the membranes to the support was tested by the “Scotch Tape Test” [26]. In

this test, a piece of Scotch Tape was applied firmly with the sticky side onto the membrane

surface and torn off rapidly. If the membrane layer was torn off together with the tape, it was

concluded that delamination had occurred. For membranes that showed no sign of delamina-

tion, the pore-size was measured with permporometry for a second time.

Specific surface area measurements were performed with the unsupported bulk membrane

material before and after SASRA, to obtain information about the stability of (doped)

γ-alumina under steam-reforming conditions.

* Model 3300, Varian, Sugarland (TX), USA.† Siemens D5000, Cu Kα radiation, Essen, Germany.

Chapter 574

Figure 1a: SEM micrograph of a delaminated conventional mesoporous γ-alumina membrane after SASRA

treatment.

Figure 1b: Cross section showing that the γ-alumina layer has detached from the α-alumina support.

The preparation and properties of hydrothermally stable γ-alumina membranes 75

3. Results

Standard γ-alumina membrane-layers on α-Al 2O3 supports not treated with MAP and prepared

as described in [10] always came off in the Scotch Tape Test after SASRA treatment. As

shown in Figure 1 (a and b), in these membranes a crack is formed in the membrane-support

interface leading to delamination. When the support was treated with MAP, however, after

steam treatment no delamination was observed.

We suggest that the beneficial effect of MAP treatment resulted from chemical bonding be-

tween the membrane-layer and the support. To investigate independently the effect of the sur-

face morphology of the support on delamination, without chemical phosphate bonding, the

roughness of the support was varied by changing the grinding procedure. It was found that

increasing the roughness of the α-alumina supports did not help avoiding delamination of the

γ-alumina layer from the support during SASRA treatment.

The concentration of the MAP solution was found to be critical. Treatment with 5 wt-%

MAP-solution gave good adherence, while a 2.5 wt-% solution resulted in some delamination,

possibly due to insufficient phosphate on the surface of the supports.

Table 1 summarises the most important results from the investigation of metal doping. In this

table the results of MAP treatment are combined with effects of firing temperature and do-

ping. As can be seen in Table 1, γ-alumina membranes with pore radii as low as 2.0 nm (Kel-

vin radius) may be obtained after firing at 600°C. Note that an instrumental standard error of

0.5 nm (90% reliability) is common in permporometry. This technique should therefore only

be used for comparison purposes and to obtain a qualitative impression of the pore-size and

pore-size distribution of the material under investigation.

Chapter 576

Membrane Support Treatment Tcalc (ºC) Test conditions rKelvin (nm)

γ None 600 None 2.0

γ None 825 None 3.6

γ None 1000 None 8.7

γ MAP 10 825 None 4.2

γ MAP 10 825 SASRA 6.2

γ MAP 10 825 2×SASRA 7.5

γ +3La None 825 None 3.3

γ + 3La MAP 10 1000 None 8.4

γ + 3La MAP 10 1000 SASRA 9.3

γ + 6La MAP 10 1000 None 6.0

γ + 6La MAP 10 1000 SASRA 6.1

γ + 6La MAP 10 1000 150ºC steam 6.0

γ + 6La MAP 10 1000 300ºC steam 6.0

γ + 6La MAP 10 1000 CO2 6.3

γ + 9La MAP 10 1000 None 8.6

γ + Ca MAP 10 1000 None 7.8

γ + Ca MAP 10 1000 SASRA 13.2

γ + Gd MAP 10 1000 None 10.3

γ + Gd MAP 10 1000 SASRA 15.8

Table 1: Influence of support treatment, γ-alumina doping, membrane firing temperature and SASRA-

treatment on the pore-size of γ-alumina. MAP 10 indicates a 10 times diluted standard MAP solu-

tions, which results in an effective MAP concentration of 5 mol-%., 3La indicates a 3 mol-%

La-doped membrane, 6La indicates a 6 mol-% La-doped membrane.

The preparation and properties of hydrothermally stable γ-alumina membranes 77

All membranes without MAP treatment delaminated after SASRA treatment. Therefore no

permporometry results for these membranes are presented. The pore-growth of undoped

γ-alumina strongly depended on temperature showing a large increase in pore-size between

825 and 1000ºC. The MAP-treated membranes had somewhat larger pores after firing at

825ºC. The cause of this effect is not clear yet. For undoped γ-alumina membranes, the pores

grew during SASRA from 4.2 to 6.2 nm, and after a second SASRA treatment to 7.5 nm.

Thus, it appears that the pore-growth continues within the time scale of our SASRA treatment

experiments.

Compared to undoped materials, 3 mol-% lanthanum doping gave hardly any beneficial ef-

fects on stability (Table 1). A significant improvement was found, however, for 6 mol-% lan-

thanum doping. For this case a pore-size of only 6.0 nm was found after firing at 1000ºC and

no pore growth during SASRA treatment was observed at all. Additionally, after SASRA

treatment, the pore-size distribution of a 6 mol-% doped γ-alumina membrane was still very

narrow, as can be seen in Figure 2.

0

1E+17

2E+17

3E+17

4E+17

5E+17

6E+17

0 10 20 30

Kelvin radius (nm)

# of

por

es

Figure 2: Pore size distribution of a SASRA-treated γ-alumina membrane. The support was treated

with 5 mol-% MAP (MAP 10). The γ-alumina was doped with 6 mol-% La and sintered at

1000ºC for three hours.

Chapter 578

In an attempt to investigate the reactions between the γ-alumina membrane material and other

components, powder mixtures of γ-alumina doped with 6 and 9 mol-% La fired at 1000°C

were studied. In Figure 3 XRD patterns of these powders as prepared, and after SASRA

treatment at 600°C are shown. After firing at 1000°C the patterns did not reveal the presence

of any α-Al 2O3, but several broad peaks that can be ascribed to transition alumina. From

comparison with the XRD pattern for LaAlO3, it can be concluded that traces of LaAlO3 may

be present. After SASRA treatment, the presence of LaAlO3 is more outspoken, suggesting

that the SASRA treatment promoted formation of this phase.

In Figure 4 XRD patterns of powders mixtures of γ-alumina with 6 mol-% La and MAP are

shown. After firing at 1000°C the XRD pattern suggests that the powder consists mainly of

AlP3O9 and AlPO4 (orthorhombic phase). After SASRA treatment of the powder at 600°C a

clear change in the XRD pattern could be observed: a single dominating phase of AlPO4 (tri-

gonal, Berlinite phase) is observed together with traces of a LaPO4 phase. The SASRA treat-

ment at 600°C thus induced structural changes in the Al-P-O phases.

0100200300400500600700

0 20 40 60 80 100

2 theta values

Inte

nsity

54321

Figure 3: XRD patterns. 1. γ-alumina with 6 mol-% La-doping after calcination at 1000°C, 2. Same powder

as in 1 after additional SASRA treatment at 600°C, 3. γ-alumina with 9 mol-% La-doping after

calcination at 1000°C, 4. Same powder as in 3, but after additional SASRA treatment at 600°C,

5. Reference LaAlO3 powder.

The preparation and properties of hydrothermally stable γ-alumina membranes 79

Table 2 shows results of specific surface area measurements of the same powders as men-

tioned in Figure 3 and Figure 4.

Powders of undoped γ-alumina fired at 1000°C had a specific surface area close to 100 m2/g.

Addition of 6 mol-% La reduced the surface area with about 20%, while the reduction is about

twice as large with a 9 mol-% lanthanum addition. For all these materials the surface area

showed an increase after SASRA treatment at 600°C. The increase was largest for 9 mol-%

La addition and smallest for pure γ-alumina. Addition of MAP reduced the surface area sig-

nificantly, and opposite to the previous observations, the SASRA treatment gave a further re-

duction. The remarkable effect of MAP in reduction of surface area is possibly due to the

formation of some reaction products of the MAP with the γ-alumina.

The effect of CO2 on the stability of the membranes was tested, because of the possible insta-

bility of lanthanum compounds towards CO2 (formation of La2(CO3)2). CO2 treatment at

600ºC for 100 hours did not result in a measurable pore-growth: the pore-size increased from

6.0 nm to 6.3 nm after treatment, which is within the measurement error.

0

200

400

600

800

1000

0 20 40 60 80

2 theta values

Inte

nsity

a a

a

a

a

ba ab a+b aa a a a a a a

b

c

c

c c c c c c c d d

a

Figure 4: XRD patterns of: 1. MAP + γ-alumina with 6 mol% La-doping after calcination at 1000°C, 2.

Same material as in 1 after additional SASRA treatment. a = Al3P3O9, b = AlPO4, c = AlPO4

(Berlinite), d = LaPO4.

2

1

Chapter 580

The treatment of the above mentioned membranes in pure steam for 100 hours at 150ºC and

300ºC did not induce any pore-growth either. This is a very important result, because such a

treatment is common in steam-sterilisation.

The encouraging results that were found for the 6 mol-% lanthanum doped samples could not

be reproduced for the materials with 9 mol-% lanthanum doping. These samples had already a

Kelvin radius of 8.6 nm before SASRA treatment. Hence no further investigations were per-

formed with such highly-doped membranes.

Compared to the 6 mol-% La-doped membranes, calcium and gadolinium-doped membranes

showed larger pores and more pore growth during SASRA treatment. This indicates that the

stabilising effect of the latter metal ions is not of the same quality as that of lanthanum. These

findings could be of interest, however, for the preparation of membranes with specific pore-

sizes.

4. Discussion & Conclusions

A clear and beneficial effect of treating the α-Al 2O3 membrane support with a MAP precursor

solution on the stability of γ-alumina membrane/α-Al 2O3 support structure was observed.

Composition T, °C SASRA treatment, 600°C Surface area, m2/g

γ-alumina 1000 no 96

γ-alumina 1000 yes 99

γ-alumina + 6 mol% La 1000 no 82

γ-alumina + 6 mol% La 1000 yes 89

γ-alumina + 9 mol% La 1000 no 60

γ-alumina + 9 mol% La 1000 yes 70

MAP + γ-alumina + 6 mol% La 1000 no 12

MAP + γ-alumina + 6 mol% La 1000 Yes 7.6

Table 2: Specific surface area of different unsupported membrane materials.

The preparation and properties of hydrothermally stable γ-alumina membranes 81

BET and permporometry results show, however, that unsupported γ-alumina sintered during

SASRA treatment even with lanthanum stabilisation. Since it can be assumed that sintering is

more pronounced in the membrane than in the support, the tensile stresses may build up in the

γ/α-alumina interface. This stress manifests itself clearly in conventional membrane structures

where the γ-alumina membrane blisters off in flakes from the support after SASRA treatment.

As shown in this study, however, a pretreatment of the support with a sufficiently concen-

trated MAP solution together with 6 mol-% La-doping of the γ-alumina results in a highly

steam stable membrane-support combination. Because steam stability is largely improved

going from a 2.5 wt-% to a 5 wt-% MAP solutions, it seems reasonable to suggest that the

number of phosphate bonds is critical in order to overcome the interfacial stress. The phase

changes in the Al-P-O phases as observed by XRD studies of powder mixtures seem to have

no effect on the mechanical stability of the phosphate bonded membrane. This could mean

that the bonding is maintained even if structural changes occur or that locally, the aluminium

phosphate phase condition has reached a saturated state.

From an industrial point of view, the MAP treatment appears very promising for stabilising

layered membrane structures that are applied in processes where water vapour is present at

high temperatures. The possible large impact of steam stable (mesoporous) membranes in in-

dustry gave rise to a patent application of the described process [27].

The second issue to address, besides bonding, is the stability of the mesopores. Our results

show that doping with lanthanum can prevent pore growth to a large extent. The effect of La-

doping on γ-alumina has been subject to many studies, particularly in catalysis. The mecha-

nisms of pore-growth and phase stabilisation, however, are not fully understood. It is usually

assumed that the presence of lanthanum on the alumina surface leads to formation of micro-

domains of lanthanum aluminate on the alumina surface, which immobilises the surface ions

and thereby reduces sintering of the material and impedes the phase transition to α-alumina

[13,14]. The interested reader is referred to the paper of Oudet et al. [13] who provided a de-

tailed description of the mechanism behind lanthanum doping of γ-alumina.

A larger amount of La-doping, e.g., the 9 mol-% doped membranes, possibly gives rise to

completely crystalline LaAlO3 clusters. The formation of such crystalline clusters during sin-

tering may destroy the mesoporous structure of the material due to stresses involved. This is a

tentative explanation for the relatively large pore-size of the 9 mol-% lanthanum-doped mem-

branes. The influence of doping with Ca2+ ions is discussed in a paper by Burtin et al. [15]

Chapter 582

who found a CaAl12O19 phase in the doped material. No explanation is given for the stabilis-

ing effect of calcium, but essentially the same mechanisms as for lanthanum doping are likely

to be present.

The fact that no reports are available on the behaviour of microporous amorphous silica mem-

branes under high temperature steam containing atmospheres is not surprising in the light of

the instability of γ-alumina which is by far the most commonly used intermediate layer for

such membranes. With the availability of more hydrothermally stable supporting structures,

however, a systematic study of the intrinsic properties of microporous silica membranes in

water vapour containing atmospheres at high temperatures has recently been initiated in our

group. Additionally, we expect significant improvements in membrane stability under milder

steam conditions than applied in this study, as often encountered in steam sterilisation and

pervaporation.

As one can see from Table 1, a spin-off result of this work is a list of recipes for the prepara-

tion of membranes with different amounts of doping, covering a complete range of pore-sizes

with a resolution of 1-2 nm. This shows that we are now able to produce membranes with a

tailor-made pore-size, which may be important for retaining certain large molecules by high-

flux nanofiltration.

5. Acknowledgement

The author wishes to thank Dr. R. Bredesen (SINTEF) for performing the SASRA treatment

and the XRD and specific surface area measurements and for the fruitful discussions about

hydrothermal stability of intermediate layers.

6. References

1. R. Bredesen, “Key Points in the Development of Catalytic Membrane Reactors”, Paper no. A7.0 in Proc.

13th Int. Congr. Chem Process Eng., August 23-28 1998, Praha, Czech Republic.

2. G. Sarraco and V. Specchia, “Catalytic Inorganic Membrane Reactors: Present Experience and Future Op-

portunities”, Catal. Rev. Sci. Eng., 36 [2] 305-84 (1994).

3. J. Zaman and A. Chakma, “Inorganic Membrane Reactors”, J. Membrane Sci., 92 1-28 (1994).

4. G.R. Gallaher, and P.K.T. Liu, “Characterization of Ceramic Membranes. I. Thermal and Hydrothermal

Stabilities of Commercial 40Å Membranes”, J. Membrane Sci., 92 29-44 (1994).

The preparation and properties of hydrothermally stable γ-alumina membranes 83

5. C.H. Chang, R. Gopalan and Y.S. Lin, “A Comparative Study on Thermal and Hydrothermal Stability of

Alumina, Titania and Zirconia Membranes”, J. Membrane Sci., 91 27-45 (1994).

6. K. Wefers and C. Misra, “Oxides and Hydroxides of Aluminium”, Alcoa Technical Paper No19 (1987).

7. H. Schaper and L.L. van Reijen, “The Influence of Dopants on the Stability of Gamma Alumina Catalyst

Supports”, Mater. Sci. Monographs, 14 173-76 (1982).

8. H. Schaper , E.B.M. Doesburg and L.L. van Reijen, “The Influencce of Lanthanum Oxide on the Thermal

Stability of Gamma Alumina Catalyst Supports”, Appl. Catal. 7 211-20 (1983).

9. H. Schaper, E.B.M. Doesburg, P.H.M. de Korte, L.L. van Reijen, “Thermal Stabilisation of High Surface

Area Alumina”, Solid State Ionics, 16 261-66 (1985).

10. Y.S. Lin, K.J. de Vries and A.J. Burggraaf, “Thermal Stability and Its Improvement of the Alumina Mem-

brane Toplayers Prepared by Sol-Gel Methods”, J. Mater. Sci., 26 715-20 (1991)

11. Y.S. Lin and A.J. Burggraaf, “Preparation and Characterisation of High-Temperature, Thermally Stable

Alumina Composite Membrane”, J. Am. Ceram. Soc., 74 [1] 219-24 (1991).

12. M.F.L. Johnson, “Surface Area Stability of Aluminas”, J. Catal., 123 245-59 (1990).

13. F. Oudet, P. Courntine, and A. Vejux, “Thermal Stabilisation of Transition Alumina by Structural Coher-

ence with LnAlO3 (Ln = La, Pr, Nd)”, J. Catal., 114 112-20 (1988).

14. B. Beguin, E. Garbowski and M. Primet, “Stabilisation of Alumina by Addition of Lanthanum”, Appl.

Catal. 75 119-32 (1991).

15. P. Burtin, J.P. Brunelle, M. Pijolat, and M. Soustelle, “Influence of Surface Area and Additives on the

Thermal Stability of Transition Alumina Catalyst Supports. I: Kinetic Data”, Appl. Catal., 34 225-38

(1987).

16. D. Lafarga, A. Lafuente, M. Menéndez and J. Santamaría, “Thermal Stability of γ-Al 2O3/α-Al 2O3 Mesopo-

rous Membranes”, J. Membrane Sci., 147 173-85 (1998).

17. V.I. Vereshchagin, V.Y. Zelinskii, T.A. Khabas and N.N. Kolova, “Kinetics and Mechanism of Conversion

of Low-Temperature Forms of Alumina into α-Al 2O3 in the Presence of Additives”, J. Appl. Chem. USSR,

1792-93 (1983).

18. W.H. Gitzen, L.D. Hart and G. MacZura, “Phosphate-Bonded Alumina Castables: Some Properties and

Applications”, Ceram. Bull., 35 [6] 217-23 (1956).

19. J.E. Lyon, T.U. Fox and J.W. Lyons, “An Inhibited Phosphoric Acid for Use in High-Alumina Refracto-

ries”, Ceram. Bull., 45 [7] 661-65 (1966).

20. C. Toy and O.J. Whittemore, “Phosphate Bonding with Several Calcined Aluminas”, Ceram. Int., 15 167-

71 (1989).

21. M.J. O’Hara, J.J. Duga and H.D. Sheets Jr., “Studies in Phosphate Bonding”, Ceram. Bull., 51 [7] 590-95

(1972).

22. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, “Formation and Characterization of Sup-

ported Microporous Ceramic Membranes by Sol-Gel Modification Techniques”, J. Membrane Sci., 99 57-

75 (1995).

23. B.E. Yoldas, “Hydrolysis of Aluminium Alkoxides and Bayerite Conversion”, J. Appl. Biotech., 23 803-9

(1973).

Chapter 584

24. H.W. Brinkman, “Ceramic Membranes by (Electro)Chemical Vapour Deposition”, PhD thesis, University

of Twente, 1994.

25. F.P. Cuperus, D. Bargeman and C.A. Smolders, “Permporometry. The Determination of the Size Distribu-

tion of Active Pores in UF Membranes”, J. Membrane Sci., 71 57-67 (1992).

26. S. Krongelb, “Environmental Effects on Chemically Vapour-Plated Silicon Dioxide”, Electrochem. Tech., 6

[7-8] 251-56 (1968).

27. A. Nijmeijer, H. Kruidhof and H. Verweij, “Scheidingsinrichting met Keramisch Membraan” Dutch Patent

Application (in Dutch), no. 1010097 (1998).

&KDSWHU

3UHSDUDWLRQFKDUDFWHULVDWLRQDQGSURSHUWLHVRI

PLFURSRURXVVLOLFDPHPEUDQHV

1. Introduction

Microporous inorganic membranes are expected to be suitable for use in gas separation and

pervaporation at high temperatures (up to 600ºC) and harsh environments. State-of-the-art

microporous silica membranes show high fluxes for small gas molecules like H2 and high se-

lectivities with respect to larger molecules [1,2]. These membranes can be used in applica-

tions like natural gas purification, molecular air filtration and selective CO2 removal. High

temperature gas separation can be performed in either stand-alone operation or in direct com-

bination with reaction in high temperature membrane reactors. In that case reactions of inter-

est include steam reforming, the water-gas shift process, the dehydrogenation of hydrocarbons

and coal gasification [3,4].

Amorphous microporous silica membranes as discussed here, consist of a macroporous

α-alumina support (pore diameter ~100 nm) with a mesoporous γ-alumina intermediate layer

(Kelvin radius of 2.5 nm) and a microporous silica top layer (pore diameter ~4 Å) [1,2].

In this chapter several experiments are described in which the silica was either modified by

increasing the firing temperature, or by doping the silica with foreign ions or atoms. The fir-

ing temperature was increased to obtain a better hydrothermal stability, while doping with

foreign ions or atoms was done for two reasons. First, partly doping of silica structures may

enhance (hydro)thermal stability of the structure, as described below, and second, doping the

silica structure may improve the transport properties of the silica membrane: e.g., by doping

with palladium, hydrogen transport through the membrane might be promoted.

The influence of firing temperature on the stability and transport properties of microporous

silica membranes can largely be explained on the basis of the concentration of Si-OH groups

Chapter 686

in the network. As described by Imai et al. [5] the densification of silica structures can be in-

fluenced by the number of OH groups present in the structure. We therefore assume that when

these OH-groups are already removed from the silica structure during synthesis, the hy-

drothermal stability under process conditions will be much improved. It is well-known from

literature [6-8] that after firing at 800ºC only traces of OH-groups are left in the silica struc-

ture.

The influence of the addition of Al3+- and Mg2+- ions to the silica structure is described by

Fotou et al. [9]. This addition improved the hydrothermal stability of unsupported micropo-

rous silica material considerably. The positive effect of doping silica with Al2O3 on the hy-

drothermal stability of the material was also recognised by Lin et al. [10], who investigated

the thermal behaviour of alumina-doped silica xerogels. Because doping with large amounts

of Al2O3 hindered the viscous sintering, the mesoporous structure remained unchanged even

after firing at 1000ºC.

In this study modified silica layers were applied on undoped, flat γ-Al 2O3 membranes, as de-

scribed in chapter 5. Rutherford Backscattering (RBS) on some selected doped membranes

was used to reveal the location of the dopants in the membrane structure.

Until now, flat silica membranes were only tested at relatively low temperatures (up to 300ºC)

because of limitations in thermal stability of the polymer sealing rings [2]. However, with the

use of dense alumina rings together with carbon sealing it is possible now to measure mem-

brane properties (e.g. permeance) at much higher temperatures (up to 600ºC), which will be

described below.

Next to the experiments with modified silica on flat supports, also coating experiments of

commercially available tubular α-alumina supports with state-of-the-art silica are described.

The intention of this chapter is to provide an overview of the research performed on silica

membranes and membrane materials during the project. Due to the fact that subjects like the

preparation of tubular supports and a hydrothermally stable intermediate layer were given a

higher priority, most of the results are preliminary and much more research will be needed to

get a clear picture of the behaviour of (doped) silica membranes under SASRA conditions.

Therefore the results provided should be seen as the result of preliminary work and might

serve as a basis for future research.

Preparation, characterisation and properties of microporous silica membranes 87

2. Theoretical background on Rutherford Backscattering

Rutherford Backscattering Spectrometry (RBS) is a non-destructive (sub)-surface analysis

technique for solid systems. In principle atomic composition and depth distribution can be

obtained for a sub-surface layer of a few microns.

2.1 Principle of method

A small beam of high energy protons or 4He+

ions is directed to the surface of the sample

(“target”) under investigation. The primary

ions, with mass M1 and energy E1 (typically in

the range of 1-2 MeV) can collide with a target

atom with mass M2. In the elastic scattering

process (Figure 1), the incident ion transfers

energy and momentum to the target atom. The

energy of the scattered primary particle depends on the mass ratio, M2/M1, and the scattering

angle, θ:

where K is the kinematic factor. Hence the en-

ergy loss of the scattered ion, measured at an

angle θ, is a measure for the atomic weight of

the target atom and consequently for its

chemical nature (assuming no isotope interfer-

ence).

Along its trajectory through the target the primary particle will also loose energy through

inelastic interactions with the target electrons (free and bound) and nuclei. For a pure element

target this energy loss over a distance x is given by:

Figure 1: Schematic representation of the scat-

tering event.

Figure 2: Schematic representation of different

pathways of scattered primary ions.

E E

M

M

M MKE3 1

2

1

2

2

2 1

2

11=

+

−

+

=cos sin

/

θ θ

Chapter 688

E EE

xx E N x

x

1 10

1'= − ⋅ = − ⋅ ⋅Id

dd ε

Here it is assumed that (for small distances from the sample surface) the energy loss per unit

length is independent of the particle energy and can be replaced by a constant, Nε, where ε is

the stopping cross-section and N is the atomic density of the target. It should be noted that the

energy loss is indirectly related to the length of the path but directly to the number of atoms

(per unit area) encountered. The stopping cross-sections have been tabulated for most ele-

ments. For composite samples the total stopping cross-section can be calculated from the

stopping cross-sections of the elements using Bragg’s rule. E.g. for a compound AmBn:

ε ε εAB A Bm n= +

The inelastic energy loss of the scattered primary ion thus serves as a measure of the distance

between the surface and the target atom. But one has to take into account the angle, α, the

primary beam makes with the surface (see Figure 2):

where t is the depth with respect to sample surface and εc the “composite” stopping cross-

section for the target. Similarly the energy loss of the scattered primary particle leaving the

sample depends on the travelled path:

where β is the angle between the detector normal and the surface of the sample, with the

scattering angle: θ = α+β.

Another important parameter is the yield. The yield of particles detected by a solid state de-

tector subtending a small solid angle, Ω, (typically Ω < 10-2 sr) is given by:

Y Q Nt= ⋅ ⋅ d

d

σΩ

Ω

Q is the number of primary ions, Nt the number of atoms per unit area, N, in a layer of thick-

ness t. dσ/dΩ is the average differential scattering cross-section. For small values of Ω the

average scattering cross-section is given by:

E E S l E Ntc

1 1 1 1 1'sin

= − = − εα

E E S l E Ntc

3 3 3 3 3" ' 'sin

= − = − εβ

Preparation, characterisation and properties of microporous silica membranes 89

d

d

σθ

θ θ

θΩ=%&'

()*

⋅− +

−

%&K

'K

()K

*K

Z Z e

E

MM

MM

1 22 2

4

2

2

2

4

4 1

1

1

2

12

1

2

12sin

[ ( sin ) ] cos

[ ( sin ) ]

For M2>M1 this reduces to:

d

d

σθΩ

=%&'

()*

Z Z e

E1 2

2 2

424

1

sin

Hence the yield measured by the detector provides quantitative information on the number of

atoms per unit area, N t (for pure elements).

The energy spread of the primary ions is small (typically < 2keV), but due to the statistical

nature of the inelastic interactions with the target this energy spread increases. This phenome-

non, called energy straggling, reduces the accuracy of the mass and depth determination of

the atoms at increasing distance from the target surface. The energy resolution of the detector

(typically 15 keV) also places a limit on the mass resolution, which becomes most distinct for

the heavy elements.

Each particle entering the detector causes an electronic pulse with a height proportional to the

particle energy. Each detector event is analysed and stored, with respect to the measured en-

ergy, in a Multi Channel Analyser (MCA) with a typical channel width of 4keV. With 512

channels an energy spectrum range can be recorded of just over 2 MeV. The position of the

surface atoms depends on the ion beam energy, E1, and the kinematic factor for the specific

atom. Hence the heaviest atoms appear in the high energy side of the spectrum. The peak will

extend to lower energies depending on the thickness of the sample or the concentration profile

of the atom in the sample.

Chapter 690

In Figure 3 three schematic examples are presented. Figure 3A shows the energy spectrum for

a thin foil of composition AB where A has the highest atomic number. For a thin foil two

separate peaks will be observed, each limited at the high-energy side by KXE1. In Figure 3B

the case of a stack of layers is presented. The energy edge for element B is shifted to lower

energy as the primary and scattered particles must travel through foil A, loosing energy

through inelastic interactions. When a thick sample of composition AB is analysed then the

contribution from the light element B is added to the spectrum of the heavy element

(Figure 3C).

For thin film samples the analysis of the RBS spectra is generally straightforward, especially

when the peaks are well separated. For bulk samples and samples with layers of different

compositions the spectrum will be a complicated sum of the individual element spectra. For

analysis, a model spectrum is generated based on assumptions about the elemental composi-

tion and element distribution in the sample. The model parameters are (manually or by a

minimisation routine) adjusted until a satisfactory agreement with the measured spectrum is

obtained.

Typical standard RBS analysis conditions [11,12] are:

• 4He+ or H+ beam energy: 1.5 to 2.3 MeV, energy spread < 2 kEv

• Beam spot size: 1 to 4 mm2, beam current: 10 to 50 nA

• Ion dose to accumulate one spectrum: 10 to 40 µC

• Incidence angle: 5 to 10° to the sample normal

• Incident beam divergence: better than 3° (full angle)

• Target vacuum: better than 10-4 Pa (10-6 Torr)

Figure 3: Examples of RBS measurements and results for different types of samples.

Preparation, characterisation and properties of microporous silica membranes 91

• Detector angle: scattering at 165 to 170° to the incident beam

• Detector solid angle: 3 to 5 msr, surface barrier detector area: 25 to 300 mm2

• Detector resolution: 15 keV, spectrum channel width: 4 keV

• Analyser data storage per spectrum: 512 channels

• Typical accumulation time: 5 to 20 minutes

2.2 Application to supported silica membranes

A remaining question is the morphology of the interface between the silica top-layer and the

intermediate γ-Al 2O3 support layer. In the dip-coating process the silica layer may either form

an almost flat layer on top of the polished support surface or partly infiltrate the pores in the

γ-layer. With conventional techniques (SEM-EDS, or TEM) the microscopic geometry of this

interface cannot be elucidated. With RBS analysis a more clear indication can be obtained on

the depth distribution of the silicon atoms with respect to the γ-alumina surface.

3. Experimental

3.1 Membrane synthesis

Supported silica membranes were prepared according to the sol-gel method first described by

De Lange et al. [13]. In the experiments on flat membranes the supporting system consisted of

a colloidal filtrated α-alumina support with a γ-alumina intermediate layer.

The support

AKP-30 supports were prepared as described in chapter 4. After firing the supports were ma-

chined to the required dimensions and polished until a shiny surface was obtained. For low

temperature measurements, up to 300ºC, the supports were used without any further sealing.

The membranes were sealed with Kalrez®* or PTFE† O-rings in a K250 testing cell. For the

* Du Pont Company, Wilmington, Delaware, USA.† Eriks, Alkmaar, The Netherlands.

Chapter 692

high-temperature measurements, the polished porous supports were glass-soldered to dense

alumina rings for application in the K500 high temperature membrane testing cell*.

The γ-alumina intermediate layer

For the application of the mesoporous γ-alumina layer, essentially the same recipe as devel-

oped by de Lange et al. [13] was used. The complete synthesis route is described in chapter 5.

Dip coating was performed under class 100 cleanroom conditions in order to minimise parti-

cle contamination of the membrane layer [1,2]. After coating, the membranes were dried in a

climate chamber† at 40ºC and 60% relative humidity for 3 hours in air, because it has been

shown [15] that the drying rate of the boehmite layer under such conditions is sufficiently low

to avoid crack formation. After drying the membranes were fired at various temperatures be-

tween 650 and 1000ºC for 3 hours with a heating/cooling rate of 1ºC/min, as described in

chapter 5. The total γ-alumina layer thickness is in the order of 3 µm with an average Kelvin

radius of 2.5-6.0 nm, depending on the firing temperature. The Kelvin radius was determined

by permporometry [16].

The (doped) silica toplayer

The microporous silica layer was coated on the γ-alumina membrane surface by dip-coating

under class 100 cleanroom conditions. A standard silica dip-coating [13] was prepared by re-

acting 21 ml of Tetra Ethyl Ortho Silicate (TEOS)‡ with 8 ml concentrated HNO3 and 3 ml

double distilled water. As solvent, 21 ml ethanol§ was used. The reaction was performed un-

der refluxing at 60ºC for 3 hours. After reaction, the resulting polymeric silica sol was diluted

19 times with ethanol to obtain the dip coating solution. After coating, the silica membranes

were fired at various temperatures (400-800ºC) for three hours with a heating/cooling rate of

0.5ºC/min.

The silica sol was doped by adding the required concentration of dopant to the sol during

synthesis after 2 hours of refluxing. The solutions of ionic dopants were prepared by dissolv-

* Velterop BV, Delden, The Netherlands.† Heraeus Vötsch, Ballingen, Germany.‡ Adrich Chemical Company Inc., Milwaukee (WI), USA.§ E. Merck, Darmstadt, Germany.

Preparation, characterisation and properties of microporous silica membranes 93

ing the required amount of metal nitrates in a mixture of 90 vol-% ethanol and 10 vol-% dou-

ble distilled water. The sol was doped by metallic platinum by adding the required amount of

platinum in the form of 1 mg/ml colloidal platinum in 20% HCl* to the sol after 2 hours of

reaction.

For various systems unsupported silica membrane material has been prepared by drying the

(doped) silica sol and subsequent firing at the desired temperatures with a heating/cooling rate

of 0.5ºC/min. Special 0.5 mol-% K-doped samples were prepared for studying the effect of

potassium on the silica membrane material. This is of special importance because the catalyst

used in steam reforming contains some potassium. Of the different unsupported silica samples

the specific surface area and XRD†-spectra were measured before and after a treatment in a

Simulated Ambient Steam Reforming Atmosphere (SASRA). The conditions of such a

SASRA treatment are: 100 hours in 3:1 steam/methane atmosphere at 600ºC and 2.5 MPa to-

tal pressure.

Tubular membranes

For the preparation of tubular silica membranes, commercially available mesoporous mem-

branes‡ [17] are used. These tubular supports have a total length of 25 cm and are enamelled

at both ends, required for a gas-tight sealing with carbon seals to the reactor, so that an effec-

tive porous length of 20 cm remains. The tube consists of 4 layers. Layer 1, 2 and 3 consist of

α-alumina with a thickness of 1.5 mm, 40 and 20 µm and a pore diameter of 12, 0.9 and 0.2

µm respectively. Layer 4 consists of γ-alumina with a thickness of 3-4 µm a Kelvin radius of

4 nm. A schematic drawing of the cross-section of a mesoporous support tube is provided in

Figure 4.

* Alfa, Johnson Matthey GmbH, Karlsruhe, Germany.† Siemens D5000, Cu Kα radiation, Essen, Germany.‡ T1-70, SCT/US Filter Membralox, Bazet, France.

Chapter 694

For the coating of the tubes a standard silica

dip-coating sol was used, prepared as de-

scribed above. This sol was either used undi-

luted, or was diluted 10 times to account for

the longer contact time during coating, com-

pared to the coating of flat membranes. The

tubes were filled with sol under class 1000

cleanroom conditions, and left standing for 1

or 2 min. after which they were emptied. After

coating, the membranes were fired at 650ºC

for 3 h. with a heating/cooling rate of

0.5ºC/min. Often, this procedure was repeated

to obtain a second silica layer. After permeance measurements, the tubes were often coated

again to repair any defects, either already present in the structure or formed during the meas-

urements.

3.2 Permeance and selectivity measurements

Gas permeance through the membranes was measured in the pressure-controlled dead-end

mode [18]. The disc-shaped membranes were placed in the commercially permeance cells,

K250 and K500 as mentioned before. Maximum operation temperatures were 300 and 600ºC

respectively. The membrane was fitted in the cell with the microporous top-layer at the gas

feed side. The pressure difference over the membrane was adjusted by an electronic pressure

controller*. The gas flow through the membrane was measured by electronic mass flow me-

ters†. A schematic representation of the permeance set-up is given in Figure 5.

* Model 5866, Brooks Instrument B.V., Veenendaal, The Netherlands.† Model 5850TR, Brooks Instrument B.V., Veenendaal, The Netherlands.

4

3

1 2

Figure 4: Schematic drawing of a cross-section

of the used commercial tubes.

Preparation, characterisation and properties of microporous silica membranes 95

The tubular supports were measured in a membrane reactor, which could also serve for steam-

reforming experiments when applicable (Figure 6). The tubes were sealed with carbon sealing

at the enamelled ends of the tubes. Permeance measurements were performed at 500ºC.

4. Results

4.1 Permeance and selectivity measurements

To obtain quantitative data about the integrity of the prepared membranes, both permeance

and selectivity measurements were performed. It was found that high quality flat membranes

Figure 5: Schematic diagram of the experimental set-up for permeance measurements.

graphite seal

enameled surface

porous surface

opposite male ridgescopper seal

Figure 6: IRC membrane reactor.

Chapter 696

could be prepared which showed a very high permselectivity. An overview of the most inter-

esting results is provided in Table 1.

The experiments on the tubular supports were more cumbersome. The increase in selectivity

was not as high as normally obtained for the flat membranes. With an undiluted sol, however,

it was possible to increase the H2/CH4 permselectivity considerably above the Knudsen perm-

selectivity (2.8). The results of two representative measurements on tubular supports are

summarised in Table 2.

Material Firing T

(ºC)

Measuring T

(ºC)

H2 Permeance

(mol/m2sPa)

Permselectivity

H2/CO2

Permselectivity

H2/CH4

St. silica 600 200 5*10-7 45 >1000

St. silica 825 200 1*10-7 1000 135

St. silica 625 400 3*10-7 17 30

St. silica 625 600 3*10-7 21 14

Si + Pt 600 200 3*10-7 70 >2000

Si + Mg/Al 600 200 1*10-7 110 >2000

Table 1: Permeance and permselectivity properties for different flat silica membranes. St. silica indicates

standard silica, Si + Pt = standard silica with 2.5 mol-% Pt metal, Si + Mg/Al = standard silica

with 0.5/0.5 mol-% Mg2+ and Al3+.

Sample no Nr. of SiO2 layers Coating time

(min)

Dilution N2 permeance

(mol/m2sPa)

Permselectivity

H2/CH4

1 2 1 10x 25*10-7 3.6

1 2+1 2 10x 3*10-7 3.0

1 3+1 1 Undil. 2*10-7 2.5

2 1 1 Undil 10*10-7 5.8

2 1+1 1 Undil. 5*10-7 3.4

Table 2: Results of permeance and permselectivity measurements for tubular silica membranes.

Preparation, characterisation and properties of microporous silica membranes 97

4.2 RBS measurements

In Figure 7 two RBS spectra are shown, one

for an untreated α/γ-Al 2O3 membrane and one

for a silica-coated membrane. The two spectra

are virtually identical, except for a small step

at the high-energy side of the spectrum of the

silica-coated membrane. This small step in the

spectrum indicates that a silica layer is formed

on top of the gamma layer. This can be seen

more clearly in an enlargement of the high-

energy region in Figure 8. Also a simulation

curve is shown (dashed line). For the simula-

tion the membrane was divided into four lay-

ers, a silica top-layer (about 8 nm thick), an

intermediate silica/γ-alumina layer of ap-

proximately 45 nm and an γ-alumina layer of

about 1 µm thick. The fourth layer is formed

by the α-Al 2O3 support.

Figure 9 shows the total spectrum for a Pt-

modified silica membrane. The first (high-

energy) peak of Pt occurs exactly at the sur-

face energy edge and is followed by a second

peak with a tail with decreasing energy. Also a

clear and separate Si peak is found at the sur-

face indicating a silica top-layer with little

penetration in the γ-alumina layer, as evidenced

by the onset of the Al-edge at an energy sub-

stantially lower than the edge for surface Al.

Figure 7: RBS spectra for a treated and an un-

treated α/γ-Al 2O3 membrane.

Figure 8: Enlargement of the high-energy region

of Figure 7.

Figure 9: RBS spectrum of a Pt-doped silica

membrane.

Chapter 698

4.3 Specific surface area measurements

The results of specific surface area measurements of different unsupported (doped) silica

membrane materials is provided in Table 3. Unfortunately due to time limitations no meas-

urements were performed on Mg/Al-doped material fired at lower temperatures.

4.4 XRD measurements

XRD-diagrams do not show any phase transformation in the SASRA treatment for the undo-

ped samples and the samples doped with Pd, Al/Mg. All these XRD-spectra showed an amor-

phous silica phase, for the undoped material the diagrams are shown in Figure 10. In the case

of K-doping, however, a phase change to a keatite [19,20] phase occurred during SASRA. As

shown in the XRD-diagram of Figure 11, a more detailed description of the results will be

provided in [21].

Material Before SASRA (m2/g) After SASRA (m2/g)

SiO2 (400ºC) 0.24 0.09

SiO2 (600ºC) 0.10 0.08

SiO2 (825ºC) 0.10 0.11

SiO2 (Mg/Al 825ºC) 0.06 <0.06

Table 3: Specific surface area of different membrane materials. The temperature between brackets indi-

cates the firing temperature of the material.

Preparation, characterisation and properties of microporous silica membranes 99

Pure silica material

0

200

400

600

800

1000

1200

1400

0 20 40 60 80 100

2 Theta

Si

IntensIty

654

321

Figure 10: XRD-diagrams of pure unsupported silica material. Diagrams 1, 2 and 3 are before and 4, 5 and 6

after SASRA treatment, respectively. 1 and 4 is silica fired at 400ºC, 2 and 5 silica fired at 600ºC

and 3 and 6 silica fired at 825ºC.

Silica with 0.5%K

0

5000

10000

15000

20000

25000

0 20 40 60 80 100

2 Theta

IntensIty

21

Figure 11: XRD-diagrams of 0.5 mol-% K-doped unsupported silica material, prepared at 825ºC. Curve 1

before and curve 2 after SASRA. The peaks observed in curve 2 indicate a keatite phase.

Chapter 6100

5. Discussion and conclusions

As the permeance and permselectivity measurements show, it is possible to prepare high-

quality doped silica membranes with excellent properties. Moreover it was possible to per-

form permeance and permselectivity measurements at temperatures up to 600ºC on flat mem-

branes. To the author’s knowledge these are the first reliable measurements ever performed

on flat membranes at such a high temperature. A more detailed discussion of the permeance

and permselectivity results follows. It must however be noted that the relatively low hydrogen

permeances obtained for the described membranes were at least partly due to the used

AKP-30 supports, which had a bare-support hydrogen permeance of ~8*10-7 mol/m2sPa.

Standard silica membranes

The very high H2/CO2 permselectivity for the 825ºC fired standard silica membrane is re-

markable. It is even more remarkable that the H2/CH4 selectivity is lower, which is contrary to

the common observation that the H2/CO2 selectivity is lower than the H2/CH4 selectivity.

The excellent separation properties of silica membranes prepared at temperatures as high as

825ºC enables their use for high temperature applications, such as the dehydrogenation of H2S

(chapter 8). Unfortunately no hydrothermal stability of the prepared layers could be tested be-

cause the mesoporous intermediate layer was not hydrothermally stable, but an indication of

the hydrothermal stability of the unsupported material could be obtained from the specific sur-

face area and XRD measurements. These measurements did not show any structural change in

the material during SASRA treatment, which is a very hopeful result for the operation of real,

supported, membranes at high temperatures and high pressures.

High temperature measurements reveal a somewhat lower permselectivity of the membranes,

but the permselectivities measured are still far above the Knudsen permselectivity. We as-

sume that the lower permselectivity is due to defects caused by the sealing of the porous sub-

strate to the dense alumina ring necessary for the application of the carbon sealing. The coat-

ing of the membrane layers on the supports was performed after sealing the dense alumina

ring to the porous substrate. When the adherence of one of the layers to the glass-seal is not

perfect due to different adherence, some defects might arise on the interface between the seal

and the porous substrate.

Preparation, characterisation and properties of microporous silica membranes 101

Doped membranes

It was shown that silica membranes doped with Mg/Al and with Pt showed very high selec-

tivity towards methane and good selectivity towards CO2. From specific area and XRD meas-

urements it is clear that after firing at 825ºC, no large structural changes occur in doped silica

material during a SASRA treatment. For the moment it is not clear whether Mg/Al doping has

a positive effect of hydrothermal stability of silica membranes. To get real insight in a possi-

ble increase in hydrothermal stability, also experiments with material fired at lower tempera-

tures should be performed, as was done for undoped silica.

Silica membrane material doped with potassium, however, showed in SASRA treatment the

formation of a crystalline keatite phase. One should keep this result in mind when deciding on

the design of a membrane steam reformer. Regarding the amount of potassium in the common

used catalysts in steam reforming (see chapter 2) and the mobility thereof under steam re-

forming conditions, it might be wise to avoid spill-over of potassium from the catalyst to the

silica layer. Thus, when applying the separative silica layer on the inside of the tube, the

catalyst should be loaded in the annular space between reactor tube and membrane tube. A

detailed description of reactor design is provided in chapter 2.

Tubular membranes

Not much research was performed on coating of tubes in this project. Results on the coating

of commercial tubes show however that it is rather difficult to coat sufficiently defect-free

membranes on these supports. As was already stated in chapter 4, the surface roughness of the

used tubes was possibly too high to coat high quality silica layers. Another possibility is that

here the same problem occurs as was encountered with the sealed flat membranes. Some de-

fects in the membrane layer might result from a bad adherence of the coated layer at the

enamel/membrane interface.

First results are, however, rather positive. It was possible to coat silica layers with a selectiv-

ity, which was somewhat above the Knudsen selectivity. This shows the presence of at least

some degree of microporosity with, fortunately, still a relatively high N2 permeance compared

to that of the centrifugal cast tubes described in chapter 4. Coating with just 1 silica layer

from an undiluted silica coating solution showed to give best results. Coating with a second

layer, did not improve selectivity, it did only reduce the N2 permeance of the membrane with

Chapter 6102

a factor of 2. This is an indication that the defects in the coated layer are of such large size

that they can not be repaired anymore by just coating another layer. It might, however, also

support the suggestion that the reduction in selectivity is due to bad adherence of the coating

to the enamel, thereby creating large defects on the enamel/membrane interface.

As can be seen from Table 2, it was not possible to coat microporous silica layers starting

with a 10 times diluted coating sol. The silica concentration in this sol might be too low to get

the joint effect (entanglement) of the silica polymers during coating, which is normally pre-

venting the silica polymers from intruding the pores in the γ-alumina layer. Silica polymers

will therefore intrude in the pores of the support, as is shown by the decrease of N2 permeance

by a factor of 10 by coating an extra silica layer on a support which was already coated with 2

silica layers.

RBS measurements

RBS analysis revealed the existence of a very thin silica layer on top of a γ-Al 2O3 membrane.

An interesting feature that was observed in the RBS measurements is the increased oxy-

gen/aluminium ratio in the γ-alumina compared to the α-Al 2O3 supports. Assuming a stoichi-

ometric ratio of 3/2 for the α-Al 2O3 support, the model used for fitting the RBS data seems to

indicate rather a 4/2 ratio for the γ-layer. This most likely indicates that the surface of the γ-

Al 2O3 particles in the γ-Al 2O3 layer consists of Al(OH)3. A rough estimate shows that, with a

particles of about 5 to 10 nm diameter covered by an Al(OH)3 surface layer would have an

avarage O to Al ratio between 3.5 and 4. Furthermore, the model used for fitting the RBS data

indicates that the top-layer contains half of the silica. The other half is mixed in with the first

part of the γ-layer. As the amount of silica present in these membranes is rather small, the

modelling parameters present rather rough estimates.

For the Pt-doped membranes, the data suggests that the Pt is deposited mostly in the gamma

layer and on top of the silica layer. The deposition of platinum on top of the silica layer is

confirmed by SEM on a 5% Pt-doped membrane (Figure 12). Moreover it seems that in the

case of the platinum doped membranes the silica layer is much more located on top of the

γ-alumina layer than in the case of the undoped membranes. The reason for this is still un-

clear.

Preparation, characterisation and properties of microporous silica membranes 103

6. Acknowledgements

The author wishes to express his thanks to: Dr. H. Weyten (VITO) for performing the high

temperature measurements, Dr. R. Bredesen (SINTEF) for performing SASRA treatments,

XRD-measurements and specific surface area measurements, Dr. J.A. Dalmon (IRC) for per-

forming permeance and permselectivity measurements on the tubular membranes and Dr. A.

Vredenburg (University of Utrecht) for performing the RBS measurements. These people are

also kindly thanked for fruitful discussions on their respective measurments.

7. References

1. R.M. de Vos and H. Verweij, “High-Selectivity, High-Flux Silica membranes for Gas Separation”, Science,

279 1710-11 (1998).

Figure 12: SEM micrograph of the surface of a 5% Pt-doped silica membrane at a magnification of 5500x.

The white areas indicate surface platinum clusters.

Chapter 6104

2. R.M. de Vos and H. Verweij, “Improved Performance of Silica Membranes for Gas Separation”, J. Mem-

brane Sci., 143 37-51 (1998).

3. G. Saracco, H.W.J.P. Neomagus, G.F. Versteeg and W.P.M. van Swaaij, “High-Temperature Membrane

Reactors: Potential and Problems”, Chem. Eng. Sci., 54 1997-2017 (1999).

4. J. Zaman and A. Chakma, “Inorganic Membrane Reactors”, J. Membrane Sci., 92 1-28 (1994).

5. H. Imai, H. Morimoto, A. Tominaga and H. Hirashima, “Structural Changes in Sol-Gel Derived SiO2 and

TiO2 Films by Exposure to Water Vapour”, J. Sol-Gel Sci, Tech., 10 45-54 (1997).

6. M.L. Hair, “Hydroxyl Groups on Silica Surface”, J. Non-Cryst. Solids, 19 299-309 (1975).

7. M.L. Hair and W. Hertl, “Reactivity of Boria-Silica Surface Hydroxyl Groups”, J. Phys. Chem., 77 [16]

1965-9 (1973).

8. J.B. Peri, “Infrared Study of OH and NH2 Groups on the Surface of a Dry Silica Aerogel”, J. Phys. Chem.,

70 [9] 2937-45 (1966).

9. G.P. Fotou, Y.S. Lin and S.E. Pratsinis, “Hydrothermal Stability of Pure and Modified Microporous Silica

Membranes”, J. Mater. Sci., 30 2803-8 (1995).

10. C. Lin, J.A. Ritter and M.D. Amiridis, “Effect of Thermal Treatment on the Nanostructure of SiO2-Al 2O3

Xerogels”, J. Non-Cryst. Solids, 215 146-54 (1997).

11. J.M. Walls (ed.), “Methods of Surface Analysis”, Cambridge University Press (1989).

12. J.R. Bird and J.S. Williams, “Ion Beams for Materials Analysis”, Academic Press/Harcourt Brace Jovano-

vich Publishers (1989).

13. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, “Formation and Characterization of Sup-

ported Microporous Ceramic Membranes by Sol-Gel Modification Techniques”, J. Membrane Sci., 99 57-

75 (1995).

14. B.E. Yoldas, “Hydrolysis of Aluminium Alkoxides and Bayerite Conversion”, J. Appl. Chem. Biotech., 23

803-9 (1973).

15. K.N.P. Kumar, “Nanostructured Ceramic Membranes, Layer and Texture Formation”, PhD Thesis, Univer-

sity of Twente (1993).

16. F.P. Cuperus, D. Bargeman and C.A. Smolders, “Permporometry. The Determination of the Size Distribu-

tion of Active Pores in UF Membranes”, J. Membrane Sci., 71 57-67 (1992).

17. R. Soria, “Overview on Industrial Membranes”, Catal. Today, 25 [25] 285-90 (1995).

18. W.J. Koros, Y.H. Ma and T. Shimidzu, “Terminology for Membranes and Membrane Processes”, Pure &

Appl. Chem., 68 1479-89 (1996).

19. P.P. Keat, “A New Crystalline Silica”, Science, 120 328-30 (1954).

20. J. Shropshire, P.P. Keat and P.A. Vaughan, “The Crystal Structure of Keatite, a new Form of Silica”, Z.

Kristall., 112 409-13 (1998).

21. A. Nijmeijer, R. Bredesen, H. Kruidhof and H. Verweij, to be published.

&KDSWHU

/RZWHPSHUDWXUH&9,PRGLILFDWLRQRI γDOXPLQDPHPEUDQHV

1. Introduction

Inorganic membranes are suitable candidates for gas separation at high temperatures in chemi-

cally aggressive environments, as in high temperature membrane reactors. In these reactors,

microporous silica membranes are of particular use to selectively remove H2 in petrochemical

processes like steam reforming, the water gas shift reaction, dehydrogenation of hydrocarbons

and coal gasification [1,2]. Selective removal of hydrogen leads to significant conversion en-

hancement provided the H2 flux and selectivity toward H2 of the membranes are high.

A common and well-known method to prepare silica membranes with molecular sieving

properties is sol-gel coating [3-5]. With this technique, microporous silica layers with a pore-

size of about 0.5 nm are dip-coated on top of supported γ-alumina membranes. The supports

are porous α-alumina disks with pore diameters in the range from 100-200 nm. On top of

these macroporous supports a 3 µm thick mesoporous γ−alumina layer is coated, with a pore

size of 3 nm.

However, for application of silica membranes in membrane reactors it might be more favour-

able to locate the silica layer inside the mesoporous γ-alumina membrane, where it is pro-

tected against mechanical damage during catalyst loading and unloading. A suitable technique

to prepare a microporous layer inside the pores of a support is Chemical Vapour Infiltration

(CVI), first reported by the group of Gavalas [6,7]. In their experiments they used mesoporous

Vycor glass membranes and excellent permselectivities as high as 3000 for H2:N2 were ob-

tained. The membranes, however, had a relatively low hydrogen permeance of

1.5*10-2 mol/m2sPa due to the large support resistance.

* Based on: A. Nijmeijer, B. Bladergroen, H. Verweij, Microporous & Mesoporous Mater, 25 179-84 (1998).

Chapter 7106

To overcome the drawback of low hydrogen permeances, Morooka et al. [8] deposited silica

layers inside a macroporous α−alumina support with pore sizes as large as 150 nm, resulting

in relatively thick silica layers. Although the membranes had a very good permselectivity of

H2 towards N2, the resulting H2 permeance of the membrane was not higher than that of the

modified Vycor membranes of Gavalas et al. [6].

To obtain thinner layers with a reasonable hydrogen permeance, Lin et al. [9] used α−alumina

supported mesoporous γ−alumina membranes with a pore size of 4 nm. However, they did not

succeed in depositing highly selective layers. Most likely this is caused by defects in the

γ-alumina layer.

In our laboratory, however, we are able to produce defect-free γ-alumina membranes [4] and

hence we decided to use the improved γ-alumina membranes for highly selective silica mem-

branes with a high hydrogen permeance by CVI.

2. Experimental

The starting material is a state-of-the-art flat γ-alumina membrane prepared by dipcoating of a

boehmite solution on a macroporous α-alumina support and subsequent firing at 600°C as de-

scribed in [4]. The α-alumina support is prepared from AKP-30 powder* by making a colloi-

dal suspension of this powder in diluted nitric acid and subsequent filtration. After filtration

the wet cake is dried overnight and sintered for 1 hour at 1100°C. The resulting flat α-alumina

supports have a mean pore diameter of 80 nm. A detailed description of the support synthesis

is provided in chapter 4.

All coating work is performed under class 100 cleanroom conditions to avoid large defects

due to dust particles, resulting in highly reproducible membrane properties, such as layer

thickness and pore-size. A typical thickness of the deposited γ-alumina layer is 3 µm and the

mean pore size is 2.5 nm as determined by permporometry.

In literature, numerous precursors for CVI of silica are proposed. Most common are silane

[6,7,10], Tetra Ethyl Ortho Silicate (TEOS) [8,9,11,12] and silicon tetrachloride [13,14]. As

oxygen source mainly pure oxygen or water vapour in a helium or nitrogen flow is used. In

* AKP-30, Sumitomo, Osaka, Japan.

Low temperature CVI modification of γ-alumina membranes 107

this work silane and silicon tetra-acetate [15] are used as precursor, because of the tempera-

ture limitations of the reaction apparatus. Oxygen diluted with nitrogen is used as oxygen

source. In the experiments, the influence of the concentration and type of precursor as well as

the concentration of oxygen on the permeance and permselectivity of the resulting CVI silica

membranes is studied.

Before performing a CVI-experiment, the flat γ-alumina membranes were clamped into a re-

action cell (Velterop K250*) a schematic drawing of which is shown in Figure 1.

As can be seen in Figure 1, the silicon precursor is fed to the reaction cell at the side where the

γ-alumina layer is situated. The oxygen is fed to the other side of the membrane. The mem-

brane is clamped gas-tight in the cell by polymer O-rings†. Before each experiment the appa-

ratus was checked for possible leaks by pressure testing. The cell was heated to the reaction

temperature at a rate of 2°C/min.

* K250, Velterop B.V., Delden, The Netherlands.† Kalrez, Du Pont de Nemours, Dordrecht, The Netherlands.

Figure 1: Schematic drawing of the K250 reaction cell.

Chapter 7108

Silane apparatus

For the experiments with silane, a mixture of 0.5 vol-% silane in helium was used. The silane

CVI-set up is shown in Figure 2. The silane/He flow is regulated by Mass Flow Controller

MFC-1, whereas the oxygen flow is regulated by MFC-3. Because silane is a hazardous gas,

the complete reaction apparatus can be flushed with nitrogen by closing the line to the silane

bottle and opening the nitrogen #1 bottle. The 0.5 vol-% silane/He mixture can be diluted

further with either helium or nitrogen as controlled by MFC-2. The oxygen flow to the reac-

tion cell can be diluted as well with either nitrogen (from bottle #2) or helium via MFC-3. The

outlets of the reactor at both the silane side and the oxygen side are connected to a wash-bottle

containing a 1 M KOH-solution to remove unreacted silane and any hazardous compounds

formed by the CVI-reaction.

After a certain reaction time, the reaction can be ended by closing MFC-1 and MFC-4. The in-

situ nitrogen permeation through the membrane can be measured by closing the outlet of the

reactor at the precursor side (V-1) and closing the inlet of the reactor at the oxygen side (V-2).

The flow of nitrogen can be controlled by MFC-1 and the resulting pressure is measured by

the Pressure Indicator PI-1 at the reaction cell outlet. From these data the permeance as a

function of the pressure difference over the membrane can be calculated. Depending on the

measured nitrogen permeance it can be decided to stop the experiment or to start the CVI re-

action again for a certain time.

Low temperature CVI modification of γ-alumina membranes 109

Silicon tetra-acetate apparatus

For the experiments with silicon tetra-acetate (SiAc4), the experimental set-up as shown in

Figure 3 was used. This apparatus can be used as a CVI reactor as well as a flow-controlled

single-gas permeation apparatus. The permeance of the different gases is measured by setting

a flow by MFC-3 and measuring the resulting pressure by the pressure indictor PI.

After each CVI experiment nitrogen was used to check the change in permeance of the mem-

brane. For several membranes also permeance measurements were performed with He, H2 or

CO2. In these cases the corresponding permselectivities could be calculated, which are a better

indication of the membrane quality than the change in nitrogen permeance only. Unfortunately

this type of measurements was not possible with our silane set-up (see Figure 2) because of

safety regulations involved.

Furnace + reaction cell

TC

N2 #1 SiH4/He

MFC-1

N2 #2 He O2

1MKOH

MFC-2 MFC-3

MFC-4

PI-1 PI-2

V-1

V-2

Figure 2: Reaction apparatus for CVI-experiments with silane. Mass Flow Controllers are indicated by

MFC, Pressure Indicators by PI, Temperature Controllers by TC and valves by V.

Chapter 7110

Air, which can be diluted with nitrogen via MFC-3, is used to oxidise SiAc4. The silicon tetra-

acetate is fed to the reactor by a helium flow via MFC-1. A PID-controller* regulates the tem-

perature of the SiAc4 bath. The bath temperature was kept 155°C in all experiments. The He-

stream was saturated with SiAc4 and therefore all lines from the SiAc4 bath to the reaction cell

were heated to avoid condensation in the lines. During the CVI experiment the pressure at

both sides of the membrane were slightly above atmospheric.

Before and after experiments the pore sizes of the membranes were measured by permpo-

rometry [16], a technique based on blocking of smaller pores by capillary condensation of cy-

clohexane and the simultaneous measurement of the permeance of oxygen gas through the

larger, open pores. The measurements are performed at 20ºC on an area of 8.5*10-4 m2. The

pore size distribution (Kelvin radii) is determined in the desorption stage using the Kelvin

equation. More details on the permporometry technique can be found in [17] and all experi-

mental details of the permporometry apparatus are presented in [16].

* Model 818, Eurotherm, Zoeterwoude, The Netherlands.

He

He

Air

N2

H2

CO2

1MKOH

Atm

SiAc 4 bath

Reactor cellMFC-1

MFC-3

MFC-2

TC

TC

PI

Figure 3: The CVI apparatus with SiAc4 as a silicon source. Abbreviations identical to the ones used in Fig-

ure 2.

Low temperature CVI modification of γ-alumina membranes 111

3. Results and discussion

The CVI-experiments with silane as precursor did not show any increase in permselectivity of

the membranes. The reaction temperature was in all cases 275°C and several oxygen pressures

were tried. In each experiment, however, white powder was obtained on the membrane sur-

face, indicating the decomposition of silane at the surface of the membrane. Reaction condi-

tions could not be chosen in such a way that a highly separative layer was obtained. This was

probably related to with the fact that the reactor temperature or the concentration of silane in

the precursor gas was too low. Safety regulations, however, prohibited an increase of the si-

lane concentration in the precursor flow.

The CVI-experiments with SiAc4, on the other hand, were very successful, even at the rela-

tively low reaction temperature of 275°C. Optimum results were found with a reaction time of

about 45 minutes, a SiAc4 bath temperature of 155°C, a He-flow of 200 ml/min and an oxy-

gen partial pressure of 0.2 bar in nitrogen with a total flow of 200 ml/min. The pressure was

kept slightly above atmospheric at both sides of the membrane during the reaction and the re-

action area was 9*10-4 m2.

During the permeance measurements no pressure dependence of the permeance was found.

The results of the experiments are given in Table 1, together with the total reaction times.

Experiment no. treaction (min) Tcell

(°C)

P(H2) P(He) P(N2) H2:N2 He:N2 H2:CO2

1 0 275 11.0 7.1 2.9 3.7 2.5 -

43 250 4.0 4.1 0.09 43 44 -

43 20 1.7 2.1 0.04 43 55 -

2 0 20 13.0 8.9 3.6 3.7 2.4 -

0 275 9.0 5.4 2.3 3.9 2.3 3.9

25 250 2.0 2.1 0.08 25 26 50

Table 1: Results from permeability measurements before and after CVI-experiments with SiAc4 as precur-

sor. The permeances (P) are given in 10-7 mol/m2sPa.

Chapter 7112

As can be seen from Table 1, the starting membrane exhibits almost pure Knudsen separation

characteristics, as can be expected from a state-of-the-art mesoporous γ-alumina membrane.

The ideal Knudsen permselectivities are H2:N2=3.74, He:N2=2.65, H2:CO2=4.69, which are in

agreement with the measured permselectivities. After 43 minutes of CVI in experiment num-

ber 1, the hydrogen permeance had decreased to 4.0*10-7 mol/m2sPa, but the permselectivity

towards nitrogen had increased to 43, which is far above the Knudsen value.

In experiment number 2 the permselectivity of H2 towards CO2 was measured as well and was

found to be 50 after only 25 minutes of CVI.

To determine the change in pore size of a mesoporous membrane during CVI, the pore-size

was measured before and after a CVI-experiment by permporometry. Typical results are

shown in Figure 4, in which raw permporometry data are shown before and after CVI with

SiAc4.

In this figure the oxygen permeation is plotted as a function of the relative pressure of cyclo-

hexane. The curve, which was obtained from the membrane before the CVI-experiment,

shows a clear transition point. This is indicative of a mesoporous material, in this case with a

calculated Kelvin radius of 2.5 nm. For the membrane after CVI no such curve could be ob-

tained and no clear transition point could be observed. This behaviour is representative for

microporous materials.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

relative pressure cyclohexane

O2

perm

eatio

n

Before CVIAfter CVItransition point

Figure 4: Typical permporometry plot before and after CVI with SiAc4.

Low temperature CVI modification of γ-alumina membranes 113

4. Conclusions

It has been shown that state-of-the-art γ-alumina membranes can be modified with low tem-

perature SiAc4 CVI to obtain silica membranes exhibiting molecular sieving properties. Al-

though the permselectivity of the membranes is not as good as reported in several other stud-

ies where permselectivities of more than 1000 are found [6,7,13], the permselectivity is still

rather high compared to common sol-gel silica membranes [3]. The main advantage of the

CVI membranes prepared in this work is the high hydrogen flux, which makes them suitable

for many industrial processes for which the permeability of a given membrane is more im-

portant than the permselectivity of the membrane.

The CVI method can also be used to repair ceramic membranes in situ in industrial processes.

In this case the defects formed in the membranes in the reactor or separation unit are repaired

by controlled CVI during a shut-down. This has the advantage that the membranes in the unit

need not to be replaced, which in most cases would mean the replacement of the complete

unit. A second advantage is that repairing the membranes is probably less time-consuming

than complete replacement.

The CVI modification of γ-alumina membranes with a 0.5 mol-% silane mixture has not re-

sulted in highly selective membranes. This might be due to an insufficiently high reaction

temperature or silane concentration. Unfortunately in this work the temperature was limited by

the reaction set up and the silane concentration by safety regulations.

The possibility to measure the permeance of several gases in-situ has proven to be very con-

venient for following the CVI-process.

5. References

1. J. Zaman and A. Chakma, “Inorganic Membrane Reactors”, J. Membrane Sci., 92 1-28 (1994).

2. G. Saracco, G.f. Versteeg and W.P.M. van Swaaij, “Current Hurdles to the Success of High-Temperature

Membrane Reactors”, J. Membrane Sci., 95 105-23 (1994).

3. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, “Formation and Characterization of Sup-

ported Microporous Ceramic Membranes by Sol-Gel Modification Techniques”, J. Membrane Sci., 99 57-

75 (1995).

Chapter 7114

4. R.M. de Vos and H. Verweij, “High Selectivity, High Flux Silica Membranes for Gas Separation”, Science,

279 1710-11 (1998).

5. R.J.R. Uhlhorn, M.H.B.J. Huis in ‘t Veld, K. Keizer and A.J. Burggraaf, “Synthesis of Ceramic Membranes.

Part 1. Synthesis of Non-Supported and Supported Gamma-Alumina Membranes without Defects,” J. Ma-

ter. Sci., 27 527-37 (1992).

6. G.R. Gavalas, C.E. Megiris and S.W. Nam, “Deposition of H2-Permselective SiO2 Films”, Chem. Eng. Sci.,

44 [9] 1829-35 (1989).

7. S.W. Nam and G.R. Gavalas, “Stability of H2-Permselective SiO2 Films Formed by Chemical Vapor Depo-

sition”, AIChE Symp. Series, 85 [268] 68-74 (1989).

8. S. Morooka, S. Yan, K. Kusakabe and Y. Akiyama “Formation of Hydrogen-Permselective SiO2 Membrane

in Macropores of α-Alumina Support Tube by Thermal Decomposition of TEOS”, J. Membrane Sci., 101

89-98 (1995).

9. C.L. Lin, D.L. Flowers and P.K.T. Liu, “Characterization of Ceramic Membranes II. Modified Commercial

Membranes with Pore Size under 40 Å”, J. Membrane Sci., 92 45-58 (1994).

10. S. Kitao and M. Asaeda, “Gas Separation Performance of Thin Porous Silica Membrane Prepared by Sol-

Gel and CVD Methods”, Key Eng. Mater., 61 & 62 267-72 (1991).

11. S. Yan, H. Maeda, K. Kusakabe, S. Morooka and Y. Akiyama, “Hydrogen-Permselective SiO2 Membrane

Formed in the Pores of Alumina Support Tube by Chemical Vapor Deposition with Tetraethyl Orthosili-

cate”, Ind. Eng. Chem. Res., 33 2096-101 (1994).

12. J.C.S. Wu, D.F. Flowers and P.K.T. Liu, “High-Temperature Separation of Binary Gas Mixtures Using Mi-

croporous Ceramic Membranes”, J. Membrane Sci., 77 85-98 (1993).

13. S. Kim and G.R. Gavalas, “Preparation of H2 Permselective Silica Membranes by Alternating Reactant Va-

por Deposition”, Ind. Eng. Chem. Res., 34 [1] 168-76 (1995).

14. M. Tsapatsis, S. Kim, S.W. Nam and G. Gavalas, “Synthesis of Hydrogen Permselective SiO2, TiO2, Al2O3

and B2O3 Membranes from the Chloride Precursors”, Ind. Eng. Chem. Res., 30 2152-59 (1991).

15. T. Maruyama and J. Shionoya, “Silicon Dioxide Films Prepared by Chemical Vapour Deposition from Sili-

con Tetraacetate”, Jap. J. Appl. Phys., 28 [12] L2253-54 (1989).

16. H.W. Brinkman, “Ceramic Membranes by (Electro)Chemical Vapour Deposition”, PhD-thesis, University

of Twente, 1994.

17. F.P. Cuperus, D. Bargeman and C.A. Smolders, “Permporometry. The Determination of the Size Distribu-

tion of Active Pores in UF Membranes”, J. Membrane Sci., 71 57-67 (1992).

&KDSWHU

7KHWKHUPDOGHK\GURJHQDWLRQRI +6LQDPHPEUDQHUHDFWRU

In this chapter, a basis is provided for a project proposal on thermal dehydrogenation of H2S

in a membrane reactor. Such a proposal uses the results of this thesis as a starting point.

1. Introduction

In a modern refinery the different hydrocarbon streams contain a variety of sulphur com-

pounds, including hydrogen sulphide, mercaptans, sulphides, disulphides, and thiophenes.

The amount and the type of the sulphur compounds depends highly on the used crude oil type,

but even the most “sweet” (low sulphur) crudes contain considerable amounts of sulphur. For

different reasons the sulphur concentration of the product streams of the crude destillation

train must be reduced:

• From a refiners point of view the primary reason to reduce sulphur concentrations is the

protection of catalysts and equipment in the downstream units, e.g., platforming units.

• From an environmental point of view the sulphur concentration must be reduced in order

to reduce SO2 formation in automotive exhaust gases [1].

To remove sulphur from the different streams in a refinery, the more complex organic sulphur

compounds are first hydrogenated to H2S over a nickel or cobalt molybdate catalyst in so-

called hydro-desulphurisation units [2]. These units are typically operated in the temperature

range of 350-400ºC. For complete desulphurisation of the process streams in a refinery, con-

siderable amounts of hydrogen are needed. The drive to use cheaper heavy crudes with a

higher sulphur content and stricter environmental regulations may even increase the amount

of hydrogen needed above the amount that is normally produced by a refinery (in e.g. the plat-

former). Going to very low sulphur levels in refinery products might need the use of excessive

extra amounts of hydrogen because of difficulties encountered in current desulphurisation

units to remove the last few ppm’s of sulphur compounds. It might therefore become neces-

sary to build special hydrogen producing plants (e.g., steam-reformers) at the refinery site.

Chapter 8116

The H2S formed in the hydro-desulphurisation process can be removed from the product

stream in a variety of ways. Commonly used methods are chemical reaction with, for exam-

ple, zinc oxide or iron oxide, caustic scrubbing and absorption processes. For the H2S decom-

position processes treated in this chapter, only the absorption/desorption methods are of im-

portance. Most used are absorber/stripper combination units with an alkanolamine as absorb-

ing compound [2].

After the absorber/stripper unit, in conventional operations the pure H2S is fed to a Claus unit

where the H2S is converted to elemental sulphur and H2O. The Claus unit can be equipped

with an after-treatment to enhance conversions. Another method to decompose H2S to less

harmful compounds is the thermal dehydrogenation of H2S to hydrogen and sulphur. Both

processes will be treated in detail in the remainder of this chapter.

2. Sulphur recovery by Claus processes

2.1 The conventional Claus process

The most commonly used process for sulphur recovery from H2S is the Claus process. This

process was developed by C.F. Claus in 1883 [3]. I.G. Farbenindustrie AG modified the proc-

ess in 1936 to its current lay-out, including a thermal and a catalytic separation step [4].

The basic reactions occurring in the Claus process are [5]:

2H2S + O2 • H2O + 2S

2H2S + 3O2 o 2H2O + 2SO2

2H2S + SO2 o 2H2O + 3S

Of these reactions, the first two take place in the thermal stage (reaction furnace) and the last

one in the catalytic step.

Basically there are two different conventional Claus processes, the straight-through process

and the split-flow process. For gas mixtures containing 50-100 mol-% H2S the straight-

through process is recommended, for the other mixtures the split-flow process is more appro-

priate. Both processes will be treated in some more detail below [2]. Several modifications to

the original Claus process have been developed in the mean time. One of them, the Superclaus

process is also treated below, for other variations on the conventional process, the interested

reader is referred to [2].

The thermal dehydrogenation of H2S in a membrane reactor 117

The straight-through process

In the normal straight-through process, see Figure 1, the complete H2S stream is fed to a

burner together with the amount of air required to burn one third of the H2S to SO2 to obtain

the required gas mixture for the catalytic step of the process [2,5]. The reaction temperature in

this thermal step is about 1000-1400ºC, and in this step 60 to 70% of the H2S in the gas is di-

rectly converted to elemental sulphur. The sulphur is condensed by cooling the gases first in a

waste heat boiler, and after that in a sulphur condenser. After elemental sulphur removal the

gas mixture is reheated to 250-300ºC and fed to a catalytic converter. Three types of catalyst

are currently used [5]:

• Standard α-alumina.

• α-Alumina promoted with alkaline or alkaline earth elements, iron or nickel sulphates or

titania.

• Non-promoted or Ca-promoted titania.

In the catalytic converter, after cooling, an additional 25% of the original gas mixture is con-

verted to elemental sulphur. The second and possibly third catalytic stages are operating at a

somewhat lower temperature: 200-250ºC. After the last sulphur condenser, the exhaust gases,

containing ~3 to 6 mol-% H2S, are incinerated and the resulting SO2 is released to the atmos-

phere.

Figure 1: Two-stage straight-through Claus plant [2].

Chapter 8118

The split-flow process

The split-flow process is used when the H2S concentration in the gas stream is too low to ob-

tain a stable combustion. In this process, one-third of the total gas stream is fed to the reaction

furnace and the rest is bypassed and directly fed to the first catalytic converter. As a conse-

quence, the production of elemental sulphur from the thermal step is less than in the straight-

through process. The remaining part of the process is essentially the same as the straight-

through process.

2.2 Claus tail gas treatment processes

The H2S concentration in the tail gas of a conventional Claus plant is still some 5%. This H2S

is normally incinerated to SO2 and released to the atmosphere. Due to stricter environmental

regulations a large number of new technologies based on Claus tail gas treatment have been

developed to minimise the SO2 exhaust from sulphur recovery units. The Superclaus process

and the Shell Claus Off-Gas Treating (SCOT) process are treated below. For descriptions of

other tail-gas processes, the reader is referred to [2].

The Superclaus process

In the so-called Superclaus process, the last catalytic converter of the conventional Claus pro-

cess is replaced by a catalytic converter containing the “Superclaus catalyst”. This catalyst

consists of an α-alumina support with iron and chromium oxides as catalytic material. The

Superclaus catalyst is highly selective for the direct oxidation of hydrogen sulphide:

2H2S + O2 • 2H2O + 2S

This reaction is considered irreversible and the catalyst will convert more than 85% of H2S to

elemental sulphur, the rest is converted to SO2 and other sulphur compounds.

Two different Superclaus processes have been developed: Superclaus 99 and Superclaus 99.5.

In the Superclaus 99 process, the plant typically consists of one thermal stage, two or three

conventional catalytic Claus steps and a Superclaus reactor. In the Superclaus 99 process

99.0% of all sulphur in the feed is recovered. Superclaus 99.5 introduces a hydrogenation

stage between the last conventional Claus reactor and the Superclaus reactor. In this hydro-

genation stage all sulphur-containing compounds of the gas stream are converted to H2S over

The thermal dehydrogenation of H2S in a membrane reactor 119

a cobalt/molybdenum catalyst. A minimum overall sulphur recovery of 99.4% is claimed

when three Claus reactors are included.

The SCOT process

The SCOT process provides an efficient way of removing sulphur-containing compounds

from the tail gas of a conventional Claus reactor. The tail gas is heated to about 300ºC and fed

to a hydrogenation reactor, where all sulphur compounds in the gas are converted to H2S. Al-

most all H2S is removed in an absorber/stripper combination and fed back to the Claus plant.

The off-gas from the absorber contains virtually no sulphur compounds (values as low as 500

ppm are reported [2]) and is incinerated in the Claus incinerator. A schematic diagram of the

SCOT process is provided in Figure 2.

3. Thermal dehydrogenation of H2S in a membrane reactor

With the drive to use more heavy crudes with high sulphur content and with acceptable sul-

phur levels in fuels decreasing further, an increasing amount of hydrogen is needed in the hy-

dro-desulphurisation units. For removing the last traces of sulphur from the product streams,

an excessively large extra amount of hydrogen is needed. Therefore, at a certain level of hy-

drogen consumption, the refinery self-supply will no longer be sufficient; several approaches

can be chosen to solve this problem:

Figure 2: The SCOT process [2].

Chapter 8120

• By building a hydrogen production plant, for example a steam-reformer, one can optimise

hydrogen recovery from process streams.

• The platformer stream, for instance, contains considerable amounts of recoverable hydro-

gen, which could be retrieved with membrane units.

• Recovery of the hydrogen, contained in the H2S. In the Claus process this is impossible,

because all hydrogen is converted to water, but in the thermal dehydrogenation of H2S,

hydrogen is produced according to [6]:

H2S o H2 + 1/xSx

This reaction is, however, strongly endothermic and is thermodynamically unfavourable

at temperatures below 1500ºC [6]. With catalysts, for example MoS2, the reaction yield

can be much improved, but a maximum reaction yield of about 30% at temperatures of

about 1100ºC does not offer very good perspectives for industrial use.

Decomposition of hydrogen sulphide over MoS2 at 800ºC with intermittent removal of

hydrogen, however, led to a 95% conversion in 10 runs [10]. By performing the dehydro-

genation reaction in a hydrogen-selective membrane reactor in which the hydrogen is

continuously removed, one would be able to increase the conversion to 95% in just one

step. The remaining H2S can be treated in a Claus-unit and the produced hydrogen can,

after final purification, be used again for hydrogenation of sulphur compounds in the hy-

dro-desulphuriser. In literature several reactors using different types of membranes have

been investigated. The results will be treated below, together with a proposal based on the

new developed silica membranes described in this thesis.

3.1 Metal membranes

Because Pd-based metal membranes, commonly used for hydrogen separation [11] are not

resistant towards sulphur, not much research has been performed on the use of such mem-

branes in H2S dehydrogenation reactors. Some success has, however, been reported by Edlund

and Pledger [12]. They developed a platinum-based layered metal membrane that could resist

irreversible attack by H2S at 700ºC. At this temperature a conversion of 99.4% was achieved

in the membrane reactor. Without hydrogen removal the conversion was only 13%. No per-

meance data is provided, but platinum-based metal membranes are known for their low hy-

drogen permeance [14]. Johnson-Matthey developed palladium composite membranes with a

hydrogen permeance of about 1*10-6 mol/m2sPa [14], but these are most probably not resis-

The thermal dehydrogenation of H2S in a membrane reactor 121

tant towards sulphur because of the well-known irreversible attack of palladium by sulphur

[12]. The low permeance of the (platinum) metal membranes would imply that for industrial

application of these membranes very large membrane areas should be installed, which would

involve large amounts of noble metals and therefore large costs.

3.2 Vycor glass membranes

Kameyama et al. [16-19] studied the use of Vycor glass membranes in a permselective mem-

brane reactor for the thermal dehydrogenation of H2S over a MoS2 catalyst. Vycor glass is a

mesoporous glass with 4 nm pores. As a result, the maximum selectivity which can be ob-

tained with Vycor glass membranes is Knudsen selectivity. The Vycor glass was found to be

stable up to 800ºC above which temperature shrinkage of the glass starts to occur. The con-

version towards hydrogen was reported to be twice as high as the equilibrium conversion

without hydrogen removal. A large drawback of this type of membranes is the low selectivity

towards hydrogen, which is reported to be of the order of 2 (H2/H2S). This low selectivity

caused the only slight increase in conversion of a factor two. In addition, the hydrogen per-

meance through mesoporous Vycor glass membranes is about a factor 10 lower compared to

the state-of-the-art supported silica membranes described in this thesis. The required mem-

brane area for a membrane reactor would therefore at least be a factor 10 larger.

3.3 Silica membranes

The above-mentioned results on metal and Vycor glass membranes made us believe that the

membranes which are developed for steam-reforming conditions, as described in this thesis,

might be used as a starting point for developing a H2-permselective a H2S dehydrogenation

membrane reactor.

State-of-the-art

As shown in chapter 6, silica membranes can nowadays be prepared at temperatures as high

as 825ºC while state-of-the-art steam-stable γ-alumina membranes are prepared at 1000ºC.

This enables the use of such membranes in the high-temperature range needed for thermal de-

hydrogenation of H2S under conditions used in literature. The permselectivity of H2/CO2 of

the silica membranes prepared at 825ºC was >100 with a hydrogen permeance of

Chapter 8122

1*10-7 mol/m2sPa measured at 200ºC. The H2/CO2 selectivity is a good measure for what can

be achieved for H2/H2S selectivity, because the kinetic diameter of CO2 is 3.3 Å whereas the

kinetic diameter of H2S is 3.6 Å [20].

Objectives

Private communication with Shell [21] revealed that industrial objectives for a membrane re-

actor for H2S dehydrogenation are:

• Operating temperature: 1000-1250ºC

• Operating pressure: 2 bar

• Capacity: 300 ton/day H2S

• H2 flux through membrane: 20 ton/day

• Needed membrane area: 600 m2 (based on a hydrogen permeance of

1*10-6 mol/m2sPa at 1000ºC)

Main innovations needed

The results obtained for microporous silica membranes in the membrane steam-reforming

project, described in this thesis, provide favourable perspectives to realise a H2-permselective

membrane reactor for the dehydrogenation of H2S. Realisation of such a reactor, however,

imposes significant scientific and technical challenges.

Partly the challenges, addressed in the proposed project lie in the field of stability of the

membranes under the harsh process conditions (1000ºC, H2S atmosphere.) Based on the re-

sults obtained with Vycor glass, mentioned above, there is no reason to doubt that the silica

membrane material is resistant towards H2S and other sulphur compounds. The high tem-

perature might, however, invoke problems. Although defect-poor membranes were prepared

at 825ºC, no thermal stability tests were performed for temperatures higher than 600ºC. Cur-

rently there are no data available on very high temperature stability, i.e., 1000ºC or even

higher, of silica membranes under process conditions. Moreover, up to date no silica mem-

branes have been prepared at temperatures as high as 1000ºC. Hence it might be necessary to

modify the silica structure of the top layer of the membrane to make silica membranes more

resistant to high temperatures.

State-of-the-art silica membranes, calcined at 825ºC, have a hydrogen permeance at 200ºC, 10

times lower than the required 1*10-6 mol/m2sPa. No activation energies for hydrogen permea-

The thermal dehydrogenation of H2S in a membrane reactor 123

tion have been measured for these membranes yet, but a factor 10 in hydrogen permeance is

easily achieved when raising the temperature from 200 to 1000ºC.

It may be necessary to improve membrane selectivities, so that further purification of the pro-

duced hydrogen before re-use in the desulphurisation units can be limited as far as possible.

Moreover the membrane reactor can be optimised for various variables, such as H2S conver-

sion, hydrogen recovery, membrane area and temperature. In a techno-economic evaluation

combined with advanced process design the impact of different operating parameters on the

investment and operating costs should be studied.

Much attention should be paid to the choice of the catalyst for the reaction. Various catalysts

of which MoS2 is the most common, are possible. In a separate subtask, the impact of the used

catalyst on the stability of the membrane should be studied. E.g., alkali doping might have a

negative effect on the stability of the silica layer due to keatite fomation as described in this

thesis.

Proper sealing of the alumina tubular substrates to the reaction will provide an important

challenge as well. As shown in the steam-reforming work, glass or glass-ceramic sealing of

tubes to units operating at high temperatures is not straightforward and might involve some

considerable technical risk.

4. Conclusions

As stated above, hydrogen management in refineries is expected to become an increasing

challenge due to extensive desulphurisation treatments in the various process streams. It

would therefore be advantageous to have a process in which part of the hydrogen used for

desulphurisation is recovered from the formed H2S in a membrane reactor for the thermal de-

hydrogenation of H2S. The unreacted H2S can be further treated in conventional (Super)Claus

units equipped with a suitable after-treatment to reach the imposed environmental regulations

on SO2 exhaust of a refinery.

Although silica membranes have not been studied for a process like the thermal dehydrogena-

tion of H2S, we believe, that silica membranes make a good change to be highly suitable for

the process, based on the knowledge obtained on membrane behaviour under steam reforming

Chapter 8124

conditions. The advantages of microporous silica membranes over other membrane materials

include:

• Ease of preparation,

• Low costs per square meter membrane area,

• Rrelatively high hydrogen permeance,

• Relatively high selectivity,

• High temperature stability, and

• Assumed high stability towards sulphur containing environments.

5. References

1. K. Harries-Rees, Sulphur, 236 20- (1995).

2. A. Kohl, R. Nielsen, “Gas Purification”, 5th edition, Gulf Publishing Company, Houston, USA (1997).

3. C.F. Claus, British Patent 5958 (1883).

4. H. Baehr, “Gas Purification by the IG Alkacid Process and Sulfur Recovery by the IG Clauss Process”, Re-

finer Natural Gas. Manuf., 17 [6] 237-44 (1938).

5. A. Piéplu, O. Saur, J.C. Lavalley, “Claus Catalysis and H2S Selective Oxidation”, Catal. Rev. Sci. Eng., 40

[4] 409-50 (1998).

6. M.E.D. Raymont, “Make Hydrogen from Hydrogen Sulfide”, Hydrocarbon Proc., [7] 139-42 (1975).

7. N.I. Dowling, J.B. Hyne and D.M. Brown, “Kinetics of the Reaction between Hydrogen and Sulfur under

High-Temperature Clause Furnace Conditions”, Ind. Eng. Chem. Res., 29 2327-32 (1990).

8. N.I. Dowling and P.D. Clark, “Kinetic Modelling of the Reaction between Hydrogen and Sulfur and Op-

posing H2S Decomposition at High Temperatures”, Ind. Eng. Chem. Res., 38 1369-75 (1999).

9. V. Kaloidas and N. Papayannakos, “Kinetics of Thermal, Non-Catalytic Decomposition of Hydrogen Sul-

phide”, Chem. Eng. Sci., 44 [11] 2493-500 (1989).

10. K. Fukuda, M. Dokiya, T. Kameyama and Y. Kotera, “Catalytic Decomposition of Hydrogen Sulphide”,

Ind. Eng. Chem. Fundam., 17 [4] 243-48 (1978).

11. F. Sakamoto, Y. Kinari, F.L. Chen and Y. Sakamoto, “Hydrogen Permeation Through Palladium Alloy

Membranes in Mixtures Gases of 10% Nitrogen and Ammonia in the Hydrogen”, Int. J. Hydrogen Energy,

22 [4] 369-75 (1997).

12. D.J. Edlund and W.A. Pledger, “Thermolysis of Hydrogen Sulfide in a Metal-Membrane Reactor”, J. Mem-

brane Sci, 77 255-64 (1993).

13. D.J. Edlund and W.A. Pledger, “Catalytic Platinum-Based Membrane Reactor for Removal of H2S from

Natural Gas Streams”, J. Membrane Sci., 94 111-19 (1994).

14. V.M. Gyraznov, “Platinum Metals as Components of Catalyst-Membrane Systems”, Platinum Metals Rev.,

36 [2] 70-79 (1992).

The thermal dehydrogenation of H2S in a membrane reactor 125

15. J.E. Philpott, “Hydrogen Diffusion Technology, Commercial Application of Palladium Membranes”, Plati-

num Metals Rev., 29 [1] 12-16 (1985).

16. T. Kameyama, K. Fukuda, M. Fujishige, H. Yokokawa and M. Dokiya, “Production of Hydrogen from Hy-

drogen Sulfide by Means of Selective Diffusion Membranes”, Hydrogen Energy Prog. Res., 2 569-79

(1981).

17. T. Kameyama, M. Dokiya, M. Fujishige, H. Yokokawa and K. Fukuda, “Production of Hydrogen from Hy-

drogen Sulfide by Means of Selective Diffusion Membranes”, Int. J. Hydrogen Eng., 8 [1] 5-13 (1983).

18. T. Kameyama, M. Dokiya, M. Fujishige, H. Yokokawa and K. Fukuda, “Possibility for Effective Produc-

tion of Hydrogen from Hydrogen Sulfide by Means of a Porous Vycor Glass Membrane”, Ind. Eng. Chem.

Fundam., 20 97-99 (1981).

19. T. Kameyama, K. Fukuda, M. Fujishige, H. Yokokawa and M. Dokiya, “Production of Hydrogen from Hy-

drogen Sulfide by Means of Selective Diffusion Membranes”, Adv. Hydrogen Energy, 2 569-79 (1981).

20. D.W. Breck, page 637 in: “Zeolite Molecular Sieves”, Wiley (1974).

21. W.S. Kijlstra, Shell Global Solutions, private communication (1999).

&KDSWHU

(YDOXDWLRQUHFRPPHQGDWLRQV

1. Introduction

In this final chapter an overview will be given of the main innovations obtained in the investi-

gations presented in this thesis. Additional recommendations for future research are provided,

while, on the basis of the experience acquired in the project, a general membrane optimisation

scheme is presented.

2. Evaluation of the project

In this section the most important experimental results obtained in the research presented in

this thesis are summarised.

• A Techno Economic Evaluation revealed the main factors influencing the costs and

thereby the cost-effectiveness of a membrane steam-reformer compared to conventional

processes (chapter 2). The evaluation showed that membrane steam reforming can be

cost-effective.

• Large effort has been put in the preparation of centrifugal cast tubes that are superior in

homogeneity, roundness, strength and smoothness to any commercially available product

(chapter 4).

• Completely hydrothermally stable intermediate layers were developed (chapter 5). Syn-

thesis routes were developed to realise mesoporous membranes with tailor-made pore-

radii ranging from 2 to 10 nm.

• By performing the coating of the membrane layers under cleanroom conditions, silica

membranes with a very high permselectivity were obtained (chapter 6).

Chapter 9128

• By improving high temperature sealing procedures, a few permeance and permselectivity

measurements of the membranes at temperatures as high as 600ºC were possible (chap-

ter 6).

• CVI silica membranes, which show reasonable permselectivity and permeance, were pre-

pared at a relatively low synthesis temperature (chapter 7).

3. Proposed general membrane optimisation scheme

On basis of the experience acquired in the current project some general recommendations can

be made to speed up further research on membranes. This resulted in the following scheme,

which does not largely depend on the intended application of the membrane.

Supports

1. For lab-scale stability and permselectivity studies use a support with a surface that is as

smooth as possible.

2. For membrane testing under process-conditions, select a commercially available macro-

porous support that meets permeance and stability requirements. Make a well-considered

choice between the different geometries available (flat, tubular, multichannel or hollow

fibre) and test the support system under process-conditions.

3. Start as soon as possible with coating experiments on the finally selected commercial

supports.

Intermediate layer

1. Do lab-scale stability tests first on unsupported membrane material. This saves time and

material-consuming coating steps on lab-scale supports. Test on changes in:

◊ Specific surface area (BET)

◊ Crystal structure (XRD)

2. With selected materials from the tests on unsupported material, coat flat membranes and

perform stability and possibly permselectivity measurements on these membranes. Test

on (changes in):

◊ Pore size (permporometry)

Evaluation & recommendations 129

◊ Integrity and adherence (Scotch Tape Test and SEM)

◊ Permselectivity (depending on application, not always necessary)

3. Possibly perform lab-scale stability tests on tubular membranes (centrifugal cast tubes).

Test on (changes in):

◊ Pore size (permporometry)

◊ Large defects (bubble point testing)

◊ Permselectivity (depending on application)

4. Coat the selected commercial supports and test them under process conditions.

Top-layer

1. Do lab-scale stability tests first on unsupported membrane material, this saves time and

material consuming coating steps on lab-scale supports. Test on changes in:

◊ Specific surface area (BET)

◊ Crystal structure (XRD)

2. With selected materials from the tests on unsupported material, coat flat membranes and

perform stability tests. Test on (changes in):

◊ Pore-size (permselectivity measurements)

◊ Integrity and adherence (SEM, TEM)

3. Possibly coat tubular centrifugal cast membranes with well-developed intermediate layers

and perform stability tests. Test on changes in pore-size (permselectivity measurements).

4. Coat commercially available supports with developed intermediate layer and test them

under process conditions.

Lab-supports should promote the development in industry of commercial supports with better

dimensional definition and surface quality, so that the extra development steps needed on the

high-quality lab-supports can be omitted.

For choosing the geometry, the following should be considered.

• When the main purpose of a project is a separation process, one has to consider using

hollow fibres, multibore tubes or flat stacks.

• When the main is an integrated membrane reactor, however, single tubes are worth con-

sidering. Using membrane modules based on single tubes, one creates the necessary space

for the catalyst. Moreover gas-phase limitations will be more important in reactors and

they are more easily solved when using single tubes.

Chapter 9130

4. Recommendations & new concepts

4.1 Membranes for steam-reforming

Considering separative properties, the quality of the membranes prepared according to the

methods described in this thesis is good. For application under steam-reforming conditions,

however, several improvements have to be made.

For the application of microporous silica membranes in steam-containing environments it is

of major importance that the silica membranes will be tested on hydrothermally stable sup-

ports. Silica membranes should be prepared on the basis of the results of the specific surface

area measurements described in chapter 6. Unsupported silica membrane material of which

the specific surface area does not change under SASRA conditions is most promising. An ex-

ample is silica fired at 825ºC (chapter 5). The need of doping the silica with foreign ions or

atoms is currently uncertain.

Apart from the above-mentioned membrane properties, also the sealing of the membranes is

of utmost importance. When going to higher temperatures, the conventional polymeric sealing

techniques do not suffice anymore and new systems like glass and glass-ceramic seals should

be used. As shown in chapter 6, high temperature seals have been developed, showing suffi-

cient stability. Large improvements are needed, however, concerning the adherence of the

membrane layers to the seal and the quality of the seal itself. In membrane applications at

very high temperatures (>800ºC), the development of the sealing might involve a very high

technical risk, as described in chapter 8. The need for more research dedicated specially to-

wards the development of new sealing materials and sealing techniques is evident.

When suitable membranes are available real process technology and catalysis-oriented re-

search can be performed. Important fields of research for membrane steam-reformers may in-

clude (see also chapter 2):

• Influence of lower hydrogen concentration on the performance of the catalyst:

◊ Coking rates.

◊ Oxidation of the catalyst.

• Conversion as a function of operation conditions:

◊ Temperature.

Evaluation & recommendations 131

◊ Pressure.

◊ Steam : methane ratio.

◊ Hydrogen concentration.

• Influence of the silica membrane on the activity of the catalyst.

• Influence of potassium from the catalyst on the stability of the silica layer (possible

keatite formation).

4.2 Other developments & recommendations

A set-up for a project proposal based on the results attained was already presented in chap-

ter 8 of this thesis, here some other developments and recommendations will be discussed.

A new development is the synthesis of hydrophobic silica membranes, as described by De

Vos [1]. These membranes have large potential for separation of gaseous streams containing

H2O. Permselectivities as high as 50 were already measured for H2/propane, though the

H2/CH4 permselectivity (=7) was slightly above the Knudsen selectivity (=2.8) at the same

conditions. It is expected that optimisation of the synthesis route will result in a large increase

in permselectivities also towards smaller molecules in the near future. An important related

objective is the preparation of an all-hydrophobic membrane, i.e., a membrane consisting of a

hydrophobic support, hydrophobic intermediate layer and hydrophobic top-layer. Some initial

experiments are currently being performed for the preparation of a hydrophobic support.

Because the currently used γ-alumina is not stable in all acid and basic environments used in

industry [2], the development of mesoporous layers other than γ-alumina deserves attention as

well. Most common materials that can be used for the mesoporous layer are: zirconia and ti-

tania [3,4], but recently also the preparation of mesoporous hafnia is described [5]. Hafnia

seems to be a very interesting membrane material, because it can, unlike zirconia and titania,

be fired up to 1850ºC without a phase transformation of its monoclinic form. Hafnia also has

a high chemical resistance toward acid and basic media. Another interesting material, cur-

rently under investigation by the group of Brinker is mesoporous silica [6,7]. This material is

especially interesting because a tailor made morphology and pore-size is possible.

When using these mesoporous layers for the preparation of microporous silica membranes,

the following requirements should be fulfilled (of course depending on the application):

• A pore-radius that is sufficiently small to enable the application of a silica layer.

Chapter 9132

• Good hydrothermal stability

• Good acid/base resistance

For applications where only the mesoporous layer is used, e.g. as a nanofiltration membrane,

the surface charge also might play an important role for the specific application. To cover a

complete range of applications, one should not only cover a complete range of pore-sizes in

the used membrane materials, but also a range in surface charge on the membrane pores.

5. References

1. R.M. de Vos, “High Selectivity, High-Flux Silica Membranes for Gas Separation. Synthesis, Transport and

Stability”, PhD Thesis, University of Twente, 1998.

2. J.M. Hofman-Züter, “Chemical and Thermal Stability of (Modified) Mesoporous Ceramic Membranes”,

PhD Thesis, University of Twente, 1995.

3. A. Larbot, J.P. Fabre, C. Guizard and L. Cot, “New Inorganic Ultrafiltration Membranes: Titania and Zir-

conia Membranes”, J. Am. Ceram. Soc., 72 [2] 257-61 (1989).

4. C.H. Chang, R. Goplan and Y.S. Lin, “A Comparative Study on Thermal and Hydrothermal Stability of

Alumina, Titania and Zirconia Membranes”, J. Membrane Sci., 91 27-45 (1994).

5. P. Blanc, A. Larbot, J. Palmeri, M. Lopez and L. Cot, “Hafnia Ceramic Nanofiltration Membranes. Part I:

Preparation and characterisation”, J. Membrane Sci., 149 151-61 (1998).

6. Y. Lu, R. Ganguli, C.A. Drewien, M.T. Anderson, C.J. Brinker, W. Gong, Y. Guo, H. Soyez, B. Dunn,

M.H. Huang and J.I. Zink, “Continuous Formation of Supported Cubic and Hexagonal Mesoporous Films

by Sol-Gel Dip-Coating”, Nature, 389 364-8 (1997).

7. A. Sellinger, P.M. Weiss, A. Nguyen, Y. Lu, R.A. Assink, W. Gong and C.J. Brinker, “Continuous Self-

Assembly of Organic-Inorganic Nanocomposite Coatings that Mimic Nacre”, Nature, 394 256-60 (1998).

6XPPDU\

In process industry hydrogen is a very important feedstock for a large number of processes

like the production of ammonia (fertiliser), various hydrogenation reactions and electricity

production in gasturbines. In future hydrogen will play an even more important role because

of the rise of the “hydrogen based society”. This becomes already visible in the fuel cell car

recently taken in development.

The classical way to produce hydrogen is steam reforming of hydrocarbons, mostly methane.

The overall steam reforming reaction is given by:

CH4 + 2H2O o CO2 + 4H2

For thermodynamical reasons equilibrium conversion reaches 90% only at a temperature as

large as 1000ºC. This makes the process energetically very unfavourable. By removing the

produced hydrogen selectively from the reaction zone (for example by using hydrogen selec-

tive membranes), the same conversion can be attained at 600-700ºC. When this can be real-

ised environmental benefits are evident.

Besides the environmental effects, also the cost of the hydrogen produced in such a membrane

reactor is of main importance. Therefore, in chapter 2 of this thesis an analysis is provided

which describes not only the requirements for the membrane, but also for the complete mem-

brane reactor process (process technology) to make such a reactor economically feasible. In

this chapter membrane-related quantities like flux, selectivity and stability (thermal, chemical

and mechanical) were treated besides aspects like sealing technology (and the costs thereof),

the catalyst used, operating temperature, etc. The conclusion was that it is virtually impossible

to make a statement beforehand on economical grounds whether a membrane reactor can be

constructed in a cost-effective way. Small changes in the operating temperature of the reactor

may have significant consequences for a membrane reactor being economically feasible or

not.

It was tried to make membranes using the specifications obtained in the technical economic

evaluation. Of main importance besides the economic considerations was of course the stabil-

ity of the membrane under steam reforming conditions. To test the stability of the prepared

membranes, so-called Simulated Ambient Steam Reforming (SASRA) conditions were used.

Summary134

These are: 30 bar total gas pressure with methane to steam ratio of 1:3 at a temperature of

600ºC. The main requirement was that a developed microporous membrane should withstand

these conditions for at least 100 hours without any signs of degradation.

To realise this, first very homogeneous α-Al 2O3 supports were prepared by using colloidal

techniques. Colloidal filtration was used for the preparation of flat supports, while a centrifu-

gal casting technique was developed to prepare highly smooth and round tubular supports.

On these supports mesoporous γ-Al 2O3 layers were dipcoated. These layers were, however,

not stable to above-mentioned SASRA conditions. After a SASRA-treatment these layers

showed serious delamination from the support. The application of a phosphorous containing

layer in between the support and the mesoporous layer showed to solve the problem of de-

lamination. The phosphate groups of this layer were found to behave as a sort of adhesive

between the support and the mesoporous γ-Al 2O3 layer. The formed bond showed to be stable

enough to withstand the SASRA conditions. To minimise pore-growth of the γ-Al 2O3 layer

during SASRA-treatment, the γ-Al 2O3 was doped with lanthanum and sintered at 1000ºC. The

resulting mesoporous γ-Al 2O3 membrane had a pores with a Kelvin radius of about 6 nm and

was completely resistant towards the SASRA conditions.

Unfortunately due to time limitations, there was no time to coat these steam stable mesopo-

rous membranes with a microporous silica toplayer. Microporous (doped) silica membranes

have, however, been applied on conventional mesoporous γ-Al 2O3 membranes which were

not stable under SASRA conditions. On these membranes permeance measurements have

been performed which showed that these membranes could be prepared with a very high se-

lectivity under cleanroom conditions. Stability measurements on (doped) silica bulk material

sintered at a high temperature (600-800ºC) showed no change in the specific surface area

during SASRA treatment. This is an indication that it should be possible to prepare a silica

membrane that shows complete stability towards SASRA conditions.

On the basis of the results obtained in the project in chapter 8 a concept proposal is provided

for the use of the developed membranes for the thermal dehydrogenation of H2S in a mem-

brane reactor.

6DPHQYDWWLQJ

In de hedendaagse industrie is waterstof al een zeer belangrijke grondstof voor tal van proces-

sen, zoals de productie van ammonia (kunstmest), diverse hydrogeneringsreacties en de pro-

ductie van electriciteit in gasturbines. In de toekomst zal onder meer door de opkomst van de

“waterstof maatschappij”, zoals nu al zichtbaar in de brandstofcel auto, de rol van waterstof

alleen maar toenemen.

De klassieke manier om waterstof te maken is de stoom reformering van koolwaterstoffen,

met name methaan. De overall stoom reformeringsreactie voor methaan wordt gegeven door:

CH4 + 2H2O o CO2 + 4H2

Energetisch gezien is dit een ongunstige reactie, want door de ligging van het evenwicht

wordt pas bij een temperatuur van 1000°C 90% conversie van methaan gehaald. Wanneer bij

stoom reformering nu selectief de geproduceerde waterstof uit de reactiezone van de reactor

kan worden afgevoerd (bijvoorbeeld door een waterstof selectief membraan), dan kan dezelf-

de conversie mogelijk bij 600-700°C gehaald worden. Als dit gerealiseerd kan worden, is het

grote milieuvoordeel evident.

Aangezien niet alleen een milieuvoordeel van belang is, maar ook de kostprijs van de gepro-

duceerde waterstof, werd in hoofdstuk 2 van dit proefschrift een analyse gemaakt aan welke

eisen een waterstof selectief membraan, maar ook de gehele membraanreactor opzet (proces-

technologie) moet voldoen om het geheel eonomisch verantwoord te maken. Hierbij kwamen

niet alleen aan het membraan gerelateerde eisen zoals, flux, selectiviteit en stabiliteit (ther-

misch, chemisch en mechanisch) om de hoek kijken, maar ook onderwerpen als sealing-

technieken en de kosten daarvan, eisen aan de gebruikte katalysator, procestemperatuur, en-

zovoort. Het bleek dat er te veel onzekerheden zijn om bijvoorbaat te zeggen, op economische

gronden, of een membraan reactor nu wel of niet economisch verantwoord gebouwd kan wor-

den. Door een kleine verandering van bijvoorbeeld de procestemperatuur van de reactor, ver-

andert het economisch plaatje compleet ten voor of ten nadele van het membraan reactor con-

cept.

Op basis van de uitkomsten van het bovengenoemde technisch economisch onderzoek werd

geprobeerd een membraan te ontwikkelen wat zoveel mogelijk aan de gestelde eisen voldeed.

Samenvatting136

Daarbij was natuurlijk van het grootste belang dat het membraan stabiel was onder stoom re-

formerings condities. Om dit te testen werden zogenaamde gesimuleerde stoom reformerings

condities (SASRA-condities) gebruikt. Deze zijn: 30 bar totale gasdruk met een methaan :

stoom ratio van 1:3 bij een temperatuur van 600ºC. Het werd als eis gesteld dat een micropo-

reus membraan in ieder geval 100 uur deze condities zonder degradatie moest kunnen door-

staan.

Om dit te realiseren werden als eerste zeer homogene dragers bereid met nieuw ontwikkelde

colloidale technieken. Colloidale filtratie werd gebruikt voor de bereiding van vlakke dragers,

terwijl met centrifugale depositie ultra gladde en ronde buisvormige membraandragers wer-

den bereid.

Op deze dragers werden mesoporeuze γ-alumina membranen gedipcoat, welke helaas niet be-

stand bleken te zijn tegen de bovengenoemde SASRA condities. Na SASRA-behandeling

bleken deze lagen ernstige delaminatie verschijnselen te vertonen. De toepassing van een op

fosfaatverbindingen gebaseerde tussenlaag bleek het probleem grotendeels op te lossen. De

fosfaatgroepen van de fosfaatlaag bleken zich te gedragen als een soort lijm tussen de drager

en de γ−alumina laag. De gevormde verbinding bleek sterk genoeg om delaminatie onder

SASRA condities tegen te gaan. Om poriegroei van de γ-alumina laag gedurende SASRA-

behandeling tegen te gaan, werd het alumina gedoopt met lanthaan en gesinterd bij 1000°C.

Het resulterende mesoporeuze γ-alumina membraan had poriën van ongeveer 6 nm en was

volledig bestand tegen SASRA condities.

Helaas was de projectduur te kort om op deze mesoporeuze membranen nog te coaten met een

microporeuze silica toplaag. Wel zijn er aan (gedoopte) silica membranen op conventionele

(dus niet SASRA-stabiele) mesoporeuze γ-alumina membranen permeatie metingen verricht,

waaruit blijkt dat membranen met een zeer hoge selectiviteit geproduceerd kunnen worden.

Uit stabiliteitsmetingen aan silica bulkmateriaal blijkt dat voor de hoog gesinterde (gedoopte)

silica materialen (600-800°C) er geen verandering meer optreedt in specifiek oppervlak van

deze materialen gedurende SASRA-behandeling. Dit is een indicatie dat het in principe mo-

gelijk is een silica membraan te maken dat volledig stabiel zal zijn onder stoom reformerings

condities.

Op basis van deze resultaten is dan ook in hoofdstuk 8 een opzet gegeven voor een project-

voorstel voor het gebruik van de hier ontwikkelde membranen voor de thermische dehydro-

genering van H2S in raffinaderijen.

/HYHQVORRS

Tot u spreekt/schrijft de auteur dezes, Arian Nijmeijer. Ik werd 14 maart 1972 geboren in Es-

pel, gemeente Noordoostpolder. Hier heb ik ook mijn lagere school doorlopen. In 1990 haalde

ik het VWO diploma aan het Zuyderzee College te Emmeloord. Hierna ben ik aan de studie

Chemische Technologie begonnen aan de Universiteit Twente te Enschede. In december 1995

ben ik afgestudeerd bij de Vakgroep Anorganische Materiaalkunde op de ontwikkeling van

nieuwe materialen voor keramische condensatoren bij Prof. Verweij. Daarna ben ik direct in

december 1995 bij dezelfde vakgroep assistent in opleiding geworden. Gedurende deze perio-

de heb ik in Europees verband onderzoek gedaan naar de ontwikkeling van nieuwe micropo-

reuze keramische membranen voor de toepassing in hoge temperatuur membraanreactoren.

Deze zijn met name bedoeld voor de stoom reformering van methaan. De resultaten van het

onderzoek staan beschreven in dit proefschrift.

'DQNZRRUG

Geachte lezer, u heeft eindelijk het einde van dit proefschrift bereikt, alhoewel… Een aselecte

steekproef onder de lezers van proefschriften doet vermoeden dat een groot deel van deze le-

zers slechts geïnteresseerd is in dit laatste hoofdstukje en daarbij alle wetenschappelijke ver-

handelingen in het voorafgaande links laat liggen. De belangrijkste vraag die men zich daarbij

stelt is natuurlijk, “ik word toch wel genoemd?”. Deze vraag is natuurlijke een zeer terechte,

temeer omdat ik het hier beschreven onderzoek niet alleen gedaan heb. In de afgelopen vier

jaar heeft een klein leger aan mensen zich met mijn onderzoek beziggehouden. Deze mensen

wil ik dan ook gaarne allemaal van harte bedanken. Op het gevaar af dat ik, onbedoeld, men-

sen vergeet te bedanken doe ik hier dan toch maar een poging in onze onvolprezen “list bul-

let” stijl:

• Als eerste wil ik natuurlijk mijn promotor Henk Verweij bedanken voor het bieden van de

mogelijkheid om in zijn groep te promoveren. Ik heb de afgelopen vier jaar zeer veel van

hem geleerd.

• Vervolgens mijn grootste leermeester in deze tak van sport, Henk Kruidhof (in de wan-

delgangen “Henk 2”). Hij heeft mij de fijne kneepjes van de anorganische materiaalkunde

bij gebracht en zonder hem was het Brite-project nooit een succes geworden.

• José voor alle bijstand gedurende het hele promotietraject. Af en toe was een luisterend

oor zeer welkom!

• Mister Inorganic Membrane, Klaas Keizer is natuurlijk ook een onmisbare factor geweest

in het geheel. Zijn eeuwige vrolijkheid werkt aanstekelijk en de 4 weken in Potchef-

stroom waren dan ook een groot succes.

• Cindy, onze “poreuze technica”. In vrijwel alle hoofdstukken is wel iets van haar werk

terug te vinden.

• Attila voor het zeer snelle opbouwen (en weer afbreken) van allerhande opstellingen die

we in ons enthousiasme hadden bedacht.

• Gerrit voor het ontwerpen van sealing technieken en de hoge temperatuur permeatie cel.

• De mensen van de glasblazerij, de fijnmechanische werkplaats en de elektronica afdeling.

In het bijzonder natuurlijk Joop en Henk voor respectievelijk de keramische bewerkingen

aan de dragers en de ontwikkeling van allerhande apparatuur en materialen nodig voor het

Dankwoord140

membraan onderzoek. Beide heren droegen mede bij aan een onvergetelijke tijd in Pot-

chefstroom.

• De D-studenten, Patrick, Marcel, Ben, Peter, Richard, Hans en Niels die bijgedragen heb-

ben aan het onderzoek, van de een is iets meer terug te vinden in dit proefschrift dan de

ander, maar dat neemt niet weg dat eenieder op zijn eigen manier een steentje bijdroeg

aan het totstandkomen van het geheel. Het was in ieder geval erg gezellig. Ik hoop dat

jullie er wat van opgestoken hebben, ik in ieder geval wel. Daarbij wil ik Ben nog harte-

lijk bedanken voor de leuke tijd in Kaapstad, Richard voor de geslaagde vakanties naar

Kos en Mexico en natuurlijk Niels voor het ontwerpen van de omslag en het omzetten van

het geheel naar pdf formaat.

• My dear project partners, Rune Bredesen (for the nice times in Oslo, we should soon have

another evening in the Microbrewery or Café Amsterdam), Christian Simon, Bente Tilset

and Reidar Haugsrud from SINTEF, Frans en Maurice Velterop, Robert Kuipers en Theo

de Beer from Velterop BV, Herman Weyten en Jan Luyten from VITO, Jean-Alain Dal-

mon and Patrice Ciavarella from IRC, Frans Janssen en Robert Meijer from KEMA and

Arne Arnundskås from Norsk Hydro.

• Bernard voor de hulp bij het stuk over de RBS-metingen.

• Mijn (ex)kamergenoten, Martijn, Baukje, Renate, Nieck, Marjan (ook voor de strelende

woorden betreffende sommige hoofdstukken: “het leest wel lekker weg”), Wolfgang en

Bas voor de gezellige tijd. Bas natuurlijk ook omdat hij met mij zich in het avontuur van

een eigen bedrijf wil gaan storten. Voor geïnteresseerden: Twente Ceramic Solutions, p.a.

Universteit Twente, kamer CT-1749, Postbus 217, 7500 AE Enschede. Tel.: 053-

4892994. Bas ik dank je natuurlijk ook hartelijk voor het kritisch doorlezen van mijn con-

cept proefschrift.

• De buurtjes van 1733: René, André, Zeger (en Nicole natuurlijk) en Sven. Waarbij opge-

merkt dient te worden dat het toch wel dankzij René is dat ik bij “Dé Groep” ben geko-

men.

• Cis, onze enige echte vakgroep moeder. Waar zouden we zijn zonder Cis?

• Wim en Jan-Willem voor het “paranimfen” en voor de leuke avondjes in Twente dan wel

in Rotterdam en de zeer geslaagde vakanties naar Kos c.q. Ierland.

• Ton voor de (ont)spannende fietstochten in de Twentse en Duitse natuur en natuurlijk het

proeven van de diverse bieren na afloop van deze tochten. De trend van Grolsch om elke

Dankwoord 141

maand met een nieuw bier op de markt te komen maakt het promoveren er niet makkelij-

ker op.

• Jasper voor de leuke reizen naar Thailand en Cuba. Zulke reizen zijn zeer zeker aan te

raden. Je leert er enorm van op eigen benen te staan. Vooral Cuba kan als een uitdaging

worden gezien.

• Ben Kokkeler voor de leuke dagen c.q. nachten in het “Euregio Publiekscentrum voor

Sterrenkunde, EPS”. Ik weet nog steeds niet veel van astronomie, maar het EPS in com-

binatie met astronomie vereniging Andromeda heeft mij wel een hoop kennis doen verga-

ren in het organiseren van evenementen en reisjes.

• De reguliere studenten: Gerald, Coen, Sjoerd, Kim, Peter, Jeroen, Thymen, Paul, Herman,

Sven, Jaco, Erik, Joy, Christiaan en Johnny (literatuurstudies uitschrijven voor studenten

is een uitstekende manier om enige orde te scheppen in de enorme hoeveelheid artikelen

die een promovendus gedurende zijn promotietijd verzamelt), de “SINTEF” stagiares,

Niels, Jeroen en Simone en de Zuid-Afrika gangers Duco en Jiri.

• Dan nu de rest van de huidige groep: Tomas, Elise, Mark, Fiona, Marco (ook voor de leu-

ke vakantie naar Mexico), Henny, Samuel, Maarten (tevens voor het kritisch bekijken van

de manuscripten), Mercedes, Nela, Monse (mede voor de leuke week in Mexico City),

Werner, Matthijs, Ning, Ben, Natscha, Herman, Louis, Manon, George, Sjoerd, Marnix,

Jurian en Ronald.

• De (ex)huisgenoten Marc, Hanneke, Marit, Bert, Joop, Louis, Paula, Roy, Lars, Jeroen,

Sascha, Jochem, Harko, Peter en Daphne.

• En verder: Peter Schrap (onze eerste “werknemer”, tevens eeuwig student), Maarten

Oosterkamp, Bart Oosterlee, Erik Mallens, Erwin v/d Vis, Eddy Brinkman en Marco Rep.

• Natuurlijk last, but certainly not least, mijn ouders, Marc en Esther!

Nogmaals, mocht ik mensen vergeten zijn, dan mijn oprechte excuses van deze plaats. In ie-

der geval allemaal heel hartelijk:

%HGDQNW

$ULDQ