bhave r.r., (ed.) inorganic membranes. synthesis, characteristics and applications (only 1,2,3,6 and...

Contents

Foreword / vii

Preface / xi

1. The Developing Use of Inorganic Membranes: A HistoricalPerspective / 1

1.1. Introduction / 11.2. The Nuclear Period / 113. The Development of Ultrafiltration and Microfiltration

Inorganic Membranes: The 1980-1990 Period / 41.4. The Third Stage in the Development of Inorganic Membranes:

Further Developments by Other Organizations / 81.5. Conclusions / 8References / 8

2. Synthesis of Inorganic Membranes / 1 0

2.1. Introduction and Overview / 102.1.1. General Background: Membrane Types and

Structures / 102.2. Basic Principles of Membrane Synthesis / 14

2.2.1. Ceramic Asymmetric Membranes / 152.2.2. Glass Membranes / 172.2.3. Anodic Membranes / 172.2.4. Track-Etch Membranes / 182.2.5. Pyrolysis / 182.2.6. Dense Membranes / 18

2.3. Packing of Particles from Suspensions / 192.3.1. Introduction and Support Systems / 1 92.3.2. Sol-Gel Process / 212.3.3. The Slip-casting of Ceramic Membranes / 23

2.4. Typical Results for Different Materials / 262.4.1. Alumina Membranes / 262.4.2. Zirconia Membranes / 342.4.3. Titania Membranes / 35

XV

xvi CONTENTS

2.4.4. Silica Membranes / 372.4.5. Binary Composite Membranes / 38

2.5. Phase Separation/Leaching Methods and Glass Membranes / 392.6. Anodic Oxidation / 452.7. Pyrolysis / 49

2.7.1. Molecular Sieve Carbons / 492.7.2. Micro/Ultrafiltration Carbon Membranes / 532.7.3. Silica Membranes / 53

2.8. Track-Etch Method / 542.9. Composite Membranes: Modification Methods / 552.10. Miscellaneous Methods and Comments / 57References / 58

3. General Characteristics of Inorganic Membranes / 64

3.1. Introduction / 643.2. Commercial Inorganic Membranes / 6433. Microstructural Characteristics / 67

3.3.1. Microscopic Morphology / 673.3.2. Thickness / 723.3.3. Pore Size / 743.3.4. Permeabilities and Retention Properties / 793.3.5. Maximum Pore Size and Structural Defects / 803.3.6. Characteristics of Pore Network / 82

3.4. Materials Properties / 833.4.1. Chemical Resistance / 833.4.2. Surface Properties / 863.4.3. Mechanical Properties / 87

Membrane Element and Module Configurations / 883.6. End-seal and Module Packing Materials / 92References / 92

4. Permeation and Separation Characteristics of InorganicMembranes in Liquid Phase Applications / 95

4.1. Introduction / 954.2. Common Terminology and Definitions / 9543. Phenomena Influencing Flux and Separation

During the Filtration Process / 964.3.1. Concentration Polarization / 97

CONTENTS xvii

4.3.2. Adsorption / 984.3.3. Membrane Surface Characteristics / 1004.3.4. Overall Transport Resistance: Determination of

Membrane Permeability / 1024.4. Microfiltration / 105

4.4.1. Models for the Prediction of Flux / 1054.5. Ultrafiltration / 107

4.5.1. Permeation Models / 1074.5.2. Flux Characterization / 1084.5.3. Solute Retention Properties / 110

4.6. Liquid Permeation and Separation with Formed-in-Place(Dynamic) Membranes /1174.6.1. Transport Characterization /1184.6.2. Separation Performance /119

4.7. Dense Membranes / 123Nomenclature /124References /126

Liquid Filtration and Separation with Inorganic Membranes: OperatingConsiderations and some Aspects of System Design /129

5.1. Introduction / 1295.2. Cross-flow Membrane Filtration / 12953. The Eifect of Operating Parameters on Membrane Filtration

and Separation Performance / 1 3 15.3.1. Membrane Pore Size / 1 3 15.3.2. Feed Pretreatment / 1325.3.3. Cross-flow Velocity / 1345.3.4. Transmembrane Pressure / 1355.3.5. Temperature /1365.3.6. pH / 136

5.4. Backflushing: Theoretical Aspects / 1385.5. Membrane Regeneration / 1405.6. Microfiltration with Uniform Transmembrane Pressure / 1415.7. Operating Configurations / 144

5.7.1. Open System / 1445.7.2. Closed System / 1455.7.3. Feed and Bleed / 145

5.8. Some Aspects of System Design and Operation / 1475.8. Transmembrane Pressure / 1 4 85.8.2. Shear Rate / 149

xviii CONTENTS

5.8.3. Temperature / 1495.8.4. System Dead Volume / 1505.8.5. Backflushing: Operational Aspects / 1505.8.6. Other System Design Considerations / 1 5 1

Nomenclature / 153References / 153

6. Gas Separations with Inorganic Membranes / 155

6.1. Introduction / 1556.2. Porous Membranes / 1 5 6

6.2.1. Gas Separation by Knudsen Diffusion / 1 5 66.2.2. Gas Separation by Surface Diffusion / 1 6 06.2.3. Gas Separation by Multilayer Diffusion and

Capillary Condensation / 1 6 56.2.4. Gas Separation by Molecular Sieving / 1 6 7

63. Dense Membranes / 1696.4. New Developments / 1706.5. Conclusions / 1 7 1Nomenclature / 1 7 2References / 173

7. Inorganic Membrane Reactors to Enhance the Productivityof Chemical Processes / 177

7.1. Inorganic Membranes for High-Temperature Applications / 1777.2. Gas (or Vapor) Phase Reactions: The Concept

of the Membrane Reactor / 18073. Fundamental Aspects of Membrane Reactors / 1 8 4

7.3.1. Separative Membranes / 1847.3.2. Nonseparative Porous Membranes / 1967.3.3. Nonseparative Dense Oxide Membranes / 201

7.4. Assessment of Commercial Possibilities / 202References / 203

8. Inorganic Membranes in Food and Biotechnology Applications / 208

8.1. Introduction / 208

CONTENTS xix

8.2. Applications of Inorganic Membranes in the Dairy Industry / 2098.2.1. Microfiltration of Milk for Bacteria Removal / 2098.2.2. Concentration of Pasteurized Skimmed Milk /2138.2.3. Concentration of Whole Milk / 2138.2.4. Microfiltration in the Processing of Whey

to Produce Whey Protein Concentrate /2178.2.5. Concentration of Serum Proteins from Whey

by Ultrafiltration / 2208.2.6. Concentration of Acidified Milk to Produce

Fresh Cream Cheese / 2228.2.7. Rheological Behavior of Concentrates During

the Processing of Lactic Curds Using InorganicMembranes / 225

83. Inorganic Membranes in the Clarification of Fruit Juices / 2338.3.1. Apple Juice Clarification Using Ceramic Membranes / 2348.3.2. Clarification of Apple Juice Using Inorganic

Membranes on Porous Metallic Supports / 2358.3.3. Processing of Cranberry Juice with Ceramic

Membranes / 2398.3.4. Clarification of Strawberry and Kiwifruit Purees / 245

8.4. Applications of Inorganic Membranes to Concentrate Proteinsin Food Industry / 2478.4.1. Concentration of Soy-milk Proteins / 247

8.5. Clarification of Fermented Alcoholic BeveragesUsing Inorganic Membranes / 2488.5.1. Colloidal and Noncolloidal Wine Components / 2518.5.2. The Role of Filtration in the Process of Wine Making / 2528.5.3. A Comparative Evaluation of the Various Filtration

Processes Used in Wine Clarification / 2528.5.4. Influence of Process Variables on Cross-flow Filtration:

Clarification of White and Red Wines / 2538.5.5. Vinegar Filtration with Ceramic Membranes / 2568.5.6. Bacteria Removal with Inorganic Microfilters to

Produce "Cold" Sterile Beer / 2588.5.7. Recovery of Beer by Clarification of Tank Bottoms / 260

8.6. Inorganic Membranes in Biotechnology Applications / 2628.6.1. Microorganism Separation and Cell Debris

Filtration / 2658.6.2. Plasma Separation by Cross-flow Filtration

Using Inorganic Membranes / 268References / 271

xx CONTENTS

9. Inorganic Membranes for the Filtration of Water, WastewaterTreatment and Process Industry Filtration Applications / 27S

9.1. Introduction / 2759.2. Filtration of Water Using Inorganic Membranes / 276

9.2.1. Cross-flow Microfiltration for the Production ofDrinking Water / 277

9.2.2. Bacteria Removal from Water byInorganic Membrane Filters / 2 8 1

93. Non-oily Wastewater Treatment With Inorganic Membranes / 2829.3.1. Treatment of Textile and Paper Industry Effluents / 2829.3.2. Concentration of Latex Wastewaters / 2859.3.3. Miscellaneous Effluent Treatment Applications / 287

9.4. Oily Wastewater Treatment with Inorganic Membranes / 2889.4.1. Cross-flow Filtration of Oily Produced Water / 2889.4.2. Treatment and/or Recovery of Oils from Oily Wastes and

Oil-Water Emulsions / 2909.5. Inorganic Membranes in Process Industry Filtrations / 293

9.5.1. Recovery of Caustic from Process Effluents / 2969.5.2. Hydrocarbon Processing Applications / 296

References / 298

Appendix / 300

Index / 305

1. The Developing Use of InorganicMembranes: A Historical Perspective

J. GILLOT*Societe des Ceramiques Techniques

(a subsidiary of Alcoa Separations Technology, Inc.) Tarbes

1.1. INTRODUCTION

To most users, inorganic membranes are a relatively new product But in fact,their development started in the 1940s and can be schematically divided intothree periods:

1. The development and mass production of membranes for the separationof uranium isotopes by the process of gaseous diffusion applied to UF 6 .

2. Starting from this basis, the development and industrial use of a newgeneration of membranes adapted to the ultrafiltration and micro-filtration of process liquid streams.

3. The more recent research work on a much broader range of membranetypes aiming at separations using a variety of basic processes, includingthe coupling of catalytic reactions and membrane separation.

Many aspects of the development of uranium enrichment membranes were,and to a large extent still are classified, the scanty traces in the public domainonly being a number of patents (CEA 1958, Clement, Grangeon and Kayser1973, Miszenti and Mannetti 1971, Veyre et al. 1977). Much of the work doneat present to develop new and improved inorganic membranes is also moreor less classified.

Although the author participated in all three periods, this historicalperspective covers only the first two periods. The perspective on the import-ant developments in the third period as described above can be foundthroughout the book (especially in Chapters 6 and 7).

1.2. THE NUCLEAR PERIOD

Naturally occurring uranium contains a very small percentage (0.7%) of thefissile 2 3 5U isotope to be used either in nuclear weapons, which require a very

* With R. R. Bhave.

2 INORGANIC MEMBRANES: SYNTHESIS AND APPLICATIONS

high concentration, or in power generating plants, most of which require aconcentration of approximately 3 % . Since the separation by mass spectro-scopy used in the Manhattan Project during World War II is prohibitivelyexpensive, nuclear industries of the major industrialized countries researchedto develop economically acceptable industrial processes.

Gaseous diffusion technology was thus developed and is still in use as theworld's largest-scale industrial application using inorganic membranes. Thisprocess uses UF 6 , which is the most practical or rather the least inconvenientvolatile compound of uranium. The gas transport occurs by the Knudsendiffusion mechanism across a porous membrane with a pore diametertypically in the range 6-40 nm (Charpin and Rigny 1990). The lighter 2 3 5 U F 6molecule flows a little bit faster than the 2 3 8 U F 6 molecule. The theoreticalenrichment factor is 1.0043. In practice, the value is somewhat lower. Thisindicates that even for separating natural uranium into an enriched fractioncontaining approximately 3% 2 3 5 U and a depleted fraction containingapproximately 0.2% 2 3 8 U , over 1000 stages will be required (e.g. there are1400 stages in the Eurodif plant). Uranium enrichment plants are gigantic.

The qualitative problems involved in this development also were formid-able. UF 6 is chemically very aggressive, which limits the choice of possiblematerials. Some metals and ceramics are among the candidate materials.Typically, tubular membranes were developed, which comprised a macropo-rous support, one or several intermediate layers of decreasing thickness andpore diameter, and the separating layer. The separating layer covered theinternal surface of the tube (Charpin and Rigny 1990).

Little has been published on the work performed in the 1940s and 1950s.The first work was performed in the U.S.A. within the framework of theManhattan Project in the 1940s (Egan 1989). In France, the Commissariat a Energie Atomique (CEA) began research on such membranes in the 1950s.At least three French industrial companies developed tubular macroporoussupports for the CEA:

1. Desmarquest, a ceramics company which now is a subsidiary ofPechiney

2. Le Carbone Lorraine, a producer of carbon and graphite products (nowalso a subsidiary of Pechiney)

3. Compagnie Generale d' Electroceramique (CGEC), a ceramics com-pany, then a subsidiary of Compagnie Generale d' Electricite (CGE),which later became Ceraver, the Membrane Department of which nowbelongs to Societe des Ceramiques Techniques (SCT), a subsidiary ofAlcoa

Simultaneously, SFEC (Societe de Fabrication d' Elements Catalytiques) wascreated as a subsidiary of CEA to develop and manufacture the separating

THE DEVELOPING USE OF INORGANIC MEMBRANES 3

layer to be deposited on the macroporous support and to assemble themembrane into large modules.

In France, the first period of industrial production (late 1960s and early1970s) was aimed at making the membranes for the Pierrelatte militaryenrichment plant. Most of the tubular membrane supports were made byCGEC, whereas the layers were made by SFEC. Since a good gaseousdiffusion membrane does not wear out, it is very noteworthy that the originalmembranes are still in operation at this plant.

After the oil crisis in 1973, the need for large enrichment capacities forsupply of fuel to the nuclear power plants became obvious and severalEuropean countries (Belgium, France, Italy and Spain) decided to build thehuge Eurodif gas diffusion plant. This plant is located in France, in the Rhonevalley, a few kilometers away from the Pierrelatte plant. Simultaneously,England, West Germany and the Netherlands (the Troika) chose to jointlydevelop the centrifugation process for uranium enrichment, which does notuse membranes.

For Eurodif and for Pierrelatte, the supports were made by privateindustrial companies, the final separating layer by SFEC and the CEAdeveloped the process and had the overall technical responsibility. A handfulof companies were competing to manufacture the membrane support struc-ture. Finally, two companies proposing ceramic oxide based supports,Ceraver (the new name of CGEC) and Euroceral (a 50/50 joint venturebetween Norton and Desmarquest) each won 50% of the market. Thishappened in 1975. Within a matter of 6 years, each company had to delivermore than 2,000,000 m2 of supports which SFEC would convert into morethan 4,000,000 m2 of membranes (Charpin and Rigny 1990). Special plantswere built at a very rapid pace. These were close to Tarbes for Ceraver, closeto Montpellier for Euroceral and close to the Eurodif site for SFEC.

The enrichment capacity of Eurodif is 10,800,000 UTS (units of separationwork). This corresponds to the fuel consumption of 90 nuclear reactors of the900 M W class. In view of all the programs for building nuclear power plantshastily set up by many countries shortly after the 1973 oil crisis, it was clearthat another uranium enrichment plant of similar size would have to be builtimmediately after Eurodif was completed. This was the Coredif project.

A few years later, one had to realize that most of the ambitious plans forbuilding nuclear power plants would be strongly delayed or even abandoned.Only France stuck to its original plans and built a large number of nuclearreactors. The Coredif project did not materialize. In 1982, the membraneproduction plants of Ceraver, Euroceral and SFEC were shut down and lateron dismantled due to the lack of demand for another enrichment plant (theservice life of a gaseous diffusion membrane is several decades). For France,this was an abrupt end of the nuclear period for membranes.


In the U.S. similar developments in the area of inorganic membranes hadtaken place earlier on a somewhat larger scale and had resulted in mem-branes that are believed not to be of a ceramic nature (Charpin and Rigny1990). Several very large gaseous diffusion plants were constructed (OakRidge and later Paducah in 1953 and Porsmouth in 1954) and are operatedby the Department of Energy (DOE). These are not in operation but have atleast twice the capacity of the Eurodif plant. At least part of the membranedevelopment was made by Union Carbide, a company which had someimpact on later developments.

In the USSR also, inorganic membranes were developed and gaseousdiffusion plants were constructed to meet the needs for enriched uranium. Forunderstandable reasons, very little is known of these developments.

In today's world, it is obvious that uranium enrichment by Knudsendiffusion has no future. Uranium enrichment by using laser technology canbe accomplished much more efficiently. Such plants are expected to be readyfor industrial-scale production by the year 2000 or 2005, when, according tocurrent estimates, new uranium enrichment plants will be needed.

During the nuclear period, it was demonstrated that inorganic membranes,particularly the ceramic membranes, can be produced on a very large scalewith exceptionally high quality to meet very stringent specifications. Further,the inorganic membranes were found to be very reliable and with long servicelife even under chemically aggressive environments. As indicated here, themembranes used in the Pierelatte plant are still performing very well evenafter more than 8 years of service. Likewise, at the time of this writing, theEurodif membranes have satisfactorily completed 20 years of service and areexpected to continue their performance for many years to come.

1.3. THE DEVELOPMENT OF ULTRAFILTRATION AND MICROFILTRATIONINORGANIC MEMBRANES: THE 1980-1990 PERIOD

The development of industrial inorganic ultrafiltration (UF) and micro-filtration (MF) membranes resulted from the combination of three factors:

1. the know-how accumulated by the companies that built the nucleargaseous diffusion plants

2. the existence of ultrafiltration as an industrial process using polymericmembranes

3. the limitations of polymeric membranes in terms of temperature, pre-ssure and durability

All major industrial participants in the developments that took place in theperiod 1980-1985, were companies which actively participated in the devel-opment and manufacture of inorganic membranes for nuclear applications,


especially in the French nuclear program. The pioneering work was per-formed by two companies that were most active in this program, namelySFEC and Ceraver.

The concept of inorganic ultrafiltration or microfiltration membranes isnot new. The basic structure of these membranes is not different from that ofthe gas diffusion membranes described in a number of patents issued in theearly 1970s (Clement, Grangeon and Kayser 1973, Miszenti and Mannetti1971). The first attempt to use the high mechanical resistance of inorganicsupports probably dates back to the 1960s, when dynamic membranes madeof a mixture of zirconium hydroxide and polyacrylic acid deposited on aporous carbon or ceramic support were developed by the Oak RidgeNational Laboratory in the U.S. (Kraus and Johnson 1966, Marcinkowsky,Johnson and Kraus 1968). These nonsintered or dynamically formed mem-branes require frequent regeneration by filtering through the support asuspension of zirconium hydroxide and polyacrylic acid. The thin cake thusformed is the separation layer. Such membranes later evolved into theultrafiltration or reverse osmosis membranes made of a dynamic zirconiumhydroxide on a stainless steel support that are now marketed in the U.S. byCARRE, a subsidiary of du Pont.

This concept later evolved into the Ucarsep membrane made of a layer ofnonsintered ceramic oxide (including ZrO2) deposited on a porous carbon orceramic support, which was patented by Union Carbide in 1973 (Trulson andLitz 1973). Apparently, the prospects for a significant industrial developmentof these membranes were at the time rather limited. In 1978, Union Carbidesold to SFEC the worldwide licence for these membranes, except for anumber of applications in the textile industry in the U.S. At that time, SFECrecognized the potential of inorganic membranes, but declassification of theinorganic membrane technology it had itself developed for uranium en-richment was not possible.

The Union Carbide ultrafiltration membrane comprised a layer of un-sintered ZrO2 particles on a tubular carbon support with 6 mm innerdiameter. Using the experience acquired in the nuclear program, SFECadded the step of sintering the ZrO2 layer, thereby permanently attaching itto the support. SFEC also designed modules, filtration systems and de-veloped ultrafiltration applications. In 1980, SFEC began selling completeultrafiltration plants under the trademark of Carbosep.

Since membranes no longer had important nuclear applications in future,SFEC was sold in 1987 by the CEA to the French company Rhone-Poulencwhich merged them with their polymeric membrane division to form the newsubsidiary, currently known as Tech Sep. ZrO2-based ultrafiltration mem-branes on 6 mm inner-diameter carbon tubes continues to be the mainproduct line of Tech Sep in terms of inorganic membranes.


Ceraver's entry into the microfiltration and ultrafiltration field followed acompletely different approach. In 1980, it became apparent that the type ofproduct made by Ceraver for uranium enrichment, which was a tubularsupport and an intermediate layer with a pore diameter in the microfiltrationrange, might be declassified. Ceraver therefore developed a range of -123microfiltration membranes on an -123 support with two key features:first, the multichannel support and second, the possibility to backflush thefiltrate in order to slow down fouling.

The use of a multichannel support made of a sintered oxide carrying aseparation layer deposited on the surface of the channels was not a newconcept. This was described in the patent literature as far back as the 1960s(Manjikian 1966). The multichannel geometry is particularly attractive interms of its sturdiness, lower production cost compared to the single tube ortube-bundle geometry and lower energy requirement in the cross-flow re-circulation loop. However, Ceraver was the first company to industriallyproduce multichannel membranes. Since 1984 these membranes, which have19 channels per element with a 4 mm channel diameter are sold under thetrademark Membralox.

The second innovative feature of the Membralox membranes was thepossibility to backflush, a feature the Carbosep membranes did not offer.The principle of backflushing was not new, but it was the first time this wasdemonstrated with an industrial cross-flow membrane module. The ability tobackflush can be very beneficial because in numerous applications foulingdecreases the flux through a microfiltration membrane down to roughlysimilar range of values as those obtained with an ultrafiltration membrane.Backflushing is thus necessary to fully exploit the possibility of high fluxoffered by the relatively larger pore diameter (0.2 /xm and larger) of micro-filtration layers.

SFEC was essentially able to market their ZrO2-based ultrafiltrationmembranes to an already existing market in the sense that these membranesreplaced polymeric UF membranes in a number of applications. They alsodeveloped a certain number of new applications. For Ceraver, the situationwas different. When the Membralox membranes were first developed,microfiltration was performed exclusively with dead-end polymeric cartridgefilters. In parallel to the development of inorganic MF membranes, Ceraverinitiated the development of cross-flow MF with backflushing as a newindustrial process.

The first generation Membralox membranes were essentially developedfor MF applications. Although -123 (also described as transition A12O3)UF layers with pore diameters suitable for UF were available, their poorchemical resistance prevented their widespread use for UF applications.


Membralox UF membranes with ZrO2 layers were commercialized in 1988.These are resistant to extreme pH values and can be backflushed.

Ceraver's business approach was, however, completely different from thatof SFEC. Ceraver's strength was primarily in the manufacture of technicalceramics. Thus, Ceraver sold membranes in the form of complete modules toequipment manufacturers who developed the filtration systems including inmost cases the filtration process itself.

In 1986, CGE, which by then had its primary focus in the energy andcommunication businesses, divested its association from materials, and soldthe ceramic part of Ceraver, including the ceramic membranes division toAlcoa. Under the name SCT it is now a subsidiary of the recently formedAlcoa Separations Technology, Inc.

A few other players in the nuclear membranes activity also developedinorganic membranes for the filtration of liquids. This was the case withNorton-USA who with the know-how of Euroceral developed MF mem-branes made of an -123 tubular support with an -123 layer. The innertube diameter was 3 mm and the outer diameter 5 mm. In 1988-1989, Nortonalso produced the multichannel membrane elements. These membranesproduced by Norton are now sold by Millipore under the trademarkCeraflo.

In the early 1980s, former employees of Euroceral founded a smallcompany located near Montpellier in France known as Ceram-Filtre. Therather less well-known Ceram-Filtre membranes comprise a multichannelsupport with 19 channels of 4 mm diameter and a microfiltration membranemade of an oxide.

Another participant in the French nuclear program, Le Carbone-Lorraine,developed inorganic membranes by combining their know-how in the field ofmembranes with their expertise in carbon. They developed tubular UF andMF membranes using a tubular carbon support (inner diameter 6 mm, outerdiameter 10 mm). The carbon support is made of carbon fibers coated withand bonded by CVD carbon, the separating layers also being made of carbon.These membranes have been marketed since 1988.

The membrane research and development activities of some universitylaboratories is also a fallout of the nuclear membrane program. The in-organic membrane work performed by the University of Montpellier origin-ated in a cooperative effort with the neighboring Euroceral plant. Thiscooperative effort continued with Ceraver subsequent to the shutdown of theEuroceral plant.

A completely different type of inorganic membrane also has its origin in thenuclear industry: the asymmetric alumina membranes obtained by the anodicoxidation of an aluminum sheet were first developed for uranium enrichment


in the 1950s in France (Charpin, Plurien and Mammejac 1958) and in Sweden(Martensson et al. 1958). Such membranes are now marketed under thetrademark of Anopore by Anotec, A British subsidiary of Alcan.

1.4. THE THIRD STAGE IN THE DEVELOPMENT OF INORGANIC MEMBRANES:FURTHER DEVELOPMENTS BY OTHER ORGANIZATIONS

In the second half of the 1980s, an increasing number of companies enteredthe field of inorganic membranes, the most significant ones being ceramiccompanies such as NGK of Japan which also developed a multichannelmembrane element (19 channels, 3 mm diameter), Nippon Cement and Totoalso from Japan and very recently Corning who also developed a multichan-nel membrane structure.

An increasingly large number of university and industrial laboratorieshave also begun exploring new techniques for producing and/or utilizingexisting inorganic membranes, including metal membranes, to develop newapplications. A variety of new separating layers are under development,including porous and nonporous glasses, layers doped with catalysts, etc. Thecharacteristics of many of these are discussed in the later chapters.

1.5. CONCLUSIONS

In summary, the development of inorganic membranes was initially orientedtowards uranium enrichment which is still by very far their most significantapplication. Some of the key participants involved in the nuclear programsfurther developed them into cross-flow filtration membranes. The recentyears have seen the start of a much broader exploration of the manyfoldpotentialities of inorganic membranes, both in terms of materials and appli-cations. Thus, a multifaceted new field of technology is emerging.

REFERENCES

CEA. 1958. Porous membranes with very fine porosity and their production process. FrenchPatent 1,197,982.

Charpin, J., P. Plurien and S. Mammejac. 1958. Application of general methods of study ofporous bodies to the determination of the characteristics of barriers. Proc. 2nd United NationsIntL Conf. Peaceful Uses of Atomic Energy, 4: 380-87.

Charpin, J. and P. Rigny. 1990. Inorganic membranes for separative techniques: From uraniumisotope separation to non-nuclear fields. Proc. 1st IntL Conf. Inorganic Membranes, 3-6 July,1-16, Montpellier.

Clement, R., A. Grangcon and J. Kayser. 1973. Process for preparing filtering elements withhigh permeability. French Patent 2,527,092.

Egan, B. Z. 1989. Using inorganic membranes to separate gases: R/D status review. Oak RidgeNational Laboratory Report ORNI/TM-11345.


Kraus, K. A. and J. S. Johnson. 1966. Colloidal hydrous oxide hypcrfiltration membrane. U.S.Patent 3,413,219.

Manjikian, S. 1966. Production of semipermeable membranes directly on the surfaces ofpermeable support bodies. U.S. Patent 3,544,358.

Marcinkowsky, A. E., J. S. Johnson and K. A. Kraus. 1968. Hyperfiltration method of removingorganic solute from aqueous solution. U.S. Patent 3,537,988.

Martensson, M., K. E. Holmberg, Lofman and E. I. Eriksson. 1958. Some types of membranesfor isotope separation by gaseous diffusion. Proc. 2nd United Nations Intl. Conf. Peaceful Usesof Atomic Energy. 4: 395-404., Geneva.

Miszenti, G. S. and C. A. Mannetti. 1971. Process for preparing porous composite membranes orbarriers for gaseous diffusion systems. Italian Patent 27802A/71.

Trulson, . and L. M. Litz. 1973. Ultrafiltration apparatus and process for the treatment ofliquids. U.S. Patent 3,977,967.

Veyre, R., S. Richard, F. Pejot, A. Grangeon, J. Charpin, P. Plurien and B. Rasneur. 1977.Process for producing permeable mineral membranes. French Patent 2,550,953.

2. Synthesis of Inorganic Membranes

A. J. BURGGRAAF and K. KEIZER*University of Twente, Faculty of Chemical Technology, Enschede

2.1. INTRODUCTION AND OVERVIEW

2.1.1. General Background: Membrane Types and Structures

The aim of this introductory section is twofold. In the first place, the largevariety of different synthesis methods and techniques will be placed againstthe background of membrane types and structures. These will be brieflysummarized with focus on their relation with synthesis aspects. This willjustify a selection of two groups of methods: those which will be discussed inmore detail (Sections 2.3-2.9) and those which will only be mentioned, butnot treated extensively (Section 2.8). In the second place, a brief summary ofthe most important aspects of membranes relating to synthesis methods willbe given. This will serve as a guideline in the more detailed treatment ofparticular synthesis methods. The field of inorganic membranes has attractedmore attention in recent years and is now rapidly developing. This is reflectedin the relatively large number of patents or patent applications in the last fewyears indicating a considerable, partly hidden, industrial activity. Further-more, the number of reviews since 1987 have sharply increased, as is theattention inorganic membranes receive in an increasing number of symposiaand conferences. Many papers have a preliminary character and mostcontributions have a strongly descriptive nature. A focus will be given onthose fields and contributions where at least a certain coupling betweensynthesis and resulting microstructure has been shown.

A membrane can be described as a semipermeable barrier between twophases which prevents intimate contact. This barrier must be permselectivewhich means that it restricts the movement of molecules in it in a very specificway. The barrier can be solid, liquid or gas. Permselectivity can be obtainedby many mechanisms:

1. Size exclusion or molecular sieving2. Differences in diffusion coefficients (bulk as well as surface)3. Differences in electrical charge4. Differences in solubility5. Differences in adsorption and/or reactivity on (internal) surfaces

* With R. R. Bhave.

10

SYNTHESIS OF INORGANIC MEMBRANES 11

The flux of liquids or gases through the membrane is in most cases driven bya pressure gradient and sometimes by an electric field gradient. Membranescan be used for:

1. Separation of mixtures (liquids, gases or liquid-solid mixtures can beseparated).

2. Manipulation of chemical reactions: shifting the equilibrium situationor manipulation of the conversion or selectivity of catalytic reactionsare two possibilities.

The effectiveness of the membrane in a certain application depends on thedetailed morphology and microstructure of the membrane system, in addi-tion to the performance of the above mentioned physicochemical mech-anisms. These are critically determined by the synthesis process and this iswhy details of the preparation procedures are so important. The mostimportant and well developed of these procedures are treated in Sections2.3-2.9.

Inorganic membranes can be categorized as shown in Table 2.1. The denseinorganic membranes consist of solid layers of metals (Pd, Ag, alloys) or(oxidic) solid electrolytes which allow diffusion of hydrogen (or oxygen). Inthe case of solid electrolytes transport of ions takes place. Another categoryof dense membranes consist of a porous support in which a "liquid" is

Table 2.1. Types of Inorganic Membranes

Dense

Dynamic

Porous metal ornonmetallic

Inorganic

CompositeModified

Main Characteristics

Metal foil(Oxidic) solid electrolyteLiquid immobilized (LIM)Permanent

Nonpermanent*

Symmetric, asymmetricSupported, nonsupportedPore shape, morphology and sizeChemical nature of

pore surfaceTwo-phase particle mixturePores in matrix (partially)

filled with 2nd phasesandwich structures

Comments

Solution/diffusion of atomic orionic species

Ion exchange in hydroxide layerson a support

Permselective diffusion affected bythe pore characteristics

Distribution, important,microparticles, props

* "Nonpermanent" means separation layer is formed during the preparation process in-situ on aporous support


immobilized. The liquid fills the pores completely and is semipermeable.Interesting examples are molten salts immobilized in porous steel or ceramicsupports and semipermeable for oxygen. Sometimes this liquid can be formedin-situ during the process under consideration. This is, for example, the caseduring the use of ceramic membranes in the decomposition of H 2 S whereliquid S is condensed in the pores and blocks hydrogen diffusion. Anotherexample is the group of the so-called dynamic membranes where a hydroxide(gel) is precipitated in or on a porous support.

The porous membranes consist of a porous metal or ceramic support withporous top layers which can have different morphologies and microstruc-tures. Their essential structural features are presented in Figures 2.1 and 2.2and are discussed later (Section 2.2).

Figure 2.1. Schematic representation of main types of pore structures and membranes. A and B:homogeneous unsupported; straight pores C: supported asymmetric, interconnected pores D: aphotograph of a membrane of the type (c). (SCT-support+Y-Al2O3 top layer UT Twente)


From Figure 2.1 it can be seen that there are three types of pore systems:straight pores running from one side of the membrane to the other with aconstant pore diameter (Figure 2.1a) or conical pores (Figure 2.1b). Thesetypes of systems can be correlated with track-etch and anodic oxidationprocesses, respectively. The system shown in Figure 2.1c consists of apercolation system of pores with more or less regular shapes or with a spongystructure. This is correlated with packing of particles and phase separation,respectively. In composite and modified membranes the top layer (with veryfine pores) shown in Figure 2.2 is modified further as schematically re-presented in Figure 2.3 or consists of an intimate mixture of two phases. Themodification technologies in most cases consist of precipitation of a phasefrom liquids or gases followed by further treatments (see for further detailsSection 2.9) and result in a decrease of the pore size or in a change in thechemical character of the internal pore surface. The obtainable pore sizes areschematically represented in Figure 2.4 together with the most importantrelated fields of application. Principles to obtain the required mean pore sizeand narrow pore size distribution are summarized below.

1. porous support (1-15 /im pores)2. intermediate layers) (100-1500 nm)3. separation (top) layer (3-100 nm)4. modification of separation layer

1+2 is microfiltrationor 'primary* membrane

1+2 + 3 is ultrafiltrationor 'secondary* membrane

1+2 + 3 + 4 is hyperfiltration or/andgas separation membrane

Figure 2.2. Schematic representation of an asymmetric-composite membrane (Keizer and Burg-graaf 1988).


a Homogenous (multi) layers in the poresb Plugs in the pores (constrictions) Plugs/layers on top of the pores

Rgure 2. Schematic representation of the microstructure of modified membrane top layers(Keizer and Burggraaf 1988).

Application fields

applicationfield

movingparticle

gas(vapor)separationreverse osmosishyperfiltration

ultrafiltr.microfiltration

0.1 1 10 100 100Cions macromol. microscopic

atoms colloidal particlesmolecules;particles

Figure 2.4. Pore size range of ceramic membranes and related application fields.

2.2. BASIC PRINCIPLES OF MEMBRANE SYNTHESIS

In this section a short introduction will be given on the synthesis of porousceramic membranes by sol-gel techniques and anodization, carbon mem-branes, glass membranes and track-etch membranes. An extensive discussionwill be given in Sections 2.3-2.8.

Terms such as symmetric and asymmetric, as well as microporous, meso-porous and macroporous materials will be introduced. Symmetric mem-branes are systems with a homogeneous structure throughout the membrane.Examples can be found in capillary glass membranes or anodized aluminamembranes. Asymmetric membranes have a gradual change in structurethroughout the membrane. In most cases these are composite membranes


consisting of several layers with a gradual decrease in pore size to the feedside of the membrane. Examples are ceramic aluminas synthesized by thesol-gel technique or carbon/zirconia membranes.

Pore diameters larger than 50 nm are called macropores, mesopores have adiameter between 2 and 50 nm and below a diameter of 2 nm the system iscalled microporous (Sing et al. 1985).

2.2.1. Ceramic Asymmetric Membranes

The asymmetric membrane system shown in Figure 2.2 consists of a poroussupport a few millimeters in thickness, with pores in the range 1-10 /an, aporous intermediate layer of 10-100 /zm thickness, with pores of 50-500 nm,and a top layer (the proper separation layer, e.g. for ultrafiltration) with athickness of 1 /zm (or smaller)-10/xm with pores of 2-50 nm. The inter-mediate layer must prevent the penetration of the precursor of the top layermaterial into the pores of the support during the synthesis and the collapse ofthe thin finished top layer into the large pores of the support. Furthermore, ithelps to regulate the pressure drop across the top layer of the membrane in

, operation. In a number of cases it can be dispensed with (e.g. in carbon-supported zirconia membranes manufactured by SFEC). As shown in Figure2.3 the pore system can be further modified in different ways. In all cases thetop layer must be defect-free (no cracks or pinholes) and have preferably anarrow pore size distribution. This sets severe demands on the quality of theintermediate layer and of the support. It may also require development ofspecial technologies to overcome inferior qualities of the support system. Byfar the most frequently used principle to meet these requirements is theformation of a layer consisting of a packing of well-ordered, uniform-sizedparticles. The size and shape of the particles determine the minimumobtainable mean size and pore size. These parameters as well as the porositycan be changed by further heat treatment. The main process for makingceramic membranes is to first prepare a dispersion of fine particles (called slip)and then to deposit the particles contained in the slip on a porous support bya slip-casting method. The capillary pressure drop created on letting the slipcome into contact with the microporous support forces the dispersionmedium of the slip to flow into the pores of the support. The slip particles areconcentrated at the pore entrance to form a layer of particles or a gel layer.When relatively large particles are used to make the support or intermediatelayer (with pore sizes > 1 /an) the particles can be precipitated from super-saturated solutions. These need to be calcined and classified by sedimenta-tion, centrifuging or sieving. A slurry of these particles can be used afterstabilization.


For making the top layers with very small pores, colloidal suspensions areused. Here we need nanometer-sized particles which are stabilized in a liquiddispersion medium by colloidal or other physicochemical methods. Theseparticles and colloidal suspensions are obtained mainly by the so-calledsol-gel methods. The essential feature of this technique is the controlledhydrolysis of an organometallic compound or salt and its subsequent peptiz-ation. Very critical is the drying step in the process because here large(capillary) forces can cause cracking of the layer very easily. This can becontrolled by particle shape, agglomeration control, addition of binders,roughness of the support, etc. The above-mentioned method will be discussedfurther in Section 2.3. A summary of presently obtainable combinations ofsupports and unmodified top layers is given in Table 2.2. The minimum sizeof pores obtainable in this way is about 2.5 nm. This number is related to thesmallest sol primary particle which can be obtained. The size of this primaryparticle is determined by seeding and crystallization parameters such as thesurface tension (TS) and the free-enthalpy difference (AG) between dissolvedand crystallized material. Stable nuclei smaller than 5 nm are hardly ob-tained. Thermal treatments always increase the pore size due to a decrease inthe free-enthalpy G.

The pore shape is determined by the particle shape. Plate-shaped particleslead to plate-shaped pores in the case of regular packing. Sphere-shapedparticles favor cylindrical or sometimes ink-bottle-type pores.

Table 2.2. Asymmetric Composite Membranes: Combinations of Substratesand Top Layers

Top Layer

SubstrateAlumina

ZirconiaTitaniaSilicaCarbonSiCSintered:SteelNickel

121

Ind.*

Lab.

ZrO22

Ind.

Lab.

Ind.

TiO23

Lab.'

Lab.

Lab.Lab.

SiO24

Lab.

5

Lab.

SiC6

Lab.

BinaryLayers

1 + 2/1 + 3

ModifiedLayers

Ag,MgOV2O3, SiO2

Ag,V 2O5

* Ind.: Industry* Lab.: Laboratory


To reduce further the pore size and/or to introduce specific interactionsbetween the solid surface and the liquid or gaseous medium in the pores,sol-gel layers need to be modified. In principle this is done by precipitation orby adsorption of components from a gaseous or liquid medium followed byheat treatment of the formed products inside the pores or the pore entrance.This will be further discussed in Section 2.7.

2.2.2. Glass Membranes

Glass membranes with an isotropic spongy structure of interconnected porescan be prepared by thermally demixing a homogeneous Na2O-B2O3-SiO2glass phase in two phases. The Na2O-B2O3-rich phase is then acid-leachedthereby creating a microporous SiO2-rich phase (Hsieh 1988). Some porousmetal membranes have been made in a similar way with a strong acid orother types of leachant (Hsieh 1988). The remaining silica glass structure isnot very chemically resistant. This has been overcome partially by surfacetreatment (e.g. by means of chemical agents) of the internal pore structurewhich make the surface hydrophobic (Schnabel and Vaulont 1978). Theadvantage of glass membranes is that capillaries (hollow fibers) can be easilyformed, and can be further modified as described above to porous hollow-fiber membranes.

2.2.3. Anodic Membranes

Pores with a linear form as shown in Figure 2.1 are produced by the so-calledanodic oxidation process (Smith 1974). Here one side of a thin high-purityaluminum foil is anodically oxidized in an acid electrolyte. A regular patternof pores is formed. The pore size is determined by the voltage used and by thetype of acid, the pore shape being always conical. The process must bestopped before the foil is oxidized completely and to avoid closure of pores.The unaffected part of the metal foil is subsequently etched away with astrong acid. The resultant structure has distinctive conical pores perpendicu-lar to the macroscopic surface of the membrane. The membranes so obtainedare not stable under long exposure to water. The stability can be improved bytreatment in hot water or in a base. Such a treatment can also be used todecrease the pore size on one side of the membrane, and as a consequence anasymmetric membrane can be produced. The disadvantage of the method isthat only unsupported membranes can be produced in the form of membranefoils. To get sufficient mechanical stability they must be supported in someway for most applications.


2.2.4. Track-Etch Membranes

Pores with a very regular, linear shape can be produced by the track-etchmethod (Quinn et al. 1972). Here a thin layer of a material is bombarded withhighly energetic particles from a radioactive source. The track left behind inthe material is much more sensitive to an etchant in the direction of the trackaxis than perpendicular to it. So etching the material results in straight poresof uniform shape and size with pore diameters ranging between 6 nm and1200 nm. To avoid overlap of pores only 2-5% of the surface can be occupiedby the pores. This process has been applied on polymers (e.g. Nucleporemembranes) and on some inorganic systems like mica. Membranes soobtained are attractive as model systems for fundamental studies.

2.2.5. Pyrolysis

Membranes with extremely small pores ( < 2.5 nm diameter) can be made bypyrolysis of polymeric precursors or by modification methods listed above.Molecular sieve carbon or silica membranes with pore diameters of 1 nmhave been made by controlled pyrolysis of certain thermoset polymers (e.g.Koresh, Jacob and Soffer 1983) or silicone rubbers (Lee and Khang 1986),respectively. There is, however, very little information in the publishedliterature. Molecular sieve dimensions can also be obtained by modifying thepore system of an already formed membrane structure. It has been claimedthat zeolitic membranes can be prepared by reaction of alumina membraneswith silica and alkali followed by hydrothermal treatment (Suzuki 1987).Very small pores are also obtained by hydrolysis of organometallic siliciumcompounds in alumina membranes followed by heat treatment (Uhlhorn,Keizer and Burggraaf 1989). Finally, oxides or metals can be precipitated oradsorbed from solutions or by gas phase deposition within the pores of analready formed membrane to modify the chemical nature of the membrane orto decrease the effective pore size. In the last case a high concentration of theprecipitated material in the pore system is necessary. The above-mentionedmethods have been reported very recently (1987-1989) and the results are notyet substantiated very well.

2.2.6. Dense Membranes

A second class of membranes are described as dense membranes. They mayconsist of thin plates of metals (Pd and its alloys, Ag and some alloys) oroxides (stabilized zirconia or bismuth oxides, cerates). These membranes arepermeable to atomic (for metals) or ionic (for oxides) forms of hydrogen oroxygen and have been studied, especially, in conjunction with chemical


reactions like (oxidative) dehydrogenation, partial oxidation etc. in mem-brane reactors. Their main drawback is the low permeability. This might beimproved by making very thin layers (micrometer to nanometer range), e.g.by deposition in a pore system. A second form of "dense" membranes are theso-called liquid-immobilized membranes (LIM). Here the pores of a mem-brane are completely filled with a liquid which is permselective for certaincompounds. In the polymer field this principle has been already investigatedextensively. In the inorganic membrane field research efforts have just begun.With molten salts incorporated in a porous matrix, one can obtain permea-bilities for oxygen or ammonia comparable with those of porous materials(Pez 1986, Dunbobbin and Brown 1987). Important parameters are thewettability of the matrix by the liquid and the morphology of the pore systembecause these determine the degree to which the liquid is captured (immobil-ized) within the membrane system.

23 . PACKING OF PARTICLES FROM SUSPENSIONS

2.3.1. Introduction and Support Systemsi

Membranes produced by packing of particles from dispersions have thegeneral structure as discussed in Section 2.2 and given in Figure 2.2. The thintop layer with (very) small pores is applied on top of a support system whichconsists of one or two much thicker layers with (much) larger pores. The mainsupport consists of a packing of rather coarse-grained material (micronrange) which is produced in a classical way by cold isostatic pressing of a dry.powder, by co-extrusion of a paste of ceramic powder with additions of

; binders and plasticizers or by slip-casting (Messing, Fuller and Housner1988). After burning away the organic material the so-called "green" compactis sintered. In order to obtain defect-free membranes, thin top layers on the

psupport system must fulfill more stringent requirements than those utilized in||the manufacture of commercially available porous tube materials. Pore size

distribution and roughness must be smaller than usual. The ways to obtain"" &se characteristics are largely classified with practically no published

formation. Some aspects have been discussed by Vuren et al. (1987) andpTcrpstra, Bonekamp and Veringa (1988). It is important to obtain a narrow

distribution. Therefore, suspensions or pastes are prepared from ahaving a narrow particle size distribution. This implies a very good

*ntrol of the agglomeration state of the material by deagglomerationttments (e.g. milling, ultrasonification) and/or removal of the fraction withlargest diameters e.g. by sedimentation. Organic surfactants are some-

added to counteract flocculation of the deagglomerated suspension.


The quality of the support is especially critical if the formation of the toplayer is mainly determined by capillary action on the support (see Section2.3.2). Then, besides a narrow pore size distribution the wettability of thesupport system plays a role (see Equation 2.1). An example of the synthesis ofa two-layer support and ultrafiltration membrane is given in the FrenchPatent 2,463,636 (Auriol and Tritten 1973). In many cases an intermediatelayer, whose pore sizes and thickness lie between those of the main supportand the top layer (see Figure 2.2), is used. This intermediate layer can be usedto improve the quality of the support system. If large capillary pressures areused to form such an intermediate layer, defects (pinholes) in the support willbe "transferred" to this layer. This can be avoided by decreasing the actingcapillary pressures or even by eliminating them. This can be done in severalways.

If a film coating technique is used, the viscosity of the system is increased tosuch an extent that none or hardly any penetration of the "liquid" in the poresystem occurs. The film thickness obtained is governed by the surface tensionand the viscosity of the suspension (Deryaquin and Levi 1964). The sameresult is obtained with co-extrusion of material of the main support and of theintermediate layer. The capillary driving force for extracting liquid from thecoating suspension into the pores of the support can be decreased by fillingthe pores of the support material with a liquid, with a small difference insurface tension with respect to that of the suspension from which theintermediate layer (or top layer) is formed (Tallmadge and Gutfinger 1967).

In another method the pores of the support are rendered nonwettable forthe coating suspension liquid. This can be obtained by forming a hydro-phobic pore surface on the support in the case of an aqueous coatingsuspension. The support surface is treated, e.g. by organic silanes which reactwith surface hydroxyl groups as described by Messing (1978). Such a surfacehas at the same time improved properties for separation of a mixture of polaror nonpolar gases or hydrophobic and hydrophilic liquids. The disadvantageof a decreased capillary action in the formation of a layer is the diminishedadherence of the support and that layer. The adherence as well as eliminationof defect formation can be improved by a smaller roughness of the supportmaterial. Gillot (1987) reports a roughness which should be less than 10% ofthe mean particle diameter of the grain size of the support material. Thisimplies again that the particle size distribution of the support should benarrow to obtain a locally well-defined roughness. Gillot uses this principle tomake an improved three-layer membrane system, with particles of 0.55 inthe top layer and using polyvinylalcohol to control the viscosity and asurfactant Darvan to avoid flocculation. The mean pore size of theresulting top layer is 0.26 /an and a small roughness is reported. To obtainsmaller pore sizes as indicated above it is necessary to use (ultra) fine-grained


powders and suspensions in the synthesis of the top layer. This is obviousbecause in well-packed systems of uniform particles the mean pore radius isminimum (about 0.4-0.7 of the mean particle radius), depending on thepacking structure. This means that (colloidal) suspensions with particlediameters ranging from 5 nm-100 nm are needed. These can be obtained indifferent ways described in recent symposia and congress proceedings (Henchand Ulrich 1984, 1986, Brinker, Clark and Ulrich 1984, 1988). The mostcommonly used route for membrane synthesis with packing methods is theslip-casting process using suspensions obtained by sol-gel processes. In theremainder of this chapter attention will be focussed on this combination.

2.3.2. Sol-Gel Process

The sol-gel process can be divided into two main routes which are schemati-cally shown in Figure 2.5. These may be distinguished as the colloidalsuspension route and the polymeric gel route. In both cases a precursor ishydrolyzed while simultaneously a condensation or polymerization reactionoccurs. The essential parameter to control is the hydrolysis rate with respectto the polycondensation rate. The precursor is either an inorganic salt or ametal organic compound. The chemistry of the initial stages has beendescribed by Livage (1986). In the colloidal route a faster hydrolysis rate isobtained by using a precursor with a fast hydrolysis rate and by reacting theprecursor with excess water. A precipitate of gelatinous hydroxide or hydra-ted oxide particles is formed which is peptized in a subsequent step to a stablecolloidal suspension. The elementary particle size ranges, depending on thesystem and processing conditions, from 3-15 nm tfnd these particles formloosely bound agglomerates with sizes ranging from 5-1000 nm. The size ofthe agglomerates can be decreased, e.g. by ultrasonification of the suspensionand by manipulation of the electrical charge on the particles. By increasingthe concentration of the suspension and/or by manipulation of the surface(zeta) potential of the sol particles the colloidal suspension is transformed to agel structure consisting of interlinked chains of particles or agglomerates(Figure 2.5). As discussed by Partlow and Yoldas (1981) the packing densityat the time of gelation (i.e. the gelling volume) can vary from a rather loose toa dense form depending on the charge on the particles. This means that thepH and the nature of the electrolyte (or anion in the peptizing acid) has animportant effect on the gelling point and volume, because they determine themutual repulsion force which is necessary to obtain a stable colloidalsuspension. The anion chosen for the electrolyte or peptizing acid must notform a complex with the metal ion of the membrane to be formed. Ininstances where the initial particle has a charge opposite to that of theelectrolyte, the gelling volume exhibits a minimum with increasing electrolyte


Colloidalgel route

SOL:

\

/

. ' ' '

' tColloidal

COLLOIDAL GEL

Metal salt or alkoxide

/ precursorf

particles liquid

DRYING andSINTERING

4

1

*~*

Polymericgel route

SOLUTION:

> / : " : :

inorganicpolymer molecule

POLYMERIC GEL

\Powder, f iber, coating, membrane, monolith

Figure 2S, Scheme of sol-gel routes. Colloidal sol-gel route and polymeric gel route (Burggraaf,Keizer and van Hassel (1989a, b).

concentration. This is, for example, the case for alumina sols. It seems thatgels with their minimum volume are better suited to obtain monolithicstructures (Partlow and Yoldas 1981, Yoldas 1975) and membranes (Leen-aars, Keizer and Burggraaf 1984, 1987, Leenaars and Burggraaf 1985). Thismeans that pH, counter-ion type and concentration must be chosen in such away that the particle is just far enough from its point of zero charge, i.e.isoelectric point (IEP), to prevent flocculation. With conditions too close tothe IEP a poorly dcnsified film will be obtained. In this way orderedaggregates of elementary particles can be obtained. In the further processingthis is the best starting situation to give defect-free membranes (or monoliths)with a narrow pore size distribution.

The hydrolysis and polymerization rate of metal organic compounds cangenerally be better controlled than those of metal salts. The chemical reactioninvolves two steps (Livage 1986):


1. The partial hydrolysis of the metal organic compound (e.g. a metalalkoxide) introduces the active functional OH groups, attached to metalatoms.

2. These then react with each other or with other reactants to form apolymeric solution which further polymerizes to form a viscous solutionof organic-inorganic polymeric molecules.

In the polymeric gel route the hydrolysis rate is kept low by addingsuccessively small amounts of water and by choosing a precursor whichhydrolyzes relatively slowly. The final stage of this process is a stronglyinterlinked gel network (Figure 2.5) with a structure different from thatobtained from the colloidal route. This can be seen from the fact that thenetwork formation takes place continuously within the liquid. The gel willform and shrink even within the liquid. It is not necessary to remove thisliquid to obtain a gel as in the colloidal route. This means that concentrationsof solid material in polymeric gels are usually smaller than in particle gels.The water necessary for the reaction can be supplied in different ways: (1)slowly adding a water or water/alcohol solution to an alcoholic solution ofthe alkoxide, (2) in-situ production of H2O through an esterification reactionby adding an organic acid to the alkoxide solution, (3) dissolving an alkalinebase or (4) an hydrated salt into the alkoxide solution in alcohol. The localwater concentration can be manipulated in this way thereby strongly influ-encing the gel volume at the gelling point. With method (4) even complexcompounds (BaTiO3) can be obtained (Livage 1986). Finally, the gelationprocess can be significantly changed by the nature of catalysis of thepolycondensation/polymerization reaction (Her 1979). A model which pre-dicts some aspects of the inorganic polymerization reactions is given byLivage and Henry (1985). Silica systems can be controlled very well and bothcolloidal and polymeric gel routes can be realized. Alumina has a strongpreference to follow the colloidal route, while titania systems behave inter-mediately.

2.3.3. The Slip-Casting of Ceramic Membranes

A common method to slip-cast ceramic membranes is to start with a colloidalsuspension or polymeric solution as described in the previous section. This iscalled a "slip". The porous support system is dipped in the slip and thedispersion medium (in most cases water or alcohol-water mixtures) is forcedinto the pores of the support by a pressure drop (APJ created by capillaryaction of the microporous support. At the interface the solid particles areretained and concentrated at the entrance of pores to form a gel layer as in thecase of sol-gel processes. It is important that formation of the gel layer starts


immediately and that the solid particles do not penetrate the pores of thesupport system. This means that the solid concentration in the slip must notbe too low, the slip must be close to its gelling state, the particle (oragglomerate) size must not be too small compared with the pore size of thesupport, unless agglomerates are formed in the pore entrance immediately atthe start of the process. Some variables such as solid concentration andparticle diameter have been given by Cot, Guizard and Larbot (1988) whohave demonstrated that the characteristics of the membrane and its forma-tion can be influenced by the agglomerate state of the slip.

The smaller and more uniform the primary particles, and the weaker theagglomerates in the sol are, the smaller the pore size and the sharper itsdistribution in the membrane will be. The thickness of the layer Lg increaseslinearly with the square root of the dipping time. The process is quantitativelydescribed by Leenaars and Burggraaf (1985). The rate of membrane deposi-tion increases with the slip concentration or with decreasing pore size of thesupport as shown below. This has been experimentally confirmed for aluminaand titania (Leenaars and Burggraaf 1985, Uhlhorn et al. 1989).

The capillary pressure drop, APC, caused by pores with an effective radius rfor each capillary is given by

APc = (2t/r)cos0 (2.1)

where t = surface tension, = contact angle between liquid and support, and

where Lg is the permeability constant of the gel layer, tj the viscosity of the slip"liquid", Kg, a constant related to the reciprocal of solid concentration andAPg the pressure drop across the gel layer. APg can be eliminated fromEquation 2.2 and expressed in terms of the above mentioned parameters.

Tiller and Chum-Dar Tsai (1986) discuss the theory of slip-casting ingeneral and they show that there is an optimum pore diameter for producinga maximum pressure drop across the formed cake (this is the gel layer in thecase of membrane formation) to give a maximum rate of cake formation. Theresults of both groups of investigators show the importance of support poresize and structure, and of the effects of the gel layer structure which isincorporated in the value of Kg. The value of Kg can be expressed in terms ofstructural constants of the membrane if the structure is known or if a model isassumed (Leenaars 1984,1985). After the gel layer is formed it is dried. This isa very critical process step because large capillary forces are set up during theremoval of the liquid. A xerogel layer is formed and large stresses due toshrinkage along the depth of the membrane occur which have to be released


in some way. If a critical stress is exceeded cracks are formed in supportedmembranes. This occurs at a critical thickness (1-10 ) of the membranewhich strongly depends on the forming conditions and on the morphology ofthe material (plates or spheres) and the support. The effect of the support isdemonstrated by the fact that nonsupported membranes can be produced inthicknesses up to 100 /im under room temperature and standard humidityconditions (about 60%). The compaction stresses can be used to order thecompact structure during the relaxation process of these stresses and thiscontributes to a narrow pore size distribution (Leenaars 1984, 1985, vanPraag et al. 1990). From the work of the group led by Burggraaf and Keizer, itemerges that the production of defect-free membranes is easier for plate-shaped particles (alumina) than for spheres (titania) (van Praag et al. 1990).

From Equation 2.2, it can be seen that the viscosity of the slip plays animportant role. It regulates the formation rate of the gel layer and helps toprevent the slip from penetrating the support pore system. In the colloidalsuspension route the evolution of the viscosity during the solvent extractionis slow during the very first steps of the process and drastically increases justbefore gelling. With the polymeric gel route a more gradual increase of theviscosity is observed. In both cases the evolution of the viscosity can bemodified by the addition of binders to the sol "slip". Different kind of bindersare chosen depending on the nature of the solvent, the compatibility with theprecursors and the viscosity of the system.

Finally, binders or plasticizers can play an important role in the preventionof cracks in the layer. As shown by van Praag et al. (1990), titania-supportedmembranes can be formed with plasticizers on normal supports, whilewithout plasticizers these can be obtained only on special supports with verysmall roughness. Frequently used binders/plasticizers include polyvinylalco-hols, cellulosic compounds and polyglycols in an aqueous medium andpolyvinylbutyral in an alcoholic medium. It is important that the organicmaterial can be completely pyrolyzed at relatively low temperature withoutleaving carbon or metal residues. After drying, the xerogel is first calcined toform an oxide structure. Further heat treatment strongly affects the final poresize of the membrane (Burggraaf, Keizer and van Hassel 1989, Leenaars et al.1984,1985, Larbot et al. 1988). Temperature and time also strongly influencethe phase compositions (e.g. alumina, titania and zirconia membranes). Atphase transitions (y-0-a alumina or anatase-rutile TiO2) there is a strongincrease in pore size (Larbot et al. 1987, Keizer and Burggraaf 1988).

It is obvious that the pore diameter can be regulated by heat treatment tovalues as small as 3-6 nm (minimum) and up to 50-200 nm depending on thematerial.

Although inorganic (ceramic) membranes offer many advantages they dosuffer from a few limitations at the present state of technology development


Table 2.3. Advantages and Disadvantages of Inorganic (Ceramic) Membranes

Advantages1. High temperature stability2. Mechanical stability under large pressure gradients (noncompressible, no creep)3. Chemical stability (especially in organic solvents)4. No ageing, long lifetime5. Rigorous cleaning operation allowable (steam sterilization, high backflush capability)6. (Electro) catalytic and electrochemical activity easily realizable7. High throughput volume and diminished fouling8. Good control of pore dimension and pore size distribution

Disadvantages1. Brittle character needs special configurations and supporting systems2. Relatively high capital installation costs3. Relatively high modification costs in case of defects4. Sealing technology for high-temperature applications may be complicated

(see Table 2.3). Some typical synthesis methods and results for differentceramic materials will be discussed in the next section.

2.4. TYPICAL RESULTS FOR DIFFERENT MATERIALS

In the discussions to follow two types of membranes must be distinguished:(1) nonsupported and (2) supported ones. Nonsupported membranes areproduced by pouring a slip onto a very smooth, dense substrate on whichgelling takes place by slow evaporation of the dispersion liquid. In this wayrather thick, crack-free films can be obtained. They are especially suitable forcharacterization purposes and structural investigations. These are evaluatedto determine whether or not the structures obtained are similar or compar-able with those of the supported ones, made from the same slip and the samematerial. The next section will also focus on the supported membranes.

2.4.1. Alumina Membranes

The mode of synthesis of alumina membranes through the colloidal suspen-sion route is given in Figure 2.6. The first step involves the preparation of aslip consisting of boehmite particles. These are plate-shaped in the form of"pennies" with a diameter of 25-50 nm and a thickness of 3.5-5.5 nm(Leenaars et al. 1984,1985). The synthesis chemistry of the colloidal boehmite(y-) solution is described in detail by Leenaars and Yoldas (1975) andto some extent by Anderson, Gieselman and Xu (1988) and by Larbot et al.(1987).

M(OR)r +R'OH: M - Al(III ) ,Tl(IV),Zr(IV)

x CHjCOOH (x

-n

z

S

33>

i

1


The process starts with the controlled hydrolysis of aluminum tri-secbutoxide (ATSB) or its dilute solution in 2-butanol in excess water at80-90C At lower temperatures other aluminum compounds are formedwhich are not easily transformed into colloidal solutions and gels. Afterremoving the alcohol from the mixture the precipitate is peptized with anappropriate amount of acid at 90-100C at a pH value lower than 1.1 (Larbotet al. 1987), and in our studies at a pH of about 3.5. Supported gel layers arenow formed by letting the colloidal solution come in contact with the poroussupport (slip-casting process). Whether or not a gel layer will be formed andwhat the casting rate (increase of thickness with time) will be, depend in acomplex way on a large number of parameters. Some of the most importantones are investigated by Leenaars et al. (1984, 1985) and are summarizedbelow

1. At a given pore size of the support a certain minimum concentration isnecessary to obtain a gel layer, otherwise "pore clogging" will occur.With an increase in the sol concentration the casting rate increases.Typical concentrations used are from 0.7-1.2 mol A1OOH/L (see (5)below).

2. In the peptizing acid series (boehmite sols) the gelling concentrationsincrease according to the order HC1 > HNO3 > HC1O4 (gelling vol-ume decreases). This implies an increasingly dense gel and a decreasingcasting rate in the same order as in this series. This has been experi-mentally verified.

3. With an increasing quantity of acid used per mole A1OOH the meanpore size of the membranes after calcination is slightly decreased.Probably the stacking density of the particles in the gel increases withincreasing concentration. Typical amounts of acid used range from0.05-0.1 mol/L. Anderson's work seems to confirm this type of depend-ence (Anderson et al. 1988).

4. The gel layer thickness increases linearly with the square root of dippingtime indicating that indeed a slip-casting process is operative. The rateconstant depends on gel structure and pore size of the support. If themodal pore size of the support is increased from 0.12/xm (type 1support) to 0.34 /an (type 2 support) the casting rate is decreased inaccordance with theory. Typical casting rates for type 1 and type 2supports are 4.4/xm/s1/2 and 2.8/xm/s1/2, respectively for HNO3-stabilized sols with a concentration of 1.22 mol boehmite/L.

5. Ageing of the sol profoundly affects the casting behavior. After ageing(e.g. one week) gel layers could be formed on type 2 supports whilebefore ageing this was not found to be the case. This points to anincrease of the agglomerate size of the boehmite particles in accordance


with light scattering experiments of Ramsay, Daish and Wright (1978).Experiments with different acids suggest an increasing agglomerate sizein the peptizing acid sequence HC1 > HNO3 > HC1O4.

In all cases the number of "pinholes" and of other types of casting defects iscritically dependent on the quality of the support. Even in cases where thesame nominal support material is used (but from different batches) varyingresults are obtained. This sensitivity of support quality could be diminishedby adding an organic additive. In our experiments we used polyvinylalcohol,PVA, with a molecular weight of 72,000 and of the type giving a very lowresidue of ash or tar on pyrolysis. A typical standard "solution" contains0.6 mol A1OOH/L (peptized with 0.07 mole HNO3 per mole A1OOH) withabout 25-30 wt.% PVA based on dry A12O3 (or 20 wt.% based on A1OOH).

As discussed by Larbot et al. (1987) and Cot, Guizard and Larbot (1988)the addition of organic "binders" has a dual function. First it regulates(increases) the viscosity of the solution and prevents the solution from beingsucked into the pores of the support before gelation starts. It promotes amore gradual increase in the viscosity during the gelation process and thusmakes the (initial) steps of the slip-casting process less critical. This, however,changes the rate constant of this process. A second function is the diminishedtendency for crack formation during the drying (and subsequent calcination)steps. The wet gel layers must be dried under carefully controlled conditionswhich are typically 3 h at 40C in an atmosphere of 60% relative humidity(Burggraaf, Keizer and van Hassel 1989) or 48 h at 40C (Larbot et al. 1987).This step is very important, because inadequate drying results in crackedlayers. This can be understood from the very large capillary and shrinkagestresses formed within the layer. (Scherer 1986, 1987). Imbalances in thesestresses must undergo relaxation to some extent to prevent crack formation.At the same time these stresses are responsible for the ordering of the initiallyrandomly packed, particles or agglomerates. In the case of plate-shapedparticles, as with y-AlOOH, the final result is represented in Figure 2.7. It canbe easily seen that in this case the pores are slit-shaped. The critical diameterof the pore is related to the thickness of the boehmite plates (of about 3 nm)with a narrow pore size distribution. Experimental confirmation of orderingduring drying has been given by Leenaars and Burggraaf (1985) and is basedon the following main observations on nonsupported membranes:

1. A strong texture can be deduced from the XRD spectrum. The intensityof the [020] reflection as compared with those of the [120] and [031]reflections in boehmite is 10 times larger than can be expected for arandom distribution of crystal directions and is in accordance with themodel shown in Figure 2.7.


Figure 2.7. Idealized model of the boehmite membrane structure, d is the distance between the 2boehmite crystals A and , is the thickness of the boehmite plates (Leenaars and Burggraaf1985).

2. Drying without the occurrence of large capillary stresses was obtainedwith supercritical drying in an autoclave. In this case a mean pore sizewas obtained which was twice that obtained under normal dryingconditions and with a broad pore size distribution in accordance withthe expectation for a noncompressed, random packing of particles.

There exist a maximum allowable thickness of the supported gel layersabove which it is not possible to obtain crack-free membranes after calcina-tion. For y-alumina membranes this thickness depends on a number of(partly unknown) parameters and has a value between 5 and 10 //m. One ofthe important parameters is certainly the roughness and porosity of thesupport system, because unsupported membranes (cast on teflon) are ob-tained crack-free up to 100 . The xerogel obtained after drying wascalcined over a wide range of temperatures. At 390C the transition ofboehmite to -123 takes place in accordance with the overall reaction

2-1 = -123 + H2OT (2.3)This transition produces an isomorphous phase and the resulting y-

alumina has the same morphology and texture as its boehmite precursor.With increasing temperature and time the mean pore diameter increasesgradually and other phases appear (8-, 9-alumina). Due to the broad XRDlines, the distinction between y- and 8-alumina cannot be made; 9-aluminaoccurs at about 900C while the conversion to the chemically very stable a-alumina phase takes place at T > 1000C. Some typical results for aluminamembranes synthesized without binders are given in Table 2.4. When PVAwas used as a binder, thermogravimetric analysis showed that, provided theappropriate binder type was used, the binder could be effectively removed atT > 400C. The ash residue is of the order of 0.01 wt.%. Mean pore size and


Table 2.4. Microstructural Characteristics of Alumina Membranes as a Functionof Calcination Time and Temperature (Leenaars et al. 1984, Burggraaf et al. 1989,

Uhlhorn et al. 1989b)

Temperature(C)

200400

500700

800900

1000550

Time(h)

3434

170850

345

120930

34343434

Phase

Y-A1OOHY-A12O3Y-A12O3Y-A12O3Y-A12O3Y/6-Al2O3Y/6-Al2O3Y/5-Al2O3Y/5-Al2O39-Al2O3-123Y-A12O3

BET Surface Area(m2/9)

3153012762492402071591491549915

147

Pore Size(nm)

2.5*2.72.93.13.23.23.84.34.85.478*6.1*

Porosity(%)

415353535451515155484159

* All pore sizes are according to the slit-shaped modelT Cylinder-shaped pore model (diameter)

, * Prepared from a sol treated in an autoclave at 200C

pore size distribution did not change on addition of the binder, the porosityhowever increased somewhat.

It can be observed from Table 2.4 that the transition to a-alumina isaccompanied by a large increase in pore diameter. Results of Larbot et al.(1987) show the same trend but larger pore diameters with more pronouncedincrease with temperature was observed (e.g. 10 nm at 900C). The rate ofincrease of pore diameter with temperature at the y-9-cx transitions was,however, smaller (e.g. 25 nm at 1100C for a-alumina membrane).

Thus, thermally stable alumina membranes can be produced with a porediameter as low as 3 nm. For long-term thermal stability the temperature ofheat treatment should be 50-100C higher than the applied temperature.Typical pore size distributions, e.g. for y-alumina membrane, are givenin Figure 2.8. If the desorption branch of adsorption/desorption isotherms isused this yields the smallest passage in a packed array. This is, however,exactly what is known as "effective" pore diameter in membranes. If foralumina membranes a slit-shaped pore model is used for the calculations, agood match is obtained between experimentally found values for pore size(distribution), porosity and BET values and those calculated from particlesize (XRD, electron microscopy) and model representations (Figure 2.7). Thisimplies that the agglomerates which are present in the colloidal suspension


(a) GAMMA-ALUMINA WITHOUT PVApore model: slit shaped

PORE RADIUS , (nm)

GAMMA-ALUMINA WITH PVApore model: slit shaped

00

1?

1T3

PORE RADIUS , (nm)

Figure 2.8. Typical examples of a pore size distribution for (a) y-alumina membranes; desorptionbranch (b) anatase titania membranes; desorption branch.

(b)

1.60


TITANIA WITHOUT BINDERSpore model: cylindrical shaped

PORE RADIUS , (nm)

t

>

TITANIA WITH PVA AND HPC (Mw = 10E5)pore model: cylindrical shaped

PORE RADIUS , (nm)Figure 2 (continued)


are "collapsed" during the type of drying process in which large capillarypressures are used. Cot, Guizard and Larbot (1988) used agglomerate size asa parameter to obtain larger pores for microfiltration membranes (seeSections 2.4.2 and 2.4.3 for zirconia and titania respectively).

2.4.2. Zirconia Membranes

Zirconia membranes on carbon supports were originally developed by UnionCarbide. Ultrafiltration membranes are commercially available now undertrade names like Ucarsep and Carbosep. Their outstanding quality is theirhigh chemical resistance which allows steam sterilization and cleaning pro-cedures in the pH range 0-14 at temperatures up to 80C. These systemsconsist of a sintered carbon tube with an ultrafiltration layer of a metallicoxide, usually zirconia. Typical tube dimensions are 10 mm (outer diameter)with a wall thickness of 2 mm (Gerster and Veyre 1985).

The fabrication of the ultrafiltration top layer is described in the patent ofCacciola and Leung (1981). The coating is applied to the macroporouscarbon support in the presence of a volatile liquid capable of drawing thecoating into the support and of desolvating the coating. The liquid isthereafter volatilized with the resulting dry ultrafiltration membrane beingcrack-free and having good mechanical and chemical stability. The supportmay be pretreated with a volatile liquid such as acetone, prior to theapplication of the coating to the support. Alternatively, the coating materialcan be dispersed in a volatile liquid, such as methanol, to form a coatingsuspension that is applied to the support.

The coating material, preferably zirconia, will have a primary particle sizein the range 20-150 nm and an aggregate size in the range 100-1000 nm. Thecoating thickness is about 20 /xm. The carbon tube support has a pore size ofabout 300 nm and a porosity of about 30%. Stability of the coating isenhanced by firing at temperatures of 400-600C. The separation propertiesdepend on the applied primary particle size and pores of 5 nm and larger canbe obtained (Veyre 1985). This type of ultrafiltration membrane is com-mercially available and is produced by Tech Sep (formerly SFEC) under thename of "Carbosep". The applications are in the field of treatment of waterstreams (oily wastes) (Bansal 1975, 1976) and food and dairy industry (e.g.cheese production, Veyre 1985a, b).

Guizard et al. (1986), Cot, Guizard and Larbot (1988) and Larbot et al.(1989) used a sol-gel method to prepare zirconia membrane top layers on analumina support. The water necessary for the hydrolysis of the Zr-alkoxidewas obtained from an esterification reaction. The complete hydrolysis wasdone at room temperature and resulted in a hydrated oxide. The precipitatewas peptized with nitric or hydrochloric acid at pH < 1.1 and the final sol


contained about 20% of metal oxide. Important parameters for the sol-geltransition are the pH and the concentration of the solution. In order to give acompact arrangement after gelation the particles should be in a maximumrepulsion state (Cot, Guizard and Larbot 1988). With a pH near theisoelectric point, larger particles are formed and poorly densified porous filmsresult. A similar effect is obtained by increasing the electrolyte concentrationin the solution. Cot, Guizard and Larbot (1988) report particle sizes of 10 nmfor a low electrolyte concentration and 60 nm for an electrolyte concentra-tion which is 100 times higher. This particle size increases further if the ZrO2concentration in the solution is increased above 1%. The particles areagglomerates whose sizes can be manipulated. After addition of a binder thezirconia membrane was formed on an alumina support using the slip-castingprocess. After drying (20-150C) and thermal treatment (400-900C) theresulting membranes showed a broad pore size distribution. The zirconiamembrane layer exists in the tetragonal phase up to 700C, above which aphase transition to the monoclinic form occurs. The pore diameter increasesgradually with increasing temperature from 6 nm (700C) to 70 nm (1200C).Results given by Cot, Guizard and Larbot (1988) show a curve with a meanpore radius of 32 nm with considerable "tails" with smaller and larger values.Jt appears that the membranes are prepared from rather large agglomerates.It is questionable if these membranes are crack- and defect-free (see Sec-tion 2.10).

2.4.3. Titania Membranes

Supported titania membranes are described by several authors (Burggraafet al. 1989, Zaspalis et al. 1989, Cot, Guizard and Larbot 1988), while thepreparation of nonsupported and supported titania membranes is reportedby Anderson et al. (1988) and Gieselman et al. (1988). Titania membranesshow excellent chemical resistance and interesting photochemical and cata-lytic properties. The results of titania membranes reported by Cot, Guizardand Larbot (1988) were obtained in the same way as described for zirconia.Particle diameters ranging from 20-170 nm are reported, depending onpreparation conditions (Cot, Guizard and Larbot 1988) and the same type ofbroad pore size distribution as reported for zirconia. Up to 500C thematerial is present in the anatase phase, above which it is transformed to therutile phase. The pore diameter increased gradually with increasing temper-ature from 6nm (500C) to 180 nm (U00C). Again it is doubtful if thesemembranes are defect-(crack) free (see below).

The results obtained by Burggraaf, Keizer and van Hassel (1989a, b) differmarkedly from those reported above by Cot, Guizard and Larbot (1988). Thepreparation of defect-free titania membranes was found to be much more


difficult than that of alumina membranes. It was necessary to use bin-ders/plasticizers and better results were obtained with very smooth supportsurfaces. Sols were obtained by hydrolysis of Ti-tetraisopropoxide dissolvedin isopropanol (0.45 M) at room temperature in an excess of a water/isopro-panol mixture (4.5 M water in isopropanol) to which a small amount ofsulfuric acid is added.

Peptization of the precipitate was obtained with nitric acid at a pH of 1.5 at70C under reflux conditions. At pH values higher than 3, peptization wasnot obtained. The peptized sols consisted of stable colloidal dispersion. Usinglight scattering technique agglomerate sizes of about 26 nm were observedwhich after ageing for two months increased to about 46 nm. On subjectingthe solution to ultrasound the agglomerate size decreased to about 9 nm.Transmission electron microscopy showed primary particle sizes of 4-5 nm(van Praag et al. 1990). Titania membranes were produced on aluminasupports from colloidal suspensions containing 0.1-0.2 mol/L TiO2 afteraddition of PVA (polyvinylalcohol) or a combination of PVA with hydroxy-propylcellulose (HPC). A layer thickness of about 1 /an was depositedfollowed by controlled drying and calcination at 450C. This procedure isrepeated until the required thickness is obtained and/or a defect-free mem-brane is obtained as indicated by gas permeation measurements. Non-supported membranes could be produced in this way with a thickness up to50 fim. The addition of SOj" ions was necessary to stabilize the anatasephase up to a temperature of 600C. Without this addition the anatase torutile phase transition takes place at 350-450C This is accompanied by adecrease of th

bhave r.r., (ed.) inorganic membranes. synthesis, characteristics and applications (only 1,2,3,6 and...

Documents