Optimising catalyst and membrane performance

and performing a fundamental analysis on the

dehydrogenation of ethanol and 2-butanol in a
catalytic membrane reactor

by

JOHAN NICO KEULER

B.Eng (Chemical)
M. Eng (Chemical)

I N ENGINEERING
(Chemical Engineering)

In the Department of Chemical Engineering

at the University of Stellenbosch

Promoters:
PROF. L LORENZEN
PROF. RD SANDERSON

STELLENBOSCB
September 2000
 DECLARATION

I hereby certifY that this dissertation is my own original work, except where specifically
acknowledged in the text. Neither the present dissertation, nor any part thereof, has previously
been submitted for a degree at any University

JNKEULER
September 2000
 SUMMARY

Stricter government regulations and higher energy costs have forced the chemical industry to
focus more on environmentally friendly processes and to reduce energy consumption. The
main goals of chemical companies are to obtain a high product yield and selectivity, and to .
reduce unwanted side products. Furthermore, if reactions can be performed at lower
temperature, while maintaining the reaction conversion, it will result in large energy savings.

Low temperature dehydrogenation reactions (below 300°C) are very selective and do not
produce many by-products, but conversion is limited by the reaction equilibrium. The
conversion limitations have resulted in the development of alternative processes in recent
years for producing alkenes from alkanes and aldehydes or ketones from alcohols. Advances
in membrane technology have created the possibility of using a new type of reactor, called a
catalytic membrane reactor, in which separation and reaction occurs simultaneously. A
catalytic membrane reactor, of the palladium composite type, can selectively remove hydrogen
and manipulate the reaction equilibrium in dehydrogenation reactions. The possibility exists
to save energy, obtain high conversions and perform very selective reactions in the catalytic
membrane reactor.

This dissertation describes a thorough investigation carried out into the design, optimisation,
operation and modelling of a catalytic membrane reactor. The two components of the
membrane reactor, i.e. the catalyst packing and the membrane structure, were optimised
individually for the dehydrogenation of ethanol and 2-butanol. The optimised catalyst and
optimised membrane were combined and their combined performance compared to a
conventional plug flow reactor. A fundamental model was developed for the catalytic
membrane reactor and a full sensitivity analysis was conducted to test the effects of membrane
parameters, reaction rate parameters and process variables on reaction conversion.

Copper-based catalysts were prepared by impregnation of alumina, silica and magnesium

oxide supports. The low surface area of MgO yielded poor conversions, while the Ah03
support favoured dehydration and not dehydrogenation. Silica supports yielded good

11
dehydrogenation results and an optimum copper loading of 15% gave the highest
dehydrogenation conversion for both the ethanol and the 2-butanol reactions. The copper-
silica catalyst was stable at 280°C and below for ethanol dehydrogenation, and at 250 °C and
below for 2-butanol dehydrogenation. At higher temperatures in ethanol dehydrogenation, the
catalyst deactivated due to both sintering and coking. Kinetic data from 200°C to 300 °C for
ethanol dehydrogenation, and from 190°C to 280 °C for 2-butanol dehydrogenation, indicated
that both reactions could be well described by the dual site, surface reaction, controlling
mechanism.

Significant advances were made in the production of very thin Pd films (1.0 to 1.5 1IDl) on the
inside of 200 run a-alumina membrane tubes (from Societe des Ceramiques Techniques). A
modified electro less plating technique was used for producing the Pd films. Hydrogen
permeances through the films varied between about 8 and 15 1IDl01lm2.Pa.s for temperatures
from 330°C to 450 °C and palladium films from 1.0 to 1.5 1IDl. Hydrogen to nitrogen
selectivity was greater than 100 for all membranes tested and greater than 400 for all but two
membranes (thickness 1.0 to 1.5 1IDl). These values are a significant improvement over other
published results. Pd membranes can only be used above 300°C and since the catalyst was
unstable in that temperature region, Pd-Ag membranes had to be prepared. Pd-Ag films of
thickness less than 2.2 Ilm were successfully synthesised and tested. Good high temperature
(500°C) and low temperature (below 300°C) stability was obtained for the Pd-Ag
membranes.

The performance of an optimised Pd-Ag membrane, packed with an optimised 14.4 wt %

copper on silica catalyst (the catalytic membrane reactor), was compared to that of a plug flow
reactor. The best results for ethanol dehydrogenation in a membrane reactor were obtained at
275°C. At this temperature, the membrane reactor improved the total exit ethanol conversion
from 45% (Plug flow reactor) to 60% at low feed flow rates and from 36% to 46% at high feed
flow rates. The maximum 2-butanol conversion in a plug flow reactor at 240°C was 80% and
that was increased to above 90% for the membrane reactor.

III
 OPSOMMING

Strenger wetgewing en hoilr energiekoste dwing die chemiese industrie om te kyk na meer
omgewingsvriendelike prosesse en om energieverbruik te minimeer. Die belangrikste
doelwitte van chemiese maatskappye is om hoe opbrengste en hoe selektiwiteite te verkry van
hul verlangde produkte. Ongewenste byprodukte moet so ver as moontlik uitgeskakel word.
Reaksies wat hoe omsettings gee by laer temperature sal lei tot groot energie besparings.

Lae temperatuur dehidrogeneringsreaksies (onder 300°C) is baie selektief met min newe
produkte, maar omsetting word beperk deur die reaksie ewewig. Ewewigsbeperkings in
konvensionele prosesse het gelei tot die ontwikkeling van nuwe prosesse in die laaste dekade
vir die produksie van alkene vanaf alkane en vir aldehiede en ketone vanaf alkohole.
Membraanontwikkeling het die moontlikheid geskep vir 'n nuwe generasie reaktore, die
katalitiese membraanreaktore, waarin skeiding en reaksie gelyktydig plaasvind. 'n Palladium-
tipe reaktor kan selektief waterstof skei en die ewewig verskuif in dehidrogeneringsreaksies.
Die moontlikheid bestaan om energie te bespaar en om hoil omsetting sowel as hoe
selektiwiteit te verkry in die katalitiese membraanreaktor.

Hierdie proefskrif beskryf 'n omvattende ondersoek van die ontwerp, bedryf, optimering en
modellering van 'n katalitiese membraanreaktor. Die twee komponente van die
membraanreaktor, nl. die katalis pakking en die membraanstruktuur, is individueel geoptimeer
vir die dehidrogenering van etanol en 2-butanol. Die optimum katalis en die optimum
membraan is gekombineer en hul gekombineerde werking is vergelyk met die werking van 'n
propvloeireaktor. 'n Fundamentele model is gefonnuleer vir die membraanreaktor en 'n
volledige sensitiwiteitsanalise is uitgevoer op die model. Die effek van die reaksietempo
parameters, membraan parameters en die proses veranderlikes op die reaksie omsetting is
ondersoek.

Koper kataliste is berei deur die impregnasie van alumina, silika en magnesIum oksied
partikels. Die lae oppervlak area van MgO het gelei tot lae omsettings terwyl alumina meer
dehidrasieprodukte as dehidrogeneringsprdukte opgelewer het. Koper op silika het die beste

IV
dehidrogeneringsresultate getoon met 'n optimum koperkonsentrasie van 15 massa % op
silika. Vir laasgenoemde koperkonsentrasie is die hoogste dehidrogeneringsomsettings verkry
vir beide die etanol en 2-butanol reaksies. Die koper-silika katalis was stabiel by 280°C en
laer vir etanol dehidrogenering en by 250°C en laer vir 2-butanol dehidrogenering. By hoer
temperature (etanol dehidrogenering) het die katalis begin deaktiveer weens sintering en
koolstofvorrning. Kinetiese data vanaf 200°C tot 300 °C vir etanol dehidrogenering en vanaf
190°C tot 280 °C vir 2-butanol dehidrogenering het daarop gedui dat die reaksies goed
beskryfkan word deur die dubbel posisie, oppervlak reaksie, beherende meganisme.

Goeie vordering is gemaak met die vervaardiging van baie dun Pd films (1.0 tot 1.5 /lm) op
die binnekant van 200 om a-alumina membraanbuise (verskaf deur Societe des Ceramiques
Techniques). 'n Gemodifiseerde elektrodelose plateringstegniek is gebruik vir die
vervaardiging van die films. Waterstof perrneasies deur die films het gewissel van 8 tot
15 /lmol/m 2.Pa.s vir temperature vanaf330 °C tot 450°C en palladium filmdiktes vanaf 1.0 tot
1.5/lm. Waterstoftot stikstofselektiwiteit was meer as 100 vir al die getoetste membrane en
meer as 400 vir almal behalwe twee membrane (dikte 1.0 tot 1.5 1J.IIl). Hierdie waardes is
beduidend beter as ander gepubliseerde resultate. Pd membrane kan slegs gebruik word bo
300°C en aangesien die kataliste onstabiel was in daardie temperatuurgebied, is Pd-Ag
membrane berei. Pd-Ag membrane met totale filmdikte kleiner as 2.2 IJ.IIl is suksesvol berei
en getoets. Die Pd-Ag membrane was stabiel tussen 200°C en 500 0c.

Die werking van 'n geoptimeerde membraan, gepak met 'n geoptimeerde 14.4 massa % koper-
silika katalis (die katalitiese membraanreaktor), is vergelyk met 'n propvloeireaktor. Die beste
resultate verkry vir etanol dehidrogenering was by 275 °e. By daardie temperatuur is die
etanol omsetting verhoog vanaf 45% vir die propvloeireaktor tot 60% vir die membraanreaktor
by lae etanol vloeitempo's en vanaf 36% tot 46% by hoe etanol vloeitempo's. Die maksimum
2-butanol omsetting vir die propvloeireaktor was 80% by 240°C en dit is verhoog tot bo 90%
vir die membraanreaktor.

v
 TABLE OF CONTENTS

CHAPTER 1: INTRODUCTION 1

CHAPTER 2: BACKGROUND AND LITERATURE SURVEY 6

2.1. MEMBRANE SEPARATION PROCESSES 6

2.2. INORGANIC MEMBRANES 6

2.2.1. Dense inorganic membranes 7
2.2.1.1. Dense metal membranes 7
2.2.1.2. Nonporous electrolyte membranes 7
2.2.1.3. Dense inorganic polymer membranes 8
2.2.1.4. Dense metal composite membranes 8
2.2.2. Porous inorganic membranes 9
2.2.2.1. Porous glass 10
2.2.2.2. Porous metal 10
2.2.2.3. Molecular sieving membranes 10
2.2.2.4. Porous ceramic and composite membranes 11
2.2.2.5. Zeolite membranes 11
2.2.3. Membrane modification 12
2.3. MEMBRANE PREPARATION AND CHARACTERISATION 13
2.4. SEPARATION MECHANISMS 14
2.4.1. Knudsen diffusion 15
2.4.2. Surface diffusion 16
2.4.3. Capillary condensation 16
2.4.4. Molecular sieve separation 16
2.4.5. Flow through non-porous membranes 17
2.5. PALLADIUM AND PALLADIUM ALLOYS 20
2.5.1. Palladium-hydrogen system 20
2.5.2. Palladium-silver systems 21
2.5.3. Palladium and copper or gold 22

VI
 2.5.4. Palladium and rare earth elements, uickel or others 22
2.6. PREPARING PALLADIUM MEMBRANES 23
2.6.1. Wet impregnation 24
2.6.2. Sol gel process 24
2.6.3. Vapour deposition techniques 25
2.6.3.1. Physical vapour deposition 25
2.6.3.2. Spray pyrolysis 26
2.6.3.3. Chemical vapour deposition 26
2.6.4. Electroplating 27
2.6.5. Electroless plating 27
2.6.5.1. Substrate pretreabnent 28
2.6.5.2. Electroless plating solution composition 29
2.6.5.3. Recent advances in electroless palladium plating 30
2.6.5.4. Electroless palladium-silver coatings 32
2.6.6. Palladium membrane temperature stability 33
2.6.7. Palladium-silver alloying 33
2.6.S. Deactivation of palladium membranes 34
2.7. APPLICATIONS OF INORGANIC MEMBRANES 35
2.S. CATALYTIC MEMBRANE REACTORS 36
2.S.1. General advantages of catalytic membrane reactors 37
2.S.2. Catalytic membrane applications 37
2.8.2.1. Membranes as distributors 38
2.8.2.2. Membranes for dehydrogenation reactions 38
2.8.2.3. Other applications 39
2.S.3. Problems with catalytic membrane reactors 39
2.9. DEHYDROGENATION OF ALKANES 40
2.9.1. General principles 41
2.9.2. Alkane dehydrogenation catalysts 41
2.9.3. Commercial dehydrogenation processes 42
2.9.3.1. FBD Process (Snamprogetti-Yarsintez) 43
2.9.3.2. UOP Oleflex process 44

Vll
 2.9.3.3. Lummus Catofin process 44
2.9.3.4. STAR process (Phillips steam active reforming) 45
2.9.3.5. Linde-BASF process 45
2.9.4. Alternative dehydrogenation processes 45
2.9.4.1. Coupled dehydrogenation and hydrogen oxidation 46
2.9.4.2. Oxidative dehydrogenation 46
2.9.4.3. Dehydrogenation in a membrane reactor 46
2.9.5. Methane steam reforming reaction 46
2.9.6. Water gas shift reaction 48
2.9.7. Ethane dehydrogenation 49
2.9.8. Dehydrogenation of propane 50
2.9.9. Dehydrogenation of butanes and butenes 52
2.9.9.1. Iso-butane dehydrogenation 52
2.9.9.2. n-Butane and butene dehydrogenation 53
2.9.10. Dehydrogenation of ethylbenzene to styrene 55
2.10. ALCOHOL DEHYDROGENATION CATALYSTS 56
2.10.1. Effect of copper percentage on catalyst performance 57
2.10.2. Catalyst preparation techniques 58
2.1 0.2.1. Precipitation 59
2.10.2.2. Urea hydrolysis 59
2.10.2.3. Electroless plating 60
2.10.2.4. Impregnation 60
2.10.3. Catalyst reduction 60
2.10.4. Catalyst deactivation 62
2.10.4.1. Sintering 62
2.10.4.2. Coking and poisoning 63
2.11. DEHYDROGENATION KINETICS 64
2.12. DEHYDROGENATION OF ALCOHOLS 65
2.12.1. Dehydrogenation of methanol to yield formaldehyde 65
2.12.2. Dehydrogenation of iso-propanol to acetone 66
2.12.3. Dehydrogenation of cyclohexanol to cyclohexanone 67

viii
 2.12.4. Dehydrogenation of C4 alcohols 69
2.12.5. Dehydrogenation of ethanol 69
2.12.5.1. Catalyst selectivity 70
2.12.5.2. Catalyst activity 72
2.12.5.3. Ethanol dehydrogenation kinetics 74
2.12.5.4. Oxidative dehydrogenation of ethanol 76
2.12.5.5. Dehydrogenation of ethanol in a membrane reactor 77
2.13. ALCOHOL DEHYDRATION REACTIONS 77
2.14. SUMMARY 78

3. EXPERIMENTAL PROCEDURES 80

3.1. OPTIMISING CATALYST COMPOSITION 80

3.1.1. Support types 80
3.1.2. Catalyst preparation 81
3.1.3. Summary of catalysts investigated 82
3.1.4. Catalyst testing for determining the optimum catalyst composition 83
3.1.4.1. Ethanol reaction 85
3.1.4.2. 2-Butanol reaction 86
3.1.5. Product analysis 86
3.1.6. Catalyst characterisation 87
3.2. KINETIC TESTING 88
3.2.1. Ethanol dehydrogenation 88
3.2.1.1. Catalyst deactivation testing 88
3.2.1.2. Characterising deactivated catalysts 90
3.2.1.3. Detennining the mass transfer limited regime 90
3.2.1.4. Determining kinetic parameters 91
3.2.2. 2-Butanol reaction 92
3.3. MEMBRANE PREPARATION AND TESTING 93
3.3.1. Membrane support 93
3.3.2. Composite membrane preparation 95
3.3.2.1. Electroless plating pre-treatment 95

IX
 3.3.2.2. Electroless Pd plating 97
3.3.2.3. Membrane cleaning 100
3.3.2.4. Electroless silver plating 101
3.3.3. Determining palladium film thickness 102
3.3.4. Membrane testing 102
3.3.4.1. Reactor temperature profiles .107
3.3.5. Membrane heat treatment and reduction 108
3.3.6. Palladium-silver alloying 109
3.3.7. Membrane characterisation 110
3.4. CATALYTIC MEMBRANE REACTOR EXPERIMENTS 111
3.4.1. Ethanol dehydrogenation 112
3.4.2. 2-Butanol dehydrogenation _ 112
3.5. SUMMARY 113

4. OPTIMISING CATALYST COMPOSITION 114

4.1. DEHYDROGENATION OF ETHANOL 114

4.1.1. MgO catalysts 114

4.1.2. Ah03 catalysts 117
4.1.2.1. BET and copper surface areas for Cu on Ah03 catalysts 120
4.1.2.2. Literature data on surface areas of copper catalysts 122
4.1.2.2.1. N20 titration 123
4.1.2.2.2. Chemisorption 123
4.1.2.2.3. Effect o/preparation method on catalyst characteristics 123
4.1.2.3. The effect of additives on ethanol conversion and product
distribution 125
4.1.2.4. Summary for using Cu on alumina catalysts 127
4.1.3. Silica catalysts 127
4.1.3.1. The effects of copper loading and temperature on
acetaldehyde yield 127
4.1.3.2. The effect of additives on product distribution and yields 129
4.1.3.3. Effect of particle size on acetaldehyde yield 131

x
 4.1.3.4. Surface area data for silica-based catalysts 132
4.1.3.4.1. TEM and XRD data/or copper on silica catalysts 134
4.2. DEHYDROGENATION OF 2-BUTANOL 136
4.2.1. MgO catalysts 136
4.2.2. Silica catalysts 139
4.2.2 .1. Effect of particle size on 2-butanol conversion . 141
4.3. SUMMARY 142
4.3.1. Ethanol reaction 142
4.3.2. 2-Butanol reaction 143

5. REACTION KINETICS 145

5.1. CATALYST DEACTIVATION TESTING 146

5.1.1. Effect of reduction temperature on ethanol reaction 146

5.1.2. Effect of reaction temperature and additives on catalyst stability 147
5.1.3. Further deactivation testing 149
5.l.3.l. Deactivation mechanism 153
5.1.4. Catalyst stability for dehydrogenation of 2-butanol 153
5.2. REACTION MASS TRANSFER RESISTANCE 154
5.2.1. Determining the mass transfer limiting regime for ethanol
dehydrogenation 154
5.2.2. Determining the mass transfer limiting regime for 2-butanol
dehydrogenation 155
5.3. KINETICS FOR ALCOHOL DEHYDROGENATION 156
5.3.1. Ethanol dehydrogenation 158
5.3 .1.1. Equation optimisation 162
5.3.2. 2-Butanol dehydrogenation .- 167
5.3.2.l. Equation optimisation 170

5.4. SUMMARY 172

6. PdMEMBRANE PREPARATION AND CHARACTERISATION 174

6.1. MEMBRANE STRUCTURE 174
6.2. SURFACE PRETREATMENT 175

Xl
 6.3. THE ELECTROLESS Pd PLATING PROCESS 176
6.3.1. Solution feeding to membrane tube 177
6.3.2. Effect of plating rate on membrane performance 179
6.3.3. Pd membrane thickness measurement 181
6.3.4. Membrane post plating treatment 181
6.4. THE EFFECT OF SUPPORT STRUCTURE ON Pd FILMS .183
6.5. STRUCTURAL CHARACTERISATION OF Pd MEMBRANES 184
6.6. PERMEANCE TESTING OF Pd MEMBRANES 186
6.6.1. The effect of.iP on H2 and N2 permeance 187
6.6.1.1. Nitrogen experiments 187
6.6.1.2. Hydrogen experiments 188
6.6.2. The effect oftemperature. on H2 and N2 permeance 190
6.6.3. The effect of film thickness on permeance 192
6.6.3.1. Hydrogen penneance 192
6.6.3.2. Nitrogen penneance 193
6.6.3.3. Membrane selectivity 194
6.6.3.4. Arrhenius parameters for hydrogen penneance 194
6.6.4. Comparison with literature data 195
6.6.4.1. Hydrogen penneances in the present study 196
6.6.5. H2 permeance employing a sweep gas 197
6.7. SUMMARY 204
7. Pd-Ag MEMBRANE ALLOYING AND STABILITY TESTING 206
7.1. Pd-Ag MEMBRANE PREPARATION 206
7.2. UNALLOYED Pd-Ag MEMBRANES 207
7.2.1. Characterising unalloyed Pd-Ag membranes 209
7.2.1.1. Unalloyed Pd-Ag membranes tested with a sweep gas 211
7.3. LITERATURE DATA ON ALLOYING Pd-Ag MEMBRANES 214
7.3.1. Alloying co-deposited Pd-Ag films 214
7.3.2. Alloying successive Pd-Ag films 214
7.3.3. Diffusion kinetics 215
7.4. ALLOYING RESULTS FOR Pd-Ag MEMBRANES 216

Xli
 7.4.1. Alloying in a tube furnace 217
7.4.1.1. Alloying at a temperature of 545°C 217
7.4.1.2. Alloying at a temperature of 530 °C 219
7.4.1.3. Alloying at a temperature of 590°C 220
7.4.2. Temperature stability of Pd-Ag membranes 220
7.4.2.1. Literature data on Pd membrane stability .220
7.4.2.2. Pd-Ag membrane stability 221
7.4.2.3. Structural changes ofPd-Ag membranes during heating 224
7.4.3. Alloying in situ in the reactor 227
7.4.3.1. Alloying in a hydrogen environment 227
7.4.3.2. Alloying in nitrogen and argon environments 229
7.4.3.3. Further treatment ofpd-Ag films 231
7.4.3.4. Suggested heat treatment procedure for Pd-Ag films 232
7.5. FULL CHARACTERISATION OF MEMBRANES N4x AND N8x 233
7.6. SUMMARY 234

8. CATALYTIC MEMBRANE REACTOR MODELLING 236

8.1. REACTANT FEED RATES 236

8.2. ALCOHOL PERMEANCE DATA 237

8.3. DEVELOPING A PROCESS MODEL 239
8.3.1. Model assumptions 239
8.3.2. Design equation 240
8.3.3. Modelling of mass transfer effects 242
8.3.3.1. Interphase mass transfer resistance modelling 242
8.3.3.1.1. Theeffectivenessfactor 246
8.3.3 .2. Intraparticle mass transfer resistance modelling 247
8.3.3.2.1. Effective diffusion coefficient (DeJ 249
8.3.3.3. Summary of modelling mass transfer 249
8.3.4. Mass transfer coefficients from experimental data 250
8.3.4.1. Ethanol dehydrogenation 250
8.3.4.2. 2-Butanol dehydrogenation 253

XIll
8.4. EXPERIMENTAL DATA FOR A PLUG FLOW REACTOR 253
8.4.1. Ethanol dehydrogenation in a plug flow reactor 253
8.4.2. 2-butanol dehydrogenation in a plug flow reactor 256
8.4.3. Summary of model for ethanol dehydrogenation 257
8.4.4. Summary of model for 2-butanol dehydrogenation 262
8.5. SELECTIVITY ANALYSIS OF THE PROCESS MODEL 266
8.5.1. The effect ofk'-values on model performance 267
8.5.1.1. Back diffusion 268
8.5.2. The effect of K. c • t on model performance 268
8.5.3. The effect of the membrane selectivity on model performance 269
8.5.4. The effect of the reaction effectiveness factor on model performance 270
8.5.5. The effect ofthe ethanol feed rate on model performance 270
8.5.6. The effect of the sweep gas on model performance 271

8.5.7. The effeCt of the H2 permeance at constant selectivity on model

performance 272
8.5.8. The effect of the H2 permeance at varying selectivity on model
performance 272
8.5.9. Model performance under non-standard conditions 273
8.5.10. Optimising membrane performance on model performance 274
8.6. SUMMARY 275

9. CATALYTIC MEMBRANE REACTOR DATA 278

9.1. ETHANOL DEHYDROGENATION 279

9.1.1. Membrane reactor experiments at 250°C 279

9.1.2. Membrane reactor experiments at 275 °C 283
9.1.3. Membrane reactor experiments at 300°C 286
9.1.4. Further discussion of ethanol dehydrogenation 288
9.2. 2-BUTANOL DEHYDROGENATION 291
9.2.1. Membrane reactor experiments at 190°C 291
9.2.2. Membrane reactor experiments at 215 °C 294
9.2.3. Membrane reactor experiments at 240°C 297

XIV
 9.2.4. Further discussion of 2-butanol dehydrogenation 300
9.3. SUMMARY 302

10. CONCLUSIONS 304

10.1. CATALYST OPTIMISATION 304

10.2. REACTION MODELLING FOR ETHANOL AND 2-BUTANOL
DEHYDROGENATION 305
10.3. MEMBRANE OPTIMISATION 306
10.4. MEMBRANE MODELLING 307
10.5. MEMBRANE REACTOR VS. PLUG FLOW REACTOR 307
10.6. MEMBRANE REACTOR MODELLING 308

11. FUTURE WORK 310

12. REFERENCES 311

LIST OF SYMBOLS 336

LIST OF ABBREVIATIONS 340

APPENDIX A: List of chemicals used 341

APPENDIX B 1: Ethanol and 2-butanol response factors as determined

with a HP G 1800A GC 343
APPENDIX B2: Ethanol and 2-butanol response factors as determined
with a HP 5850 GC 348

APPENDIX C1: Equations for determining ethanol dehydrogenation kinetics 351

APPENDIX C2: Equations for determining 2-butanol dehydrogenation kinetics 355

APPENDIX D1: Hydrogen and nitrogen permeance data for Pd films 358
APPENDIX D2: Hydrogen and nitrogen permeance data for Pd-Ag films 378

APPENDIXE: Literature data on Pd and Pd-alloy thin films 394

xv
APPENDIX Fl: Plug flow reactor data and membrane reactor data for
ethanol dehydrogenation 402
APPENDIX F2: Plug flow reactor data and membrane reactor data for
2-butanol dehydrogenation 410

APPENDIXG: Sensitivity analysis for the catalytic membrane reactor model 420

XVI
 LIST OF FIGURES

CHAPTER 1:
Figure 1.1: Project flow diagram 5

CHAPTER 2:
Figure 2.1: Separation by a semi-penneable membrane 6
Figure 2.2: Knudsen diffusion 15
Figure 2.3: Surface diffusion 15
Figure 2.4: Capillary condensation 15
Figure 2.5: Molecular sieve separation 15
Figure 2.6: Hydrogen and oxygen flow through a non-porous membrane 17
Figure 2.7: Equilibrium solubility isothenns ofPdHn for bulk Pd (Shu et aI., 1991) 20
Figure 2.8: Penneability ofH2 through Pd alloy membranes at 350°C and 2.2 Mpa 23
Figure 2.9: Pore blocking by deposited copper 58
Figure 2.10: The effect of temperature on copper surface structure 61

CHAPTER 3:
Figure 3.1: Set-up used for testing the kinetics of the catalyst at the CNRS, France 84
Figure 3.2: SCT membrane structure 94
Figure 3.3: Curing process for enamelled membranes 94
Figure 3.4: Pre-treatment set-up 96
Figure 3.5: Electroless Pd plating set-up 99
Figure 3.6: Membrane reactor used to test the membrane penneance 103
Figure 3.7: Set-up used for membrane testing at room temperature 103
Figure 3.8: Set-up used for high temperature (> 300°C) hydrogen and nitrogen
penneance testing 105
Figure 3.9: Set-up used for testing ethanol penneance 106
Figure 3.10: Reactor temperature profiles at different oven temperatures 108
Figure 3.11: Membrane packed with catalyst III

XVll
CHAPTER 4:
Figure 4.1: Total ethanol conversion for Cu on MgO catalysts 115
Figure 4.2: Ethanol to acetaldehyde yield (Cu on MgO catalysts) 115
Figure 4.3: Acetaldehyde selectivity for Cu on MgO catalysts 115
Figure 4.4: Total ethanol conversion (Cu on alumina catalysts) 118
Figure 4.5: Ethanol to ethene yield (Cu on alwnina catalysts) 118
Figure 4.6: Ethanol to di-ethyl ether yield (Cu on alwnina catalysts) 118
Figure 4.7: Ethanol conversion to di-ethyl ether (0% Cu on alumina) at
different W IF ratios 118
Figure 4.8: Acetaldehyde yields for Cu on alumina catalysts (WIF = 393 kg.s/mol) 119

Figure 4.9: Butene yields for Cu on alumina catalysts (average WIF) 119

Figure 4.10: Effect of Cu% on catalyst surface area for Cu on alumina catalysts 120
Figure 4.11: Effect of Cu% on Cu surface area for alumina-based catalysts 120
Figure 4.12: Pore size distribution for Cu on alwnina catalysts 121
Figure 4.13: Effect ofCr, Ni and Co (on alumina) on total ethanol conversion 125
Figure 4.14: Effect ofCr, Ni and Co (on alumina) on ether yield 125
Figure 4.15: Effect ofCr, Ni and Co (on alwnina) on ethene yield 126
Figure 4.16: Total ethanol conversion (Cu on silica catalysts) 128
Figure 4.17: Ethanol to acetaldehyde yield (Cu on silica catalysts) 128
Figure 4.18: Ethanol to acetaldehyde selectivity (Cu on silica catalysts) 128
Figure 4.19: Total ethanol conversion (15 wt % Cu on silica catalyst) 129
Figure 4.20: Ethanol converted to acetaldehyde (IS wt % Cu on silica) 129
Figure 4.21: Effect of Co, Cr orNi additives on total ethanol conversion (Cu on silica) 130
Figure 4.22: Effect of Co, Cr or Ni additives on acetaldehyde yield (Cu on silica) 130
Figure 4.23: Effect of Co, Cr or Ni additives on total ethanol conversion (Cu on silica) 131
Figure 4.24: Effect of Co, Cr or Ni additives on acetaldehyde yield (Cu on silica) 131
Figure 4.25: Effect of catalyst particle size on total ethanol conversion 131
Figure 4.26: Effect of catalyst particle size on acetaldehyde yield percentage 131
Figure 4.27: Effect ofCu% on catalyst surface area for Cu on silica catalysts 133
Figure 4.28: Effect of Cu% on Cu surface area for silica-based catalysts 133
Figure 4.29: Pore size distribution for copper on silica catalysts 133

XVI11
Figure 4.30: TEM images of the unused 14.4 wt % copper on silica catalyst 135
Fig 4.30a: TEM (20 nrn) 135
Fig 4.30b: TEM (20 nrn) 135
Fig 4.30c: TEM (20 nrn) 135
Fig 4.30d: TEM (20 nrn) 135
Fig 4.30e: TEM (20 nrn) 135
Fig 4.30f: TEM (20 nrn) 135
Fig 4.30g: TEM (50 nrn) 135
Fig 4.30h: TEM (200 nrn) 135
Fig 4.30i: TEM (200 nrn) 135
Figure 4.31 : Total2-butanol conversion for Cu on MgO catalysts 137
Figure 4.32: Total butene yield for Cu on MgO catalysts 137
Figure 4.33: MEK yield for Cu on MgO catalysts 138
Figure 4.34: MEK yield for a 16.9 wt % Cu on MgO catalyst 138
Figure 4.35: MEK selectivity for a 16.9 wt % Cu on MgO catalyst 138
Figure 4.36: Total2-butanol conversion for Cu on silica catalysts 139
Figure 4.37: MEK yield for Cu on silica catalysts 139
Figure 4.38: Butene yields for Cu on silica catalysts 140
Figure 4.39: MEK yield for a 15 wt % Cu on silica catalyst 141
Figure 4.40: MEK selectivity for a 15 wt % Cu on silica catalyst 141
Figure 4.41 : Effect of catalyst particle size on total 2-butanol conversion 141
Figure 4.42: Effect of catalyst particle size on MEK yield percentage 141
Figure 4.43: Equilibrium ethanol conversion vs. measured values for a 15 wt %
Cu on silica catalyst 143
Figure 4.44: Equilibrium 2-butanol conversion vs. measured values for a 15 wt %
Cu on silica catalyst 144

CHAPTERS:
Figure 5.1: Effect of reduction T on acetaldehyde production rate over time
(14.4 wt % Cu on silica) 147

XIX
Figure 5.2: Effect of reduction T on acetaldehyde production rate over time
(14.4 wt % Cu on silica) 147
Figure 5.3: Acetaldehyde production rate as a function of time
and temperature (14.4 wt % Cu on silica) 148
Figure 5.4: Effect of additives (Cr and Co) on acetaldehyde production rate 149
Figure 5.5: Effect of additives (Cr and Co) on acetaldehyde production rate at 280°C 149
Figure 5.6: TEM images of a 14.4 wt % copper on silica catalyst after
being in use 400°C lSI
Figure 5.6a: TEM (20nm) 150
Figure 5.6b: TEM (20nm) 150
Figure 5.6c: TEM (50 nm) 150
Figure 5.6d: TEM(50nm) 150
Figure 5.6e: . TEM(100nm) 150
Figure 5.6f: TEM(100nm) 150
Figure 5.6g: TEM (100 nm) lSI
Figure 5.6h: TEM(200nm) 151
Figure 5.6i: TEM (1 fUll) 151
Figure 5.7: Catalyst activity (14.4 wt % Cu on silica) after re-oxidation 152
Figure 5.8: MEK production rate as a function of time for a 14.4 wt %
Cu on silica catalyst 154
Figure 5.9: The effect of feed flow rate on acetaldehyde production rate ISS
Figure 5.10: The effect of 2-butanol feed flow rate on the MEK production rate 156
Figure 5.11: Linear fits of reaction rate data at I atm. total pressure and 225°C 158
Figure 5.12: Linear fits of reaction rate data at I atm. total pressure and 300 °C 158
Figure 5.13: % Catalyst deactivation after 24 hours of operation 160
Figure 5.14: Parameters for ethanol reaction equation as a function of temperature 161
Figure 5.15: Difference between model 1 rates and actual rates at different
temperatures 164
Figure 5.16: Difference between model 2 rates and actual rates at different
temperatures 164
Figure 5.17: Percentage deviation between model 1 rates and measured rates 165

xx
Figure 5.18: Percentage deviation between model 2 rates and measured rates 165
Figure 5.19: Modell rates vs. actual rates at different rate values 165
Figure 5.20: Model 2 rates vs. actual rates at different rate values 165
Figure 5.21: Percentage deviation between model values and measured values
at 573 K 166
Figure 5.22: Linear fits of reaction rate data at I atm. total pressure and 220°C 168
Figure 5.23: Linear fits of reaction rate data at 1 atm. total pressure and 280 °C 168
Figure 5.24: Parameters for 2-butanol reaction equation as a function oftemperature 170
Figure 5.25 Comparison between measured reaction rates and model reaction rates 172

CHAPTER 6:
Figure 6.1: Cross section view of a three layer SCT membrane 175
Figure 6.2: Top view (20 OOOx) of a three layer SCT membrane 175
Figure 6.3: Top view (5 OOOx) of a three layer SCT membrane 175
Figure 6.4: Pd concentration in solution after repeated plating sessions 178
Figure 6.5: Cross section of membrane (a) (10 OOOx) 179
Figure 6.6: Top view of membrane (a) (25 OOOx) 179
Figure 6.7: Top view of membrane (a) (50 OOOx) 180
Figure 6.8: Cross section of membrane (b) (10 OOOx) 180
Figure 6.9: Top view of membrane (b) (25 OOOx) 180
Figure 6.10: Top view of membrane (b) (5000x) 181
Figure 6.11: Top view of membrane (b) (25000x) 181
Figure 6.12: Effect of oxidation post treatment on H2 permeance
(membrane 2, 1.43 jl111 Pd) 182
Figure 6.13: Effect of oxidation post treatment on selectivity
(membrane 2, 1.43 jl111 Pd) 182
Figure 6.14: Cross section of membrane (3b) (10 OOOx) 184
Figure 6.15: Top view of membrane (3b) (5000 x) 185
Figure 6.16: Top view of membrane (3b) (25 OOOx) 185
Figure 6.17: Cross section of membrane (11) (10 OOOx) 185
Figure 6.18: Top view of membrane (11) (5000 x) 186

XXI
Figure 6.19: Top view of membrane (11) (25 OOOx) 186
Figure 6.20: Effect of pressure on N2 permeance for a 1.47 micron Pd film (6) 188
Figure 6.21: Effect of pressure on N2 permeance for a 2.4 micron Pd film (3a) 188
Figure 6.22: Effect of pressure on N2 permeance for a 3.08 micron Pd film (la) 188
Figure 6.23: Effect of pressure on N2 permeance for a 4.43 micron Pd film (3b) 188
Figure 6.24: Effect of pressure on H2 permeance for a 1.47 micron Pd film (6) 189
Figure 6.25: Effect of pressure on H2 permeance for a 2.4 micron Pd film (3a) 189
Figure 6.26: Effect of pressure on Hz permeance for a 3.08 micron Pd film (la) 189
Figure 6.27: Effect of pressure on Hz permeance for a 4.43 micron Pd film (3b) 189
Figure 6.28: Selectivity data for membrane (3a) 190
Figure 6.29: Selectivity data for membrane (N8) 190
Figure 6.30: Selectivity data for membrane (N2) 191
Figure 6.31: Hydrogen permeance in J.lmollmz.Pa.s (membrane 2a) 191
Figure 6.32: Hydrogen permeance in J.lmollm 2.Pa.s (membrane N7) 191
Figure 6.33: Hydrogen permeance for Pd films from 1 to 6.5 micron thickness 192
Figure 6.34: Hydrogen permeance for Pd films from 1 to 1.5 micron thickness 193
Figure 6.35: Nitrogen permeance for Pd films from 1 to 6.5 micron thickness 193
Figure 6.36: Nitrogen penneance for Pd films from 1 to 1.5 micron thickness 193
Figure 6.37: Hz to N z selectivity for Pd films from 1 to 6.5 micron thickness 194
Figure 6.38: Hz to Nz selectivity for Pd films from 1 to 1.5 micron thickness 194
Figure 6.39: Arrhenius parameters for hydrogen permeance (1 to 6.5 micron Pd films) 195
Figure 6.40: Arrhenius parameters for hydrogen permeance (1 to 1.5 micron Pd films) 195
Figure 6.41: % Hz permeated with N2 sweep gas and space time = 2.37 s 201
Figure 6.42: % Hz permeated with N z sweep gas and space time = 1.19 s 201
Figure 6.43: Hydrogen flow in tube (cmJ/min) as predicted by model for
N7 (t = 2.37 s) 202
Figure 6.44: Hydrogen flow in tube (cmJ/min) as predicted by model for
N7 (t = 2.37 s) 202
Figure 6.45: Hydrogen flow in tube (cml/min) as predicted by model for
N7 (t = 1.19 s) 202

XXll
Figure 6.46: Hydrogen flow in tube (cm3/min) as predicted by model for
N7 (,t = 1.19 s) 202
Figure 6.47: % H2 permeated with N z sweep gas and space time = 2.37 s 203
Figure 6.48: % H2 permeated with N2 sweep gas and space time = 1.19 s 203
Figure 6.49: Hydrogen flow in tube (cm3/min) as predicted by model for
N4 (1: = 2.37 s) 203
Figure 6.50: Hydrogen flow in tube (cm3/min) as predicted by model for
N4 (1: = 2.37 s) 203
Figure 6.51: Hydrogen flow in tube (cm 3/min) as predicted by model for
N4 (1:= 1.19 s) 204
Figure 6.52: Hydrogen flow in tube (cm3/min) as predicted by model for
N4 (1:= 1.19 s) 204

CHAPTER 7:
Figure 7.1: Hydrogen permeances forPd and Pd-Ag membranes (8, NI, 8b, Nib) 209
Figure 7.2: Hydrogen permeances for Pd and Pd-Ag membranes (N3, N4, N3b, N4b) 210
Figure 7.3: Measured vs. calculated % H2 permeance for 8b 212
Figure 7.4: Measured vs. calculated % H2 permeance for NI b 212
Figure 7.5: Measured vs. calculated % H2 permeance for N3b 212
Figure 7.6: Measured vs. calculated % H2 permeance for N4b 212
Figure 7.7: % H2 permeated with N2 sweep gas and space time = 2.37 s 213
Figure 7.8: % H2 permeated with N2 sweep gas and space time = 1.19 s 213
Figure 7.9: Hydrogen flow in tube (cm3/min) as predicted by model for
N3b (1: = 2.37 s) 213
Figure 7.10: Hydrogen flow in tube (cm3/min) as predicted by model for
N3b (,t = 2.37 s) 213
Figure 7.11: Hydrogen flow in tube (cm3/min) as predicted by model for
N3b (F 1.19 s) 213
Figure 7.12: Hydrogen flow in tube (cm3/min) as predicted by model for
N3b (1: = 1.19 s) 213

XX111
Figure 7.13: Heating times required to obtain similar Pd-Ag diffusion at different
temperatures 216
Figure 7.14: Effect of heating time in argon on hydrogen permeance through a
Pd-Ag film (membranes N3b to N3e) 208
Figure 7.15: Effect of heating in argon on Hz permeance for N4 220
Figure 7.16: Effect of heating in argon on N z permeance for N4 222
Figure 7.17: Effect of heating in argon on nitrogen permeance (N3) 223
Figure 7.18: Effect of heating in argon on selectivity (N3) 223
Figure 7.19: Membrane dl (25 OOOx) 225
Figure 7.20: Membrane d2 (25 OOOx) 225
Figure 7.21: Membrane d3 (25 OOOx) 225
Figure 7.22: Membrane dl (50000x) 225
Figure 7.23: Membrane d2 (50 OOOx) 225
Figure 7.24: Membrane d3 (50 OOOx) 225
Figure 7.25: Membrane dl (5 OOOx) 226
Figure 7.26: Membrane d2 (5 OOOx) 226
Figure 7.27: Membrane d3 (5 OOOx) 226
Figure 7.28: Membrane d3 (2000x) 226
Figure 7.29: Side view of d2 226
Figure 7.30: Side view of d3 226
Figure 7.31: Effect of heating time in Hz on Hz and N2 permeance for N4x 228
Figure 7.32: Effect of heating time in H2 at 590°C on Hz permeance for NIx 228
Figure 7.33: Effect of heating time in Hz at 590°C on Nz permeance for NIx 228
Figure 7.34: Effect of temperature on Hz and N z permeance for NIx 229
Figure 7.35: Effect of heating time in N2 and Ar at 500 and 550°C on Hz
permeance for N2x 230
Figure 7.36: Effect of heating time in N2 at 500°C on N2 permeance for N2x 230
Figure 7.37: Effect of heating time in Ar at 500 and 550°C on Ar permeance for N2x 231
Figure 7.38: Effect of temperature on H2 permeance for N2x 231
Figure 7.39: Hz and N z permeances for N4x 233
Figure 7.40: Hz and N2 permeances for N8x 233

XXIV
Figure 7.41: Measured vs. calculated % Hz permeance for N4x 234
Figure 7.42: Measured vs. calculated % Hz permeance for N8x 234

CHAPTER 8:
Figure 8.1: Theoretical effect of feed rate on equilibrium conversion 236
Figure 8.2: Comparison between Hz to N z and Hz to ethanol selectivities
for a 1.5 flm Pd film 238
Figure 8.3: Ethanol to nitrogen permeance ratio for a 1.5 flm Pd film 238
Figure 8.4: Description of catalytic membrane reactor process 241
Figure 8.5: Effect of Reynolds number on effectiveness factor 251
Figure 8.6: C! vs. inverse Re at different temperatures 252
Figure 8.7: C! vs. inverse ReO. 8 ! at different temperatures 252
Figure 8.8: Plug flow reactor data for ethanol dehydrogenation at 250°C 254
Figure 8.9: Plug flow reactor data for ethanol dehydrogenation at 275 °C 254
Figure 8.10: Plug flow reactor data for ethanol dehydrogenation at 300°C 254
Figure 8.11: Selectivity towards acetaldehyde production for a plug flow reactor 254
Figure 8.12: Re as a function of temperature and conversion
(feed rate = 4.77*10. 5 molls) 255
Figure 8.13: Re as a function oftemperature and conversion
(feed rate = 9.54*10. 5 molls) 255
Figure 8.14: Re as a function of temperature and conversion
(feed rate = 1.43*10.4 molls) 255
Figure 8.15: Plug flow reactor data for 2-butanol dehydrogenation at 190°C 256
Figure 8.16: Plug flow reactor data for 2-butanol dehydrogenation at 215°C 256
Figure 8.17: Plug flow reactor data for 2-butanol dehydrogenation at 240°C 256
Figure 8.18: Selectivity towards MEK production for a plug flow reactor 256
Figure 8.19: Effect of sweep gas to standard feed molar ratio on ethanol conversion
and ethanol losses 275

xxv
CHAPTER 9:
Figure 9.1: Ethanol conversion at 250°C vs. sweep gas flow rate
(Fe' = 2.39*10.5 molls) 280
Figure 9.2: Ethanol conversion at 250°C vs. sweep gas flow rate
(Fe' = 4.77*10-5 molls) 280
Figure 9.3: Measured acetaldehyde selectivity (250°C) 282
Figure 9.4: Ethanol conversion at 275°C vs. sweep gas flow rate
(Fe' = 4.77*10"5 molls) 283
Figure 9.5: Ethanol conversion at 275°C vs. sweep gas flow rate
(Fe' = 9.54*10- 5 molls) 283
Figure 9.6: Ethanol conversion at 275°C vs. sweep gas flow rate
(Fe' = 1.43*10-4 molls) 284
Figure 9.7: Acetaldehyde yield at 275°C for a membrane reactor 284
Figure 9.8: Measured acetaldehyde selectivity (275°C) 286
Figure 9.9: Ethanol conversion at 300°C vs. sweep gas flow rate
(Fet = 4.77*10- 5 molls) 287
Figure 9.10: Ethanol conversion at 300°C vs. sweep gas flow rate
(Fe' = 9.54*10-5 molls) 287
Figure 9.11: Measured acetaldehyde selectivity (300°C) 288
Figure 9.12: Measured acetaldehyde selectivity at a constant feed rate
(Fe' = 9.54*10-5 molls) 290
Figure 9.13: 2-Butanol conversion at 190°C vs. sweep gas flow rate
(FZBu,= 1.52*10-5 molls) 292
Figure 9.14: 2-Butanol conversion at 190°C vs. sweep gas flow rate
(FzBu ,= 3.04*10-5 molls) 292
Figure 9.15: 2-Butanol conversion at 190°C vs. sweep gas flow rate
(F2Bu,= 4.56* 10-5 molls) 292
Figure 9.16: MEK yield at 190°C 292
Figure 9.17: Measured MEK selectivity at 190°C 294
Figure 9.18: 2-Butanol conversion at 215°C vs. sweep gas flow rate
(FzBu,= 1.52*10-5 molls) 295

XXVI
Figure 9.19: 2-Butanol conversion at 215°C vs. sweep gas flow rate
(F 2But = 3.04*10-5 moUs) 295
Figure 9.20: 2-Butanol conversion at 215°C vs. sweep gas flow rate
(F 2But = 6.08*10-5 moUs) 296
Figure 9.21: MEK yield at 215°C 296
Figure 9.22: Measured MEK selectivity at 215°C 297
Figure 9.23: 2-Butanol conversion at 240°C vs. sweep gas flow rate
(F 2But = 1.52*10-5 moUs) 298
Figure 9.24: 2-Butanol conversion at 240°C vs. sweep gas flow rate
(F 2But = 3.04*10.5 moUs) 298
Figure 9.25: 2-Butanol conversion at 240°C vs. sweep gas flow rate
(F 2But = 6.08*10. 5 moUs) 299
Figure 9.26: MEK yield at 240°C 299
Figure 9.27: Measured MEK selectivity at 240°C 300
Figure 9.28: MEK yield % VS. flow rate and temperature 301
Figure 9.29: Effect of sweep ratio on MEK yield % 301

XXVll
 LIST OF TABLES

CHAPTER 2:
Table 2.1: The effect of repairing of electro less Pd plated coatings (Li et aI., 1999) 31
Table 2.2: Operating conditions for various dehydrogenation processes 42
Table 2.3: Process efficiency for propane and iso-butane dehydrogenation 43
Table 2.4: Catalyst supports (Carrizosa and Munuera, 1977; Rosynek et aI., 1990) 57
Table 2.5: Possible rate equations for solid catalysed dehydrogenation reactions 64
Table 2.6: Common supports for copper catalysts 67
Table 2.7: Selectivities for ethanol dehydrogenation (Iwasa and Takezawa, 1991) 71
Table 2.8: Effect of Cr and Co addition to Cu on ethanol dehydrogenation
(Church et aI., 1951) 71
Table 2.9: Reaction rate parameters for ethanol dehydrogenation (Tu et aI., 1994b) 72
Table 2.10: Activities and TOFs for ethanol dehydrogenation at 190°C
Kanoun et al. (1991, 1993) 73

CHAPTER 3:
Table 3.1: Characteristics of alumina and silica supports used 81
Table 3.2: Different copper loadings investigated for copper-supported catalysts 83
Table 3.3: Other copper supported catalysts 83
Table 3.4: Reactor conditions used to test Cu on Si02 and MgO supports in
the ethanol dehydrogenation reaction 85
Table 3.5: Reactor conditions used to test Cu on Al20 3 supports in the ethanol
dehydrogenation reaction 86
Table 3.6: Reactor conditions used to test Cu on SiOz and MgO supports in
the ethanol dehydrogenation reaction 86
Table 3.7: Different combinations of reduction and reaction temperatures studied 89
Table 3.8: Summary of catalyst deactivation tests for ethanol dehydrogenation 89
Table 3.9: Determining the region free of interphase mass transfer resistance 91
Table 3.10: Conditions investigated for determining interphase mass transfer
limited regime for 2-butanol dehydrogenation 93

XXVlll
 Table 3.11: Membrane layer characteristics of a SCT membrane 94
Table 3.12: Compositions of pre-treatment solutions 96
Table 3.13: Stirring sequence and times used in pre-treatment 97
Table 3.14: Composition of the Pd plating solution per litre (for 2.00 g/litre Pd
in solution) 98
Table 3.15: Plating procedure used for producing Pd films 100
Table 3.16: Composition of silver plating bath per litre of plating solution 101
Table 3.17: Plating procedure used for producing Ag films 102

CHAPTER 4:
Table 4.1: BET surface areas for Cu on MgO catalysts in m2/g 116
Table 4.2: Yield matrix indicating main products and by-products for Cu/MgO 117
Table 4.3: Surface areas of unsupported copper catalysts 123
Table 4.4: Surface areas of supported copper catalysts 124
Table 4.5: BET and copper surface areas for Cu/alumina catalysts with additives 126

CHAPTERS:
Table 5.1: Summary of catalyst deactivation at different reaction temperatures 148
Table 5.2: Performance ofCU/Cr/Co on silica catalyst 149
Table 5.3: XRD and TOC results of tested catalysts 150
Table 5.4: Copper surface areas for 14.4 wt % Cu on silica catalysts 152
Table 5.5: Reaction rate parameters for ethanol dehydrogenation 159
Table 5.6: Kinetic model parameters for ethanol dehydrogenation 164
Table 5.7: Reaction rate parameters for 2-butanol dehydrogenation 168
Table 5.8: Kinetic model parameters for 2-butanol dehydrogenation 171

CHAPTER 6:
Table 6.1: Composition (per litre) of plating solutions for membrane plating 177
Table 6.2: Experimental H2 permeances vs. calculated values for membrane (N7) 199
Table 6.3: Experimental H2 permeances vs. calculated values for membrane (N4) 200

XXIX

",!,
CHAPTER 7:
Table 7.1: Alloying procedures used for Pd-Ag membranes 208
Table 7.2: N2 permeance and selectivity data for 8b, N1 b, N3b and N4b 211
Table 7.3: Ag-Pd diffusion coefficients 215
Table 7.4: Effect of heating on hydrogen permeances for membrane N3 219
Table 7.5: Nitrogen permeances of membranes (8c) and (N6) after heating at 600°C 223
Table 7.6: Description of different membrane stages (d l-d3) 224
Table 7.7: Hydrogen permeances at 410°C 231
Table 7.8: Hydrogen and nitrogen permeances at 410°C after oxidation 232

CHAPTER 8:
Table 8.1: Different liquid hourly space volumes (LHSV) used for
dehydrogenation reactions in a membrane reactor 237
Table 8.2: Surface reaction rates at different temperatures 251
Table 8.3: Ethanol reaction rate parameters 261
Table 8.4: Permeance data for N8x at 250, 275 and 300°C 261
Table 8.5: Parameters for solving ethanol dehydrogenation model 261
Table 8.6: Reaction rate coefficients for 2-butanol dehydrogenation 265
Table 8.7: Permeance data for N8x at 190, 215 and 240°C 265
Table 8.8: Parameters for solving 2-butanol dehydrogenation model 265
Table 8.9: Parameters for ethanol dehydrogenation at 275°C 266

CHAPTER 9:
Table 9.1: Model differences for ethanol conversion at 250 DC 281
Table 9.2: Improvements in total Xet for the membrane reactor at 250°C 282
Table 9.3: Model differences for ethanol conversion at 275 °C 285
Table 9.4: Improvements in total Xet for the membrane reactor at 275°C 285
Table 9.5: Model differences for ethanol conversion at 300 °C 287
Table 9.6: Improvements in total Xet for the membrane reactor at 300 DC 288
Table 9.7: Improvements in total X2But for the membrane reactor at 190°C 293
Table 9.8: Model differences for 2-butanol conversion at 190 DC 293

xxx
Table 9.9: Improvements in total X28ut for the membrane reactor at 215°C 296
Table 9.10: Model differences for 2-butanol conversion at 215°C 296
Table 9.11: Improvements in total X 2But for the membrane reactor at 240°C 299
Table 9.12: Model differences for 2-butanol conversion at 240°C 299

XXXI
 ACKNOWLEDGEMENTS

I want to give special thanks to the following people:

JG Keuler and SCJ Keuler (my parents),
C Nel (my grandmother),
All the other members of my family,
C Brown and her family,
D W Mouton for hours of assistance, and
For every friend that listened and/or gave advice.

For technical support I want to thank:

Prof. L Lorenzen,
Prof. R Sanderson,
Prof. J Dalmon,
Dr. S Miachon, and
Dr. B Jager.

I want to thank those who assisted me with analyses:

HBothaand
MWaldron

For financial support I want to thank:

Sasol,
FRD,and
CNRS

And most of all, I want to thank the LORD for the gift of life.

XXXll
 1. INTRODUCTION

In this study the possibility of dehydrogenating alcohols in a palladium-based membrane

reactor was investigated. Recent developments in membrane technology have opened the
way for performing conventional reactions in a new type of reactor called a catalytic
membrane reactor. This type of reactor offers the possibility of doing separation and
reaction simultaneously in a single process unit. The Pd membrane reactor offers several
theoretical advantages over the conventional plug flow reactor.
• In equilibrium restricted reactions, the equilibrium can be shifted, resulting in higher
conversions at the same operation temperatures.
• Selectivity is usually better for the membrane reactor than for the plug flow reactor.
• The reactions can be performed at lower temperatures, which will increase catalyst
life by reducing catalyst deactivation.
• Lower operating temperatures will reduce energy consumption, and finally,
• For selective Pd-based membranes, high purity hydrogen can be separated through
the membrane. Hydrogen is an important commodity on chemical plants.

Traditional dehydrogenation reactions have become less favourable due to high energy
costs and have been replaced by alkene oxidation processes to yield the same products.
An example is the oxidation of ethylene (Wacker process) to produce acetaldehyde. By
exploiting advantages offered by membranes, dehydrogenation in a catalytic membrane
reactor becomes a very competitive alternative to alkene oxidation. This process is more
selective and provides high purity separated hydrogen as a by-product.

The six main goals of this dissertation were:

• To optimise the performance of an alcohol dehydrogenation catalyst,
• To model the kinetics of an alcohol dehydrogenation reaction,
• To optimise the composition and thickness of Pd and Pd-alloy films for hydrogen
separation,

1
• To model the membrane separation process with a sweep gas and a pressure
differential,
• To compare the performance of a membrane reactor consisting of the optimised
catalyst and optimised membrane with a plug flow reactor, and
• To model the membrane reactor.

The dehydrogenation of ethanol to yield acetaldehyde was chosen as model reaction. A

model was developed for this reaction and the validity of the model was tested with the
dehydrogenation of 2-butanol to give 2-butanone, also known as methyl ethyl ketone
(MEK). Acetaldehyde is an important intermediate for producing other chemicals. The
dehydrogenation of 2-butanol is a well known industrial process and the main process
used for producing the solvent MEK. The latter is of great industrial importance.

A flow diagram for the project is shown in Figure 1.1. The project can be divided into
three sections: firstly, membrane development and modelling; secondly, some catalyst
development and modelling and thirdly, a final section in which catalysis and separation
are combined in the membrane reactor. For Pd plating (block 1), the composition of the
plating solution, the plating conditions and the solution feeding mechanism to the inside
of the membrane tube were investigated. The thickness of the Pd film was minimised to
obtain films with very high hydrogen permeances, while retaining good hydrogen to
nitrogen selectivities. Films were tested from 330°C to 450 °C (block 2) under positive
pressure and by using a sweep gas. Palladium thin films were used as supports for
depositing silver (block 3). Pd-Ag films were heat treated in different gas environments
and at different temperatures. The stability of the films was determined as a function of
time.

Catalysts were prepared (blocks 7 and 8) using the impregnation technique. The effects
of support type, copper loading, added stabilisers and catalyst particle size were studied
for the dehydrogenation of ethanol and 2-butanol. For each catalyst, experiments were
performed over a wide range of operating conditions and both the conversion and
selectivity towards the desired product were optimised. The optimised catalyst was then

2
used for determining the reaction kinetics of the dehydrogenation of ethanol and the
dehydrogenation of 2-butanol.

The optimised catalyst and optimised membrane were combined to form a catalytic
membrane reactor. Experiments in the membrane reactor were performed at different
temperatures, feed flow rates and sweep gas to feed molar ratios. The results were
compared to those obtained in a conventional plug flow reactor.

In this dissertation, the following contributions were made to existing work from other
researchers:
Catalyst development: The contribution made in this area was not large. The
deactivation of a copper-based catalyst during ethanol dehydrogenation was studied in
depth and a more accurate deactivation mechanism was developed. Reaction rate
mechanisms and kinetic parameters were determined for ethanol and 2-butanol
dehydrogenation with the optimised catalyst. Kinetic data was used for modelling
purposes.

Membrane development: Significant advances were made in this field. Much thinner,
yet very selective, Pd films were prepared on the inside of alumina membrane tubes with
a modified electroless plating technique. The Pd film tickness in this study was less than
half of the lowest thickness previously reported (see Appendix E) for the same
membrane configuration. Thin Pd-Ag films were successfully prepared on the inside of
membrane tubes with thickness less than 2.2 microns. The stability and alloying of Pd-
Ag membranes were investigated thoroughly, something that has not been done
previously. A method was developed for improving the hydrogen permeance through Pd-
Ag films.

Membrane reactor work: Significant advances were made in the dehydrogenation of

alcohols in a membrane reactor. Previously, ethanol dehydrogenation was studied by a
few groups in a membrane reactor under very limited conditions (see 9.1.4). The high
quality membranes prepared in this dissertation, made it 'possible to study the ethanol

3
dehydrogenation reaction at higher ethanol feed flow rates, yielding very good results. A
thorough investigation was conducted into 2-butanol dehydrogenation in a membrane
reactor. This, to our knowledge, has not been investigated previously. A full sensitivity
analysis on the membrane reactor model gave more insight into the effects of different
parameters on the dehydrogenation process.

4
Developing 1 Characterising Pd plated 2
Pd electroless
Plating process
membranes
330-450°C ~ 6
Membrane separation
modelling with positive
3 4 AP and a sweep gas
Developing Alloying and characterising
Ag electroless ~
Pd/Ag plated membranes V~ 15
Plating process 250-450°C 14
Performing catalytic Developing and
~ 5 membrane reactor testing cataly-
Testing stability of experiments with
~
. tic membrane
PdAg membranes optimised catalyst reactor model
and membrane
.~
j
16 Performing a
Developing suitable 7 Determine kinetic 11 sensitivity
catalyst for the dehy- / parameters for etha- analysis on
drogenation of ethanol noldehydrogenation the model
9 ...,.
~
Testing stability of Modelling of 13
optimised catalysts pure kinetic

~
data
8 , 10 ..------'
Developing suitable Further catalyst Determine kinetic 12
catalyst for the dehy- deactivation parameters for 2-buta
drogenation of 2-butanol studies noldehydrogenation

Figure 1.1: Project flow diagram

5
 2. BACKGROUND AND LITERATURE SURVEY

2.1. MEMBRANE SEPARATION PROCESSES

A membrane is any semi-permeable active or passive barrier, which separates particles,
molecules or any other species in a gaseous and/or liquid mixture (see Figure 2.1) when
certain driving forces are applied. The driven force that causes preferential movement of
species through the membrane may be a pressure difference, concentration difference or
voltage difference across the membrane .

•••
o· 0 • .0. •
.0

-
o •
o
0
••
- ..
.0
0 •
0
0 ••
.-0
••
00.0. ••
Figure 2.1: Separation by a semi-permeable membrane (right)

Membrane processes are classified according to the separation method and the size of the
separated species. Basic processes are micro filtration, ultrafiltration, reverse osmosis,
dialysis, electrodialysis and gas separation.

2.2. INORGANIC MEMBRANES

All membranes are either organic (polymeric) or inorganic. There are two types of
inorganic membranes: dense and porous. Dense membranes can further be classified into
either metal membranes or solid electrolyte membranes. They are prepared by different
methods and from different materials. The preparation method has a definite effect on
the pore structure. Membranes with a uniform pore structure across the thickness of the
membrane and made in a single step are called symmetric membranes. Single step
membranes with a changing structure throughout the thickness are asymmetric. When a

6
membrane consists of two or more layers, prepared in consecutive steps, it is called a
composite membrane. For composite membranes, the initial layer usually provides
mechanical strength and acts as a support on which further layers are deposited on. The
second layer and subsequent layers, determine the membrane's separation properties.

2.2.1. DENSE INORGANIC MEMBRANES

There are several classes of dense inorganic membranes. They are either oxide based,
metal based or a combination of the two. A new class of inorganic polymers has also
been studied in recent years.

2.2.1.1. Dense metal membranes

Dense metal membranes are mostly made from palladium and its alloys. Pd can be
alloyed with Ag, Ru, Rh, Ni or Au. Ag is, however, most frequently added to Pd to
prevent hydrogen embrittlement of pure Pd that occurs below 300 0c. Johnson Matthey
has used palladium-silver (77 wt %, 23 wt %) alloy membranes for hydrogen purification
since the early 1960s (Hsieh, 1996). Unit production is about 56 m% (Hsieh, 1996).
Union Carbide installed a Pd alloy membrane plant in the 1960s to separate hydrogen
from refinery off gas (McBride and McKinley, 1965). Membrane cost, durability and
poisoning by carbon and sulphur compounds have restricted large scale progress (Armor,
1989). Membrane permeance tends to be low due to thick alloy layers. Alternative
metals like niobium, tantalum and vanadium have been investigated for hydrogen
separation (Buxbaum and Kinney, 1996). Silver membranes have been used on a much
smaller scale to selectively separate oxygen from other gases.

2.2.1.2. Nonporous electrolyte membranes

Solid electrolytes are impervious to gases and liquids, but allow some ions to pass
through their lattices under an applied voltage difference or a chemical potential
difference. Calcium-stabilised zirconia allows for oxygen transport, while other gases
(!toh, 1990) cannot pass through. PbO selectively separates oxygen from other gases.
Other electrolytes under investigation (Hsieh, 1996) include simple or complex halides

7
(RbAg415), simple or complex oxides (~-aluminas) and oxide solid solutions (ZrOz-Y20],
Zr02-CaO, Th0 2-Y 20]).

Silica deposited on porous Vycor glass gives high hydrogen separation factors.
Hydrogen passes through defects in the silica network. Gavalas et al. (1989) obtained
Knudsen separation values for hydrogen and nitrogen at room temperature, but the value
increased to over 2000 at 450 DC. The main problem with silica deposited membranes is
that they have a very poor stability.

2.2.1.3. Dense inorganic polymer membranes

These types of membranes have been developed for separation and reaction at
intermediate temperatures (up to 200 DC for long periods of time). Organic membranes
cannot withstand such high temperatures. Polyphosphazenes (Hsieh, 1996) are
amorphous rubbery polymers which exhibit higher permeances but lower selectivities
than glassy polymer membranes. They are very selective in separating acidic (carbon
dioxide and hydrogen sulphide) and non-acidic gases (e.g. methane). They consist of
alternating phosphorous and nitrogen double and single bonds in a polymer network.
Polysilazanes, containing silicon and nitrogen bonds, is another class of organometallic
polymers that can be employed as membranes.

2.2.1.4. Dense metal composite membranes

This category of membranes has a dense metal substrate as support with some sort of
palladium modification. Dense Pd-porous stainless steel membranes thus fall outside this
group and are discussed separately. Refractory metals like vanadium, tantalum and
niobium have very high hydrogen permeances, they are cheap compared to palladium,
easy to fabricate into tubes and they are stronger than palladium. They are, however,
much more prone to hydrogen embrittlement. Niobium must operate above
420 DC and tantalum above 350 DC in hydrogen. Buxbaum et a1. (1993, 1996) have done
extensive research on refractory metals coated with palladium for hydrogen separation.
A palladium coating is necessary to reduce surface poisoning. They used commercial

8
niobium (150 Ilm thickness) and tantalum (75 Ilm thickness) tubes coated with palladium
by electro less plating. The membranes were very stable over time.

Peachey et al. (1996) tested composite palladium-tantalum membranes. Tantalum foils of

7.5, 10 and 12.5 !J.ffi were used with I Ilm Pd deposited on both sides by electron beam
evaporation. They found that by cleaning the tantalum foils prior to Pd deposition they
could obtain much higher hydrogen fluxes through the composite membranes. Cleaning
consisted of bombarding the Ta foil with argon ions (ion milling). Edlund (1993a,
1995a) made many multi-layered membranes based on vanadium as substrate. A typical
example was Pd-Si02-V-Si0 2-Pd (25-25-30-25-25 Ilm). The Si0 2 was compared to
many other oxides in the 5-layer membrane, but gave the best hydrogen penneance
results.

2.2.2. POROUS INORGANIC MEMBRANES

Advances and research in the field of porous inorganic membranes have been dramatic in
recent years (Hsieh, 1989; Soria, 1995). Industrial application of inorganic membranes
started during the post World War II period in the field of nuclear power. The uranium
isotope, 235U , was enriched from 1% to between 3 and 5% for fuel in nuclear reactors or
up to 90% for nuclear weapons. Since the 1940s, membranes have played an important
role in gaseous diffusion, with France, the United States and the Soviet Union leading the
way. More than 100 million porous membrane tubes have been made employing French
technology (Hsieh, 1996). Composite membranes have been in very high demand. Thin
separation layers allow for high fluxes, while the support gives mechanical strength. The
pores in porous membranes can be divided into three classes:
• Macroporous >50nm,
• Mesoporous 2 nm < pore size < 50 nm, and
• Microporous <2nm.

Some general membrane applications will be mentioned later (see Section 2.7).
2.2.2.1. Porous glass
Macroporous Vycor glass membranes became available in the 1940s. They are made by
acid leaching one of the phases in the glass. Currently these membranes can be prepared
with pores as small as 4 nm. Kameyama et al. (1981) claimed to produce porous Vycor
glass membranes with 86% of their pore diameters within 1 nm of the 4.5 nm average
pore diameter. The brittleness and loss of microstructure upon heating for long periods at
elevated temperatures (> 300°C), limit their application.

2.2.2.2. Porous metal

Porous silver membranes were commercialised in the 1960s, but their use has been
limited. Porous stainless steel membranes have been employed as high quality filters for
many years. Porous stainless steel membranes can be used as supports for preparing
composite membranes. The large pore size of these membranes and the possibility of
inter metal diffusion at higher temperatures require some substrate modification. This is
discussed in Section 2.2.3. Many researchers have used porous stainless steel as
SUbstrate to prepare high selectivity composite membranes (Edlund, 1996; Shu et aI.,
1996a; Nam et aI., 1999).

2.2.2.3. Molecular sieving membranes

Molecular sieving membranes have pore sizes ranging from 0.2 to 1 nm. Carbon
molecular sieves, silica molecular sieves and zeolites are the most widely known. Carbon
molecular sieves can separate molecules differing by as little as 0.02 nm in critical
dimensions (Hsieh, 1996). They are prepared by pyrolysis of the membrane material
between 500°C and 750 °C. The pyrolysis temperature and conditions determine the
pore size. These membranes are usually formed as hollow fibres (outer diameter between
5 microns and Imm). Haag and Tsikoyiannis (1991) prepared aluminophosphate
molecular sieve membranes and Raman and Brinker (1995) prepared molecular sieving
silica membranes.

10
2.2.2.4. Porous ceramic and composite membranes
Ceramics have several properties that make them the superior choice for inorganic
membranes. Ab03 remains stable up to 800°C without degradation of the pore structure,
(Kameyama et a!., 1983), it is resistant to corrosive environments, it is mechanically
stable and can withstand pressure drops of up to 1.5 MPa. Metals and oxides can easily
be dispersed on the membrane surface and into the pores to add catalytic properties. The
acidity of the support must be taken into account and modified if it catalyses undesirable
reactions.

Ceramics are mainly used as composite membranes, where several layers with decreasing
pore sizes are deposited on one another. The final or permselective layer is typically a
few microns thick and allows for high fluxes. A common example is one or more a-
alumina support layers with a final y-alumina separation layer, yielding a membrane with
4 to 5 nm pores. The top layer determines the characteristics (permeance and selectivity)
and the pore size of the membrane. Top layers that have been deposited and studied
include y-alumina, zirconia, titania, oxide mixtures, zeolites, silica, metals and metal
alloys. Each of these top layers will result in different pore sizes, with the aim being to
make the membrane very selective (very small pore sizes in the Angstrom range) and
allowing for a high flux to pass through the membrane (very thin selective layers in the
nanometer range).

2.2.2.5. Zeolite membranes

Zeolite membranes are composite membranes, where a thin zeolite layer is deposited on a
support (usually a-alumina with or without y-alumina modification). Jansen et a!. (1998)
prepared a specific zeolite structure (called a MFI-type zeolites) on steel, silicon and
quartz. This type of membrane has received much attention in the last decade, because
zeolites can separate molecules in the Angstrom range. Possible applications are for use
in isomerisation processes, hydrogen separation, water-alcohol separation and separation
of organic compounds.

II
Crystals can be grown parallel or perpendicular to the support. Ishikawa et al. (1989)
deposited zeolite coatings on porous Vycor glass and obtained water to butanol
separation factors of 2700. Suzuki (1987) prepared A, X, Y, L, FU-I, ZSM5 and
silicalite zeolites on porous Vycor glass, stainless steel, nickel and alumina. Jia et al.
(1993) prepared silicalite zeolite membranes in situ on a porous ceramic support, while
Sano et al. (1994) prepared a similar zeolite on stainless steel.

2.2.3. MEMBRANE MODIFICATION

Porous ceramic membranes are often multi-layered, with two or more a-alumina support
layers, as mentioned earlier. A y-alumina layer can be deposited on the a-alumina layers
to reduce the final pore size. With a smaller pore size, the final Pd layer can be made
thinner, but metal to ceramic adhesion can become poor.

In recent years, special attention has been given to porous stainless steel membranes.
Their larger pore size (0.2-0.5 J.Ul1), compared to ceramic membranes, and the wider pore
size distribution are, however, some drawbacks for depositing thin films on them. A
further problem is the diffusion of palladium or its alloy into the stainless steel and a
gradual decline in hydrogen flux upon use. Nam et al. (1999) modified a porous stainless
steel support with a nickel deposit to reduce pore size. Jemaa et al. (1996) used a shot
peening technique to reduce the pore size of stainless steel membranes. During shot
peening, iron particles of less than 125 ~m were fired onto the support. Jemaa et al.
(1996) obtained favourable results and observed a reduction in pore size.

To reduce inter metal diffusion, Gryaznov et al. (1993) introduced a diffusion barrier
between the stainless steel and the palladium layer. It was mainly iron that diffused into
the palladium. They achieved a constant hydrogen permeability for 1000h at 800°C with
an intermediate layer of tungsten (0.8 J.Ul1), tantalum oxide (0.1 J.Ul1), magnesia (0.5 J.Ul1)
and zirconia (1.0 ~m), each applied individually. Shu et al. (1996a) applied a 0.1 J.Ul1
titanium nitride layer onto porous stainless steel. Their Pd-AglTiN/stainless steel
membrane was thermally stable up to 700°C.

12
Edlund and McCarthy (1995b) observed a very rapid hydrogen flux decline in their
palladium-vanadium composite metal membranes at 700°C. Vanadium diffused into
palladium and the process was accelerated in a hydrogen atmosphere. They introduced a
250 !-1m thick porous alumina diffusion barrier between palladium and vanadium.
Thereafter, the hydrogen flux remained constant at 700 °C for the duration of the testing
time (76h).

2.3. MEMBRANE PREPARATION AND CHARACTERISATION

Several techniques for preparing dense and porous membranes are discussed by Hsieh
(1989), Hsieh (1996) and Keuler (1997a). These include:
• Conventional powder metallurgy methods such as pressing, extrusion and slip casting
followed by sintering,
• Anodic oxidation (Hsieh, 1989; Nourbakhsh et aI., 1989),
• Pyrolysis (Linkov, 1994; Soria, 1995),
• Phase separation and leaching (Soria, 1995), and
• Cold rolling and annealing.

When composite membranes are prepared, additional thin film formation techniques are
required to modify the membrane support. Thin film formation techniques will be
discussed in greater detail later.

Several structural characterisation techniques are discussed by Mulder (1991), Hsieh

(1996) and Keuler (1997a, 1999a). They include:
• Scanning electron microscopy (Hearle et aI., 1974),
• Transmission electron and atomic force microscopy,
• X-ray diffractometry (Cullity, 1978),
• Pore size measurement techniques, and
• Particle induced X-ray emission (Johansson et aI., 1988).

13
Membranes can also be characterised by the transport mechanism through the membrane
pores. In the next section, separation mechanisms, which are determined by pore size and
pore structure, will be discussed.

2.4. SEPARATION MECHANISMS

There is only one basic separation mechanism for flow through dense membranes and
this method is restricted to hydrogen and oxygen permeance. For porous membranes, the
flow mechanism can be viscous flow, Knudsen diffusion, capillary condensation, surface
diffusion or molecular sieving (see Figures 2.2 to 2.5, Noble and Stem, 1995). Viscous
flow does not yield any separation and thus is not relevant for membrane separation
processes. In most cases Knudsen diffusion dominates, but more than one flow
mechanism can occur simultaneously.

In the literature the following terms are used for flow through a membrane.
2
• Permeability, in mol.m1(m .s.Pa),
2
• Permeance, in mol/(m .s.Pa),
• Flux, permeation flux or permeation rate, in mol/(m2.s), and
• Flow rate, in mol/so

The selectivity of two components is the ratio between their permeation rate. The
separation factor (a) is a similar parameter and defined as:

(2.1)

with F the flow rate, x the high pressure side, y the low pressure side and i and j are the
components being separated. When the pressure drop across the membrane is small, back
diffusion will take place and it will reduce the separation factor. The separation factor
further depends on pore size distribution, temperature, and interaction between the gases
being separated and the membrane surfaces.

14
 r-r-" ~"'~ ~~ ""'Jr"''''' 1'~~'1
" 1 o
'.
...
~ ____
.• . . . .
'" , _ '" ~"~~ ~"-d
o

.0
.~.7 '.~.T. 0
o 0

Figure 2.2: Knudsen diffusion Figure 2.3: Surface diffusion

o
o 0
o
o
o
o o
o

Figure 2.4: Capillary condensation Figure 2.5: Molecular sieve separation

2.4.1. KNUDSEN DIFFUSION

When viscous flow dominates, molecules collide more with one another than with the
pore walls of the membrane (see Figure 2.2). The pore diameter is large compared to the
mean free path of the molecule and no separation can take place. By decreasing the pore
size, separation can occur when molecules collide more with the pore walls than with one
another. The flux (J) through a membrane of thickness I is (Noble and Stem, 1995):

J = GfS, t.Pi (2.2)

PJY1iRoT I

with G r the geometric factor accounting for porosity and tortuosity, t.P the pressure
difference across the membrane, M the molecular weight and Sc the Sievert's constant.
The separation factor for an equimolar gas mixture diffusing by Knudsen diffusion is the
square root ofthe ratio of the molar masses:

(2.3)

15
Separation by Knudsen diffusion is limited in membrane reactors, since a lot of the feed
is lost through the membrane's pores, which reduces the product yield. The best
separation is obtained for light components like hydrogen.

2.4.2. SURFACE DIFFUSION

Surface diffusion is more important at low temperatures. Surface diffusion is an
adsorption-dependant process (see Figure 2.3). It can occur in parallel with Knudsen
diffusion but at higher temperatures Knudsen diffusion dominates, as molecules desorb
from the surface. Molecules adsorb onto the pore wall and migrate along the surface of
the membrane pore. The permeability of the more strongly adsorbed molecule is
increased. Many examples of this phenomenon are documented in literature (Hwang and
Karrunermeyer, 1966; Kameyama et ai., 1979; Weaver and Metzner, 1966). Reported
permeability ratios increase between a factor of 1 to 2.5, compared to Knudsen diffusion
ratios, for examples encountered in the literature. Surface flow has been expressed
mathematically (Uhlhorn et al., 1989), but parameters must be determined
experimentally.

2.4.3. CAPILLARY CONDENSATION

Condensable vapour components in a mixture can condense in pores and block gas-phase
diffusion through it if the pores are small enough (see Figure 2.4). The condensate will
evaporate on the low pressure side of the membrane. The result is that the permeance of
other components will be slow and limited by their solubility in the condensable
component. In studies involving capillary condensation, separation factors of 80 for
C3H6 and N2 (Uhlhorn et al., 1990), 460 for water and air, 1000 for S02 and H2 (Keizer et
ai., 1982) and 120 for methanol and water (Hsieh et al" 1988), have been reported.

2.4.4. MOLECULAR SIEVE SEPARATION

Molecular sieve membranes allow for molecular sieve separation (see Figure 2.5). Pore
sizes are less than 1 nm and allow for diffusion of only very small molecules. Uhlhorn et
ai. (1990) used Si02-modified alumina membranes with pores of less than 1 nm to
separate hydrogen and propylene. Separation factors exceeding 100 were obtained at

16
200°C. In another study by Suzuki (1987), an equimolar ratio of methane, ethane and
propane at 15 bar yielded a mixture of73.5% methane, 26% ethane and 0.5% propane as
the permeate in a porous stainless steel membrane with a zeolite layer.

2.4.5. FLOW THROUGH NON-POROUS MEMBRANES

Hydrogen and oxygen transport through a non-porous membrane is illustrated in Figure
2.6. For hydrogen permeance there are several transport steps (Ward and Dao, 1999). A
mathematical description of each process has been described by Ward and Dao (1999).
These processes include:

1. Molecular transport from the bulk to the surface film layer.

2. Dissociative adsorption on the membrane surface.
3. Atomic hydrogen dissolves in the membrane.
4. Diffusion of hydrogen through the bulk membrane.
5. Transition from the bulk to the surface on the low pressure side.
6. Hydrogen atoms recombine to form molecules and desorb on the other side of the
membrane.
7. Gas transport from the membrane surface into the bulk gas.

YoO
11

2e"+O ..
,--- 01"
adsorption
1

Figure 2.6: Hydrogen and oxygen flow through a non-porous membrane

The permeation flux (1) can be expressed using Fick's law (Buxbaum and Kinney, 1996):

17
J = D (C;,] - C;,,)
(2.4)
I

The diffusivity (D) is an Arrhenius function:

(2.5)

The hydrogen surface concentration (C) is the product of the Sievert's constant (Sc) and
the hydrogen pressure ( PH, ):

(2.6)

When Sievert's law applies, n = Yo. The conditions for Sievert's law have been discussed
by Shu et al. (1991) and Ward and Dao (1999). In general, as films get thicker (above
10 flm) they approach Sievert's law and n = Yo. Diffusion becomes the rate limiting step
in hydrpgen permeation. For very thin films, in the order of a few microns, the value of n
approaches one. Hydrogen chemisorption on the palladium surface becomes the rate
limiting step (Nam et aI., 1999; Yan et aI., 1994). Surface poisoning, grain boundaries
and external mass transfer will cause further deviations from Sievert's law. The limiting
transport mechanism is very temperature dependent. Ward and Dao (1999) concluded the
following after an intensive investigation into hydrogen transport:
• Diffusion is likely to be rate limiting above 300°C, even for thin membranes
(approaching 1 flm).
• Desorption is likely to be rate limiting at lower temperatures.
• Adsorption is likely to be rate limiting for low hydrogen partial pressure and high
surface contamination.
• For thin films (much less than 10 flill), external mass transfer becomes important,
especially on the low pressure side.
• The membrane fabrication technique plays a significant role in permeation, which is
probably related to the microstructure.

18
Furthennore, the penneability (Per) expressed in mol.mJ(m 2.Pa.s) is defined as:

Per = Sc D 0 e-ED/R,T = Poe -ED/R,T (2.7)

The flux equation can now be expressed in tenns of pressure difference and penneability.
Substituting equations (2.5) and (2.6) into (2.4), and then (2.7) into the result gives:

(2.8)

And the penneance (Pm) in moll(m2 .Pa.s) is:

P = Pee (2.9)
m I

The hydrogen flux is very high through palladium and palladium alloys, mainly because
palladium has a high hydrogen solubility. Do and ED values for the different palladium
phases and at different temperatures have been given by Shu et ai. (1991).

Oxygen penneance through silver is similar to that of hydrogen through palladium. The
value of n can be taken as Y, (Gryaznov et ai., 1986a). Competitive adsorption by other
gases in a gas mixture on silver, reduces the oxygen penneability. For nonporous silica
glass, the activation energy for hydrogen penneance is significantly higher than for
palladium. For palladium it is in the order of 20-25 kJ/mol (Shu et aI., 1991) and for
silica about 35 kJ/mol (Gavalas et ai., 1989). The flux of oxygen through solid
electrolytes has been reported to be proportional to p\\ (Itoh, 1990).

19
2.5. PALLADIUM AND PALLADIUM ALLOYS
Palladium and certain palladium alloys have a high hydrogen permeance. The
characteristics and performance of palladium and palladium alloys have been studied for
many years.

2.5.1. PALLADIUM-HYDROGEN SYSTEM

An in-depth study on the palladium-hydrogen system was conducted by Lewis (1994).
The solubility characteristics of hydrogen in small palladium crystallites (nm range) are
different to those in bulk palladium (Boudart and Hwang, 1975). Structural changes for
palladium in hydrogen presented in this study is for bulk palladium or palladium films
and not palladium crystallites. At temperatures below 298°C and pressures below
2.0 MPa, the ~ phase of palladium will co-exist with the a phase in a hydrogen
atmosphere (see Figure 2.7 from Shu et aI., 1991). There is a considerable difference in
lattice expansion of the two phases, for example a hydrogen to palladium ratio of 0.5
results in an expansion of about 10% in volume (Knapton, 1977). Severe strains are
induced by the nucleation and growth of the ~phase in thea phase matrix.

"
a.
::;:
.
:I:
a.
'"

Figure 2.7: Equilibrium solubility isotherms ofPdHn for bulk Pd (Shu et aI., 1991)

20
De Ninno et a!. (1997) discussed the stress fields that are created when hydrogen
dissolves in palladium. The results were hardening, embrittlement and distortion of the
film, which led to cracks in the membrane after a few hydrogenation-dehydrogenation
cycles. To avoid these negative effects, the palladium must be kept in the a phase above
300°C at all time. Lewis (1994) studied the irreversible effects that took place near the
phase transition in a hydrogen atmosphere.

Alternatively, the palladium can be alloyed to suppress a to ~ phase transitions and avoid
distortion. The permeability of the alloy should be comparable to or better than that of
the pure palladium, have high mechanical strength and be resistant to poisoning.

Yan et a!. (1994) and Aoki et a!. (1996), performed temperature cycling tests on
palladium. Thin Pd films « 1 J.lm) prepared by chemical vapour deposition remained
stable for many temperature cycles between 100°C and 300 °C.

2.5.2. PALLADIUM-SIL VER SYSTEMS

Hunter (1956, 1960) discovered that silver addition to palladium improved dimensional
stability. The addition of about 20% silver depressed the a to ~ phase transition to below
room temperature. Further studies (Makrides, 1964; Jewett and Makrides, 1965; Holleck,
1970; Knapton, 1977) determined hydrogen diffusion coefficients and hydrogen
solubility coefficients in Pd-Ag systems. The addition of silver reduces the diffusion
coefficient. There is, however, a very sharp optimum in hydrogen solubility when the
silver content increases. The net result is an optimum hydrogen permeability (P er=SD H)
at a Ag percentage of 23 wt %. This permeability value is about 1.7 times higher than
that of pure palladium. The Pd-Ag (77 wt %, 23 wt %) alloy was used for commercial
hydrogen separation by Johnson Matthey. Several researchers have used Pd-Ag films to
study the performance of catalytic membrane reactors (Sathe et a!., 1994; Clayson and
Howard, 1987; Shu et a!., 1994, Uemiya et a!., 1991a). Ali et al. (1994a) investigated the
deactivation and regeneration of palladium-silver membranes.

21
The morphological changes of the Pd-Ag system upon hydrogen permeation was
described by Shu et al., 1997a. They used SEM images to study changes in a 50 Ilm Pd-
Ag (75:25 wt %) foil before and after hydrogen permeation. Several defects such as
trans granular cracks, intergranular cracks and hydrogen blisters were identified under
different conditions. Lattice strains caused minor cracks, which later developed into
large trans granular cracks.

2.5.3. PALLADIUM AND COPPER OR GOLD

Knapton (1977) reported that gold-containing alloys show higher resistance to sulphur
poisoning that silver-containing alloys do. Weyten et a!. (1997) reported that Pd-Ag
membranes have a low chemical resistance for H2S. H 2S gradually destroys the Pd-Ag
alloy film. The problem can be solved to an extent, by covering the Pd-Ag film with a
thin layer of gold to improve their resistance to sulphur. Maestas and Flanagan (1973)
studied hydrogen diffusion rates in palladium-gold alloys. Up to 20 wt % gold, the
diffusion rate was relatively unchanged from that of pure palladium. For a higher gold
content, the diffusion rate decreased logarithmically with gold increase.

For copper-containing alloys, there is an optimum copper concentration for maximum

hydrogen permeance. This maximum permeance occurs at 40% copper, which is
associated with the formation of an ordered P-CuPd phase. This alloy has a higher
hydrogen diffusivity than pure palladium, but the low hydrogen solubility results in
weaker permeances than for pure palladium. Uemiya et a!., (199Ib) compared a Pd-Cu
(94:6 wt %) and Pd-Ag (94:6 wt %) membrane and found that the first one had a much
lower hydrogen permeance.

2.5.4. PALLADIUM AND RARE EARTH ELEMENTS, NICKEL OR OTHERS

Knapton (1977) also studied palladium-cerium and palladium-yttrium systems. Both
showed high permeance values. Cerium is prone to poisoning by oxidation, but not
yttrium. Figure 2.8 (Shu et a!., 1991) show permeance data for various palladium alloy
systems. Binary alloys ofPd with B (McKinley, 1967), Ni (McKinley, 1967, Roshan et
a!., 1983), Rh (Rodina et aI., 1968), Ru (Armor, 1995), Cr, V and Fe (see Shu et a!.,

22
1991) have been investigated as well as ternary alloys ofPd-Ag with Ni, Rh, Pt and Au
(Knapton, 1977).

6
5.5
5 I
... Y

U!
~
4.5 I
E 4 I

~" 3.5 I
.l!. 3 :
c:
.,
0

.'" 2.5
2
.'
I
Ce
.... -.)." .. .' .
.-' .....AD
"
,
..
E
Q.
1.5
;,~:,"." \
,, ,
Au
". .. ....
Cu
,,.,
\
1
" .. ' - (

'.
0.5
0
'
--- -- -- ...... " , ~ ~

~ --
0 10 20 30 40 50 60
% of other component

Figure 2.8: Permeability ofH2 through Pd alloy membranes at 350°C and 2.2 Mpa

The addition of elements to palladium not only suppresses the phase transition, but can
also improve strength. Rodina et aI. (1968) tested the strength implications of the
addition ofNi and Au to palladium and found Ni to be very effective. The improvements
in catalytic properties of palladium with Ni (Smirnov et aI., 1978; Bulenkova et aI.,
1978), Ru (Gryaznov, 1986b; Skakunova et aI., 1988), Rh (parfenova et aI., 1983;
Gryaznov et aI., 1986c), Sn, Sb, Cu and Mo have been studied.

2.6. PREPARlNG PALLADIUM MEMBRANES

Initial work on palladium membranes used foils, typically 50 flm or thicker. The
advances made in preparing inorganic membranes have shifted research away from foils
towards composite membranes with much thinner palladium layers. Not only is this
cheaper, but it also allows for a large increase in hydrogen flux through the film.
Composite palladium membranes are prepared by depositing palladium or palladium
alloys on a multi-layer inorganic membrane support. Several thin film deposition
techniques have been developed to deposit thin palladium layers with minimum defects.

23
1"'"

Prior to any deposition, the membrane support needs to be thoroughly cleaned. Keuler
(1997a) discussed substrate cleaning. Different cleaning methods for porous glass,
porous stainless steel and porous alumina were presented. Different thin film deposition
techniques will be discussed briefly in the next few paragraphs. More detail on preparing
thin films can be found in Keuler (1997a).

2.6.1. WET IMPREGNATION

This technique is not suitable for preparing dense metal layers on inorganic supports.
The metal is deposited in the pores of the membrane and often these membranes are used
as contactors. Champagnie et al. (1992) prepared Pt-impregnated membranes and
Cannon et al. (1992) prepared Pd-impregnated porous Vycor glass membranes. The
change in membrane pore size was minimal. Porosity control and uniform impregnation
were some problems encountered. The metal served as a catalyst and not as a separator.
Uzio et al. (1993) found that the membrane permeability was not changed after
depositing Pt through ion exchange on an alumina membrane (Societe des Ceramiques
Techniques or SCT multi-layer membrane with a 4 nm y-alumina toplayer). For the
membranes tested by Uzio et al. (1993), the diffusion mechanism remained Knudsen
diffusion after depositing Pt on an alumina membrane. Jun and Lee (1999) used a
Pd(C3H3)(CsHs) organometallic precursor to impregnate a porous stainless steel
membrane modified with sintered nickel. They claim to have obtained a hydrogen to
nitrogen selectivity of 1600 after annealing the Pd composite membrane. Kizling et al.
(1997) used a similar pore plugging technique to prepare Pd-Ag (77:23 wt %) on alumina
membranes, but found it difficult to remove the surfactants after deposition.

2.6.2. SOL GEL PROCESS

Livage et al. (1988), Brinker and Scherer (1990) and Julbe and Guizard (1993) described
sol gel chemistry. Particles of a few nanometer in size can be made and deposited on a
support membrane. The gel is applied onto the membrane by slip casting. The organic
components in the gel are burned off during the firing stage, and the result is an inorganic
membrane support with a metal or oxide layer deposited on it.

24
 Several attempts have been made to prepare Pd or Pt composite membranes by the sol gel
method. Hongbin et al. (1995) prepared Pdly-alimuna membranes of which the pore size
varied between 5.5 and 6.5 nm for the different Pd concentrations employed, indicating
that there was little success in reducing the pore size. Vitulli et al. (1995) prepared a
PtlSi0 2 layer by the sol gel process on a SCT multi-layer alumina support. The resultant
membrane showed Knudsen
, diffusion properties with a hydrogen to nitrogen selectivity
. ofless than 3. Zhao (1997) was able to obtain a hydrogen to nitrogen selectivity of23 for
his composite membrane. He used an AlOOH sol to prepare Pdly-alumina membranes.

2.6.3. VAPOUR DEPOSITION TECHNIQUES

There are two basic types of vapour deposition techniques for preparing thin films:
chemical and physical vapour deposition. Physical vapour deposition can be either
through evaporation or sputtering.

2.6.3.1. Physical vapour deposition

With physical vapour deposition, complex alloys can be prepared, the deposition rate can
be accurately controlled and very thin films « 1 J.lll1) can be prepared, but deposition on
non-flat surfaces poses problems. Chemical vapour deposition can be performed inside
tubes, but there is a large loss of vapour through the membrane in the initial stages
(Morooka et aI., 1995).

Evaporation can be performed with resistive heating, but it is far less common than
sputtering. During sputtering, atoms from the target are dislodged through ion
bombardment by an inert gas and deposited on the substrate. Argon is most frequently
used. A magnetron sputtering setup is depicted in Xomeritakis and Lin (1997) and
sputtering conditions for palladium (Xomeritakis and Lin, 1997) and palladium-silver
(Gobina and Hughes, 1996) have been reported. Key parameters during sputtering are
the sputtering time, plasma power, substrate temperature and target to substrate distance.
Jayaraman et al. (1995a, b) investigated the effects of some of those parameters on the
quality and permeance of the sputtered film. Gryaznov et a1. (1993) prepared complex

25
alloys of palladium and one or more ofRu, Co, Pb, Mn and In on porous metal discs. All
examples encountered in literature for palladium sputtering on porous supports used discs
as the substrate.

2.6.3.2. Spray pyrolysis

Spray pyrolysis is similar to sputtering. Li ZY et al. (1993) rotated capillary membranes
in a high temperature flame to deposit palladium and silver on the outer surface of the
membrane. Palladium and silver nitrate were atomised and the aerosol fed with oxygen
to a hydrogen-oxygen flame. The metal condensed on the membrane to form a metal
layer.

2.6.3.3. Chemical vapour deposition

For chemical vapour deposition (CVD), a metal salt is heated and deposited on the
substrate. Palladium acetate is commonly used (Morooka et aI., 1995; Yan et ai., 1994)
as the metal salt. CVD reactors are described by Aoki et aI. (1996) and Xomeritakis and
Lin (1996). Typical sublimation conditions for palladium acetate are temperatures
between 400°C and 500 °C and a reduced pressure in an argon atmosphere. Palladium
chloride can also be used for CVD. The detailed experimental conditions have been
given by Xomeritakis and Lin (1996). PdCl2 was reduced with hydrogen. The reduced
pressure was applied on the one side of the tube and the layer deposited on the other side
of the tube. There was some deposition of palladium inside the pores. Layers prepared
by CVD are typically thicker than those prepared by sputtering. Some examples are:
• 4!-lm Pd on porous a-alumina, with 150 nm pores (Huang et aI., 1997; Yan et ai.,
1994),
• 4.4!-lm Pd on porous a-alumina ,with 150 nm pores (Morooka et aI., 1995; Aoki et
ai., 1996), and
• 1 I!ffi Pd on porous a-aluminaly-alumina, with 5 nm pores (Xomeritakis and Lin,
1996).

26
2.6.4. ELECTROPLATING
Metals and alloys can be plated on a conducting substrate that acts as a cathode.
Ceramics and plastics need to be treated before they can be electroplated. The metal
cations are suspended in solution and reduced by an external current passing through the
electrolyte. The cation concentration, bath temperature and current density determine the
deposition rate. Even deposition on large surfaces is difficult due to a variance in current
density and a declining metal ion concentration in the plating bath. Kikuchi (1988)
electroplated Pd-Cu alloys on a porous support. The Pd content was varied between
71 and 94 wt % and the Cu content between 6 and 29 wt %. A method for plating Pd and
its alloys on porous supports was also described by Itoh and Govind (I 989a).

In a more recent study, Nam et al. (1999) used a vacuum electroplating technique to
deposit palladium on a modified porous stainless steel support. A submicron Ni layer
was dispersed on the surface of the porous stainless steel support (0.5 f.tlll pore size)
under low vacuum and then sintered at 800°C for 5h under high vacuum. A thin copper
layer was deposited on the Ni and finally a Pd layer was electroplated on the copper
under vacuum. The resultant film was about 1 f.tlll thick with 78 wt % Pd and 22 wt %
Ni. Hydrogen to nitrogen selectivities varied between 500 and 5000 at temperatures over
350°C.

2.6.5. ELECTROLESS PLATING

Electroless plating is an autocatalytic oxidation-reduction reaction in which metal ions
are reduced and deposited as metal atoms. It is similar to electroplating, but no external
current is supplied. A detailed discussion can be found in Keuler (1997a). It can be
applied onto any material that has been properly pretreated. Some materials that have
been electroless plated are porous Vycor glass (Yeung et aI., 1995a; Uemiya et aI.,
1991b), porous stainless steel (Shu et aI., 1993) and porous alumina (Collins and Way,
1993a; Yeung and Varma 1995b). The main advantages and disadvantages of electroless
plating are listed in Keuler (1997a). The advantages of this process can be summarised
as:
• The technique is quick, simple and inexpensive.

27
• Dense, non-porous films of even thickness can be prepared on any shape.
• There is good metal to ceramic adhesion.

The main disadvantages are:

• Impurities might form in the metal layer when using certain reducing agents. Using
sodium hypophosphite as reducing agent causes some phosphor deposition
(Loweheim, 1974, reported a 1.5% phosphor deposition) and using boronhydride
results'in a 3-8% boron deposit (Shipley, 1984). Hydrazine gives very pure deposits,
but the deposition rates ten~ to be slow (Athavale and Totlani, 1989).
• Thickness control is difficult and costly losses might occur due to the decomposition
of the plating solution (Shu et aI., 1991).
• Today, co-deposition with other metals to form alloys has not been very successful.
Deposition of separate metal layers and subsequent alloying has also proven to be
very difficult.

2.6.5.1. Substrate pretreatment

As mentioned earlier (see Section 2.6) the substrate needs to be thoroughly cleaned
before any thin film deposition technique can be successfully applied. For electro less
plating on non-conducting surfaces (ceramics and plastics), the surface needs to be
activated prior to plating. There are two procedures for catalysing the surface to be
plated (Feldstein N, 1974). Both processes employ palladium and tin salts. In the older
process the substrate is first placed in a tin chloride solution (sensitising step) and then in
a palladium salt solution (activation step). For the exchange process, a colloidal solution
containing both palladium and tin salts is required. The substrate is placed in this
solution and then in an acid solution containing HCI or H2S04 (Cohen and Meek, 1976).

Palladium ions are reduced and Pd nuclei are deposited on the substrate. Models for
nuclei growth on the substrate have been developed by Cohen et al. (1971, 1973).
Several pretreatment solutions are listed (Osaka and Takematsu, 1980) in literature and
have been tested and evaluated (Horkans, 1983). The two step process deposits more

28
metal than the exchange process does and there is a higher Pd content (Horkans, 1983) in
the deposit. This is favourable for preparing high purity deposits.

2.6.5.2. Electroless plating solution composition

An electroless plating solution has a few basic components:
• a metal salt of the required metal than needs to be deposited,
• a reducing agent,
• a pH regulator, and
• a stabiliser that forms a complex with the metal ions and allows for a slower metal
release from the solution.

Not all metals can be electroless plated, but metals that form good hydrogenation-
dehydrogenation catalysts can be plated. A universal plating mechanism is described by
Van den Meerakker (1981). Ethylene di-amine tetra acetate (EDTA) is most commonly
used as stabiliser with hydrazine or sodium hypophosphite as the reducing agent. Ohno
et al. (1985) lists five reducing agents that can be used for various metal depositions. The
amine complex of palladium is used for electro less plating: (NH3)4PdX, with X = Ch or
N03 .

Rhoda (1959) and Athavale and Totiani (1989) reported the chemical reactions involved
in electroless palladium plating based on hydrazine and sodium hypophosphite as
reducing agents, respectively. Various factors need to be taken into consideration to
ensure a stable plating bath, an even thickness coating and an adequate plating rate.
These are:
• bath temperature,
• solution pH,
• Pd ion concentration,
• reducing agent concentration and stabiliser concentration, and
• solution volume to plated area ratio.

29
Furthermore, different plating characteristics are observed when plating the inside and the
outside of porous tubes. Pearlstein and Weightman (1969) studied some of the plating
variables. Keuler et a!. (1997b) investigated the interaction between various plating
variables and their effect on solution stability.

2.6.5.3. Recent advances in electroless palladium plating

With the conventional electroless plating technique, membrane selectivity drops fast
when palladium films are thinner than 5 ).tm. The deposit tends to be colurnnlike and
defects are present in the thinner films. Research has focussed on trying to make films
thinner, yet maintaining high selectivity.

Yeung et a!. (1995a,b) studied the application of osmotic pressure during electroless
plating. They used porous Vycor membranes as well as aly-alumina membranes as
supports. They found that modifying the plating solution with between 1 and 2 ml
formaldehyde per litre made the films denser, with smaller grain sizes and higher scratch
resistance. A higher formaldehyde concentration inhibited plating, while a lower one
showed no improvement. The plating solution was pumped through the membrane tube
with air, distilled water, sucrose solution or CaCh solution on the other side of the
membrane. By having plating solution on the inside of the tube and either sucrose
solution or CaCh solution on the other side, an osmotic pressure was created from the
inside to the outside. The pressure strength was dependent on the solute concentration.
Yeung and co-workers found that the osmotic pressure made the Pd coatings more dense,
nonporous, thinner and with a smoother surface morphology.

Li et a!. (1997, 1999) used a similar approach to repair defects in their electro less plated
Pd films. Porous stainless steel (0.1 ).tID pore size) and a-alumina (0.16 ).tm pore size)
membranes were used as supports. An initial Pd coating was applied and then one or
more coatings were added under osmotic pressure with NaCI as solute. This resulted in
film densification and defect minimisation. Table 2.1 (Li et a!., 1999) shows how the
selectivity of the Pd film improved after it was repaired.

30
Zhao et al. (1998) used a different pretreatment process to the traditional Pd/Sn activation
and sensitisation process. The porous alumina substrate was activated by a Pd(II)
modified boehmite sol. The gel-coated substrate was dried, calcinated at 600°C and then
reduced in hydrogen at 500 °C. Electroless plating was performed on those activated
substrates which they claimed had a smoother surface and more uniform distribution than
those prepared by conventional pretreatment. After using very high hydrazine
concentrations, they observed that the electro less Pd coating consisted of much finer
particles and this resulted in a more compact film.

Table 2.1: The effect of repairing of electroless Pd plated coatings (Li et aI., 1999)
Plating Film thickness (Jlm) Total Hz permeance HzIN z ratio
Jlmollmz.s.Pa
Original 7.6 6.24 10.3
I" Repair 9.2 3.24 96.5
2"u Repair 10.3 2.68 970

Paglieri et a1. (1999) investigated the effects of pretreatment on hydrogen permeance for
coatings prepared by electroless plating. They tested tin sensitisers of different
concentration as well as a new approach based of palladium acetate without any tin.
They found that high tin concentrations deposited in the pretreatment step led to poor
membrane stability at temperatures exceeding 500°C. Hydrogen to nitrogen selectivity
declined quite rapidly with time. One possible explanation is that tin, with its low
melting point (232°C), could be enhancing metal diffusion along the grain boundaries.
This creates wider channels for diffusion of gases. A pretreatment method based on
palladium acetate gave a better high temperature stability. Pretreatment consisted of
dipping the membrane in a Pd acetate, chloroform solution, drying the membrane by
heating it up to 500°C and reducing it in hydrogen at 500 °C before cooling it down to
room temperature.

31
2.6.5.4. Electro less palladium-silver coatings
Palladium-silver coatings can either be made by consecutive palladium and silver coating
steps or by a single co-deposition technique. Some attempts have been made on Pd-Ag
co-deposition (Shu et aI., 1993; Yeung and Varma, 1995b). The co-deposition process is
very complex and the success has been limited. Silver is much less stable than palladium
in solution, thus very dilute solutions have to be used. Furthermore, silver passivates
palladium deposition, but not vice versa. During the plating process, more and more
silver will be deposited with the net result that the final coating composition varies
significantly from the initial plating solution composition (Pd to Ag ratio). Shu et al.
(1993) investigated the electrochemistry of palladium-silver co-deposition, but many
questions remained unanswered.

An easier approach is to deposit palladium and silver separately. Keuler (1997a) has
given the composition of a typical silver plating bath. The textural differences after
changing the deposition order (pd on Ag or Ag on Pd) were discussed by Keuler et al.
(1999a, b). Results indicated that better metal to ceramic adhesion could be obtained by
depositing silver on palladium rather that palladium on silver. Concentration profiles
across the thickness of the Pd-Ag films, after initial alloying attempts, were also
discussed.

Cheng and Yeung (1999) attempted to model palladium, silver and palladium-silver
electroless plating on porous Vycor tubes (5 nm pores). Separate palladium and silver
plating was well modelled by their equations. Only the metal ion concentration and
hydrazine concentration were, however, treated as variables. Ammonia and EDTA
concentrations and the temperature were fixed. They were unable to properly describe
palladium-silver co-deposition. In general, higher temperatures, lower ammonia
concentrations and higher hydrazine concentrations favoured palladium deposition from a
palladium-silver plating bath. These conditions may result in poor bath stability. Silver
inhibited palladium plating and influenced the film microstructure.

32
2.6.6. PALLADIUM MEMBRANE TEMPERATURE STABILITY
There has been some work done on the long term stability of metal composite membranes
(Buxbaum and Kinney, 1996), where palladium is coated onto other refractory metals.
Very little has been published on the long term stability of palladium or palladium alloys
deposited on porous supports and on alloying procedures.

Paglieri et aI. (1999) recently studied the high temperature stability ofPd composite films
prepared by electro less plating. Plating was performed on the inside of a 200 nm
a-alumina support. At temperatures of 550°C and above, the membranes failed after a
few days and the separation factors dropped to the Knudsen level. Removing tin from the
pretreatment procedure in electroless plating reduced the problem of membrane failure,
but substantial selectivity decline still occurred above 550°C. At 450 °C and 500 DC, the
membranes remained fairly stable for a number of weeks, and the time of stability
depended on the Pd film thickness. It was found that the amount of time to fail was
proportional to the Pd film thickness and that the same failing mechanism prevailed for
ail thicknesses. Possible reasons for failing were:
• Impurities might be trapped at the Pd-alumina interface during pretreatment and
plating, which later result in pore formation.
• Differences in thermal expansion ofPd and alumina can cause cracking.
• Residual porosity in the Pd film can transform into pores.

2.6. 7. PALLADIUM-SILVER ALLOYING

Both Keuler (1997a) and Shu et al. (1996a) annealed their silver on palladium coatings in
hydrogen, for 5h at 650 DC and above 500°C, respectively. Keuler (1999a, b) used
alumina-zirconia supports and Shu et al. (1996a) used porous stainless steel supports with
a diffusion barrier. In both cases, silver diffused into the palladium matrix. When silver
was deposited on palladium, an annealing time of 5h did not produce a constant Pd:Ag
ratio across the thickness of the film. The film was asymmetric with a high silver content
at the film's outer edge and a low silver content near the support. The alloy composition
was not uniform. Unfortunately, neither Keuler (1997a) nor Shu et al. (1996a) included
any permeance data comparing coatings before and after annealing.

33
Uemiya et aI., (199Ia) coated the outside of a 200 run porous a-alumina membrane with
palladium and silver. They heat treated it at 900°C in argon for 12h and observed the
formation of a Pd-Ag alloy. Hydrogen permeance was 2.7 times higher after heat
treatment than before heat treatment. They reported no tearing or loosening of the alloy
film from the support. Kikuchi (1995) used the same process for heat treatin& their Pd-
Ag coating on the outside of a porous alumina tube. In both instances the heating process
was only vaguely described; no heating and cooling rates were supplied and no long term
stability test results were reported.

No data could be found on alloying processes for coatings on the inside of porous
ceramic tubes.

2.6.8. DEACTIVATION OF PALLADIUM MEMBRANES

Palladium and palladium alloy membranes perform well when exposed to only pure
hydrogen. The presence of other gases may severely impair hydrogen transport through
the membrane. Although this field has not been extensively studied, some investigators
have reported important findings. McBride and McKinley (1965) studied the effects of
about 50% CO, HzS, CH4 and CZH4 in hydrogen. They reported that all gases showed
some decrease in hydrogen permeance, with HzS giving the worst result. They concluded
that at lower temperatures, molecules adsorb on palladium to decrease the sites available
for hydrogen adsorption. At high temperatures a thin contaminant layer (coking) may
form on the palladium. Antoniazzi et al. (1989) studied membrane deactivation caused
by HzS. They concluded that HzS poisoning was irreversible and that the reduction in
hydrogen permeance through the Pd foil fell by about I % for every ppm H2 S present in
the feed.

A number of studies have focussed on the effects of carbon monoxide on hydrogen

permeance (Yoshida et al. 1983; Chabot et al. 1988; Sakamoto et al. 1996) through
palladium and palladium alloys. The general conclusion was that the operating
temperature of the membrane in the presence of CO should be above 350°C, to prevent

34
CO adsorption and loss of hydrogen flux. Jorgensen et al. (1997) reported that CO
adsorption could be used favourably during Pd membrane start-up and shutdown. By
covering the Pd surface with CO at low temperatures (during Pd membrane start-up and
shutdown), the solution of hydrogen in Pd was substantially reduced and the risk of
hydrogen embrittlement restricted.

Jung et al. (2000) studied hydrogen permeance through palladium in the presence of
steam, methane, propane and propylene. Propane and methane had a negligible effect on
the hydrogen flux through the palladium film. Propylene caused severe flux decline,
which dropped further with time. A carbonaceous layer was formed on the Pd due to the
dehydrogenation of propylene. Steam had both a positive and a negative effect. Steam
adsorbed strongly on palladium to decrease the available surface for hydrogen adsorption
and thus the hydrogen flux through the film. On the other hand, steam volatilised carbon
species on the palladium surface to reduce coking and improve the hydrogen flux. The
findings by Li et al. (2000) on palladium deactivation by steam and CO were in line with
those of other studies. Steam adsorbed more strongly on palladium than CO did and
caused greater reduction in hydrogen permeance than CO did.

2.7. APPLICATIONS OF INORGANIC MEMBRANES

There are many advantages and disadvantage of using inorganic membranes for
separation and reaction. Keuler (1997a) gave an extensive list of these advantages and
disadvantages. More advantages and disadvantage were given by Armor (1995).

Inorganic membrane technology has been fully commercialised and is used in many
different industries. These include food processing, processing of beverages, drinking
water, biotechnology, pharmaceuticals, waste oil treatment and petrochemical processing.
Hsieh (1989) and Hsieh (1996) discussed these and many other specialised applications.
Pd alloy metal membranes are commercially used as hydrogen purifiers (Hsiung et aI.,
1999). The presen~ study will concentrate on a class of modified inorganic membrane

35
called catalytic membranes, and more specifically on palladium-based catalytic
membranes.

2.8. CATALYTIC MEMBRANE REACTORS

A catalytic membrane reactor combines the functions of a membrane and a catalyst in a
single unit. Reaction and sep.aration can thus be performed in one step. Commercial
applications are still limited and most of the work in this field is of a fundamental nature.
Membrane tubes with an inside diameter of larger than 6 mm are mostly used. Capillary
membranes (outer diameter = 1 to 3 mm) and hollow fibres (outer diameter < 1 mm) have
been successfully produced, but not widely studied in catalytic membrane reactors due to
mechanical problems.

The catalyst can either be deposited into the membrane pores (Cannon et aI., 1992) or the
membrane can be packed (Tsotsis et aI., 1992; Yeung et aI., 1994) with a catalyst. A
third alternative is to deposit the catalyst only on the inner or outer membrane· surface
next to, or as part of, the separation layer. In the latter case the catalytic surface area is
very small and not effective unless the catalyst is on the inside of a membrane with a very
small inner diameter (hollow fibre). Membrane tubes need to be packed to provide
sufficient catalyst surface area. There are two flow possibilities. One or more reactants
can enter the membrane reactor on the same side (either both shell side or both tube side)
and one or more of the products are separated by the membrane. The second alternative
is to feed reactants into both the shell side and the tube side of the reactor. One of the
reactants moves through the membrane, which acts as a distributor, to react with the other
one(s) on the opposite side of the membrane. This is often used in hydrogenation and
oxidation reactions and in such a case the catalyst is usually in the membrane pores.

To create a driving force for the components being separated by the membrane, either a
sweep gas or a pressure difference is used. The sweep gas enters the shell and tube
reactor on the opposite side of the membrane than the reactant( s), and the sweep gas can
be inert or active. In the case of an active sweep gas, the sweep gas will react with the

36
component permeating through the membrane, for example a oxygen sweep (Itoh and
Govind, I 989b) in dehydrogenation reactions. Furthermore, the sweep gas can be co- or
counter current. The reactor can be adiabatic or isothermal. The effects of different flow
patterns and reactor configurations on reaction conversions have been well documented
(Shu et aI., 1991; Itoh, 1995a; Ross and Xue, 1995).

2.S.1. GENERAL ADVANTAGES OF CATALYTIC MEMBRANE REACTORS

Noble and Stem (1995) have proposed several applications in which catalytic membranes
have a potential edge over other processes.
• The equilibrium conversion of an equilibrium-restricted reaction can be exceeded by
removing one of the reaction products. This allows for higher conversions at the
same operating temperature. By performing the reaction at slightly lower
temperatures, the reaction selectivity and catalyst life can be increased and the energy
input reduced.
• In consecutive reactions the selectivity towards intermediate products can be
increased (Agarwalla and Lund, 1992). Undesirable products can be removed or
decomposed before they poison the catalyst.
• When one of the reactants is fed through the membrane in a controlled way, the
concentration can be manipulated to limit side reactions. Thermal runaway can be
eliminated in highly exothermic reactions.
• Feeding a reactant through the membrane can yield a higher concentration of that
reactant on the catalyst surface and improve mass transfer.
• Separation and catalysis can be performed in a single unit, which might be cheaper
than two separate units.
• Exothermic and endothermic reactions can take place at the same time on opposite
sides of the membrane. This is favourable from an energy point of view and only
requires one reactor instead 0 f two.

2.S.2. CATALYTIC MEMBRANE APPLICATIONS

There are two main areas of catalytic membrane applications. They can either be used
for equilibrium-restricted reactions, to shift the equilibrium, or as contactors for better

37
feeding control. The latter is mainly used for oxidation and hydrogenation reactions,
while the first application is for dehydrogenation reactions.

2.8.2.1. Membranes as distributors

When feeding oxygen or hydrogen through the membrane to react on the other side,
separation of unreacted H2 or O2 can be eliminated. Hydrogenation or oxidation reaction
rates are higher and the reactions are more selective. An impure H2 or 02 feed can be
used on the one side of the membrane if the membrane is very selective for H2 or O2
permeance. A uniform concentration of H2 or O2 can be obtained along the length of the
membrane. Nagamoto and Inoue (1981, 1985) compared the reaction rates of olefin
hydrogenation when separated feeds and mixed feeds were used. Separated feeds gave
better yields of the desired products in the olefin hydrogenation reactions. Ciavarella et
al. (1997) used a SCT membrane (a-alumina with y-alumina toplayer impregnated with
Pt) as contactor for hydrogenation reactions. The membrane reactor had a higher kinetic
activity, due to higher hydrogen coverage of active Pt sites, compared to the conventional
reactor. Shirai et aI. (1998) used a Pd membrane reactor as a hydrogen distributor for the
hydrosulphurisation of thiophene. They found that hydrogen atoms permeating through
the palladium film gave higher desulphurisation activity as compared to a system where
thiophene and hydrogen co-adsorb on the active sites. Itoh and Sathe (1997) tested the
validity of the bubble model and the solution model for the gas to liquid phase transport
of hydrogen through a palladium membrane during liquid phase hydrogenation reactions.

Gryaznov et al. (1986a) studied the oxidation of alcohols with oxygen fed through a
100 ~m silver film. Zaspalis et al. (1989) used microporous membranes for feeding
oxygen and observed an improved catalyst life in some cases. Many more oxygen
reactions are given by Noble and Stem (1995).

2.8.2.2. Membranes for dehydrogenation reactions

A membrane with a high separation factor is necessary to be effective in enhancing
conversion in an equilibrium-restricted reaction. For effective operation, the membrane
type must be non-porous. Palladium alloy, composite membranes are mostly used where

38
,.,,

a thin layer of palladium is deposited on an inorganic support. Most reversible

dehydrogenation reactions are either alkane or alcohol dehydrogenations. The literature
contains many examples of reactions that have been studied (Shu et al. 1991; Agarwalla
and Lund, 1992; Noble and Stem, 1995). A more detailed discussion on alkane and
alcohol dehydrogenation reactions will follow (Sections 2.9 and 2.10). The membrane
reactor can also be operated in such a way that simultaneous reactions take place on.
opposite sides of the membrane. For example, a dehydrogenation reaction may take
place in the membrane tube and the hydrogen that passes through the membrane may be
used for a hydrogenation reaction on the other side.

2.8.2.3. Other applications

Another class of equilibrium-restricted reactions that can benefit from a membrane
reactor is decomposition reactions, to remove impurities that can poison catalysts
downstream. These include the decomposition of water (Compagnie des Metaux
Precieux, 1976), ammonia (Collins, 1993b) and hydrogen sulphide (Edlund and Pledger,
1993b; Ohashi et aI., 1998).

Some niche applications of palladium membranes that have been investigated include: the
production of hydrogen from methanol for fuel cell powered vehicles (Jorgensen et aI.,
1997) and for low CO 2 power generation (Weyton et aI., 1997).

2.8.3. PROBLEMS WITH CATALYTIC MEMBRANE REACTORS

There are many problems associated with catalytic membrane reactors, both from a
mechanical and a membrane structural point of view. Some mechanical problems are
(Armor, 1995):
• For catalytic membrane operation, membrane tubes need to be packed with catalyst,
which can damage the alloy film. Replacing catalyst after a long operating time is
often very difficult.
• To obtain a large membrane surface area, capillary membranes or hollow fibres need
to be used. These membranes suffer from poor mechanical strength, and coating a
uniform dense Pd layer on the inside of these tubes will be very difficult.

39
,l'i

• Methods have been developed for sealing ceramic tubes (10 mm outer diameter) at
high temperatures. Sealing capillaries or hollow fibres at temperatures above 400°C
will pose problems and replacing broken membranes will be time consuming.
• High temperature operation of Pd composite, ceramic membranes might reqUire
special reactor construction materials. The difference in thermal expansion of the
ceramic membranes and, for instance, a steel reactor will cause damage to the
palladium film deposited on the ceramic supports.
• Films are sensitive to start-up and shutdown procedures. It should be performed
slowly and proper purging procedures must be followed.

Besides these mechanical problems, there are numerous other membrane structural and
related problems that still need to be addressed.
• The membrane cost is very high and needs to be reduced by making the supports
cheaper, as well as producing thin very selective Pd alloy films.
• Membrane poisoning and competitive adsorption on palladium sites reduce hydrogen
permeance.
• Pd alloying and the long term performance of palladium alloy membranes, need to be
studied further.
• Migration of elements in the palladium alloy matrix takes place, resulting in loss of
uniformity.
• Membrane regeneration procedures need to be improved.

2.9. DEHYDROGENATION OF ALKANES

Ethylene and propylene have traditionally been produced by mainly steam cracking,
while gasoline production utilises the fluid catalytic cracking process. The disadvantage
of both processes is the simultaneous production of a range of other products, which
reduces the yield of the desired product. Alkanes have become an alternative feedstock
for alkenes due to their good availability from wet natural gas. Typical commodity
chemicals produced by dehydrogenation include butadiene, styrene and alkylbenzene

40
derivatives. Dehydrogenation not only exploits the cheap and abundant natural gas
liquids, but it also offers improved selectivity towards the desired products.

2.9.1. GENERAL PRINCIPLES

Dehydrogenation reactions are endothermic and reqUIre large amounts of energy to
remove hydrogen molecules. To obtain conversions in the region of 50% for C2 to C5
paraffins, reaction temperatures of between 530 °C and 730 °C must be employed (Ertl et
a!., 1997). For lighter paraffins even higher temperatures are required. High
temperatures cause thermal cracking, lowering selectivity. A compromise between
selectivity and conversion must be found when choosing the operating temperatures.

Unwanted side reactions produce heavy aromatics and coke during dehydrogenation.
These compounds are irreversibly adsorbed on the catalyst and can cause very significant
catalyst deactivation, even though they make up only a fraction of the reaction products
(1 % or less). Catalysts are frequently exposed to an air atmosphere to bum off coke.

According to stoichiometry, a pressure increase has a negative effect on reaction

conversion. Processes often operate at lower pressures to increase conversion. The
hydrocarbon partial pressure is reduced by co-feeding it with a diluent such as steam or
hydrogen, and thereby shifting the equilibrium. Alternatively the process is operated in a
vacuum.

2.9.2. ALKANE DEHYDROGENATION CATALYSTS

Ertl et al. (1997) cited three basic catalyst classes for paraffin dehydrogenation, with
further differentiation between each class depending on the support type. Catalysts
contain a support, doped with alkali metals to reduce acidity, and one or more different
active sites.
1. Chromium oxides on zirconia and alumina supports with promoters,
The Snamprogetti-Yarsintez FBD (fluidized bed dehydrogenation) process and
the Lummus Catofin processes use a catalyst consisting of an alumina support
doped with an alkali metal and chromium oxide as the active sites.

41
 The higher thermal stability and lower acidity of Zr02 compared to alumina has
resulted in further investigation into this possibility (Ertl et a!., 1997).

2. Supported noble metal catalysts (usually platinum) with additional promoters, or

The UOP Oleflex process (James and Castor, 1994) employs an alumina support
with alkali metals and PtlSn active sites.
The Phillips STAR (steam active reforming) process uses ZnAh04 or MgAh04
supports doped with alkali elements and PtlSn active sites.

3. Miscellaneous catalysts such as supported nickel sulphides and vanadium or

molybdenum oxides supported on alumina. The alumina is doped with alkaline
metals. The addition of sulphur to nickel improves selectivity significantly (Lundin et
a!., 1993). Without sulphur, nickel becomes non-selective for dehydrogenation
reactions.

2.9.3. COMMERCIAL DEHYDROGENATION PROCESSES

Low alkanes (propane, butane, iso-butane and isopentane) are currently dehydrogenated
in chemical plants all over the world. Table 2.2 and 2.3 (from Ertl et a!., 1997) indicate
the process conditions and efficiency for propane and iso-butane dehydrogenation.

Table 2.2: Operating conditions for various dehydrogenation processes

Process Operating Operating temperature (0C) Feed rate
pressure (kPa) LHSV (h- I )

FBD 110-150 547-597 0.4-2.0

Oleflex 200-500 547-597 4

Catofin 33-50 587-647 0.4-2
Star 300-800 477-597 0.5-10

LHSV is the liquid hourly space volume, i.e. the time it takes to process one reactor
volume of feed liquid

42
Table 2.3: Process efficiency for propane and iso-butane dehydrogenation
Process Feed Conversion (%) Selectivity (mol%)
FBD Propane 40 89
Iso-butane 50 91
Oleflex Propane 25 89-91
Iso-butane 35 91-93
Catofin Propane 48-65 82-87
Iso-butane 60-65 93
Star Propane 30-40 80-90
Iso-butane 40-55 92-98
Linde Propane 30 90

2.9.3.1. FBD Process (Snamprogetti-Yarsintez)

The reaction is carried out in a bubbling fluidised bed reactor (multi-stage) without steam
addition and at atmospheric pressure. A chromialalumina catalyst is continuously cycled
to a regenerator unit and back (Sanfilippo et aI., 1992). The catalyst has the following
properties:
• microspheroidal (average particle < 0.1 mm),
• high attrition resistance, and
• the support is a mixture of aluminas containg Cr10J (12-20 wt %), K10 (1-2 wt %)
and Si02 (1-2 wt %).

The catalyst recycle rate is 5-15 kg catalyst per kg iso-butane. The process is quite safe,
since the catalyst regeneration zone (oxidation) and reaction zone are separated. The
absence of fired heaters and the presence of highly effective dust filters make the process
environmentally friendly. The well controlled temperature profile allows for the highest
temperature and maximum conversion at the reactor outlet.

43
2.9.3.2. UOP OIeflex process
This process uses separate units for reaction and regeneration to dehydrogenate mainly
C3 and C4 paraffins. The reaction is performed in moving bed reactors (stacked and
adiabatic) with external heating between stages (Pujado and Vora, 1990). Hydrogen is
used as a diluent. The catalyst properties are:
• spherical pellets,
• a y-alurnina support with platinum (0.1-1 wt %), tin (0.1-4.0 wt %) and alkali metals
(0.1-4.0 wt %), with a surface area of between 50 and 120 m2/g, and
• the operational lifespan for the catalyst is between 1 and 3 years.

Tne hydrogen to hydrocarbon ratio is between I and 10. The dehydrogenation process is
continuous and the catalyst retains its activity for a long time period, giving a constant
yield. The positive reaction pressure improves safety by reducing air leakage and allows
for smaller equipment to be used. The separated reaction and regeneration units also
ensure safe operation.

2.9.3.3. Lummus Catofin process

A cyclic multi-reactor is used to dehydrogenate C3 and C4 paraffins at sub-atmospheric
pressure. Reaction and catalyst regeneration are performed in the same vessel. During
reaction the temperature is gradually raised towards the end of the cycle, which lowers
the selectivity. The catalyst has the following characteristics:
• cylindrical pellets,
• y-alurnina support (120 m2/g) with Cr203 (17-20 wt %) and alkaline metals
(1-2 wt %), and
• a catalyst life of 1 to 2 years.

Due to the low operating pressure, this process gives a high selectivity and a high yield
per pass of the required dehydrogenated product. Heat is supplied by the exothermic
regeneration reactions as well as by external heating. The catalyst is not only thermally
very stable, but has a high resistance to breaking and good tolerance to poisons.

44
2.9.3.4. STAR process (Phillips steam active reforming)
This process is suitable for both the dehydrocyclization of C6 and C7 alkanes and the
dehydrogenation of shorter alkanes. A multitude of fixed beds operate in fired furnaces
with steam added as a diluent. The catalyst properties can be summarised as:
• a support ofzinc or magnesium aluminate with a calcium aluminate binder,
• platinum (0.01-5 wt %) as active component with tin (0.1-5 wt %) as promoter (alkali
metals are optional), and
• the catalyst is very stable in a steam atmosphere and its life is between 1 and 2 years
(Dunn et a!., 1992).

The steam to hydrocarbon ratio is 4 to 5. Steam reduces the hydrogen and hydrocarbon
partial pressures, which shifts the reaction equilibrium towards the product side.
Furthermore, it supplies heat to the reaction zone and reduces coking through the steam
reforming reaction. Coke formation is also limited by the use of fired reactors and the
absence of high temperature pre-heating sections. The positive pressure reduces leakage
risks.

2.9.3.5. Linde-BASF process

An iso-thermal fixed bed reactor is used without a diluent to dehydrogenate light alkanes
at pressures exceeding atmospheric pressure. The catalyst is a chromialalumina catalyst
(Bolt and Zimmermann, 1991).

Cracking is minimised by isothermal conditions, giving maximum selectivity. Positive

pressures improve safety and the absence of a diluent reduces equipment size.

2.9.4. ALTERNATIVE DEHYDROGENATION PROCESSES

Dehydrogenation processes other than those described in Section 2.9.3 are discussed
here.

45
 2.9.4.1. Coupled dehydrogenation and hydrogen oxidation
The oxidation of hydrogen is an exothermic reaction, which can be used to supply heat to
the endothermic dehydrogenation reaction. This process can take place in two separate
steps or in a single step. The latter is referred to as oxidative dehydrogenation (see
Section 2.9.4.2.). Hydrogen must first be removed in the dehydrogenation step and then
oxidised in the oxidation step. The catalyst must be selective towards hydrogen oxidation
without oxidising the hydrocarbon. The catalyst must also be very stable in steam and air
or molecular oxygen.

Single multi-functional (Imai and Schmidt, 1989) and dual catalysts (Imai and Schmidt,
1989; Imai, 1983) have been proposed, but without commercial application.

2.9.4.2. Oxidative dehydrogenation

This is similar to coupled dehydrogenation and hydrogen oxidation, but takes place in a
single step. Oxygen removes the hydrogen from the hydrocarbon molecule to form
water. Thermodynamic equilibrium restrictions are avoided and the exothermic reaction
supplies heat to the reaction zone. Catalysts generally suffer from low selectivity, since
carbon oxides and other oxygenates are formed in significant amounts.

2.9.4.3. Dehydrogenation in a membrane reactor

Dehydrogenation in a membrane reactor has several theoretical advantages over
performing the reaction in a conventional reactor. In theory, simultaneous reaction and
separation can occur, shifting the equilibrium towards the product side and giving higher
conversion at the same operating temperature. Some dehydrogenation reactions that have
been studied will be discussed.

2.9.5. METHANE STEAM REFORMING REACTION

In the 1980s, refineries were net producers of hydrogen (Courty and Chauvel, 1996).
However, in the 1990s they became net consumers as the world demand grew. Large
hydrogen plants were, and still are being, built to meet this ever growing need. Hydrogen

46

,'i
is the number one choice for desulphurising fuels and as environmental considerations
increase, more hydrogen is required (Armor, 1998).

Methane steam reforming is one of the most widely applied commercial processes for the
production of hydrogen from synthesis gas. The reactions (2.10 and 2.11) are very
endothermic and require operating temperatures up to 850°C.

CH4 + H20 (g) B CO + 3H2 Llli~98 = -206 kJ / mol (2.10)

CH4 + 2H 20 (g) B C02 + 3H2 t.H~98 = -165 kJ I mol (2.11)

The reaction requires a nickel supported catalyst operating at 800-850 °C under pressures
ranging from 1.6 to 4.1 MPa and with steam to methane ratios of between 2 and 4 (Shu et
aI., 1995). Under these conditions conversion is around 78%. Methane steam reforming
kinetics and carbon deposition reactions for methane steam reforming are discussed
extensively by Shu et al. (1994).

Oertel et al. (1987) obtained 90% methane conversion below 850°C in a palladium
membrane reactor (Pd thickness was 50 Iffil). Shu et al. (1995) indicated, through
mathematical modelling, that hydrogen removal through the membrane could best be
exploited at temperatures between 500 and 600°C. Using porous stainless steel
supported Pd and Pd-Ag membranes, packed with Ni/Alz0 3 catalyst, significant
improvements with the membrane reactor were obtained by Shu's group. Methane
conversion was 1.4 times higher in the membrane reactor at 500°C, 136 kPa and a steam
to methane ratio of 3. Uemiya and co-workers (1991c) achieved 80% conversion at
500°C and 100 kPa with a Ni/Alz03 catalyst and a steam to methane ratio of3 compared
to an equilibrium value of 42%. They used a membrane consisting of a 80 Ilm Pd77 -Ag23
alloy, coated on a porous glass tube. In similar work a team from Haldor Topsoe
(Jorgensen et aI., 1995; Armor IN, 1998) achieved 51% conversion (equilibrium = 21%)
at 500°C and 6 atm pressure with a 100 Iffil Pd-Ag alloy tube and commercial NilMgO
catalyst.

47
The group from Haldor Topsoe performed an economic analysis of this process and
indicated that the membrane process was not yet competitive with the conventional
process. This was due to a low hydrogen flux through the membrane, poor membrane
selectivity, membrane sealing problems and the absence of data regarding carbon
formation on the membrane during reaction.

Oxidation of methane can be considered as an oxidative dehydrogenation process. Total

oxidation must be avoided to prevent the formation of useless carbon dioxide.
Membranes can be used for achieving partial oxidation and yielding C2s (Cheng and
Shuai, 1995) or formaldehyde (Yang et aI., 1998). Methane and oxygen enter the reactor
at different sides of the membrane tube i.e. shell and tube side. Oxygen is added to
methane by permeating it through the porous membrane. The trans-membrane pressure
controls the permeance. This process is called dosing. Alternatively, dense oxygen
conducting membranes can be used.

2.9.6. WATER GAS SHIFT REACTION

This reaction is also important for hydrogen production or CO removal. Natural gas, oil,
coal and, more recently, biomass are hydrogen sources. These sources can produce a
CO/hydrogen mixture (syn gas) and the water gas shift reaction can be used to adjust the
gas composition for downstream processing or increase the hydrogen concentration.
Kinetic expressions for the exothermic water gas shift reaction can be found in Shu et al.
(1994).

Mf~98 =+41 kJ/mol (2.12)

Increasing temperature decreases CO converSIOn. Higher hydrogen yields can be

obtained using intercooling between multiple reactors, cooling with inert gas addition,
increasing the steam to CO ratio and shifting the equilibrium to the product side
(Burggraaf and Cot, 1996). The catalyst employed needs to be active and selective at low
steam to CO ratios as well as resistant to catalyst poisons. Three types of catalysts that

48
 have received attention are Fe-Cr catalysts (high temperature), Cu-Zn (low temperature)
and the more expensive Pt on ZnO catalysts (Bracht et aI., 1997). The latter one shows
higher sulphur resistance, higher activity and very good selectivity compared to the
"

commercially available Fe-Cr catalyst, which is inactive below 300°C. A newly

"

developed Co-Mo catalyst looks promising (Burggraaf and Cot, 1996).

Basile et al. (1996) reported conversions below equilibrium values with their Pd coated
alumina membranes. The 0.2 J.lIll Pd coating was too thin and resulted in poor membrane
selectivity and low reaction yield. With a 20 ).1m Pd coating on porous glass, Uemiya and
co-workers (1991d) achieved a conversion of 88%, compared to a 78% equilibrium
value. They used a Fe-Cr catalyst at 400°C, 1 bar and a steam to CO ratio of 2: I. The
catalyst unit and membrane separation unit can be separated to simplifY catalyst and
membrane regeneration (Ross and Xue, 1995). The units can then be operated at
different temperatures and membrane sealing is simplified. The additional cost of a
separate unit and the smaller driving force in the reaction are some disadvantages to be
considered.

Van Veen and co-workers (Burggraaf and Cot, 1996) calculated that the optimum H2 to
C02 membrane selectivity required for this process, is about 40. The high pressure, high
temperature membrane sealing and the large membrane surface areas required (1500 m2
for a 300 MWe power plant), remain problematic.

2.9.7. ETHANE DEHYDROGENATION

The dehydrogenation of ethane is an important industrial reaction for ethylene
production, which is a valuable commodity especially in the plastics industry. High
temperature, homogeneous thermal cracking of ethane is the predominant ethylene
production process. The production of methane, acetylene and higher hydrocarbons as
by-products, reduces selectivity to about 80% with recycle. At 730°C the equilibrium
conversion for the ethane dehydrogenation reaction is still only about 50% (Ertl et aI.,
1997).

49

<ii:
,;!\:,., "
I~.y'.
illl,
 Mi~" = + 137 kJ I mol (2.13)

Platinum (Champagnie et ai., 1992) and palladium (Gobina and Hughes, 1994) supported
on alumina catalysts can be used for ethane dehydrogenation. Champagnie and co-
workers performed ethane dehydrogenation experiments in a multi-layer alumina
membrane (MEMBRALOX) impregnated with Pt. At 550°C, a trans-membrane.
pressure of 1 bar and a sweep gas to feed ratio of 2, they obtained about 19% conversion
compared to the equilibrium value of just under 10%. A good fit was obtained with a
model developed by them. Gobina and co-workers (1994, 1995a, 1995b) used a 6 )lm
Pd77 -Ag23 deposited on a porous Vycour glass membrane, for studying ethane
dehydrogenation. The membrane was housed inside a stainless steel reactor and packed
with a 0.5 wt % Pd on y-alumina catalyst. They studied different parameters including
the type of sweep gas used, sweep gas to feed flow ratio and feed contact time, to draw
up a mathematical model for the process.

2.9.8. DEHYDROGENATION OF PROPANE

Most propylene in Europe and the Far East is produced from gasoil and/or naphtha in
steam cracker plants. About 90% of the propylene in the United States originates from
refinery operations or steam crackers. Catalytic dehydrogenation plants account for the
balance. Propylene consumption is estimated to grow annually at 3-4% world-wide
(Burggraaf and Cot, 1996). Additional demand cannot be met with traditional methods,
and dehydrogenation will become increasingly important.

Dehydrogenation can be performed in adiabatic reactors usmg the Snamprogetti-

Yarsintez FBD process, Lummus Catofin or the UOP Oleflex process. Alternatively it
can be performed in isothermal reactors using the Phillips STAR process or the Linde-
BASF process. Two endothermic reactions are of importance, i.e.:

t,H~" = +124 kJ I mol (2.14)

t,H~98 = +81 kJ I mol (2.15)

50
The latter is favoured from a thermodynamic point of view, but should be suppressed by
using a very selective catalyst. Thermal cracking at high temperature should also be
avoided. Different types of catalysts can be used depending on the desired result. For the
oxidative dehydrogenation of propane to propene in a conventional reactor a V -Mg-O
catalyst gave 65% selectivity and 10% conversion at 540 °C (Chaar et a!., 1987). Cadus
et a!. (1996) studied Mg-Mo-O catalysts for the same reaction, obtaining very good
selectivity, but poor conversion. Saracco et a!. (1996) applied the process of dosing to
the oxidative dehydrogenation of propane. Air was fed in a controlled manner through a
Pt-impregnated tubular alumina membrane to react with propane on the opposite side of
the membrane.

Aromatisation of propane can be performed at 500-600 °C with a gallium-silicate catalyst

(Kusakabe et a!., 1996) yielding mainly benzene, toluene and xylene. For this, a
membrane reactor consisting of a 4.4 IJ.l11. Pd layer deposited on an a-alumina membrane
support was used. Benzene, toluene and xylene yields increased by about 10%
(compared to a conventional reactor) to 47% at 500 °C in the membrane reactor.
Sheintuch and Dessau (1996) used a 254 11m thick (2% RulPd) tube and a 76 IJ.l11. thick
(25% Ag-Pd ) tube packed with 0.52 wt % Pt supported catalyst to study propane
dehydrogenation. At 550 °C, 70% propane conversion was obtained compared to 23%,
which is the equilibrium value. The flow rates used were very low to ensure equilibrium
conversion for the plug flow reactor (see discussion in section 8.1). Increasing the flow
rate reduced the conversion to values below the equilibrium value. The thick tubes
resulted in very low hydrogen fluxes and the low hydrogen partial pressure in the product
stream accelerated membrane coking and bed deactivation. Ziaka et a!. (1993) obtained
48% conversion at 580 °C compared to an equilibrium value of 31 % with a ceramic
membrane.

Burggraaf and Cot (1996) performed modeling calculations and did an economic analysis
on the use of membranes for the dehydrogenation of propane. Membrane selectivity
should be significantly greater than Knudsen diffusion selective membranes to prevent

51
reactant and product loss through the membrane pores. High hydrogen diffusion rates
must be obtained to reduce the membrane surface area. The use of Knudsen diffusion
membranes improve the return on investment for both the Catofin and Oleflex processes
by about 3% and I %, respectively. However, the process seems to be uneconomical with
a price difference ofless than $250-300 between a ton of propylene and a ton of propane.

2.9.9. DEHYDROGENATION OF BUTANES AND BUTENES

Commercially produced butenes are predominantly obtained as by-products of thermal or
catalytic cracking reactions. Catalytic dehydrogenation of n-butane yields a mixture of
1- and 2-butenes. This process is still used to some extent in Eastern Europe where
butenes serve as intermediates in the Phillips two-step process for obtaining butadiene
from butane (Ullmann, 1987). In the 1970s this process was frequently used in the USA,
but in recent years it has been replaced by butadiene from stearn cracking. I-Butene and
2-butene are either purified and separated or fed to the second stage in the Phillips
process for butadiene production.

Iso-butene demand is increasing due to changes in the Clean Air Act passed by the US
Congress in 1990 (Udomsak and Anthony, 1996). Iso-butene is one of the reactants for
producing the octane booster, methyl tert-butyl ether (MTBE).

2.9.9.1. Iso-butane dehydrogenation

The endothermic iso-butane dehydrogenation reaction is performed between 500 and
600°C and yields a variety of products. The purpose of the catalyst is to maximise both
conversion and selectivity towards iso-butene production. Isomerisation products, coking
products and other butenes must be minimised. Chromia type catalysts supported on
alumina, titania or silica (Udomsak and Anthony, 1996) and supported noble metal
catalyst with or without alkali metal promoters are generally used (Loc et aI., 1995).

The dehydrogenation of iso-butane in a membrane reactor has been studied by several

research tearns. Sheintuch and Dessau (1996) used a 254 !J.Il1 thick (2% RuJPd) tube and
a 76 /lm thick (25% Ag-Pd ) tube packed with 0.52 wt % Pt-supported catalyst to study

52
the reaction. The best yield obtained was 76% at 500°C towards total butenes (68% iso-
butene) with a reaction conversion of 81%. Increasing the flowrate reduced the
conversion to that of the equilibrium value (32.4%), due to poor hydrogen permeance
through the relatively thick Pd membrane. A hydrogen feed concentration of about 2%
gave better results due to less catalyst deactivation. Matsuda et al. (1993) obtained iso-
butane conversions exceeding the equilibrium value at temperatures between 350°C and
450 °C with a Pd (deposited by electroless plating) on alumina membrane. Membrane
deactivation was reduced by co-feeding iso-butane with a small percentage of hydrogen.
Shu et al. (1997b) improved iso-butane conversion from 12% to 32.7% and iso-butene
yield from 8.8% to 30.8% at 450°C with a membrane reactor. Their membrane consisted
of a 2 11m Pd coating on a multi-layer (a-alumina support and y-alumina toplayer) SCT
membrane.

Zhu et al. (1993) impregnated a 4 nm y-alumina membrane with Cr203 and used it to test
iso-butane dehydrogenation, while Ioannides and Gavalas (1993) used a membrane
consisting of a CVD deposited Si02 layer on porous Vycor glass for the same reaction.
Zeolite-coated membranes, for iso-butane dehydrogenation, have also been studied
(Casanave et aI., 1995).

2.9.9.2. n-Butane and butene dehydrogenation

Prior to the 1940s many catalytic processes were already known for preparing butadiene
from butenes and butane (Egloff and Hulla, 1942). During World War II the single step
Houdry Catadiene process (Forster, 1973; Craig and Dufallo, 1979) and the two step
Phillips process (Reidel, 1957; Husen et aI., 1971; Hutson et a!., 1974) were
commercially introduced for producing butadiene from butane. Later the Lummus
Catofin process was derived from the Houdry process (Craig and White, 1980). The
Phillips process uses oxidative dehydrogenation for butadiene production. Similar
oxidative dehydrogenation processes were developed in latter years by BP Chemicals
(Newman, 1970) and Petro-Tex (Welch et aI., 1978).

53
Dehydrogenation has been studied widely and modelled in fixed bed reactors (Acharya
and Hughes, 1990). Chromialalumina is the most common catalyst used for butane
dehydrogenation (Happel et a!., 1966; Hakuli et a!., 1996) and the kinetics of the reaction
have been documented (Noda et a!., 1967). Since catalyst coking is such an important
phenomenon, many researchers have investigated chromialalumina coking during butene
dehydrogenation (Hughes and Koon, 1994; Brito et a!., 1995, 1996). Coking kinetics has
been formulated (Romero et a!., 1996; Pena et a!., 1993). Other types of catalysts that
have also received attention include iron oxides supported on titania (Boot et a!., 1994,
1996), calcium-nickel-phosphate catalysts (Arnold, 1961; Swift et a!., 1976) and platinum
on alumina catalysts (Loc LC et a!., 1993, 1996).

Oxidative dehydrogenation of butenes takes place mainly on ferrites (Xu et a!., 1992;
Yang et a!., 1984) or magnesium ferrites (Gibson and Hightower, 1976; Yang et a!.,
1991). Iron-zinc oxides (Armendariz et a!., 1992) and vanadium-magnesium oxides
(Bhattacharyya et a!., 1992) are just some of the other catalysts that have also been
studied for this reaction.

Gobina and Hughes (1996) studied butane dehydrogenation in a membrane reactor

packed with a 0.5 wt % Pdlalumina catalyst. The membrane had a 6 ).lm Pd-Ag coating
on porous Vycor glass. Using reactive sweep gases, the equilibrium conversion of 5% at
400 DC was surpassed. A 21 % oxygen sweep gas gave a reaction conversion of 39% and
a II % carbon monoxide sweep gas gave 26% conversion. Rezac and co-workers (1994,
1995) used a thermally stable polymer-ceramic membrane for hydrogen extraction from
the product mixture. They improved conversions from 22% to 33% for reaction
temperatures between 480 and 540°C. Zaspalis et al. (1991) used an a-alumina
membrane with a y-alumina top layer (4-5 run pore size) and a Pt/Si0 2 catalyst for
performing butane dehydrogenation reactions. At 500°C, they achieved a 15%
conversion, exceeding the calculated equilibrium conversion by 6%.

54
2.9.10. DEHYDROGENATION OF ETHYLBENZENE TO STYRENE
The worldwide demand for styrene reached 18.2 million tons in 1992. This figure is
estimated to grow at 3-5% per annum, to reach 23.9 million tons in 2000 (Burggraaf and
Cot, 1996). Styrene is converted to polystyrene (> 65%) and co-polymerised with
butadiene (13%) for rubber elastomer production. A fraction is used for styrene-
acrylonitrile production (9%) and mixed with unsaturated polyester resins (James and
Castor, 1994). The endothermic reaction is usually performed at temperatures between
550°C and 650 °C, 0.5-1.0 aim pressure and with steam as a diluent. Ertl et aL (1997)
lists many reasons for adding steam. The main by-products of the dehydrogenation
reaction are toluene and benzene, with smaller amounts of ethylene, methane and coking
products.

Potassium-promoted iron oxide catalysts are mostly used for this reaction and have been
studied extensively (Muhler et aI., 1990, 1992; Addiego et aI., 1994). Catalyst activity
decays due to changes in the surface concentration of the potassium promoter (Matsui et
aI., 1989, 1991). The kinetics of the dehydrogenation reaction (Abdalla et aI., 1994a,
1995; Coulter et aI., 1995) and modelling of commercial fixed bed reactors (Abdalla and
Elnashaie, 1995; Elnashaie et aI., 1993) have received significant attention. Oxidative
dehydrogenation allows for lower reaction temperatures (around 450°C), to be used but
catalysts usually suffer from poor selectivity. Magnesium vanadates (Chang et aI., 1995),
Zn-Fe-Cr catalysts (Jebarathinam et aI., 1996) and carbonaceous catalysts (Drago et aI.,
1994) are some that have been investigated for use in oxidative dehydrogenation.

Most membrane studies regarding ethylbenzene dehydrogenation have used Knudsen

separation membranes (40-50 A pore size) made of a-alumina with a y-alumina toplayer
(MEMBRALOX). Yang et aL (1995) obtained 4% and 11 % yield improvements over a
conventional reactor at 620°C with different membrane reactor configurations. Gallaher
et aL (1993) also observed a 4% yield increase at 600°C with a MEMBRALOX
membrane. Both his group and Tiscareno-Lechuga et aL (1993) observed significant
carbon deposition on the membrane during operation. Jiang and Wang (1998) obtained a
10% increase in yield between 560 and 600°C using a similar Knudsen separation

55
membrane. Becker et al. (1993) reported improvements as high as 20%. Wu and Liu
(1992) as well as Abdalla and Elnashaie (1994b) have modelled the membrane reactor for
ethylbenzene dehydrogenation. Gobina et al. (1995c) compared the effects of the
separation factor of different microporous membranes and dense Pd-Ag membranes on
styrene yield. There seems to be an optimum membrane thickness for microporous
membranes, which depends on the porosity of the separation layer. Pd-Ag membranes of
less than 50 IJ.l11 thickness outperform microporous membranes.

In a feasibility study on the dehydrogenation of ethylbenzene in a membrane reactor,

performed by Burggraaf and Cot (1996), they concluded that the process was not feasible.
With the exception of palladium-based membranes, other inorganic membranes caused
high ethylbenzene permeation through the membrane resulting in large reactant losses.
Higher reaction rates, 'higher hydrogen permeances and good selectivity needed to be
achieved to make this process profitable. The high steam to ethylbenzene ratio in the
feed, resulted in a low hydrogen partial pressure on the product side. This reduced the
driving force for hydrogen permeance through the membrane.

2.10. ALCOHOL DEHYDROGENATION CATALYSTS

The industrially used alcohol dehydrogenation catalysts are copper and/or zinc-based.
Some oxidative dehydrogenation processes employ silver as a catalyst. Copper-based
catalysts can either be unsupported or supported. Most are of the supported type, where
the support provides a large surface for the copper to be deposited on. Unsupported
copper catalysts have a much smaller surface area. Catalyst supports can be basic, acidic
or both. The acidity of the support, determines whether the dehydration or the
dehydrogenation reaction will be favoured. Basic supports (high pH) favour the
dehydrogenation reaction, while acidic supports (low pH) favour the dehydration
reactions. Table 2.4 lists some of the metal oxides which are often used as catalyst
supports.

56
Table 2.4: Catalyst supports (Carrizosa and Munuera, 1977; Rosynek et a!., 1990)
Basic oxides Acidic oxides
Pure silica Alumina
MgO Silica-alumina (zeolites)
ZnO Ti0 2
Zr02
Alkiline earth oxides

It is not only the acidity of the support that is important. Other factors which are also
very important are:
• surface area,
• abrasion resistance and crushing strength,
• temperature stability and resistance to fouling, and
• toxicity.

Silica and alumina have very high surface areas compared to the other oxides (typically
in the hundreds of m 2/g area). High copper surface areas can be obtained by depositing
copper on these supports. The activity of the catalyst is usually proportional to the
surface area of the active sites and thus a large copper surface area will yield a more
active catalyst.

2.10.1. EFFECT OF COPPER PERCENTAGE ON CATALYST PERFORMANCE

The percentage copper on the support has an effect on both reaction conversion and
selectivity. Figure 2.9 indicates what happens when the copper concentration on the
support is increased. There is a decrease in the number of exposed surface sites of the
support as the copper concentration increases, and the total surface area of the catalyst
declines. For the copper, there is an optimum concentration, which yields a maximum
copper surface area. In the second image (b) of Figure 2.9 the copper concentration and
the copper surface area is higher than in the first image (a). When the copper
concentration is increased further, some of the pores in the support becomes blocked

57
(image (c». The total copper surface area available for reaction declines as no reactant
can enter the blocked pores to react.

+Cu + Cu
) )

(a) (b) (c)

Figure 2.9: Pore blocking by deposited copper

Sivaraj and Kantarao (1988a) prepared copper-supported on y-alumina catalysts by a

precipitation technique. For the 240 m2/g y-alumina support, a copper loading of 20 to
25 wt % gave an optimum copper surface area of about 40 m2/g catalyst. Chang and
Saleque (1993) determined cyclohexanol conversion on copper/a-alumina catalysts.
Maximum conversions were obtained with copper loadings of about 12-17 wt % prepared
by electroless plating, 12-17 wt % (prepared by precipitation) and about 10 wt %
(prepared by impregnation). Chang and Saleque (1994) obtained a maximum
cyclohexanol conversion on copper/y-alumina with 18-20 wt % copper loadings. For
alumina-based catalysts, the selectivity towards aldehyde or ketone formation increases
with increasing copper loading (Sivaraj et aI., 1990; Chang and Saleque, 1994). The
reason is that the acidic sites, which reduce dehydrogenation selectivity, become covered
and thus neutralised. Jeon and Chung (1994) observed a continuous decrease in
cyclohexanol conversion with an increase in copper loading on copper/silica catalysts.
The selectivity remained constant (very high) over a wide range of copper values.

2.10.2. CATALYST PREPARATION TECHNIQUES

The are a few basic methods by which to prepare copper catalysts. Unsupported catalysts
are usually prepared via a co-precipitation technique, while several methods can be used
for preparing supported catalysts viz. precipitation, impregnation, urea hydrolysis,
electroless plating or ion exchange.

58
2.10.2.1. Precipitation
This technique is suitable for preparing copper-supported or unsupported catalysts (Opitz
and Urbanski, 1958). A copper salt (usually copper nitrate) is dissolved in distilled water
in a stirring vessel. The support is added to the vessel while stirring and sodium
carbonate is added as a precipitant. The rate of addition is controlled to keep the solution
at a constant pH, usually pH=8.0. The precipitation temperature and precipitation time
both have an effect on catalyst performance. Increasing the precipitation temperature
from 20 °C to 90°C (Jeon and Chung, 1994) resulted in a catalyst that gave a much
higher cyc1ohexanol conversion at the higher precipitation temperature. From TEM
measurements, Jeon and Chung (1994) observed that copper particles precipitated at
20°C were about 7 nm to 15 nm in size, while those precipitated at 90°C were about
1.7 nm to 7 nm in size. Conversion was a weak function of precipitation time, with about
4 hours precipitation time giving the best conversion.

Unsupported copper catalysts are prepared in a similar way, with other metals (for
example chromium and/or cobalt) added in the correct mass ratio. The precipitate is
filtered, washed with ample amounts of water and then dried overnight at about 100 °C.
Thereafter, it is calcinated at temperatures exceeding 450°C. Kanoun et al. (1991,1993)
described methods for preparing vanadium-copper-zinc and copper-chromium-aluminum
catalysts via precipitation. Ammonia was added to precipitate the hydroxides. Further
ammonia was added to redissolve the hydroxides as amine complexes. Ammonia was
then removed by heat to yield the precipitate with the desired structure.

2.10.2.2. Urea hydrolysis

This method is a precipitation technique in which urea is used as the precipitant (Chang
and Saleque, 1994). An aqueous solution of copper nitrate, urea and the support are
added together in a stirring vessel. The solution is then heated to between 90 and 95°C
(Sivaraj et aI., 1988a, 1990). The pH stabilises at about 7 to 7.5 after the deposition is
complete. The precipitate is then filtered, washed and dried.

59
2.10.2.3. Electroless plating
Electroless copper plating has been used on occasion to deposit copper on the support
(Chang and Saleque, 1993, 1994). The support is first pretreated with palladium and tin
chloride solutions before copper plating. Palladium seeds on the support provide
catalytic centres for copper to plate on. Copper plating requires an alkaline solution and
is performed at high temperatures (typically> 70°C). Either formaldehyde or hydrazine .
is used as reducing agent and EDTA as a stabiliser. The concentration of the plating
solution or the volume of the plating solution can be altered to give different copper
loadings on the support. After plating the catalyst is filtered, washed, dried and
calcinated.

2.10.2.4. Impregnation
This is a quick and easy way to prepare copper-supported catalysts with or without
additional elements like chromium, cobalt or rhodium. The support particles are dried at
about 200°C overnight to remove water vapour from the pores. The particles are then
cooled in a desiccator. Either a copper solution or copper mixed with additives in the
desired mass ratio are added to the support particles. The mixture is well stirred while
adding the solution. The paste is then dried at 90°C to 110°C for several hours and later
calcinated at above 400°C in air for at least 5 hours. Iwasa and Takezawa (1991) used
this technique to impregnate Ah03, SiOz, MgO, zrOz and ZnO with copper. Shiau and
Liaw (1992) prepared Ba-copperlSi0 2 catalysts and Mendes and Schmal (1997) prepar.ed
Rh-copperly-alumina catalysts by impregnation for alcohol dehydrogenation.

2.10.3. CATALYST REDUCTION

The reduction of copper oxide by hydrogen is a necessary step for copper activation. It is
now readily accepted that it is Cuo that is the active species on the catalyst, and not Cu2+
and Cu 1+. The reduction reaction is highly exothermic.

CuO (s) + H2 --- Cu (s) + H20 (g) LlH~20K = -85 kJ I mol (2.16)

60
The temperature inside the catalyst particle can rise considerably, causing sintering and
reducing the surface area of the copper. On the other hand, poor reduction will occur if
the reduction temperature is not high enough. From literature (Bart and Sneeden, 1987),
it appears as if standard prescribed reduction methods for CU/ZnO catalysts result in poor
reduction. Reduction temperatures in excess of 300°C give a better reduction percentage
fr9m Cu2+ to CuD, but give rise to more sintering of copper clusters. Lower temperatures
cause less sintering, but also poor reduction.

Furthermore, the copper concentration on the support influences the reduction ability. At
low Cu concentrations, the Cu-ion to support bond is very strong. Bart and Sneeden
(1987) reported that for 5-10 wt % CuO on ZnO, the surface content of copper was not
altered by hydrogen reduction for 2h at 250°C. Cu2+ was converted to both Cu 1+ and
CUD. For copper concentrations up to 35 wt % CuO, Cu 1+ was stilI detected.

Tohji et al. (1985) developed a structural model for copper particles in hydrogen during
reduction (Figure 2.10). In the first step, below 127°C, a two dimensional copper layer
develops over the ZnO support. Between 127°C and 277 °C copper metal clusters are
formed reversibly on the support. Above 327°C, large copper clusters are formed
irreversibly due to sintering.

C2CX:XJC)
127°C
~e------~ fmD 400°C ) @
Zno Zno Zno

Figure 2.10: The effect of temperature on copper surface structure

Mendes and Schmal (1997) reported that there is a strong interaction between alumina
and CuO, for low CuO percentages (0.5 wt % in his study). They could only obtain 57%
CuO reduction even if the reduction temperature was increased up to 500°C. Tu et al.
(1994a) reduced their unsupported Cr/Cu catalysts at 200°C for 6 hours in a
hydrogen/argon mixture. Some sintering occurred during the reduction process. A

61
proper dispersion of Cr203 promoter slowed down sintering during reduction. At high Cr
to Cu ratios (20:40), the reducibility of the catalyst at 200°C was very poor. Kanoun et
al. (1991) reduced their vanadium-copper-zinc catalyst for 3 hours at 300°C. Chung et
al. (1994) reduced their silica-supported copper catalyst for 4 hours at 250°C with
reduction percentages of 65% and higher. Iwasa and Takezawa (1991) reduced their
copper-supported catalyst for 18 hours at 250°C.

2.10.4. CATALYST DEACTIVATION

There are three processes which cause catalyst deactivation, namely, sintering, coking or
fouling and poisoning. Sintering is dependent on the reaction temperature, while coking
depends on the reaction process and poisoning is determined by the composition of the
feed. In most high temperature, solid-catalysed reactions, more than one deactivation
process will take place simultaneously.

2.10.4.1. Sintering
Irreversible sintering of copper catalysts at temperatures exceeding 250°C is a major
cause of catalyst deactivation. Sintering will increase gradually as the temperature is
raised (see Section 2.10.3) and the degree of sintering depends on the metal to support
physical and chemical bond as well as the copper crystallite size. Bart and Sneeden
(1987) reported that copper sintering for CU/ZnO/Ah03 catalysts is less than expected on
the basis of Tamman temperatures, due to strong bonding between Cu and ZnO, which
reduces copper mobility.

Church et al. (1951) found that Cr and Co addition improve long term copper catalyst
stability. After 100 hours of ethanol dehydrogenation, the activity of a Cu-asbestos
catalyst was 60% of the original activity, for a Cu(95%)-Co(5%)-asbestos catalyst the
activity was 70% of the original and for a Cu(93%)-Co(5%)-Cr(2%)-asbestos catalysts
the activity was 83% of the original. The support also proved to have a large effect on
catalyst life. Under identical conditions a Cu-Cr-Co/asbestos catalyst retained 80%
activity and a Cu-Cr-Co-pumice catalyst retained only 32% of initial activity after
reaction at 275°C for 48 hours.

62
i'
I
,

Tu et al. (1994a, 1994b ) detennined copper surface areas of unsupported Cr/Cu catalysts
before and after eight hours of use at 310°C (ethanol dehydrogenation reaction). For
catalysts both before and after use, the Cu surface area passed through a maximum at a Cr
to Cu ratio of 4:40. There was a significant decline in surface area after use. For pure
copper, the decline was from 9 to 6 m2/g copper and for Cr:Cu = 4:40, the decline was
from 19 to 17 m2/g copper. They found the deactivation kinetics to be second order and
concentration independent, which is similar to common sintering kinetics (Fogler, 1992).
Chinchen et al. (1988) concluded that in the absence of catalyst poisoning, sintering is the
major deactivation process for copper-based methanol synthesis catalysts.

2.10.4.2. Coking and poisoning

Coking is a common phenomenon in hydrocarbon reactions, where a carbonaceous layer
is deposited on the catalyst. Coking increases with time and is related to the feed
composition and the selectivity of the reaction. Polymerisation of reaction products may
result in the fonnation of heavy molecules, which adsorb on the catalyst and reduce
catalyst activity. Franckaerts and Froment (1964) reported that polymerisation of
acetaldehyde on the catalyst surface during the dehydrogenation of ethanol causes
gradual deactivation. Uemichi et al. (1995) stated that the deposition of cyclohexanone
oligomers on active copper sites during cyclohexanol dehydrogenation cause deactivation
of copper catalysts. Coking is a much more serious problem for oxidative
dehydrogenation reactions. The selectivity towards the desired product is usually lower
for oxidative dehydrogenation and CO, C02 and other carbonaceous product are fonned
which results in coking.

Poisoning is similar to c~king. The pOlsonmg molecules become irreversibly

chemisorbed onto active sites, reducing the number of sites available for reaction and
thus lowering activity. The poison may be a reactant, a product or an impurity in the
feedstream. Poisoning is usually associated with impurities, while reaction products
cause coking. Chinchen et al. (1988) reported that the rate of sulphur adsorption on
copper-based, methanol synthesis catalysts is proportional to the copper surface area.

63
Copper-zinc-alumina was very resistant to sulphur poisoning, with 2% sulphur in the feed
only causing a 20% decline in catalyst activity. For copper-alumina catalysts, 0.2%
sulphur in the feed caused total deactivation of the catalyst.

2.11. DEHYDROGENATION KINETICS

Perry and Chilton (1973) listed six possible rate equations for solid-catalysed
dehydrogenation reactions (Table 2.5).

Table 2.5: Possible rate equations for solid-catalysed dehydrogenation reactions

Chemical equation Catalytic steps Rate equation
A ..... R+S A+s ..... As k(C A -CRCs/K';) (2.17)
r=
1+ KRSC RCs + K. C. + KsC s
As + s ..... Rs+ Ss k(C A - cRCs I K.. ) (2.18)
r=
(l+K ACA +K,C R+KsCsY

Rs ..... R+s k(C A -C.Cs/K~) (2.19)

r
C.[I+K,C, +(K""c,/c.)+K.c.l
Ss ..... S +s

A ..... R+S A+s ..... As k(C A -CRCs/Keq) (2.20)

r=
I+K RC R +KRSCRC S
As ..... Rs+S k(C A -CRCs/K,q) (2.21)
r=
I+K A C A +KRC R
Rs ..... R+s k(C, -C.C,/K .. ) (2.22)
r=
CS [1+K,C A +(KASCA/C s )]

For solid-catalysed reactions there are five basic steps:

1. diffusion of the reactants from the bulk fluid to the external catalyst surface and into
the catalyst pores,
2. adsorption of the reactant onto the catalyst surface (exterior and in pores),
3. reaction of the adsorbed reactants to form product molecules,
4. desorption of the product molecules from the surface, and

64
 5. diffusion of the products from inside the pores and close to the external surface into
the bulk fluid.

2.12. DEHYDROGENATION OF ALCOHOLS

The catalytic dehydrogenation of primary and secondary alcohols yields aldehydes and
ketones and is an industrial process. The corresponding reaction is:

(2.23)

with R2 = H for primary alcohols and R2 = alkyl or aryl for secondary alcohols. Alcohol
dehydrogenation reactions are endothermic and are usually performed at temperatures
between 250 DC and 450 DC. Reactants are fed in the gas phase with conversion being
favoured by low operating pressures. The principal side reaction is dehydration of the
alcohol to yield the alkene and the di-ether. Oxidative dehydrogenation is an exothermic
reaction and requires strict temperature control to prevent total combustion and the
formation of carbon oxides.

Very few studies have dealt with membrane assisted dehydrogenation of alcohols. Most
membrane reactor experiments have thus far focussed on alkane dehydrogenation.

2.12.1. DEHYDROGENATION OF METHANOL TO YIELD FORMALDEHYDE

The older process combines dehydrogenation and oxidative dehydrogenation, using a
silver catalyst in the presence of water. The result is a 40% formaldehyde solution. The
more recent process employs an iron-molybdenum catalyst for methanol
dehydrogenation, yielding formaldehyde gas, which is dissolved in water. The catalyst
consists of about 19 wt % Fe20J and 81 wt % MoO) with cobalt and chromium oxide
promoters (Ert! et aI., 1997). Vanadium supported on silica catalysts have also been
investigated (Sorensen and Weber, 1993) for this reaction. The two reactions are:

65

,
i
CH3-0H (g) ..... HCHO + H2 L1H~98 = +84 kJ I mol (2.24)

CH3-0H (g) + O2 ---> HCHO + H20 (g) L1H~98 = -159 kJ Imol (2.25)

Reaction temperatures vary between 647°C and 717 DC, giving a formaldehyde yield of
about 87% (Ertl et a!., 1997) for the older process. For the newer process, reaction
temperatures vary between 350°C and 450 DC, with a formaldehyde yield of about 90%.

Zaspalis et al. (1991) used an a-alumina membrane with a y-alumina toplayer (4-5 nm
pore size) and a commercial ZnO catalyst for methanol dehydrogenation. At 500°C, the
membrane reactor gave a 19% formaldehyde yield compared to the 14.4% yield of the
conventional reactor. The disadvantage of this Knudsen separation membrane IS,

however, the loss of reactant through the pores due to poor separation characteristics.

The decomposition of methanol in a Pd/Ru/ln membrane reactor to yield CO and

hydrogen has been studied recently (Hara et aI., 1999).

2.12.2. DEHYDROGENATION OF ISO-PROPANOL TO ACETONE

The dehydrogenation of iso-propanol to acetone is endothermic, yielding acetone as the
main product. The main side reaction is the dehydration reaction, yielding propene.

L1H~98 = +84 kJ I mol (2.26)

The equilibrium constant Kp (in bar) is as follows (with T in K):

log Kp = -2764/T + 1.526*logT + 1.765 (2.27)

Many different catalysts have been tested for use in this reaction, the most common being
copper and/or zinc-based. For ZnO the reaction is performed at 300-400 °c with a
selectivity of 90% and a conversion of 98%. Cunningham et al. (1986) studied
unsupported CuO, CU20 and Cu metal, while Szabo et aI. (1975) determined activation

66
energies and Arrhenius parameters for MgO, CaO and srO. Yashima et al. (1974)
studied iso-propanol dehydration and dehydrogenation on alkali exchanged zeolites and
found the dehydration behaviour to be proportional to the acidity of the catalyst. Other
catalysts tested include lead, manganese oxide, tin-iridium complexes (Matsubara et aI.,
1991) and supported platinum and rhodium catalysts. Oxidative dehydrogenation of iso-
propanol uses similar catalysts (Gil et aI., 1996a, 1996b).

2.12.3. DEHYDROGENATION OF CYCLOHEXANOL TO CYCLOHEXANONE

The dehydrogenation of cyc1ohexanol to cyc1ohexanone is an important industrial
reaction for the production of caprolactam, which in turn is used for producing nylon.

L\H~98 = +80 kJ / mol (2.28)

The reaction becomes pseudo first order when surface adsorption is weak. The use of
several catalysts has been studied, usually in the temperature range between 200°C and
400 0c. The most common catalysts are copper-supported or unsupported catalysts, as
indicated in Table 2.6.

Table 2.6: Common supports for copper catalysts

Support Additives Reference
Ah 0 3 ZnO Sivaraj et aI., 1988b
ZnO - Lin et aI., 1988
y-Ah0 3 Rh Mendes and Schmal, 1997
MgO, ZnO, Fe304, - Jeon and Chung, 1994
Si0 2
a-Ah0 3 - Chang and Saleque, 1993
a-Ah0 3 - Sivaraj et aI., 1990
a-Ah03, y-AI20) - Chang and Saleque, 1994

Other catalysts that have also been studied for cyclohexanol dehydrogenation include
alumina-supported Pt-Co catalysts (Reddy et aI., 1997), carbon-supported cobalt catalysts

67
(Uemichi et aI., 1995), tin oxides (Hino and Arata, 1990) and zinc phosphate complexes
(Aramendia et aI., 1995).

The activity and selectivity of copper-based catalysts depends on several factors. The
more important ones are:
• acidity ofthe support,
• preparation technique,
• copper loading,
• calcination temperature, and
• reduction temperature.

Jeon and Chung (1994) prepared copper-supported catalysts usmg a deposition

precipitation method. They studied the effects of precipitation time, precipitation
temperature and copper loading on cyclohexanone yield. Cyclohexanol conversion
increased with catalysts prepared at higher precipitation temperatures. There was an
optimum precipitation time, while conversion decreased with increasing copper content.
Selectivities and conversions of MgO, ZnO, Fe)04 and Si02 were compared after catalyst
deactivation.

Chang and Saleque (1993) studied three preparation methods viz. electroless plating,
impregnation and deposition precipitation to determine the effects of the preparation
method on cyclohexanol conversion. Each method showed an optimum copper
percentage for maximum cyc1ohexanol conversion. Chang and Saleque (1993) listed.
activation energies and Arrhenius frequency factors for various catalysts. Catalysts
prepared by electro less plating gave the highest BET surface area and the best
conversion. Chang and Saleque (1994) also compared the activities of electro less plated
Cu on a-Ah03, and y-Ah03 supports, while Sivaraj et al. (1990) determined the
relationship between dehydrogenation selectivity and catalyst acidity.

68
 2.12.4. DEHYDROGENATION OF C4 ALCOHOLS
The dehydrogenation of butanol yields butyraldehyde, an important intermediate in the
manufacture of solvents, plasticisers, synthetic resins and rubber vulcanisation
accelerators. This process is not performed on an industrial scale, but several researchers
have investigated the reaction. Copper, copper-chromium and zinc-based catalysts have
been studied (Rao et aI., 1969; Purohit and Gandhi, 1975).

Shiau and Liaw (1992) used copper-barium supported on silica to investigate the kinetics
of the oxidative dehydrogenation of n-butanol between 230°C and 290 0c. Sintering
occurred and the addition of oxygen increased catalyst coking. Raizada et al. (1993)
determined reaction rate parameters for the dehydrogenation of n-butanol over a zinc
oxide catalyst at temperatures between 350°C and 450 0c. Copper, zinc or copper-zinc
catalysts were most widely used.

The dehydrogenation of sec-butyl alcohol (iso-butanol) to yield methyl ethyl ketone

(MEK) is an important industrial process. MEK is a widely used industrial solvent.
Perona and Thodos (1957) determined the kinetics of iso-butanol dehydrogenation
between 340°C and 400 °C and 1 atm, using a brass catalyst (65 wt % Cu, 35 wt % Zn).
Using a similar catalyst, Thaller and Thodos (1960) studied the reaction between 290°C
and 370 °C at pressures up to 15 atm. Ford and Perlmutter (1964) observed a change in
reaction mechanism with temperature. Below about 320°C and above about 425 °C, the
single site surface reaction was rate limiting. In the temperature range in between,
especially from 350°C to 400 °C, the adsorption of the alcohol was the controlling
mechanism.

2.12.5. DEHYDROGENATION OF ETHANOL

Prior to the 1970s the dehydrogenation of ethanol to prodl.\ce acetaldehyde was a very
widely practised commercial process, but it has become less attractive due to high energy
costs. In 1989 more than 98% of the world's 2.5 million ton acetaldehyde production
was from the oxidation of ethylene (Raich and Foley, 1998). This is the so-called
Wacker process. It, however, produces some chlorinated wastes, which are undesirable.

69

,!,
Combining membrane technology with ethanol dehydrogenation results in advantages
that could not be exploited previously. The added value of separated hydrogen as a co-
product and a dehydrogenation processes that operates at a lower temperatures, give more
credibility to ethanol dehydrogenation.

There are two main catalyst classes for ethanol dehydrogenation; zinc or zinc.oxides and
copper-based catalysts. With lnO and Cu catalysts, the ethanol dehydrogenation reaction
proceeds via different mechanisms on the catalyst surface (Chung et ai., 1993). Copper is
more active and reaction temperatures are usually in the region of 250°C to 300 0c.
Higher temperatures cause catalyst deactivation and promoters like cobalt and chromium
are often added to copper to improve catalyst stability. The lower activity of zinc-based
catalysts lead to higher operating temperatures, typically 350°C to 450 °C. Although
zinc-based catalysts are thermally more stable, they have other disadvantages. Higher
temperatures decrease reaction selectivity and cause thermal cracking of reactants and
reaction products.

The dehydrogenation reaction is:

Llli~98 = +52 kJ / mol (2.29)

The ethanol dehydrogenation capabilities of some materials, including pure magnesium

oxide (Takezawa et ai., 1975), silica pillared rectorite (Hao et al., 1994), highly
dehydrated silica (Matsumura et aI., 1989) and silicalite-l (Matsumura et aI., 1990,
1991), but excluding copper and zinc, have been tested. All the materials tested, suffer
from low activity and the selectivity towards acetaldehyde production is generally poor as
well.

2.12.5.1. Catalyst selectivity

Iwasa and Takezawa (1991) found the dehydrogenation of ethanol to acetaldehyde to be
most selective on Si02 and MgO supports (see Table 2.7), but MgO has a much smaller
surface area than Si02. lnO and lr02 gave a high selectivity towards ethyl acetate

70
fonnation, while Ah03 supports resulted in C4 and di-ethyl ether fonnation at 220°C.
Furthennore, they found that high copper percentages suppress ethyl acetate fonnation.

Table 2.7: Selectivities for ethanol dehydrogenation (Iwasa and Takezawa, 1991)
Support Active T P Conver- Se1ecti-
Materials (oq (kPa) sion vity
AIz 0 3 30%Cu 220 20.5 50* 54.1
Si0 2 30%Cu 220 20.5 50* 77.9
MgO' 30%Cu 220 20.5 50* 74.2
Zr02 30%Cu 220 20.5 50* 57.3
ZnO 30%Cu 220 20.5 50* 67.3
- Cu 220 20.5 50* 70.6
...
*ConverslOn was fixed at 50% to test selectIvItIes.

Church et al. (1951) found an improvement in both conversion and selectivity when Co
and Cr were added to Cu supported on asbestos (see Table 2.8).

Table 2.8: Effect ofCr and Co addition to Cu on ethanol dehydrogenation (Church

et aI., 1951)
Support Active T P Conver- Selecti-
Materials ceq '(kPa) sion vity
- CU,7%ZnO 335 - 80 78
- Cu,5%MgO 332 - 78 78

Asbestos Cu 328 - 79 90
Asbestos Cu, 5% Co 337 - 94 89
Asbestos Cu, 5% Co, 2% Cr 330 - 93 83

71
2.12.5.2. Catalyst activity
To test catalyst activity, the reaction is usually performed under differential reactor
conditions, i.e. very low reactant conversion. Under such conditions, the selectivity
towards acetaldehyde production is high and usually not considered.

Tu et a!. (l994b) tested the effects of chromium addition to unsupported copper catalysts
for reaction temperatures ranging from 250 DC to 310 DC. There appeared to be an
optimum Cr to Cu ratio of 4 to 40. That composition gave the highest activity over time
with the smallest deactivation rate. The high activity was related to the highest copper
surface area at that Cr to Cu ratio. Table 2.9 (Tu et a!., 1994b) gives the reaction rate
parameters for the first order dehydrogenation reaction of ethanol over Cr:Cu = 0:40 and
Cr:Cu = 4:40 catalysts. There was a significant improvement in the reaction rate for the
optimum Cr to Cu ratio compared to that for pure copper.

Table 2.9: Reaction rate parameters for ethanol dehydrogenation (Tu et aI., 1994b)
Catalyst T (oq k (dm'gcaf'.h"') A (dm'gcaf' ,h"') E. (cal/mol)
Cr:Cu= 0:40 250 14.39 1.12*10° 12100
280 23.36 " "
310 51.97 " "

Cr:Cu- 4:40 250 47.58 8.39*10' 11510

280 70.12 " "
310 131.60 " "
" The same as above value

The catalyst activity for dehydrogenation is also strongly dependent on the copper
loading. Sivaraj and Kantarao (1988a) investigated the effects of the copper percentage
on y-alumina for 4 to 34 wt % Cu. They found an optimum copper surface area in the
region of 20-25 wt % Cu. The overall reaction rate dropped sharply, to about 16 wt %
copper for temperatures from 250 DC to 300 DC, thereafter it dropped only marginally and
became almost constant above 27 wt % Cu. Below 20 wt % copper, the acidic

72
y-alumina support caused a sharp shift in product distribution at 275°C. Dehydration
products increl).sed sharply and dehydrogenation products decreased sharply when the
copper content dropped below 20 wt %.

Kanoun et ai. (1991, 1993) determined the activities of V-Cu-Zn and Cu-Cr-Al catalysts
for the dehydrogenation of ethanol at 190°C. Changes in total catalysts surface area and
copper surface' area, with an increase in Cu percentage, were discussed. They also
calculated the molecules of acetaldehyde produced per surface copper atom per second
(turnover frequency or TOP) as indicated in Table 2.10. For the Cu-Cr-Al systems the
activation energy varied between 20 and 22 kcallmol, which was much higher than values
ofll to 12 kcallmol obtained by Tu et ai. (1994a).

Table 2.10: Activities and TOPs for ethanol dehydrogenation at 190°C Kanoun et al.
(1991, 1993)
Catalysts A (mol.kg TOF Catalysts A (mol.kg TOF
carl.h- I) x 10
3
cat-I.h- I) x 10 3
Cu 5.2 45 Cu-Cr 4.34 1.8
V-Cu 2.89 2.8 Cu-CrO.9AlO.l 4.32 1.9
Zn-Cu 1.59 6.5 Cu-CrO.7AlO.3 4.16 1.9
V-Zn 0 - Cu-CrO.5AIO.54.34 4.07 1.6
V-CuO.9ZnO.l 3.93 2.4 Cu4.25-CrO.3AIO.7 4.34 1.9
V-CuO.8ZnO.2 3.84 2.1 Cu-CrO.lAlO.9 4.25 1.5
V-CuO.6ZnO.4 3.98 2.1 Cu-Al 4.43 4.9
V-CuO.5ZnO.5 4 2.1
V -CuO.4ZnO.6 3.84 2.2
V-CuO.2ZnO.8 2.84 2.1
V -CuO.lZnO.9 2.43 1.8

Por the catalysts in mentioned inTable 2.10 the following two criteria apply:
• Cu/(Cr+Al) = 1
• V/(Cu+Zn) = 1

73
Cu-Al and Cu-Cr were the best performing binary systems where the ratio of Cu to other
metal is 1 to I. 'Varying the Cr to Al ratio did not have a significant effect on catalyst
activity. For V -Cu-Zn catalysts, an equal amount of copper and zinc combined with
vanadium gave the highest dehydrogenation activity.

2.12.5.3. Ethanol dehydrogenation kinetics

Franckaerts and Froment (1964) performed a kinetic study on the dehydrogenation of
ethanol between 225°C and 285 °C, with a CuO(94%)-CoO(5%)-Cr203(i %) on asbestos
catalyst. Only the main reaction was considered and the formation of ethylacetate or
other products were not taken into account, since the selectivity towards acetaldehyde
production was very high. They constructed initial reaction rate vs. flowrate curves and
used numerical differentiation to determine the kinetic mechanism. They found that the
following equation best fitted the data, which is similar to equation 2.19 but uses partial
pressure instead of concentration:

(2.30)

with A, R, S, W = ethanol, acetaldehyde, hydrogen and water.

The water term was added for when water might be present in the feed, but for modelling
purposes it was excluded. The calculated values for the constants (equations 2.31 to
2.34) between 225°C and 285 °C were:

(2.31)

(2.32)

(2.33)

74
InK =6850_ 7.18 (2.34)
S R oT

For equations (2.31) to (2.34) the units are:

k [moVg cat.h.atm]
K [atm· l ]
T [K]

Peloso et al. (1979) also studied dehydrogenation kinetics between 225°C and 280 °C
with the following catalyst: 41.2% CuO, 33.4% Cr203, 9.3% Si02, 3.3% Na20 and 12.8%
binder. Of the mechanisms investigated the dual site reaction mechanism was once again
rate controlling (see equation 2.19).

(2.35)

with A, R, S, W= ethanol, acetaldehyde, hydrogen and water.

The constants were:

Ink = - 5810.5 + 17.90 (2.36)

T

InKA = 1166.6 1.175 (2.37)

T

InKRS = 690.2 + 1.057 (2.38)

T

InK,q = 6189.1 +11.82 (2.39)

T

with,

k [mol/kg cat.h.atm]

75
Tu et al. (1994b) found the dehydrogenation reaction to be pseudo first order and of the
form:

(2.40)

with k = 1.12 X 106 exp(-12100IRT) for pure copper (2.41)

and k = 0.84 X 106 exp(-11510IRT) for Cu:Cr = 4:40 (2.42)
(k = dm 3 .gcaf i .h- i )

2.12.5.4. Oxidative dehydrogenation of ethanol

The oxidation of ethanol is usually carried out in the vapour phase, over silver (Faith et
aI., 1.957; Sanford, 1963) or copper-based catalysts. Commercial operating temperatures
range from 400°C to 500 °C with about 80% conversion per pass. Some studies have
been done on the oxidative dehydrogenation of ethanol in the liquid phase over Pt-
graphite catalysts at temperatures between 30°C and 50 °C (Vanden Tillaart et aI., 1994,
1996).

Cullis and Newitt (1956) investigated ethanol oxidation between 270°C and 370 °C.
Initially acetaldehyde was formed, but then further oxidation occurred, yielding
methanol, formaldehyde, methane and carbon oxides. Rao et al. (1991) tested a medium
pore, titanium silicate molecular sieve catalyst for the oxidative dehydrogenation of
ethanol. At 300°C the selectivity towards acetaldehyde production was over 90%, but
the conversion was only about 20%. At 400°C the conversion increased to above 90%,
but the selectivity dropped to below 70%. Quaranta et al. (1994) investigated the
possibility of using vanadium complexes for oxidative dehydrogenation.

76
2.12.5.5. Dehydrogenation of ethanol in a membrane reactor
Deng et al. (l995) modified alwnina membranes (500 nm pore size) with a y-alumina
layer containing Pd, Pt, Cu or Ni. The net pore diameter varied between 3 nm and 9 nm.
Hydrogen-argon selectivities for the Pt and Pd-based membranes were higher than the
Knudsen values and for the others lower than the Knudsen selectivities. Ethanol
dehydrogenation was studied in the temperature range from 250°C to 310 cC, employing
a Cu-P/Si02 catalyst. The acetaldehyde yield for the conventional reactor was slightly
below the equilibriwn value, while the values for the alwnina membrane were higher
than the equilibrium value. Cu and Ni-modified alwnina yielded results similar to
alumina membranes. The best results were obtained with the Pd and the Pt-modified
alumina membranes. Acetaldehyde yields were further improved by increasing the space
time and/or the sweep gas flow rate.

Raich and Foley (1998) studied ethanol dehydrogenation in a Pd tube with a wall
thickness of 76 11m. The operating temperature varied between 175°C and 225 °C and
the tube was packed with Cu or Pt on silica catalysts. The best results were obtained with
a copper on silica catalyst prepared by ion exchange followed by copper on silica
prepared by impregnation. The latter catalyst gave higher selectivity, but lower activity
and lower overall yield. Ethyl acetate was the main by-product at lower temperatures.
They compared a palladium reactor packed with copper on silica catalyst (prepared by
ion exchange) with a conventional reactor and obtained the following results: conversion
increased from 60% to 90% and selectivity from 35% to 70%.

2.13. ALCOHOL DEHYDRATION REACTIONS

Alcohol dehydration is an important side reaction of dehydrogenation reactions. The
acidity of the catalyst determines whether, and to what extent, dehydration of the alcohol
will occur and the reaction temperature dictates which dehydration reaction will take
place. It can either be uni-molecular dehydration to form olefins or bi-molecular
dehydration to form ethers. For primary alcohols, the di-ether is easily formed at lower

77
temperatures, while for secondary and tertiary alcohols, olefin formation dominates also
at the lower temperatures.

De Boer and Visseren (1971) observed that the rate constant for ethanol dehydration is
proportional to the amount of aluminium (acidic sites) on the catalyst's surface. For
ethanol the following was found:

CH3CH20H ..... CH2CH z + H 20 at temp. > 300°C (2.43)

2CH3CH zOH ..... CH3CHzOCH zCH 3+ H 2 0 at temp. < 260°C (2.44)

Several papers describe the dehydration of ethanol over zeolitic and non-zeoli tic
molecular sieves (De las Pozas et aI., 1993; Teo and Ti, 1990). For n-butanol
dehydration, the alumina percentage in silica-alumina catalysts has a significant effect on
the dehydration products, with a higher alumina percentage favouring di-butyl ether
formation (Berteau et aI., 1991). C4 dehydratio~ has been studied over various types of
zeolites (Makarova et aI., 1994; Williams et aI., 1991) and over y-aIumina (Lu et aI.,
1995).

2.14. SUMMARY
This chapter has described the various types of inorganic membranes that are currently
available. The main advantages and disadvantages of using inorganic membranes as well
as. the areas of application have been mentioned. Inorganic membranes were separated
into three classes i.e. dense, porous and composite membranes. The different separation
mechanisms through porous membranes were discussed. Palladium and palladium alloy
membranes were discussed in more detail, with specific attention being given to the
palladium-hydrogen system and the effects of alloying on palladium stability in
hydrogen. Different methods by which to prepare composite palladium membranes were
described. The more important methods were listed and compared. Electroless plating
was discussed in detail, covering substrate cleaning, pretreatment and the actual plating
process. Recent advances in electroless plating were mentioned. Techniques available

78
for alloying palladium-silver membranes and what deactivation processes occur during
operation have also been described.

The main advantages of using catalytic membranes are for manipulating the reaction rate
in dehydrogenation reactions and as gas distributors to improve mass transfer. The
various dehydrogenation reactions that have been studied in membrane reactors,
including both alkane and alcohol dehydrogenation, have been discussed. For alcohol
dehydrogenation, catalyst preparation techniques, catalyst reduction and catalyst
deactivation due to sintering and coking, were discussed. Dehydrogenation kinetics was
listed and the kinetic parameters for ethanol dehydrogenation, obtained by previous
researchers, were also summarised.

79
 3. EXPERIMENTAL PROCEDURES

The chapter on experimental procedures will be divided into four separate sections. The
first section (3.1) will focus on catalyst preparation and testing. The second section (3.2)
discusses the procedures followed for determining reaction kinetics. The third section
(3.3) will focus on Pd composite membrane preparation using a modified electro less .
plating technique and permeance testing of the membranes with hydrogen and nitrogen.
In the last section (3.4), the optimised catalyst and a suitable membrane are combined to
perform catalytic membrane reactor experiments. The contributions made to the field by
performing these experiments were discussed in the introduction (Chapter 1).

3.1. OPTIMISING CATALYST COMPOSITION

Catalysis experiments were conducted in two phases. A variety of catalysts were tested
at various operating conditions to determine the most suitable ones for ethanol and 2-
butanol dehydrogenation. Once the most suitable catalyst was established, an in depth
study was performed with that catalyst to determine the reaction kinetics of each reaction.
More than 25 catalysts were tested during the screening phase. The variables that were
investigated include:
• support type,
• copper loading,
• effect of additives, and
• catalyst particle size.

3.1.1. SUPPORT TYPES

Alumina and silica catalyst supports were supplied by Engelhard. Table 3.1 tabulates the
characteristics of the supports. The surface areas and average pore sizes are not the
values from their data sheets, but values determined from laboratory analysis
experiments. There was a small difference between the figures supplied on data sheets
and the figures determined in the laboratory for the surface area of the alumina support

80
« 10%). The difference was more significant for the silica support (about 50%). The 'as
received' pellets were crushed and then sieved to obtain different particle size fractions.

A commercial magnesium oxide (MgO) powder (surface area = 27.4 m 2/g) was mixed
with a binder and pressed into extrusions. The extrusions were heated to 1200 °C to
agglomerate the powder particles. The extrusions were then crushed and sieved into only
a 300 to 850 Ilm fraction. The surface area of the particles was 16.7 m 2/g. Most of the
experimental work focussed on the silica and the alumina supports. Only a few
experiments were conducted with MgO and thus not much time was spent on trying to
prepare a catalyst with a larger surface area.

Table 3.1: Characteristics of alumina and silica supports used

Alumina Silica
Product code AL-3996R C500-234
Purity > 99% alumina 99.5 % silica
max 0.3% alumina
Shape Extrusions spheres
Dimensions (mm) 3.5 outer diameter 3-5
1.5 inner diameter
Bulk crush strength (MPa) 0.80 1.40
Bulk density( glcm3) 0.50 0.45
2
BET surface area (m /g) 215 446
Average pore size (A) 132 88

3.1.2. CATALYST PREPARATION

Metals were deposited onto the silica and alumina supports via impregnation. The low
porosity of the MgO support made impregnation unsuitable hence adsorption was used
for depositing copper.

The MgO support was introduced into a flask containing a copper nitrate solution of a
specific concentration. The flask was placed on a magnetic stirrer and the solution stirred

81
for two hours. Thereafter the Cu-MgO particles were filtered, washed and dried at 90°C.
The catalyst was calcinated at 500°C and then reduced in hydrogen, in situ, at 350 °C for
two hours.

The silica and alumina supports were dried at 200°C for at least two to three hours and
then stored in a desicator. The dried supports were then placed in heated copper solutions
of different concentrations. The amount of solution required to impregnate a fixed mass
of either silica or alumina was experimentally determined. An excess of about 10%
solution was prepared in each case. The copper solution was kept warm on a hotplate
while the support was added. The support-solution mixture was stirred throughout while
adding the support particles. The hotplate was kept at about 80°C to evaporate the
remaining solution. The paste was stirred every few minutes. When all the water had
evaporated, the catalyst was dried in an oven at 120°C for at least four hours. The
catalyst was then ca1cinated at 500°C and reduced in situ in hydrogen at 350 °C for two
hours. When preparing Cu-Cr, Cu~Co and Cu-Ni-supported catalysts, copper nitrate was
mixed with the nitrate of the other metal in the same Cu to other metal ratio as required
on the support.

The catalysts used for determining catalyst stability and dehydrogenation kinetics were
prepared in a similar manner. The only difference was the reduction temperature.
Different reduction temperatures were investigated to determine the effect of sintering
during reduction.

3.1.3. SUMMARY OF CATALYSTS INVESTIGATED

Table 3.2 lists pure copper-supported catalysts and Table 3.3 lists the other catalysts that
were prepared and investigated in the screening stage to determine the most suitable
catalysts for ethanol and 2-butanol dehydrogenation. For all the catalysts, the eu
percentage is the Cu weight percentage of the total mass of the sample. It is the copper
mass divided by the support mass + copper oxide mass (pre-reduced state).

82
The copper percentage and percentage of the other metal were determined by atomic
adsorption (AA). The catalyst (0.100 g) was dissolved in warm aqua regia. When all
metal had dissolved, distilled water was added to yield exactly 1000 ml solution. The
ppm Cu reading ofthe solution gave the Cu percentage.

Table 3.2: Different copper loadings investigated for copper-supported catalysts

Silica support Alumina support MgO support
(300-S50 11m) (S50-11S0 11m) (300-S50 11m)
O%Cu O%Cu O%Cu
4.2% Cu S.8%Cu 8.4% Cu
9.2% Cu 13.2% Cu 13.7% Cu
11.7% Cu 18.5% Cu 20.8% Cu
15%Cu 24.0%Cu
18.6% Cu
27.7% Cu
33.5% Cu

Table 3.3: Other copper-supported catalysts

Silica support Silica support Alumina support
(300-S50 11m) Different fractions (S50-11S0 11m)
13.5% Cu; 1.5% Co (9:1) 14.1 % Cu (150-300 )lm) 17.7% Cu; 2.09% Cr (9:1)
13.5% Cu; 1.5% Cr (9:1) 15.0% Cu (300-850 )lID) 17.8% Cu; 1.73% Co (10:1)
13.5% Cu; 1.5% Ni (9:1) 13.1% Cu (850-1180)lID) 16.1 % Cu; 2.0% Ni (8:1)
14.25% Cu; 0.75% Co (19:1) 14.5% (3000 )lID)
14.25% Cu; 0.75% Cr (19:1)
14.25% Cu; 0.75% Ni (19:1)

3.1.4. CATALYST TESTING FOR DETERMINING THE OPTIMUM CATALYST

COMPOSITION
The experimental set-up used for catalyst testing was similar to that used for kinetic
experiments (see Figure 3.1). Heated entry and exit lines were used to vaporise the feed

83
 Thennocouple

4mmid
quartz
U-tube
Liquid

~ I )I~:r'e
collector
i and gas
vent Quartz wool
Infors AG Basel
Perfusion pump Tube furnace
Catalyst bed

Enlargement
Quartz wool

MFC: Brooks mass flow controller (5850 TR); 0-100 cm3/min

T: Thennocouple
.... Ball valve
~ Needle valve

Figure 3.1: Set-up used for testing the kinetics of the catalyst at the CNRS, France

84
and keep the products in the gas phase. A gas sample was extracted at the sample point
with a heated syringe. The syringe was kept inside a stainless steel tube and the
temperature of the syringe was controlled at about 100 DC. The first set of experimental
tests was performed to determine the optimal catalyst composition and was conducted at
the laboratories of the University of Stellenbosch (Stellenbosch, South Africa). The
kinetic testing was conducted at the laboratories of the IRC-CNRS (Institut de
Recherches sur la Catalyse, Centre National de la Recherche Scientifique) in Lyon,
France. Figure 3.1 is the set-up used at the CNRS. The set-up at Stellenbosch
University differed in the following way: Hastings flow controllers (HFC 202C) were
used in stead of Brooks, the inner diameter of the quartz tube was 8mm (10mm outer
diameter) and a Braun perfusion pump was used.
II
'Ii

3.1.4.1. Ethanol reaction I

,

The use of Ah03, Si02 and MgO supports were investigated for the ethanol reaction. All
the catalysts listed in Tables 3.2 and 3.3 were tested. Tables 3.4 and 3.5 list the matrix
of flow rates and temperatures used to test for Si02 and MgO (Table 3.4) and for Ah03
(Table 3.5) supported catalysts. The flow rates indicated in the tables are the liquid feed
flow rates of only the ethanol. In both cases the ethanol was diluted with nitrogen in the
molar ratio ethanol:N2 = 1:4 and then passed over the catalyst bed in the quartz tube. For
Si02 and MgO catalysts, 1 g of catalyst was used for each set of runs (6 temperatures at 4
flow rates = 24 runs). For AI20 3 catalysts, 3 g of catalyst was used for each set of runs (6
temperatures at 3 flow rates = 18 runs).

Table 3.4: Reactor conditions used to test Cu on Si0 2 and MgO supports in the ethanol
dehydrogenation reaction
0.6 mllh; 240 DC 1.6 mllh; 240 DC 3.2 mllh; 240 DC 6.4 mllh; 240 DC
0.6 mllh; 280 DC 1.6 mllh; 280 DC 3.2 mllh; 280 DC 6.4 mllh; 280 DC
0.6 mllh; 320 DC 1.6 mllh; 320 DC 3.2 mllh; 320 DC 6.4 mllh; 320 DC
0.6 mill); 360 DC 1.6 mllh; 360 DC 3.2 mllh; 360 DC 6.4 mllh; 360 DC
0.6 mllh; 400 DC 1.6 mllh; 400 DC 3.2 mllh; 400 DC 6.4 mllh; 400 DC
0.6 mllh; 440 DC 1.6 mllh; 440 DC 3.2 mllh; 440 DC 6.4 mllh; 440 DC

85
Table 3.5: Reactor conditions used to test Cu on AhO) supports III the ethanol
dehydrogenation reaction
1.6 mllh; 180°C 3.2 mllh; 180°C 6.4 mllh; 180°C
1.6 mllh; 220°C 3.2 mllh; 220°C 6.4 mllh; 220°C
1.6 mllh; 260°C 3.2 mllh; 260°C 6.4 mllh; 260°C
1.6 mllh; 300°C 3.2 mllh; 300°C 6.4 mllh; 300°C
1.6 mllh; 340°C 3.2 mllh; 340°C 6.4 mllh; 340°C
1.6 mllh; 380°C 3.2 mllh; 380°C 6.4 mllh; 380°C

3.1.4.2. 2-Butanol reaction

The 2-butanol reaction was tested using only Si02 and MgO supports. In both cases 1 g
of catalyst was used for each set of runs and the 2-butanol to nitrogen molar feed ratio
was 1 to 4. The reaction conditions are summarised in Table 3.6.. The flow rates
indicated in the table are the liquid feed flow rates of only the 2-butanol.

Table 3.6: Reactor conditions used to test Cu on Si02 and MgO supports in the ethanol
dehydrogenation reaction
1.6 mllh; 240°C 3.2 mllh; 240°C 6.4 mllh; 240°C
1.6 mllh; 270°C 3.2 mllh; 270°C 6.4 mllh; 270°C
1.6 mllh; 300°C 3.2 mllh; 300°C 6.4 mllh; 300°C
1.6 mllh; 330°C 3.2 mllh; 330°C 6.4 mllh; 330°C
1.6 mllh; 360°C 3.2 mllh; 360°C 6.4 mllh; 360°C
1.6 mllh; 390°C 3.2 mllh; 390°C 6.4 mllh; 390°C

3.1.5. PRODUCT ANALYSIS

The ethanol and 2-butanol reaction products were analysed with a HP G1800A gas
chromatograph, equipped with a mass spectrometer (for experiments conducted at
Stellenbosch University) and flame ionisation detector. A 50 m capillary column
(50QGI.5IBPI PONA from SGE) was used. Very good product separation was obtained
with an inlet temperature of 250°C, detector temperature of 300 °C and helium as carrier

86
gas. Response factors of the main products were calculated after inj ecting numerous
liquid mixtures and constructing response factor curves. Response factors were not taken
as linear functions, but were determined by fitting data of many (at least 6) different
compositions of each binary mixtures. The ethanol and 2-butanol response factors are
listed in Appendix B1 and B2. A short description is also given in Appendix B1 on how
the response factors were determined experimentally.

The following system was used for determining reaction rate kinetics at the CNRS in
Lyon, France. Carbon-containing products were analysed on a HP 5850 gas
chromatograph equipped with a FID detector. Two capillary columns, a 30 m HP
Innowax column and a 30 m HP Plot/Ah03 column, were used in series. Hydrogen
analysis was done on a similar GC with a TCD detector. A Porapak Q column and a
molecular sieve column were operated in series.

3.1.6. CATALYST CHARACTERISATION

BET (Brunauer-Emmett-Teller) and chemisorption experiments were performed on all
catalysts listed in Tables 3.2 and 3.3. No chemisorption was performed on Cu/MgO
catalysts, because of the low catalyst surface area. A Micromeritics ASAP 2010 was
used to determine both total catalyst surface area as well as copper surface area.
Determining BET surface areas presented no problems. Samples were dried at 300 DC in
situ overnight (14 to 16 hours). Determining copper surface areas was more difficult and
resulted in several problems. Prior to analysis, the copper was reduced, in situ, in
hydrogen for 2 hours at 260°C. Both CO and H2 were used as analysis gasses, but H2
gave more reliable and reproducible chemisorption results. The copper and catalyst
surface areas are listed and discussed in Chapter 4. Further catalyst characterisation was
performed only on the catalyst used for kinetic testing. This included transmission
electron microscopy (TEM), X-ray diffraction analysis (XRD) and total organic carbon
analysis (TOC) of samples used.

87
3.2. KINETIC TESTING
The kinetics of the dehydrogenation of ethanol and the dehydrogenation of 2-butanol
were studied using a 14.4 wt % copper on silica catalyst. This was found to be the
.optimal catalyst from experiments conducted in the catalyst screening phase. All catalyst
testing was performed with the set-up shown in Figure 3.l.
To determine accurate kinetic data, interphase and intraparticle mass transfer resistance
needed to be eliminated or minimised. The data must preferably be gathered in the flow
regime, free of interphase mass transfer resistance. Intraparticle mass transfer resistance
was minimised by using small catalyst particles. A 300-425 Ilm particle size was used
for all experiments.

Several other important factors had to be taken into account when designing an
experiment for determining the kinetic parameters of a reaction. They were:
• The reactor had to be operated as a differential reactor (see Chapter 5) to be able to
accurately calculate reaction rates. A differential reactor is similar to a plug flow
reactor, but the overall conversion is kept low (typically less than 10%) by using very
small amounts of catalyst or very high feed flow rates.
• The catalyst bed had to be at a constant temperature.
• If the reaction rate is independent of the feed flow rate, but the conversion is still
high, then the catalyst mass can be changed. The amount of catalyst can be reduced,
while maintaining the same flow rate, thus lowering the conversion.

3.2.1. ETHANOL DEHYDROGENATION

Catalyst stability tests were performed to determine the temperature range free of
significant catalyst deactivation, and to determine the deactivation mechanism.

3.2.1.1. Catalyst deactivation testing

An unused 14.4 wt % copper on silica catalyst was reduced at three different
temperatures to determine the effect of reduction temperature on catalyst stability. For
deactivation testing, the reaction was continued for 24 hours at the setpoint temperature

88
and the conversion documented with time increase. An undiluted 99.8% ethanol (from
Prolabo) feed was used.

Table 3.7 indicates the different reduction temperatures evaluated. After these initial
experiments, all further catalyst reductions were perfonned at 255°C for two hours in
hydrogen (25 cm3/min). The catalysts were heated up at 8 °C/min in 50 cm3/min of
nitrogen before being reduced at the reduction temperature.

Table 3.7: Different combinations of reduction and reaction temperatures studied

Reduction T (0C) 400 400 340 340
Reaction T (0C) 400 255 340 255

Table 3.8 summarises all the stability tests perfonned on the dehydrogenation of ethanol.
Catalyst (a) is the standard 14.4 wt % Cu on silica catalyst, while catalyst (b) is a
13.5 wt % Cu, 1.0 wt % Cr and 0.5 wt % Co on silica catalyst. Cr and Co are often added
to copper to reduce sintering at a high temperature. Both catalysts were prepared in
South Africa from copper nitrate as source. The support used was a silica support from
Engelhard. Co and Cr were obtained from their nitrate salts.

Table 3.8: Summary of catalyst deactivation tests for ethanol dehydrogenation

Catalyst (a)
Reduction T ceC) 400 340 280 220
Reaction T (0C) 255 255 255 255

Catalyst ~}
Reduction T ceC) 400 340 280 -
Reaction T ceC) 255 255 255 -

The 14.4 wt % Cu on silica catalyst (catalyst (a» was reactivated after 24 hours and

89
48 hours of operation at 400°C. This was done to determine whether coking, sintering or
both were the dominating deactivation mechanism. Reactivation consisted of the
following:
• Oxidising the catalyst at 400°C in oxygen for 4 hours,
• Cooling the catalyst down to 255 °C in oxygen,
• Reduction for 2 hours in hydrogen, and
• Heating up to 400°C in nitrogen.

The reaction was then continued at 400 °C for the next 24 hours.

3.2.1.2. Characterising deactivated catalysts

Catalysts were well dried in nitrogen for 4 hours at the reaction temperature after the
24 hour stability tests. The following techniques were then employed to determine
whether sintering and/or coking was the major catalyst deactivation process for copper on
silica catalysts:
• XRD analysis with a Bruker AXS D5005 X-Ray Diffractometer.
• TEM analysis with a JEOL 2010 using a LaB 6 source and 200 kV ultra high
resolution (cs = 0.5 mm). EDS was performed with an Isis Link Oxford. Samples
were prepared by ultrasonic dispersion in ethanol, followed by deposition on an Al
grid (200 mesh) and covering with carbon.
• Total organic carbon (TOC) determination with a Shimadzu SSM 5000A.

3.2.1.3. Determining the mass transfer limited regime

The ethanol reaction rate was studied from 200°C to 300 °C, at intervals of 25°C. At
each temperature a series of experiments was performed to try and achieve a constant
reaction rate as a function of flow rate. This was achieved for temperatures up to 250°C,
but for higher temperatures some difficulty in eliminating interphase mass transfer
resistance was experienced. Table 3.9 summarises the catalyst sample masses and flow
rates tested at each temperature.

90
Table 3.9: Conditions investigated for detennining interphase mass transfer limited
regime for ethanol dehydrogenation
Reaction T (OC) Catalyst mass (g) Min feed flow (mllh) Max feed flow (mllh)
200 0.35 1 14

225 0.35 1 16

250 0.20 2 20
275 0.10 2 20
300 0.10 2 25

3.2.1.4. Determining kinetic parameters

Ethanol was co-fed to the catalyst with nitrogen (an inert gas), hydrogen or acetaldehyde.
The nitrogen and hydrogen concentrations were varied by changing the flow rate of the
mass flow controller (see Figure 3.1). In order to keep the total molar feed rate to the
catalyst and the space time constant for all the runs, the ethanol feed was changed
accordingly. Mixtures of acetaldehyde and ethanol, with varying acetaldehyde mass
percentages, were used to test the reaction at different acetaldehyde feed concentrations.

When co-feeding ethanol with hydrogen or nitrogen, the products were analysed with a
FID detector and reaction rates calculated from acetaldehyde production. A TCD
detector was used to determine the amount of hydrogen produced when ethanol was co-
fed with acetaldehyde. In the latter case, hydrogen production was used to detennine
reaction rates.

The availability of equipment limited the tests which could be performed at feed rates
higher than 14 mllh, mainly due to the size of the mass flow controller and difficulty with
temperature unifonnity in the catalyst bed at higher flow rates. All experiments were
thus done at feed rates of 14 mllh and reaction rates had to be adapted, using curves
constructed from data in Table 3.9, to compensate for mass transfer resistance. For the
ethanol dehydrogenation reaction, the reaction rate was determined at 70 different
conditions. At least three injections were made into the GC at each condition and the

91
average value was used for modelling. The conditions for determining the kinetic
parameters were as follows:
• T = 200, 225, 250, 275, 300°C, and

• at each temperature:
• N2 molar % in feed = 15, 30, 45, 57 mole %,
• H2 molar % in feed = 15, 30, 45, 57 mole %, and
• Acetaldehyde mass % in feed = 4.4, 11.6,21.3,29.8,50 mass %.

Every day a new catalyst sample was used for experiments. The reaction rates for the
catalyst (at the operating temperature) were determined using pure ethanol as feed. All
data was determined relative to a set of reference values. Reference values at the same
temperature were averaged and each data set normalised, relative to the global average at
each temperature. This yielded more accurate and consistent data for modelling.

3.2.2. 2-BUTANOL REACTION

The same equipmerit and steps used for studying the ethanol reaction were used for
studying the 2-butanol reaction. The catalyst was also a 14.4 wt % copper on silica
catalyst. Deactivation was determined at 250 °C and 310 °C. Further kinetic
experiments were performed at 190,220,250 and 280 °C. Table 3.10 is a summary of
reaction conditions investigated to determine the interphase mass transfer limited regime
for 2-butanol dehydrogenation.

A feed flow rate of 12 mllh was used for all further experiments with 2-butanol. The
following mixtures were fed to the catalyst at each of the reaction temperatures:
• N2 molar % in feed = 10, 25, 40, 55, 70 mole %,
• H2 molar % in feed = 10,25,40,55,70 mole %, and
• Methyl ethel ketone mass % in feed = 6.7,10.6,20.1,40.3,62.8 mass %.

92
Table 3.10: Conditions investigated for determining interphase mass transfer limited
regime for 2-butanol dehydrogenation
Reaction T (0C) Catalyst mass (g) Min feed flow (ml!h) Max feed flow (ml!h)
190 0.15 2 12
220 0.15 2 16
250 0.075 2 16
280 0.075 2 16

Data was gathered in a similar marmer to that for ethanol dehydrogenation. Reaction
rates were determined relative to a set of reference values, which was then used to
calculate more accurate and consistent data for modelling. A GC, with a TCD detector
and packed columns, was used for determining hydrogen production rates when MEK
was co-fed with 2-butanol. For all other experiments, products were analysed with a FID
detector and capillary columns.

3.3. MEMBRANE PREPARATION AND TESTING

This section describes all aspects of pure Pd composite membrane preparation and
testing. Preparing and testing Pd alloys will also be discussed.

3.3.1. MEMBRANE SUPPORT

The membranes employed in this study were purchased from SCT (Societe des
Ceramiques Techniques) and the membranes consisted of three a-alumina layers with
decreasing particle sizes. The structure is illustrated in Figure 3.2.

The membranes had a length of 250 mm, an outside diameter of 10 mm and an inside
diameter of 7 mm. The final layer had a pore size of 200 nm. Further characteristics
regarding pore diameter, thickness and porosity of each layer are presented in Table 3.11.

93
 4: Pd rmallayer

3: Alfa alwnina layer 3

2: Alfa alwnina layer 2

1: Alfa alwnina layer 1

Figure 3.2: SCT membrane structure

Table 3.11: Membrane layer characteristics of a SCT membrane

Pore diameter Layer thickness a-alumina particle Porosity (e)
(11 m ) (11m) size (11m)
Layer 1 11 1500 53 0.26
Lay~r 2 0.64 40 4 0.24
Layer 3 0.2 20 0.9 0.25

constant 30 min
12200 C - - - - - - - - - - - - - _..- : - - - - - - - - ,

llOO ·C
oven swi1l:hed
off

20 0 C/min

25°C

Figure 3.3: Curing process for enamelled membranes

94
The outside membrane ends had to be sealed with an enamel supplied by SCT prior to
plating in order to achieve proper membrane reactor sealing. This process fiHed the
outside membrane pores with a non-porous material, thus preventing gas leaks along the
outside membrane surface during the testing stage. Enamel was applied along a length of
10 mm at the membrane ends by dipping the membranes in the enamel slurry. Enamelled
membranes were then placed in a high temperature furnace (from Vecstar Furnaces) .and
cured according to the procedure suggested by SCT and indicated in Figure 3.3. Two or
three layers of enamel were applied on the membrane ends to obtain good sealing.

3.3.2. COMPOSITE MEMBRANE PREPARATION

A modified electroless plating process was used for preparing composite membranes.
Membranes were cleaned and pretreated prior to electro less plating. After plating, the
membranes were stirred in ammonia and dried.

3.3.2.1. Electroless plating pre-treatment

Upon completion of the enamelling process, no further cleaning was performed and the
masses of the membranes were recorded. Some membranes were supplied with
enameHed endings. Those samples were stirred in distilled water for 30 minutes using a
RW II basic stirrer from lKA Labortechnik. The samples were then placed overnight
(14 to 16 hours) in an oven at 200°C. Their masses were recorded the foHowing
morning, after cooling of the membranes in the oven down to 60°C.

Prior to electroless plating, the membrane surface needed to be activated to provide

catalytic centres for the plating process. A two step process using a palladium salt and a
tin chloride salt was used. Previously, it was found that the conventional composition of
the pre-treatment solutions deposited small amounts of tin. Keuler (1997a) found about
0.2 to 0.4 % Sn (with particle induced X-ray emission) in a 5 micron Pd film prepared
with the conventional solutions. For thin Pd films of less than 2 microns this amount
would become more significant and might cause a decrease in the hydrogen permeance
rate through the film. A higher Pd concentration and a lower Sn concentration were
chosen by the author for pre-treatment and are indicated in Table 3.12. The conventional

95
composition of the pre-treatment solution was similar to that of Shu, Grandjean et al.
(1993).

Table 3.12: Compositions of pre-treatment solutions

Conventional composition This study
Sensitisinl: solution (l!er litre}
35 wt % HCI (ml) I -
SnCb.2H 20 (g) I 0.45
Activation solution (l!er litre}
35 wt % HCI (ml) I -
10 wt % (NH3)4Pd(N03)2 (g) 1.5 -
23 wt %PdCh - 1.4

Sn to Pd molar ratio 8.8 1.1

The outside surface of the membrane tubes were wrapped with PTFE tape so that only the
insides of the tubes would be catalysed. The set-up for pre-treatment is shown in Figure
3.4. The stirrer was a RW 11 basic stirrer from IKA Labortechnik (0 to 2000 rpm) and
the stirring speed was set at about 1200 rpm.

\-=+__ Rotating
device

_ _ _ Membrane

_ _ _ Pretreatment
solution

Fil:ure 3.4: Pre-treatment set-up

96
The procedures and their sequence for pre-treatment are tabulated in Table 3.13.
Approximately 270 ml of each solution was used. Fresh tin solution was prepared for
every membrane, while the Pd solution was changed after every three membranes
catalysed.

Table 3.13: Stirring sequence and times used in pre-treatment

Pd solution Distilled water Sn solution Distilled water
Repeat 3 times 10 minutes Dip into 10 times 10 minutes Dip into 10 times
Repeat 3 times 5 minutes - 5 minutes Dip into 10 times

After pre-treatment, the teflon tape was removed and the membrane stirred in clean
distilled water for an additional half an hour to remove any solution in the membrane
pores. The membrane was then placed in an oven at 200°C overnight and the mass
recorded the next morning after cooling: The mass increase varied between 13 and 16
mg for different membranes prepared using a 0.45 g per litre SnCh sensitising solution.
Some membranes were pre-treated with a 0.25 g per litre SnCh sensitising solution, in
which case the mass increase was about 8 to 10 mg.

3.3.2.2. Electroless Pd plating

A 2.00 gliitre (2000 ppm) Pd solution was used for electroless plating. The composition
of the solution is given in Table 3.14. The Pd solution, without the hydrazine in it, was
allowed to stand (stabilise) for at least 12 to 16 hours prior to plating. Hydrazine was
added just before plating and only to the solution that was used for plating. For some of
the initial membranes that were prepared at the laboratories of the CNRS (Lyon, France),
a three times higher initial hydrazine concentration was used. The values in Table 3.14
represent the composition of the plating solution in the latter stages of the project, which
produced good quality and reproducible coatings. The hydrazine concentration in
Table 3.14 represents the initial amount that must be added to the palladium solution on a
per litre basis. Table 3.15 lists how the hydrazine was added to the palladium solution.

97
Table 3.14: Composition of the Pd plating solution per litre (for 2.00 gllitre Pd in
solution)
(NH3)4PdCIz.H20 (g) 4.94
28 wt % Ammonia (ml) 400
EDTA(g) 80
35 wt % hydrazine (ml) 0.65 (hydrazine:Pd - 0.35:1)
increased with time
Temperature (oq 71-73

Three different methods for feeding plating solution to the inside of the membrane tube
were tested:
I. Plating solution was pumped continuously through the inside of the tube at flow rates
varying between 90 and 120 ml per hour.
2. A membrane was covered on the outside with teflon tape and stirred in the plating
solution.
3. A batch process was used in which between 10 and 12 ml plating solution was
introduced into the sealed tube at a time and allowed to react for a fixed period of
time.

The third process was chosen for preparing all further membranes (see discussion in
Chapter 6). The plating set-up is shown in Figure 3.5. The membrane was sealed in a
teflon reactor with O-rings. The reactor had a single shell side outlet allowing for a
vacuum to be pulled on the shell side. A defect-plugging technique was developed to
produce thin films « 2 microns). An initial I micron Pd base was deposited on the inside
of the membrane tube using 35 ml plating solution without any vacuum applied. The
membrane was closed off at the bottom with a 10 mm silicon tube which was closed at
one side. A silicon tube of about 15 cm was also placed over the membrane at the top.
Between 11 and 12 ml plating solution was introduced into the membrane tube fixed in
the reactor at a time.

"'i

98
 Control volve
ShellsUie
vacuum --<>1<:1-----, ,------,

Liquid collector

1IIi"<--+--. Phililg solution

O-rings

Sealed membrane

Teflon plaililg
reacior
Water bath
(71.73 ·C)

Closed off
membrane

Figure 3.5: Electroless Pd plating set-up

Three to four plating sessions were performed with the same plating solution. Repeated
plating sessions ensured that all the Pd in solution was deposited on the membrane. The
hydrazine concentration was increased after each plating session to compensate for
thermal decomposition of hydrazine. The plating procedure as outlined in Table 3.15, is
essentially a batch process repeated several times. When preparing a 1 micron film, each
session in Table 3.15 was repeated 3 times (11.5 ml *3= 34.5 ml solution). The first
0.15 ml of 1.75 wt % hydrazine that was added to the 1l.5 ml plating solution (see Table
3.15) is equivalent to 0.65 ml of 35 wt % hydrazine per litre of plating solution, as
mentioned in Table 3.14.

Membranes were then cleaned and dried overnight before the next layers were applied.
The effect of Pd film thickness on hydrogen and nitrogen permeances was investigated.
For thicker films (> 2 microns), a thicker initial base was applied, but for membranes of
between 1 and 2 microns total thickness, the first layer was always about 0.9 to 1.1
microns. Membrane cleaning or post-treatment will be discussed in the next section.

99
Table 3.15: Plating procedure used for producing Pd films
Plating session Reaction time for Vol. 1.75 wt %
11.5 ml plating hydrazine added for
solution (min) 11.5 ml solution (ml)
I 20 0.15
2 20 3 drops
3 20 0.5

After the initial Pd layer was deposited, an additional one, two or three layers were
deposited to obtain the final product. The thickness of each layer depended on the
required thickness of the final Pd film: for example a film with a final thickness of
1.5 !lm required two extra layers of 0.25 !lm each. After the second layer, the membrane
was once again cleaned and dried overnight before applying the third layer.

For the second and third Pd layers, a vacuum was applied on the shell side of the teflon
reactor. Pd solution will concentrate in the more permeable or defected areas in the
membrane. More plating will occur in weakly plated (defected) areas, .film defects will
be covered with palladium and a film with less defects will result.

Pd membrane preparation focussed on preparing a thin film of thickness less than 2 !lm.
Some films of between 2 and 5 microns were also prepared to study the effect of film
thickness on hydrogen and nitrogen permeance through the film.

3.3.2.3. Membrane cleaning

The same equipment used for pre-treatment, was used for membrane cleaning (Figure
3.4). The membrane was removed from the reactor and placed in a cylinder containing
270 ml 15 wt % ammonia solution. The membrane was stirred at a rate of 1200 rpm for
one hour. This process was repeated with fresh ammonia solution and finally the
membrane was stirred in a similar volume distilled water for half an hour. The
membrane was then placed overnight in an oven at 240°C. In the latter stages of the

100
work, the membranes were dried at 100°C and again (two or three times) stirred in
ammonia solution and water before drying them at 240°C. This was an attempt to try
and extract more EDTA trapped in the membrane pores.

3.3.2.4. Electroless silver plating

The same equipment and procedures used for palladium plating was also used for silver
plating. Two or three silver layers were deposited onto the palladium layer in sequential
steps. The mass of silver plated was determined from the membrane mass increase after
each plating sessions. After each layer of silver had been deposited, the membrane was
washed and dried overnight at 240 DC. If the amount of silver was less than the desired
amount (less than 20 wt % Ag in the Pd-Ag film), an additional layer was deposited. The
composition of the silver plating bath is given in Table 3.16.

Table 3.16: Composition of silver plating bath per litre of plating solution
Ag (ppm) 1000
AgN03 (g) 1.576
EDTA(g) 40
Ammonia (28%) (ml) 200
Hydrazine (3.5 wt %) (ml) 8.50

The hydrazine concentration (for Ag plating) was increased with time, similar to the
procedure for Pd plating (see Table 3.17). For every 11 ml fresh plating solution, the
procedure in Table 3.17 was performed. Generally, a maximum of about 75% of silver
in solution was deposited on the palladium. An excess of 30% of the required amount of
silver that needed to be deposited on the membrane was used in solution for plating.
Silver plating was performed and if the deposited silver was not sufficient, a second layer
was deposited. For the second layer, the initial silver plating solution was diluted to
ensure a volume of 11 ml plating solution. This was necessary because if the silver
plating solution (original) was less than 10 ml it would not wet the entire membrane
surface and thus not plate over the entire surface.

101
Table 3.17: Plating procedure used for producing Ag films
Plating session Reaction time for Vol. 0.35 wt %
11 ml plating hydrazine added for
solution (min) 11.5 ml solution (ml)
I 20 0.4
2 20 0.4
3 20 0.2 ml, 1.75 wt %
hydrazine

3.3.3 DETERMINING PALLADIUM FILM THICKNESS

Two methods were used to determine the amount of Pd deposited on the membrane
supports. The membranes were weighed after pre-treatment and drying at 200°C
overnight to get the initial mass. The membranes were weighed again, after testing of the
membranes were completed, to get the [mal mass. The difference between the initial
mass and final mass was taken as the amount ofPd deposited.

ICP (with a Spectroflame Modula from Analytical Instruments) analysis of the plating
solution was performed after plating to determine the amount of palladium deposited.
Since the initial mass of the plating solution was known, the mass of the deposited Pd
could be calculated. The average of this mass and the measured mass was used to
calculate Pd film thickness. The calculated film thickness is an average thickness value.

3.3.4. MEMBRANE TESTING

Membrane permeance testing was performed in the reactor shown in Figure 3.6. The
reactor was made from stainless steel and graphite rings were used for obtaining an
effective membrane to reactor seal. The rings had dimensions of 10.4 by 17.9 mm, a
thickness of 5 mm and a density of 1.6 glcm3. They were purchased from Coltec
Industries (Le Carbone Loraine). Nuts turned into the membrane ends, pushing two
fittings onto the graphite rings. Those fittings slid over the membrane ends and were cut
with an angle at the bottom edge to be able to wedge into the graphite. When the nuts
were tightened, the fittings forced the graphite rings forward and sideways, pushing them

102
against th~ enamelled membrane endings. If the nuts were sufficiently tightened, a very
good seal was obtained. Over tightening could result in either breaking or cracking of the
membrane.

Sweep gas Sweep gas

Membrane (tube side)
exit ~ ===~ ~ entry

r
Tube side
--t
exit

Tube side
Graphite s ea1 Shell side Enamelle d endings
entry
Figure 3.6: Membrane reactor used to test the membrane permeance

Some difficulty was, however, still experienced when trying to obtain good reactor to
membrane seals. The main reason was that the enamel on the outside membrane surface
was not always of uniform thickness. The equipment shown in Figure 3.7 was used to
minimise the leak resulting from membrane sealing.

Nitrogen
Exit side
closed off
Membrane sealed
at both ends

MFM: 0·5.00 cm3/min Brooks mass flow meter

MFM: 0-100.0 cm3/min Brooks mass flow meter
pc: 0-500.0 mbar pressure displayer (WEST 6100)
and controller
P: Keller probe for pressure measurement

3 way valve

Figure 3.7: Set-up used for membrane testing at room temperature

103
The membrane was placed inside the reactor, with graphite rings at the edges (see
Figure 3.6) and the nuts tightened moderately. The reactor (with the exit side closed off)
was then connected to a pressure controller and two mass flow meters as indicated in
Figure 3.7. The pressure controller was set at 400.0 mbar and the value of the mass flow
meters monitored while reactor nuts were tightened. Tightening was stopped when, upon
further tightening of the nuts, there was no further decline in the reading of the mass flow
meters. The nuts on opposing sides were not necessarily turned or tightened equally, to
obtain the best seal. Sometimes the reactor had to be re-opened, the membrane shifted
slightly and nuts re-tightened to obtain a good seal.

Figures 3.8 and 3.9 show the. equipment used for high temperature membrane testing
with hydrogen, nitrogen and ethanol as feeds, respectively. In both instances the reactor
was wrapped with heating wire and insulated. A thermocouple was placed in the centre
ofthe membrane tube to record the temperature.

For hydrogen and nitrogen testing (Figure 3.8), one of the two shell side tubes of the
reactor was closed. The reactor was operated in the dead end mode, in other words, the
exit tube side was closed and the feed gas forced through the Pd film. The temperature
inside the reactor was varied between 330°C and 450 °C, using a temperature controller.
The flow rate of the permeated gas was measured using two bubble flow meters. A 0 to
100 ml flow meter was used for hydrogen measurements and a 0 to 4 ml flow meter for
nitrogen. The effect of differential pressure on hydrogen and nitrogen permeance was
studied. The mass flow controllers were set on different flow rates and the differential
pressure recorded. Initial testing was conducted at the laboratories of the CNRS (Lyon,
France). For hydrogen permeance, the maximum differential pressure that could be
tested was limited by the mass flow controller (0 to 600 cm'/min). For films in the order
of 1.5 flm, this maximum differential pressure was less than 100 mbar at 450°C. The
pressure probe had a maximum measuring ability of 2000 mbar, which was the limit for
testing nitrogen permeance.

104
 Closed

Hydrogen r+_i~~:-~T:cl
I I I

Nitrogen
I ,
L.. ______ ..J
Tube side closed
Heated zone
with sealed Shell side
membrane exit

MFC: Brooks mass flow controller (5850 TR) for hydrogen; 0-600 cm3/min
MFC: Brooks mass flow controller (5850 E) for nitrogen; 0-100 cm3/min
P: Keller tube side pressure probe (0-2000 mbar)
with WEST (8100) displayer
T: Thermocouple
TC: Temperature controller for heated zone
BFM: Bubble flow meter for nitrogen (0-2 ml)
BFM: Bubble flow meter for hydrogen (O-IDO ml)
...... Ball valve
~ Needle valve

Figure 3.8: Set-up used for high temperature (> 300°C) hydrogen and nitrogen permeation testing
 Ethane sweep gas
r-----iT,ci
_1 __ ,
I
Safety
pressure
gauge
L.. _ _ _ _ _ _ l.-_--I~ Tube side closed
Autoclave with
30 mI ethanol Heated zone
with sealed Exit to vent
temp erature membrane '-------ll GC
control

MFC: Brooks mass flow controller (5850 TR) for ethane; 0-50 cm3lmin
P: Keller tube side pressure probe (0-2000 mbar) with WEST (8100) displayer
T: Thermocouples
TC: Temperature controller for heated zone
..... Ball valve •
r:ki Needle valve
** All tubing containing ethanol is heated to 130°C using heating wire .*

Figure 3.9: Set-up used for testing ethanol penneance

106
Further testing on both pure palladium films and Pd-Ag alloy films was conducted at the
University of Stellenbosch, South Africa. At Stellenbosch, Hastings flow controllers
(HFC 202C) were used instead of Brooks and they had a maximum capacity of 1000
cm'/min. The pressure probe also had a maximum of2000 mbar.

The apparatus used for ethanol permeance testing is shown in Figure 3.9. An ethane
sweep gas was used and fed to a gas chromatograph after moving pass the membrane. By
measuring the ethanol content in the sweep gas, the ethanol permeance could be
calculated. All lines containing ethanol were wrapped with heating wire and insulated.
The temperature of the heated lines was kept between I 10°C and 130°C to prevent any
condensation in the lines. An autoclave was used to feed the ethanol to the reactor. The
autoclave was well insulated and heated on a temperature controlled hotplate from lKA
Labortechnik. The desired tube side pressure could be obtained by correctly setting the
temperature of the hotplate.

The autoclave was filled with 30 ml 99.8% ethanol. The testing pressure was stable and
fluctuated by less than I %. The slow ethanol permeance ensured a constant tube side
pressure.

After testing, the Pd membrane was kept in nitrogen at 330°C for 2 hours and then
cooled at a rate of2 °Clmin to room temperature (also in nitrogen).

3.3.4.1. Reactor temperature profiles

Temperature profiles across the length of the reactor were determined at various oven
temperatures. A constant temperature over the length of the reactor was necessary to
assume isothermal conditions for modelling purposes. The heating wire, around the
reactor and insulation material, was adjusted until satisfactory profiles were obtained.
The final profiles are presented in Figure 3.10.

107
 000,--------------------------------,

· ................ . ....
Avenlge." 301 "C
. ... - ........... - '--""'"'' .....

Awnlge • 21M ·c
_crv.n: 100'C . - - - - - - - - - - - . - - - - - - - - - - - ... _____ _
- . - o..n._200"C
...... o..n_300'C
100 -"... 0\I..,=400'Cr+--------+--------+-_ _ _ _ -l
_o.toro:440'C

o~-- __----~----~--__----~----~
o 2 4 6 8 10 12
Distance from reactor centre (em)

Figure 3.10: Reactor temperature profiles at different oven temperatures

There were two thermocouples for measuring temperature. One was situated inside 'the
membrane tube and one was situated on the outside of the reactor, next to the reactor
wall. The temperature controller was connected to the thermocouple on the outside of the
reactor wall, since that temperature provided for more stable temperature control than the
thermocouple inside the membrane tube. The temperature inside the membrane tube was
measured, but not used for control. The temperature difference between the two
thermocouples was taken into consideration when setting the reaction temperature. In
other words, to control the reaction temperature at for example 300°C, the oven was
adjusted by the difference and set at 291 °C. At 300°C and below, the temperature
profiles were excellent from the centre of the reactor up to a distance of 9 cm from the
centre. The variance from the average temperature was less than 5°C. In the last 3.5 em,
the temperature dropped by between 10 and 13 °C. The total membrane length was
25 cm. Catalytic membrane reactor experiments were performed below 300°C and the
variance in reactor temperature was small enough to assume isothermal conditions. At
400 °c and above, the difference between the oven and reactor temperatures increased.
The variance around the average was also larger and increased to between 10 and 15°C.

3.3.5. MEMBRANE HEAT TREATMENT AND REDUCTION

The membrane reactor was heated from room temperature to 450°C at 2.5 °C/min in a
nitrogen atmosphere. The membrane was then reduced in a hydrogen atmosphere at

108
450°C for one and a half hours. A very high EDT A concentration (80 g per litre) was
used for plating. The cleaning process and overnight oxidation at 240°C did not remove
all the precursor. It was found that even after reduction not all the carbon in the
membrane pores was removed either by oxidation or by thermal decomposition. The
heating process was then changed and it produced membranes with higher hydrogen
permeances and selectivities. The process was changed as follows:
• Heat the membrane in nitrogen from room temperature to 320°C, at 2.5 °C/min.
• Switch from using nitrogen to oxygen and force 10 cm3 /min oxygen from the tube
side through the membrane pores to the shell side. Cary out oxidation for 2 hours.
• Switch back to nitrogen and heat from 320°C to 450 °C, at 2.5 °C/min.
• Reduce at 450°C in hydrogen for 1.5 hours.

3.3.6. P ALLADIUM-SIL VER ALLOYING

Silver was deposited on the palladium coating to form two separate layers. The hydrogen
and nitrogen permeances for pure palladium, palladium-silver without heat treatment and
palladium-silver with heat treatment were compared. Different factors were investigated
during the alloying procedure to obtain a thorough understanding of what takes place on
the membrane surface. Since the number of membranes available for testing was limited,
not all the variables could be tested in detail, but a significant amount of data was
nonetheless obtained. The variables that were important during alloying included:
• Heating procedure: Membranes can be heated in situ in the reactor (fixed at both
ends) or they can be heated in free mode in a tube furnace. In the latter method the
membrane is not fixed to a reactor and free to expand along its axis.
• Heating environment: Different heating environments, for example hydrogen, argon
or nitrogen, may have different effects.
• Alloying temperature.
• Alloying time.

The alloying experiments were performed on membranes with metal coatings of similar
thicknesses. This was important, to be able to compare results of different membranes.
Both heating methods were investigated. Membranes were placed either in a constant

109
temperature tube furnace or in the reactor (Figure 3.6) and heated at a rate of 1.5 DC/min
to the required temperature. Initial experiments were performed with the membranes in
the free mode under the following conditions:
• argon and hydrogen atmospheres,
• alloying temperatures ranging from 520 DC to 600 DC, and
• alloying times up to 150 hours.

The amount of data obtained from membranes heated in the tube furnace, was limited.
The oven had to be cooled down (at 1.5 DC/min) to room temperature after each
experiment, and the membrane placed in the reactor for testing, before data could be
gathered. Not only did the thermal cycling weaken the coating, but the data was also of a
discontinuous nature.

Most of the alloying was performed in situ in the reactor. That allowed for continuous
monitoring of the process. Permeance and selectivity data could be obtained at any time.
The following conditions were tested during alloying in the reactor:
• argon, nitrogen and hydrogen atmospheres,
• alloying temperatures from 500 DC to 600 °C, and
• alloying times from 10 hours to more than 100 hours.

3.3.7. MEMBRANE CHARACTERISATION

Besides permeance testing, the palladium plated SCT membranes were studied using
scanning electron microscopy (SEM). SEM was conducted at the Microscope Unit,
University of Cape Town, with a Cambridge Stereos can 440. For top view images, the
samples were mounted on aluminum stubs and then gold plated. For cross-section views,
the membranes were set in a resin, sanded to expose fresh cross section and then polished
with an alumina slurry.

110
3.4. CATALYTIC MEMBRANE REACTOR EXPERIMENTS
Catalytic membrane reactor experiments were performed in a set-up similar to that shown
in Figure 3.1. The quartz tube in Figure 3.1 was replaced by the reactor in Figure 3.6,
with temperature profiles as in Figure 3.10. The alcohol feed passed through a 1 meter
coil, which was placed in a pre-heating oven and connected to the reactor inlet. The pre-
heating oven ensured that the feed entered the reactor at the reaction temperature. Exit
lines were heated with heating wire to keep the products in the gas phase. A heated
syringe (120°C) was used for taking gas samples and injecting them into a Gc. The
products were analysed with a HP G 1800A gas chromatograph, equipped with a mass
spectrometer and a flame ionisation detector (for more details see section 3.1.5).

For both the ethanol and 2-butanol reactions, the membrane was packed with a 14.4 wt %
Cu on silica catalyst. The catalyst particle size fraction was 500 to 850 microns. The
catalyst was kept in position with quartz wool at the edges of the membrane (see
Figure 3.11).

Pd-Ag coated membrane

Quartz
wool
Figure 3.11: Membrane packed with catalyst

During the start-up procedure, the membrane reactor was heated in nitrogen at 2 °C/min
from room temperature up to 275°C. The membrane and the copper catalyst were then
reduced in a hydrogen atmosphere (flow of 50 cm3/min) for 1.5 hours. After 1.5 hours,
hydrogen was replaced with nitrogen for a further 10 minutes and then the reactor was
either heated or cooled at 2 °C/min to the required reaction temperature. At the required
reaction temperature, the alcohol was introduced at 10 mllh for 1.5 hours before analysis.

111
3.4.1. ETHANOL DEHYDROGENATION
The ethanol dehydrogenation reaction in the catalytic membrane reactor was investigated
from 250°C to 300 °c. The feed rate depended on the reaction temperature, but varied
from 5 mIlh to 30 mllh. The membrane was packed with 3.00 g catalyst. The following
conditions were used for testing ethanol dehydrogenation in a membrane reactor:
• 250°C: 5,10 mIlh feed rate,
• 275.o C: 5, 10, 20, 30 mIlh feed rate, and
• 300°C: 10, 20 mIlh feed rate.

Experiments were first conducted without a sweep gas, to obtain conversions

representing a plug flow reactor. An inert sweep gas (nitrogen) was then introduced into
the shell side and increased from a sweep gas to feed molar ratio of 1:5 to 4: I. In all
experiments co-current flow was used.

3.4.2. 2-BUTANOL DEHYDROGENATION

The dehydrogenation of 2-butanol was conducted In the same manner as the
dehydrogenation of ethanol. The membrane was packed with 3.50 g catalyst. The
following conditions were used for testing 2-butanol dehydrogenation in a membrane
reactor:
• 190°C: 5, 10, 15 mIlh feed rate,
• 215°C: 5, 10, 15,20 mIlh feed rate, and
• 240°C: 5, 10,20,30,40 mIlh feed rate.

At each feed flow rate, four or five different sweep gas flow rates were tested and the
dehydrogenation results compared to results obtained when no sweep gas was used (i.e. a
plug flow reator). Nitrogen was used as the sweep gas and the flow was co-current for all
experiments.

112
3.5. SUMMARY
This chapter covered all experimental work performed. There were three basic sections:
• Catalyst optimisation and kinetic testing,
• Membrane optimisation and permeance testing, and
• Catalytic membrane reactor experiments.

The' equipment and procedures used in each section were described in detail. The
variables that were investigated, were listed. The values at which each variable were
tested, were also indicated.

113
 4. OPTIMISING CATALYST COMPOSITION

Both the dehydrogenation of ethanol and 2-butanol were investigated. For each reaction,
different catalysts were investigated at a range of operating conditions. The composition of
the catalysts and the conditions at which they were tested were discussed in Chapter 3. All
the data presented in this chapter is for average WfF values at specific temperatures and
copper .loadings. The WfF ratio is the catalyst mass divided by the feed flow rate. A low
WfF value indicates a fast feed rate and a high WfF value indicates a slow feed rate for a
constant mass of catalyst. Each experiment was performed at either three or four different
feed flow rates. The conversions, yields and selectivities discussed are average values, unless
otherwise stated. The following definitions were used:

total moles feed reacted

total feed conversion = - - - - - - - - - (4.1)
total moles feed fed

moles feed converted to product

product yield = - - - - - - - - - - - - (4.2)
total moles feed fed

moles feed converted to product

product selectivity = - - - - - - - - , - - - - - , - - - (4.3)
total moles feed reacted

yield = selectivity x total feed conversion (4.4)

4.1. DEHYDROGENATION OF ETHANOL

The dehydrogenation of ethanol was tested over magnesium oxide, silica and alumina
catalysts impregnated with copper. The results for each catalyst will be discussed below.
The particle sizes used for each support were listed in Tables 3.2 and 3.3.

4.1.1. MgO CATALYSTS

Figures 4.1 to 4.3 show the reaction results for the reaction of ethanol over MgO catalysts. A
range of products formed, depending on the reaction temperature and the copper content.
The main products were: ethene, propene, butadiene, acetaldehyde, di-ethyl ether, acetone,

114
I-butanone, 2-butanone and I-butanol. Below 360°C there was very little reaction of
ethanol (Figure 4.1). At low temperatures, acetaldehyde production was low (Figure 4.2),
but the selectivity towards acetaldehyde was high (Figure 4.3). When the temperature
increased, both the total ethanol conversion and the conversion to acetaldehyde (this is called
the acetaldehyde yield) increased sharply.

Figure 4.1: Total ethanol conversion for Figure 4.2: Ethanol to acetaldehyde yield
Cu on MgO catalysts (Cu on MgO catalysts)

Figure 4.3: Acetaldehyde selectivity for Cu on MgO catalysts

The difference between the values in Figure 4.1 and Figure 4.2 is the ethanol converted to
other products and this difference is expressed via the selectivity curve. At 440°C the

115
selectivity dropped to between 25% and 35% for the vanous copper percentages. The
majority of ethanol was not converted to acetaldehyde, but to I-butanol. Higher temperatures
not only favoured acetaldehyde formation, but also C4 formation.

J'he Cu content on the catalyst has an effect on the BET surface area of the catalyst. When
the Cu content increased, the BET surface area decreased and less support sites were
available for reaction (Table 4.1). With no copper on the catalyst, the main product was
ethene and acetaldehyde was the main by-product. As soon as copper was deposited on the
support, ethene production ceased and acetaldehyde production increased. Figure 4.2 shows
an increase in acetaldehyde yield with an increase in Cu content. The selectivity towards
acetaldehyde formation (Figure 4.3) remained fairly constant with an increase in copper.
The ethene that was formed in the absence of copper, was replaced by I-butanol in the
presence of copper. Table 4.2 gives a summary of the main product and by-product at
different conditions.

Table 4.1: BET surface areas for Cu on MgO catalysts in m2/g

Cu% BET (m"/g)
surface area
0 27.4
6.7 18.5
10.3 15.0
16.9 9.3
24.9 3.6

Using MgO catalysts, both total ethanol converSIOn and acetaldehyde yield were poor,
because of the low BET surface area of the catalysts (see Table 4.1). Iwasa and Takezawa
(1991) studied ethanol dehydrogenation with a 30 wt % Cu on MgO catalyst at 220°C. The
selectivity towards acetaldehyde production was 74% with ethyl acetate and other C4 species
the 1!lain by-products. Takezawa et aL (1975) concluded that acetaldehyde was the main
product when ethanol reacted over pure MgO between 340°C and 360 °C. The selectivity
under differential conditions was above 90% at those temperatures. A reaction mechanism for

116
pure MgO catalysts was discussed by Takezawa et al. (1975). Results in this dissertation
differ from the results obtained by Takezawa et al. (1975). For a pure MgO catalyst the main
product was ethene at 340°C and the selectivity towards acetaldehyde poor.

Table 4.2: Yield matrix indicating main products and by-products for CulMgO
320/360°C 440°C
O%Cu Ethene (1.6%) Ethene (17.8%)
Acetaldehyde (0.8%) Acetaldehyde (9.1 %)

10% Cu Acetaldehyde (0.8%) I-Butanol (9.6%)

I-Butanol (0.8%) Acetaldehyde (7.2%)

25%Cu Acetaldehyde (1.5%) Acetaldehyde (14.8%)

I-Butanol (1.3%) I-Butanol (5.0%)

4.1.2. Ah03 CATALYSTS

Alumina is an acidic catalyst, and in theory, the dehydration reaction should be favoured
above the dehydrogenation reaction. Figures 4.4 to 4.6 indicate total ethanol conversion and
the formation of reaction products as a function of temperature and copper loading. Both the
total ethanol conversion (Figure 4.4) and ethene yield (Figure 4.5) were very dependant on
temperature. An increase in temperature increased the total ethanol conversion and ethene
formation became the dominant reaction. For di-ethyl ether formation (Figure 4.6), there
was an optimum temperature at which a maximum yield was obtained. The optimum
temperature was dependent on the copper percentage on the alumina support. The optimum
temperature shifted to higher values for higher copper percentages on the support. The
highest di-ethyl ether yields were obtained with no copper on the catalysts:
• For 0% Cu: Optimum temperature of220 °C gave 59% ethanol conversion to ether.
• For 13.2% Cu: Optimum temperature of260 °C gave 51 % ethanol conversion to ether.
• For24%Cu: Optimum temperature of 260/300 °C gave 42% ethanol conversion to
ether.

117
Figure 4.7 shows the effect of flow rate and temperature on di-ethyl ether yield for a 0"10 eu
on alumina catalyst. The curves have similar profiles. With a decrease in WIF (due to an
increase in F), the optimum ether yield shifts towards the higher temperatures.

10 0

" eo
~ eo
t
<! ~o
~
.0

Figure 4.4: Total ethanol conversion Figure 4.5: Ethanol to ethene yield
(eu on alumina catalysts) (eu on alumina catalysts)

70.0 --------------------------.----------- ----------... ----._._--------------.---------------,

60.0
...-
+----,;'::,-:''-.-------------i
~ SO.O
A"\
+--1/,,~:,-:".~\'-:>'\--"---l~_=w;;;;IF:::,-;;;393;;:..:::.-~r--i,
~ 40.0 .y/'-i-
"i----\-4--1 _WIF::I 197 kg.shnaI H

!.~ tf:I;. .___\~~'\~. .~.~=·=·=·W=IF=.=~=..=MM==~

30.0 /:' \ ; ..
ill H-----\---"""._ _ _ _ _ _---!
20.0

10.0
\ ,',
t-------\:--';'.~=--...;:::_--i
_____
'---"""
0.0 +--_--~----'_-~.__-""'"
1eo 220 260 300 380
Reaction T (·C)

Figure 4.6: Ethanol to di-ethyl ether yield Figure 4.7: Ethanol conversion to di-ethyl
(eu on alumina catalysts) ether (0% eu on alumina) at different WIF
ratios

The copper concentration on the support did not have much of an effect on the total ethanol
conversion, but had a definite effect on the product distribution, Pure alumina gave the
highest ethene and di-ethyl ether yields. The reason was that pure alumina represented the

118
maximum available acidic sites, resulting in the maximum dehydration products. The ethene
and the di-ethyl ether yields dropped when the catalyst was impregnated with copper. For
10% Cu loading and above, the yields remained within a narrow band, with not more than 5%
variation from the average.

The main by-products were butenes, acetaldehyde and a small amount of hexenes. Figures
4.8 and 4.9 show acetaldehyde and butene yields with temperature increasing from left to
right. For acetaldehyde production there was an optimum copper loading on the alumina
(13.2 wt %), which resulted in the maximum acetaldehyde production. The butene
production increased with temperature and was the highest for 8.8 wt % Cu and 13.2 wt %
Cu on alumina. Higher and lower Cu loadings caused a decline in butene production.

T'" - .. -

YIoId(%1

Figure 4.8: Acetaldehyde yields for Cu on Figure 4.9: Butene yields for Cu on alumina
alumina catalysts (W/F = 197 kg.s/mol) catalysts (average W/F)

De Boer and Visseren (1971) observed a proportionality between the dehydration rate
constants and the amount of aluminum on the surface. Iwasa and Takezawa (1991) tested an
alumina-based catalyst with a high copper loading (30 wt %) at 220°C. They reported 54%
selectivity towards acetaldehyde production, 19% towards di-etheyl ether and 22% towards
C4 production.

Copper alumina catalysts have often been used for cyclohexanol dehydrogenation. The
selectivity towards the dehydrogenation reaction was higher for cyciohexanol than for
ethanol on alumina-based catalysts. The effects of copper concentration on the selectivity

119
and activity towards cyclohexanone fonnation have been investigated (Chang and Saleque,
1993, 1994; Sivaraj et a1., 1990).

4.1.2.1. BET and copper surface areas for Cu on Alz<lJ catalysts

The total catalyst surface area and the copper surface area were detennined to investigate
their effect on product yield and product distribution. Figure 4.10 shows total catalyst
surface area as a function of copper loading and Figure 4.11 shows copper surface area as a
function of copper loading.

220 1.110 20.00

I
""- ~
-BET ...
I
I • AVl!ll'agepo~elQo I
•
'" 1.70

x ~
1000

".
.
16.00

g ..
:!f 1.60 , ,,
"'-
14.00

~
155 E 1.50 ,, U
:
""'.. -" ""'- ""'-
12.00
,
". i ,!!I 1.40 , 10.00
i•
'"
"E
'".. l '
---- 8.00
<!•
o ---
-,. "'.
-;; 1.30

. ~ , .i '.20
I
'.00

". '.00
• • '" 1.10
2.00
'00
• ". 1.00 '.00
10
Cu%
15
" " • 12
"
Cu%
20
"
Figure 4.10: Effect ofCu% on catalyst Figure 4.11: Effect ofCu% on Cu surface
surface area for Cu on alumina catalysts area for alumina-based catalysts

The total surface area (Figure 4.10) declined from 215 ml/g at 0 wt % Cu to 120 ml/g at
24 wt % Cu as the deposited copper filled the micro- and mesopores in the support. Ethene
and di-ethyl ether fonned mainly on the acidic sites, because ethene and di-ethyl ether yields
dropped (Figures 4.5 and 4.6) when copper was added to alumina, to reduce the number of
acidic sites. Even though the catalysts' surface area was almost halved for 0 to 24 wt % Cu
loadings, the ethanol converted remained constant. The deposited copper sites provided
additional catalytic activity and resulted in more products being formed (mainly butenes and
acetaldehyde) to compensate for the loss of dehydration products (ethene and di-ethyl ether).
The added copper sites were much more active than the alumina sites, since the small
increase in copper surface area (Figure 4.11) fully compensated for the large decline in total
surface area (Figure 4.10).

120
Copper surfaces areas were detennined as described in Sections 3.1.6 and 4.1.2.2.2. The
copper surface area (in ml/g sample) had an optimum value at 13.2 wt % Cu on alumina. At
that value, the maximum acetaldehyde and butene yields were observed (Figures 4.8 and
4.9). From this it could be concluded that copper on alumina catalyses the formation of
acetaldehyde and butenes.

35.00

"iii 30.00
Copper % on support
!. 25.00 increases for lines from
tDpm boIlcIm at 100A

..i!i!! 20.00

&.
"ii 15.00

1
.E
10 00
.

5.00

10 100 1000
Awrage pore alze {AI

Figure 4.12: Pore size distribution for Cu on alumina catalysts

Figure 4.12 clearly shows how the pore size distribution changed when the copper loading
on the support increased. The y-axis is the incremental pore area for every pore fraction from
very large to very small pores. The curve with the highest differential area values is the 0%
Cu line. As the Cu % increases the differential areas decrease (lines drop on the y-axis for
example at 100 A pore size). The line with the lowest differential area values is the 24 wt %
Cu one. Figure 4.12 shows both a shift in the pore size distribution to higher pore sizes and a
decline in the values on the differential area axis with an increase in Cu loading. The average
pore size changes from 132 A for a 0 wt % Cu on alumina catalyst to 166 A for a 24 wt % Cu
on alumina catalyst.

Result were in line with those of other researchers (see Tables 4.3 and 4.4) who also obtained
a optimum copper percentage (on alumina) giving a maximum copper surface area on their
catalysts.

121
4.1.2.2. Literature data on surface areas of copper catalysts
There are two methods for determining copper surface areas. One method is based on N 20
titration and the other method on either CO chemisorption or H2 chemisorption. For both
methods, the catalyst must be reduced in hydrogen prior to analysis. The catalyst reduction
step has been discussed in Chapter 2 (2.10.3). Bonding between copper species and the
support sites, interactions between different copper species and the formation of amorphous
phases are only a few of the problems encountered when reducing the copper ions to metallic
copper. The result is very poor reduction. Better reduction requires higher reduction
temperatures, which result in sintering (see 2.10.4.1) and further deviation from a true copper
surface area. Experimentally-determined copper surface areas thus have a significant margin
of error.

4.1.2.2.1. N 20 titration
Sivaraj and Kantarao (1988a) described the N 20 titration method. The nitrous oxide
decomposes according to the following equation:

N 20 (gas) + 2Cus -> (CUs)20 + N2 (gas) (4.5)

The metallic surface area (SH) is determined from

(4.6)

where nm is the total amount of nitrous oxide molecules that decompose, Xm is the
chemisorption stoichiometry and ns is the amount of copper metal atoms per unit surface area
(1.47*10 19 m,2). The catalyst is first reduced in hydrogen for 5 hours at 250°C, evacuated (to
10'6 Torr) for 2 hours and then reacted with N20 at 200 Torr and 90°C. The reaction does
not cause a pressure change. After 5 hours of reaction, the remaining N 20 is frozen out in a
nitrogen trap and the pressure difference between the initial pressure and final pressure is
used to calculate the amount ofN20 that reacted.

122
4.1.2.2.2. Chemisorption
A detailed discussion on this process can be found in Jeon and Chung (1994). The same
apparatus used for measuring BET surface areas is used for chemisorption experiments. The
catalyst is first reduced and then either CO or H2 is used as the analysis gas. The method
works on the principle that these two gases adsorb reversibly on CuD, but not on Cu l +.
Chemisorption is carried out and the first adsorption isotherm constructed. This value
represents adsorption onto both Cuoand Cu l +. The catalyst is then evacuated and the gas
desorbs from CUD, but not from Cu l +. The experiment is repeated, giving a second adsorption
isotherm. The second isotherm represents CuD, while the difference between the isotherms
represents CUI +.

4.1.2.2.3. Effect ofpreparation method on catalyst characteristics

Tables 4.3 and 4.4 list BET and copper surface areas for various copper catalysts. The
amount of data is not enough to make accurate conclusions regarding copper on silica
catalysts.

Table 4.3: Surface areas of unsupported copper catalysts

Catalyst Analysis Reduction Catalyst BET Copper area Reference
2
preparation method conditions Composition (m2/g) m /g sample
Co- N 20 3h in H2 at Cu 1 1 (1) Kanoun
precipitation titration 300°C Vso-Cuso 89 10.2 (29) (1991)
unsupported VSO-Cu2SZn2S 146 18.2 (107)
Znso-Cuso 15 2.6 (6.5)
Co- N 20 3h in H2 at Cuso-Crso 163 24 (80) Kanoun
precipitation titration 300°C Cuso-CrsA4s 188 27 (102) (1993)
unsupported Cuso-Also 214 8 (14)
Co- N 20 Ih in 10% Cu (9) Tu (1994a)
precipitation titration H2/Ar at Cr9Cu91 (18.8)
unsupported 200°C Cr 13CuS7 (14.5)
..
Catalyst composItIOn IS III mol percent
Copper surface area: value in brackets is for m2/g Cu

123
 Many authors have reported either BET or chemisorption data, but not both. In most cases
the optimum copper area that was observed, was in accordance with what was discussed in
section 2.10.1. An exception was Jeon and Chung (1994), who prepared silica-supported
catalysts.

Table 4.4: Surface areas of supported copper catalysts

Catalyst Analysis Reduction Catalyst BET Copper area Reference
preparation method conditions Composition (m2/g) m 2/g sample
Urea N 20 5h inH2 at Cu l2-alumina 3.9 Sivaraj
hydrolysis titration 250°C CU24-alumina 12.0 (1990)
CU31-alumina 7.0
Impreg- 10 wt % Co on 150 Reddy
natidn y-alumina (1997)
Electro1ess Alumina 22.6 Chang
plating 19.5 wt % Cu 23.8 (1993)
Precipitation 19.5 wt % Cu 20.5
Impregnation 12.5 wt % Cu 20.7
Urea N 20 5h in H2 at Alumina 242 0 Sivaraj
hydrolysis titration 250°C Cll4.2- alumina 250 13 (308) (1988a)
CU24.4-alumina 103 41.6 (170)
CU34.2-alumina 63 36.7 (108)
Electroless N 20 y-alumina 138.3 0 Chang
plating titration CUIO.o-yAI 20 3 145.4 4.5 (1994)
CUI8.3-yAIz03 101.7 14.2
CU24.4-yAIz03 82.5 9.8
Precipitation CO 4h inH 2 at CU24-si1ica 2.8 Jeon
chemi- 250°C CU24-silic a 10.3 (1994)
sorption CU24-silica 12.6
.
..
Catalyst composItIOn IS m mass percent
Copper surface area: value in brackets is for m 2/g Cu.

124
Chang and Saleque (1993) compared the use of electroless plating, impregnation and
precipitation for preparing copper on alumina catalysts. The total BET area declined in that
same order, but the BET values showed only a small variance. Similar preparation methods
resulted in significant differences in the copper surface areas. Compare values from:
• Kanoun et al (1993) and Tu et al. (1994a),
• Sivaraj and Kantarao (1988a) and Sivaraj et al. (1990), and
• Pure copper values from Kanoun et al. (1991) and Tu et al. (1994a).

The data presented from Jeon and Chung (1994) in Table 4.4, show similar Cu percentages
on Si02 . Cu surface areas were obtained for different pH-values of the precipitate.

4.1.2.3. Tbe effect of additives on etbanol conversion and product distribution

Cobalt, chromium or nickel was added to the copper to investigate the effects on product
distribution. The effects of these additives on catalyst stability were not investigated at this
stage. This was done later (see Cbapter 5), when conditions were chosen to determine
reaction kinetics. Figure 4.13 shows that at 260°C and below, Cu-CrtAh~ outperformed
Cu on alumina catalysts, giving higher total ethanol conversion. For Ni and Co additives, the
difference in performance at 260°C and below compared to pure Cu was small. At 300°C
and above, the pure Cu catalyst performed the best, followed by the Cu-CrtAh03 catalyst.

" C!J18.S'WoCU 1118.5" cu

5O+---~_ 1---- .17.~CU;2.CJ8')1,CO
~ 70 1!117.a%CU;1.7l'KCr '---4-- 1117.8" Cu: 1.T.nI. cr
goo D 18.1% Cu; 2.0% NI
I 50 +----l0r- "
t,..f.---::
•8 .. +-------il~
~ ,.+----
~ 2O.f.--
... ,..f.............--

'80
Tempo_.. rC) "" TemperatuN (OC)
380

Figure 4.13: Effect ofCr, Ni and Co (on Figure 4.14: Effect ofCr, Ni and Co
alumina) on total ethanol conversion (on alumina) on ether yield

125
 70

618.5%CU

_17.7% CU; 2.09% Co

017.8% "", 1.73% '" i-------;;;;;----Ilt----I
I e18.1% CU: 2.0% N!

!~r----------__--
~ 20 +------------IM--

180 '20 ,,,

'"

Figure 4.15: Effect ofCr, Ni and Co (on alumina) on ethene yield

Co and especially Cr addition made the catalyst more selective for di-ethyl ether formation
(Figure 4.14), with significant improvements at 220°C and 260 °C. For ethene formation
(Figure 4.15), pure copper on alumina resulted in the best performance at all temperatures,
followed by Cu-Ni and Cu-Cr. The addition of additives did not improve the acetaldehyde
yields for copper on alumina catalysts. The acetaldehyde yields were below 8% for the
different catalysts tested in Figures 4.13 to 4.15.

There was little difference in the BET surface areas for the different catalysts as indicated in
Table 4.5. The larger difference in copper surface areas can only be due to a difference in
reduction percentages. Chromium addition makes it more difficult to reduce the eu2+ ions
with hydrogen. Some copper will be present as Cu2+ and not Cuo Less Cuo sites will be
detected, which results in a lower copper surface area.

..
Table 4 S· BET and copper surface areas for copper/alumina catalysts with additives
Cu% Additive BET area Cuarea Cu area
(m2jg) (m2/g sample) (m1/gCu)
18.5 - 140 1.58 8.54
17.7 2.09% Co 151 1.67 8.42
17.8 1.73% Cr 157 1.10 5.65
16.0 2.0%Ni 137 1.40 8.72

126
4.1.2.4. Summary for using Cu on alumina catalysts
This type of catalyst is not suitable for acetaldehyde production via the dehydrogenation
reaction. Very good yields of ethene and di-ethyl ether were obtained by suitable choice of
the copper loading, addition of the correct additives and performing the reaction at the correct
operating conditions. Alumina-based catalysts favoured the dehydration reaction. At higher
temperatures (see Figure 4.9), butene production became significant for catalysts with the
highest copper surface areas.

4.1.3. SILICA CATALYSTS

The variables that were investigated to determine maximum acetaldehyde yield are:
• copper loading,
• temperature,
• feed flow rate,
• effect of additives, and
• effect of catalyst particle size.

The different catalysts investigated were listed in Tables 3.2 and 3.3. The operating
conditions were listed in Table 3.4.

4.1.3.1. The effects of copper loading and temperature on acetaldehyde yield

Figures 4.16 to 4.18 show total ethanol conversion, acetaldehyde yield and acetaldehyde
selectivity over the entire range of copper percentages on the silica support and for
temperatures ranging from 240°C to 440 °C. Both total ethanol conversion and acetaldehyde
yield increased sharply with temperature. For both the total ethanol conversion and
acetaldehyde yield there was an optimum copper concentration that resulted in a maximum
ethanol conversion (Figure 4.16) and a maximum acetaldehyde yield (Figure 4.17). The
optimum copper loading on the catalyst was 15 wt %. At 440°C (with a 15 wt % Cu on
silica catalysts) the total ethanol conversion was 89% and the acetaldehyde yield 74% at an
average WfF.

127
The amount of acetaldehyde formed on a pure silica catalyst was negligible. For pure silica,
ethene was the main product with no significant amount of by-products. Ethanol conversion
to acetaldehyde (acetaldehyde selectivity) was not very dependant on the copper loading
(Figure 4.18) for copper loadings from 4 to 34 wt %. At 440°C the selectivity varied
between a maximum of 88% (for a 18.6 wt % Cu on silica catalyst) and a minimum of 79%
(for a 4.2 wt % Cu on silica catalyst). The selectivity was more dependent on temperature.
Up to 320°C the selectivity remained in the mid 90% level, but it dropped to about 80"10 at
440 DC. The main by-products were: ethene, di-ethyl ether, etoxy ethane, I-butanone and
ethyl acetate.

\()J sO

~g: : ~
~.\O
(jJ

....t~
40
~o
~ ~o

Figure 4.16: Total ethanol conversion Figure 4.17: Ethanol to acetaldehyde yield
(Cu on silica catalysts) (Cu on silica catalysts)

Figure 4.18: Ethanol to acetaldehyde selectivity (Cu on silica catalysts)

128
Figures 4.19 and 4.20 show the effect of the feed flow rate on total ethanol conversion and
acetaldehyde yield for the optimum copper loading. The total ethanol conversion
(Figure 4.19) remained high, with a decrease in WfF (increase in F) to between 131 kg.sJmol
and 66 kg.sJmol. At higher flow rates, i.e. lower WfF values, the total ethanol conversion
dropped due to a shorter catalyst contact time. For the shorter contact time, the reaction
could not go to completion. There was an optimum W fF ratio for a maximum acetaldehyde
yield (Figure 4.20). At the highest W fF values (slowest feed rates) the acetaldehyde yield
decreased (Figure 4.20), due to a very long residence time that increased the formation of
propene, di-ethyl ether, I-butanone and ethyl acetate. For high flow rates (WfF < 66
kg.sJmol) the lower total ethanol conversion caused a lower acetaldehyde yield.

Figure 4.19: Total ethanol conversion Figure 4.20: Ethanol converted to

(15 wt % Cu on silica catalyst) acetaldehyde (15 wt % Cu on silica)
)
Iwasa and Takezawa (1991) obtained 78% selectivity towards acetaldehyde formation at 220
°C with a 30 wt % Cu on silica catalyst. The main by-products were C4 species and ethyl
acetate.

4.1.3.2. The effect of additives on product distribution and yields

Three additives, Co, Cr or Ni, were added to Cu in 1: 9 and I: 19 mass ratios, to determine
their effect on catalyst activity and selectivity. The results are shown in Figures 4.21 to 4.24.
Below 400 °C, the addition of additives inhibited ethanol conversion and the conversion was

129
lower than for pure copper catalysts (Figures 4.21 and 4.23). The addition of Co resulted in
the poorest conversions at all temperatures. A higher Ni content (Figure 4.21) resulted in
better conversions than a lower nickel content (Figure 4.23). At both ratios, the Cu-Cr
catalysts performed similarly, with Cu:Cr = 9: 1 being slightly better. At 400 DC and 440 DC,
Cu-Cr catalysts and the one Cu-Ni catalyst (13.5 wt % Cu, 1.5 wt % Ni) gave conversions
which differed only marginally from those obtained with a pure copper catalyst.

The addition of additives made the dehydrogenation reaction far less selective. Acetaldehyde
yields decreased sharply, even at high temperatures, as indicated in Figures 4.22 and 4.24.
The by-products formed on each of the three catalyst were:
• Cu-Cr: ethene, propene, butenes, di-ethyl ether, I-butanone, MEK, ethyl acetate,
• Cu-Ni: ethene, propene, di-ethyl ether, I-butanone, and
• Cu-Co: ethene, propene, butenes, di-ethyl ether.

Kanoun et aI. (1993) calculated catalyst activities and turnover frequencies (TOFs) for Cu-
Cr-AI catalysts prepared by precipitation. Variations in the Cr content didn't have a
significant effect on the TOFs or activities. Tu et aI. (1994a) prepared unsupported Cu-Cr
catalysts via precipitation. Their results indicated an optimum ethanol dehydrogenation
activity for a Cr:Cu = 4:40 (on mol basis) catalyst. Selectivity data was not included

100
90

,,-I---------i'iIo~l!II_ so
eJ 15.0% cu
E'l15.O%Cu 70
!!6 .. +---~.-~~~
• 13,S% Cu: lao
.• ,,+----lli~_li\. 1.5% Co
"so -I-----§---I'!1--.Jii!-,=--lil---I 1113,5% Cu:

~40~-_-~~~ 1!l13.5"" Cu;

1,5'IIoCr
Ii
;:'" 1.5'IIoCr
.,3.5%Cu;
<3 "" '\---f4---lIl- 1113.15% CI,I:
30 1.5%NI
1.5'IIoNl
20
10
,
o
240 280 320 36Q

Reaction T I"C)
-400 4040 2<, ,., 32'
""
Reaction T (-C)
400 ...
Figure 4.21: Effect of Co, Cr or Ni additives Figure 4.22: Effect of Co, Cr or Ni additives
on total ethanol conversion (Cu on silica) on acetaldehyde yield (Cu on silica)

130
!II'

100
"'"90
eo
"eo
~ 70 III 15.I)%Cu
70 1!I1S.O%CU
ceo
.2 .1425%CU; l
~" 0.75% Co 50
11 t----;::----tI---1H!--li-1a--I-fa-i1
•c
'" ~14.25%CU:
O.75%Cr >"" iii 14.26% Cu;
D.7a%er
8" • 14.25'1' CU; .14.~Cu:
2D Q,15%Nl 20 O.75'l1oNi

10 10

240 280 320 380

0
...,
Reaction T ('C)
'" ''''' "" 320 300
Reaction T ('C)
"""

Figure 4.23: Effect of Co, Cr or Ni additives Figure 4.24: Effect of Co, Cr or Ni additives
on total ethanol conversion (Cu on silica) on acetaldehyde yield (Cu on silica)

4.1.3.3 Effect of particle size on acetaldehyde yield

Figures 4.25 and 4.26 show the effect of catalyst particle size on ethanol conversion and
acetaldehyde yield. In theory, larger particles cause greater mass transfer resistance from the
bulk fluid into the pores and may slow down the reaction rate if the rate is mass transfer
limited. Smaller particles reduce mass transfer resistance and improve the reaction kinetics.
Experimental results (Figure 4.25) were consistent with the theory and the total ethanol
conversion declined as the size of the catalyst particles increased.

100 100

"so
so 1lI1&J.300
.,""'"
.- .-
micrmlll; mlaOllIll;
14.1%C\I 70 14.1% OJ
l
j mlCf1)l'lS; ~eo mia-ens;
1S.Q%CU 15.0%0"
~50 :l!"
• 1i1850-1180
C
iJ85D-1180 ):40 IIIlerm.;
microns:
ll.ncu
8" .>
13.~CU

3000
30
.> 3000
20 mlCfons;
",Iaoo.;
13.7%Cu 10
13.7% eu

240 280 320 360 "'" .., 280 320 3110 400 «0

Reaction T ('CI Reaction T ('C)

Figure 4.25: Effect of catalyst particle size Figure 4.26: Effect of catalyst particle size
on total ethanol conversion on acetaldehyde yield percentage

Two effects are responsible for the decrease in ethanol conversion using the larger particles:
channelling of the feed gas and intra-particle mass transfer resistance. Channelling usually

131
starts occurring when the particle diameter is less than one tenth of the reactor diameter.
Some channelling occurred for the 3 mm catalyst spheres, because the inside diameter of the
quarts tube that housed the catalyst was only 8 mm. For the smaller particles (1.2 mm and
less) channelling was less dominant. The decrease in ethanol conversion with the larger
catalyst particles confirmed that those two effect were present during the ethanol reactions.
Figure 4.26 indicates that at 360°C and above, the acetaldehyde yield was similar for all
particles up to a size of 1180 I-lm. The mass transfer limitation for ethanol dehydrogenation
to acetaldehyde at 360°C and above was not significant for particles less than 1180 I-lm. A
further observation that could be made when comparing Figures 4.25 and 4.26, was that the
selectivity towards acetaldehyde decreased at high temperatures as the catalyst particles
became smaller.

4.1.3.4. Surface area data for silica-based catalysts

Figures 4.27 and 4.28 show BET and copper surface areas as a function of copper loading.
Figure 4.29 indicates the change in pore size distribution when the copper content on the
catalysts was increased. The BET surface area declined from 446 m2/g and an average pore
size of 87 A for pure silica, to 201m2/ g and an average pore size of 94 A for a 33.5 wt % Cu
on silica catalyst. The surface area was about double that of the alumina catalysts (see
Figure 4.10). The copper surface area increased with increasing copper loading and levelled
off above 27 wt % Cu on silica. The values were lower than those for alumina catalysts
(Figure 4.11) and also much lower than values from literature (Tables 4.3 and 4.4).
Figure 4.29 indicates how pores were filled when additional copper was deposited on the
catalysts. (The line with the highest differential areas is the one with the no copper.) As the
copper percentage on the silica support increases, the values on the differential area axis drop.
The line with the lowest differential area at 94 A is the 33.5 wt % Cu on silica. The change
in average pore size is marginal, but the change in BET surface area is significant, moving
from 0% Cu to 33.5 wt % Cu.

132
!'"

480 •• -.-•••• -.-•• 95 1.2 5.0

-i'<---I
r _BET.r.. I1------"'--1 4.5

'-
430
.- \ I • A........ pare size I • 93 4.0

i 380"'--.... 91 g
3.5
S'

~ 330 ". ·m '. 3.0

U

'k
t '-=,-""'-..o--------i 89~
2.5

.. 2lIO j------ L
2.0
J
~
1.5
230 j -_ _ _ _ _ _ __ _---=:""'"____I 87

180 +--__-_-_-_-_'-....--185 +--__- _ - _ - _ - _ - - - + 0 . 0

o 10 15 ~ ~ ~ ~
0.0
o , 10 15 20 25 30 35
Cu% CU%

Figure 4.27: Effect ofCu% on catalyst Figure 4.28: Effect of Cu% on Cu surface
surface area for Cu on silica catalysts area for silica-based catalysts

.0,-------------------,

Copper % on support
Increases for linea from
top to bottom at 90 A

0~~~------_ _----~~~--_4
10 1(,) 1000
Average pore size (A)

Figure 4.29: Pore size distribution for copper on silica catalysts

The acetaldehyde yield data (Figure 4.17) and copper surface area data (Figure 4.28) do not
correlate well. Many studies have shown that dehydrogenation activity is proportional to the
copper surface area (Tu et al., 1994b; Jeon and Chung, 1994), but in the present study there
was a difference between the two. The best explanation for this is the unreliability in copper
surface area data. Jeon and Chung (1994) did an in depth study on copper on silica catalyst
characterisation. They experienced similar problems with determining copper surface areas
on silica supports and drew the following general conclusions:
• Copper reduction ranges from 80% in the best cases to as low as 40%, depending on the
copper percentage on the silica support. For low copper content catalysts, the strong
silica-copper bond makes reduction very difficult.

133
• Silica solubility increases with solution temperature. An amorphous copper silicate phase
forms or can form. This contaminates copper sites, which can inhibit the
dehydrogenation reaction and prevent chemisorption.
• Copper surface areas determined by chemisorption, X-ray line broadening and N 20
titration varied by up to a factor of 3, depending on the method employed.
• Copper particles range from small to very large. The catalyst prepared by precipitation is
not very homogeneous.

In the present study silica was impregnated with copper solution at about 80°C, which could
lead to the formation of a copper-silicate phase. Since both copper-silica and silica are
amorphous, it is difficult to detect the different phases with X-ray diffraction analysis. The
existence of a copper-silicate phase and the poor reduction at the lower copper loadings could
possibly be responsible for the low copper surface areas and the difference in acetaldehyde
yields and copper surface areas. Further TEM and X-ray analyses were performed on a 14.5
wt % Cu on silica catalyst to investigate copper crystallite size and crystallite distribution.

4.1.3.4.1. TEM and XRD data/or copper on silica catalysts

Figures 4.30a to 4.3Oi show nine TEM images of a 14.5 wt % Cu on silica catalyst reduced
at 260°C in hydrogen for 2 hours. Black spots indicate copper on the support. For the first
six images the scale is 20 nm, for the 7th one 50 nm and for the last two 200 nm.

TEM images clearly indicate the non-homogeneity of silica-impregnated copper catalysts.

Figure 4.30a indicates that some areas of the support had no copper on. Most areas had,
however, copper crystallites ranging from as little as a few nm to about 20 nm (Figures 4.30b
to 4.30d). Figures 4.30e to 4.30g show areas with copper crystallite agglomerates in the
hundreds of nanometer. The last two images (Figures 4.30h and 4.30i) show very large
copper clusters, up to the micrometer range. The non-homogeneity of the catalysts makes it
very difficult to estimate average crystallite size. X-ray diffraction analysis was performed to
obtain a better estimate of copper crystallite size.

134
Fig 4.30a: TEM (20 nm) Fig 4.30b: TEM (20 nm) Fig 4.3Oc: TEM (20 nm)

Fig 4.30d: TEM (20 nm) Fig 4.30e: TEM (20 nm) Fig 4.3Of: TEM (20 nm)

Fig 4.30g: TEM (50 nm) Fig 4.30h: TEM (200 nm) Fig 4.30i: TEM (200 nm)

Figure 4.30: TEM images of the unused 14.4 wt % copper on silica catalyst

135
X-ray line broadening can be used to calculate average crystallite size, employing Schearer's
equation (equation 4.7). The crystallite diameter, D in A, can be expressed as (Tu et aI.,
1994b):

D _ 1.542
hkI- (4.7)
xp cos(2B)

with x the peak width in radians at 2/3 of the peak height and 29 the angle in degrees at which
the peak is recorded. The peak width was 0.20 degrees at a 29 angle of 43.3 degrees. This
yielded an average crystallite size of about 60 nm. For such large crystals the Schearer's
equation was not very accurate, but still gave a fair estimate of crystallite size.

4.2. DEHYDROGENATION OF 2-BUTANOL

2-Butanol dehydrogenation was studied using MgO and silica catalysts. The compositions of
the catalysts were listed in Tables 3.2 and 3.3. The conditions under which the catalysts
were tested were listed in Table 3.6.

4.2.1. MgO CATALYSTS

When 2-butanol reacted over MgO catalysts, the products were a mixture of butenes and
methyl ethyl ketone (MEK). There was less than I % additional by-products. Figures 4.31 to
4.33 indicate total 2-butanol conversion and butene and MEK yields as a function of
temperature and copper loading. In all three cases the values increased with increasing
temperature. The total 2-butanol conversion was the highest for pure MgO, because of the
large increase in butene yield in the absence of copper on the catalyst. The addition of copper
to the MgO support suppressed butene formation (Figure 4.32) at all temperatures. The
reaction of 2-butanol to butene took place on the MgO sites. The decline in BET surface
areas, when copper was added, was shown in Table 4.1. Reaction can take place either on
copper sites or MgO sites. The addition of copper sites compensated for the large loss of

136
MgO sites, with the net result of only a slight decrease in the total 2-butanol reaction rate or
conversion (Figure 4.31).

The 2-butanol to MEK reaction took place on both the MgO and the copper sites. With no
Cu on MgO (Figure 4.33), similar amounts of butenes and MEK were formed at all
temperatures. MEK formation had to be on the MgO sites. When copper was added to the
support, the MEK yield increased to an optimum yield at 16.9 wt % Cu on MgO. For that
catalyst, the total surface area was about a third of the value of the pure MgO catalyst (see
Table 4.1). Further copper addition to the support caused both a decrease in available copper
area and available MgO area and resulted in a decline in MEK yield.

The copper sites have much higher catalytic activity for MEK production than the MgO sites
do. A simple calculation can verify this:
At 390 °C: 0% Cu on MgO 35% MEK yield
16.9% CuonMgO --+ 47% MEK yield
but MgO surface area is about 1/3 of pure MgO
thus 12% MEK yield from MgO sites
35% MEK yield from Cu sites.

10
(fJ
~ 50
~ 40
~ il
~ ZO
-r ",,,
10

Figure 4.31: Total2-butanol conversion Figure 4.32: Total butene yield

for Cu on MgO catalysts for Cu on MgO catalysts

137
Figure 4.33: MEK yield for Cu on MgO catalysts

Figures 4.34 and 4.35 show the effect of flow rate on MEK yield and selectivity for the best
performing catalyst (16.9 wt % Cu on MgO). The selectivity declined with temperature as
more butenes were formed. The longer residence time at the low flow rates (high WIF
values) improved the MEK yield (Figure 4.34), but resulted in a sharp decline in selectivity.
The total2-butanol conversion dropped from 65% to 35% when WIF decreased from 206 to
51 kg. slmol. The higher flow rates meant that insufficient time was allowed for the 2-butanol
to fully react, while the slower rates did. This is also clear in Figure 4.34 where the MEK
yield dropped as WIF decreased.

(fJ
,0
~40
~'jO
l ~o
!- 10
~

Figure 4.34: MEK yield for a 16.9 wt % Figure 4.35: MEK selectivity for a
Cu on MgO catalyst 16.9 wt % Cu on MgO catalyst

138
4.2.2. SILICA CATALYSTS
The silica catalysts that were employed for 2-butanol dehydrogenation were similar to those
used for ethanol dehydrogenation. The operating conditions have been listed in Table 3.6.

Figure 4.36 shows total 2-butanol conversion as a function of temperature and copper
loading. The main product was MEK (Figure 4.37) and the only significant by-products
were a mixture ofbutenes (Figure 4.38). Total 2-butanol conversion increased sharply with
temperature, but started levelling off at about 360°C. For MEK production, there was an
optimum temperature. This temperature was between 300°C and 360 °C for average WIF
values. The optimum temperature did not vary with copper loading. The maximum MEK
yield was very dependent on the copper content on the support. At low copper loadings the
MEK yield was lower due to butene formation (Figure 4.38). As the copper loading
increased, butene formation decreased. Butene formation also increased with temperature.

100
() sO
~ (fJ
!It
if- AO
op

Figure 4.36: Total 2-butanol conversion Figure 4.37: MEK yield for Cu
for Cu on silica catalysts on silica catalysts

Butene formation mainly took place on the silica sites, while MEK formation took place on
the copper sites. The optimum MEK yield occurred at 15 wt % copper on silica. This was in
good agreement with the results obtained for the dehydrogenation of ethanol (see 4.1.3.1)

139
MEK yield data (Figure 4.37) suggested that the maximum copper surface area should be at
15 wt % Cu on silica.

The optimum catalyst for 2-butanol dehydrogenation was the same as for ethanol
dehydrogenation. For a discussion on the characteristics of this catalyst see 4.1.2.4.

\~
"
fIJ
i
~
6<1
aO
1jI: ].0
1 .."p

Figure 4.38: Butene yields for Cu on silica catalysts

Figures 4.39 and 4.40 show the MEK yield and selectivity for the best performing catalyst
(15 wt % Cu on silica) as a function of temperature and 2-butanol feed flow rate. As
mentioned previously the yield increased with temperature. The optimum yield shifted to
higher temperatures when the flowrate increased (yVIF decreased). For example:
• WIF = 206 kg.s/mo\: 93-91% yield at 270-300 °C,
• WIF = 103 kg.slmol: 93-92% yield at 300-330 °e, and
• WIF = 51 kg.slmol: 89-87% yield at 330-360 °C.

The selectivity towards MEK production (Figure 4.40) decreased with an increase in
temperature and decreased very marginally with an increase in WIF. Longer residence times
(slower feed rates) stimulated butene formation. At 390°C the selectivity varied between
83% and 86% for the different feed flow rates.

140
Figure 4.39: MEK yield for a 15 wt % Figure 4.40: MEK selectivity for a 15 wt %
eu on silica catalyst eu on silica catalyst

4.2.2.1. Effect of particle size on 2-butanol conversion

Figures 4.41 and 4.42 show the effects of particle size on total 2-butanol conversion and on
MEK yield. At 270 De and above, catalyst particles up to 1180 ~ gave similar values for
total 2-butanol conversion. The high conversions indicate very little mass transfer resistance
with an increase in particle size. Particles of 3000 ~ gave lower 2-butanol conversion due
to channelling of the alcohol feed gas (see Section 4.1.2.3) and minor mass transfer
resistance.

100 100

"
80
!!iJ1!5G-300
mlQ-ons;
13.",,"Cu
~70
.~ eo
.300-850
mlc:rone; .,.,.".
mlerOM;
1!5.O%CU
!!50 15.O%Cu
!'I6lIO-1180
~40 milJ"DnI;
13.9'!(.CII
rl85tJ..1180
mierCII.;
830 13.9'\I.Cu

20 .'''''''
mia>ons: .'3000
microns;
13.~Cu
13.7" C..
10

240 270 300 330 360 390 2-40 710 300 330 380 390
Reaction T ("C) Reaction T (OC)

Figure 4.41: Effect of catalyst particle size Figure 4.42: Effect of catalyst particle size
on total 2-butanol conversion on MEK yield percentage

141
The MEK yields (Figure 4.42) were the highest for the 300-S50 flll1 fraction. Smaller
particles (150-300 11m) and particles ranging from 850-IISO 11m gave high butene yields,
which decreased the MEK yield.

4.3. SUMMARY

4.3.1. ETHANOL REACTION

The use of alumina supports in the ethanol reaction resulted in the formation of mostly
dehydration products (ethene and di-ethyl ether), even at high copper loadings. Some
acetaldehyde and butenes were formed with copper on alumina catalysts. The highest
acetaldehyde yields were obtained with the catalysts with the highest copper surface area.
The use of MgO-based catalysts did lead to the formation of acetaldehyde but, due to the low
surface area of MgO supports, the reaction was slow and the conversion low.

Copper-impregnated silica catalysts gave very high acetaldehyde yields. The selectivity
towards acetaldehyde formation was high (above 7S% in the worst case) for temperatures up
to 440°C. Characterisation of the catalysts indicated that impregnation leads to the
formation of a very non-homogeneous catalyst. The copper surface areas for catalysts
prepared by impregnation were lower than for catalysts prepared by precipitation or urea
hydrolysis. For copper on silica catalysts, the addition ofNi, Co or Cr had a negative effect
on acetaldehyde yield. In contrast with this observation, additives to copper on alumina
catalysts had a positive effect on di-ethyl ether formation. An increase in particle size of
silica-based catalysts caused a decrease in total ethanol conversion. This indicated that mass
transfer resistance was important for the reaction of ethanol over Cu-silica. The acetaldehyde
yield, however, varied little at 320°C and above for catalyst particles up to 11S0 11m. The
mass transfer resistance was more important for other ethanol reactions (for example
dehydration and recombination), than for dehydrogenation.

142
The best dehydrogenation catalyst for ethanol dehydrogenation was a 15 wt % Cu on silica
catalyst. Figure 4.43 compares the equilibrium ethanol conversion for the dehydrogenation
reaction with results obtained at different ethanol feed flow rates. For the equilibrium curve
the ethanol partial pressure was taken as 0.2 bar, since the feed (1 bar pressure) consisted of
nitrogen and ethanol in a molar ratio of 4 to J (see section 3.1.4.1). From 320°C to 400 °C
the acetaldehyde yields were close to the theoretical values for WIF = 131 and 66 kg. slmo!.
The difference between the theoretical equilibrium value and the measured values was due to
either incomplete conversion of ethanol (for high flow rates, WIF = 33 kg.slmo!) or a drop in
acetaldehyde selectivity below 100% (for low flow rates, WIF = 349 kg.slmol).

100 . -----.--... ----.---.----..- ......---.-----------------.....---.--.-.-.----------.

90
~.. ·:';:..::::"O·-·-··"--·:::··"':"".·I
l,,7080 ".r.l--:r"
/"...-
-' - - - -

1i_ ~ \AI /- ." / ' ./' __ equlMbrlum

~ /' / /.F __ WIF:126kg.Slmol
i. 50 ./ / r - 11.. WIF =335 kg.slmot ~ ".
j 40 J./ ... _ - - WIF· 671 kg.slmol 1---
..!! 30 '" - WfF::;. 1342 kg.slmol

! 20 ~'"'-------.,....."-------
10t----~~·~~·--------
o ~.--.--.- ..
240 280 320 360 440
Reaction T (OC)

Figure 4.43: Equilibrium ethanol conversion vs. measured values for a 15 wt % Cu on silica
catalyst (Pdhanol = 0.2 bar for equilibrium curve)

4.3.2. 2-BUTANOL REACTION

The reaction of 2-butanol over MgO and Si02 impregnated with copper yielded MEK and
butenes. MEK was the main product (except for catalysts without copper), with a mixture of
butenes as the by-product. Similar to the ethanol reaction, MgO catalysts gave worse results
due to their low surface area.

143
 ... --:: .......... .:.,··,······-···a._ .
•
I

, .'

, _equi~bfium

---+-WIF '" 206 kg.slmoI r-

270 300 330 360 390

Reaction T rC)
Figure 4.44: Equilibrium 2-butanol conversion vs. measured values for a 15 wt % Cu on
silica catalyst (P2-bulano! = 0.2 bar for equilibrium curve)

For silica catalysts there were an optimum copper concentration on the support (15 wt %),
which gave the highest MEK yields. Catalyst particles in the range of 300-850 !l11l gave the
highest MEK yields. Smaller or larger particles produced increasing amounts of butenes.
For a 15 wt % Cu on silica catalyst the selectivity towards MEK production was close to
100% at 240°C, but declined to between 83% and 86% at 390 °C for various WfF ratios.
Figure 4.44 compares the equilibrium conversion (theoretical) of 2-butanol to MEK, at a
partial pressure of 0.2 bar, with experimental values. For WfF = 51 kg.slmol, the MEK yield
differed considerably from the equilibrium value below 330°C, while for
WfF = 103 kg.s/mol the deviation was below 270°C. Above these temperatures the
experimental values approached the equilibrium values, but were lower due to less than 100%
selectivity towards acetaldehyde production. The reaction conversion will be similar to the
equilibrium value if the reaction goes to completion at a specific temperature. If the flow rate
is too high, however, the reaction will not go to completion, since the reaction time will be
too short.

144
 5. REACTION KINETICS

The conditions required for catalyst testing were described in section 3.2. The reactions must
be performed in the steady state regime. This implies that there should be little or no
catalysts deactivation during testing, hence the catalyst deactivation conditions must be
determined. The reactor must be isothermal and the catalyst particles free from interphase
and intra-particle mass transfer resistance. The reactor must be operated as a differential
reactor. For a plug flow reactor, the relationship between the reaction rate and the conversion
is given by the following equation.

w
F
= r~
-rA
(5.1)

Under differential conditions the conversion is kept low (typically below 10%). Equation
5.1 can then be simplified to equation 5.2.

w= X (5.2)

The difficulty with the integration term is removed, while still maintaining very high
accuracy. The following sequence was performed to determine the reaction rate kinetics of
both ethanol and 2-butanol dehydrogenation.
1. Determine the temperature range which results in an acceptably low deactivation rate.
2. Determine the flow rate which results in low conversion (below 10%) and which
minimises interphase mass transfer resistance.
3. Determine reaction rate kinetics at the above temperatures and flow rates.

Small catalyst particles in the range of 300 to 425 !lm were prepared for all experiments to
minimise intra-particle mass transfer resistance. The mass of catalyst for testing was varied
according to the reaction temperature. At high temperatures, a smaller amount was used to

145
reduce the total conversion and keep it below 10%. At low temperatures, the catalyst mass
was increased to ensure a detectable conversion.

For all the data presented in Chapter 5, the following units were used for the reaction rate
parameters:
k': mol/kg.cat.h.kPa
K.q : kPa'l
Ki: kPa'l
r: mollkg.cat.h

Initial kinetic model fitting was performed with the above units and then the final step was to
convert the parameters back to the standard units. The standard units were used for process
modelling in Chapters 8 and 9.

5.1. CATALYST DEACTIVATION TESTING

The effects of reaction temperature and reduction temperature on catalyst deactivation were
investigated. For both ethanol and 2-butanol dehydrogenation a 14.4 wt % copper on silica
catalyst was employed. Experimental procedures and the testing conditions were discussed
in Section 3.2.

5.1.1. EFFECT OF REDUCTION TEMPERATURE ON ETHANOL REACTION

Different researchers suggest the use of different reduction temperatures for copper· supported
catalysts (see discussion in 4.1.2.2). Three reduction temperatures were tested and two sets
of experiments were performed to determine the effect of reduction temperature on catalyst
stability. The results are shown in Figures 5.1 and 5.2.

Reduction at 340 DC and 400 DC resulted in low initial reaction rates. Since the initial
reaction rates were much lower than the initial reaction rates for catalyst reduced at 255 DC, it
was concluded that reduction at 340 DC and 400 DC caused changes in the copper particle
size. Figure 5.1 compares the curves for reduction at 400°C and 255 DC. The initial activity
of the catalyst reduced at the higher temperature was similar to the activity of the catalyst

146
reduced at the lower temperature after about 12 hours reaction. This was considerably longer
than the reduction time (2 hours) of the catalyst reduced at the higher temperature.

100 70

90 b----f.:R.;;;"";di_;;;;T";T~;;;.,,;;;C·~C·-;;;_;;;j,;;;ooT:T~;;;400;;;C";c] I . Re6.ldlon T = 3«!'C; r..ction T = 340'C ~

80 '.,.~ .,~...... I . Redudlon T" 2505 oC; ntKtlon T. 400 ~
60

I\,: 1 • Redudlon T = 255 'C; ruction T" 340'C

';:C 70 ..;-...... .~
~ 1;1 60 +-------'."'"''-''",---------
~ ~ 50 +------',::,'''''.~------­ ..
~o ~~ 40 11,C-
•• -----'-,.-',,"'-"'=------
f\ '.
--""~
It - 30 "
'\
•_ _ \
20+-~.~~.-------------2-
~
10+----~----------

o+---~--~-~--~-~
o 5 10
Time on stream (h)
15 20 25
10

,
, ----
10 15
n,. on stream (h)
20 30

Figure 5.1: Effect ofreduction T on Figure 5.2: Effect of reduction T on

acetaldehyde production rate over time acetaldehyde production rate over time
(14.4 wt % Cu on silica) (14.4 wt % Cu on silica)

The reducing temperature, the reducing environment and the effect of water formation from
the reduction reaction all have an effect on the change of copper particles during reduction.
Although there was initially no coking, it did occur over time. From Figures 5.1 and 5.2 it is
clear that sintering of copper occurred at reduction temperatures of 340°C and 400 DC. To
avoid the loss in catalyst activity all further reduction was performed at temperatures between
250°C and 260 DC.

5.1.2. EFFECT OF REACTION TEMPERATURE AND ADDITIVES ON CATA-

LYST STABILITY
Tests to determine the stability of the catalysts were conducted at 400, 340, 280 and 220°C.
Results are presented in Figure 5.3. At 400 DC, the main by-product was ethylene and the
selectivity towards acetaldehyde remained high (> 90%) over the tested 24 hour time period.
At 340°C, the main by-products were MEK and butyraldehyde and the selectivity towards
acetaldehyde was greater than 90%. At 280°C, the selectivity was on average about 95%
and at 220 DC, the selectivity was 100%. Table 5.1 summarises the reaction rates as a
function of time.

147
 90.0 '.'.'.''.
'{. ~ Reaction T =400 ·C
80,0 ~
• Reaction T =340 ~C
70.0 .~
.a .. .. Reac:tion T =280 'C
1!5 60 ·0
c i t _ .......
•
_: ~ • Reaction T =22O'C
-

'~2 i~ 4.0t~:=.~.!.~.~.~~"~;;=;:::~:~
50.0
_____ ...
CL - 30.0 t-:-:-----:--;---------------=::~~.t::;::=-l
20.0 t-1~~~·i·~~~======~======~~~.L4
10.0 •••• ••••• ••

0,0 -I-'---~--_--~--~--___j
.., ..

o 5 10 15 20 25
Time on stream (h)

Figure 5.3: Acetaldehyde production rate as a function of time and temperature (14.4 wt %
Cu on silica)

..
Table 5 1· Summary of catalyst deactivation at different reaction temperatures
Reaction T (0C) Reaction rate (O.5h) Reaction rate (24h) % decline
moIlkg cat. h moIlkg cat. h
400 87.7 22.6 74
340 54.5 37,0 32
280 23.3 20.5 12
220 7.0 6.3 10

At 220°C and 280 °C, the catalyst remained stable throughout the 24 hour testing period, In
absolute terms there was very little change in reaction rate.

Chromium and cobalt were added to copper to try and improve catalyst stability. Stability
tests were repeated with a 13.5 wt % Cu, 1.0 wt % Cr and 0.5 wt % Co on silica catalyst. The
results are shown in Figures 5.4 and 5.5. Similar stabilities to that of a pure copper catalyst
were achieved. The catalysts with additives did, however, show poorer activity towards the
dehydrogenation of ethanol. At each temperature tested, the acetaldehyde production rate
was lower over time, on average, as indicated in Table 5.2. The CuiCr/Co on silica catalyst
was also less selective towards acetaldehyde production. At 400°C, the main by-products

148
were ethylene, MEK and butyraldehyde (similarly at 340°C and 280 0c). Selectivities are
indicated in Table 5.2.

100 30
90 .'. • 14.4 'JI.CuCI'llllca, r_400"C

80
~
.13.5%01, 1%Cr, O.5%Coon wilk:II; T" 4OO"C
..1<t4%CUontU.... T=340OC f- .,..,.. .
~~~~-.----------------~
.s - 70
! ~ 80 !~. . .13.5%01,1% Cr, O.5%Coon liIica; T" 3IIO"C
i~" I,".~ ••
.
'>..
..
'....... . .
gu
il ,. 50 :..... .:~- ..........
gu I_~'~::~==========~~~
11 ~0 -'"
.t.e. 30
1'(

.
-......... :---- ,..
.. il~15+-
• 0
.

£.510 r-~=.=;::~=;:::;:;===;-----j
----....
-- I I
20 !d4.4"AoCUonsllC<l,T_280·C
10 5
o~ r-l.
13.S%Cu, '-.Cr,o.ncoondlca;r_280'C/!----i
______________________ ~

o 4 , 12 16 20 24
o 8 12 16 ,.
TIme on stream (h) Time on stram (h) "
Figure 5.4: Effect ofadditives (Cr and Co) Figure 5.5: Effect of additives (Cr and Co)
on acetaldehyde production rate on acetaldehyde production rate at 280°C

..
Table 5 2· Performance ofCU/Cr/Co on silica catalyst
Temperature (OC) Average decline in reaction Selectivity (%) towards
rate over time (%) acetaldehyde
400 22 70-80
340 35 > 88
280 23 70-90

5.1.3. FURTHER DEACTIVATION TESTING

Further deactivation studies were conducted using XRD, TEM and TOC measurements. The
results from XRD measurements were used to calculate crystallite size before and after
reaction. Table 5.3 lists the results obtained. Crystallite size was calculated using Schearer's
relation. The peak widths were less than 0.2°, which made XRD not very suitable for
detecting sintering. The broader the peaks, the smaller the crystallite sizes. In general,
sintering is indicated by peak narrowing for tested samples, compared to a freshly reduced
sample. In this study, crystallite sizes measured by XRD were too large to be able to draw
accurate conclusions from XRD data.

149
 ..
Table 5 3· XRD and Toe results of tested catalysts
Temperature Peak width from Crystallite size Carbonwt %
(OC) XRD (degrees) (nm)
Reduced, but 0.20° 60 0.065
unused
220 0.15° 80 3.213
280 0.15° 80 3.386
340 0.14° 80 3.693
400 0.15° 80 3.737

.............,<-== TEM (20 nm) Figure 5.6c: TEM

150
f"

Figure 5.6: TEM images of 14.4 wt % copper on silica catalyst after being in use at 400°C

TEM studies showed a very wide size distribution of copper particles in the utiused catalyst.
Some areas had copper agglomerates in the 1 micron range, down to 40 to 50 nm. Other
areas had small copper crystallites of 4 to 5 nm, evenly distributed over the support (see
Figures 4.30a to 4.30i). After reaction at 400°C, there was definite agglomeration of small
crystals (4 to 5nm) to yield bigger polycrystalline particles (Figures 5.6a to 5.6i).

Some areas had no copper (Figure 5.6a), similar to observations made of the catalyst before
reaction. The small crystallites, in the order of 4 to 5 nm (Figures 4.30b to 4.30d),
agglomerated to form particles in the 20 nm range (Figures 5.6b to 5.6d). Larger
agglomerates were more frequently observed after reaction than before reaction (see
Figures 5.6f, 5.6h and 5.6i). This confirmed that some sintering occurred during reaction at
400°C.

Chemisorption results are shown in Table 5.4. In each case the surface area was the value
after 24 hours at the reaction temperature. The values indicated a decline in copper surface
area with an increase in reaction temperature, confirming TEM results. Sintering occurred
even at the low temperatures.

151
Table 5.4'. Copper surface areas for 144 wt % Cu on silica catalysts
Reaction T (DC) Copper area Copper area % Decrease
(m1/g sample) (m1/g Cu)

Fresh catalyst 0.56 3.76

220 0.47 3.15 16
340 0.44 2.96 21
400 0.37 2.49 34

Catalyst reactivation studies were performed to obtain more information regarding the
deactivation mechanism. The 14.4 wt % Cu on silica catalyst was reactivated after 24 hours
and 48 hours of operation at 400°C. This was done to determine whether coking, sintering,
or both, were the dominating deactivation mechanism. Figure 5.7 indicates a decline in
initial acetaldehyde production rate (about 25% to 30"/0) after the first re-oxidation step,
compared to the first run. This is in good agreement with results in Table 5.4. After the
second re-oxidation step, the same initial activity as after the first re-oxidation step was
obtained. From this experiment it could be concluded that some irreversible sintering
occurred during the first 24 hour reaction period. Deactivation that occurred in the second 24
hour period could be recovered by oxidation and hydrogen reduction.

120
i
• fresh catalyst
100 \ • after 1st oxidation ~
.\
,
... after 2nd oxidation
!

\
"". •
~
!
;
;

20
~• •
• ,,

o
,;
o 12 16 20 24
nnw on stream (h)

Figure 5.7: Catalyst activity (14.4 wt % Cu on silica) after re-oxidation

152
5.1.3.1. Deactivation mechanism
The following mechanism for catalyst deactivation is proposed: At all temperatures above'
220°C there is some initial sintering within the first 24 hours of use and the sintering rate
increases with temperature. At 400°C, the reduction in copper surface area after 24 hours of
use is twice the amount of the catalyst used at 220 °C for 24 hours. Results of re-oxidation
experiments indicate that most of the sintering occurs within the first 24 hours of use.

Sintering and carbon deposition on the catalyst takes place simultaneously. At the higher
temperatures, mainly coke-like carbon is deposited, which reduces catalyst activity and
prevents ethanol molecules from reaching the active copper sites. At the lower temperatures
(280°C and lower) carbon is present in molecular form, probably as oligomers and polymers
that do not prevent ethanol from reaching the active sites.

Coking contributes more to catalyst deactivation than sintering. Figure 5.7 indicates a
decrease in acetaldehyde production rate of 80% for the second 24 hour run at 400°C after
the first re-oxidation step. This value is considerably higher than the decrease in
acetaldehyde production due to sintering.

5.1.4. CATALYST STABILITY FOR DEHYDROGENATION OF 2-BUTANOL

A 14.4 wt % copper on silica catalyst was employed for the dehydrogenation of 2-butanol.
Stability was tested at 250°C and 310°C. Selectivity towards MEK production was very
high (> 99%) at both temperatures. At 250°C, the catalyst was very stable over a 24 hour
period, but deactivation took place when the temperature was increased to 310°C. Further
deactivation testing was not conducted. The same sintering effects that were present for the
ethanol catalyst are also valid for this reaction, as the same catalyst was used. Deactivation
due to coking will differ, because different organic species are present in this reaction, giving
rise to different coking kinetics.

153
 .
160

..
~-
140

120
.~

~~
\
- .T=310·C
_T=310·C
J.T=250"C
r---
I--
c 5 100 ..

~::s :1at
~:l:0
,..s
e.
80

60
•
, .
•
~
W 40
:IE
20
o
o 5 10 15 20 25
Time on stream (hI

Figure 5.8: MEK production rate as a function of time for a 14.4 wt % Cu on silica catalyst

5.2. REACTION MASS TRANSFER RESISTANCE

To obtain accurate kinetic data, the reaction must be operated in the region free from mass
transfer resistance. Mass transfer resistance is dependent on the particle Reynolds number,
which in tum is a function of the linear gas velocity past the catalyst particles. The linear gas
velocity can be increased to eliminate mass transfer resistance by increasing the feed flow
rate and/or decreasing the inside diameter of the quartz tube housing the catalyst. A quartz
U-tube with small inside diameter (4 mm) was used for all experiments.

5.2.1. DETERMINING THE MASS TRANSFER LIMITING REGIME FOR

ETHANOL DEHYDROGENATION
The liquid feed flow rates were increased from 2 ml/h to 25 ml/h using a perfusion pump.
The experimental procedures used for determining the mass transfer limiting regime were
described in section 3.2.1.3 and the mass of catalyst used at each temperature was listed in
Table 3.9. The results are presented in Figure 5.9.

There was significant interphase mass transfer resistance for ethanol dehydrogenation at the
higher temperatures. At higher temperatures, the difference between the reaction rate
limitations and the mass transfer limitations increased, because the reaction rate became
faster (less resistance). A small increase in the feed flow rate (which determines interphase

154
mass transfer resistance), would thus result in a larger increase in acetaldehyde production
rate at high temperatures as compared to lower temperatures.

350

.T-200"C
300 r-- -T-225°C ,
,.
_T_250·C
,•
.
I--
-T-275"C
I ~
I-- "T-300·C
,
,
.t---~
.. . •
··V • •

50
V: // ...
!-_.-- -
:
--'.
•
~""t-~~~",=-.- . .;:::-::r
o
o 5 10 15 20 25 30
Feed flow rate (mllll)

Figure 5.9: The effect of feed flow rate on acetaldehyde production rate

At temperatures of250 °C and below, the reaction rate was fairly constant above a feed flow
rate of 14 mVh. At 275°C and 300 °C it appeared, from the data, that the region free of mass
transfer resistance commenced at a feed flow rate of30 mVh. Due to limitations of the feed
pump, it was not possible to do full kinetic studies at such high flow rates. All data for
determining kinetic parameters was obtained at a feed rate of 14 mllh. In the modelling
calculations the final rate data at the higher temperatures was to be fitted to the curves in
Figure 5.9, to compensate for rate reduction from mass transfer resistance.

5.2.2. DETERMINING THE MASS TRANSFER LIMITING REGIME FOR 2·

BUTANOL DEHYDROGENATION
Flow rates of2 to 16 mllh were tested (see Table 3.10). For the 2-butanol dehydrogenation
reaction, there was no significant interphase mass transfer resistance (Figure 5.10). The
MEK production rate remained fairly constant with an increase in 2-butanol feed flow rate at
all temperatures tested. There were some fluctuations in the MEK production rates, with a
slight downward trend in production rates at the lower temperatures. For modelling
purposes, rates at different feed flows were averaged at each temperature. Raizada, Tripatbi
et al. (1993) reported similar results on interphase mass transfer resistance for the

155
dehydrogenation of n-butanol over zinc oxide. They concluded that bulk diffusion was not
significant.

250
,
~

200 •
-------- .. ..-- ..• --=1
;,
--~-.-."

•
.'H'" • !
"- ,,I +19O"C
.. ..J 150
~
,,
Ii
•
"
.. !
,
.220 o e
A250"C
100
'U
... -
i
<lIIo 280 "C

50
• •
• • .. •
i
!
0
0 2 , 6 8 10 12 l' 16
Feed now rate (mllh)

Figure 5.10: The effect of2-butanol feed flow rate on the MEK production rate

5.3. KINETICS FOR ALCOHOL DEHYDROGENATION

Kinetic data for ethanol dehydrogenation was fitted to the dual site reaction equation
(equation 2.18). Previous researchers had found this equation to best fit kinetic data for
ethanol dehydrogenation over supported copper catalysts (see Section 2.12.5.3), The validity
of this equation for the reaction conditions ranging from 200°C to 300 °C was also tested.
The reaction equation for the duel site surface reaction mechanism is:

(5.3)

with A, R, S = ethanol, acetaldehyde, hydrogen

Numerical values for the rate and adsorption coefficients could be calculated using least
square analysis. Equation (5.3) must be linearised in the unknown coefficients. Rearranging
equation (5.3) yielded:

156
,<

or (5.4)

y = a + bP A + CPR + dPs with (5.5)

y= and (5.6)

(5.7)

The constants k' and KA were expressed by:

k'=~ and (5.8)

a'

K -~a
A - (5.9)

The constants a and b were determined first (using a pure ethanol feed at different pressures)
from the y-axis intercept and the gradient of the best linear fit to the data. The ethanol partial
pressure was varied by feeding it with an inert gas (nitrogen). Thereafter the ethanol was fed
with hydrogen and acetaldehyde separately, while keeping the total pressure constant at one
atmosphere. Different molar feed ratios were used. Equation (5.5) can be expressed as:

y = a + bP A + d{ I 0 1-PA) = (a + 10 I d) + {b-d)P A (5.10)

where d and c are used depending on whether acetaldehyde or hydrogen is used. From the y-
axis intercept and the gradient of the best linear fit at each temperature the remaining
constants could be determined (in 5.7). The process was repeated at each temperature, to
determine the Arrhenius dependence of each adsorption constant. The Arrhenius equation is:

157
 (5.11)

Linearising equation (5.11) allows for A and E to be determined.

Ln(k) = In(A) -EIRoT (5.12)

5.3.1. ETHANOL DEHYDROGENATION

The parameters in equation (5.4) were determined at temperatures of 200°C to 300 °C with
the dual-site reaction mechanism. The linear fits to the data, which were used to determine
the parameters, are given in Appendix Ct. Figures 5.11 and 5.12 show typical linear fits at
225°C and 300 DC, respectively.

y" '1.01£-011 +,.

1eEtOl 'I" .;l.ll!f4tJ: +l.!I2!:oIOO
R!"9.77E-Dl ~"UH-Dl

>0 4.0 :oil I ...... I •••

y" ..2.O!EoOax 1" 1B6E+OO y '" 8.12Eo04ll + UeI!-o'I

R''''.2tE-Dl Rl"&.aef.e1 .. d'rJ ......

" 'I=2.09E-03II:+l.ll3E-Dl _ _ _ _ ...

Y=4.3n!-Ok+1.05E+OtI l1li .allnfln 1t1"8.l1I!!:0(11
Rl =U7E.(Il
,.,.J-~~~~~~~~~~~~r---<
~ ~ ~ ~ 00 ~ m ~ ~ ~ w ~ ~ 1~ ~ ~ ~ ~ 00 ~ m ~ ~ ~ ~ ~ 100 1~

Ethanol pRaure (kPa, Ethanol pressure (kPa)

Figure 5.11: Linear fits of reaction rate data Figure 5.12: Linear fits of reaction rate
at I atm. total pressure and 225°C data at 1 atm. total pressure and 300 °C

The y-value was proportional to the inverse square root of the reaction rate (equation 5.6). A
decrease in the reaction rate would thus cause an increase in the y-values. For the ethanol-
acetaldehyde feed there was a sharp increase in the y-values (decrease in reaction rate) with
an increase in acetaldehyde concentration. This indicated that the reaction rate was most
sensitive to the acetaldehyde concentration. Acetaldehyde adsorbed strongly on the catalytic
sites and reduced the dehydrogenation reaction rate. Hydrogen adsorption on the surface was
not very strong and ha only a small effect on the reaction rate.

158
Good linear fits for the acetaldehyde-ethanol feed in Figures 5.11 and 5.12 indicated that the
reaction rate was not dependant on the acetaldehyde pressure, but on the square of the
acetaldehyde pressure. Table 2.5 listed other possible reaction rate mechanisms. Only
equations. (2.18) and (2.19) had the acetaldehyde pressure square term. Furthermore,
equation (2.19) did not have adsorption coefficients for ethanol, hydrogen and acetaldehyde,
but only for either ethanol and hydrogen or ethanol and acetaldehyde. The data confirms that
equation (2.18) is most suitable for describing the rate mechanism.

The value of each parameter at the different temperatures is given in Table 5.5.

Table 5.5: Reaction rate parameters for ethanol dehydrogenation

T k KA Ks KR
eC) moVkg cat.h.kPa kPa'! kPa-! kPa-!
200 0.9615 0.00707 0.00752 0.1197
225 0.9086 0.00417 0.00593 0.1038
250 1.8699 0.00514 0.00324 0.0694
275 4.5697 0.00432 0.00291 0.0657
300 6.8212 0.00546 0.00368 0.0818
(A = ethanol, S = hydrogen, R = acetaldehyde)

The adsorption coefficients (KR and Ks) decreased with an increase in temperature from 200
°e to 275°C. The value of KA should theoretically also decrease and the calculated values
were in reasonable agreement with the theory, excluding the value at 300°C. The reaction
rate coefficient (k') increased with temperature (as expected). This was typical for a reaction
which follows the dual site, surface reaction mechanism (Peloso et aI., 1979). At 300°C, the
adsorption coefficients deviated from the downward trend. The values were higher than at
275 °e. To comply with the dual site, surface reaction controlling mechanism, the adsorption
coefficients must decrease with temperature, and they must be positive (Peloso et aI., 1979).
The increase in the calculated adsorption coefficients at 300 oe, compared to the values at
275°C, is illustrated in Figure 5.13.

159
.'""
.. '

75

65 /
l 55 L
=
~
~

g 45 /
"~35
~
/
-5'
<0;
25 /
15
....- /:
5
220 240 260 280 300 320 340 360 380 400
T ('C)

Figure 5.13: % Catalyst deactivation after 24 hours of operation

There was little change in the deactivation rate of the catalyst at temperatures from 200°C to
275 °C (from 10 % to about 11.2 % over a 24 hour period). At 300°C, the deactivation rate
dropped to 16.5 % after 24 hours of operation. During the testing period at 300 °C (about
two days), some deactivation would thus have taken place. This would result in the reaction
rate values being lower than the true values, due to catalyst deactivation. These lower
'falsified' reaction rate values are reflected in the rate mechanism as higher adsorption
coefficients, Stronger adsorption or higher adsorption coefficients lead to slower rates,
Nonetheless, the model still holds at higher temperatures, even though the changes in catalyst
activity gave rise to conflicting parameter values.

The Arrhenius parameters for k', KA, KR and Ks were determined only at temperatures from
200°C to 275 °C (thus excluding values at 300°C) using equation (5.12). AU adsorption
1
coefficients (KA, KR and Ks) are given in kPa- • The foUowing expressions were obtained
(with Tin K):

For T = 473 K to 548 K:

Ln(k) = 11.291-5491.1/T (R2 = 0.825) (5,13)
Ln(KA) = 135 1.1 IT - 7.941 (R2 = 0.487) (5,14)
Ln(Ks) = 3586.6/T - 12.444 (R2 = 0.939) (5.15)
Ln(KR) = 2288IT - 6,940 (R2 = 0,927) (5.16)

160
The data is shown in Figure 5.14.

,,---------------------------------,
•

l-, __========~======~y;·~~~1.'~x+~"~.~:'~~~
O.
o
18 0.0019
•

0.002
R1 • 0.8253

0.002

., y::; 2288x - 6.9401

Rl::; 0.9271
>-
-3 • •
-4
y = 1351.1x~ 7.9413
Rl = 0.4869 , ...
,
. •
y = 3586.6x-12.444
Rl::; 0.9393
-7 - - --..... ----.----. -_ .......-_.......... __ ..... _- --.. ---..-._---.----_...--.... --- .. --.. --_..------.---..--.
1fT (K")

Figure 5.14: Parameters for ethanol reaction equation as a function of temperature

Keq is a constant and could be found in the literature as an empirical correlation (peloso et al.,
1979) or calculated using firsts principles via the Van't Hoff's equation (Fogler, 1992).
Many of the correlations were very old and their accuracy doubtful, with the result that an
expression was reduced from first principles.

AG~ = G~, + G~ot - G~t (5.17)

AG=AH-TAS (5.18)

K =ex
p p
[-AG~(T)]
RT (5.19)
o

Kp in equation (5.19) is defined in atmospheres. Using thermodynamic data for ethanol and
acetaldehyde, Kp (in atmospheres) could be calculated at different temperatures. Kp values
were changed to K.q values in kPa (to standardise all K-values) and then the Arrhenius
equation could be fitted through the different Keq-values to obtain the necessary Arrhenius
parameters (similar to 5.11 and 5.12). The equilibrium constant is then:

161
Ln(Keq) = 19.014 - 8503.3/T (5.20)

The reaction rate from 473 K to 548 K and at 1 atmosphere feed pressure could now be
expressed as:

. _ 8.014xlO xe
4 -S49I. lf ( P -P P /(1.810X10 xe
IJ -8503.~)
)
A R S
- r _ - - -_ _ _ _ _ _ _----'--.:..:._-'..:--"--_ _ _ _ _ _.L-_ _ _--::- (5.21)
A (
1+3.557xlO·'xe
13Sl.){
xP +3.941xlO·'xe
3586.%
xP +9.683 x lO·'xe
228~ xP ) 2
A S R

with pressures in kPa and temperatures in K. The accuracy of equation (5.21) was tested for
all the data from 200 °C to 275 °C (Figures 5.15 and 5.17).

5.3.1.1. Equation optimisation

Equation (5.21) represents the parameters obtained using mUltiple linear regression analysis.
This will be called model 1. A second model was developed to minimise the error between
actual measured reaction rates and the calculated reaction rates.

The structure of the error function was of great importance. Either the model percentage
deviation could be used or the actual model value difference. In mathematical format:
• 100*(model value - measured value)/measured value or
• model value - measured value.

The reaction rate increased from 33 (mol/kg cat.h) at 200°C to 227 (mol/kg cat.h) at 275°C.
At high temperatures, a percentage deviation of 20 %, for example, would result in a large
difference in measured reaction rates, while the difference in measured reaction rates at low
temperatures and at the same percentage deviation, would be small. It was more important to
correctly predict reaction rates at the higher temperatures than at the lower temperatures, thus
the actual deviation and not the percentage deviation was used in the error function. Since
the deviations in actual rates at the higher temperatures were so much larger than at the lower
temperatures, different weights were applied at each temperature to the error function. The
error function is as follows:

162
!i'

EF = ~(CV
L.. _ MV)'.
T_472K +
(CV - MV»)' + (CV - MV»)'
1 1.125 T.498K 1.25 T.S2JK

+ (CV - MY»)' (5.22)

2 T.548K

with EF = error function, CV = calculated values, MV = measured values and n the number
of experimental points.

The model deviation (MD) from the measured data is simply the following:

MD = I:(CV - MV)' (5.23)

There were 62 experimental reaction rate values. The values in equation (5.21) were used as
the starting values for further optimisation. The parameters were varied using an 8
dimensional matrix (varying each one of the eight parameters in equations 5.13 to 5.16). The
smaller the error function, the closer the calculated values were to the measured values and
the better the fit. As the error function approached the minimum value, the step size for each
parameter was decreased to obtain more accurate values. The final step size employed was
0.2 % steps for each of the 8 parameters. A Turbo Pascal program was compiled to
determine the error function and calculate the parameter values. The best fit was called
model 2. Both model I and model 2 were only for data obtained from 200 °c to 275 °c. The
parameter values are listed in Table 5.6 and the models are compared in Figures 5.15 to
5.20.

There was a significant improvement in the model deviation and correlation coefficient
(R 2-value) of model 2 compared to model 1. When comparing Figures 5.15 and 5.16, it
could be seen that model 2 gave a much better prediction of values at 548 K (275 0c) than
model 1. For model I, the model values at 548 K were much lower than the measured
values. At 523 K, both models gave similar results. At 498 K, model 2 was slightly better,
while at 473 K (200 0c) model I gave better predictions that model 2. The error function

163
was structured in such a way as to minimise the difference between the model values and the
measured values. The main reason for the improvement obtained by model 2 over model I,
was due to better fitting of measured data at the higher temperatures.

Table 5.6: Kinetic model parameters for ethanol dehydrogenation

Parameters Modell Model 2
A 8.014*10" 4.6841*10'
AA 3.557*10~ 1.0078* 1O~'

AR 3.941 *10-0 6.2135*10-0

As 9.683*10-" 6.1854*10-

-EIRo -5491.1 -6424.6

-EAIRo 1351.1 3187.4
-ERIRo 3586.6 3390.0
-EsIRo 2288.0 2245.5

MD 8606.49 4914.21
R> 0.928 0.966
..
A IS In [mollkg cat.h.kPa]; -EfRo is in K

30
•
30
· -----.
.!
20

10
. i l!
"
10
·•
I -----!"
1i
~
0
I * ~
1i 0 I I

1l'" 4 '01 '0) 400 rOO 5" 5'" 530 540 ~ .a

:.: ·10
4
,
o. ..0
"" ~ 510 520 530
""
·10
I I -f.
~ ~ -20
~
'i
·20

·30
•; .."'& ...,
~o
.
~ •
-40

• '" ~o _._---
•
Temperature (K)
... Temperature (K)

Figure 5.15: Difference between model 1 Figure 5.16: Difference between model 2
rates and actual rates at different rates and actual rates at different
temperatures temperatures

164
In Figures 5.17 and 5.18, the percentage deviation between model values and actual values
are shown. For model 2 there are a few points at 498 K and above with very large percentage
deviations (greater than 50%). Those are reaction rates at very high acetaldehyde
concentrations. The reactions rates at high acetaldehyde concentrations are very low and a
small difference between the predicted and measured values cause huge percentage
deviations.

OJ 250

60 ---.. 200 _ _ _ _ ._, _____ . ....!.... __

•
40 !--__ .- -._---. · ,
20 I I
,oo~-----~,--
•
• ••
'" 0
4
•
"" 490 500 "0 520'
· "" 540 , 0
~~--.----------~
0;
·• I :
·20

....
:; · :• ...,°41-0·'--""'---490---l~f.-OO--5-"---52-0~-53-0
"
540"10

.., • .'00 L-_ _ _ _ _ _ _ _ _ _ _ _ _ _---1

Temperature (K) Temperature (K)

Figure 5.17: Percentage deviation between Figure 5.18: Percentage deviation between
model 1 rates and measured rates model 2 rates and measured rates

Figures 5.19 and 5.20 compare the measured reaction rates with the modelled reaction rates.
The solid diagonal line on both graphs indicates a perfect fit.

250
""
200 /. 200 /.
... -

..
~t
'8 ~
;;-
';;;:'" 150
~

100
/....
/. •
•
.h
!
N

.~
5'~
•
150

'00
/..
/" ,

- - ' - ..
'.

•
"- .i:/ • . ? ---
" • 5O~ ....
0 1/ . 0
0 '00 'so 200 250
0 50 '00 '50
Mea.ured rate (mollkg.cat.h)
200
"" 50
Measured rate (mollkg.cat.h)

Figure 5.19: Modell rates vs. actual Figure 5.20: Model 2 rates vs. actual
rates at different rate values rates at different rate values

165
I'

Figure 5.19 indicates that model I fits data well when the reaction rate is below about SO
(mol/kg cat.h), which is the reaction rate at the lower temperatures. At high reaction rates
(high temperatures), the model values are significantly lower than the measured values. For
model 2, the calculated values show more variance (compared to model I) for reaction rates
below 50 (mol/kg cat. h). For reaction rates above SO (mol/kg cat.h), the model values are,
however, much closer to the measured values than for model 1.

It must be stressed that the models were formulated for data from 200°C to 275 °C (473 K to
548 K). Measured reaction rates at 300°C (573 K) were compared to predictions from both
model I and model 2. The difference between model values and measured values are
summarised in Figure 5.21. For both models, the model values were generally higher than
the measured values (see Figure 5.13 and the discussion thereof). Model I predicted
measured reaction rates at 573 K far better than model 2 did. The correlation coefficients are:
• Modell: R2 = 0.953 and
• Model 2: R2 =0.918.

160
• •
140 - • Model 1
~

~ 120 • Model 2
~
c:
0 100
:;:;
''>" 80
•
"
'tI
60 · •
"
C>
~
'"
c: 40 • -.- • • • •
"I:! 20 • • •
a.."
• • •
• • •
0 • • •
·20
2
• • 6 a 10 1"2 14 16 1

Runs

Figure 5.21: Percentage deviation between model values and measured values at 573 K

The reason why model 1 predicted reaction rates at 300°C better than model 2 did, was due
to the fact that catalyst deactivation occurred at 300°C. The measured reaction rates were
much lower than the true reaction rates at 300 °C. Since model I generally under predicted

166
reaction rates, it thus gave a better prediction of the 'lower' reaction rates at 300 °e that
resulted due to catalyst deactivation.

5.3.2. 2-BUTANOL DEHYDROGENATION

Table 2.5 listed possible rate equations for reversible reactions. Of equations 2.17 to 2.21,
only 2.18 and 2.19 show an inverse quadratic relationship towards the MEK partial pressure
(or concentration). Reaction rate data in this study indicated a strong inverse quadratic
relationship between the observed reaction rate and the MEK partial pressure. Perona and
Thodos (1957) determined reaction kinetics for the dehydrogenation of 2-butanol between
343 °e and 399 °e over solid brass spheres (65% copper and 35% zinc). Under those
conditions, they found the desorption of hydrogen from a single site (equation 2.21) to be rate
limiting. Ford and Perlmutter (1964) used a brass tube (60% copper and 40% zinc) as
catalyst and carried out the dehydrogenation reaction at temperatures between 316 °e and
427 °e. From 350 °e to 400 °e alcohol adsorption was rate limiting, while at both higher
and lower temperatures the single site surface reaction was rate limiting. Thaller and Thodos
(1960) performed experiments with smaller brass catalyst particles (50 to 60 mesh; 65%
copper and 35% zinc) with a larger surface area. Below 300 °e the reaction was dual site,
surface reaction controlling (equation 2.18), while at higher temperatures the reaction was
dual site, hydrogen desorption controlling.

For equation 2.19 to be applicable, the initial reaction rates for different hydrogen-2-butanol
feeds had to be independent of 2-butanol pressure (Thaller and Thodos, 1960). That was not
true and equation 2.19 could not be used. Equation 2.18 provided a reasonable fit to the data
obtained in this study. The parameters in equation (5.4) were determined at temperatures
from 190 °e to 280°C for the dual site, surface reaction controlling mechanism. The linear
fits to the data, which were used to determine the reaction rate parameters, are given in
Appendix C2. Figures 5.22 and 5.23 show typical linear fits at 220°C and 280 °e,
respectively.

167
The reaction rate was most sensitive to the MEK concentration. The large negative slope
indicates strong MEK adsorption: The adsorption coefficients at the different temperatures
are listed in Table 5.7 .

• , 2-8ut_I.MEK

y .. -e.8llll&02lo.+1.126EotOD 3

,.., RI.9.844&01
y· .... l11E-02Jc +4.71BE..oQ
R' .. I.MlS-Ot
2 y" 2.21SS03I<+ 1.138E+OO
rf .. 5.747&01 y .. , .7ME-03k + 6A18E-01
2-aut.wN. NIIfogIwI R' .. 8.7406-01
'I .. 3.574&0310. + 11._6-01
P;'-7.9IIEI&Ol

~ • • • '" ro '"
2-Butanol partial pressure (kPa)
'" .00 •• o 20
• eo eo
24*1nol partial pressure (kPa)
". ".

Figure 5.22: Linear fits of reaction rate data Figure S.23: Linear fits of reaction rate
at 1 atm. total pressure and 220°C data at I atm. total pressure and 280 °C

Table 5.7: Reaction rate parameters for 2-butanol dehydrogenation

T k KA Ks KR
(OC) mol/kg cat.h.kPa kPa-1 kPa-1 kPa-1
190 0.339 0.001433 -0.00129 0.11735
220 0.772 0.001195 -0.00126 0.06075
250 1713 0.002987 -0.00225 0.05232
280 3.342 0.003225 -0.00109 0.06432
(A = 2-butanol, S = hydrogen, R = MEK)

The trends in the k' and KR values were in line with the theory (peloso et al., 1979), except at
280°C where KR showed an increase instead of a decrease. From initial experiments, the
same conclusion could be drawn for this reaction as for the ethanol dehydrogenation reaction
(see 5.3_1). Very little catalyst deactivation took place at 250°C and below (see Figure 5.8),
but at 310°C (Figure 5.8) significant deactivation took place. Both sintering and coking
played an important role in catalyst deactivation at the higher temperatures. The observed

168
reaction rate would be lower due to changes in catalyst activity and not due to stronger MEK
adsorption.

Both KA and Ks (the adsorption coefficients for 2-butanol and hydrogen) were negligible
compared to the adsorption coefficient of MEK (KRJ When adsorption took place, the
reaction rate slowed down. This was because diffusion resistance of the feed molecules to
the active sites increased. Negative adsorption coefficients indicated an increase in reaction
rates. The negative hydrogen adsorption coefficients (Ks-values) contradicted the theory of
the dual site, surface reaction controlling mechanism, but the values were so small that the
equation still fitted the data very well.

The reasons for the increase in reaction rate with hydrogen in the feed have been documented
for other dehydrogenation reactions and was not unexpected where coking tended to
deactivate the catalysts. Sheintuch and Dessau (1996) cited many references where hydrogen
was co-fed with either an alcohol or an alkane and where improved dehydrogenation activity
was reported. Hydrogen in the feedstream reduced coking (Sheintuch and Dessau, 1996) and
it reduces the partial pressure of the alkane or the alcohol, which is favourable for higher
conversions (Ertl et aI., 1997).

KA and Ks were taken as zero for a first approximation. The following Arrhenius expressions
were formulated for data obtained at temperatures from 190°C to 250 °C (463 K to 523 K)
by using equation (5.12), (temperature is in K). Figure 5.24 represents the data graphically.

For T = 463 K to 523 K:

Ln(k') = 13.628-6903/T (R2 = 0.944) (5.23)
Ln(KR) = 3298/T - 9.3377 (R2 = 0.905) (5.24)

The equilibrium constant for 2-butanol dehydrogenation could be expressed by the following
equation (Kolb and Burwell, 1945):

Log(:Kp) = -2790/T + 1.51 *logT + 1.865 (5.25)

169
 2

1 ...
~ y - -6903, + 13.628
R2 =0.9441
0

>-
0.0 17 0.0018 0.0019
~21 0.0 22

·1
--.....:
·2.
Y- 3298, . 9 . 3 3 7 7 _ _ . _ _ _ _ _
R2 .. 0.9047
·3 •
-4
1fT (11K)

Figure 5.24: Parameters for 2-butanol reaction equation as a function of temperature

Kp, in atmospheres, was solved as a function of temperature and changed to ~q in kPa. The
function was then expressed in the exponential form (see equation 5.11) and substituted into
the rate equation (5.3) to yield equation (5.26).

The reaction equation from 190°C (463 K) to 250°C (523 K) could now be expressed as:

(5.26)

with pressures in kPa and temperatures in K.

5.3.2.1. Equation optimisation

The reaction rate parameters for the dehydrogenation of 2-butanol were optimised in the
same manner as those of the dehydrogenation of ethanol reaction. A detailed explanation of
the steps was discussed in section 5.3.1.1. For the 2-butanol reaction the error function was
taken as:

170
EF = I(cv - MV)~=463K + (CV - MV)]' +(CV - MV)]' (5.27)
I 1.125 T=493K 1.25 T=",K

The different model parameters are listed in Table 5.8, where model 1 is the values
determined with multiple linear regression and model 2 is the optimised values. Both models
were only valid from 190°C to 250 °C and at a 2-butanol feed pressure of 1 atmosphere.

Table 5.8: Kinetic model parameters for 2-butanol dehydrogenation

Parameters Modell Model 2
A 8.290*10" 1.240*10"
AR 8.804*10'0 8.924*10'0

-EIRo -6903 -6042

-EAIRo 3298 3165

MD 12724 3162
R- 0.9478 0.9430
..
A IS III [moVkg cat.h.kPa], and
-EIRQ is in K

The most important indicator of model performance, relative to measured data was the model
deviation (MD value). The lower the deviation, the better the model. The optimised values
of model 2, compare to modell, gave a more accurate prediction of reaction rates. The
exclusion of adsorption coefficients for 2-butanol and hydrogen in the reaction rate equation
was an acceptable simplification. Predictions remained accurate without these parameters in
the rate equation. Figure 5.25 compares the performances of model 2 and model 1. Model 2
is more accurate at the higher reaction rate values.

171
 160 ...........-- -- -"-.-... --.. -.---. -_........ _................... ----_........ -_•.....•......•.....
•
'40 • •
:c
oJ
120

3'00

~
'" 80
g
Ii 60

~ 40

20

o ~ 40 80 80 '00 ,~ ~ '50
Measured value. (mollkg c:ath)

Figure 5.25: Comparison between measured reaction rates and model reaction rates

5.4. SUMMARY
Catalysts reduced at 340°C and 400 °C showed a sharp decline in activity after reduction
compared to catalysts reduced at 255°C, mainly due to sintering. At 280°C and below, a
14.4 wt % copper on silica catalyst remained stable over a 24 hour period. At
340°C and higher, the catalyst deactivated by more than 30"10 over a 24 hour period. The
addition ofCr and Co to Cu did not improve the stability of the 14.4 wt % copper on silica
catalyst. The activity of the CulCr/Co on silica catalyst was similar to that of pure copper on
silica, but the selectivity towards acetaldehyde production was significantly lower at all
temperatures tested. Further deactivation testing indicated that during the first 24 hour
period, both sintering and coking occurred during the dehydrogenation of ethanol.
Thereafter, coking was the main deactivation mechanism.

For the dehydrogenation of ethanol, there was strong interphase mass transfer resistance,
while for 2-butanol dehydrogenation there was no clear indications of interphase mass
transfer resistance.

The controlling reaction mechanisms and reaction rate parameters for both the
dehydrogenation of ethanol and 2-butanol were determined. Ethanol dehydrogenation was
studied from 200°C to 300 °c and 2-butanol dehydrogenation from 190 ° to 280°C. Both

172
reactions could be well described by the dual site, surface reaction controlling mechanism.
The reaction rate coefficients and the adsoption parameters for each reaction were
determined. In both reactions the organic product (either acetaldehyde or MEK) had a
dominant adsorption coefficient.

173
 6. Pd MEMBRANE PREPARATION AND
CHARACTERISATION

This chapter will present results on electroless Pd plating. Comments will be made on the
various plating steps and the composition of the plating solution. Membrane characterisation
results will be presented and discussed. This will include surface characterisation with SEM
and permeance testing with both hydrogen and nitrogen. For permeance testing, a positive
feed pressure or a sweep gas was employed alternatively.

Permeance results were compared to literature data and the effect of film thickness on
permeance parameters will be discussed. Arrhenius parameters were determined by
performing experiments at different temperatures.

6.1. MEMBRANE STRUCTURE

A cross section view of the SCT membrane was shown in Figure 3.2 (as reported here).
Figures 6.1 to 6.3 show the scanning electron microscope images of the same membrane
structure obtained from a

4: Pd rmallayer

3: AIfa alumina layer 3

2: AIfa alumina layer 2

1: AIfa alumina layer 1

Figure 3.2: SCT membrane structure

174
Figure 6.1: Cross section view of a three layer SCT membrane

In the cross section view there were the three, clearly visible, layers (Figure 6.1). Figures
6.2 and 6.3 show that the surface had a smooth structure, with plenty of pores. The latter
were very suitable for electroless plating.

Figure 6.2: Top view (20 OOOx) of a Figure 6.3: Top view (S OOOx) of a
three layer SCT membrane three layer SCT membrane

6.2. SURFACE PRETREATMENT

Experimental details of the pretreatment process were discussed in section 3.3.2.1. A few
remarks warrant re-emphasised here. The conventional pretreatment solution (as proposed by

175
Shu et aI., 1993) deposited a small amount of tin. Keuler (1997a) found, by PIXE analysis,
between 0 and I % tin in Pd films of 5 microns. Such low tin percentages would increase by
about 3 times when the film thickness is reduced to 1.5 microns, since the pretreatment
procedure remained the same. The result is that in this study a tin to palladium molar ratio
that was about eight times less than the conventional ratio was employed. This was done by
both increasing the palladium concentration and reducing the tin concentration in the
pretreatment procedure.

The tin in the film should not have a significant effect on the hydrogen permeance
parameters. The main reason for reducing the tin was to try and improve its high temperature
membrane stability. In the time of this study, another research group had similar ideas to
improve the membrane stability. Paglieri et al. (1999) started experimenting with
pretreatment procedures without tin. They speculated that tin, with its low melting point (505
K), could enhance metallic diffusion at the grain boundaries and lead to an increase in
defects. They concluded that the presence of tin at the alumina-palladium interface
contributed to selectivity decline.

6.3. THE ELECTROLESS Pd PLATING PROCESS

The composition of the plating solution and the plating kinetics for the plating of the inside of
a membrane tube vary from plating on the outside of a tube. Table 6.1 indicates the
differences between the plating solution composition for plating the inside and the outside
surfaces of a membrane. The optimised plating solution composition for plating outside
surfaces was taken from Keuler et al. (1997b).

Keuler (l997a) used 27.5 g of 10 wt % solution (NH3)J'd(N03)2 per litre of plating solution
for plating the outside of membrane tubes (selective layer on the outside). In the present
study, 4.96 g of (NH3)4PdCh.H20 per litre of solution was used for plating the inside surface
of membrane tubes (selective layer on the inside). When plating the inside of tubes, a much
higher Pd concentration in the plating solution can be used as compared to plating on the
outside. Initially the hydrazine:Pd molar ratio was I to I at the start for plating the inside of

176
tubes. Later it was reduced to the value given in Table 6.1 to slow down the plating rate. In
no experiment did decomposition of the plating solution occur while plating the inside of the
membrane tube, even though a higher Pd concentration was used, in addition to a much
higher hydrazine concentration, initially (hydrazine:Pd of I: 1). The reasons are that the
catalysed membrane surface is the only one available for plating and the volume to available
plating area ratio is much smaller for tubes on the inside. When plating on the outside of the
membrane, plating solution is also in contact with the plating reactor, thus increasing the
available area for Pd deposition.

Table 6.1: Composition (per litre) of plating solutions for membrane plating
Coml!onents Outside of tube Inside of tube
Pd (g) 1.47 2.00
Ammonia (28 wt %) (ml) 200 400
EDTA(g) 100 80
Buffer pH - 10 (ml) 100 -
Hydrazine:Pd molar ratio about 0.7 0.3 5 at start
Temperature (0C) 72 72

6.3.1. SOLUTION FEEDING TO MEMBRANE TUBE

Stirring the tubular membrane, covered with teflon tape on the outside, in the plating solution
at 72 °C resulted in very poor plating on the inside of the tube. There was insufficient
circulation of solution through the tube and the large volume to active surface area also
promoted solution decomposition.

The second solution feeding method tested was continuous pumping of solution through the
tube fixed in the reactor and placed in the water bath. A flow rate of 120 mVh was used and
the same solution was pumped through the tube several times, increasing the hydrazine
concentration after each run. The sharp decline in Pd concentration with time, made this
method unsuitable. At a flow rate of 120 mVh, it took the solution 5 minutes to pass from
end to end in the membrane tube. During the first run the Pd concentration dropped by about
30% during that time, resulting in a film of non-uniform thickness over the length of the

177
membrane. The film was thicker at the entry point than at the exit point, because of the
declining Pd concentration and decreasing reaction rate. For this method to be successful ,
much higher feed flow rates have to be used and the feed direction reversed every few
minutes. It was decided to use a batch process to produce coatings of even thickness along
the full length of the tube.

Figure 6.4 indicates the decrease in Pd concentration in solution as a function of the number
of plating session. Values were obtained using rep analysis. Hydrazine was added after each
plating session. The initial hydrazine:Pd molar ratio was I to 1. Half the initial volume of
hydrazine was added after 10 minutes of plating. After a further 15 minutes of plating
(25 minutes total plating), the original volume ofhydrazine was added and then after a further
20 minutes (45 minutes total plating), a few drops of35 wt % hydrazine was added. Reaction
continued for an additional 30 minutes (75 minutes in total). The results are presented in
Figure 6.4.

2000

1800

1600

1400

1200
"E
Il.
1000
Co
c- eoo
eoo
80 to 91%
400

200
• 95 to 97% 98 to 99.5% 99 to 99.5% 99.5%

0
• :>

0 234 5 6
Plating session

Figure 6.4: Pd concentration in solution after repeated plating sessions

The amount ofPd deposited after the first 10 minute session varied between 80 and 91 % of
the available amount in the 11.5 ml solution. Four different membranes were tested this way.
After 4 plating sessions, 99% of the available Pd was extracted from solution and deposited
on the membrane. Poor plating was observed in two cases and the hydrazine concentration
reduced to the values listed in Table 3.14. The plating procedure was changed to that

178
tabulated in Table 3.15. After the first 20 minute session, between 50 and 60% of the Pd was
deposited on the membrane. After three 20 minute sessions, more than 98% of the palladium
was deposited.

6.3.2. EFFECT OF PLATING RATE ON MEMBRANE PERFORMANCE

The quality of the Pd coating is very dependent on the plating rate. If the plating rate is too
fast, then the coating shows poor selectivity characteristics. The plating rate will become too
fast when the hydrazine concentration is too high, the temperature is too high and/or the
EDTA concentration is too low. The only variable that was changed, was the hydrazine
concentration. Initially, at the start of plating experiments, the hydrazine:Pd molar ratio was
1 to 1, with plating characteristics as in Figure 6.4. Ten coatings were prepared in this way
on 200 nrn Cl-alumina membranes. Two were defective, because Pd bubbles formed on the
membrane surface, causing leaks. After reducing the hydrazine concentration this problem
was solved. Figures 6.5 to 6.11 show SEM images of the two poorly performing
membranes. The calculated Pd thickness of membranes (a) and (b) were about 1.5 microns.
SEM images (Figures 6.5 and 6.8) show similar thicknesses. Appendix DI lists all the
membranes on which extensive permeance testing was performed. Extensive permeance
testing was not performed on membranes (a) and (b) due to their poor selectivity, and therefor
they are not listed in Appendix Dl.

Figure 6.5: Cross section of Figure 6.6: Top view of membrane (a)
membrane (a) (10 OOOX) (25000x)

179
Membranes (a) and (b) in Figures 6.5 to 6.11, had two different surface structures and in
both cases there were clear defects. The cross section view of membrane (a) (Figure 6.5)
indicates poor adhesion of the metal film to the alumina support. The Pd film is the thin layer
on top of the alumina base. The dense layer higher up is part of the resin. The surface is not
smooth (Figure 6.6), but consists of small metal clusters scattered over the surface area.
Under high magnification (Figure 6.7), there are tiny pores visible in the metal particles. It is
those pores or defects that caused poor selectivity.

Figure 6.7: Top view of membrane (a) (50 OOOx)

Figure 6.8: Cross section of membrane Figure 6.9: Top view of membrane (b)
(b) (10 OOOx) (25000x)

180
Figure 6.10: Top view of membrane (b) Figure 6.11: Top view of membrane (b)
(SOOOx) (2S000x)

The surface of membrane (b) seems dense both on the cross section view (Figure 6.8) and on
the top view (Figure 6.9) images. There were no continuous defects in the structure. Upon
further investigation some areas in the coating were identified where the coating was clearly
porous (Figure 6.10). These defected areas were spread out over the surface. Furthermore,
the defects did not seem to be inside metal clusters as in the case of membrane (a), but
between metal particles (Figure 6.11).

6.3.3. Pd MEMBRANE THICKNESS MEASUREMENT

Membranes were tested as discussed in Chapter 3. The thicknesses of the deposited layers
were calculated from the mass increase (mass after testing minus mass before electro less
plating). The mass before plating was taken as the mass of the clean membrane plus the mass
after pretreatment, divided by two. By doing that, it was assumed that half of the metal
deposited during the pretreatment stage penetrated into the membrane's pores and the other
half deposited on the outer surface. The samples that were used for testing as well as all their
characteristics are listed in Appendix Dl. Two sets of experiments were performed: Those
in which a positive feed pressure was used and those in which a sweep gas was used.

6.3.4. MEMBRANE POST PLATING TREATMENT

After plating, the membranes were stirred in ammonia to dissolve EDTA in the membrane
pores. Stirring in ammonia and heating overnight at 240°C was not sufficient to remove all

181
carbon frQm the membrane pores. BrQwn spots were visible in some areas on the outside
membrane surface, indicating the presence Qf carbon. Two. PQs~ibilities exist: either 240 DC
was too Iowa temperature for full QxidatiQn to. take place Qr the Qxygen to carbQn CQntact in
the PQres behind the dense palladium layer was very PQQr.

Hydrogen and nitrQgen permeance tests were perfQrmed Qn a membrane withQut and with
additiQnal QxidatiQn treatment. The results are shQwn in Figures 6.12 and 6.13.

1.4E.Q5

I
H
• Without oxidation
1.2E.Q5
I
------
• Witt! OXidation

~ 1.0E.Q5

~
1 1..-
I!!II ..... 8.OE.(]6

~
a.N·
51 ~ 6.0E-06 ./
es
J.
:z:
4.0E·06

2.0E-06
-------
D,OE+OO
330 350 370 390 410 430 460
Reactor temperature (OC)

Figure 6.12: Effect Qf QxidatiQn PQst treatment Qn H2 permeance

(membrane 2a, 1.43 fUll Pd)

1200

• Without oxidation
1000
_With oodation
~--------------~

.~
II
800

i N
600
Z
S '"'0
£
200

0
330 350 370 390 410 430 450
Reactor temperature ("C)

Figure 6.13: Effect Qf QxidatiQn PQst treatment Qn selectivity

(membrane 2b, 1.43 fUll Pd)

182
The characteristics of membranes (2a) and (2b) are listed in Appendix Dl. There was a
significant increase in hydrogen permeance after oxidation at 320°C. These results
confirmed the presence of either an EDTA or carbon layer in the pores behind the Pd film.
Reduction in hydrogen at up to 500°C did not thermally decompose the layer. When oxygen
was forced through the defects in the Pd film and the membrane support pores under pressure
(between 1.0 and 2.5 bar, depending on the membrane selectivity) at 320°C, most of the
remaining precursor was removed. That resulted in higher hydrogen permeances through the
Pd membrane and improved selectivity (see Figures 6.12 and 6.13).

In certain cases (membrane N7 and N8, see Appendix Dl), even after oxidation there was
some carbon present in the fihn. A good indicator of the presence of carbon in the film was
the rate at which steady state was obtained after the membrane was reduced and switching
from nitrogen to hydrogen during analysis. If steady state was obtained quickly (in less than
3 to 5 minutes) it indicated a pure fihn. When carbon was present in the film it could take 10
to 15 minutes (or even longer) for the hydrogen flux to stabilise (especially at the lower
hydrogen feed pressures).

When carbon was present in the fihn it could be expected that the film would show poorer
stability over time at high temperatures compared to pure films. This assumption was,
however, not further investigated.

6.4. THE EFFECT OF SUPPORT STRUCTURE ON Pd FILMS

Several Pd plating experiments were performed on SCT membranes with a finalS nm pore
size layer of y-alumina, to produce Pd films of less than 1 micron. The plating steps were
successful and thin fihns could be synthesised. Upon heating, even at the drying stage
(240°C), bubbles and cracks formed in the film. All selectivity was lost. The loss of
selectivity was a combination of poor metal to ceramic adhesion and a difference between
thermal expansion coefficients for Pd and y-alumina.

183
For those experiments, the higher hydrazine concentration was used (a hydrazine:Pd molar
ratio of 1: 1). Electroless plating with a lower hydrazine concentrations was not tested on the
y-a1umina membranes. A lower hydrazine concentration (a hydrazine:Pd molar ratio =

0.3 5: 1) should yield better results.

6.5. STRUCTURAL CHARACTERISATION OF Pd MEMBRANES

The structure of Pd membranes was investigated with a scanning electron microscope.
Membrane (3b) and membrane (11) were studied. Their theoretical thicknesses were:
• 3b: two layers totalling 4.4 microns, and
• 11: 1 layer of 1.5 microns.

Data for all tested (permeances) Pd membranes are listed in Appendix Dl. Membrane (11)
broke in the reactor before permeance testing was performed and there is therefor no
permeance data in Appendix Dl for membrane (11).

Figure 6.14: Cross section of membrane (3b) (10 OOOX)

The side view image of (3b) (Figure 6.14) clearly shows two Pd layers on the alumina
support. Since no alloying was performed after the second coating, a single layer did not
form at that stage. After application of the fIrst layer (membrane 3a), the membrane was

184
tested. A second layer was applied and the membrane (membrane 3b) tested again. The
combined thickness of the two layers should be about 4.4 microns. Figure 6.14 shows,
however, a total thickness of closer to 7 microns. This was the only membrane that showed
deviation between the calculated Pd thickness and the SEM determined Pd thickness. For
other membranes tested by SEM (a, b, c, d and 11), the results of the calculated thicknesses
and the SEM determined thicknesses, were a good agreement. Top view images of (3b)
(Figures 6.15 and 6.16) show a dense structure without any pores or defects. Selectivity data
on (3b) confirmed a dense and compact film with very little defects (see Appendix Dl).

Figure 6.15: Top view of membrane (3b) Figure 6.16: Top view of membrane (3b)
(5000 x) (2500Ox)

Figure 6.17: Cross section of membrane (11) (10 OOOX)

185
The calculated thickness of membrane (11) (1.45 microns) was in good agreement with the
SEM determined thickness (from 1.0 and 1.7 microns). Figure 6.17 shows a very dense
layer on top of the 200 urn a-ruumina support. Top view images (Figures 6.18 and 6.19)
confirmed this. Under high magnification (Figure 6.19) it appears as if there are more grain
boundaries for this thinner film compared to the thicker film of (3b) (Figure 6.16).

Figure 6.18: Top view of membrane (II) Figure 6.19: Top view of membrane (II)
(5000 x) (2500Ox)

6.6. PERMEANCE TESTING OF Pd MEMBRANES

Membranes were tested under positive feed pressure conditions and with a sweep gas. Under
positive feed pressure conditions, the effects of temperature, pressure difference and film
thickness were studied. When a sweep gas was employed, the effects of temperature, space
time and sweep gas ratios were tested. All the data is listed in Appendix Dl.

The hydrogen permeance process was described mathematically in Chapter 2. The main
equations were:

(2.7)

(2.8)

186
Pressure data is necessary to calculate the value of n, which is an indication of the flow
process through the film. Temperature data is necessary to calculate the Arrhenius
parameters (Po and ED) in equation (2.7).

The parameters of each tested membrane were calculated and are listed in Appendix Dl.
The values will be discussed in the followings sections.

6.6.1. THE EFFECT OF AP ON Hz AND N z PERMEANCE

Membrane selectivity was determined using hydrogen and nitrogen as testing gases.
Nitrogen permeance is an indication of membrane defects or leaking. There are three factors
that contribute towards the measured nitrogen permeance. They are:
• Leaking through defects in the electroless plated film,

• Leaking at the membrane reactor, graphi~e ring, enamel interfaces, and

• Leaking at the porous membrane, non-porous enamel and Pd film interfaces.

The contribution of the final two factors cannot be quantified, but from experience it is
known that there is at least some leakage between the membrane and the reactor seal. The
measured nitrogen permeance represents the worst case scenario or the maximum value.

6.6.1.1. Nitrogen experiments

Figures 6.20 to 6.23 show nitrogen permeances as a function of pressure and temperature at
different Pd thicknesses (1.47 !lm to 4.43 !lm). The average pressure between the tube and
shell side was used, it being the sum of the absolute pressures on the shell and tube sides
divided by two. Theoretically, the average pressure should not have any effect on the
nitrogen permeance (in mol/m2.Pa.S) if it is Knudsen flow. For nitrogen, the permeance is
proportional to the amount of defects. The more the defects, the higher the rate. Thinner
films thus have a higher nitrogen permeance than thicker films. The effect of film thickness
on membrane performance will be discussed in more detail later (section 6.6.3). Figures
6.20 and 6.23 indicate that the nitrogen permeance varied little with an increase in average
pressure, confirming that the flow through the membrane was Knudsen flow.

187
 '16.0 •••.••••.•.••••••••••.•.••••.• ---- .• -------------.-----.-------------------.-------.----- '.0

\
•.•--------,~-
~~ot-------;-------,~ . . 7.0 t-----------------------------;;---
l ~ot_----------------~--------~ 1iii
8.0
i
t---:---le-----~-----------=-'
~ ~
__
l! ~0t_------------r============]­
2!.ot_---------------j .T .. 450"C _T=410"C f--
i~ 4··00t======::J~~~~~~~
. f--- .T=4&O"C .T=410"C

1~O~============~.T~.~m;~~~';T.~~~~~~t=
8
I
+- 15 0
:
t-----------------------------
10..0
I
i ,o~----========-
I---
3.0 t_-------------/ • T = 370"C xT. 330·C

Z 5.0 t----------------------------- .z: 1.0 t-----------------------------

0.0 +-------__--__----__----__----_ 0.0 +-----__--____--__--__----______
1000 1100 1200 1300 1<'" 1500 1000 1200 1«10 1t!OO 1800
AVllrage pr... u... (mbar) AVllrq. pr... u ... (mbar)

Figure 6.20: Effect of pressure on N z Figure 6.21: Effect of pressure on N z

permeance for a 1.47 micron Pd film (6) permeance for a 2.4 micron Pd film (3a)

.ItO
, 1.' ,
"i!
~
35.0
, ~

•
,
.
• :
'iii
:. 1.2
1.4
,
.
,
.
30.0
] • • • •
o
E
25.0 0
E
1.0
c c
- ~o ;- 0.8 .T=450"C .T=.10"C -
I 150 .. T=450"C .T=410·C ---;
~
m 0.' .T"370"C ><T"'330"C -
i 100 .T=370"C xT=330"C
---j ~ 0
1 .•

Z '.0 Z 0.2
0.0 1150 1200
001100 1500 1700 11000 ,,,.
1060 1100
AVllrage pressu... (mbar)
"" 13'"
'''' Average pressure (mbar)

Figure 6.22: Effect of pressure on N z Figure 6.23: Effect of pressure on N z

permeance for a 3.08 micron Pd film (la) permeance for a 4.43 micron Pd film (3b)

6.6.1.2. Hydrogen experiments

The value of n in the flux equation (2.8) was calculated for every membrane and results are
listed in Appendix Dl. The RZ-value is an indication of the fit between the measured and
calculated data. A value close to 1 indicates a very good fit. The value of n was assumed to
be 1 and then the calculated values were compared to the measured values. The R2-values
were also found to be close to 1, indicating that the calculated and measured values were
indeed similar. A n-value of 1 indicates a permeance which stays constant at different
differential pressures. It must, however, be mentioned that the differential pressures at which
the hydrogen permeances were determined were small (typically less than 150 mbar). When

188
the differential pressures are small, changes in the value of n have little effect on the quality
ofthe fit.

'4.0 -------------------------------,--------------------------------------------.-.----.-.---~ 10.0 -----_._----_._----_._----------_._._--------

!
! , . o + - - - - . , . - - - - - -_ _ _ _~
'iii' 12.0+--------.--------------i
:: ~ 8.0t--------~--~-~
"E 10.0 +-------"'0.- - - - - - - - - - - - - i "E 7.0 t------;--~--~--.._------i
!'il ..o+-------""---~---~,_i 1.ot--~--~--~--.._-__i
- 5.0 +------'."-------'-----'-,_ _~
~
I
..
8.0

4.0
.T=4&lOC

.T"370"C
,T:410"<:

"'T=330"C
~
~! II .0
3.0
t+-::::::::::::::::::::::::::::::::::::-_::::--Ir======;-"1
2.0 + - - - - - - - - - - j • T= 370"C
.T:450"C H .T~410"C

xT. 33O"C ~
i: 2 . 0 t - - - - - - - - - - - - - - j ;
i: 1.0j-------"======"'---i
a.o+----_--_--_-_---.--J 0.0+----_-_--_-_-_--1
o 2Q 40 60 8tI 100 o 20 40 8Q 80 100
Dll'ferentlal pressure (mbar) DIra,ential prliSsurelmbat)

Figure 6.24: Effect of pressure on Hz Figure 6.25: Effect of pressure on Hz

permeance for a 1.47 micron Pd film (6) permeance for a 2.4 micron Pd film (3a)

9.0 _._------------------------------------------------------------.-------,-- '.0 !

!,
_ '.0 t---.------,-------:----_I
•
:. '.at------------;--~-_I
OJ '.0
:. · · !,

1
"E ! '.0 • ,
·• !

a
6.0
5.0t------------::........_I
0
E
at •
·• • ,i
I .0 t---"-------'~-----"--~-_I c
I '.0
3.0 •

~ 3.aj-------;=====:::;: ~ .r=450"C .r-"10"C H

!. 2.04-------1 .T=«!iO*C .T=410'C !. .T=370°C xT=330"C
'£ 1.0 H
£ 1.0t-----------i .T=370"C xT=330"C
a.aL-_-_--r'====::::;::=:::J120
o 20 40 80
Dlfh,rential pressure Imbar)
80 100
0.0
0 . so
DIf'f..-.ntJaI pr.ssur. {mbarJ
120 OJ

Figure 6.26: Effect of pressure on Hz Figure 6.27: Effect of pressure on H2

permeance for a 3.08 micron Pd film (la) permeance for a 4.43 micron Pd film (3b)

The hydrogen permeance (Figure 6.24 to 6.27) did not vary considerably with pressure.
Differential pressures below 40 mbar caused a slight deviation (see Figures 6.25 and 6.27)
from the trend. The reason for this was that the error of the pressure probe was between 2
and 4 mbar and thus, at low pressures, there was some error in the measured values. A n-
value of 1 indicates that hydrogen chemisorption on the palladium surface is the rate limiting
step (Nam et aI., 1999; Yan et aI., 1994). Sievert's law, where n = \1" is not applicable to the
thin films synthesised in this study. Diffusion is not the rate limiting step.

189
The film thickness and the permeance temperature have a very significant effect on the
hydrogen permeance as will be discussed in the next sections.

6.6.2. THE EFFECT OF TEMPERATURE ON H2 AND N2 PERMEANCE

If the flow through a membrane is Knudsen flow, the flux through the membrane must
decline when the temperature increases (see equation 2.2).

J = GfS, AP; (2.2)

• ~2nMiRoT I

In section 6.6.1.1 pressure data for nitrogen permeance suggested that Knudsen flow might
have been the mechanism of nitrogen transport through the defects in the Pd film. This
indicated that the defects were in the lower nanometer range. For each of the fifteen Pd films
tested, the nitrogen permeance was plotted as a function of temperature (see Appendix Dl).
For ten of the films, the nitrogen permeance declined with an increase in temperature (see
Figure 6.28 for a typical example). In four cases there were no clear permeance trend with
temperature change (Figure 6.29) and in one case the permeance increased with increasing
temperature (Figure 6.30).

. "" ""

..-
..t------------I
.,• . ,. ,
__ NIIrogeo pennance

.... t
;;-
N

~
E

,
. ." HycftlgllnJNItrog.n

",,"
"
"'"
ft'

~I
2.8

ut--~~-~------I
". """"

-
6600

I
.I
0 1.3 U
~,"'" /" -il
... f
E
~
sc "
• U ',. ~.
E
18
'
:i'-.. ".
".
- .--
/
.,/ ;j

I ... "
", ~-
--~-'"

,..
.x. HI

i!
t=====1'~"~"~~~~-~":"':-::::'tj ....
U

1/'" z 1.4
u
I,'
." ". T('C)
.., .'"""" ...
~---_--_----4~
." T(OC) '" ...
Figure 6.28: Selectivity data for Figure 6.29: Selectivity data for
membrane (3a) membrane (N8)

190
 22 '00
....• .•...
~
-E
21
/,>- 730

~ 20
E ....... ..... .' 710 ~
S
'

..• - •
8
I•
"
"
- .'
.. /

'" I
i 870
z I --:---:~:=-~_I
17

"330 370
TC'C)
.10 ...'"
Figure 6.30: Selectivity data for membrane (N2)

The temperature data confirmed that some kind of Knudsen flow dominated when nitrogen
passed through the palladium film defects. The reason for the decline in permeance was that
the greater vibrational energy of the N2 molecules at the higher temperature resulted in more
resistance to flow through tiny pores and thus a decrease in permeance.

Hydrogen temperature data was fitted to the Arrhenius equation (2.7). Arrhenius parameters
2
for each film are listed in Appendix Dl. The high R -values of the Arrhenius fits indicate
that the data fitted the equation well. The hydrogen permeance increased with temperature,
as predicted by equation (2.7). Figures 6.31 and 6.32 show typical increases in hydrogen
permeance with an increase in temperature.

8
1
'8 6
i 6
~. ,
4

...
~
...01>
-<>
~ . ..t>

':::- ol>
~.gP

Figure 6.31: Hydrogen permeance Figure 6.32: Hydrogen permeance

in llffiol/m2.Pa.s (membrane 2a) in llffiol/m2 .Pa.s (membrane N7)

191
6.6.3. THE EFFECT OF Fll..M 1IDCKNESS ON PERMEANCE
The effect of film thickness on hydrogen permeance, nitrogen permeance, membrane
selectivity and Arrhenius parameters are depicted in Figures 6.33 to 6.40. The hydrogen
permeance should be inversely proportional to the Pd film thickness (equation 2.9).

P =p.. (2.9)
m I

6.6.3.1. Hydrogen permeance

Data in Figure 6.33 shows a decrease in hydrogen permeance with an increase in film
thickness up to a thickness of about 4.5 microns. Thereafter, the permeance started to level
off. The permeailce increased with temperature, as mentioned previously (see 6.6.2). The
decrease in permeance was not directly proportional to the inverse thickness. For example;
taking values from the quadratic fit at 450°C at thicknesses of 1.0 and 3.0 microns, yielded
permeances of about 7.0 and 13.0 !llll0I/m2 Pa.s, respectively. This ratio (almost 1:2) is
different from the theoretical prediction (1:3). The reason for this is that the model equations
(2.8 and 2.9) were formulated for thick foils in the tens and hundreds of microns range.
When films become very thin, surface structure and morphology effects come into play,
which cause deviation from the model equations. Figure 6.34 depicts permeances of Pd
films from 1.0 to 1.5 micron thickness. The values ranged between 15 and 6 !llll0I/m2 .Pa.s
for temperatures from 330°C to 450 °C.

16.0

";! 14.0 t--7------I

..
•
---" - ---" ----.------......... ------.------.-------------.----.------.------.------,

.T=450°C .T::410·C

&: _T=370·C *T::33Q°C

N "O~~~----~============~
!~ 10.0 t--,:;J~"",.___---------____i

~
~ 8.0 t-~~~~""""~-------____i

X.~ 6.0 t---=.:.=...-~""'=''''=:-...::".~-----____i

•
-£ 4.0t======:~::~!~~~;:J
2.0
0.50 1.50 2.50 3.&1 4.50 5.50 6.50
Pel ftlm thickness (microns)

Figure 6.33: Hydrogen permeance for Pd films from 1 to 6.5 micron thickness

192
 18.0 ------------.-------------------------------------------------.-.--. ___ •______________________________ _
• •
ii' 14.0 t-----""'--S.=:i.[--;----~--------___1

~
f'I' 12.0 t-~fL--~-~-~·~~~~~-~~~"'-~~--._-___1
~_§. ~=:"--:.:.=. ~~'~i• '~'='.:..~.~~.-~..;-~-:~:-:~~-~-::.~~~.
10.0 ='
-.... -- ~
.--
8.0 • -.

~N •.0
:
•E .T=450~C .T=410°C
•
1-_______---1
"~ '.0
N
eT=370°C :r:T=330"C
:t 2.0

0.0
0.95 1.05 1.15 1.25 1.35 1.46
Pd film thickness (microns)

Figure 6.34: Hydrogen permeance for Pd films from I to 1.5 micron thickness

6.6.3.2. Nitrogen permeance

Nitrogen permeance generally decreases with an increase in Pd film thickness. Nitrogen
permeance as a function of film thickness is given in Figures 6.35 and 6.36.

~
100
.:. . I
·T=450.C
·T=370·C

1E
o I'
1 E
o I'
c ~
! 10t-------~~----~ S 10t-------------~
••

I
£
:
i
£

0.50 1.50 2.50 3.50 4.50 5.50 6.50 0.95 1.05 1.15 1.25 1.35 1.45
Pel tim thlc:kness tmlc:rons) Pel film thick,.... (microns)

Figure 6.35: Nitrogen permeance for Pd Figure 6.36: Nitrogen permeance for Pd
films from I to 6.5 micron thickness films from I to 1. 5 micron thickness

For the first membranes that were prepared (membranes Ia, Ib and Ic), the plating process
was not well refined and the selectivity was poor as shown in Figure 6.35 at thicknesses of
3.1, 3.9 and 6.2 microns. If those values were excluded then the decline in nitrogen
permeance with an increase in film thickness would be clear. The worst membrane (1.0 !J1Il)
had a nitrogen leak rate of 70 nmoVm2 .Pa.s, while the best ones had values close to

193
1 nmollm2Pa.s. Figure 6.36 indicates that the majority of the Pd films with thicknesses
ranging from 1.0 to 1.5 fUll had a nitrogen leak rate of between 20 and 2 nmol/m2 .Pa.s.

6.6.3.3. Membrane selectivity

The selectivity data, which is the hydrogen permeance divided by the nitrogen permeance, is
presented in Figures 6.37 and 6.38. At all the temperatures tested the selectivity remained
above 100, which is an indication of very good membranes. All films of thickness ranging
from 1.0 to 1.5 fUll had a selectivity of at least 400, except the 1.0 fUll film and a 1.47 fUll film
(see Figure 6.38).

10000 10000

"'.

j
~ 1000 J-------'-~-_:_---------
;: ~t :
.,
'T:450"C ·T"'410"C I
I ·T=3roOC .T;330"C I
100 +--_-_---.::.........--~-_--
0,50 1.50 2,50 3.50 4.50 5.50 6.50 1.05 1.15 1.25 1.35 1.45
Pd film thic:knen (microns) Pd film thickness (microns)

Figure 6.37: H2 to N2 selectivity for Pd Figure 6.38: H2 to N2 selectivity for Pd

films from 1 to 6.5 micron thickness films from 1 to 1.5 micron thickness

6.6.3.4. Arrhenius parameters for hydrogen permeance

There was significant variance in the Arrhenius parameters for hydrogen permeance.
Figures 6.39 and 6.40 were constructed to illustrate this. The permeance (Pm = P..II) and not
the permeability (P... = Pm*1), was plotted on the right-hand side axis, to be able to compare
values for different Pd thicknesses. A decrease in Poll and/or an increase in ED indicate
slower permeance. Both Figures 6.39 and 6.40 show an increase in activation energy and a
very slight decline in Poll as expected. The permeance decreased with increasing film
thickness.

194
 """'" . 2.0£-<>< 20000 2.~

",m
• .--- 2.0E-04 10100
W·Loft ..·.1
• Rightaxi.
•
2.0E-04

•
. ~ .•
;;-
•
. ~
I·loft ... ,l
i"12000 1.5E-G4 i"12000
. .. ..
1.5&04
'E
.8_
~ ----
'$'
. . . • R1d1tultl
1.0E-04
'6
oS
~
.8_ · ~
!
.....
. ~ . . .. ·
1.0E-04
~
4000
. 5.0E-05 4000
· 5.0E-05

0 O.OE+OO 0 O.OE.OO
0.50 1.50 2.50 3.50 4.00 '.50 ...0 0." 1.05 1.16 1." 1." 1."
Pd film thickness (microns) Pd film thkltnass (merons)

Figure 6.39: Arrhenius parameters for H2 Figure 6.40: Arrhenius parameters for H2
permeance (1 to 6.5 micron Pd films) permeance (1 to 1.5 micron Pd films)

6.6.4. COMPARISON WITH LITERATURE DATA

During the course of this project an extensive database on hydrogen permeance data through
palladium films was accumulated. This data was divided into different categories, for
example:
• Pure Pd or Pd alloys,
• Substrate type (alumina, porous glass, porous stainless steel or refractory metal),
• Deposition method (electroless plating, CVD, wet impregnation, electroplating etc.), and
• Deposition position (inside tube, outside tube or on flat disk).

The following general remarks summarised in Appendix E can be made regarding hydrogen
permeance through Pd films:
1. Unsupported Pd films or foils are thick (typically 24 J.UIl and thicker) and hydrogen
permeance poor. The best permeance value from literature was 1.2 J.UIlollm2 .Pa.s at 350
°C (Hurlbert and Konecny, 1961), but usually the values were lower than I J.UIlollm2.Pa.s
even at much higher temperatures.
2
2. Pd-Ta-Pd (15 J.UIl) foils had permeance values of up to 1.76 J.UIlollm .Pa.s at 340°C
(peachey et aI., 1996).
3. CVD and wet impregnation were successfully used to produce Pd films ofless than I J.UIl.
4. The surface structure of Pd films prepared by CVD appeared to be unsuitable for high
hydrogen permeance. Typically, films had a hydrogen permeance of less than

195
 2 .
1 flmol/m .Pa.s, except for those prepared by Yan et al. (1994) which had values of up to
4 flmol/m 2.Pa.s.
5. The best values for hydrogen permeance were obtained with alumina substrates. Most
hydrogen permeances of films on porous glass were less than I flmol/m 2 .Pa.s. For films
on porous stainless steel permeances were less than 1 flmol/m 2.Pa.s and for films on
refractory metals permeances were less than 2 flffiol/m 2.Pa.s.
6. The best values obtained with alumina supports and electroless plating were:
2
• 9.75 flffiol/m .Pa.s at 450°C for plating on a disc by Zhao et al. (1998), but the H2 to
N2 selectivity of the film was only 23.
2
• 2.86 flmol/m .Pa.s at 400°C for plating on the outside of a tube by Kikuchi (1995).
The selectivity was not mentioned.
• 5.27 flffiol/m 2.Pa.s at 500°C for plating on the inside of a tube by Shu et al. (1997b).
The selectivity was not mentioned.

The best overall hydrogen permeance values that have been published were for Pd films on
modified porous stainless steel discs, prepared by wet impregnation. Film thickness varied
from 0.5 to 0.8 flm. The hydrogen permeance was 15.8 flmol/m 2 Pa.s at 450°C (Iun and Lee,
1999) and 17.8 flffio!/m 2.Pa.s at 550°C (Nam et aI., 1999). In both cases the H2 to N2
selectivity was above 1000.

6.6.4.1. Hydrogen permeances in the present study

The geometry of membrane tubes is preferable to that of membrane discs for practical
installation. Whether the selective layers and the Pd film should be on the inside of the tube
or the outside of the tube is debatable. If the Pd film is on the inside of the tube, it is more
protected against scratching and damage during installation. From past experience it was
found to be easier to coat tubes on the inside than on the outside, because of greater solution
stability. Pd films on the outside of tubes have the advantage of a larger surface area per
membrane length, compared to coatings on the inside. ECN produces membrane with
selective layers on the outside, while SCT produces membranes with selective layers on the
inside of the tube.

196
During the present study some important advances have been made in the preparation of Pd
membranes. Pd film thickness on the inside of alumina tubes has been reduced to between
1.0 and 1.5 /lm, while maintaining H2 to N2 selectivities exceeding 400 for the majority of the
films. An important cost advantage is that the cheaper 200 nm a-alumina support was used
successfully. The more expensive y-a-alumina support was not necessary. The thinnest Pd
films prepared previously by electro less plating on the inside of membrane tubes were:
• a 2.0 /lm film on the inside of an assymetric aly-alumina membrane with a 5 nm pore
size (from SeT) by Shu et al. (1996b), and
• a 2.1 /lm film on the inside of an assymetric aly-alumina membrane with a 3-4 nm pore
size (from SeT) by Shu et a!. (1997b).

The highest permeance obtained for a Pd film on the inside of a tube (see section 6.6.4),
excluding the results from Zhoa et al. (1998), due to poor selectivity, and results from Shu et
al. (1997b) and others, where selectivity was not mentioned, is 2.68 /lmolJm2 .Pa.s at 467 °e
(Li A et a!., 1999). In the present study, hydrogen permeances ofPd films from 1.0 to 1.5 /lm
varied between about 8 and 15 /lmol/m2 .Pa.s for temperatures from 330 °e to 450 °e and
(Figures 6.33 and 6.34). These values are a significant improvement over other published
results.

2
Only the results of Jun and Lee (1999): 15.8 JlIDol/m .Pa.s at 450 °e, and Nam et al. (1999):
17.8 JlIDol/m2.Pa.s at 550 °e, are comparable to values in this study, but they used the
unfavourable disc membranes.

6.6.5. H2 PERMEANeE EMPLOYING A SWEEP GAS

Hydrogen permeance through palladium films were also tested usmg an atmospheric
hydrogen feed on the tube side of the membrane and an inert sweep gas (nitrogen) on the
shell side of the membrane. Equation (2.8) cannot be used in that format. The membrane
was treated as a plug flow reactor without chemical reaction, but with permeance, to derive
the model equation (6.1). Equations were derived froni the basic plug flow reactor equations
(Fogler, 1992).

197
 (6.1)

where L is the axial position along the membrane. The axial position is made dimensionless
by dividing both sides of (6.1) by the reactor length, Lo. Substituting parameters yields:

(6.2)

For hydrogen, the flow was from the tube side (high H2 pressure) to the sweep side (low H2
pressure), while for nitrogen the flow was from the sweep side (high N2 pressure) to the tube
side (low N2 pressure). Partial pressures was expressed in terms of flow rates. Substituting
molar flow rates in (6.2) yielded two coupled differential equations:

(6.3)

(6.4)

F H(O) is the hydrogen molar feed rate on the tube side and FN(o) is the nitrogen molar feed rate
on the shell side. The atmospheric pressure was taken as 100 000 Pa. The boundary
conditions for (6.3) and (6.4) were:

At (LlLo) = 0, FH = FH(O) and FN = FN(O) (6.5)

198
Equations (6.3) and (6.4) were solved using average hydrogen and nitrogen permeances
(found in Appendix 01) and employing Euler's method (Fogler, 1992) for first order
differential equations. A very small step length of 0.001 was chosen.

The amounts of hydrogen that permeated through the Pd films at different sweep gas ratios,
hydrogen space times and different temperatures are given in graphical format in Appendix
Dl. Hydrogen flow profiles along the axis of the membrane were calculated for membranes
(N4) and (N7), to compare the experimental data with the calculated data. To solve equations
(6.3) and (6.4) the following assumptions were made:
• The reactor was isothermal, and
• Hydrogen and nitrogen permeances were independent of pressure, even at low pressures.

Table 6.2 shows that the error between the experimental values and calculated values was
typically less than 1.5%, indicating very high accuracy. It was concluded that the model
assumptions were valid and that the model predicted hydrogen permeance very well.

Table 6.2: Experimental H2 permeances vs. calculated values for membrane (N7)
Space time = 2.37 seconds (200 cm'/min Hz feed)
Temperature eC)
Sweep gas mol 450°C 410°C 370 °C 330°C
% of Hz feed % Hz permeated
5 96.8 96.7 96.3 96.0
(98.2) (98.0) (97.6) (97.0)
10 97.7 97.5 97.2 96.9
(99.0) (98.9) (98.7) (98.4)
20 98.5 98.2 98.1 97.8
(99.5) (99.4) (99.3) (99.1)

199
 Space time - 1.19 seconds (400 cm'/min Hz feed)
Temperature ("C)
Sweep gas mol 450°C 410°C 370°C 330 °C
% of Hz feed % Hz permeated
5 96.7 95.9 93.3 87.9
(98.1) (97.6) (93.9) (85.7)
10 98.3 98.1 97.9 97.6
(99.0) (98.9) (98.7) (98.4)
20 98.9 98.7 98.6 98.6
(99.5) (99.4) (99.3) (99.1)
Values III brackets are the model values or calculated values

Table 6.3 tabulates a similar set of data for membrane (N4). Only at the very low sweep gas
flow rates (2.5% of H2 feed) did the experimental and calculated values deviate by more than
2%.

Table 6.3: Experimental H2 permeances vs. calculated values for membrane (N4)
Space time = 2.37 seconds (200 cm'/min H2 feed)
Temperature ("C)
Sweep gas mol 450°C 410°C 370 °C 330°C
% ofH2 feed % H2 permeated
5 97.8 97.5 97.7 97.4
(97.0) (96.8) (96.6) (95.7)
10 98.6 98.4 98.6 98.3
(98.4) (98.3) (98.2) (97.7)
20 99.2 99.0 99.1 99.0
(99.1) (99.1) (99.0) (98.7)

200
 Space time -1.19 seconds (400 cm'/min Hl feed)
.

Temperature (0C)
Sweep gas mol 450°C 410°C 370°C 330°C
% ofH1 feed % Hl permeated
2.5 94.7 92.8 91.4 87.2
(92.0) (90.3) (87.4) (81.2)
5 98.0 97.7 97.6 96.8
(97.0) (96.8) (96.6) (95.6)
10 98.9 98.8 98.8 98.6
(98.4) (98.3) (98.2) (97.7)
Values In brackets are the model values or calculated values

Figures 6.41 and 6.42 show the effects of temperature and sweep gas flow rates on hydrogen
permeance through the Pd film of membrane (N7). Both an increase in temperature and
sweep gas flow caused an increase in hydrogen permeance. The increase was more sensitive
to the sweep gas flow rate (see Figure 6.42).

99.0 100.0
i
98.• t-------------""'"
~; ------:::: 98.0
-:?</
... . ~ .'
\

1i 97.' +-----/"""""?/""-:,.~~,.....,
/-<~.~~·:;. >·t 1
98 0 ".0
I ;
.
...~:~. ~.~.~'"'-=.-,..-.".~; ~
".0
, ,I
8.
-£ 97.0 +----,r-:7..'.p.. "-
.. 7,""','-o.rr"'-----------j :£
92.0
,
~T·"·~H
I--T=«iO"C
96.5i----::-..'/,-.L--;::::======:-i ~
<I-
".0 +-
__----.o'<-'_·_--i1 - -r :4&O"Cr;;410-C:
__

I" ..· r-370"e ..... ·T.330"C i

Hi 90.0

8M
! I.".· T=370"C ..... ·T=330"C

.... +---_----'=r====;==='----' 88.0

5 10 15
Nt sweep rate as R1CJIar %of H,; feed
20 0
• 10
Nt sweep rata as molar %of H,; fead
15 20

Figure 6.41: % H2 permeated with N2 Figure 6.42: % H2 permeated with N2

sweep gas and space time = 2.37 s sweep gas and space time = 1.19 s

The effects of space time, temperature and sweep gas flow rate are clearer in Figures 6.43 to
6.46, where hydrogen flow along the membrane axis is plotted. By comparing Figures 6.43
and 6.44 it can be seen that increasing the sweep gas rate was more effective for removing

201
hydrogen than increasing the temperature was. For all experiments with a hydrogen feed rate
3
of 200 cm /min (space time = 2.37 s), equilibrium was achieved at some point along the
membrane axis. For a feed flow rate of 400 cm3/min (Figures 6.45 and 6.46), the permeance
rate at low temperatures and low sweep gas flow rates was too slow to allow all the hydrogen
too permeate through the Pd film. The large effect of sweep gas flow rate on hydrogen
permeance can be seen in Figure 6.46. Hydrogen permeance was incomplete at a sweep gas
flow rate of 20 cm3/min, but increased to full completion at four tenths of the membrane
length for a sweep gas flow rate of80 cm3/min.

200

"I) T_450'"C, 410-e. S70"C.nd330"C s..., 1M nII" • .to, 20 W1d 10

;; moving from .... to right cnfh'nln IIKnIing from Id to right
~ ,20
e ~
\\
8 eo
£'"
\

.. \,

"

0
0 0.2 0.4 0.8 0,8 o 0.2 0.4 0.8 0.8
DifTW!nslonllss .......ne J.ngth Dlrmnstonllss . .mbrane length

Figure 6.43: H2 flow in tube (cm3/min) as Figure 6.44: H2 flow in tube (cm3/min) as
predicted by model for N7 (T = 2.37 s) predicted by model for N7 (T = 2.3 7 s)

40J .---,'----.---------.---- .-.•.. ---.---.--.• ------,---------------_ •. -•. ----.•.• - 40J ------.,---.------.. -._-.-----•. ------.----.--------•. ----.. ---•.... ---.-----.-------- ----l

T-330°C

320
T_460"C,41D"C,370 -e.nd".,'C . . . . . . . _·80,40 ... 20

1 ,.. \
:5 I1'IO'oW18 from I.n: to rlghI: ;; cnI'/mkt ft!KMnO from IIIItD right
E ~
i 240
e
~
o ". ! ,0>
'"£ £'"
" '~
'-'.~.-".~.. ::~~
"
0 0
0 0.2 0.4 0,8 0.8 0 0.2 0.4 0,8 o.a
Dlnwnstonlus membrane length DI....n5kx11nli n.rnbrane a.ngth

Figure 6.45: H2 flow in tube (cm3/min) as Figure 6.46: H2 flow in tube (cm3/min) as
predicted by model for N7 (T = 1.19 s) predicted by model for N7 (T = 1.19 s)

202
Figures 6.47 to 6.52 are a repetition of Figures 6.41 to 6.46, but for membrane (N4).
Membrane (N4) had significantly higher hydrogen permeance values than (N7) (see
Appendix Dl).

From Figures 6.49 to 6.52 it can be seen how, when compared to data of (N7), all the curves
were shifted to the left. Complete hydrogen permeance for membrane (N4) was achieved
much quicker along the membrane axis i.e. at lower dimensionless reactor lengths, than for
membrane (N7).

100.0

.... ,.. ;,tr.

100.0

'
-.---- --
.... ~'l

J ,., 89.0
I .... Ii" I: I
,.
/"

i I 920
J .... !
%:
". 8B.O
%:
". ,
& "
I---T=450"C __ T"'410-C
~;' r 90.0

i17.S t-_'-----j
_T=460"C -+-T=.10-C
, I
1···T.moe .... ·T=330-C

rI / .•. T:::370"C __ -T=330"C sa.

~

g7.0+-~-_-~-~~-_-~_ ....
5 10 15 ~ ~
Hz: sweep rate as molar" f1l H,; feed
M ~ ~ 0
• 10
N,; swap me as molar % of H,; fHd " "
Figure 6.47: % Hz permeated with N z Figure 6.48: % Hz permeated with N z
sweep gas and space time = 2.37 s sweep gas and space time = 1.19 s

200 ,, ""
IwMP a- t.ecI_ 10 I;ftlllmln T_moe
'0]
!
, '0]
nI_."
:e
~ ".
1\ T-450"'C,41D"C, 3'ftJ '"CandlSO"C
mo'otng from ... to light
!,
,
:eE
11>' \\\
:!!weep u- 20 md 10
cm·..... 1NnIng from "'10 right

e
~ so
~ !
Ii
~ eo \\\
~
'"%:
%:
40
!
! . \\\
\~ ,~~
i
i \\\ ~ -- --- ----,. . _., ......
•• ...
_.,-,

, •• ...

,
'.2
Dlrmnstonless lM~n. length
••• ••• 0.2 0.'
DkNnslonless nw:....n. t.ngth
0.' 0.'

Figure 6.49: Hz flow in tube (cm3/min) as Figure 6.50: Hz flow in tube (cm3/min) as
predicted by model for N4 (I: = 2.37 s) predicted by model for N4 (I: = 2.37 s)

203
 ... ...
s..., .. fMcI_ 20 cm'hnln T.3SO"C
1\ l-
>20
\ T. 45D "C, 410 "Co '70 "C lOCI !SO "C "" \~ aw...,.,.. .... -10, 4O.,d 20
:5 :5
~ 240
mollWlU frvm , ... 10 right

i 2~ \\ cm'.mIn mD'lllno "'om "'Ia right

.!i.
~ 100 ~ ! S\~
\\\, ~
1GO
'"-£ '~ -£
80 eo
'~~ "
~
0
.... :-::.~
, 0
\ "'" " ,

0 0.2 0.' 0.' 0.8 1 0 0.' 0.' 0.' 0.' 1

DlRWlnsionless nwmtnne .....glh Dln.nslonless membrane length

Figure 6.51: H2 flow in tube (cm3/min) as Figure 6.52: H2 flow in tube (cm3/min) as
predicted by model for N4 (. = 1.19 s) predicted by model for N4 (. = 1.19 s)

6.7. SUMMARY
Each step in the e1ectroless plating process must be carefully optimised to produce thin,
highly selective, Pd films. For pretreatment, a low Sn to Pd ratio was employed to limit Sn
deposition and increase thermal stability of the film. The quality of the Pd film was very
dependent on the plating rate. A high plating rate, due to: a high hydrazine concentration, a
low EDTA concentration and/or a high plating temperature, must be avoided to produce
selective thin films. In some cases the Pd films had defects due to high plating rates. The
last critical step in the membrane production process was the post plating cleaning. After
plating, membranes were stirred in ammonia solution for several hours, dried and then further
oxidised at 320°C in pure oxygen before reduction.

Pd films, of thicknesses down to 1 fllll, were deposited on the inside of asymmetric SCT
a-alumina membranes (200 nm pore size). Hydrogen permeances ofPd films from 1.0 to
1.5 fllll, varied between about 8 and 15 fllllol/m2.Pa.s for temperatures from 330°C to
450 °c. Hydrogen to nitrogen selectivity was> 100 for all membranes tested and > 400 for
all but two membranes (thickness 1.0 to 1.5 fllll) tested. These values are a significant
improvement over other published results.

204
Hydrogen permeance fitted the flux equation well, with the permeance constant at different
differential pressures. This implied a n-value of I in the flux equation. Temperature data
fitted the Arrhenius equation with high accuracy. For the majority of the membranes,
nitrogen flow through defects in the films showed signs of Knudsen flow. Finally, hydrogen
permeance was tested using an inert sweep gas. Two coupled differential equations were
formulated for modelling the membrane as a plug flow reactor with permeance, but without
reaction. The equations provided very good fits for the experimental data.

205
 7. Pd-Ag MEMBRANE ALLOYING AND STABILITY
TESTING

Pd membranes can only be used above 300°C. Pd-Ag membranes can, however, be used at
lower temperatures since to the a to ~ phase transition in the Pd crystals is suppressed by the
silver addition. This a to ~ phase transition in the Pd crystals is what causes defects and loss
of selectivity in Pd films below 300°C.

In Chapter 7 the preparation and testing of Pd-Ag membranes will be described. They will
also be compared to pure Pd membranes. The membranes were heat treated at high
temperatures in an effort to form a homogeneous alloy between the Pd and Ag. The stability
of the membranes was monitored while the alloying process was in progress. The Pd-Ag
membranes were also tested at temperatures below 300°C.

7.1. Pd-Ag MEMBRANE PREPARATION

For supported films, the Pd and Ag deposition processes have an effect on alloying
conditions. Pd and Ag can either be deposited as separate successive layers or by a co-
deposition process. Different methods for preparing Pd-Ag membranes are listed in
Appendix E. Co-deposition processes include:
• Sputtering (dc or magnetron) on flat discs and sometimes on the outside of tubes,
• Spray pyrolysis on discs or on tlie outside of tubes, and
• Electroless plating on any surface.

The main advantage of sputtering is that complex alloys can be prepared. Metals can be
deposited in any ratio and with high accuracy. The disadvantage is that the technique is not
suitable for depositing metal on the inside of a membrane tube. Even depositing metal on the
outside of tubes may cause problems with thickness control and composition uniformity.

206
Only electroless plating can be used for co-depositing Pd and Ag on the inside of a membrane
tube, but the process control is poor. It is very difficult to obtain the desired Pd to Ag ratio on
the membrane surface (Shu et aI., 1993; Cheng and Yeung, 1999).

The plating kinetics for Pd and Ag in solution differs. Their stabilities also differ, with Pd
being more stable in solution than Ag. This resulted in very diluted concentrations having to
be used for co-deposition. The Pd to Ag ratio in the starting solution differs considerable
from the deposited composition on the membrane. Co-deposition is discussed in more detail
in Keuler (1997a), Cheng and Yeung (1999) and Shu et al. (1993).

In the present study, Pd-Ag membranes were prepared by successive Pd and Ag plating. The
Pd layer was deposited first, followed by the Ag layer. The plating procedure was discussed
earlier in section 3.3.2.4. The composition of the plating solution was given in Table 3.16
and the plating procedure given in Table 3.17.

Before the silver layer was deposited, the membrane,. covered with teflon tape on the outside,
was dipped twice in both the pretreatment solutions. Post plating cleaning was similar to that
after Pd deposition.

7.2. UNALLOYED Pd-Ag MEMBRANES

Firstly, unalloyed Pd-Ag membranes were compared to pure Pd membranes and then the
effects of alloying studied. Table 7.1 was taken from Appendix D2 and summarises the
alloying conditions.

In the table, when two temperatures and two times are given in a row, the first heating time is
for the first temperature and the second heating time is for the second temperature.

207
Table 7.1: Alloying procedures used for Pd-Ag membranes
Name Elements Alloying Heating Gas Temp ("C) Time (b)
system
8 Pd No
8b Pd+Ag No
8e Pd+Ag Yes Tube oven Ar 550,600 15,25

NI Pd No
NIb Pd+Ag No
NIx NewPd+Ag Yes Reactor H2 590 10

N3 Pd No
N3b Pd+Ag No
N3e Pd+Ag Yes Tube oven Ar 545 50
N3d Pd+Ag Yes Tube oven Ar 545 100 (total)
N3e Pd+Ag Yes Tube oven Ar 545 150 (total)

N4 Pd No
N4b Pd+Ag No
N4e Pd+Ag Yes Tube oven Ar 530 30
N4x NewPd+Ag Yes Reactor H2 540 30

N6 Pd+Ag Yes Tube oven Ar 550,600 15,25

N2x NewPd+Ag Yes Reactor Ar, N2 500,550 50,25

N8x NewPd+Ag Yes Reactor Ar 550 10

208
7.2.1. CHARACTERISING UNALLOYED Pd-Ag MEMBRANES
Four Pd-Ag membranes were tested prior to alloying. They were: 8b, NIb, N3b and N4b
(see Table 7.1). The metal layer, at that stage, consisted of a Pd film and a second Ag film.
The hydrogen permeances of the membranes are depicted in Figures 7.1 and 7.2. The
addition of silver resulted in a sharp decline in hydrogen permeance at all temperatures. This
confirmed that silver has a very poor hydrogen permeance and that separated Pd and Ag
layers in the film should be avoided.

15 ........................................................-.......................................-.... -j

t..
12 ~
__-- --------
_- .. "'" - .. ~i
N

~
E
.
__ ., .• - - -
-+-1.1 "mPd(8)
'

[ 9 t-"----'-.-'_ _ _ _-I ·.·1.54"mPd-Ag(25%) (8b) H

- .... 1.12~mPd(N1) !

IE 6~-------------------------I
~ 1.4 ~m Pd-Ag(22%) (N1b)

!. ~.",.~..;--.=.t
-£ t::. :." :;:.;:." ~
3+---------~------~--------~
330 370 410 450
T (.C)

Figure 7.1: Hydrogen permeances for Pd and Pd-Ag membranes (8, NI, 8b, NIb)

Figure 7.2 provides a good example of how silver influences the hydrogen permeance.
Metal films of membranes (N3) and (N4b) were of similar thickness. Membrane (N3)
consisted of pure Pel, while N4b had a Pd layer of 1.18 J.Un and a silver layer of
approximately 0.25 J.Un. The hydrogen permeance for membrane (N4b) was more than three
times lower than that of (N3), even though this silver layer was very thin.

Pd and Ag can be considered as two resistances in series. The purpose of alloying was to
obtain a homogeneous alloy. The two resistances in series would change to two resistances
in parallel and the hydrogen permeance would increase.

209
 15 .................................................................................................................. .,
~
In
oj - ..,a..- - - -

~ 12+---------~r_~~----------------~

1ia 9~~~-=-~-=-~4~====;::;~-~~======~
~ --+-1.471J1Tl Pd (N3)
3 ---2.141'"1 Pd--Ag(24%) (Nab) .
m 6 + - - - - - - - - - - - - - - - - - 1 - .... 1.181'"1 Pd (N4) r-j
~ 3 -+-1.43"m Pd·Ag(23%) (N4b) ,

~- .--::.
o+---------~----------.---------~
----
330 370 410 450
T (OC)

Figure 7.2: Hydrogen permeances for Pd and Pd-Ag membranes (N3, N4, N3b, N4b)

Table 7.2 lists nitrogen permeance and selectivity data for membranes 8b, Nib, N3b and
N4b. For all the membranes except (NIb), the nitrogen permeance decreased after the silver
layer was deposited. That was expected, since silver plugged defects that were present in the
Pd layer. Thicker films tend to have less defects, as discussed in Chapter 6. Membrane
(NI) had a very low nitrogen permeance and high selectivity. After silver deposition, the
film showed poor characteristics. That was the only silver plating experiment that resulted in
this behaviour. Other unalloyed Pd-Ag membranes in Appendix D2 (8b, N3b and N4b)
showed good selectivity after silver deposition. Two possible explanations are offered for
this behaviour:
• Some kind of physical damaged to the film might have taken place.
• There might have been hydrogen left in the system during cooling after testing (NI). The
hydrogen would then have caused some embrittlement of the pure Pd film on which the
silver was deposited.

The selectivities of the Pd-Ag films were lower than those of pure Pd films. The loss of
hydrogen permeance was far greater than the gain in reducing defects or the decline in
nitrogen permeance. The net result was a decline in the H2 to N2 ratio.

210
Table 7.2: N2 permeance and selectivity data for 8b, NI b, N3b and N4b
Temp. (0C) N2 permeance Selectivity Nz permeance Selectivity
(nmollm 2.Pa.s) (nmollm2.Pa.s)
MembraneS MembraneSb
450 21.42 671
410 21.85 627 11.48 472
370 22.42 553 11.90 394
330 23.07 457 12.24 336

MembraneNl Membrane NIb

450 3 4533
410 2.97 4251 57.09 91
370 2.84 3784 45.27 99
330 3.08 2953 48.03 77

MembraneN3 Membrane N3b

410 5.17 2090 3.22 999
370 5.22 1851 3.72 758
330 5.65 1479 3.66 632

MembraneN4 Membrane N4b

450 20.64 671
410 20.64 631 11.98 239
370 20.06 591 10.57 226
330 21.83 462 10.19 180

7.2.1.1. Unalloyed Pd-Ag membranes tested with a sweep gas

The equations derived in section 6.6.5 were used for calculating the hydrogen permeances of
membranes 8b, Nib, N3b and N4b. The calculated vs. the measured values of each
membrane are compared in Figures 7.3 to 7.6.

211
 100 100

•0
0

••
95 A ,
B
"~
•
95 r-,o--------_____ _
90 ' -_ _ _ _ _ _ __

E
•a. 90 / &Mt-------------
1

;E
1f.
""~ "
/ ,
;E
1f.
!
80

75
t----------
t--------- •
"

;;
;!
u•
80 / ,
]

~
10 t--------:,.-L- .--~-.- -, _._ .. -

/
65 /'

8O~
" 75 so
" 90 95 100
Measured V. Hl parmeance Measured % Hz penneance

Figure 7.3: Measured vs. calculated % Hz Figure 7.4: Measured vs. calculated % Hz
permeance for 8b permeance for Nl b

100
100

•
0

:"
90 /- ,
0 •
••"
90 ~
E
• /- E
• /._---
/,
0. 80 a. OJ
;E
1f.
70
;E
1f.
70 /-. --- --
""-!i ~•
.
;;
0
eo / ;!
•
80 /'. --- .- --

~
u
V
U

so 50
so 60 70 60 90 100 so eo 70 eo 90 100
Measured %Hz penneanee Measured % Hz penneance

Figure 7.5: Measured vs. calculated % Hz Figure 7.6: Measured vs. calculated % Hz
permeance for N3b permeance for N4b

The model equations (listed in section 6.6.5) predicted Hz permeances well, as long as the
nitrogen permeances were low (less than 20 nmol/mz.Pa.s). For membrane (Nlb),the error
between measured values and calculated values was large due to very poor selectivity of the
membrane (high nitrogen permeances). For the other membranes (N3b, N4b and to a lesser
extent 8b), the error between the actual and measured values was small. Most of the time the
calculated percentage of hydrogen permeated was less than the measured percentage.

Figures 7.7 and 7.8 show the measured hydrogen permeances as a function of temperature
and sweep gas flow rate for membrane (N3b). Figures 7.9 to 7.12 show hydrogen flow
profiles along the axis of the membrane for (N3b).

212
 100.0 -------.---------------- ------------------------------------- -.----------.---------------..
~--....--.---------~-~
100.0
~ -,:-..--.""1"1""----_.-..-..._-----
-:'
".0 ".0
90.0 ,-
".0
: . r •rJ
I .
~
".0

I
: I I :
94.0 BO.O
,I r
1
-£
920

90.0
t-__f-/'_ _ _ _---il-+-T:::.UO-c ."110- T:370"C ~ ;£ ro.O
75.0
I:
I: i
: r
;I. I ;I.
".0 . I 1---'T=330"C ".0
80.0
.' j_+_T;;410"C

__ 'T"330'C
-"110.- T=370"C ~

'6.0 55.0

".0
0 10 15 ~ ~ ~ ~ ~
50.0
0 , 10
" " 25 30 35
"
NI; SWeep rata IS molar % of ~ fHd NI; SWHP rate as molar % or HI; fIIltd

Figure 7.7: % Hz penneated with N z Figure 7.8: % Hz penneated with N z

sweep gas and space time = 2.3 7 s sweep gas and space time = 1.19 s

200 ------."----------------------------------_.-.------------.-----.-------------------.. -----.--,

""
I~
T_330"C

'80HIr-----------------i '0)

s T" 410 "C. 370"C -.td 33O"C

movintI trvrn Id to rW"t C \\~
\~"
s.Np . . ~ .1...... 40 end 28
cnlhNn _lIInfft1mll.tt to I1ght

~
1•
0
120

80
e
.,•
0
1aJ

80
\0>'~
.\
\
'
...........

";£ ,£ \,~""'"
,\ '...... ~
., ~

......
'"
\\,~
'--
............................

0
0 0.2 0.4 0.11 0.8
0
0 0.2 0.' 0.' 0.' ,
DImensionless membrane length Dimensionless membrana length

Figure 7.9: Hz flow in tube (cm3/min) as Figure 7.10: Hz flow in tube (cm3/min) as
predicted by model for N3b (1: = 2,37 s) predicted by model for N3b (1: = 2.37 s)

400 ----------------------.---------------------.------.----.-------------------------------------
.00

'20
T .. 410 '"C, 310 "C MIl S30 "C
'" Sw. . 0-"" .1,,10, 4O..-.d 20
C moWlt from ,... to right c anI""", -*'Ibm'" to righI:

~ "0 ~ 240

e e
~ 160 •
~ 160
"'
;£ ;£
80 80

0 0
0 0.2 0.4 o.e 0.8 0 0.2 0.4 0.8 0.8
Dimensionless membrana length DlnansfonlHs memtnnaltnglh

Figure 7.11: Hz flow in tube (cm3/min) as Figure 7.12: Hz flow in tube (cm3/min) as
predicted by model for N3b (1: = 1.19 s) predicted by model for N3b (1: = 1,19 s)

213
Figures 7.7 to 7.12 show that, in most cases, at 330°C the hydrogen penneance did not go to
completion along the length of the membrane tube. When Figures 7.9 to 7.12 are compared
with Figures 6.43 to 6.46 and Figures 6.49 to 6.52 (pure Pd films), it can be seen how far the
curves shift to the right when silver was added. Much higher sweep gas feed rates were
needed to extract a high percentage of hydrogen through the Pd-Ag films.

7.3. LITERATURE DATA ON ALLOYING Pd-Ag MEMBRANES

In section 7.1 it was mentioned that there are two procedures available for preparing Pd-Ag
membranes: either co-deposition or successive deposition. The thickness of the supported
film is important when choosing the most appropriate alloying conditions. There is not much
literature available on alloying Pd-Ag films; what there is, is mentioned in the next section.

7.3.1. ALLOYING CO-DEPOSITED Pd-Ag FILMS

Cheng and Yeung (1999) co-deposited a 1.2 I-Ull Pd gg -Agl2 film on the outside of a tubular
5 urn porous Vycor glass membrane with electroless plating. They annealed the membrane in
hydrogen at 400°C and 500 °C, respectively, for a total of 8 hours. After annealing, they
found a single phase (obtained from XRD data). They obtained, for the alloyed Pd-Ag
membrane, an increase of 1.4 to 1.7 times in hydrogen penneance compared to a pure Pd
membrane. The values for both were, however, very low:
• 0.10 l-Ullol/m2.Pa.s for the Pdgg-Ag film at 500°C, and
12

• 0.06 Ilmol/m2.Pa.s for the Pd film at 500°C.

Other references in Appendix E do not specifically state if and how they annealed their Pd-
Ag membranes.

7.3.2. ALLOYING SUCCESSIVE Pd-Ag FILMS

Uemiya et al. (199Ja) deposited successive Pd and Ag layers by electroless plating on the
outside of a membrane tube. The support was an asymmetric alumina tube with a 200 nm
pore SIze. Total film thickness ranged from 4.5 to 6.4 11m, with the Ag percentage ranging

214
from 11 to 30.5 wt %. Membranes were heat treated at 500°C, but the double layer structure
was not changed to a single alloy layer. Membranes were then treated in argon for 12 hours
at 800°C to 900 °C to yield a miscible palladium-silver alloy. They claimed to have 100%
selective membranes with a hydrogen permeance varying between 1.45 and
2
2.24 ).tmol/m .Pa.s at 400°C. The highest value was obtained for a 23 wt % silver film. The
reactor they used had alSO °C temperature difference between the inlet and the centre.

Shu et al. (1996a) prepared Pd-Ag membranes on porous stainless steel disks (200 nm pore
size) with successive electroless plating steps. The film consisted of a 2.8 ).tm Pd-Ag layer.
The Pdso -Ag20 film was annealed in hydrogen at 500°C for 5 hours. The Tamman
temperatures were 630°C for Pd and 344°C for Ag. They claimed that annealing at 500°C
would cause significant vibration and migration of silver atoms into the palladium lattice,
resulting in interdiffusion. Silver penetrated into the Pd layer to 1, 1.3 and 1.46 ).tm after
annealing 5 hours at 400°C, 500 °C and 600°C, respectively. Results were obtained by
Auger electron depth profiling. No hydrogen permeance or selectivity data was provided.

7.3.3. DIFFUSION KINETICS

Diffusion kinetics of thin films has been discussed in Brandes and Brook (1992). Shu et al.
(1996a) also summarised the relevant equations. Little data is available on diffusion
coefficients of Ag into a Pd matrix. Table 7.3 lists some diffusion coefficients for the Ag-Pd
system.

Table 7.3: Ag-Pd diffusion coefficients

Diffusion process T range (K) Do (m"/s) ED (kJ/mol) Reference
Bulk diffusion in 988-1215 4.52*10'0 183 Brandes [1] (1992)
Ag78 .2Pd
Chemical diffusion; 873-1173 7.4*10' lu 103 Brandes [2] (1992)
50% Ag in Pd
Interdiffusion in 773-1073 2.5* 1O"IT 72 Shu [3] (1996a)
layered Pd-Ag

215
The Arrhenius equation is applicable for diffusion data:

D=Doexp(- ED)
RoT
(7.1)

Figure 7.13 was constructed to compare diffusion rates (of Table 7.3) at different
temperatures by applying equation (7.1). The figure must be interpreted in the following
way: The times on the y-axis are indications of the time required to obtain the same amounts
of diffusion as after 100h at 500 °e, but at different temperatures. There is a significant
difference in the values from the three different sets of diffusion coefficients.

100 .................................................................................................................. .

~ Brand.. [1J (1992)

........ Brandes [2) (1992)

... Shu [3] (19960)

g ~+---~~~~~----------------~
..
E
~ ~+-----~~----~.---~~~----~

20+-------------~~------~~~~

o+-----.-----~----~----~----~
500 520 540 560 sao 600
Temperature 1°C)

Figure 7.13: Heating times required to obtain similar Pd-Ag diffusion at different
temperatures

7.4. ALLOYING RESULTS FOR Pd-Ag MEMBRANES

There seems to be significant differences in diffusion procedures described in section 7.3.
Uemiya et al. (1991a) only observed alloying after 12 hours at temperatures above 800 °e for
films exceeding 5 microns. Shu et al. (1996a), on the other hand, measured significant
interdiffusion of Ag into Pd after 5 hours at temperatures as low as 500 °e. Their film had a
thickness of 2. 8 microns.

216
In the present study the temperature range used for alloying was from 500°C to 600 °C.
Different alloying times and gas environments were employed. No literature data could be
found on alloying films on the inside of a membrane tube. A major problem associated with
online diffusion analysis is that the membrane has to be broken to reach the inner surface for
determining alloying (XRD analysis) and performing a depth analysis (Auger depth profiling,
PIXE analysis or EDX). Once a membrane is broken, it cannot be used again. A different
approach was taken in the present study. The permeance and the selectivity of the Pd-Ag
films were monitored to determine the effect of annealing on the membrane performance.
From the changes in hydrogen permeance and membrane selectivity, certain conclusions
could then be drawn.

Two alloying setups were used:

1. A constant temperature tube furnace with the membrane not fixed at either end, and
2. A reactor with the membrane fixed at both ends.

Section 3.3.6 described the advantages of both methods and the different variables
investigated.

7.4.1. ALLOYING IN A TUBE FURNACE

Membranes that were heat treated in the tube furnace (8c, N3b-N3e, N4c and N6) are
indicated in Table 7.1. Permeance and selectivity data are listed in Appendix D2. Most
membranes were tested with positive feed pressures and with a sweep gas.

7.4.1.1. Alloying at a temperature of 545°C

The first set of experiments was performed at 545 °C in argon. The oven had to be cooled
down after the alloying time to remove the membrane and place it in the reactor. The
membrane was tested after SOh, 100h and ISO hours and the results compared to those of both
. a pure Pd film and an unalloyed Pd-Ag film. Results are presented in Figure 7.14.

217
 7.0 .---------------... ------------------------------------_.-.. _._ .. -----------------.-------------------

.. 6.0 /

~ //
~ 5.0 //-/

! W/ 4.0

~ V 3.0

~ 2.0 t------------f'----:-+---;Tc::=-;:41~0o;;:C-,---!
~ 1.0 +------------1 ~T=370°C
........ T=33Q°C
-

0.0 +-----~----~-------1
o 50 100 150
TIme (hI

Figure 7.14: Effect of heating time in argon on hydrogen permeance through Pd-Ag
(membranes N3b to N3e)

Figure 7014 indicates a sharp increase in hydrogen permeance after the first 50 hours of heat
treatment. The values were more that double those for the unalloyed Pd-Ag film. After 100
hours, the hydrogen permeance showed a slight decline, but values changed little from 100
hours to 150 hours. The increase in hydrogen permeance confirmed that silver diffused into
the palladium matrix to yield a structure with less resistance than the separate silver and
palladium layers. The decline in hydrogen permeance at 100 hours and 150 hours might be
due to one of two reasons:
• The surface structure and morphology of the film changed after prolonged heating,
causing a decline in hydrogen permeance, and/or
• Silver diffused through the Pd layer and started accumulating next to the alumina
interphase. The structure then started approaching separate layers again, but with a high
silver concentration next to the membrane support.

Table 7.4 compares hydrogen permeabilities for the pure Pd film with those of the Pd-Ag
films. Permeability (Pm*l) was used rather than permeance to compare values, since
permeance did not allow direct comparison of films with different thicknesses. The Pd film
had a thickness of 1.47 !J.IIl and the Pd-Ag films a thickness of2.14!J.1Il. The silver content of
the film was 24%.

218
Table 7.4: Effect of heating on hydrogen permeances for membrane N3
Membrane Heating T - 410°C T-370°C T - 330°C
time (h)
Permeability (llffiol.J.ll11/m".Pa.s)
N3 0 15.89 14.19 12.27
N3b 0 6.87 6.03 4.96
N3c 50 14.10 12.82 11.09
N3d 100 12.54 11.24 9.18
N3e 150 12.90 11.58 9.82

Values in Table 7.4 indicate that after silver was added to palladium, the hydrogen
permeance did not return to the values obtained from the pure palladium. At all heating times
(from 0-150 hours), the hydrogen permeances were lower than for the pure palladium. This
contradicted results obtained by Cheng and Yeung (1999) and Uemiya et al. (1991a). They
recorded improvements for Pd-Ag films compared to Pd films.

Reasons for this contradiction could not be obtained from permeance and selectivity data
alone and could not be fully explained without further investigation. It must be noted,
however, that the Pd-Ag film produced on membrane (N3) had a hydrogen permeance that
was three times higher than that ofUemiya et al. (1991a) and more than 60 times higher than
that of Cheng and Yeung (1999). From this initial experiments, it could be concluded that the
surface structure and morphology of the metal film had a greater impact on hydrogen
permeance than the film composition when the hydrogen permeance was very high and the
film very thin (less than 2.5 microns).

7.4.1.1. Alloying at a temperature of 530°C

Membrane (N4b) was a 1.43 11m Pd77Ag23 membrane. The effect of heating it in argon for
30 hours at 530°C is shown in Figure 7.15. There was no real improvement in hydrogen
permeance after heat treating the Pd-Ag film in argon for 30 hours at 530°C.

219
 14

"iii' 12

:.
~ 10

~ 8
-=.
I:! •
ill
E 4
1N
::c 2

o
410 370 330
Permeance temperature (OC)

Figure 7.15: Effect of heating in argon on H2 permeance for N4

7.4.1.3. Alloying at a temperature of 590 °C

Membranes (8b) and (N6) were heat treated in argon for 15 hours at 550 °e, whereafter the
temperature was increased to 600 °e for a further 25 hours. Hydrogen permeance testing was
not performed on these membranes, because their selectivity became too poor. The effect of
temperature on membrane selectivity will be discussed in the following section.

7.4.2. TEMPERATURE STABILITY OF Pd-Ag MEMBRANES

Membrane stability can be expressed in terms of both nitrogen and hydrogen permeances. If
both the hydrogen permeance and the nitrogen permeance remain constant over time, the
metal film is concidered stable. If the hydrogen permeance decreases and/or the nitrogen
permeance increases, then the film quality is deteriorating.

7.4.2.1. Literature data on Pd membrane stability

Depending on the support structure, different mechanisms are responsible for membrane
deterioration. Some work has been performed on the stability of Pd deposited on metal
substrates (porous stainless steel or a refractory metal). Interdiffusion between Pd and the
support metal (see Shu et a1., 1996a) resulted in a decline in hydrogen permeance over time
(Buxbaum and Kinney, 1996). The effect on nitrogen permeance was not explicitly stated.
No interdiffusion occurs when Pd is deposited on an alumina support. Differences in thermal
expansion coefficients and changes in film morphology at high temperature could result in

220
film cracking and pore formation, leading to an increase in the amount of defects in the film
and an increase in the nitrogen permeance. Paglieri et al. (1999) performed the first
comprehensive study on the effects of film temperature on the film stability of pure Pd film
deposited on the inside of alumina membrane tubes. The majority of the films had
thicknesses between 8 and 27 micron.

Paglieri et al. (1999) tested membranes from 400°C to 600 DC, and some even up to
temperatures of 800 DC. They showed that the same deactivation occured in films of all
thicknesses. The thicker the film, the longer it took for the selectivity to fall at 550 DC.
Initial hydrogen to nitrogen selectivities were only about 30 to 80 for most membranes. After
several days at 550 DC, selectivity declined to a Knudsen value of about 4. They obtained
significant improvements in stability by omitting tin from the pretreatment process and
replacing it with Pd acetate. They concluded that the stability of the film was influenced by:
• Tin deposition during pretreatment,
• Components of the electroless plating solution (mainly EDTA) trapped between the
alumina surface and the Pd film, causing pore formation upon heating, and/or
• A difference in the thermal expansion coefficients of alumina and Pd.

After plating, the membranes were soaked overnight in water at 70°C. There was a very
good probability that there was EDTA trapped in the 200 nm pores of their membranes. In
the present study, plated membranes were rotated at high speed in an ammonia solution for
several hours. (EDTA has a high solubility in an ammonia solution, but a very low solubility
in water.) Thereafter, membranes were dried and oxidised at 320 DC in pure oxygen. Even
then, carbon was still present in some samples.

7.4.2.2. Pd-Ag membrane stability

No data exists on Pd-Ag membrane stability, for films of less than 2.5 microns, deposited on
the inside of alumina membrane supports. A Sn to Pd molar ratio, which was about 8 times
less than the conventional concentration for pretreatment (Shu et aI., 1993), was used for
pretreatment. The ratio in the present study was about 12 times lower than that for the full

221
strength sensitiser tested by Paglieri et al. (1999). Furthermore, a lot of attention was given
to post plating cleaning (see sections 3.3.2.3 and 3.3.5).

Figure 7.16 shows the increase in N2 permeance (at the permeance temperatures) of
membrane (N4) after 30 hours of alloying in argon at 530°C. The permeance remained
below that of pure Pd, because of the extra thickness of the film. The decrease in permeance
was about 50"10. The membrane retained good selectivity at 530°C.

2STr====================~----------'
iii I~ Pure Pd ED Pd-Ag Ell Pd-Ag, after 30h at 530"C I
oj
D. 20

1E~ 1S
I:
~

g 10

~
l
N
S
Z
o
410 370 330
Permeance temperature rCJ

Figure 7.16: Effect of heating in argon on N2 permeance for N4

Figures 7.17 and 7.18 show nitrogen permeance and selectivity data for membranes (N3b) to
(N3e). The heating temperature was 545°C in argon. Figure 7.17 shows a steady increase
in nitrogen permeance as the amount of defects in the film increased. Nitrogen flow
increased with decreasing temperature, indicating that the defects that formed in the film
were of Knudsen dimension (5 to 30 nm range). The selectivity of the membrane decreased
to values between 110 and 75 for temperatures from 410°C to 330 °C. A disadvantage of the
heating system was that it required cooling before the membrane could be tested. After each
set of tests, the membrane had to be reheated to 545°C to be alloyed further. This repeated
temperature cycling, between room temperature and 545 °C, could contribute significantly to
the decrease in membrane stability. The complete temperature cycling for membrane (N3)
was:
• 25°C to 450 ° to 25 °C for testing the Pd film,

222
• 25 °C to 410 to 25 °C for testing the Pd-Ag film,
• 25 °C to 545 to 25 °C for alloying the Pd-Ag film (50 hours at 545 0C),
• 25 °C to 410 to 25 °C for testing the Pd-Ag film,
• 25 °C to 545 to 25 °C for alloying the Pd-Ag film (50h; total = 100 hours at 545 0C),
• 25 °C to 410 to 25 °C for testing the Pd-Ag film,
• 25 °C to 545 to 25 °C for alloying the Pd-Ag film (50h; total = 150 hours at 545 0C), and
• 25 °C to 410 0 to 25 °C for testing the Pd-Ag film.

1200 .•• _ .....•.- .... _ . _ .....• - ...• -_ ...••.....••• '1

;; 8 0 + - - - - - - - - - - - - . . . . . " . 1000 k-------------j!
~

I :+-------#"'-------1 8001-~-----_1~~~r_~
r --T:4,O.C

~> ,OOf~~~----~~~T~.3~70~.C~

i:t-----;&:2:::.--i-~;::Tr;-.41i370;:;.cOl
-T=33Q"C

) <00 t---~,;:---------j
z 10~~~-----~~~T~.~~~.c~
o~---_---_---~
100 150 o 50 100 150
Time (hI Time (hI

Figure 7.17: Effect of heating in argon Figure 7.18: Effect of heating in argon
on nitrogen permeance (N3) on selectivity (N3)

Membranes (8c) (1.54 J.UI1 Pd75-A~5) and (N6) (2.05 J.UI1 Pd7"Ag21 ) were heat treated in
argon for 15 hours at 550 °C and then for a further 25 hours at 600 °C. The results are shown
in Table 7.S. A temperature of 600 °C is too high to tolerate for films of thickness in the
order of 2 J.UI1. Selectivity will decrease rapidly, making the membrane unsuitable for
separation at those high temperatures.

..
Table 7 S' Nitrogen permeance of membranes (8c) and (N6) after heating at 600 °C
Sample Nl permeance before Nl permeance after
heating (nmoVm1.Pa.s) heating (nmoVm1.Pa.s)
8c 12.2 (330 0c) 114 (room)
N6 8.5 (room) 151.4 (room)

223
7.4.2.3. Structural changes of Pd-Ag membranes during heating
A scanning electron microscope was used to analyse the surface structure of a Pd-Ag
membrane (membrane d) after various heating times. Top view images were taken to detect
pore formation and possible clustering with time. Side view images were taken to detect
peeling. The film had a theoretical thickness of 1.45 ).1m, which correlated well with the SEM
determined thickness. The silver content was 20%. Table 7.6 describes the different
membrane stages.

Table 7.6: Description of different membrane stages (dl-d3)

Stage Heat treatment
dl 50h in argon at 545°C
d2 100h in argon at 545°C
d3 135h at 545°C + 25h at 600°C in argon

Figures 7.19 to 7.30 compare membrane (d) at different heating stages and different
magnifications. When comparing Figures 7.19 to 7.24, it can be seen that there was an
increase in cluster size with an increase in heating time. Small clusters tended to agglomerate
to form bigger clusters. Some holes were visible in (d3) (Figures 7.21 and 7.24). As
agglomeration and segregation took place during prolonged heating at high temperatures, the
grain boundaries decreased. To compensate for this decrease, small pores or holes formed.

Figures 7.25 to 7.28 give very clear pictures of membrane changes during heating. After
50 hours of heating, there were no pores visible on the membrane surface. Some pores
started to develop after 100 hours of heating. In the middle of the left side of Figure 7.26
some holes can be seen. Membrane (d3), heated for 135 hours at 545°C and a further 25
hours at 600 °C, showed two different deterioration mechanisms. There were many pores
scattered all over the membrane surface (Figure 7.27). Furthermore, some porous clusters
started to develop. Two clusters can be seen in the middle left and middle right sides of
Figure 7.28. What is interesting is that those clusters had a large height dimension, many
times the thickness of the film. For the film to have had such an increase in height, the
increase in pore area had to be very large.

224
Figure 7.19: Membrane dl (25 OOOx) Figure 7.20: Membrane d2 (25 OOOX) Figure 7.21: Membrane d3 (25 OOOX)

Figure 7.22: Membrane dl (5000Ox) Figure 7.23: Membrane d2 (50 OOOX) Figure 7.24: Membrane d3 (50 OOOX)

225
Figure 7.25: Membrane dl (5 OOOX) Figure 7.26: Membrane d2 (5 OOOX) Figure 7.27: Membrane d3 (5 OOOx)

Figure 7.28: Membrane d3 (200Ox) Figure 7.29: Side view of d2 Figure 7.30: Side view of d3

226
Side view images (Figure 7.29 and 7.30) were taken to determine whether thermal expansion
caused the film to detach from the support. After 100 hours and after 170 hours, the figures
showed that film adhesion to the ceramic support remained strong.

7.4.3. ALLOYING IN SITU IN THE REACTOR

Membranes (NIx), (N2x), (N4x) and (N8x) were heat treated in the reactor. Heat treatment
in the reactor had the advantages of continuous analysis and little temperature cycling
compared to heating the membranes in the tube furnace. The disadvantage was that each
membrane was subjected to high temperature axial stress in the reactor, since both ends of the
membrane were fixed to the reactor.

7.4.3.1. Alloying in a hydrogen environment

Experiments were performed in Ar, H2 or N 2. For membranes (NIx) and (N4x), only
hydrogen was employed. Membrane (N4x) had a thickness of 2.13 lJ.IIl and a silver content
of20 wt % Ag. It was heated in hydrogen at 540°C for 30 hours. The hydrogen permeance
increased with time, from about 4.1 lJ.IIlol/m2.Pa.s to 5.7 Ilmol/m2.Pa.s (Figure 7.31) during
the alloying stage. The nitrogen permeance also increased. The sharp initial increase in N z
permeance, from 0 to 2.5 hours, was similar to the increase in hydrogen permeance. From
2.5 hours to 30 hours, the increase in nitrogen permeance was much slower than during the
first 2.5 hours. This indicates a gradual increase in membrane defects in a hydrogen
environment at 540°C.

Figures 7.32 and 7.33 show hydrogen and nitrogen permeances for membrane (NIx) heated
in hydrogen at 590°C, as a function of time. (NIx) had a thickness of 2.13 11m and a silver
content of22 wt % Ag. Unlike with (N4x), the hydrogen permeance decreased with time at
590°C (Figure 7.32).

227
 6 ................................................................................................. 10

-~ 5.8t===~~~.==l··
. J .. ' ,.'
5.6 .'
9
8
Ui'
.;

l 5.4 t--,.-"-,- ..-.. ,---=/...---;~------'----------f 7

I>.

~
[ ..",0::--/ - r - - - - - - - - - - - - t 6 E
5.2 i----,/,-;;

i 5+7,+'--~-----------------------+ 5 S-
f!r::
= 4.8 ttt--------;====;-----t 4
III
E 4.6 -1'-1---------1
1 ~h"'rog"" 11----+ 3 E
..1>.4.4 -r--I---------L1 .-...:.:nitroge=n..J-1_--t 2 8.
~

----=-

J: 4.2
N
+-+_______________+ 2
N

4+--·--r_--~--~----r_--._--_+ 0
o 5 10 15 20 25 30
Time (hI

Figure 7.31: Effect of heating time in H2 on H2 and N2 permeance for N4x

6 -----.----... ---.--..--........-----.------_._._._._._ ..... _._._-------------... _."'----'1

'" . .e·- .
~----
270

~.5t--~~---------~
..
2S .'
240

'l 4 t--------'=-,-==~===l
,,r' .... >~. .---,.... .---~- ..--
~
."
'

~ 20
c ,; i
~ • IL ! 210 ]
= ~ N'

~
EEls
K"
! iI '. = N

1
,
. 180 Z
-2 10 . .2N

r:
2

t----------------i S L~' __ H
". . .
.. 150
l:
1
,·, ...-'1
0
o 2 4 , 10 0 2 4
• • 10
120

li. . (h) lime (h)

Figure 7.32: Effect of heating time in H2 Figure 7.33: Effect of heating time in H2
at 590 °C on H2 permeance for Nix at 590 °C on N2 permeance for Nix

The nitrogen permeance increased sharply (Figure 7.33), indicating that defects developed
quickly at 590 °C in a hydrogen atmosphere. More defects should favour an increase in
hydrogen permeance rather that a decrease in hydrogen permeance. The decrease in
hydrogen permeance confirmed that structural and morphological changes took place on the
metal alloy surface. There was a decrease in grain boundaries, which resulted in a decrease
in hydrogen permeance. Figure 7.34 shows the effect of temperature on the Pd-Ag film after
heat treatment in hydrogen at 590 0c. Both nitrogen and hydrogen permeances were poor
compared to the permeances for pure palladium films.

228
 4.50 40

4.00
38
/
.. .,,; . .,
3.50

U
3.00 ---- /' --
36
U
,;
"E"' "-. 2.50 ~---- /_------ "
34 CIS CL

/
"
~ .§
N
E N'E
"'" -0E
N
2.00

/'
32:;
'"
N
0
::::
E
:I: .a 1.50
,/' 30 Z .s
1.00
1 ....... hydrogen I
0.50 I ...- nitrogen I
• 28

0.00 28
250 300 350 400 450 500 550 600
Permeance temperature (Oe)

Figure 7.34: Effect of temperature on Hz and N z permeance for Nix

7.4.3.2. Alloying in nitrogen and argon environments

Extensive testing was performed on membrane (N2x). The membrane had a thickness of
2.02 !-1m and a silver content of 25 wt %. The data is depicted in Figures 7.35 to 7.38. For
the first 25 hours, the membrane was heated in nitrogen at 500 cC. Hydrogen permeance was
determined by switching from nitrogen (or argon) to hydrogen and allowing for one hour in a
hydrogen atmosphere before the reading was taken. From hours 25 to 50, the environment
was changed to argon at 500 ° C. For hours 50 to 75, the temperature was increased to
550 °C, while maintaining the argon environment.

Experiments in a nitrogen atmosphere (0 to 2Sh) indicated that nitrogen had a passivating

effect on the film. The hydrogen permeance decreased with time (Figure 7.35). Apart from
the expected increase in nitrogen permeance in the first hour, the nitrogen permeance
remained constant at 500°C for 25 hours. The increase in the first hour was because the
surface was activated with the first introduction of hydrogen and any oxides on the surface
were reduced.

With the introduction of argon, the hydrogen permeance started increasing at 500°C. Argon
did not cause passivation of the Pd-Ag surface, as was the case with nitrogen. Not many

229
defects fonned in the film at 500°C from 25 to 50 hours. There was only a very slight
increase in argon penneance during this period (Figure 7.37).

From 50 to 75 hours, there was a sharp increase in hydrogen penneance at 550°C in argon
(see Figure 7.35). Values increased from 3.0 jlmol/m2.Pa.s to 5.3 jlmol/m 2 Pa.s. Many
defects were introduced into the film during this time (see Figure 7.37). The stepwise
increase in argon flux indicates that heating in hydrogen was the main reason for the increase
in defects. Hydrogen was used for reduction at the following times and the argon penneances
were measured before and after hydrogen introduction:
• 50-51 hours,
• 55.5-56.5 hours,
• 63.5-64.5 hours, and
• 73.5-75 hours.

Figure 7.38 shows the effect of temperature on the membrane after 75 hours of heat
treatment.

..
6.0 25.0

•• 5.0
---- C. 20.0 { .-_.. -
0,

le 4.0 / .'"e
~ I °e 15.0

-=• 3.0 .:.

u
--- g
•••
10.0

E
2.0
••
E
! 1.0 !l 5.0
,;: 2:
0.0
0 15 30
Time (h)
45 .. 75
0.0
0 5 10
Time (h)
15 20 25

Figure 7.35: Effect of heating time in N2 Figure 7.36: Effect of heating time in N2
and AI at 500 and 550°C on H2 at 500°C on N2 penneance for N2x
penneance for N2x

230
 30.0 6.0
or,;
0,
26.0 / or
,; 5.0 t----'--------------,A
/
}:-T
~0
22.0
E

~
oS
~
0 16.0 ------------
••E ~

-
Co 14.0
«
.- .. ---~--------- ----_._-- --- k : ------_____ . _
8.
r."
1.0

10,0 0.0 '---_ _ _ _ _ _- _ - - _ - - - "

25 35 45 55 65 75 250 300 350 400 450 SOO 550
Time (h) T (OC)

Figure 7.37: Effect of heating time in AI at Figure 7.38: Effect of temperature on H2

500 and 550 °e on AI penneance for N2x penneance for N2x

7.4.3.3. Further treatment ofPd-Ag films

Table 7.7 was compiled to compare the hydrogen penneances of all heat treated Pd-Ag films
at 410 °e.

Table 7.7: Hydrogen penneances at 410 °e

Membrane Thickness (/lm) %Ag H2 permeance at 410°C
(/lmol/m 2.Pa.s)
N3c 2.14 24 6.59
N4c 1.43 23 3.02
NIx 2.13 22 2.52
N2x 2.02 25 2.95
N4x 2.13 20 3.16

Values in Table 7.7 were similar to those of (N3b) and (N4b), where no heat treatment was
perfonned. The exception was (N3c), which had a hydrogen penneance about twice that of
the other membranes. The difference between (N3c) and the other membranes was that after
50 hours of heat treatment in argon, (N3c) was oxidised at 320 °e and then reduced. The
other membranes were not re-oxidised after heat treatment.

231
This phenomenon was investigated further. Membrane (N4x) was oxidised after heat
treatment (see 7.4.3.1). (N8x) was prepared, heat treated for 10 hours in argon and then
oxidised. The effects of further oxidation on hydrogen permeance for those two membranes
are summarised in Table 7.8.

After oxidising and reducing the membranes agam (after heat treatment), the hydro,gen
permeances increased. In the case of (N8x), the H2 permeance more than doubled. On the
negative side, additional oxidation weakened the film and created defects. Two different
oxidation temperatures and times were tested. Oxidation at 350°C was too severe.
Oxidation for 1 hour at 310 °C yielded much better results. Further oxidation and reduction
changed the surface morphology and structure of the film. These changes promoted
hydrogen movement through the film.

Table 7.8: Hydrogen and nitrogen permeances at 410°C after oxidation

Membrane Oxidation Oxidation Hl permeance Hl permeance
Temp (0C) Time (h) Before oxidation After oxidation
(IlmoIlm2.Pa.s) (JlffioIlml.Pa.s)
N4x 350 3 3.16 4.62
N8x 310 1 2.85 6.03

N2 permeance Nl permeance
Before oxidation After oxidation
(nmoIlml.Pa.s) (nmoIlm2.Pa.s)
N4x 350 3 9.6 24.5
N8x 310 1 13.1 18.2

7.4.3.4. Suggested heat treatment procedure for Pd-Ag films

The following procedure is suggested for heat treating Pd-Ag films ofless than 2.5 f.lm.
• Heat the Pd-Ag membrane in argon up to a temperature of between 540°C and 550 °C,
• Heat treat in argon for 10 hours at the above temperature,
• Cool down in argon to 310°C and oxidise in oxygen for 1 hour,

232
• Heat up to 400°C in argon, and
• Reduce in hydrogen for 2 hours.

The stability of the Pd-Ag membranes at temperatures from 200°C to 300 °C were monitored
during the catalytic membrane reactor experiments and will be mentioned in Chapter 9.

7.5. FULL CHARACTERISATION OF MEMBRANES N4x AND N8x

Membranes (N8x) and (N4x) were used for catalytic membrane reactor experiments. All H2
and N2 permeance data for membranes (N8x) and (N4x) is listed in Appendix D2. They
were tested at temperatures ranging from 250°C to 410 °C. Figures 7.39 and 7.40 plot the
hydrogen and nitrogen permeances of membranes (N8x) and (N4x), from 250°C to 410 °C.
Appendix D2 contains the calculated Arrhenius parameters.

5.'
__ HyG'ogen perm-.c:e
28 '.5
"
..- ;;
.•
-+--Hylt'ogM pIIrmunDII
if OJ '.0 ,
l, "
•.,
- tI- Nltragen pBIIINnCl!

.'
, ..-
27 &: ,
- tI- Nllrvslen plIITTI_nce
.// l
".'"••
/'/'
1; ,
1; 5.5

'"E•
,
,,
'"~
"
" -_.¥'. 21 ~
" / ......- 5.0
"
~
.; "
'" s. .; /' oS
3.5
" •uc ~- ... .:~--
3
c
I 3.'
.. "
./
:<
"~
"
" ,
,
3c
•• =
~
4.5

, ~/
.'
1.
:
=
•
~
7 E
I
"
4.0
1 2.5
,
, ,
25
•... &
:i! 3.5 ./
..-
:£
%: -£
3.0 1.5
2.0
250 290 370 " 250 290 330 370 410
'"
TrC)
410
T I'C)

Figure 7.39: H2 and Nl permeances for N4x Figure 7.40: Hl and Nl permeances for N8x

Calculated values correlated well with measured values when a sweep gas was employed
(Figures 7.41 and 7.42). For membrane (N8x), the error was typically less than 5%, but for
N4x it was sometimes slightly larger. This was due to the higher N2 permeance of (N4x).

233
 100 100

~ 95
c 1-- ~ •c "
0 .---.-~
:"
•• •• 90
i 90 '.
A'" E / .. •
.-.---.--.-~

,;: .. /.: • "

0-
,;: / . --~-----

/. .
eo
~ ~
L .
. '"
!l
•
i
1;1
75
/.
/. .. ~ L
"5 70 ----
] 0
1i
~ ~

U•
75

70 V . U

60
60 65 70 75 80
----" --

85 90 95 100
70 75 80 85 go 95 100
Measured V. H3 permeance Measured % H2 permeance

Figure 7.41: Measured vs. calculated Figure 7.42: Measured vs. calculated
% H2 permeance for N4x % H2 permeance for N8x

7.6. SUMMARY
Pd-Ag membranes of less than 2.2 IJlIl thickness were prepared by successive electroless
plating steps. The permeances of those films at 410°C ranged from 3.0 to 5.5 lJlIlol/m2.Pa.s
before heat treatment. That was significantly lower than the values for pure Pd films.

Different heat treatment methods were employed to improve the hydrogen permeance of the
Pd-Ag film. In all the gas environments tested, the films weakened at and above 590°C.
Hydrogen created defects in the film at a moderate rate at 550 °C and at a fast rate at 590°C.
Furthermore, hydrogen passivated the film at 590 °C but not at 540°C. Film passivation
resulted in a decline in hydrogen permeance. Continuous thermal cycling contributed
towards film defection.

Some defects formed in argon at 545°C after 100 hours of heating, but very few defects
formed at 500°C in either nitrogen and argon. Nitrogen also passivated the film and should
not be used for heat treatment. The best conditions for heat treatment were:
• an argon environment,
• at a temperature of 550 °C, and
• a heating time of about 10 to 15 hours.

234
After heat treatment, the film has to be oxidised at 310 °C for not more than 1 hour and then
reduced to obtain the best combination of hydrogen permeance and selectivity. This
additional oxidation-reduction step resulted in a great improvement in hydrogen permeance.
This phenomenon has not been observed previously by any other research group.

235
 8. CATALYTIC MEMBRANE REACTOR MODELLING

This chapter will focus on the development of a catalytic membrane reactor model. All
aspects of the model will be investigated and a complete sensitivity analysis will be
performed to determine the effect of each parameter on the performance of the reactor. The
effect of the feed rate on reaction conversion is discussed. Experimental data for a plug flow
reactor at different feed flow rates are presented.

8.1. REACTANT FEED RATES

Most work performed to date on dehydrogenation in catalytic membrane reactors has been
done with very low feed flow rates. This approach exaggerates the theoretical advantages of
the membrane reactor and provides very good results. In Figure 8.1 the region marked by
block 1 is the region of very low feed rates. Feed rates increase from left to right and the
space times increase from right to left. Data gathered in the region of 1 is of very little
practical importance. Such low feed flow rates result in an extremely low product production
rate even though the conversion is increased.

~
equilibrium
L--
• •
1 2
•
•
Feed rate ---4

+ - - - Space time

Figure 8.1: Theoretical effect of feed rate on equilibrium conversion

In the present study high feed rates were employed to determine the practical importance of a
catalytic membrane reactor. Experiments were performed in the region marked by block 2.

236
If a catalytic membrane reactor can significantly improve reaction conversion when the feed
rate is fast enough not to limit the reaction by the equilibrium conversion, then the reactor
becomes of practical importance. Table 8.1 summarises some experimental conditions
employed by other researchers for dehydrogenation reactions with liquid feeds in a
membrane reactor. The feed rates used in the present study are included, to illustrate the
where the present study 'fits in'.

Table 8.1: Different liquid hourly space volumes (LHSV) used for dehydrogenation
reactions in a membrane reactor
Dehydrogenation Temperature range LHSV (h) Reference
reaction (0 C)

cyclohexane <200 48-263 Itoh (1987)

methylcyclohexane 300-400 2-12 Ali (1994)
cyclohexane 185 Approx.42 Cannon (1992)
ethanol 250-310 5-22* Deng (1995)
ethanol 175-275 110** Raich (1998)

ethanol 250-300 0.75-0.25 The present study

2-butanol 190-240 0.4-1.5 The present study

* Assummg a catalyst mass of 3 g and reactor volume of 8 cm.>

** Using an argon feed flow of 6.5 mVmin and a reactor volume of 8 cm)

8.2. ALCOHOL PERMEATION DATA

The permeance of ethanol through film defects was determined for a pure Pd film and the
result compared to the nitrogen permeance of the same film. The experimental set-up and
procedure used have been discussed in section 3.3.4. Figures 8.2 and 8.3 show ethanol
selectivity as a function of temperature, compared to nitrogen selectivity. Ethanol had a
lower permeance through the film than nitrogen, because of the larger molecule size. This
resulted in a higher hydrogen to ethanol selectivity compared to the hydrogen to nitrogen
selectivity. The nitrogen to ethanol selectivity varied between 3.1 and 4.4, with temperature.

237
To simplify calculations, an average of 4 was assumed for modelling. This value was
different from what would be expected using the Knudsen theory, which suggests a nitrogen
to ethanol selectivity of about 1.3. Other transport effects through the defects in the
membrane occurred simultaneously with Knudsen flow. Since the measured nitrogen to
ethanol selectivity was much larger than the Knudsen value it shows that many defects were
smaller than the Knudsen defects. The presence of a large amount of molecular sieving
defects would explain the difference between the nitrogen and ethanol permeances. The
ethanol molecules were too large to pass through many defects where nitrogen molecules
could pass through.

The permeances of other organic species were not measured. The worst case scenario was
used for determining permeances for modelling. The selectivities were, taken as the ratio
between the molecular masses of each species (Knudsen theory) relative to ethanol. This
could be expressed as:

Pm,1 =Pm,2 M ~M", (8.1)

.,2

The permeance for 2-butanol will thus be 1.27 times lower than that of ethanol.

3300 .................................................................. ~ ... '.5

~
2"'"

2300
r
+-_________f • "2,N2
~
L
~ '"
• "2j-_~\-_--
!,.+-------\\c-------
J
1800
I- H2:.tlwloI r-
1300
! ,.+---------''r-\----
iii ,.,+-_______~..._---_
'---...
80IJ

300
,.,+--------------
320
"" 360 380 400
Temperature ('C)
420 440 460
". "" .20

Figure 8.2: Comparison between H2 to N2 Figure 8.3: Ethanol to nitrogen

and H2 to ethanol selectivities for a permeance ratio for a 1. 5 J.Ul1 Pd film
1.5 J.Ul1 Pd film

238
8.3. DEVELOPING A PROCESS MODEL
A preliminary model was constructed to mathematically describe the dehydrogenation of
alcohols in a catalytic membrane reactor. The process used a nitrogen sweep gas to create a
partial pressure differential across the membrane. The flow of the sweep gas was co-current
to that of the feed. A practical sweep gas in industry could be steam, which would supply
heat to the endothermic dehydrogenation reaction and create the hydrogen concentration
gradient across the membrane. Alternatively, the shell side of the reactor could operate under
vacuum conditions.

8.3.1. MODEL ASSUMPTIONS

The following assumptions were made:
• isothermal operation,
• plug flow on both the tube and the shell side,
• isobaric conditions, in other words negligible pressure drop on the shell and tube sides,
• no radial concentration gradients in the catalyst bed, and .
• steady state operation.

Both the shell and the tube sides were operated under atmospheric conditions. The pressure
drop across the length of the packed membrane tube (length of catalyst bed = 20 cm) was less
than 15 mbar. Isobaric conditions along the membrane's axis was therefore a good
assumption. To incorporate an axial pressure drop, the Ergun equation (8.2) could be used
(Fogler, 1992).

(8.2)

where

P = pressure [Pal
= void fraction of packed bed

239
gc = conversion factor (1 for SI units)
Dp = particle diameter [m]
Jl = gas mixture viscosity [kg/ms]
L = length down the tube [m]
U = empty column velocity [mls]
p = gas density [kg/m3 ]
G = empty column mass velocity = pU [kg/m2s]

For porous membranes (non palladium based) exhibiting Knudsen diffusion, the penneance
equation cannot be used and a radial pressure drop model must be employed (Deckman et a!.,
1995).

Radial concentration models have been studied for a catalytic membrane reactor (Becker et
a!., 1993; Gobina et aI., 1995a-c). Gobina et al. (1995b) found that the radial concentration
change was negligible for a membrane with a 7.8 mm inner diameter.

The final assumption of steady state was not very accurate, but was made to simplify
calculations. The data was gathered within a 24 hour period to give a better approximation of
steady state. Thereafter, both the catalyst and the membrane were re-activated before the
next set of tests. Non-steady state is due to both catalyst and membrane deactivation with
time, which influences both the kinetics and the separation aspects of the process.

8.3.2. DESIGN EQUATION

A catalytic membrane reactor IS a plug flow reactor with separation. The process IS

illustrated in Figure 8.4.

The flow on the tube side can be expressed by:

dF J
-=r-- (8.3)
dV L
or in dimensionless fonn, with the reaction at bulk conditions

240
 (8.4)

mainly
nitrogen sweep hydrogen and
penneation nitrogen

alcohol - - ---
-- .
_ reaction - _
_
reaction _
alcohol +

- --_--» products

) mainly
nitrogen sweep penneation hydrogen and
nitrogen

Figure 8.4: Description of catalytic membrane reactor process

Similarly, on the shell side:

dQ; =V[~J]
L
d-
R'
m
(8.5)

La

In (8.4), i = alcohol feed, reaction products and hydrogen, while in (8.5), i = nitrogen sweep
gas. The boundary conditions for co-current flow are:

At L= 0, F; = F;o. (8.6)
At L= 0, Qi = Q;o. (8.7)

The flux equations were derived in section 6.6.5. For equations (6.1) to (6.5), the right hand
side of the equations represents the penneance component, expressed in tenns of feed flow
rates.

241
A kinetic expression for ethanol dehydrogenation was derived in equation (5.21), with
optimised parameters in Table 5.6. The 2-butanol dehydrogenation expression was given in
equation (5.26) with optimised parameters in Table 5.8. The permeance and kinetic
expressions will be summarised in a later section.

8.3.3. MODELLING OF MASS TRANSFER EFFECTS

Interphase and/or intraparticle mass transfer resistance reduce the actual reaction rate,
because the reactant concentration within the catalyst particle becomes lower than the bulk
concentration. It is important to know whether the process is reaction rate limited or mass
transfer limited, or both. Furthermore, it is important to know which mass transfer process is
limiting. Sometimes all limitations can be present at once. Mass transfer is dependent on
physical mixture properties like diffusion coefficients, density and viscosity. As the reaction
takes place, the mixture composition changes continuously along the reactor's axial position.
The limiting step in the reaction can change with position.

In most instances, correlations do not exist to predict properties of ternary and higher order
mixtures. Furthermore, the available correlations may contain significant errors.

8.3.3.1. Interphase mass transfer resistance modelling

The first step in modelling mass transfer is to calculate the interphase mass transfer
coefficient (kc). Correlations have been formulated to relate this coefficient to various
physical properties of the liquid or gas flowing through a packed bed reactor. The most
common equation is the Thoenes-Kramers correlation (Fogler, 1992) for flow through a
packed bed.

(8.8)

mass transfer coefficient [mls]

particle diameter (equivalent diameter of sphere of the same volume) [m]
= [(6ht)(volume ofpellet)t3 [m]

242
Cb = void fraction of packed bed
y = shape factor (external area divided by 7rD/)
JJ. = viscosity [kg/m.s]
p. = fluid density [kg/m 3 ]
u = JJ./p = kinematic viscosity [m 2/s]
DA,m = gas phase diffusivity of A in bulk fluid (mixture) [m 2 /s]
U = Empty column velocity [rnis]

Subscript m refers to mixture properties.

The equation is valid for: 0.25 < Cb < 0.5; 40 < Re' < 4000 and 1 < Sc < 4000, with

DpU
Re=-- (8.9)
v
v
Sc=-- (8.10)
DAm
Re
Re'= (8.11)
(1- Gb)Y

Sh'= (Re,)15 (SC).K (8.12)

ShGb
Sh'= (8.13)
(I-Gb)Y

For equations (8.9) to (8.13), mixture properties also apply.

Accurate correlations exist for determining physical properties of mixtures at various

temperatures and pressures. Empirical estimation methods have been discussed in Millat et
al. (1996).

Diffusion coefficients in multi-component gas mixtures are difficult to calculate. Many

accurate correlations exist for determining binary diffusion coefficients (Perry and Chilton,
1973), but not for higher order mixtures. Some studies on diffusion in three component gas

243
mixtures (Toor, 1957; Duncan and Toor, 1962) and multi-component gas mixtures (Curtiss
and Hirschfelder, 1949; Wilke, 1950; Fairbanks and Wilke, 1950) have been performed. In
the simplest form, the diffusion coefficient can be estimated by the following:

1- Y A
D A •m = Y Y Y (8.14)
_B_ + _C_ + _D_ + ...
DAB D AC DAD

For laboratory reactors with low flow rates and thus low Reynolds numbers, the Thoenes -
Kramers equation is invalid. Satterfield (1980) listed some alternative correlations for
determining the mass transfer coefficient.

(8.15)

If 3 < Re < 2000 and 1.8 mm < Dp < 9.4 mm

(8.16)

For: 0.1 < Re < 10

The value of the constant, C 1, differs depending on the literature source. Satterfield (1980)
suggested a value of 0.07, while Collins (1993b) used a value of 0.015, determined from
experimental data.

With kc known, the Mears' criterion can be used to determine whether external diffusion is
rate limiting (Fogler, 1992). When

(8.17)

244
where

n = reaction order
Dp = particle diameter [m]
Pb = bulk density of catalyst bed [kg/m l ]
CA,b = bulk concentration of A [kmoVml ]
ko = mass transfer coefficient [mls]

then external mass transfer effects can be neglected. It is clear from (8.17) that external mass
transfer will become more important as the feed concentration drops along the membrane
axis and CA,b decreases.

To link the reaction rate coefficient and the mass transfer coefficient, a mass balance can be
performed around the outside of the catalyst surface.

(8.18)

koAp(P A,b - PA,surf)/8.314T = 1]rnr' A,surf (8.19)

where
Ap = external surface area of the catalyst
= 6m(1- eb)fPbD p [m2]
m = mass of catalyst [kg]
r' A,surf the surface reaction rate [moVkg.catalyst.s)
= the reaction rate in the absence of any mass transfer limitations
1] = effectiveness factor

The bulk concentration of the reactant is known, but not the surface concentration or the
surface rate. Equation (8.18) or (8.19) must be used to determine the surface concentration or
surface pressure in terms of the bulk concentration or bulk pressure. The difference between

245
the two is due to the external mass transfer resistance. When no external mass transfer
resistance is present, the surface concentration is equal to the bulk concentration. It is thus
very important to obtain kinetic data free from mass transfer limitations, because if mass
transfer limitations are present, the measured rate will be different from the surface rate and
calculations will be faulty.

In the cases of ethanol and 2-butanol dehydrogenation, the reaction rate on the right hand side
of (8.18) or (8.19) must be replaced by either equation (5.21) or (5.26): The surface
concentration or pressure of the alcohol must then be detennined in tenns of the overall
effectiveness factor and the bulk concentrations or bulk pressures of all species. The
dehydrogenation rate equations include surface partial pressures for the feed as well as for the
products. As a first approximation, it can be assumed that the interphase mass transfer
resistance of all the species to reach the surface is the same, and then the following
simplification can be made:

p
p _ A,surf P
i.surf - P i.b
(8.20)
A.b

Equation (8.20) is substituted into the rate equations (5.21 or 5.26) and the rate equation is
then substituted into (8.19). The result is an equation with three unknowns: kc, 11 and PA,sun'
Mathematically PA,surf can be expressed by:

(8.21)

To solve equation (8.21), k, kc and 11 must first be detennined. All the parameters can be
obtained from experimental data by perfonning experiments in the absence of intraparticle
mass transfer resistance (see 8.3.4).

8.3.3.1.1. The effectiveness/actor

The definition and application of the effectiveness factor can cause great confusion. In the
present study the effectiveness factor is taken as the difference between the measured

246
reaction rate and the rate that exists on the catalyst surface. Both the effects of interphase and
intraparticle mass transfer resistance are incorporated into TJ.

- r~.b = 7]( -r~,sun) (8.22)

The best approach is to determine TJ experimentally. Very small catalysts particles

« 0.6 mm) must be used to limit internal mass transfer resistance and prevent channelling of
the feed gas through the reactor (see 4.1.3.3). The reaction rate should then be measured in
the region where the reaction rate does not change with feed flow rate (Reynolds number).
The ratio between reaction rates at low feed flow rates and very high feed flow rates gives an
experimental value for TJ. The value of TJ is dependent on the feed flow rate or Reynolds
number. The effectiveness factor can also be estimated theoretically, as will be discussed in
section 8.3.3.2.

8.3.3.2. Intraparticle mass transfer resistance modelling

In section 4.1.3.3 the effect of particle size on ethanol conversion was discussed. Figures
4.25 and 4.26 show that for particles smaller than 850 j.lIIl, the ethanol dehydrogenation
results did not vary too much and thus, it is acceptable to ignore intrapartic1e mass transfer
resistance for the experiments in the present study, The theory of intraparticle mass transfer
resistance modelling will only be discussed briefly.

The catalyst particle is modelled as a sphere. The concentration of the reactant declines from
the outside surface into the pores of the particle, The following expression can be used to
describe the radial concentration profile in the particle (Fogler, 1992).

d[-D, ~(r')2]
(8.23)
dr'

For non-reversible nth order reactions, the equation can be written in dimensionless form and
solved algebraically (Fogler, 1992). For complex reactions, the problem must be solved

247
munerically. Collins (1993b) formulated equations for solving a reversible reaction. In
dimensionless form the following equation was obtained:

(8.24)

mole fraction of A in catalyst

where X fA =--:---::-~.------,----~ (8.25)
. mole fraction of A in bulk

and

(8.26)

Equations (8.24) to (8.26) apply for both the dehydrogenation of ethanol and 2-butanol.

Boundary conditions and material balances for each speCIes In the reaction have been
summarised by Collins (1993b). The effectiveness factor, 11, is the actual overall rate of
reaction divided by the rate of reaction at bulk conditions (CA,b)' Some authors define 11 as
the actual reaction rate divided by the rate that would result if the entire interior surface was
exposed to the surface conditions (Fogler, 1992, calls it an internal effectiveness factor). This
definition was, however, not used by Collins (1993b). He defined 11 as an overall
effectiveness factor. The effectiveness factor at a specific axial position along the
membrane's axis can be expressed by:

(8.27)

248
Equation (8.27) must be evaluated numerically. Bulk conditions are used for solving
equation (8.27). With the value of 11 known, the surface concentrations or partial pressures
can be calculated with equation 8.21. With the surface properties known, the surface reaction
rate can be calculated by using equation (8.18) or (8.19). Finally the reaction rate at bulk
conditions can be calculated:

- r~.b = 7]( -r~.,wf) (8.22)

The main goal of the modelling exercise was to calculate the value of 11. The surface reaction
rate could be measured under special conditions and once 11 was known, the reaction rate at
any bulk conditions could be calculated.

The Weisz-Prater criterion (Fogler, 1992) for internal diffusion can be used to detennine
whether there are significant internal diffusion limitations. The Weisz-Prater parameter is
define by:

(8.28)

If CwP « 1 then there are no internal diffusion limitations.

8.3.3.2.1. Effective diffusion coefficient (DeJ

In equations (8.23) and (8.24) a new coefficient, the effective diffusion coefficient, was
introduced. This coefficient describes the movement of molecules inside a catalyst's pores.
Movement can be either through bulk diffusion inside the pores or by Knudsen diffusion.
Both these mechanisms are incorporated into the effective diffusion coefficient. Empirical
correlations for detennining Dei are listed in Satterfield (1980).

8.3.3.3. Summary of modelling mass transfer

Incorporating mass transfer limitations into the reaction rate equation can be summarised by
the following steps.

249
1. Detennine the dimensionless parameters, Re and Sc.
2. Calculate the mass transfer coefficient, kc, using one of the equations (8.8), (8.15) or
(8.16). In the case of (8.16), C1 will have to be detennined experimentally.
3. Detennine whether the external mass transfer is rate limiting by testing Mears' criterion
(equation 8.17).
4. Calculate the effective diffusion coefficient defined by Satterfield (1980).
5. If external mass transfer is not rate limiting, then surface conditions can be taken as bulk
conditions and the Weisz-Prater criterion can be tested to detennine if internal mass
transfer is limiting. If both the interphase and itraparticle mass transfer resistances are
negligible, then the effectiveness factor equals one.
6. If external mass transfer is rate limiting then the Weisz-Prater criterion cannot be tested
since the surface concentrations are not known. The reactant surface concentration or
pressure must be calculated in tenns ofT] by using equation (8.18) or (8.19).
7. The effectiveness factor must be detennined from equation (8.27).
8. With kc (after point 3), PA,surf (from point 6) and T] (from point 7) known, the surface
reaction in equation (8.19) can be calculated.
9. The reaction rate using the bulk conditions can now be detennine from equation (8.22),
since both the effectiveness factor and the surface reaction rate are known.

8.3.4. MASS TRANSFER COEFFICIENTS FROM EXPERIMENTAL DATA

Mass transfer coefficients for the dehydrogenation of ethanol and 2-butanol were detennined
experimentally. As mentioned previously, small catalyst particles were employed to limit
internal mass transfer resistance and prevent channelling of the feed gas through the reactor.
Only external mass transfer resistance was considered as a first approximation.

8.3.4.1. Ethanol dehydrogenation

The first step in detennining the mass transfer coefficient was to detennine the Reynolds
number and thus the viscosity of the gas mixture. The viscosity is dependant on the mixture
composition and the temperature at low pressures. The method of Reichenberg (Millat et ai.,
1996) was used to determine pure gas viscosities and the method of Davidson to detennine
the viscosity of the gas mixture (Millat et ai., 1996). Binary diffusion coefficients were

250
determined by Wilke-Lee's (Millat et aI., 1996) method and the ethanol diffusion coefficient
in the mixture was determined by Wilke's method (Fairbanks and Wilke, 1950). A
simplification was made for determining mixture properties. The mixture was assumed to
only consist of the three main components, ethanol, acetaldehyde and hydrogen. The ethanol
selectivity towards acetaldehyde production was close to 100% for kinetic experiments
(Chapter 5) and thus the assumption was very good. Figure 5.9 was taken from Chapter 5
to show the effects of interphase mass transfer resistance. Figure 8.5 converts data from
Figure 5.9 to a plot of the effectiveness factor vs. Reynolds number. The surface reaction
rates (r'A,,) used for calculating the effectiveness factors (see Figure 5.9) are listed in Table
8.2.

Table 8.2: Surface reaction rates at different temperatures

Temperature Surface reaction rate
(OC) (mollkg cat.h)
200 33.5
225 47.8
250 88
275 227.3
300 275.8

1.20

1.00 +------,--.,----:;=""""------1
'ff?
ooo+__------~~~-----------

)_o~+------~~~~·-/-··-·--T-=-~-O-.c-w-~-o-.c-----1
/
, .., +__--i#"-------------------
If
O~+__~---------------------

,~~~;~~.~:~.---------.~----------
, ,.
000+----__- -__- -__- -__- -__- - _
5 10 15
_flow_I....)
2D 30 O.OOQ 0.020 0.040 0.080
~Id.
0.080
ftlilllHr
0.100
'12'

Figure 5.9: The effect offeed flow rate on Figure 8.5: Effect of Reynolds number on
acetaldehyde production rate effectiveness factor

251
For the low Reynolds numbers in Figure 8.5, the format of equation (8.16) is best suited to
determine the mass transfer coefficient as a function of feed rate, but C 1 is unknown.
Equation (8.19) is also valid, but in this case Cs is unknown and kc cannot be determined.

D
k ,=C~R
I e (8.16)
Dp

(8.19)

(8.29)

€b was taken as 0.4

Pb was taken as 600 kg/m3

In the region of strong interphase mass transfer resistance, CA.surf «< CA.b. As the Reynolds
number increases, the surface concentration will start increasing. An assumption was made
that CA,,,," «< CA,b in the region where TJ < 0.8. Equation (8.19) could be solved at different
temperatures and different Reynolds numbers. In Figure 8.6, the constant in equation (8.16)
is plotted against inverse Reynolds number at different temperatures.

0.045 0.02

0.04

0.035
7A--=-....-::~
~ - 0,016

0.03
..... 111"'0'---, 0.012
0.025
U T = 200 ·Cto 3OO·C U T =200 ·C to 300 ·C
0.02
0.'"
0.015

0,01 0,004

0.005

0
0 20 40 60 60 100 120
0
, 10 15 ~ ~ ~ ~ ~ ~

Inverse Re Inwrs.Ra"'1

Figure 8.6: C l vs. inverse Re at different Figure 8.7: C l vs. inverse Reo. 8l at different
temperatures temperatures

252
The value of C I is not supposed to change when the temperature and/or the feed flow changes
(C I is a constant). Figure 8.6 indicates that CI was not constant at different inverse Reynolds
numbers. A much better correlation was obtained when C I was plotted against the inverse of
the Reynolds number to the power of 0.81 (Figure 8.7). Equation (8.16) reduces to:

(8.30)

8.3.4.2. 2-Butanol dehydrogenation

For the dehydrogenation of 2-butanol, the reaction rate did not increase with an increase in
feed flow rate (see F!gure 5.10). Therefore, it could be concluded that the interphase mass
transfer resistance was small compared to the reaction rate. The effectiveness factor for 2-
butanol dehydrogenation could be taken as I with the reaction rate at bulk conditions
equalling the surface reaction rate.

8.4. EXPERIMENTAL DATA FOR A PLUG FLOW REACTOR

Both the ethanol and 2-butanol dehydrogenation reactions were studied at different
temperatures and different feed rates. Plug flow reactor data is listed in Appendix Fl and
F2.

8.4.1. ETHANOL DEHYDROGENATION IN A PLUG FLOW REACTOR

Figures 8.8 to 8.10 compare equilibrium ethanol conversion to acetaldehyde at different
temperatures with results obtained when the reactor was operated as a plug flow reactor.
Data was taken from Appendix Fl. Equilibrium values were calculated as described in
section 5.3.1. At all temperatures the ethanol conversion to acetaldehyde declined with an
increase in the feed flow rate. This was consistent with the theory discussed in Figure 8.1.
The conversion of ethanol to acetaldehyde was lower than the equilibrium conversion. At
low feed rates the total ethanol conversion was higher than the equilibrium conversion,
because the selectivity towards acetaldehyde production was less than 100%. Some of the
ethanol was converted to other products, which increased overall ethanol conversion.

253
 34 ....••.........•••••.•.•.•.••..•...•.•...•..•.•.•...•....••.•.....•.•....•.•..•. _ ......••.• 50 ............................................................... __ •.....•......•......•...•..

~t_--------~.~---------------

"
... . ... . ... .•. . . . -.......• .. t_------~.--------~--------­
•
1
30

~~t_--------~~---------------
•
8 •
f ,. t1rI=:H;::;...:;:...=""'==:=~="=••::.~===:;---------­
i 24 H .EthII\OI conwnlonto .~ •
• ~lIbrium cortYWIion to -mldahydll

~~============~---------
~~--------------------------~ O.OE1-00 3.0E-05 8.0E-05 1.2E-D4 1.5E-04
O.OE+OO 2.0E-D5 4.0E-D5
F(moIIs) F(moVs)

Figure 8.8: Plug flow reactor data for Figure 8.9: Plug flow reactor data for
ethanol dehydrogenation at 250°C ethanol dehydrogenation at 275°C

eo 100

" •
j
~"
•
8
-_ .. _. - --.....• -_ .. _ .. _ ... -_
•
..
•0 ..
'I! •
! .. r-- • TalaIlIIh.,GI con......lon
• Btl.,'" ~onto~ya.
•

..
O.OE+OO
.t.~llbrIumCQ"""""tJI'Ito~

2.0E-0!5 4.0E·D5 8.0E·05 S.OE-O!I '.1JE.04 4.005-05 8.005-05 1.205-04 1.606-04

F(moVs) F(moUs)

Figure 8.10: Plug flow reactor data for Figure 8.11: Selectivity towards acetaldehyde
ethanol dehydrogenation at 300°C production for a plug flow reactor

Figure 8.11 shows selectivity towards acetaldehyde production at different temperatures and
feed rates. The selectivity dropped at the lower feed rates as more unwanted products were
formed. Selectivity was less dependent on temperatures than on reaction time. Side products
were formed due to the reaction between acetaldehyde and ethanol. Acetic acid ethyl ester
was one of the main by products. Further products that formed were due to acetal and
hemiacetal formation (Streitwieser et aI., 1992). The acetal, 1,1'-diethoxyethane and the
dehydrated acetal, ethoxyethene were present in small quantities in most experiments.
Butanals were also present in low concentrations. These products were similar to those
observed by Raich and Foley (1998).

254
At higher feed rates or shorter catalyst contact times, the reaction time was too short for the
C4 by products to form and the selectivity towards acetaldehyde increased. Under
differential conditions, the selectivity of ethanol dehydrogenation towards acetaldehyde was
close to 100"10 (Chapter 5).

Reynolds numbers were determined at each temperature to determine whether the

experiments were in the mass transfer limited regime. The Reynolds number is dependent on
the reaction conversion and was calculated at different conversions to obtain the boundaries
of the Reynolds number. The data for temperatures from 200°C to 300 °C is shown in
Figures 8.12 to 8.14.

0.0300 O.OISOD •• ---.-.--••• -----------------•••• -------••••• -----------... ----------•• ____________ •_______ .

::s·~-.--~~ --._--....- 0.0300+--------------

Ethanol conv.raton -100, 0, 10%
---
&! 0.0150 +----::::----:-----:----:::::-:c-::=--------I from top to bottom
Ethanol c:onv«Slon ·'00, 0, 50%
from top to bottam •." " " ' + - - - - - - - - - - - - - -
0.0100+------------------1
•.0 1 0 0 + - - - - - - - - - - - - - -
0.""'+------------------1
o.""'l--_ _-_--_--_-~ •.0000+--_ _- _ - - _ - - _ - _
20.
200
'" '" 300 200
'" Trcl
300

Figure 8.12: Re as a function oftemperature Figure 8.13: Re as a function oftemperature

and conversion (feed rate = 4.77*10-' moVs) and conversion (feed rate =9.54*10-' moVs)

0.0800

0.0700t====:~~==
F=-=
11.0Il00
----- - ------
0.0500+---------------=
~ O.IMOO +---IEI'"h"'an"'oIr;;
.....
;;;;;;;;.iiiiOftii•.,.~I.""
...... ~._
.• "' .
, .,.----
= _____
+-___---""''":::.m;;;:.DP'''to'''b=......
D.....,

D."""'+--------------
0.0100+--------------
o.ooool--_--_--_--_-_
200 280 300

Figure 8.14: Re as a function of temperature and conversion (feed rate = 1.43*10-4 moVs)

255
At all tested feed flow rates, the Reynolds numbers were less than 0.07 and thus in the region
of strong interphase mass transfer resistance according to Figure 8.5.

8.4.2. 2-BUTANOL DEHYDROGENATION IN A PLUG FLOW REACTOR

In Figures 8.15 to 8.17 the equilibrium 2-butanol conversion data is compared to measured
data for a plug flow reactor. Data was taken from Appendix F2.

eo .---------------------------------------------------.-----------------_.-.---- 75 ._._---_._._---_ .. _--.-. -----...------------------------ --._--------------..... _------_._ .

•• ~ _. M ~ __________ • _. -.-A
~t_--~----_r-----------------
a 7Or---~--------------~----y_-
•
II~t_--------------------------- j"r----"----"-----"----.-----------

• Tat:ll2-butlnol CIOIlWrIIon
.2-Bubrlol conVlQlon 10 MEK
" r 1r=;.T;...;......
•
55
;;;.;~;n.;.!III;oo=::;-------------
.2..eutmoI CIIInWI-.Ian to MEK
" EquHlbriwn COIIYIIrSkrI to MEK
.. E~ CCIIIWI'SICf'I to MEK

35L'=::=:~==~~~ 50 +-----__--__----__----__ ----~

0.0&00 2l1E-05 ...-os 0..... O.OEtOO 2.CE-05 •. IE-05 e.Df-05 a,CEo05 1./JE-04
F(molls) F(molls)

Figure 8.15: Plug flow reactor data for Figure 8.16: Plug flow reactor data for
2-butanol dehydrogenation at 190°C 2-butanol dehydrogenation at 215°C

.. 102 ._--_ .. _... --.. -----.------------------------------------------------------------------1

!
82
" ''''' • • •
" .............
............ _-_ !
iao
c
8
• a
.i
~
> • " •

r
•
,.
•
•
.. Tahll2-butw101 ~on
.2-1kitmoI currvw-!lon to MEK
.. ~IbI'I\rnCCll'l\ltlnlontoMEK
:l
.. " !
I
.T= 190'C
.T=215"C
... T"240"C

9.
"O.OE+OO 2JlE-OS ""DE-OS &'oE-OS 0.06-05 1.0E-G4 '.2!;<4 ,...... O_OOE+OO 4_00E-05 B.OOE-05 1.Z0E-04 1.60E-04

F(molls) F (maUs)

Figure 8.17: Plug flow reactor data for Figure 8.18: Selectivity towards MEK
2-butanol dehydrogenation at 240°C production for a plug flow reactor

The measured conversions at 190°C (Figure 8.15) and at 215 ac (Figure 8.16) declined with
an increase in the feed flow rate. The reaction time became too short at the higher flow rates

256
for full conversion to take place. At 240°C, the reaction rate was very fast compared to the
rate at the lower temperatures. Even at the maximum feed flow rates tested, the 2-butanol
conversion was still close to that of the equilibrium value and did not decline significantly.
Selectivity for this reaction was very high. Figure 8.18 indicates that the selectivity was
more than 96% for all experiments conducted. The main by-product was 3-octanol.

8.4.3. SUMMARY OF MODEL FOR ETHANOL DEHYDROGENATION

Physical properties of the ethanol-acetaldehyde-hydrogen mixture were determined as
described in section 8.3.4.1. The viscosity of the mixture is dependent on both the
temperature and the composition of the mixture. The pressure was kept constant at one
atmosphere. The effectiveness factor (Figure 8.5) was expressed as a function of Reynolds
number at different temperatures. The following equations were obtained:

At 300°C:
11 = -224.00*Re3 - 69.20*Re2 + 18.60*Re + 0.053 for 0.01 < Re:S 0.08 (8.31)
11 = 1 for Re > 0.08

At 275°C:
11 = -241.18*Re 3 - 36.65*Re2 + 15.89*Re + 0.039 for 0.01 < Re:S 0.085 (8.32)
11 = 1 for Re > 0.085

At 250°C:
11 = 363.16*Re3 - 189.62*Re2 + 26.34*Re - 0.080 for 0.01 < Re:S 0.08 (8.33)
11 = 1 for Re > 0.08

It should be emphasised that these effectiveness factors will only be true for small particles
without intraparticle mass transfer resistance. For larger particles, equation (8.27) must be
solved using the kc value determined in equation (8.30). The Reynolds number is:

(8.34)

257
u = 8.314T(FEt.t + FH"t + FA"t + FN"t)
(8.35)
Pt(nR~)

_ 46PEt ,t + 44PA"t +2PH "t + 28PN "t

(8.36)
Pm - 1000*(8.314)T

with

p, = total pressure [Pa]

T = reaction temperature [K]

On the tube side:

For ethanol

with (8.37)

(8.38)

(8.39)

For hydrogen:

with (8.40)

(8.41)

258
For acetaldehyde:

with (8.42)

(8.43)

For nitrogen:

(8.44)

On the shell side:

For nitrogen sweep gas:

dFN", =2nR L P [P -P
d( ~o) m 0 m,N, N,,' N,,!
(8.45)

For ethanol:

(8.46)

For hydrogen:

d(dF 2
H "
~)
2nR LP
m 0 m.H,
[
P 1
-P
H,.! H 2 .,
(8.47)

For acetaldehyde:

(8.48)

259
The initial conditions were:
AtL=O: FE.,. = FE',t(O) FE." = 0

FH ,.! = 0 FH , = 0
"
F Ac ,. = 0 FAc ,. = 0

FN,.t = 0

Finally, the partial pressures need to be expressed in tenns of flow rates:

On the tube side:

(8.49)

(8.50)

(8.51)

(8.52)

On the shell side:

(8.53)

(8.54)

P =p FH", ] (8.55)
H,.. !,' [ F
Et,s
+F
Ac,s
+F
H1,s
+F
N 1,5

P =p FN". ] (8.56)
N,.. !,' [ F
Et,s
+F
Ac.s
+F
H 2 .S
+F
N 2 ,s

260
To solve the model, kinetic coefficients and gas permeances must be known. Membrane
(N8x) was used for catalytic membrane reactor experiments. Permeance properties of the
membrane are listed in Appendix D2. Measured reaction rate coefficients will be used
instead of fitted values (from Chapter 5) at 250 °e, 275 °e and 300 °e, respectively. Table
8.3 lists reaction rate coefficients. Reaction rate coefficients were taken from Table 5.5 and
changed to standard units i.e. [Pa], [Pa'!] and [mol/kg.cat.s.Pa]. The permeance data is listed
in Table 8.4 and further data necessary to solve the model is given in Table 8.5.

Table 8.3: Ethanol reaction rate parameters

T K. q k KEt KHydrogen K Ac

(oq Pa mollkg cat.s.Pa Pa'l Pa'l Pa'!

250 10918 5.19E-07 5.14E-06 3.24E-06 6.94E-05
275 24083 1.27E-06 4.32E-06 2.91E-06 6.57E-05
300 56478 1.89E-06 5.46E-06 3.68E-06 8.18E-05

Table 8.4: Permeance data for N8x at 250,275 and 300 °e

T (oq Hz permeance Nz permeance Ethanol Acetaldehyde
(Ilmollml.Pa.s) (nmollm2.Pa.s) permeance permeance
(nmollm1.Pa.s) (nmollm1.Pa.s)
250 3.37 22.71 5.68 5.68
275 3.78 21.29 5.32 5.32
300 4.21 20.15 5.04 5.04

Table 8.5: Parameters for solving ethanol dehydrogenation model

Rm [m] 3.5*10'>
Lo [m] 0.18
V [m] 6.93E-06
Pb [kglm'] 430
Eb 0.4
m [kg] 0.003

261
8.4.4. SUMMARY OF MODEL FOR 2-BUTANOL DEHYDROGENATION
The model fOf 2-butanol dehydrogenation is similar to that of the ethanol dehydrogenation
model. Figure 5.10 shows that the dehydrogenation rate changes very little with feed flow
rate, indicating that the interphase mass transfer resistance is small compared to the reaction
rate. The effectiveness factor can be taken as 1 and Reynolds numbers do not have to be
calculated. The model can be expressed as:

On the tube side:

For 2-butanol

dF'Bu,
d(~') =
'v[·
- f
Pb 'Bu"b77 - Rm2 J 'Bu' ] with (8.57)

Lo .

(8.58)

(8.59)

For hydrogen:

with (8.60)

(8.61)

ForMEK:

(8.62)

262
 (8.63)

For nitrogen:

(8.64)

On the shell side:

For nitrogen sweep gas:

(8.65)

For 2-butanol:

(8.66)

For hydrogen:

d(elFtH",J=2nR L P
m 0 m,H,
[
H"I
l
P -P
H",
(8.67)

ForMEK:

(8.68)

The initial conditions were:

AtL=O: F2Bul,1 = F2Bul,I(O) FBut,s = 0

FH "I =0

263
 FMEK,t = 0 FMEK,t = 0
FN , ,I =0

Finally, the partial pressures needed to be expressed in terms of flow rates:

On the tube side:

(8.69)

(8.70)

(8.71)

P =P FN"I ] (8.72)
N"t ![ (F'BUI,! + FMEK,! + FH,,') + (FN"s(O) - FN,,,)

On the shell side:

P = Pt,s [ F F2But ,s ] (8.73)

2But,s +F +F +F
2But,s MEK,s H2,s Nz,s

pMEK,s = p t,s [ FMEK,s ] (8.74)

F2Bu!" + FMEK" + FH", + FN",

(8.75)

P
Nz's
=P
t,s
FN,,' ] (8.76)
[ F2But" + FMEK" + FH", + FN",

Table 8.6 lists reaction rate coefficients. Values were taken from Table 5.7 and changed to
standard units i,e. [Pal. [Pa· l ] and [mol/kg.cat.s.Pa]. Membrane (N8x) was used for
experiments. The permeances were determined from data in Appendix D2. Values were

264
estimated at 190°C, 215 °C and 240°C, respectively, by fitting an equation through the
hydrogen and nitrogen permeance data for membrane (N8x) .

Table 8.6: Reaction rate coefficients for 2-butanol dehydrogenation

T K. q k KR
(0C) Pa mol/kg cat.s.Pa Pa-1
190 73132 9.42E-08 1.17E-04
220 161176 1. 94E-07 7.02E-05
250 330120 3.89E-07 5.51E-05

Table 8.7: Permeance data for N8x at 190, 215 and 240°C
T (0C) H2 permeance N2 permeance 2-Butanol MEK permeance
(Ilmollm 2.Pa.s) (nmollm2.Pa.s) permeance (nmollm2.Pa.s)
(nmollm 2.Pa.s)
190 2.40 26.61 5.25 5.25
215 2.81 24.80 4.89 4.89
240 3.22 23.21 4.57 4.57

The other parameters necessary to solve the model are listed in Table 8.8.

Table 8.8: Parameters for solving 2-butanol dehydrogenation model

Rm [m] 3.5*10"
Lo [m] 0.20
V [m] 6.93E-06
Pb [kg/mJ ) 430
tb 0.4
m [kg] 0.0035

265
8.5. SELECTIVITY ANALYSIS OF THE PROCESS MODEL
The standard values are the measured reaction rate and membrane parameters for ethanol
dehydrogenation at 275°C (Table 8.9).

Table 8.9: Parameters for ethanol dehydrogenation at 275 °C

Reaction T (0C) 275
Ethanol feed flow rate (mIlh) 10
(molls) 4.771 *1 0.5

N2 sweep to ethanol feed molar ratio I

Shell pressure [Pa] 100000
Tube pressure [Pa] 100000

Reactor length [m] 0.18

Rm [m] 0.0035
J
Pb [kglni ] 430
m [kg] 0.0030
eb 0.4

k [moIlkg cat.s.Pa] 1.27*10'0

Ke'hanol [Pa"] 4.32*10'u

Kacetaldehyde [Pa"] 2.91 *10'0

Khydrogen [pa"] 6.57*10'"

Ke q [Pal 24080
1
"
Pm.hydrogcn [J.UlloIlm" .Pa.s] 3.78
Pm,nitrogen [nmoIlm".Pa.s] 21.29

Pm,ethanol [nmoIlm" .Pa.s] 5.32

Pm,acetaldehyde [nmoIlm" .Pa.s] 5.32

266
The performance of the membrane reactor can be judged by several factors. The most
important factors are the exit alcohol conversion, the fraction of ethanol feed lost through the
membrane, the fraction of produced hydrogen permeated to the shell side and the purity of
the hydrogen on the shell side. Process conditions, reaction parameters and membrane
parameters that yield high alcohol conversion and pure hydrogen on the shell side, represents
optimum conditions. The effect of each parameter on the dehydrogenation of ethanol at
275°C is presented graphically in Appendix G. It should be kept in mind that the
equilibrium ethanol conversion for a plug flow reactor is 40% at 275°C and 1 atmosphere
total pressure.

Model equations (8.37) to (8.56) were solved numerically with Euler's method. A very small
step length of 0.001 (LILo)was employed to obtain results of high accuracy.

8.5.1. THE EFFECT OF k'-VALVES ON MODEL PERFORMANCE

The reaction rate k' -value is one of the most important parameters for manipulating the
overall ethanol conversion. Overall conversion goes from 27% to 54% when the k' -value
increases from 0.5 to 8 times the standard value (Appendix G.l). This represents a very
large increase in conversion. Up to a 2 times increase in k' -value, the conversion increases
sharply, but thereafter there is little gain in conversion for higher k' -values. The amount of
ethanol permeated to the shell side remains low and drops when the reaction rate increases
(larger k' -value). For twice the k' -value and larger, the hydrogen purity on the shell side is
above 86% and changes little with larger k' -values. The tube side composition also remains
constant at the larger k' -values.

In the very first part of the reactor, the hydrogen production rate is much faster than the
permeance rate. The driving force shoots up from zero to a maximum value. The faster the
rate (larger the k' -value), the higher the maximum hydrogen driving force. The hydrogen
driving force then drops quickly to very low values further down the axis of the membrane.
At different k' -values there is almost no change in the percentage of produced hydrogen that
permeates to the shell side. The percentage of produced hydrogen that permeates to the shell
side, increases sharply in the first part of the reactor and then declines slowly as back

267
diffusion takes place. The ethanol flow rate on the tube side also drops quicker along the
membrane's length as the k'-value increases.

An optimum k' -value would be about four times the standard rate. Higher values will not
improve the ethanol conversion significantly.

8.5.1.1. Back diffusion

The term back diffusion will be used to describe the situation where hydrogen no longer
permeates from the tube side to the shell side. Permeation is reversed and hydrogen
permeates from the shell side to the tube side. It happens when the hydrogen driving force
becomes negative. The reason why it happens is as follows.

The sweep gas permeance (nitrogen) is much higher than the permeances for ethanol and
acetaldehyde. A fraction of the nitrogen sweep gas will permeate from the shell side to the
tube side and this fraction is much larger than the amount of ethanol and acetaldehyde that
will permeate from the tube side to the shell side. The nitrogen partial pressure will drop on
the shell side and increase on the tube side. Once the produced hydrogen reaches a maximum
amount permeated, no more hydrogen goes from the tube side to the shell side and back
diffusion of hydrogen will commence. The hydrogen partial pressure on the shell side will
increase due to a decline in nitrogen partial pressure, which is larger than the combined
increase in ethanol and acetaldehyde partial pressures on the shell side. Similarly, the
hydrogen partial pressure on the tube side will decrease due to an increase in nitrogen partial
pressure. The hydrogen driving force is reversed and hydrogen will permeate back from the
shell side to the tube side.

8.5.2. THE EFFECT OF K atet ON MODEL PERFORMANCE

Kacet is the acetaldehyde adsorption coefficient and it affects the reaction rate. The higher the
value, the slower the reaction rate. The value of K acet plays a significant role in the overall
ethanol conversion (see Appendix G.2). A twenty fold decrease in Kacet improves ethanol
conversion from 44% to 53%. The measured acetaldehyde adsorption coefficient was about
twenty times larger than the other adsorption coefficients. The lost ethanol feed declines with

268
a lower Kacer, but remains at acceptable levels for all Kacet-values. The variance in ethanol
conversion for changes in K acet is much smaller than the variance when k is changed. For
Kaceevalues ofless than five times the standard value, the improvement in ethanol conversion
becomes very small for a further decline in Kacet.

Hydrogen purity on the shell side varies between 85% and 88%. On the tube side the
nitrogen content is constant at 10%. The initial hydrogen driving force is larger for small
Kacet-values. The driving force either goes to very low values or to below zero along the axis
of the membrane. The lack of a strong driving force results in a low hydrogen percentage
permeated to the shell side. The value is 45% and changes little with changes in the value of
K acet . This result is similar to that obtained for changing the k' -value.

An optimum Kacecvalue would be at least 5 times smaller than the standard value.

8.5.3. THE EFFECT OF THE MEMBRANE SELECTIVITY ON MODEL PERFOR-

MANCE
This situation is where there is a constant hydrogen permeance, but changing permeances for
all the other species to obtain a changing selectivity. The selectivity has a negligible effect
on reaction conversion at standard conditions (see Appendix G.3). Conversion remains
between 44% and 45% for a variety of selectivities. At different conditions, this might not be
true and will be investigated later. Poor selectivity, however, has a very negative effect on
the ethanol feed lost through the membrane and the purity of hydrogen on the shell side. For
a hydrogen to nitrogen selectivity of less than 150, the hydrogen purity on the shell side drops
sharply from just over 80% (at selectivity = 150) to about 30% (at selectivity = 10).

The nitrogen fraction on the tube side increases from 5% at a selectivity of 320 to over 40%
at a selectivity of 10. With such an increase in nitrogen content on the tube side, a significant
amount of back diffusion can be expected. This is indeed the case and the hydrogen driving
force becomes negative at low dimensionless length values for low selectivities. The low
percentage of produced hydrogen that permeates to the shell side, indicates very strong back

269
diffusion. At a selectivity of 10, only 20% of the produced hydrogen permeates to the shell
side and this percentage increases to 47% at a selectivity of320.

A selectivity of at least 150 must be used to ensure that not more than 3% of the ethanol feed
is lost through the defects in the Pd-Ag film.

8.5.4. THE EFFECT OF THE REACTION EFFECTIVENESS FACTOR ON MODEL

PERFORMANCE
The effectiveness factor is a measure of the decrease in reaction rate due to interphase and
intraparticie mass transfer resistance. The factor varies between 0 and I, where 1 indicates
zero mass transfer resistance. A decrease in the effectiveness factor will cause the same
effects as a decrease in the k' -value, since both will reduce the overall reaction rate. Figures
in Appendix G.4 are similar to those for k' -values (Appendix G.!) in the region where the
k' -values are less than the standard k' -value.

The best results will be obtained with the maximum value of the effectiveness factor, i.e.
when the effectiveness factor approaches one.

8.5.5. THE EFFECT OF THE ETHANOL FEED RATE ON MODEL PERFOR-

MANCE
The ethanol conversion decreases with an increase in ethanol feed rate, because the contact
time decreases and the reaction cannot go to completion along the length of the membrane.
The conversion declines from 50% to 34% for flow rates of 0.5 to 2.5 times the standard
value (Appendix G.5). The higher the feed rate, the smaller the fraction of feed lost through
the membrane and the better the purity of the hydrogen on the shell side.

The ethanol fraction at the tube side exit increases with an increase in feed flow rate due to a
decrease in conversion. At the higher flow rates, a small hydrogen driving force remains
over the full length of the membrane and a maximum percentage (47%) of produced
hydrogen permeates to the shell side. At a low flow rate (0.5 multiple), back diffusion takes

270
place in the second half of the reactor. The hydrogen driving force shifts along the axis of the
membrane when the feed rate increases.

The value of the feed rate should be determined by all other parameters. The quality of the
membrane and the speed of the reaction rate will determine which feed rate is best suited for
the model. For example, if the reaction rate is very slow, it is of no use to employ a very fast
feed rate, since the benefits of the membrane reactor will be cancelled.

8.5.6. THE EFFECT OF THE SWEEP GAS ON MODEL PERFORMANCE

The sweep gas flow rate is a very important parameter for the membrane reactor model (see
Appendix G.6). At a constant temperature only the sweep gas rate and the feed rate can be
changed. Reaction rate parameters and membrane parameters are fixed. The sweep gas is
necessary to create the hydrogen driving force. Alternatively, a pressure difference between
the tube and the shell side can be used. Modelling calculations in the present study will only
focus on sweep gas experiments, since reaction rates were determined at atmospheric
pressure and not higher pressures.

Ethanol conversion increases from 42% to almost 50% for sweep gas to feed ratios of 0.25 to
8. The ethanol feed lost through the membrane increases slightly, because of a higher ethanol
driving force, but it remains at acceptably low values. A hydrogen purity of above 90% is
obtained on the shell side for sweep to feed ratios greater than 4. There is a sharp decline in
hydrogen content (from 25% to 5%) on the tube side with an increase in sweep rate. The
acetaldehyde fraction increases on the tube side as more ethanol is converted and the nitrogen
fraction increases as more nitrogen permeates through the membrane at higher sweep rates.

A higher sweep rate reduces the hydrogen partial pressure on the shell side. The result is a
higher hydrogen driving force and a higher percentage of produced hydrogen that permeates
the membrane. A high sweep rate is very effective for extracting both pure, and a very high
percentage of, produced hydrogen. The percentage of produced hydrogen permeated,
increases from less than 15% at a sweep to feed rate of 0.25 to over 85% for a sweep to feed
rate of8.

271
A sweep to feed ratio of at least 4 should be used, but this depends to a large extent on the
feed rate and must be chosen in conjunction with the feed rate.

8.5.7. THE EFFECT OF THE H2 PERMEANCE AT CONSTANT SELECTIVITY ON

MODEL PERFORMANCE
In this scenario the permeances of other species increase with the permeance of hydrogen to
maintain a constant selectivity. The effect of hydrogen permeance on ethanol conversion is
negligible. The conversion changes by little more than I % over a wide range of hydrogen
permeances (Appendix G.7). With an increase in hydrogen permeance, more ethanol is lost
through the membrane as the ethanol permeance also increases. The hydrogen fraction on the
shell side decreases sharply and the nitrogen fraction on the tube side increases sharply with
an increase in hydrogen permeance. The main reason is not so much the increase in
hydrogen permeance, but the increase in the permeances of other species to maintain the
constant selectivity. At a very low hydrogen permeance (0.25 multiple), the hydrogen
driving force remains high over the full length of the membrane. In this case, the percentage
of hydrogen permeated, increases steadily along the length of the membrane. For higher
hydrogen permeances (2 and 4 times the standard rate), back diffusion occurs in the latter part
of the membrane.

When comparing the data in Appendix G.3 and Appendix G.7, it is clear that the membrane
selectivity plays a more important role in the membrane reactor's performance than the
hydrogen permeance does. This statement is true under standard conditions and in the ranges
that each parameter was investigated. With the standard selectivity, the hydrogen permeance
should not move outside the range of 0.5 to 1.5 times that of the standard permeance.

8.5.8. THE EFFECT OF THE H2 PERMEANCE AT VARYING SELECTIVITY ON

MODEL PERFORMANCE
For this set of modelling conditions the nitrogen, ethanol and acetaldehyde permeances are
kept constant at the standard values. The hydrogen permeance is varied and the selectivity
changes as the hydrogen permeance changes. The overall conversion remains close to 44%

272
(Appendix G.8). The overall ethanol conversion in Appendix G.8 is similar to that in
Appendix G.7, but there is very little change in ethanol feed lost through the membrane. The
hydrogen composition on the shell side exit remains constant at just above 88%. The
composition of the tube side exit mixture also varies little with changes in the hydrogen
permeance. The hydrogen driving forces in Appendix G.8 are similar to those in Appendix
G.7, but less back diffusion of hydrogen occurs, since less nitrogen permeates from the shell
side to the tube side.

The percentage of produced hydrogen that permeates from the tube side to the shell side
starts declining at hydrogen permeances of less than 0.5 times the standard value. At lower
hydrogen permeances, the reactor length is too short to reach equilibrium for hydrogen
permeance. The result is a decline in hydrogen permeance as compared to when equilibrium
is reached.

At standard selectivity, the hydrogen permeance should be at least one half or more of the
standard hydrogen permeance for optimum membrane performance.

8.5.9. MODEL PERFORMANCE UNDER NON-STANDARD CONDITIONS

Two situations are presented, in Appendix G.9 and G.IO, where the reaction rate is four
times the standard rate and the sweep gas to feed ratio is four times higher than the standard
ratio, respectively. In Appendix G.9 the hydrogen permeance is varried, while permeances
for the other species remain constant (varying selectivity). In Appendix G.! 0 the hydrogen
permeance as well as the permeances for the other species are changed (constant selectivity).
The results are very similar for the two scenarios except for the fact that less ethanol feed is
lost to the shell side when selectivity increases with increasing hydrogen permeance
(Appendix G.9). In this case, the hydrogen purity on the shell side is above 93% for
hydrogen permeances of more than 0.5 times the standard value. The hydrogen driving force
is similar for the two sets of data and in both cases above 70% of produced hydrogen
permeates through the membrane for hydrogen permeances of 0.5 times the standard value,
and more.

273
In both cases the ethanol conversIOn reaches a mronmum of 64% when the hydrogen
permeance is increased. The gain in ethanol conversion is small for hydrogen permeances
larger that 0.5 times the standard value. At smaller values the ethanol conversion will drop
sharply towards the equilibrium value or lower in the case of higher feed rates. The
equilibrium ethanol conversion at 275 °C is 40%. Hence the membrane reactor offers a
significant improvement in ethanol conversion. This improvement is more than 50%.

8.5.10. OPTIMISING MEMBRANE PERFORMANCE ON MODEL PERFOR-

MANCE
Two sets of conditions were compared to optimise the membrane performance. The sets
were defined as:
• Standard conditions, as in Appendix G, and
• Two times standard k'-value (k = 2.54*10-6 mollkg cat.s.Pa); 0.5 times standard Kact -
value (Kact = 3.29*10-5 Pa- I); 11 = 0.9; Hz permeance = 3 Ilmol/mz.Pa.s; Hz:N z selectivity =
150. Other conditions are similar to the standard conditions.

The second set of conditions represents a slight improvement in catalyst activity. Membrane
parameters are similar to those of the optimised membrane. For the standard conditions, the
ethanol conversion is 44.5% and the ethanol feed lost through the membrane is 2.33%. For
the second set of conditions, the ethanol conversion is given as a function of sweep gas to
feed ratio in Figure 8.19. Conversion increases sharply with an increase in sweep gas flow
rate. The membrane reactor is very effective if a large hydrogen driving force across the
membrane can be maintained along the length of the membrane.

To optimise the membrane reactor's performance, two main conditions must be satisfied.
1. The dehydrogenation rate and the hydrogen permeance through the membrane should be
similar.
2. A high hydrogen driving force must be maintained along the full length of the membrane.

274
 80 -------------------------------.. ---------------------_._----------.-.-------...---------------------.-------.. _- 2

75 =:__: :_~_~_~__~__:-:_j I
~rr-<O::--'i·::::--:-:--::::---:;;...;::-:::-~:::--:-::--:::-~--:.:-::-;:::_;::;_::;::__~_~__=__::=__ 1.6

i
70
J
i I
: /

65 I 1.2

i..u'ot~~----------~============~__j
55
fI I ~COn I .
I -0-
....lon(IoII ....'
LosttMd(rlght ....,
0.8 ~
.;.
i 50 I
0.' ~
45 t---------------------------1 ..
40~- __--~-__- __- - _ -__- __- _ 4 0
o 4 8 12 16 20 24 28 32
Sweep gas to standard feed molar ratio

Figure 8.19: Effect of sweep gas to standard feed molar ratio on ethanol conversion and
ethanol losses

For the membrane parameters and reaction rate parameters measured in this study, the
hydrogen permeance was much faster than the reaction rate for the ethanol dehydrogenation
reaction. A more active catalyst will improve the overall ethanol conversion significantly,
since the ethanol conversion in the membrane reactor was mainly limited by the reaction rate.
The hydrogen driving force may be improved by using a vacuum on the shell side or a
positive feed pressure on the tube side instead of a sweep gas. The larger the pressure
difference between the shell and the tube sides, the better the overall reaction conversion will
be if the reaction rate stays constant (Figure 8.19). Higher feed pressures will, however,
have a negative impact on the reaction rate and cause a decline in the reaction rate.

8.6. SUMMARY
The effects of feed flow rate on reaction conversion were discussed and the low flow rates
that have been investigated in membrane reactors by previous authors, mentioned.
Membrane permeance data for ethanol was presented. It was found that the nitrogen to
ethanol selectivity of the Pd film membrane was about four to one.

A process model was developed for the catalytic membrane reactor. All model conditions
were covered, including the effects of pressure drop along the packed membrane and

275
interphase and intraparticle mass transfer resistance. The reaction temperature was assumed
to be constant and there were no radial concentration profiles in the catalysts bed. Equations
were formulated to calculate the mass transfer coefficient for ethanol dehydrogenation at
different Reynolds numbers. Experimental plug flow reactor data was compared to
equilibrium data at different feed flow rates. At low flow rates, the alcohol conversion was
similar to the equilibrium conversion at the tested temperatures. Flow rates were determined
where the conversion started decreasing to below the equilibrium values. For the catalytic
membrane reactor to be applicable, an improvement in reaction conversion must be obtained
in the higher flow rate region.

A thorough sensitivity analysis on catalyst parameters, membrane parameters and process

conditions was conducted. The main findings were:
• The reaction rate parameters (k' -value; Kacet and 11) have a very significant effect on the
overall ethanol conversion. A high k-value, a low K.cet-value and/or a high effectiveness
factor will increase the reaction rate and improve the reaction conversion. The reaction
rate parameters have no effect on the percentage of produced hydrogen that exits the shell
side. The percentage of produced hydrogen that permeates to the shell side remains
similar for different values of each parameter. The hydrogen driving forces have similar
profiles and the maximum value increases with reaction rate. The reaction rate
parameters have a small effect on the amount of ethanol feed that is lost through the
membrane and the purity of hydrogen on the shell side. When the reaction rate is slow,
the ethanol driving force will be larger, resulting in more ethanol permeance through the
membrane and a decline in hydrogen purity on the shell side.
• Membrane selectivity (IIz;N2 ) should be at least 150 to prevent large losses of ethanol
feed through the membrane and a low hydrogen purity on the shell side. At low
selectivities, hydrogen back diffusion occurs and the amount of produced hydrogen that
permeates to the shell side decreases sharply. It is more important to have a high
selectivity than a high hydrogen permeance. Very thin films with higher hydrogen
permeance, but poorer selectivity should be avoided. There is a negligible improvement.
in ethanol conversion with higher membrane selectivity and/or higher hydrogen
permeance at low sweep rates and low reaction rates. At higher reaction rates and higher

276
 sweep rates, the membrane parameters play very important roles and can result in large
improvements in alcohol conversion.
• The two process parameters, the feed rate and the sweep rate, must be chosen in
conjunction with one another. In both a plug flow reactor and a membrane reactor, the
alcohol conversion will drop with an increase in feed flow rate, because the reaction time
becomes shorter. The feed rate has a effect on the percentage of produced hydrogen that
permeates to the shell side. At low feed rate, back diffusion of hydrogen will occur. At
high feed rates, the purity of hydrogen on the shell side is high. The sweep to feed molar
ratio is very important for extracting hydrogen from the tube side to the shell side. At
high sweep rates, most of the produced hydrogen can be extracted from the reaction side
(tube side) to the shell side. The reaction equilibrium is shifted and the reaction
conversion increases.

277
 9. CATALYTIC MEMBRANE REACTOR DATA

The experimental procedures for catalytic membrane reactor testing were discussed in
section 3.4. The Pd-Ag membrane (N8x) was packed with the optimised catalyst. The
optimised catalyst consisted of 14.4 wt % copper on silica, with a particle fraction ranging
from 500 to 800 microns. The performance of the reactor with and without an inert sweep
gas was compared. The effect of the sweep gas flow rate on reaction conversion and
selectivity was studied at different temperatures. Results are listed in Appendix F.
Experimental data was compared to model predictions for both the dehydrogenation of
ethanol and the dehydrogenation on 2-butanol. All daia presented was at the exit conditions
of the reactor. Experiments conducted to optimise the catalyst composition (Chapter 4) were
also performed in a plug flow reactor, but the values in Chapter 4 were not directly
comparable to the plug flow reactor data in Chapter 9. In Chapter 4 a diluted feed (feed to
nitrogen molar ratio of 1 to 4) was employed, but in this chapter an undiluted feed was used.
Conversions for the plug flow reactor will thus be higher in Chapter 9 than in Chapter 4,
due to higher feed concentrations.

For ethanol dehydrogenation, two models were constructed. Model 1 used the measured
reaction rate and membrane parameters for the reaction at the specific temperatures. In
model 2, the reaction rate parameters were changed to better fit the experimental data. Model
2 does not have significant physical importance, but indicates how the parameters must be
altered for the model to better predict the experimental data. For 2-butanol dehydrogenation,
only one model was developed at every temperature. The model did not use measured kinetic
data, since the model then under predicted the experimental 2-butanol conversion
significantly. Instead the model used altered parameters to improve the prediction of
experimental values (similar to model 2 for ethanol dehydrogenation).

It should also be emphasised that the reaction rate models developed in Chapter 5 are
independent of the membrane reactor models in this chapter. This implies, for example, that
model 2 (ethanol dehydrogenation reaction) in this chapter does not use the reaction rate
parameters of model 2 in Chapter 5, but the reaction rate parameters defined in this chapter.

278
9.1. ETHANOL DEHYDROGENATION
The results of ethanol dehydrogenation in a membrane reactor are listed in Appendix Fl.
The results at each temperature will be discussed separately (sections 9.1.1 to 9.1.3). The
following definitions were mentioned in Chapter 4 and will also be used in this chapter.

total moles feed reacted

total feed conversion = - - - - - - - - - (4.1)
total moles feed fed

moles feed converted to product

product yield = - - - - - - - - - = - - - - - (4.2)
total moles feed fed

moles feed converted to product

product selectivity = ------------ (4.3)
total moles feed reacted

yield = selectivity x total feed conversion (4.4)

Standard parameters (measured reaction rate and membrane parameters) for modelling were
listed in Tables 8.3 to 8.5. Included in the standard parameters is an effectiveness factor of 1.

9.1.1. MEMBRANE REACTOR EXPERIMENTS AT 250°C

Figures 9.1 and 9.2 show the effects of sweep gas flow rate on exit ethanol conversion at two
different ethanol feed flow rates. At a lower flow rate of 5 mllh (Figure 9.1), the plug flow
reactor (where the sweep gas to feed ratio is 0) operated at the equilibrium conversion. The
overall ethanol conversion was slightly higher than the equilibrium conversion, because
ethanol was also consumed to produce other products besides acetaldehyde. The total ethanol
conversion increased with an increase in sweep gas flow rate, which is in line with the model
equations. At the higher flow rate of 10 mllh (Figure 9.2), the improvements in total ethanol
conversion for the membrane reactor were less than at the lower flow rates. The plug flow
reactor in Figure 9.2 yielded an ethanol conversion below the equilibrium value. The

279
membrane reactor provided a good increase in ethanol conversion even under the higher feed
flow rate conditions.

55 40 ------------ .. --.".-.... ---------- -...... --.,._---------" .-.-----.--............ ---.. ----.

~r_--------------------------.
-- - - .• 36
• - --
lfl.
~r_-----+--~-=-~
•.~---------------~.
.... ...•.. ,," ,..... •..... . ..• .,. • --- -
-.-
· .. -.•
~."~ ... ,__: . I
32
•.. . ..,
:.!; 40 •
, . •
~
'~'

•
., ......
••
8~~~~~----------------------
,:'~i··

+ .
28
8
..
"f=:=~~~~~~~ Bhlnoi to aCllblld!ltyda

H
• Tatalllthlnol ODIWIII'tIon
ett..01 tv.~

H
• Tut.!.tn.IoI oonv.llon
25 ." Ethtnol conv.-lIiIon (mcrMl1) -+. EthmolllOllVW$lon (m~ 2) 24 .,."' Elhnol con""ulon (modal 1) - ... EthlnollXln\oWtlon (model 2)
-Eq.IiibriLftl ecnveriecn -~lbrilm conv.illQ1

20~~~~:::=:::==:=~ t...=:::::~~::::::::=::=::::==:=='--, 20
o 2 3 456 7 o M 1 1~ 2 U 3 ~ 4
Sweep gas to teed molar ratio SWMp gas to fMd molar ratio

Figure 9.1: Ethanol conversion at 250°C vs. Figure 9.2: Ethanol conversion at 250°C vs.
sweep gas flow rate (F.. = 2.39*10-5 mol/s) sweep gas flow rate (F.. = 4.77*10-5 mol/s)

Two models were formulated to estimate total ethanol conversion for the membrane reactor.
Both models only take the dehydrogenation of ethanol into account. They do not account for
side reactions. They predict the total amount of ethanol converted and not the acetaldehyde
yield. The acetaldehyde yield will be lower than the total ethanol conversion due to the
formation of side products.

According to data for a differential reactor (see 8.3.4.1), the feed flow rates in both these
figures (Figures 9.1 and 9.2) are in the mass transfer limited regime, resulting in a
effectiveness factor of less than one. Modelling calculations with an effectiveness factor of
less than 1 and model parameters from Chapter 8, totally under predicted the ethanol
conversion. The reaction rate for the packed bed, plug flow reactor was much higher than the
values obtained from kinetic experiments with a differential reactor. The higher reaction rate
could be due to a higher k' -value, lower Kac<t-value or a higher TJ-value.

The individual changes in each parameter could not be obtained from plug flow reactor
experiments in this study, but the combined effect of the three parameters could be predicted
fairly accurately. The hydrogen permeance in the reaction mixture also varied significantly
from pure component data. Hydrogen permeance through the Pd-Ag film decreased due to

280
competitive adsorption on Pd sites. Other components adsorb on Pd and fewer sites are
available for hydrogen adsorption. The result was a decrease in hydrogen permeance.

For model 1 in Figures 9.1 and 9.2, standard membrane and kinetic parameters were used,
but with an effectiveness factor of 1 (see Tables 8.3 to 8.5). For model 2, the following
values were used for modelling:
• k'-value: 1.5 times the standard value,
• hydrogen permeance: 0.5 times the standard value,

• TJ: 1,
• other permeances: 0.5 times standard values to maintain standard selectivities, and
• all other data: similar to standard values.

Model 2 gIVes a much better prediction of the measured data. Table 9.1 indicates the
difference in total ethanol conversion between the model values and the measured values.
The average absolute error is defined as the difference between the measured total ethanol
conversion and the predicted values, averaged over all the sweep gas flow rates.

Table 9.1: Model differences for ethanol conversion at 250°C

Feed (molls) Average absolute error in total Average absolute error in total
conversion for model 1 (%) conversion for model 2 (%)
2.39*10"Y 5.3 2.9
4.77*10" 3.0 3.0

Figure 9.3 shows the selectivity of acetaldehyde as a function of the sweep gas flow rate. At
250°C, there was no improvement in selectivity with an increase in sweep gas flow rate.
Selectivity was between 80% and 95%. The acetaldehyde selectivity was most sensitive to
the flow rate and a higher flow rate resulted in a higher acetaldehyde selectivity. By-products
(see section 8.4.1) required a longer reaction time to form than the acetaldehyde did, and thus
decreased with an increase in feed flow rate.

281
 98

96

94
• -'"
..
...... .... . .. ......• __ F=2.39E-5 movs
~
..•.. F-4. T7E-5 molfs !
92
~ 90 ,
~ •\.~ i
> sa
~
1J
III
.
8.
I'\.
\ -= •
!
;,

82
\ / ,I

80
V
78
0 1 2 3 5
•
Sweep gas to feed flow molar ratio
• 7 8

Figure 9.3: Measured acetaldehyde selectivity (250°C)

The absolute and relative improvements obtained in total ethanol conversions by the
membrane reactor are summarised in Table 9.2. Absolute and relative improvements are
defined by:

Absolute improvement = MRV - PFRV (9.5)

Relative improvement = (MRV - PFRV)IPFRV (9.6)

With MRV the maximum membrane reactor value and

PFRV the plug flow reactor value.

.
Table 9.2· Improvements in total x.. for the membrane reactor at 250°C
Flow rate Sweep to feed Absolute Relative
(molls) molar ratio improvement in improvement in
conversion (%) conversion (%)
2.39*10·' 8 19.7 60.8
4.77*10·' 4 5.7 19.8

282
9.1.2. MEMBRANE REACTOR EXPERIMENTS AT 275°C
Figures 9.4 to 9.6 show measured and model data for ethanol dehydrogenation at 275 °C.
The feed flow rate in Figure 9.4 resulted in the equilibrium acetaldehyde yield for the plug
flow reactor. The equilibrium ethanol conversion and the measured acetaldehyde yield
differed by only about 2%. As the feed flow rate increased, the acetaldehyde yield decreased
and the difference between the equilibrium ethanol conversion and the measured
acetaldehyde yield increased (see Figures 9.5 and 9.6). In Figure 9.6 the exit ethanol
conversion dropped to below the equilibrium value due to the higher feed flow rate. The
same principles that were discussed in section 9.1.1 apply to the experiments performed at
275°C.

65 -' "
•
60
.. - - -
5Or-----~.--------------------~ •
55 - - - -- - - • • • - -! - .. ___ - - - - .. _- - - .. - -t
• -.- - - -'--"- -
,

'if. ,-
, 'if. .. r------_~.~-~==i.c__-----------
c
f!•
50

~ .. J.-.--
• ....
." ..;0,"

•
u
.. ~~
• .- ........
o·

• Eth...cd \Q ilCCltaloWlydll • T01.1 .th...", eonv.rslon )~

35 """Bh"oI~(model1) - ... EthWtoi con ......on (mocW 2)
-Equllbl1um oonverlscn
30
0 0.5 1 1.5 2
Sweep gas to feed molar ratio
2.5 3 3.5
• o 0.5 1 1.5 2 2.5
SWeep gas to feed molar ratio
3 3.5 4

Figure 9.4: Ethanol conversion at 275°C vs. Figure 9.5: Ethanol conversion at 275°C vs.
sweep gas flow rate (F", = 4.77*1O-s moVs) sweep gas flow rate (F", = 9.54*1O-s moVs)

Model I in Figures 9.4 to 9.6 used the standard parameters with an effectiveness factor of 1.
Model 2 had the following parameters:

• k'-value: 2 times the standard value,

• hydrogen permeance: 0.5 times the standard value,

• Tt: 1,

• other permeances: 0.5 times standard values to maintain standard selectivities, and

• all other data: similar to standard values .

283
For the highest tested feed flow rate at 275 °C (Figure 9.6), model 2 started under predicting
values at the higher sweep rates. For lower feed rates (Figures 9.1,9.2,9.4 and 9.5), model 2
was accurate in predicting total ethanol conversion at the higher sweep rates. A very likely
explanation for this is the presence of interphase mass transfer resistance. If interphase mass
transfer resistance was present, the value of lJ would increase with an increase in feed flow
rate. Since the lJ-value was fixed at 1 for all calculations, the model would under predict total
ethanol conversion at high feed flow rates in the presence of interphase mass transfer
resistance. Not enough data was available to determine the k' -value, K.... and lJ,
individually, and therefore values were fixed at each temperature and not treated as variables.
The model cannot account for changes in reaction rate parameters with changes in feed flow
rate.

50

30 j...~~·"'··C· ':c.:'~----.C.~--.-~
.. . . . -"'~"~~~~~~---1

o 0.5 1 1.5 2 2.5 3

SWnp gas to feed molar ratio

Figure 9.6: Ethanol conversion at 275 °C vs. Figure 9.7: Acetaldehyde yield at 275 °C
sweep gas flow rate (Fa = 1.43*10-4 mol/s) for a membrane reactor

Figure 9.7 shows the combined effect of feed flow rate and sweep gas to feed molar ratio on
total ethanol conversion. The feed multiple is the multiple of the standard feed rate
5
(4.77*10- mol/s). Total ethanol conversion increased more sharply at the lower feed rates
with an increase in the sweep gas flow rate. Towards the higher feed rates, the total ethanol
conversion for the plug flow reactor (sweep gas to feed molar ratio = 0) dropped slightly.

284
Model 2 gave accurate predictions of the exit membrane conditions. Table 9.3 compares the
model values with measured values for data at 275 DC.

Table 9.3: Model differences for ethanol conversion at 300 DC

Feed (moVs) Average absolute error in total Average absolute error in total
conversion for model! (%) conversion for model 2 (%)
4.77*10. 0
8.0 3.3
9.54*10. 0
11.9 4.3
1.43*10· 9.9 2.3

Table 9.4 indicates the measured improvements obtained by using the membrane reactor at
275 DC. The higher the equilibrium conversion, the more difficult it becomes to push the
reaction's equilibrium further towards the product side. The improvements in total ethanol
conversion with a membrane reactor will become srnaller at higher temperatures.

Table 9.4: Improvements in total Xet for the membrane reactor at 275 DC
Flow rate Sweep to feed Absolute Relative
(moVs) molar ratio improvement in improvement in
conversion (%) conversion (%)
4.77*10. 0
4 13.8 29.9
9.54*10. 0
4 7.0 16.0
1.43*10· 3 10.4 28.9

Figure 9.8 indicates an improvement in acetaldehyde selectivity at the lower feed rates with
an increase in sweep rates. Selectivity ranged between 80% and 95%, with the highest feed
rates giving the best selectivities at the lower sweep gas ratios. At the higher sweep gas
ratios, selectivity improved to above 90% for all feed rates.

285
 96 ------------------- ------------------------------------------------________________________ ·---..-..------------0-'
""..---.... ~-... ~ ................
94 --A_
~-- ... ----- ·········/i
92

~OO
•
:f 88 r.~ ....•.
~ 86
jiB4 -+-F=4.77E-5 molls
..•.. F=9.54E-5 molls H
82 t-------------------~~-.~-£~~1~.4~~4~mo~~~~,

80

78
0 0.5 1 1.5 2 2.5 3 3.5 4
Sweep gas to feed now molar ratio

Figure 9.8; Measured acetaldehyde selectivity (275 °C)

9.1.3. MEMBRANE REACTOR EXPERIMENTS AT 300 °C

Measured and model data at 300°C is presented in Figures 9.9 and 9.10. Modell predicts
total ethanol conversions with the standard parameters at 300°C and an effectiveness factor
of 1. Model 2 used the following parameters:
• k'-value: 2.5 times the standard value,
• hydrogen permeance: 0.5 times the standard value,

• T]: 1,
• other permeances: 0.5 times standard values to maintain standard selectivities, and
• all other data: similar to standard values.

At the lower ethanol flow rate (Figure 9.9), the exit ethanol conversion was the equilibrium
conversion, but at the higher flow rate the exit conversion dropped to below the equilibrium
5
value. At a flow rate of 9.54*10- moVs and a reaction temperature of 275 °C (Figure 9.5),
the total ethanol conversion was above the equilibrium value. To be consistent with data at
275°C, the total ethanol conversion should not have been below the equilibrium value at
300°C for a plug flow reactor and a feed rate of 9.54*10-5 moVs (Figure 9.10), due to the
higher reaction rate. The conversion was, however, below the equilibrium value. This
suggested that the reaction rate was slowed down at 300°C. This finding is consistent with
the catalyst stability data in Table 5.1. Between 280 °C and 340°C there was a sharp
decrease in catalyst stability, mainly due to coking. In the membrane reactor at 300 DC,

286
coking must have occurred during the testing period. Total testing took about 16 hours.
After testing at a flow rate of 9.54*10.5 molls (20 mVh), the feed rate was increased to 30
mllh, but the plug flow reactor results became very inconsistent.

60

- - - -- •
"r---------------._~._=_=_=~~----~ __

· r
l
an.,01 to 1e«lldllhydlo • Tobit eth...111 converllion I 40 ,," ........ _£_.
45 ., .'. Ettn!lnol CO!WI.rllion (model 1) -+,' Shanol conv.llioo (model 2)
- EIJoIIIbr11.1Tt c::onverlson
'" L.:=;:~~=::::=:==::==.::==:j
o 0.5 1 1.!I 2 2.5 3 3.5 " 1.5 2 2.5 3 3.5 "
SWeap gas to feed molar ratio SWeep gas to 1Hd mollr ratio

Figure 9.9: Ethanol conversion at 300 °e vs. Figure 9.10: Ethanol conversion at 300 °e vs.
sweep gas flow rate (Fe< = 4.77*10. 5 molls) sweep gas flow rate (Fe< = 9.54*10-5 molls)

Table 9.5 tabulates the accuracy of models 1 and 2 at 300 °e. Once again model 2 is very
accurate at the tested sweep flow rates, but model 1 under predicts the measured values due
to a reaction rate that is too slow in model 1.

Table 9.5: Model differences for ethanol conversion at 300°C

Feed (molls) Average absolute error in total Average absolute error in total
conversion for model 1 (%) conversion for model 2 (%)
4.77*10·' 1l.5 l.9
9.54*10·' 12.7 2.3

The improvements in total ethanol conversion using a membrane reactor at 300 °e are
indicated in Table 9.6.

Acetaldehyde selectivity varied between 85% and 96% (Figure 9.11). For the plug flow
reactor, the selectivity was the highest at the highest feed flow rate. At low sweep rates, the
selectivity dropped for both feed flow rates tested and then increased again towards the

287
 higher sweep gas flow rates. In both cases the selectivity was above 90% at the higher sweep
gas flow rates.

Table 9.6: Improvements in total X.. for the membrane reactor at 300°C
Flow rate Sweep to feed .. Absolute Relative
(mol/s) molar ratio improvement in improvement in
conversion (%) conversion (%)
4.77*10·) 4 8.1 14.3
9.54*10·) 4 7.7 16.3

9. ,
,!
98
• ,,
94
!
i'-
~92
__-----1
····················T
.~ ................ ~
......
¥
11
90
!
til
88 /' :,
88 1\ I V
I
-+-F=4.77E-5 rnoVs
..•.. F 9,54E-5 moVs
!

,
,!
84 , , , ,!,
0 0.5 1 1.5 2 2.5 3 3.5 4
Sweep gas to feed now molar ratio

Figure 9.11: Measured acetaldehyde selectivity (300°C)

9.1.4. FURTHER DISCUSSION OF ETHANOL DEHYDROGENATION

The theory (see 2.9.4.3) suggests a shift in the reaction equilibrium in a membrane reactor,
resulting in higher conversions at the same operating temperature. Catalytic membrane
reactor experiments carried out to determine ethanol dehydrogenation confirmed the theory.
In section 8.1, the effect of feed rate on reaction conversion was discussed. It was said that
for the membrane reactor to be of use, the improvements in conversion had to be obtained in
region 2 of Figure 8.1. For ethanol dehydrogenation in a catalytic membrane reactor,
reaction conversion improvements were indeed obtained in this region, and as follows:
• 250°C: 5.7% absolute or 20% relative improvement, and
• 275°C: 10.4% absolute or 29% relative improvement.

288
I
I

I
At 300°C, coking began to take place and it lead to inaccurate data. From an operating point
of view, there seemed to be an optimum working temperature. If the temperature was too
low, the reaction kinetics was slow and the conversion poor, even with a membrane reactor.
Organic molecules adsorb more strongly onto the Pd-Ag membrane film at low temperatures
than at higher temperatures and reduce hydrogen permeance more substantially at lower
temperatures. Higher temperatures improve hydrogen permeance in two ways: by increasing
the hydrogen permeance kinetics and by reducing the organic molecule adsorption onto the
Pd-alloy surface.

On the negative side: at high temperatures, catalyst deactivation occurs quickly, which
reduces the activity of the copper based catalyst and leads to poor conversion. Coking will
most probably also occur on the membrane surface and thus reduce hydrogen permeance
severely. The sensitivity analysis in Chapter 8 indicated that the performance of the
membrane reactor was more sensitive to reaction rate parameters than to membrane
parameters. To obtain large improvements in ethanol conversion over the conventional plug
flow reactor, the catalyst activity must remain high.

A temperature of 275°C gave the best ethanol dehydrogenation results. The selectivity
towards acetaldehyde formation was above 80% at all times. Selectivity was more dependent
on feed flow rate than on temperature or sweep gas flow rate. The selectivity increased with
an increase in feed flow rate and also, to a lesser extent, with an increase in sweep gas flow
rate. Selectivity was above 90% for all experiments at the highest tested sweep rates, except
for the very low feed flow rate of 2.39* 10.5 molls. Figure 9.12 indicates some selectivity
profiles at different temperatures. There was a lot of variance in the data at the lower sweep
rates, but the final selectivities at the higher sweep rates were not very dependent on
temperature.

Total ethanol conversions were modelled with two different models. The first model used
standard conditions (as defined in Tables 8.3 to 8.5) and an effectiveness factor of 1. In most
instances the model under predicted the measured data. This indicated that the reaction rates

289
determined in kinetic experiments in a differential reactor were much lower than the values in
the plug flow reactor. Higher k' -values were employed in the second model, keeping all
adsorption coefficients constant The effectiveness factor was fixed at 1 and the permeances
for each component in the reaction was halved. The net result was a model that predicted
total ethanol conversion to within 4.5%.

96 ....................................................................................................................... .

94f---I-~-----=~'======::::!
--,:,:::
/ ,--- --'
........

-T-250°C -

M+---------------------4 -1l- T=300°C -

o 0.5 1 1.5 2 2.5 3 3.5 4

Sweep gas to feed flow molar ratio

Figure 9.12: Measured acetaIdehyde selectivity at a constant feed rate

(Fet = 9.54*10-5 mol/s)

Deng et aI. (1995), Liu et aI. (1997) and Raich and Foley (1998) studied ethanol
dehydrogenation in a membrane reactor, under conditions that differed greatly from those
used in the present study. The feed rates that they employed were more than an order of
magnitude lower than the feed rates used in this study (see Table 8.1). Liu et aI. (1997)
tested ethanol dehydrogenation as a function of feed rate. At a similar WIF ratio of
1000 g.min/mol to Liu et aI. (1997), this study achieved a maximum of 65% conversion at
300°C compared to their 30%.

Acetaldehyde selectivity in this study was significantly higher than the values obtained by
Raich and Foley (1998). This is to be expected at the higher feed rates. They obtained higher
conversions because they used a very low feed rate and a very diluted feed. When the partial
pressure of ethanol was reduced (by using a diluted feed), the equilibrium conversion for the
ethanol dehydrogenation reaction increased.

290
9.2. 2-BUTANOL DEHYDROGENATION
Results of 2-butanol dehydrogenation in a membrane reactor are listed in Appendix F2.
Catalytic membrane reactor experiments were performed at temperatures ranging from
190°C to 240 dc. Standard reaction rate parameters and membrane parameters used for
modelling were listed in Tables 8.6 to 8.8. Catalytic membrane performance at 190 DC,
215 °C and 240°C will be discussed separately. For the three dimensional figures shown in
this section, the standard feed rate was 3.04*10-5 molls and the feed multiple was the multiple
of the standard feed rate.

9.2.1. MEMBRANE REACTOR EXPERIMENTS AT 190 °C

Figures 9.13 to 9.16 show 2-butanol conversion as a function of sweep gas flow rate at
different feed rates. The measured exit 2-butanol conversion dropped below the equilibrium
value for feed flow rates of3.04*10- 5 molls and higher. The exit conversion was modelled in
a manner similar to that for the ethanol dehydrogenation reaction. Initial modelling
experiments were performed with reaction rate parameters determined from differential
reactor experiments in Chapter 5. The reaction rate parameters of Table 8.6 substituted into
equations (8.57) to (8.76) yielded 2-butanol conversions much lower than the measured
values. For further model calculations, the parameters were changed to the following:
• k'-value: 2 times the standard value at 190°C,
0.1 times the standard value at 190 DC,

· ,.,:
• permeances:
fixed at 1 for all flow rates, and
the same as in Table 8.7 at 190°C.

Increasing the model's k- value for 2-butanol dehydrogenation alone was insufficient to
obtain predicted conversions close to the measured conversions for the plug flow reactor.
The MEK adsorption coefficient was also reduced by a factor 10 for modelling. This may
seem exaggerated, but it should be noted that the individual adsorption coefficients were
determined for pure component data. For example, the MEK adsorption coefficient was
determined when only 2-butanol and MEK were passed over the catalyst bed in the
differential reactor. The conversion was kept very low and very little hydrogen formed. In

291
the plug flow reactor, 2-butanoi was converted to MEK and hydrogen in equal molar
amounts. The MEK was essentially in a MEK-hydrogen environment and not in a MEK-2-
butanol environment. The MEK adsorption behaviour could thus be expected to be
significantly different from the behaviour in a MEK-2-butanol mixture.

80

"
..• "
•

"'70
.. . .. .
, .. -'
~t-----------------
•

•= • • ~ .~
='" j----l!-_---~==-~=:c::-..::.-.---
~"
> •
• 2-.... ~. ~ MEK • T""2-IM""""""", ~ j ... .
860
1- .' -2-eut.nol COfIva'elon (1) -EquAlbrium o::onv.,.;$Q'I
8~~~==========~
/. . . . :. . . I • 2-BubnoiIoMEK

"+7"---1I .•
" 2-BWnoIllIIfl'.'WnIon(1) -E4II~~

50 ,,~----------"------, , ,
0 2 4 6
SWeep gas to teed molar ratio
6 10 12 14 o 2 4
Sweep gas to fad molar ratio
•
Figure 9.13: 2-Butanol conversion at Figure 9.14: 2-Butanol conversion at
190°C vs. sweep gas flow rate 190°C vs. sweep gas flow rate
(F2But= 1.52*10-5 moVs) (F2But= 3.04*10-5 moVs)

eo ,
. ,
•
'j" 52
.. ....•
~
•>
~ .. ! .
....... ... '" ."

. I '~_ ~
U

.. I
•

·
2-Bl.Mnailo MEK

2-8utMoi conv.-elon (1)

• T."2·...
-E~lbriumc:ortV«l""" ,i
.., , ,
0 1 2
SWHP gas to feed molar ratio
4
• 7

Figure 9.15: 2-Butanol conversion at 190°C vs. Figure 9.16: MEK yield at 190°C
sweep gas flow rate (FzBut= 4.56*10-5 moVs)

The 2-butanol conversion was sensitive towards the feed rate. There was a drop in 2-butanol
conversions at 190°C with an increase in feed rate. Figure 9.16 illustrates this very clearly.

292
 Optimum conversions were obtained at low flow rates and high sweep ratios as predicted by
the membrane reactor model.

At 190°C, the model predicted the measured values well. There was little difference
between the total 2-butanol conversion and the MEK yield, indicating very high selectivity.
The best 2-butanol conversion in the membrane reactor at a feed flow rate of 1.52 * 10. 5 molls
was 79% and it dropped to 53% at a feed rate of 4.56*10. 5 molls.

The improvement in 2-butanol conversIOn by the membrane reactor is summarised in

Table 9.7. There was a large improvement in conversion at the lower flow rate. The absolute
differences between the model conversions and the measured 2-butanol exit conversions are
listed in Table 9.8.

Table 9.7: Improvements in total X2But for the membrane reactor at 190°C
Flow rate Sweep to feed Absolute Relative
(molls) molar ratio improvement in improvement in
conversion (%) conversion (%)
1.52*10"0 12.6 24.3 44.3
3.04*10"' 6.3 8.8 16.1
4.56*10"' 6.3 8.0 17.7

Table 9.8: Model differences for 2-butanol conversion at 190°C

Feed (molls) Average absolute error in total
conversion for model 1 (%)
1.52*10"0 3.5
3.04*10"0 3.3
4.56*10"' 0.9

As mentioned previously (section 8.4.2), the selectivity towards MEK production at 190°C
was very high. Figure 9.17 plots selectivity as a function of feed rate and sweep gas to feed

293

i .
ILl.
molar ratio. Selectivity was above 99% for all experiments and increased towards 100"10 at
the higher flow rates. The only by-product observed in some experiments was 3-octanol.

100
..
99.6
•• • •
;!.
~992
•
~
OJ 98.B
'" .F=1.52E·5 molls
• f=3.04E-5 moVs -
98.4
.. F=4.56E-5 molfs

9B
0 1 2 3 4 5 6 7
Sweep gas to _ !low molar ratio

Figure 9.17: Measured MEK selectivity at 190°C

9.2.2. MEMBRANE REACTOR EXPERIMENTS AT 215 °C

Figures 9.18 to 9.21 summarise 2-butanol conversion at 215°C in a membrane reactor.
Figures 9.18 to 9.20 show the classic signs of mass transfer resistance for the model values,
i. e. an over prediction of conversions at low flow rates and an under prediction of values at
high flow rates. The same model parameter ratios employed at 190°C were employed for
modelling calculations at 215 °C. The standard parameter values at 215°C are listed in
Tables 8.6 and 8.7.

The model did not take mass transfer resistance into account, since the effectiveness factor
was fixed at I. Kinetic data obtained in the differential reactor indicated a reaction free from
interphase mass transfer resistance (see discussion in section 5.2.2). The plug flow reactor
data contradicted the differential reactor data regarding interphase mass transfer resistance.
One possible reason is the difference in the Reynolds numbers of the two situations. For
differential reactor experiments, a 4 mm inside diameter tube was used, resulting in much
larger gas velocities as compared to the 7 mm inside diameter tube used for plug flow and
membrane reactor experiments. By assuming 11 as a variable dependent on feed flow rate, a
better model fit might be obtained, but as a first approximation this was not investigated.

294
If 1] decreases with a decrease in feed flow rate, the k' -value must be even higher than the
value used for this model. In such a case, the model should be fitted to the measured data at
the maximum feed rate with a high 1]-value to determine the k' -value and KMEK. The 1]-value
must then be reduced at the lower flow rates to yield lower overall reaction rates and lower
model conversions at the lower feed rates.

" ........... -.....

•.. j i 90

00
, ........ . •
•
• 85~------------------------~'~~
.• .

i: .' ./
;1.85 ;I. ...•...... ' .. ' .' ....
e

!.~_d~MEK • Tm.~~~_ ~
.2 ••
~80
•
•0 " ~--.·~-II • :z..Butanoi to MEK
~
U
• L""'" conversion (1) -
2-Bllt.nol
• Tobd 2-8ut (:anv.1'$Ian
Equllbftum convtlrison n 8 . . .> . I. ...
2-BubI'IOI converlion (1) -Eq.JIlbriIm converitol'l !i

70 t--~======~;l
1
70
;
,

85~---,---__- -__- -__- -__--~--~i

o 1 2 3 " 5 6 7
Sweep gas to feed molar ratio
a 9 10 11 12 13 2 3 "
Swaep gas to fHd molar ratio
5 • 7

Figure 9.18: 2-Butanol conversion at Figure 9.19: 2-Butanol conversion at

215°C vs. sweep gas flow rate 215°C vs. sweep gas flow rate
(F = 1.52*10-s molls)
2But (F2But= 3.04*10-s molls)

At a feed rate of 1.52*1O-s molls and 3.04*10-5 molls, the exit 2-butanol conversion for the
plug flow reactor was similar to the equilibrium conversion. The model values were larger
5
than the equilibrium values. At the highest feed rate tested (6.08*10- molls), the 2-butanol
exit conversion in the plug flow reactor dropped below the equilibrium value and the reaction
moved into block 2 as described in Figure 8.1. Figure 9.21 indicates the difference in 2-
butanol conversion at different feed rates in the plug flow reactor (at zero sweep gas). At
high sweep rates, the membrane reactor improved 2-butanol conversion for all feed flow rates
tested. The highest conversions were in the region of 90%, with the equilibrium value being
70%.

295
 15
90

.. ..
, ...
'

6O+-t-_-_-_-_-_-_-~
o 234 5
Sweep gas to feed molar raHo
•

Fil!Ure 9.20: 2-Butanol conversion at 215°C vs. Figure 9.21: MEK yield at 215°C
sweep gas flow rate (F2But= 6.08*10.5 mol/s)

The performance of the membrane reactor at 215°C is listed in Table 9.9 and the accuracy of
the model is summarised in Table 9.10.

Table 9.9: Improvements in total X2But for the membrane reactor at 215°C
Flow rate Sweep to feed Absolute Relative
(molls) molar ratio improvement in improvement in
conversion (0/0) conversion (%)
1.52*10·> 12.6 21.2 30.5
3.04*10·' 6.3 IS.2 26.6
6.0S*10·> 6.3 7.4 1l.S

.
Table 910·. Model differences for 2-butanol conversion at 215°C
Feed (molls) Average absolute error in total
conversion for model 1 (%)
l.52*10·> 6.6
3.04*10·> 1.9
6.08*10·> 2.5

For model differences in Table 9.10 and in all other tables, the absolute differences are listed.
There is no distinction between model values higher or lower than the measured values, only

296
the difference between them is stated. The model loses accuracy at the lower feed rates, with
the difference between measured values and model values increasing to above 6%.

The MEK selectivity at 215°C (Figure 9.22) was high (above 96%), but slightly lower than
the values obtained at 190 °C. The selectivity improved with an increase in feed flow rate
and at the highest feed flow rate tested, the selectivity reached 100%.

100

•
99 •
.... 98
• •
~ •
.a:
j
.,.. 97
•
• • • F=1.52E-5 molls

98 • F=3.04E-5 moVs ---1

.. F=6. oaE-5 molls

95
0 2 3 4 5 6 7
Sweep gas to feed flow molar ratlo

Figure 9.22: Measured MEK selectivity at 215°C

9.2.3. MEMBRANE REACTOR EXPERIMENTS AT 240°C

The highest temperature at which 2-butanol dehydrogenation was tested was 240°C. The
results are presented in Figures 9.23 to 9.26. Modelling parameters at 240 °C were:
• k'-value: 2 times the standard value at 240°C,
0.1 times the standard value at 240 °C,

• 1]:
fixed at 1 for all flow rates, and
• permeances: the same as in Table 8.7 for 240°C.

At 240°C, the exit 2-butanol conversion remained at the equilibrium value for all flow rates
4
tested in the plug flow reactor, up to a feed flow rate of 40 mllh (1.22*10- molls).
Membrane reactor experiments were only performed at feed flow rates of up to 20 mllh
(6.08*10-5 molls) due to limitations of the sweep gas flow meters. At the low flow rates
(Figure 9.23 and 9.24) the model over predicts the measured values and at the higher flow

297
rates, the model under predicts the measured values. The same explanation supplied in
section 9.2.2 is applicable at 240°C. Initial modelling results indicated the presence of
interphase mass transfer resistance at low flow rates for 2-butanol dehydrogenation. The
effectiveness factor was not constant with feed flow rate, but declined as the feed flow rate
declined. Measured 2-butanol conversions were at or above 90% for all flow rates tested.

'00 '00

os .. ' .
--, ....... "

ae
~t-----------------~~~-----
............................
....
"

...........•..... •• & ...•• " .,.. •.••• ,

'"
c
.2
00
••
.
5 90 t-----.
..•.,-'."'----------------------~-
;> ./

0
c
0
"
• i8 "
Mt---,--~---------------------
•
80

o 2 4 5 6 7
2 4 II e
SWeep gas to fAd molar ratio
Sweep gas to fHd molar ratio

Figure 9.23: 2-Butanol conversion at Figure 9.24: 2-Butanol conversion at

240°C vs. sweep gas flow rate 240°C vs. sweep gas flow rate
(F2But= 1.52*10-5 mol/s) (F2But= 3.04*10-5 mol/s)

For the plug flow reactor in Figure 9.26, the measured data confirmed, to some extent, the
model prediction of interphase mass transfer resistance. At a zero sweep gas to feed molar
ratio, the total 2-butanol conversion increased very slightly with an increase in feed flow rate.
The improvements in 2-butanol conversion in the membrane reactor for high sweep rates
were similar at all feed multiples. If higher feed rates were to be employed, the benefits of
the membrane reactor would start declining.

The improvements obtained by the membrane reactor are listed in Table 9.11. As mentioned
previously, improvements become smaller at the higher temperatures, because the
equilibrium conversion becomes very high. The model predicts the measured values at lower
feed rates with less accuracy than at the higher feed rates (see Table 9.12). The difference
between the model predictions and the measured data increases above 5% at the lower feed
rates.

298
1'l~!II!' ',

" • ~Bubnollo MEK I

.'"
• Tot.I2-EM conven:lgn

I...... 2-But.,.01 co~on (1) -ECf,llHbr1um cOl'l'llll'blgn

t-============~-
!
I •
r
. .... . ... .. . ......•
............
.............
....•', ....

8D~=============--

o 2 3 4 5 ,
SWftp gas to fMcI motIr ratio

Figure 9.25: 2-Butanol conversion at 240°C vs. Figure 9.26: MEK yield at 240°C
sweep gas flow rate (F2But= 6.08*10.5 molls)

.
Table 9 11'. Improvements in total X2But for the membrane reactor at 240 °C
Flow rate Sweep to feed Absolute Relative
(molls) molar ratio improvement in improvement in
conversion (%) conversion (%)
1.52*10·' 12.6 16.8 21.4
3.04*10·' 6.3 11.9 15.1
6.08*10·' 6.3 8.2 10.0

Table 9.12: Model differences for 2-butanol conversion at 240 °C

Feed (molls) Average absolute error in total
conversion for model! (%)
1.52*10·' 5.5
3.04*10·' 5.3
6.08*10·' 3.1

Selectivity didnot decrease with an increase in temperature up to 240 °C, it remained above
97% (Figure 9.27). At higher temperature, butenes will start forming and the selectivity will

299
III!!

drop. Figure 4.41. shows the MEK selectivity as a function of temperature up to 390°C,
Selectivity at 240°C was very good but then gradually declined towards the higher
temperatures.

100 ".--------------.. '-.------------.. --. --------"._-----.---.-------.-----------.. -----. ----""----.---.---..----

•
99,5 •
• •
•
'#-
99
•
~
>
~ 98.5
•
'iI
U)
98 • F 1.52E-5 molls -

97.5
• • F=3.04E~5

... F=6.0BE·5 moVs

moVs
-

97
0 2 3 4 5 6 7
Sweep gas to feed now molar rstlo

Figure 9.27: Measured MEK selectivity at 240°C

9.2.4. FURTHER DISCUSSION OF 2-BUTANOL DEHYDROGENATION

For 2-butanol dehydrogenation, the conversion in a plug flow reactor increased sharply with
an increase in temperature and decreased with an increase in feed flow rate. The higher the
operating temperature and the reaction rate, the higher the feed rates that could be employed
before the conversion started decreasing (Figure 9.28), due to a decrease in reaction time.

Catalytic membrane reactor experiments were performed with feed flow rates resulting in
both equilibrium restricted and non-equilibrium restricted 2-butanol conversion (at 190°C
and 215 0c). For high flow rates resulting in the non-equilibrium restricted conversions, the
theoretical benefits of the membrane reactor were confirmed by experimental data. At
240°C, membrane experiments were only performed for feed flows resulting in the
equilibrium restricted conversion.

300
Figure 9.28: MEK yield % vs. flow rate and temperature
(feed mUltiple of one equals 3.04*10.5 molls (lOml/h»

Figure 9.29 shows the improvements in total 2-butanol conversion obtained from a
membrane reactor. Figure 9.29 illustrates results for a feed rate of3.04*10· 5 molls (10 m1/h).
To achieve a high MEK yield, a high operating temperature must be used and the temperature
should be kept at or below 250°C to improve catalyst life. Figure 5.8 indicates catalyst
deactivation over a 24 hour period. at different operating temperatures. At 250°C, the
catalyst was stable, but at higher temperatures the activity of the catalyst declined over time.
At 240°C, the membrane reactor gave an increase in MEK yield of about 10"10.

Figure 9.29: Effect of sweep ratio on MEK yield %

301
No data on 2-butanol dehydrogenation in a membrane reactor could be found in literature. It
was not possible to compare results on 2-butanol dehydrogenation with data obtained from
other researchers.

9.3. SUMMARY
Theoretical benefits of a catalytic membrane reactor as compared to a plug flow reactor were
realised for the dehydrogenation of ethanol and 2-butanol. For ethanol dehydrogenation,
catalyst deactivation above 280°C limits the upper operating temperature of copper based
catalysts. For 2-butanol dehydrogenation, the copper based catalyst was stable at 250°C and
lower. Coking at higher temperatures will deactivate the copper on silica catalysts and result
in a rapid decrease in catalyst activity.

The optimum working temperature for ethanol dehydrogenation with copper on silica
catalysts is between 270 °C to 280°C. The equilibrium conversion in this temperature range
is just over 40% for an undiluted feed at atmospheric pressure. The membrane reactor
improved the total ethanol exit conversion from 45% (plug flow reactor) to 60% at low feed
flow rates and from 36% to 46% at high feed flow rates. The acetaldehyde selectivity of the
reaction was above 80% for all experiments performed. The selectivity was most sensitive to
the feed flow rate and the selectivity improved with an increase in feed flow rate. The
selectivity also improved with an increase in sweep gas flow rate, but changed little with
temperature from 250°C to 300 °C. At a high sweep gas flow rate, the acetaldehyde
selectivity was above 90% for the majority of the experiments.

The membrane reactor is less suited for 2-butanol dehydrogenation, because there are less
temperature limitations for this reaction. Improved conversion of 2-butanol can be obtained
by increasing the operating temperature up to 250°C, without any significant deactivation of
the catalyst. It is more economical to increase the reaction temperature than to use a
membrane reactor at a lower temperature to obtain the same results. The membrane reactor
does have some benefits in that it can obtain a near complete conversion at the higher
temperatures. Maximum plug flow reactor conversion at 240 °C is 80%. This can be

302
increased to above 90% at 240"C by employing a membrane reactor. The exit 2-butanol
conversIOn in the membrane reactor was also modelled with equations developed in
Chapter 8. A proper model could only be formulated by using a k'-value twice (or more) the
value obtained from the differential reactor experiments and a MEK adsorption coefficient
one tenth of the value obtained from the differential reactor data. The model predicted the
experimental data to within 6.5% of the measured exit 2-butanol conversion. Selectivity
towards MEK production was high under all conditions (> 96%) and increased towards 100%
at the higher feed flow rates.

The Pd-Ag membrane was stable between 190 °C and 300 "C in both an ethanol and a 2-
butanol environment for about 30 working days (450 hours) in total. The same membrane
was used for both reactions and the membrane reactor maintained its superior performance as
compared to the plug flow reactor for the full duration of testing.

303
 10. CONCLUSIONS

In Chapterl, the six main goals of the project were stated. They were:
• To optimise the performance of an alcohol dehydrogenation catalyst,
• To model. the kinetics of an alcohol dehydrogenation reaction,
• To optimise the composition and thickness of Pd and Pd-alloy films for hydrogen
separation,
• To model the membrane separation process with a sweep gas and a pressure
differential,
• To compare the performance of a membrane reactor consisting of the optimised
catalyst and optimised membrane with a plug flow reactor, and
• To model the membrane reactor.

The main conclusions to the work performed to reach each goal are summarised in the
following few pages.

10.1. CATALYST OPTIMISATION

All catalysts were prepared by impregnating a support with copper. The performances of
the catalysts were tested for the dehydrogenation of ethanol and 2-butanol. Three support
types were tested and it was found that:
• The alumina support yielded mainly dehydration products and not the required
dehydrogenation products,
• The MgO support had a very low surface area and that resulted m very low
conversion of ethanol to acetaldehyde, and
• The silica support, impregnated with copper, gave high conversions of both ethanol
and 2-butanol towards the desired dehydrogenation products.

For silica based catalysts, there were an optimum copper percentage of 15 wt % on silica
to obtain the highest conversion of the alcohol towards the desired product. The addition
of cobalt, chromium or nickel stabilisers to copper on silica catalysts had negative effects

304
on acetaldehyde yield. More by products fonned and the yield of acetaldehyde decreased
due to a decline in acetaldehyde selectivity. When the same stabilisers were added to
copper-alumina catalysts, the effect was positive and the production of di-ethyl ether (the
desired product) increased. A catalyst particle size of up to 1180 microns had a very
small effect on product yield, indicating that intraparticle mass transfer resistance for both
ethanol and 2-butanol dehydrogenation were negligible. The production of methyl ethyl
ketone from 2-butanol and acetaldehyde from ethanol, approached the equilibrium
conversion at low feed flow rates.

10.2. REACTION MODELLING FOR ETHANOL AND 2-BUTANOL

DEHYDROGENATION
A 14.4 wt % copper on silica catalyst was used for all kinetic and stability testing. The
stability of the catalyst was detennined prior to kinetic testing to detennine a stable
region for perfonning kinetic experiments. Copper catalysts was reduced in hydrogen
prior to using them. The reduction temperature is critical to prevent sintering of copper
particles. Reduction at 400°C and 340 °C caused severe sintering, but not at 255°C.
For ethanol dehydrogenation, catalyst stability tests were perfonned from 220°C to
400 °C and for 2-butanol dehydrogenation, stability tests were perfonned at 250°C and
310°C. For ethanol dehydrogenation, the catalyst was stable up to 280 DC. Higher
temperatures caused both sintering and coking of the catalysts even if additives (Co, Cr or
Ni) were added to the copper. For 2-butanol dehydrogenation, the copper on silica
catalyst was stable at 250°C.

The ethanol dehydrogenation reaction exhibited strong interphase mass transfer

resistance, while for 2-butanol dehydrogenation there was no clear indication of
interphase mass transfer resistance. Ethanol dehydrogenation kinetics were detennined
from 200°C to 300 °C and 2-butanol dehydrogenation kinetics from 190°C to 280 DC.
Both reactions could be described by the dual site, surface reaction controlling
mechanism. The organic products (MEK or acetaldehyde) had dominant adsorption
coefficients.

305
10.3. MEMBRANE OPTIMISATION
A palladium membrane is sensitive towards the operating conditions of the membrane. If
the membrane is to be operated above 300°C, a pure palladium membrane can be used iri
a hydrogen environment, but if the membrane is to be operated below 300°C, a
palladium alloy must be used. Pure palladium embrittles quickly below 300°C in
hydrogen, and thus becomes unusable.

Palladium films were prepared be electro less plating. To produce a high quality and very
thin palladium film, supported on a membrane, each production step must be performed
with care. Prior to plating, the asymmetric alumina membrane support must be pretreated
with palladium and tin solutions. A high palladium to tin ratio gives a more stable film.
Alternatively, new tin free pretreatment solutions can be employed. The initial plating
rate should be slow to prevent the formation of defects in the film. The plating rate was
controlled by the addition of hydrazine at a constant plating temperature of 71°C.
Thorough cleaning of the membrane after plating is crucial to remove EDTA (from the
plating solution) trapped in the membrane pores. Membranes were stirred in ammonia
solution, dried at 240°C and further oxidised at 320°C to remove organic components in
the membrane pores.

Palladium films of thicknesses down to 1 J.l.m were deposited on the inside of asymmetric
SCT a-alumina membranes (200 nm pore size). Hydrogen permeances varied between
about 8 and 15 J.l.mollm2.Pa.s for temperatures from 330°C to 450 °C and palladium films
from 1.0 to 1.5 !lm. Hydrogen to nitrogen selectivity was above 100 for all membranes
tested and above 400 for all but two membranes (thickness 1.0 to 1.5 ).Lm). These values
are a significant improvement over other published results.

Silver was deposited on palladium to form palladium-silver films of less than 2.2 !lm
thick. The separated layers caused a sharp drop in hydrogen permeance through the film.
The bi-Iayer was heat treated to improve hydrogen permeance and form a homogeneous
alloy. In a hydrogen atmosphere at 590°C, defects formed in the Pd-Ag film within

306
10 hours. The hydrogen also passivated the film at 590°C. At 550 °e, defects still
fonned in hydrogen, but at a much slower rate. Some defect fonned in an argon
atmosphere at 550°C after more than 100 hours. Few defects fonned in both argon and
nitrogen at 500°C. Nitrogen passivated the film and thus argon was chosen for
perfonning alloying experiments.

Alloying should be perfonned at 550°C in argon for 10 to 15 hours. Thereafter, the

membrane must be cooled down to 310°C in argon, oxidised in oxygen for 1 hour and
then reduced in hydrogen. This additional oxidation reduction step has not been
proposed before in literature and a significant improvement in the hydrogen penneance
was obtained after this step was perfonned.

10.4. MEMBRANE MODELLING

For both Pd and Pd-Ag membranes, the hydrogen penneances were accurately described
by the flux equation with a value of 1 for the pressure exponent (n-value). Temperature
data fitted the Arrhenius equation well and nitrogen flow through defects in the palladium
film was Knudsen diffusion. A model was fonnulated for describing hydrogen
penneance in the presence of an inert sweep gas. The model provided a very good fit to
the experimental data. Pure Pd membranes were tested at temperatures from 330°C to
450 °e and Pd-Ag membranes were tested at temperatures from 250°C to 450 °e with
and without a sweep gas.

10.5. MEMBRANE REACTOR VS. PLUG FLOW REACTOR

Flow rates were detennined where the plug flow reactor conversion started decreasing to
below the equilibrium values. For the catalytic membrane reactor to be applicable,
improvement in reaction conversion must be obtained in the higher flow rate region.
Theoretical benefits of the catalytic membrane reactor were realised for both ethanol and
2-butanol dehydrogenation in a catalytic membrane reactor, even at the higher flow rates.

307
For ethanol dehydrogenation in a membrane reactor, the best results were obtained at
275°C. At 250 °C, the reaction rate was too slow to obtain large improvements in
conversion for the membrane reactor compared to the plug flow reactor. At 300°C,
coking of the copper on silica catalyst took place, which deactivated both the membrane
and the catalyst and which led to poor results. At 275 °C, the membrane reactor
improved the total ethanol exit conversion from 45% (Plug flow reactor) to 60% at low
feed flow rates and from 36% to 46% at high feed flow rates. Modelling the process with
kinetic data obtained from differential reactor experimentss, under predicted the
experimental data. By increasing the reaction rate (k' -value and effectiveness factor), a
very good fit for the experimental data was obtained.

The dehydrogenation of 2-butanol in a catalytic membrane reactor also resulted in much

higher conversions than in the plug flow reactor. The maximum 2-butanol conversion in
a plug flow reactor at 240°C was 80% and that increased to above 90% for the
membrane reactor. The process model also under predicted the measured conversions
when kinetic data from differential reactor experiments was used for calculating reaction
rate parameters. Higher k' -values and lower KMEK-values resulted in a model that fitted
the experimental data well.

10.5. MEMBRANE REACTOR MODELLING

A membrane reactor model was developed for constant temperature operation, with no
pressure drop across the catalyst bed and no radial concentration profiles in the
membrane tube. Small catalyst particles were used to exclude intraparticle mass transfer
limitations from the model. Interphase mass transfer limitation was expressed as a
function of the Reynolds number.

A sensitivity analysis of the membrane reactor model (for ethanol dehydrogenation) was
performed to test the effects of membrane parameters, reaction rate parameters and the
operating variables on the overall ethanol conversion. Membrane data from the

308
optimised membrane and reaction rate data from the optimised catalyst were used as
standard inputs for the model.

The reaction rate parameters have a significant effect on the overall conversion, but have
no effect on the percentage of produced hydrogen that penneates to the shell side of the
reactor. A high k' -value, a low Kace,-value and/or a high effectiveness factor will increase
the reaction rate and improve the reaction conversion, while a slow reaction will increase
the ethanol losses through the Pd-alloy film defects. The membrane selectivity (H 2 :N2)
should be at least 150 to prevent large losses of ethanol feed through the film defects, and
to prevent severe back diffusion of hydrogen from the shell side to the tube side. The
feed flow rate and the sweep gas to feed ratio must be chosen in conjunction with one
another. The higher the pressure difference (i.e. the higher the sweep gas flow rate), the
better the hydrogen driving force and the higher the overall conversion.

309
 11. FUTURE WORK

')

Few alcohol dehydrogenation reactions have been studied in a membrane reactor. Of

those reactions that have been studied, the feed flow rates have been unrealistically low
and/or the hydrogen permeance has been very poor due to thick palladium films. This
dissertation concentrated on performing reactions at realistic feed flow rates III a
membrane reactor with very high hydrogen permeance and very high selectivity.

The performance of palladium and palladium alloy membranes have been optimised to
such an extent that little further work in this area is necessary. Model calculations have
shown that little will be gained by further improving membrane selectivity and/or
hydrogen permeance. The main challenge now is in improving catalyst activity and
stability. Work performed in this dissertation has proved that Pd-based membranes can
be used successfully from 190°C and upwards, without deterioration of the film. From a
thermodynamic point of view, higher temperature reactions are more favourable, since a
high temperature will improve hydrogen permeance through Pd or its alloy.

The future lies in identifying high value added alkane or alcohol dehydrogenation
reactions. Such reactions will typically be found in the pharmaceutical industry and the
perfume industry, in the production of pesticides and insecticides and other speciality
chemicals. Higher temperatures are beneficial for endothermic reactions to improve
conversions and also to improve hydrogen permeance, but result in a decrease of
selectivity. Cracking and polymerisation side reactions must be avoided to limit catalyst
coking.

After identifying suitable reactions, the catalysts for those reactions should be optimised
to maintain high selectivity and high activity over long periods of time. The performance
of the membrane, packed with the most suitable catalyst, should then be monitored over a
long time period. This step was not part of this dissertation. The final step would then be
to scale up the technology.

310
 12. REFERENCES

Abdalla BK, Elnashaie SS, Alkhowaiter S et aI., Applied Catalysis A: General, 113
(1994a) 89-102

Abdalla BK and Elnashaie SS, AIChE Journal, 40 (1994b) 2055-2059

Abdalla BK and E1nashaie SS, Journal ofMembrane Science, 101 (1995) 31-42

Acharya DR and Hughes R, Canadian Journal of Chemical Engineering, 68 (1990)

89-96

Addiego WP, Estrada CA, Goodman DW et aI., Journal of Catalysis, 146 (1994) 407-414

AgarwaUa S and Lund CRF, Journal of Membrane Science, 70 (1992) 129-141

Ali JK, Newson EJ and Rippin DWT, Journal ofMembrane Science, 89 (1994a) 171-184

Ali JK, Newson EJ and Rippin DWT, Chemical Engineering Science, 49 (1994b)
2129-2134

Antoniazzi AB, Haasz AA and Strangeby PC, Journal of Nuclear Materials, 162-164
(1989) 1065

Aoki K, Yokoyama S, Kusakabe K et aI., Korean Journal of Chemical Engineering, 13

(1996) 530-537

Aramendia MA, Borau V, Jimenez C et aI., Journal of Catalysis, 151 (1995) 44-49

Armendariz H, Aguilar-Rios G, Salas P et aI., Applied Catalysis A: General, 92 (1992)

29-38

311
Annor IN, Applied Catalysis, 49 (1989) 1-25

Annor IN, Catalysis Today, 25 (1995) 199-207

Annor IN, Journal of Membrane Science, 147 (1998) 217-233

Arnold TH, Chemical Engineering, (October 30, 1961) 90-92

Athavale SN and Totlani MK, Metal Finishing, 87 (1989) 23-27

Athayde AL, Baker RW and Nguyen P, Journal of Membrane Science, 94 (1994)

299-311

Bart JCJ and Sneeden RPA, Catalysis Today, 2 (1987) 1-124

Basile A, Drioli E, Santella F et a!., Gas Separation and Purification, 10 (1996) 53-61

Becker YL, Dixon AG, Moser WR et aI., Journal of Membrane Science, 77 (1993)
233-244

Berteau P and Delmon B, Applied Catalysis, 70 (1991) 307-323

Bhattacharyya D, Bej SK and Rao MS, Applied Catalysis A: General, 87 (1992) 29-43

Bolt M and Zimmermann H, AIChE Spring Meeting, Houston, (1991), Paper 26b

Boot LA, Van der Linde SC, Van Dillen AJ et aI., Studies in Surface Science and
Catalysis, 88 (1994) 491-498

Boot LA, Van Dillen AJ, Geus JW et aI., Journal of Catalysis, 163 (1996) 195-203

312
 Boudart M and Hwang HS, Journal of Catalysis, 39 (1975) 44-52

l_ Bracht M, Alderliesten PT, Kloster R et al., Energy Conversion Management, 38 (1997)

I

S159-S164

Brandes EA and Brook GB, Smithells Metals Reference Book, 7th Edition, Butterworth-
Heinemann Ltd., London (1992), Chapter 13, 1-99

Brinker CJ and Scherer GW, Sol-gel Science, Academic Press, New York (1990) 1-40

Brito A, Arvelo R, Villarroel R et aI., React. Kinet. Catal. Lett., 54 (1995) 77-83

Brito A, Arvelo R, Villarroel R et al., Chemical Engineering Science, 51 (1996)

4385-4391

Bulenkova LF, Shimanskaya MV, Gryaznov VM et al., Chemical Abstracts, 92: 180393m
(1978)

Burggraaf AJ and Cot L, Fundamentals of Inorganic Membrane Science and Technology,

Elsevier, Amsterdam, The Netherlands (1996) 1-690

Buxbaum RE and Marker TL, Journal ofMembrane Science, 89 (1993) 29-38

Buxbaum RE and Kinney AB, Industrial and Engineering Chemistry Research, 35

(1996) 530-537

Cadus LE, Abello MC, Gomez MF et al., Industrial and Engineering Chemistry
Research, 35 (1996) 14-18

Cannon KC and Hacskaylo JJ, Journal ofMembrane Science, 65 (1992) 259-268

313
 Canizosa I and Munuera G, Journal of Catalysis, 49 (1977) 189-200

Casanave D, Giroir-Fendler A, Sanchez J et a!., Catalysis Today, 25 (1995) 309-314

Chaar MA, Patel D, Kung MC et a!., Journal of Catalysis, 105 (1987) 483

Chabot J, Lecomte J, Grumet C et a!., Fusion Technology, 14 (1988) 614

Champagnie AM, Tsotsis TT, Minet RG et a!., Journal of Catalysis, 134 (1992) 713-730

Chang Hand Saleque MA, Applied Catalysis A: General, 103 (1993) 233-242

Chang H and Saleque MA, Journal ofMolecular Catalysis, 88 (1994) 223-238

Chang WS, Chen YZ and Yang BL, Applied Catalysis A: General, 124 (1995) 221-243

Cheng S and Shuai X, AIChE Journal, 41 (1995) 1598-1601

Cheng YS and Yeung KL, Journal ofMembrane Science, 158 (1999) 127-141

Chinchen GC, Denny PJ, Jennings JR et a!., Applied Catalysis, 36 (1988) 1-65

Chung M, Han S, Park K et a!., Journal ofMolecular Catalysis, 79 (1993) 335-345

Church 1M and Joshi HK, Industrial and Engineering Chemistry, 43 (1951) 1804-1811

Ciavarella P, Moueddeb H and Dalmon J, Workshop: 'Applications and future

possibilities of catalytic membrane reactors', Tumhout (16-17 Oct 1997) 121-124

Clayson DM and Howard P, UK Patent, GB 2,190,397A (1987) 1-4

314
 Cohen RL, D' Amico JF et aI., Journal Electrochemical Society: Electrochemical Science
and Technology, 118 (1971) 2042

Cohen RL and West KW, Journal Electrochemical Society: Electrochemical Science and
Technology, 120 (1973) 502

Cohen RL and Meek RL, Plating, 63 (1976) 47

Collins JP and Way JD, Industrial and Engineering Chemistry Research, 32 (1993a)
3006-3013

Collins JP, Catalytic decomposition of ammonia in a membrane reactor, PhD Thesis,

Oregon State University, USA, (1993b) 1-196

Compagnie des Metaux Pn:cieux, French Patent 2,302,273 (1976) I

Coulter K, Goodman DW and Moore RG, Catalysis Letters, 31 (1995) 1-8

Courty P and Chauvel A, Catalysis Today, 29 (1996) 3

Craig RG and Dufallo 1M, CEP, (Feb 1979) 62-65

Craig RG and White EA, Hydrocarbon Processing, 59 (1980) 111-114

Cullis CF and Newitt EJ, Procedures Royal Society A, 237 (1956) 530-542

Cullity BD, Elements of X-ray diffraction, 2nd Edition, Addison-Wesley Publishing

Company Inc., Massachusetts, USA (1978) 1-555

315
 Cunningham J, AI-Sayyed GH, Cronin JA et ai., Journal of Catalysis, 102 (1986)
160-171

Curtiss CF and Hirschfelder JO, The Journal of Chemical Physics, 17 (1949) 550-555

De Boer JH and Visseren WJ, Catalysis Reviews, 5 (1971) 55-66

Deckman HW, Corcoran EW, McHenry JA et aI., Catalysis Today, 25 (1995) 357-363

De las Pozas C, Lopez-Cordero R, Gonzalez-Morales JA et ai., Journal of Molecular

Catalysis, 83 (1993) 145-156

Deng J, Cao Z and Zhou B, Applied Catalysis A: General, 132 (1995) 9-20

De Ninno A, Violante V and La Barbera A, Physical Review B, 56 (1997) 2417-2420

Drago RS and Jurczyk K, Applied Catalysis A: General, 112 (1994) 117-124

Duncan JB and Toor HL, AIChE Journal, 8 (1962) 38-41

Dunn RO, Schuette GF, Brinkmeyer FM et ai., Procedures De Witt: Petrochemical

Review, Houston (1992) 1

Edlund DJ, US Patent 5,259,870 (1993a) 1-10

Edlund DJ and Pledger WA, Journal ofMembrane Science, 77 (I 993b) 255-264

Edlund DJ, US Patent 5,393,325 (1995a) 1-14

Edlund DJ and McCarthy J, Journal ofMembrane Science, 107 (1995b) 147-153

316
r:
I

Edlund DJ, US Patent 5,498,278 (1996) 1-18

EgloffG and HulJa G, The Oil and Gas Journal, 41 (1942) 36-51

Elnashaie SS, Abdalla BK and Hughes R, Industrial and Engineering Chemistry

Research, 32 (1993) 2537-2541

Ertl G, Knozinger H and Weitkamp J, Handbook of Heterogeneous Catalysis, Volume 5,

VCH, Weinheim, Germany (1997) 2140-2165

Fairbanks DF and Wilke CR, Industrial and Engineering Chemistry, 42 (1950) 471-475

Faith WL, Keyes DB and Clark RL, Industrial Chemicals, 2nd Edition, John Wiley &
Sons Inc., New York (1957) 2-3

Feldstein N, Journal Electrochemical Society: Electrochemical Science and Technology,

121 (1974) 738-744

nd
Fogler HS, Elements of Chemical Reaction Engineering, 2 Edition, Prentice-Hall, New
Jersey (1992) 1-837

Ford FE and Perlmutter DD, Chemical Engineering Science, 19 (1964) 371-378

Forster HA, Hydrocarbon Processing, 52 (1973) 119-122

Franckaerts J and Froment GF, Chemical Engineering Science, 19 (1964) 807-818

Gallaher GR, Gerdes TE and Liu PKT, Separation Science and Technology, 28 (1993)
309-326

317
'lr=
Gavalas GR, Megiris CE and Narn SW, Chemical Engineering Science, 44 (1989)
1829-1835

L
l Gibson MA and Hightower JW, Journal of Catalysis 41 (1976) 420-430

Gil A, Ruiz P and Delmon B, Canadian Journal of Chemical Engineering, 74 (1996a)

600-605

Gil A, Ruiz P, Delmon B, Journal of Catalysis, 159 (1996b) 496-499

Gobina E and Hughes R, Journal ofMembrane Science, 90 (1994) 11-19

Gobina E, Hou K and Hughes R, Catalysis Today, 25 (1995a) 365-370

Gobina E, Hou K and Hughes R, Chemical Engineering Science, 50 (1995b) 2311-2319

Gobina E, Hou K and Hughes R, Journal ofMembrane Science, 105 (1995c) 163-175

Gobina E and Hughes R, Applied Catalysis A: General, 137 (1996) 119-127

GrashoffGJ, Pilkington CE and Corti CW, Platinum Metals Review, 27 (1983) 157-168

Gryaznov VM, Vedemikov VI and Gul'yanova SG, Trans. Kinet. Katal., 27 (1986a) 142

Gryaznov VM, Platinum Metals Review, 30 (1986b) 68-72

Gryaznov VM, Mishchenko AP and Sarylova ME, French Patent 2,595,093 (1986c) 1

Gryaznov VM, Serebryannikova OS and Serov YM, Applied Catalysis A: General, 96

(1993) 15-23

318
Haag WO and Tsikoyiarmis JG, US Patent 5,019,263 (1991) 1

Hakuli A, Kytokivi A, Outi A et aI., Journal o/Catalysis, 161 (1996) 393-400

Hao Y, Tao Land Zheng L, Applied Catalysis A: General, 115 (1994) 219-228

Happel J, Blanck Hand Hamill TD, I&EC Fundamentals, 5 (1966) 289-294

Hara S, Xu W, Sakaki K et aI., Industrial and Engineering Chemistry Research, 38

(1999) 488-492

HearIe JWS, Sparrow JT et aI., The use 0/ the scanning electron microscope, Pergamon
Press, Oxford, Great Britain, (1974) 1-278

Hino M and Arata K, Catalysis Letters, (1990) 1737-1740

Holleck GL, Journal 0/ Physical Chemistry, 74 (1970) 503-511

Hongbin Z, Guoxing X, linghua G et aI., Catalysis Today, 25 (1995) 237-240

Horkans J, Journal Electrochemical Society: Electrochemical Science and Technology,

130 (1983) 311-317

Hsieh HP, Bhave RR and Fleming HL, Journal o/Membrane Science, 39 (1988) 221-241

Hsieh HP, AIChE Symposium Series, 85 (1989) 53-67

Hsieh HP, Inorganic membranes for separation and reaction, Elsevier Science Publishing
Company, New York, USA (1996) 1-591

Hsiung TH, Christman DD, Hunter EJ et aI., AIChE Journal, 45 (1999) 204-208

319
!'!'
I

Huang L, Chen CS, He ZD et aI., Thin Solid Films, 302 (1997) 98-101

Hughes Rand Koon CL, Applied Catalysis A: General, 119 (1994) 153-162

Hunter JB, US Patent 2,773,561 (1956) 1

Hunter JB, Platinum Metals Review, 4 (1960) 130-131

Hurlbert RC and Konecny 10, The Journal of Chemical Physics, 34 (1961) 655-658

Husen PC, Deel KR and Peters WD, The Oil and Gas Journal, 69 (Aug 1971) 60-61

Hutson T, Skinner RD and Logan RS, Hydrocarbon Processing, 53 (1974) 133-135

Hwang ST and Kammermeyer K, Canadian Journal of Chemical Engineering, 44 (1966)

82-89

Ilias S, Su N, Udo-Aka UI et aI., Separation Science and Technology, 32 (1997) 487-504

Imai T, US Patent 4,418,237 (1983), UOP, 1-2

Imai T and Schmidt RJ, US Patent 4,886,928 (1989), UOP, 1

Ioannides T and Gavalas GR, Journal ofMembrane Science, 77 (1993) 207-220

Ishikawa A, Chiang TH and Toda F, Journal Chemical Society: Chemical

Communications, (1989) 764

Itoh N, AIChE Journal, 33, (1987) 1576-1578

320
 Itoh N and Govind R, AIChE Symposium Series, 85 (1989a) 10-17

Itoh Nand Govind R, Industrial Engineering Chemical Research, 28 (1989b) 1554-1557

Itoh N, Studies in Surface Science and Catalysis, 54 (1990) 268-269

Itoh N, Catalysis Today, 25 (1995a) 351-356

Itoh N, Machida T, Xu W et a!., Catalysis Today, 25 (1995b) 241-247

Itoh N and Sathe AM, Journal ofMembrane Science, 137 (1997) 251-259

Iwasa Nand Takezawa N, Bulletin Chemical Society Japan, 64 (1991) 2619-2623

James DH and Castor WM, Ullmann's Encyclopaedia Industrial Chemistry, 5th Edition,
25 (1994) 329-344

Jansen JC, Koegler JH, Van Bekkum H, et a!., Microporous and Mesoporous Materials,
21 (1998) 213-226

Jayaraman V, Lin YS, Pakala Met a!., Journal ofMembrane Science, 99 (1995a) 89-100

Jayaraman V and Lin YS, Journal ofMembrane Science, 104 (1995b) 251-262

Jebarathinam NJ, Eswaramoorthy M and Krishnasamy V, Applied Catalysis A: General,

145 (1996) 57-74

Jemaa N, Shu J, Kaliaguine S et a!., Industrial and Engineering Chemistry Research, 35

(1996) 973-977

Jeon GS and Chung JS, Applied Catalysis A: General, 115 (1994) 29-44

321
Jewett DN and Makrides AC, Trans Faraday Soc., 61 (1965) 932-939

Jia MD, Peinemann KV and Behling RD, Journal of Membrane Science, 82 (1993) 15-26

Jiang Z and Wang J, Separation Science and Technology, 33 (1998) 1379-1385

Johansson SAB, Campbell JL et aI., PIXE: A novel technique for elemental analysis, John
Wiley and Sons, Great Britain (1988) 1-347

Jorgensen LS, Nielsen PH and Lehrmann P, Catalysis Today, 25 (1995) 303

Jorgensen LS, Mogensen G, Lehrmann P et aI., 4th Workshop: 'Optimisation of catalytic

membrane reactor systems', Oslo (30-31 May 1997) 51-57

Julbe A and Guizard C, Journal ofMembrane Science, 77 (1993) 137-153

Jun C and Lee K, Journal ofMembrane Science, 157 (1999) 107-115

Jung SH, Kusakabe K, Morooka S et aI., Journal ofMembrane Science, 170 (2000) 53-60

Kameyama T, Dokiya M, Fukuda K et aI., Separation Science and Technology, 14 (1979)

953

Kameyama T, Dokiya M, Fujishige M et aI., Industrial and Engineering Chemistry

Fundamentals, 20 (1981) 97-99

Kameyama T, Dokiya M, Fujishige M et aI., International Journal Hydrogen Energy, 8

(1983) 5

Kanoun N, Astier MP and Pajonk GM, Applied Catalysis, 70 (1991) 225-236

322
Kanoun N, Astier MP and Pajonk GM, Journal of Molecular Catalysis, 79 (1993)
217-228

Keizer K, Leenaars AFM and Burggraaf AJ, Eur. Coli. on Ceramics in Adv. Energy Tech.
(1982) 367

Keuler IN, Preparation and Characterisation of Palladium Composite Membranes,

M.Eng Thesis, University of Stellenbosch, Stellenbosch, South Africa (l997a) 1-203

Keuler IN, Lorenzen L and Sanderson RD, Plating and Surface Finishing, 34 (1997b)
34-40

Keuler IN, Lorenzen L, Sanderson RD et aI., Nuclear Instruments and Methods in

Physics Research E, 158 (l999a), 678-682

Kikuchi E, Japanese Patent 88,294,925 (1988) 1

Kikuchi E, Catalysis Today, 25 (1995) 333-337

Kizling ME, Gustafsson J, Von Krusenstiema 0 et al., 4th Workshop: 'Optimisation of

catalytic membrane reactor systems', Oslo (30-31 May 1997) 197-201

Knapton AG, Platinum Metals Review, 21 (1977) 44-50

Kolb HJ and Burwell RL, Journal of the American Chemical Society, 67 (1945) 1084

Kusakabe K, Yokoyama S, Morooka Set a!., Chemical Engineering Science, 51 (1996)

3027-3032

Lewis FA, Platinum Metals Review, 38 (1994) 112-118

323
Li A, Liang Wand Hughes R, Workshop: 'Applications and foture possibilities of
catalytic membrane reactors', Turnhout (16-17 Oct 1997) 106-109

Li A, Liang W and Hughes R, Separation and Purification Technology, 15 (1999)

113-128

Li A, Liang W and Hughes R, Journal ofMembrane Science, 165 (2000) 135-141

Li ZY, Maeda H, Kusakabe K et ai., Journal ofMembrane Science, 78 (1993) 247-254

Lin Y, Wang I and Yeh C, Applied Catalysis, 41 (1988) 53-54

Linkov VM, Preparation and application of hollow fibre carbon membranes, PhD
Thesis, University of Stellenbosch, Stellenbosch, South Africa (1994) 1-126

Liu B, Cao Y and Deng J, Separation Science and Technology, 32 (1997) 1683-1697

Livage J, Henry M et aI., Prog. Solid State Chemistry, 18 (1988) 259-341

Loc LC, Gaidai NA, Kipennan SL et aI., Kinetics and Catalysis, 34 (1993) 451-455

Loc LC, Gaidai NA, Kipennan SL et a!., Kinetics and Catalysis, 36 (1995) 504-510

Loc LC, Gaidai NA, Kipennan SL et a!., Kinetics and Catalysis, 37 (1996) 790-796

Loweheim FA, Modern Electroplating, John Wiley & Sons, New York (1974) 342-357,
739-747

Lu M, Xiong G, Zhao H et ai., Catalysis Today, 25 (1995) 339

324
 Lundin J, Holmlid L, Menon PG et ai., Industrial and Engineering Chemistry Research,
32 (1993) 2500-2505

Maestas S and Flanagan TB, The Journal of Physical Chemistry, 77 (1973) 850-854

Makarova MA, Paukshtis EA, Thomas JM et ai., Journal of Catalysis, 149 (1994) 36-51

Makrides AC, The Journal ofPhysical Chemistry, 68 (1964) 2160-2169

Matsubara T, Saito Y, Yamakawa T et ai., Journal of Molecular Catalysis, 66 (1991)

171-181

Matsuda T, Koike I, Kubo N et ai., Applied Catalysis A: General, 96 (1993) 3

Matsui J, Sodesawa T and Nozaki F, Applied Catalysis, 51 (1989) 203-211

Matsui J, Sodesawa T and Nozaki F, Applied Catalysis, 67 (1991) 179-188

Matsumura Y, Hashimoto K and Yoshida S, Journal of Catalysis, 117 (1989) 135-143

Matsumura Y, Hashimoto K and Yoshida S, Journal of Catalysis, 122 (1990) 352-361

Matsumura Y, Hashimoto K and Yoshida S, Journal of Molecular Catalysis, 68 (1991)

73-86

Mendes FMT and Schmal M, Applied Catalysis A: General, 151 (1997) 393-408

McBride RB and McKinley DL, Chemical Engineering Progress, 61 (1965) 81

McCool B, Xomeritakis G and Lin YS, Journal of Membrane Science, 161 (1999) 67-76

325
McKinley DL, US Patent 3,350,845 (1967) 1

Millat J, Dymond JH and Nieto de Castro CA, Transport Properties of Fluids,

Cambridge University Press, Cambridge (1996) 296-308

Morooka S, Yan S, Yokoyama S el a!., Separation Science and Technology, 30 (1995)

2877-2889

Muhler M, Schiilze J, Wesemann M el a!., Journal of Catalysis, 126 (1990) 339-360

Muhler M, Schlagl R and Ertl G, Journal of Catalysis, 138 (1992) 413-444

Mulder M, Basic principles of membrane technology, 2nd Edition, Kluwer Academic

Publishers, The Netherlands (1996) 1-564

Nagamoto H and Inoue H, Journal Chemical Engineering Japan, 14 (1981) 377-382

Nagamolo H and Inoue H, Chemical Engineering Commun., 34 (1985) 315-323

Nam S, Lee S and Lee K, Journal ofMembrane Science, 153 (1999),163-173

Neely SD, US Patent 3,106,581 (1963) 1-10

-,

Newman FC, Industrial and Engineering Chemistry, 62 (1970) 42

Noble RD and Stem SA, Membrane Separations Technology: Principles and

Applications, 2, Elsevier Science Ltd., (1995) 1-66

Noda S, Hudgins RR and Silveslon PL, Canadian Journal of Chemical Engineering, 45

(1967) 294-299

326
Nourbakhsh N, Champagnie A and Tsotsis TT, AIChE Symposium Series, 85 (1989)
75-84

Oertel MM, Schmitz J, Weirich VW et aI., Chemical Engineering Technology, 10 (1987)

248

Ohashi H, Ohya H, Aihara Met ai., Journal ofMembrane Science, 146 (1998) 39-51

Ohuo I, Wakabayashi 0 and Haruyama S, Journal Electrochemical Society:

Electrochemical Science and Technology, 132 (1985) 2323-2330

Opitz W and Urbanski W, US Patent 2,861,106 (1958) 1-6

Osaka T and Takematsu H, Journal Electrochemical Society: Electrochemical Science

and Technology, 127 (1980) 1021-1029

Paglieri SN, Foo KY, Way JD et aI., Industrial and Engineering Chemistry Research, 38
(1999) 1925-1936

Parfenova NI, Polyakova VP, Gryaznov VM et aI., Chemical abstracts 100:1l269d

(1983)

Peachey NM, Snow RC and Dye RC, Journal of Membrane Science, 111 (1996) 123-133

Pearlstein F and Weightman RF, Plating, 56 (1969) 1158-1161

Peloso A, Moresi M, Mustachi C et aI., Canadian Journal of Chemical Engineering, 57

(1979) 159-164

Perra JA, Monzon A and Santamaria J, Journal of Catalysis, 142 (1993) 59-69

327
 Perona JJ and Thodos G, AIChE Journal, 3 (1957) 230-235

Perry RH and Chilton CH, Chemical Engineers' Handbook, 5th Edition, McGraw Hill,
Tokyo, Japan, 1973, Chapter 3,1-250; Chapter 4,1-75

Pujado PR and Vora BV, Hydrocarbon Processing, (1990) 65-70

Purohit MA and Gandhi AN, Indian Chemical Engineering, 17 (1975) 26

Quaranta NE, Martino R, Gambaro L et aI., Studies in Surface Science and Catalysis, 82
(1994) 811-818

Raich BA and Foley HC, Industrial and Engineering Chemistry Research, 37 (1998)
3888-3895

Raizada VK, Tripathi VS, Lal D et aI., Journal of Chemical Technology and
Biotechnology, 56 (1993) 265-270

Raman NK and Brinker CJ, Journal ofMembrane Science, 105 (1995) 273-279

Rao PR, Thangaraj A and Ramaswamy AV, Journal Chemical Society, Chemical
Commun., (1991) 1139-1140

Rao UR, Kumar R and Kuloor NR, Industrial and Engineering Chemistry Proc. Des.
Dev., 8 (1969) 9

Reddy GK and Rao PK, Catalysis Letters, 45 (1997) 93-96

Reidel Ie, The Oil and Gas Journal, 54 (1957) 166-171

Rezac ME, Koros WI and Miller SJ, Journal ofMembrane Science, 93 (1994),193-201

328
Rezac ME, Koros WJ and Miller SJ, Industrial and Engineering Chemistry Research, 34
(1995) 862-868

Rhoda RN, Transactions o/the Institute 0/ Metal Finishing, 36 (1959) 82-85

Rodina AA, Gurevich MA and Doronicheva NI, Russian Journal Physical Chemistry, 42
(1968) 959-960

Romero E, Rodriguez JC, Pena JA et aI., Canadian Journal o/Chemical Engineering, 74

(1996) 1034-1038

Roshan NR, Mishchenko AP, Polyakova VP et aI., Journal Less-Common Metals, 89

(1983) 423-428

Ross JRH and Xue E, Catalysis Today 25 (1995) 291-301

Rosynek MP, Koprowski RJ and DeliiSante GN, Journal o/Catalysis, 122 (1990) 80-94

Sakamoto Y, Chen FL, Kinari Yet aI., International Journal 0/ Hydrogen Energy, 210
(1996) 1017-1024

Sanfilippo D, Buonomo F, Fusco G et aI., Chemical Engineering Science, 47 (1992)

2313-2318

Sano Y, Mizukami F, Yagishita H et aI., Japanese Patent 06,099,044 (1994) 1

Saracco G, Veldsink JW, Versteeg GF et aI., Chemical Engineering Comm, 147 (1996)
29-42

329
 Sathe AM, Itoh Nand Tominaga M, Indian Chemical Engineering, Section A, 36 (1994)
131-135

Satterfield CN, Heterogeneous Catalysis in Practice, McGraw-Hili Book Company, New

York (1980) 313-369

Sheintuch M and Dessau RM, Chemical Engineering Science, 51 (1996) 535-547

Shiau C and Liaw S, Journal of Chemical Technology and Biotechnology, 53 (1992)

13-19

Shipley CR, Plating and Surface Finishing, 71 (1984) 92-99

Shirai M, Pu Y and Arai M et aI., Applied Surface Science, 126 (1998) 99-106

Shu J, Grandjean BPA, Van Neste A et aI., Canadian Journal of Chemical Engineering,
69 (1991) 1036-1060

Shu J, Grandjean BPA, Ghali E et aI., Journal ofMembrane Science, 77 (1993) 181-195

Shu J, Grandjean BPA and KaJiaguine S, Applied catalysis A: General, 119 (1994)
305-325

Shu J, Grandjean BPA and Kaliaguine S, Catalysis Today, 25 (1995) 327-332

Shu J, Adnot A, Grandjean BPA et aI., Thin Solid Films, 286 (1996a) 72-79

Shu J, Grandjean BPA, Kaliaguine S et aI., Journal Chem. Society Faraday Trans., 92
(1996b) 2745-2751

330
 Shu J, Bongondo BEW, Grandjean BPA et a!., Journal of Materials Science Letters, 16
(1997a) 294-297

Shu J, Grandjean BPA, Kaliaguine S et a!., Canadian Journal of Chemical Engineering,

75 (1997b) 712-720

Sivaraj CH and Kantarao P, Applied Catalysis, 45 (1988a) 103-114

Sivaraj CH, Reddy BM and Rao PK, Applied Catalysis, 45 (1988b) Lll

Sivaraj C, Srinivas ST, Rao VN et a!., Journal of Molecular Catalysis, 60 (1990)

L23-L28

Skakunova EV, Ermi!ova MM and Gryaznov VM, Izv. Akad. Nauk SSSR. Ser. Khim., 5
(1988) 986-991

Smimov VS, Gryaznov VM, Miropo!'skaya MA et a!., Soviet Patent 437,743 (1978) 1

Sorensen eM and Weber RS, Journal of Catalysis, 142 (1993) 1-17

Soria R, Catalysis Today, 25 (1995) 285-290

Steitwieser A, Heathcock CH, Kosower EM, Introduction to Organic Chemistry, 4th

Edition, MacMillan Publishing Company, New York (1992) 379-438

Suzuki H, US Patent 4,699,892 (1987) 1

Swift HE, Beuther H and Rennard RJ, Industrial and Engineering Chemistry Prod. Res.
Dev., 15 (1976) 131-136

SzabO ZG, Jover Band Ohmacht R, Journal of Catalysis, 39 (1975) 225-233

331
r
I
Takezawa N, Hanamaki C and Kobayashi H, Journal of Catalysis, 38 (1975) 101-109

Thaller LHand ThodosG, AIChE Journal, 6 (1960) 369-373

Teo WK and Ti HC, Applied Biochemistry and Biotechnology, 24/25 (1990) 521-532

Tiscareno-Lechuga F, Hill CG and Anderson MA, Applied Catalysis A: General, 96

(1993) 33-51

Tohji K, Udagawa Y, Mizushima T et aI., Journal Physical Chemistry, 89 (1985) 567l

Toor HL, AIChE Journal, 3 (1957) 198-207

Tsapatsis M, Kim S, Nam SW et aI., Industrial and Engineering Chemistry Research, 30

(1991) 2152-2159

Tsotsis TT, Champagnie AM, Vasileiadis SP et aI., Chemical Engineering Science, 47

(1992) 2903-2908

Tu Y, Chen Y and Li C, Journal ofMolecular Catalysis, 89 (1994a) 179-190

Tu Y, Li C and Chen Y, Journal of Chemical Technology and Biotechnology, 59 (1994b)

141-147

Udomsak S and Anthony RG, Industrial and Engineering Chemistry Research, 35 (1996)
47-53

Uemichi Y, Shouji K, Sugioka M et aI., Bulletin Chemical Society Japan, 68 (1995)

385-387

332
 Uemiya S, Matsuda T and Kikuchi E, Journal ofMembrane Science, 56 (1991a) 315-325

Uemiya S, Sato N, Ando H et aI., Journal of Membrane Science, 56 (1991 b) 303-313

Uemiya S, Sato N, Ando H et aI., Industrial and Engineering Chemistry Research, 30

(1991c) 585-589

Uemiya S, Sato N, Ando H et aI., Applied Catalysis, 67 (1991d) 223

Uhlhorn RJR, Keizer K, Burggraaf AJ, Journal of Membrane Science, 46 (1989) 225-241

Uhlhorn RJR, Keizer K and Burggraaf AJ, Presented at the 1990 International Congress
on Membranes and Membrane Processes (ICOM '90), Chicago, lL, Session 25, 451

Ullmann, Ullmann's Encyclopaedia of Industrial Chemistry, A4, Fifth Revised Edition,

VHC Publishers, Germany (1987) 483-494

Uzio D, Peureux J and Giroir-Fendler A, Applied Catalysis A: General, 96 (1993) 83-97

Van den Meerakker JEAM, Applied Electrochemistry, 11 (1981) 395

Van den Tillaart JAA, Kuster BFM and Marin GB, Applied Catalysis A: General, 120
(1994) 127-145

Van den Tillaart JAA, Kuster BFM and Marin GB, Catalysis Letters, 36 (1996) 31-36

Vitulli G, Pitzalis E, Salvadori P et aI., Catalysis Today, 25 (1995) 249-253

Ward TL and Dao T, Journal ofMembrane Science, 153 (1999) 211-231

Weaver JA and Metzner AB, AIChE Journal, 12 (1966) 655

333
 Welch LM, Groce LJ and Christmann HF, Hydrocarbon processing, 57 (1978) 131-136

Weyten H, Prevost I, Duxbury J et aI., Workshop: 'Applications and futurepossibilities

of catalytic membrane reactors " Turnhout (16-17 Oct 1997) 41-44

Wilke CR, Chemical Engineering Progress, 46 (1950) 95-104

Williams C, Makarova MA, Malysheva LV et aI., Journal of Catalysis, 127 (1991)

377-392

Wu JCS and Liu PKT, Industrial and Engineering Chemistry Research, 31 (1992)
322-327

Xomeritakis G and Lin YS, Journal ofMembrane Science, 120 (1996) 261-272

Xomeritakis G and Lin YS, Journal afMembrane Science, 133 (1997) 217-230

Xu W,Yin Y, Li G et a!., Applied Catalysis A: General, 89 (1992) 131-142

Yan S, Maeda H, Kusakabe K et aI., Industrial and Engineering Chemistry Research, 33

(1994) 616-622

Yang BL, Kung MC and Kung RH, Journal of Catalysis, 89 (1984) 172-176

Yang BL, Cheng DS and Lee SB, Applied Catalysis, 70 (1991) 161-173

Yang C, Xu N and Shi J, Industrial and Engineering Chemistry Research, 37 (1998)

2601-2610

Yang W, Wu J and Lin L, Catalysis Today, 25 (1995) 315-319

334
Yashima T, Suzuki H and Hara N, Journal of Catalysis, 33 (1974) 486-492

Yeung KL, Aravind R, Zawada RJX et aI., Chemical Engineering Science, 49 (1994)
4823-4838

Yeung KL, Sebastian 1M and Varma A, Catalysis Today, 25 (1995a) 231-236

Yeung KL and Varma A, AIChE Journal, 41 (1995b) 2131-2139

Yoshida H, Konishi S and Naruse Y, Journal Less-Common Metals, 89 (1983) 429

Zaspalis VT, Keizer K, Van Praag W et aI., Brit. Ceram. Soc. Proc., 43 (1989) 103

Zaspalis VT, Van Praag W, Keizer K et aI., Applied Catalysis, 74 (1991) 223-234

Zhao HB and Baron GV, Workshop: 'Applications and future possibilities of catalytic
membrane reactors', Turnhout (16-17 Oct 1997) 52-58

Zhao HB, Pflanz K, Gil JH et aI., Journal of Membrane Science, 142 (1998) 147-157

Zhu Y, Minet RG and Tsotsis TT, Catalysis Letters, 18 (1993) 49-50

Ziaka ZD, Minet RG and Tsotsis TT, Journal of Membrane Science, 77 (1993) 221-232

335
 LIST OF SYMBOLS

A' = pre-exponential factor for reaction rate constant [mollkg cat.s.Pa]

Am = surface area of metal film [m 2]
Ap = external surface area of the catalyst [m 2]
C = concentration [mol/m 3]
C1 constant for determining Icc
C = total molar density of reaction side gas [mol/m 3]
C WP = Weisz-Prater parameter
D = coefficient for diffusion through a membrane film [m2/s]
Do rre-exponential factor of diffusion coefficient [m2/s]
DA,m = gas phase diffusivity of A in bulk mixture [m2/s]
DAB = binary diffusion coefficient of A in B [m2/s]
De = effective diffusivity [m%]
Dhkl = Schearer's particle diameter [Angstrom]
Dp = catalyst particle diameter [m]
E = activation energy for the rate constant [J/mol]
ED = diffusion activation energy [J/mol]
F = flow rate on the tube side of reactor [mol/s]
gc = conversion factor in Ergun equation, 1 for SI units
G empty column mass velocity [kglm 2 s]
GO = Gibbs free energy [J/mol]
Gf geometric factor
H = enthalpy [J/mol]
J = flux [mol/m 2 ,s]
k = reaction rate constant [mol/m 3.Pa,s]
k' = reaction rate constant [mol/kg cat.Pa.s]

kc = mass transfer coefficient [mls]

K = adsorption coefficient [Pa- I ]

Ke q = equilibrium constant [Pal

Kp = equilibrium constant [bar]

336
I = film thickness [m]
L = distance from reactor inlet [m]
Lo = reactor length [m]
m = catalyst mass [kg]
M = molecular mass [g/mol]
n = pressure exponent or reaction order
nm = number ofN20 molecules that decomposes
ns = eu metal atoms per surface area [1.47*10- 19 m- 2]
P = pressure [Pal
Po = pre-exponential factor of permeability coefficient [moLmlm2 Pa,s]
Per = permeability coefficient [moLmlm 2 ,Pa,s]
Pm = permeance [mol/m 2 ,Pa,s]
P, = total pressure [Pal

Q = flow rate on the shell side of the reactor [molls]

r = dimensionless radius Cr' IRp)
r' = radial position in catalyst [m]
rA = rate of generation for component A in reaction [mollm 3,s]
r'A = rate of generation for component A in reaction [mol/kg cat.s]

Ro = Universal constant [8.314 J/mol.K]

Re = Reynolds number
Rm = inner radius of membrane tube [m]
Rp = radius of catalyst particle [m]
S = entropy [J/moLK]
Sc = Sievert's constant
Sc = Schmidt number
SH = metallic surface area [m2]
Sh = Sherwood number
T = temperature [K]
U = empty column velocity [mls]
V = reactor volume [ml]

W = catalyst mass [kg]

337
I"~

Xp = peak width [radians]

XA = conversion of A
Xrir) = normalised mole fraction of j in catalyst (Y/Yj,b)
Xm = chemisorption stoichiometry
Yj = mole fraction of j in catalyst
Y.b
J, = mole fraction of j in bulk fluid

Greek symbols
a = separation factor
eb = void fraction of packed bed

11 = overall effectiveness factor

P = fluid density [kglm 3 ]

Pb = bulk density of catalyst bed [kg/m3]

3
Pp = catalyst particle density [kglm ]

't = space time [s]

J.L = viscosity [kglm.s]

Vi = stoichiometric coefficient for component i in reaction
u = J.LIp = kinematic viscosity [mlls]
y = shape factor for Thoenes-Kramers correlation

SubscriI!ts
0 = inlet conditions
2But = 2-butanol
A = feed (ethanol or 2-butanol depending on reaction)
Ac = acetaldehyde
b = at bulk conditions
Et = ethanol
Hl = hydrogen
= component i

338
J = componentj
m = mixture properties
MEK = methyl ethyl ketone
Nz = nitrogen
obs = observed
R = main products (acetaldehyde or MEK depending on reaction)
s = on the shell side ofthe reactor
surf = conditions at the catalyst surface
S = hydrogen
t = on the tube side of the reactor
W = water
x = high pressure side
y = low pressure side

Note:
• The reaction rates and reaction rate parameters (k, K and r) described in Chapter 2 were
taken from published work in their original fonnat and not transfonned to standard units.
In each case the published units were stated.
• For kinetic testing in Chapter 5, the reaction rates were expressed in (mollkg.cat.h) and
the parameters used kPa instead of Pa for their units. Once the parameters were
detennined in kPa, it was changed to Pa for modeling in Chapters 8 and 9.

339
/
I

I LIST OF ABBREVIATIONS

i AA atomic adsorption
BET Brunauer-Emmett-Teller
CNRS Centre National de la Recherche Scientifique
CVD chemical vapour deposition
dc direct current
ECN Energy Corporation of the Netherlands
FID flame ionisation detector
GC gas chromatograph
HP Hewlett Packard
ICP inductively coupled plasma
IRC Institut de Recherches sur la Catalyse
MEK methyl ethyl ketone
PIXE particle induced X-ray emission
SCT Societe des Ceramiques Techniques
SEM scanning electron microscope
TCD thermal conductivity detector
TEM transmission electron microscope
TOC total organic carbon
XRD X-ray diffraction

340
r

APPENDIX A

List of chemicals used

341
r
I
I

Name Purity Supplier

Ammonia solution 25 wt% Saarchem
AgN0 3 99.9999% Aldrich
2-Butanol > 99% (0.1 % HzO) Riedel-de-Haen
Co(N03)z.6H zO >99% Merck
Cr(N03)3. IOHzO >97% Fluka
Cu(N03)z.3H zO >99% ACE
Ethanol 99.8% (0.02 % HzO) Merck
Hydrazine 35 wt% solution Aldrich
MgO >97% Merck
NazEDTA.2HzO >99% Saarchem
(NH3)4PdC/z.HzO 99.99% Aldrich
Ni(N03)z.H zO 98% Saarchem
PdC/z 59% Merck
SnC/z.2HzO >98% Fluka

342
 APPENDIXBl

Ethanol and 2-butanol response

factors as determined with a
HPG1800AGC

343
Apparatus: HP GI800A GC and
50 m capillary column (50QGI.5IBPI PONA from SGE)

This apparatus was used for all experiments where the catalyst composition was
optimised (Chapter 4) and for membrane reactor experiments (Chapters 8 and 9). The
response factors discussed in Appendix Bl are only for the work in those two chapters.

The way to interpret the response factors is as follows:

,
The GC area % of the other component must be calculated relative to the ethanol area %.
The response factor is then used to convert the GC area % of the other component to the
true mass % of the other component relative to the ethanol mass %. This is done for
every component in the mixture and then the true mass percentages are normalised so that
they add up to 100%. Mass percentages are then converted to mole percentages to
calculate conversions.

Ethanol-acetaldehyde response factors were determined from liquid mixtures kept at

about 4°C (in ice water) to keep acetaldehyde in the liquid phase. Ethanol-ethene-
mixtures were prepared by filling the reactor in Figure 3.8 at 200°C (without the
membrane in it) with ethene at atmospheric pressure. The reactor was purged for several
minutes to remove air in the reactor and to ensure that there was only ethene in the
reactor. Thereafter, all reactor inlets and outlets were closed. The one reactor outlet was
modified to allow ethanol injections into and gas extractions from the reactor cavity
through a sealed septum. Care was taken to heat and insulate the whole system to
prevent any condensation of the ethanol. Ethanol was injected into the ethene cavity to
obtain different ethanol-ethene mass ratios and the mixtures were then analysed with the
GC to calculate the response factors.

344
Ethanol response factors:

GC response factors for binary mixtures

with ethanol
100
90
80

...
~
70
60
~

•E 60
,
~ 40
~ 30
20
10
0
0 ~ • 80 80 100
GC area % of other component

GC response factors for ethene-ethanol

100
90
~ 80
~
~
.= 70

...".. 80
Ethane

.... 50
40
E

-
~
30. y = -2.72E.07x 3 • 4.61E-03x 2 -+ 1.46E+OOx -+ 7.87E-01

..." ~ . R2 = 9.gsE.Ol

10
0
0 20 40 80 80 100
GC area % ethene

Excluding acetaldehyde, di-ethyl ether and ethene, the majority of other by-products that
formed for all ethanol catalyst optimisation experiments (Chapter 4) were below 5 GC
area %. Exceptions were for MgO catalysts, where the following maximums were
detected in different experiments:
propene: max. 10 GC area %
butenes: max. 15 GC area %
I-butanol: max. 16 GC area %

345
For some Cu on silica optimisation experiments (for ethanol dehydrogenation), up to 12
GC area % propene and 9 GC area % I-butanone were detected. For Si0 2 and MgO
supports, those high percentages were detected at and above 400°C and at very low
ethanol feed flow rates.

For area percentages below 5%, the response factors were determined or estimated. The
measured response factors correlated reasonably well with the molecular mass ratio
between ethanol and the other component. This observation was used as a very rough
approach to estimate the other response factors as indicated.

Correlations for low GC area % (0 to 5% range)

Component Mr(ethanol)/ Measured Factor to convert
Mr(component) (mass %)/(GC area % to true
area %) mass %
Ethene 1.65 1.65 * graphic fit
Propene 1.1 ? 1.1
Butenes 0.8 ? 0.8
Acetaldehyde I 1.2 * graphic fit
Acetone 0.8 0.8 0.8
Di-ethyl ether 0.6 0.5 * graphic fit
Etoxi-ethene 0.6 ? 0.6
MEK 0.6 0.6 0.6
I-butanone 0.6 0.6 0.6
I-butanol 0.6 0.55 0.6
Ethyl acetate 0.5 ? 0.5
Heavier <0.5 ? 0.5
products
* mterpolated from measured data presented m the figures
? values were not measured

346
The error that resulted from this simplification will have little if any effect on the
membrane reactor work and model development. The error was restricted to the catalyst
optimisation stage, where the same calculation procedures were used for all catalysts
tested. Data obtained during the catalyst optimisation stage were not used for
fundamental calculations and model development, but only used to compare catalysts.
Since the error will be similar for the different catalysts tested, the simplifications in the
response factors will not have a significant effect in identifying the optimum catalysts.

For membrane reactor work, the acetaldehyde selectivities were above 80% and no
alkanes or heavy products formed. Ethyl acetate was the only component of more than 2
GC area %, which did not have a measured response factor. Even a very large difference
between the estimated response factor and the true response factor for ethyl acetate, will
have a negligible effect on the total ethanol conversion and the acetaldehyde yield.

2-Butanol response factors:

GC response factors for binary mixtures

with 2·butanol

60

...
W 60 MEK

~
~ 40
2
I-
20

o 20 40 60 eo 100
Gearea %

The 2-butanol-butene response factors were determined in the same way as that of the
ethanol-ethene response factors. MEK and butenes were the only products that formed in
the catalyst optimisation experiments. For membrane experiments, 3-octanol was the
only by-product that formed, with a maximum value of up to 3.5 GC area %. The
response factor was taken as 0.5.

347
 APPENDIXB2

Ethanol and 2-butanol response

factors as determined with a
HP 5850 GC

348
Apparatus: HP 5850 GC equipped with a FID detector and
30m HP Innowax and a 30m HP Plot/AhO) columns in series

This apparatus was used for al experiment where the kinetics of the optimised catalysts
were investigated (Chapter 5).

Ethanol response (actors:

Acetaldehyde-ethanol response for

acetaldehyde area% < 10%
18r------------------------------------------,
16
y = 4.82E~03x3 - 4.62E-02x' + 1.35E+OOx + 8.97E·02
~ 14
R2 = 9.96E-01
E 12
., 10
~
.g
8
6
~u 4
« 2
o~~-- __----__----__----__- - - - - - - - - - - J
o 2 4 6 6 10 12

Acetaldehyde area%

Acetaldehyde-ethanol response for

acetaldehyde> 10 area%
100
90
•
SO
..E
en
en
70
., 60
,..
"C

..
.s::.
"C
50
40

-..
ii

~
30
20
y = -1.21 E-Q4x' + 1.29E-02x' + S.72E-01x + 3.96E+OO
R' = 9.93E-01

10
0
10 20 30 40 50 60 70 SO 90 100
Acetaldehyde area%

349
For the kinetic experiments, the selectivity of the reaction towards acetaldehyde
formation was above 97%. The ethanol-acetone response factor was measured as I. All
response factors excluding ethanol-acetaldehyde were taken as I; in other words, the true
mass % was taken as the GC area %.

2-Butanol response factors:

The 2-butanol-MEK response factor was equal to lover the entire range of MEK GC
area percentages. The selectivity of the reaction towards MEK formation was above 98%.
All either response factors were taken as I; in other words, the true mass % was taken as
the GC area %.

Note:
For all gas samples, i.e. for determining response factors as well as for all reactor
experiments, a heated syringe was used for injections into the GC. A heated syringe was
used to prevent condensation of the products in the syringe. A very small tube furnace
was made by wrapping heating wire around a ceramic tube and insulating it. The syringe
was placed inside the tube prior to taking a gas sample and the temperature inside the
tube was controlled at 110°C. Only the bottom halve of the syringe was heated to still be
able to handle it at the top end. A very small gas sample was extracted (about one
fifteenth of the syringe volume). This procedure gave very reproducible results when the
same gas mixture was injected into the Gc. When the syringe was not properly heated,
there was significant variation in the results due to condensation of some of the products
in the syringe.

For ethanol-acetaldehyde mixtures (liquid), the syringe was not cooled, but used at room
temperature when the response factors were detennined. At high acetaldehyde
concentrations (> 80%) some flashing of acetaldehyde from the syringe might have
occlUTed and the true response factors might differ marginally from the measured values.

350
 APPENDIXCI

Equations for determining ethanol

dehydrogenation kinetics

351
The linear equations determined by least square analysis at each temperature are:

At T = 200 °c:
For an ethanol-N2 feed:
y= 7.21 *10-3x + 1.02

For an ethanol-H2 feed:

y = 3.70*10-5 x + 1.80 (R2=6.77*10-5)

For an ethanol-acetaldehyde feed:

y = -1.14*1O- l x + 13.3 (R2=O.974)

AtT =225 °c:

For an ethanol-Nz feed:
y = 4.37*1O- 3x + 1.05 (R2=0.988)

For an ethanol-H2 feed:

2
y = -3.l4*1O-3x + 1.75 (R =0.428)

For an ethanol-acetaldehyde feed:

y = -1.07*10-lx + 11.8 (R2=0.9l8)

AtT = 250°C:
For an ethanol-N2 feed:
y= 3.76*1O-3x+0.73l

For an ethanol-H2 feed:

y = 1.32* 10·3 x + 0.963 (Rz=0.667)

352
For an ethanol-acetaldehyde feed:
y = -4.76*1O-zx + 5.78

AtT = 275°C:
For an ethanol-Nz feed:
y = 2.02* I 0-3 X + 0.468

For an ethanol-H2 feed:

y = -6.95*10·4 x + 0.609

For an ethanol-acetaldehyde feed:

y = -2.89*1O-zx + 3.55

AtT =300 °C:

For an ethanol-N2 feed:
y = 2.09*1O-3 x + 0.383

For an ethanol-H2 feed:

y = -8.1 2* 1O-4 x + 0.538

For an ethanol-acetaldehyde feed:

y = --2.95* 1O-2x + 3.52

K-values
The K-values were detennined using both the gradients and the y-axis intercepts of the
fitted lines. The values are summarised in Table Cl.

353
Table Cl: K-values for ethanol dehydrogenation
T (0C) Ks(I) Ks(2) Ks KR(I) K R(2) KR
1 1 1 1 1
kPa- kPa- kPa- kPa- kPa- kPa- 1
200 0.00772 0.00732 0_00752 0.1182 0.1212 0_1197
225 0.00660 0.00715 0.00688 0.1014 0.1061 0.1038
250 0.00314 0.00334 0.00324 0.0684 0.0703 0.0694
275 0.00298 0.00283 0.00291 0.0652 0.0661 0.0657
300 0.00401 0.00334 0.00368 0.0811 0.0825 0.0818

354
 APPENDIXC2

Equations for determining 2-butanol

dehydrogenation kinetics

355
The linear equations detennined by least square analysis at each temperature are:

AtT = 190°C:
For an 2-butanol-Nz feed:
y = 2.461 *10-3x + 1.717 (RZ = 0.452)

For an 2-butanol-Hz feed:

y = 4.745*1O-3x + 1.501 (RZ = 0.935)

For an 2-butanol-MEK feed (0-20.5 wt% MEK):

y = -1.991 *1O. 1x + 22.06 (R z = 0.993)

AtT = 220°C:
For an 2-butanol-Nz feed:
y = 2.213*10-3x + 1.138 (R Z = 0.575)

For an 2-butanol-Hz feed:

y = 3.574*10·3x + 0.986 (RZ = 0.800)

For an 2-butanol-MEK feed (0-63.5 wt% MEK):

y = -6.686*1O-zx + 8.125 z
(R = 0.994)

AtT = 250°C:
For an 2-butanol-Nz feed:
y = 2.282*10-3x + 0.764 (RZ = 0.636)

For an 2-butanol-Hz feed:

y = 3.845*10-3x + 0.575 (R2 = 0.902)

356
For an 2-butanol-MEK feed (0-41.0 wt% MEK):
y = -3.824*10·zx + 4.745 (Rz = 0.955)

AtT = 280°C:
For an 2-butanol-Nz feed:
y = 1.764*10·3x + 0.547 (RZ = 0.874)

For an 2-butanol-H z feed:

y = 2.402* 10-3 x + 0.491 (RZ = 0.695)

For an 2-butanol-MEK feed (0-41.0 wt% MEK):

y = -3.379*1O-zx + 4.063 (Rz = 0.980)

K-values
The K-values were determined using both the gradients and the y-axis intercepts of the
fitted lines. The values are summarised in Table C2.

Table C2: K-values for 2-butanol dehydrogenation

TeC) Ks(I) Ks(2) Ks KR(I) K R(2) KR

1 1
kPa- kPa- 1 kPa- 1 kPa- kPa- 1 kPa- 1

190 -0.00125 -0.00133 -0.00129 0.11730 0.11740 0.11735

220 -0.00132 -0.00120 -0.00126 0.06079 0.06070 0.06075
250 -0.00245 -0.00205 -0.00225 0.05159 0.05304 0.05232
280 -0.00101 -0.00117 -0.00109 0.06364 0.06500 0.06432

357
 APPENDIXDI

Hydrogen and nitrogen permeance

data for Pd films

358
Membrane Thickness T Avg.H2 Avg. N2 Selectivity Arrhenius parameters for R value for n-I
(micron) ('C) penneance penneance hydrogen penneance in permeance
(J.Ullollm1pa.s) (nmollm2.Pa.s) equation

1a 3.08 450 7.68 31.95 240 Po 5.243E-1O 0.9995

410 6.73 33.45 201 Eo- 18485 0.9994
370 5.45 34.13 160 R- 0.9918 1
330 4.19 35.19 117 0.9853

1b 3.86 450 6.44 9.67 666 Po- 1.956E-l0 0.9908

410 4.93 9.99 493 En 12737 0.9994
370 4.75 10.22 464 R'~ 0.8981 0.9981
330 4.06 10.80 376 0.9997

1c 6.19 450 5.14 5.97 860 Po - 5.578E-10 0.9993

410 4.21 6.21 678 En 17282 0.9995
370 3.59 6.21 578 R'· 0.9903 0.9992

2a 1.43 450 6.90 9.61 718 Po 7. 127E-1O 0.9963

410 4.53 9.61 472 ED - 26182 0.9994
370 3.59 9.87 363 R' 0.9601 0.9994
330 2.82 9.98 283 0.9994

2b 1.43 450 11.67 12.01 971 Po 3.205E-10 0.9951

410 9.93 12.50 794 En- 17799 0.9976
370 7.65 12.81 597 R- 0.9832 0.9998
330 6.60 13.25 498 0.9968

3a 2.40 450 8.37 5.72 1465 Po 1.752E-IO 0.9997 I

410 6.88 6.00 1147 En 13223 0.9993

370 5.97 6.28 952 W· 0.9643 0.9997
330 5.37 6.40 839 0.9987
--

359
Membrane Thickness T Avg. H, Avg.N, Selectivity Arrhenius parameters for IV value for n-I
(micron) (0C) penneanee permeance hydrogen permeance in penneance
2
(f1II1oUm' .Pa.s) (nmoUm Pa.s) equation

3b 4.43 450 4.86 1.13 4297 Po - 1.697E-1O 0.9984

410 3.93 1.19 3306 Eo~ 12673 0.9748
370 3.47 1.21 2869 R'- 0.9424 0.9866
330 3.16 1.45 2179 0.9998
5 1.00 450 15.00 65.67 228 Po - 7.504E-II 0.9983
410 12.96 68.62 189 Eo- 9778 0.9986
370 12.32 70.54 175 R"~ 0.9676 0.9969
330 10.64 70.44 151 0.9997

6 1.47 450 11.30 38.26 295 Po - 8.178E-II 0.9987

410 10.13 39.36 257 Eo- 9654 0.9999
370 9.27 39.88 232 R"- 0.9969 0.999\
330 8.16 39.25 208 0.9988

8 1.10 450 14.37 21.40 671 Po - 7.779E-II 0.9992

410 13.70 21.90 627 Eo- 9435 0.9989
370 12.40 22.40 553 R'- 0.9642 0.9996
330 10.54 23.10 457 0.9997

NI 1.12 450 13.59 3.00 4533 Po - 1.238E-1O 0.9999

410 12.61 3.00 4251 Eo- 12483 0.999
370 10.75 2.80 3784 R"- 0.9872 0.9995
330 9.08 3.10 2953 0.9995

360
Membrane Thickness T Avg. H, Avg. N, Selectivity Arrhenius parameters for R value for n-I
(nticron) (0C) penneance permeance hydrogen permeance in permeance
(l1fIloVm' .Pa.s) (nmoVm'.Pa.s) equation
N2 1.08 450 15.11 20.80 728 Po - 5.267E-II 0.998
410 14.37 19.50 737 En- 6984 0.9994
370 13.38 18.10 738 R'· 0.9857 1
330 11.99 17.20 698 0.9999

N3 1.47 410 10.81 5.20 2090 Po 1.118E-1O 0.9992

370 9.65 5.20 1851 En 11065 0.9999
330 8.35 5.60 1479 R'- 0.9989 0.9998

N4 1.18 450 13.84 20.60 671 Po - 8. 137E-II 0.9996

410 13.03 20.60 631 En 9541 0.9986
370 11.85 20.10 591 R'- 0.9745 0.9992
330 10.09 21.80 462 I

N7 1.23 450 8.28 7.40 1119 Po - 9.2IIE-II 0.9996

410 7.41 7.30 1021 En 13164 0.9997
370 6.51 7.50 865 W· 0.9926 0.9998
330 5.34 7.80 683 0.9998

N8 1.19 450 10.89 2.30 4830 Po 5.815E-1\ 0.9999

410 10.05 2.20 4529 En- 8993 0.9989.
370 9.19 1.70 5364 R'· 0.9958 0.9999
330 8.07 2.20 3638 0.9997

361
 Name: la (3.08 pm)

Mass (g) (mm) H2 permeance parameters

Clean 29.6650 Plated length 240 Po (mol.m1m".Pa.s) 5.243E-I0
After - Permeable length 230 ED (J/mol) 18485
pretreatment Inside diameter 7 R" for Arrhenius fit 0.9918
AfterPd 29.8607

Hydrogen perrneance in micromoleJPamA 2.s (1a)

Selecitivity data for 1a

36.0 250

~ .' - -" --
-
35.0
---
-----
220

~
C
0:
•
34,0
~ --
:: ...D.:
-,,- --
EE
~i5
E
330

---.-
---- - ------- ~
__ Nftrogen permsanco
Z S 32.0
,,- -
"
--- -.-. Hydrogen/Nitrogen
''''
--
31.0
selectivity

30.0 100
330 370 410 450
T('C)

362
 Name: Ib (3.86 Jim)

Mass (g) (mm) Hz permeance parameters

Clean 29.6650 Plated length 230 Po (moLmlm·.Pa.s) 1.956E-1O
After - Permeable length 230 ED (J/mol) 12737
pretreatment Inside diameter 7 R' for Arrhenius fit 0.8981
AfterPd 29.9100

Hydrogen penneance in micromoleJPa.m A 2.s (1b)

Selecitivity data for 1b

11.0 -----.----------.. ------- ---------_.----_.-------------.----------.+-------------_._----.. 700

10.6~
,,,~~ 650

t--~--=::".-=:::::::::===-7'---'-'-1:·"
t:e:I ~
; 10.2
/
:t::
>

ii
z.s
9.8
/
---------------~
-. .,
J __ . :

... ~ -+- Nitrogen permeance

/- 400
9.4 ~"-------j f---j
-.-, HydrogenlNitrogen 350
selectMty

330 370 410 450

T('CI

363
 Name: Ie (6.19 Jim)

Mass (g) (mm) H2 permeance parameters

Clean 29.6650 Plated length 230 Po (mol.mlm".Pa.s) 5.578E-I0
After - Permeable length 230 ED (J/mol) 17282
pretreatment Inside diameter 7 R· for Arrhenius fit 0.9903
AfterPd 30.0577

Hydrogen penneance in micromolelPa.m A 2.s (ic)

Selecitivlty data for 1c

... _ - - ----.-- -- -- ----------------.--------------..----------- 900

'.3
BOO

.n'."
u
u
c
•
•
~
'.1

q 5."
zE.,
600
5.7
••-. HydrogenINitrogen
r-- 550
selectivity
5.5 l-------------__============~~~o
370 410
T('C)
...

364
 Name: 2a (1.43 Jim)

Mass (g) (mm) H2 permeance parameters

Clean 29.2217 Plated length 250 Po (moLmlm'.Pa.s) 7.127E-10
After - Permeable length 230 ED (J/mol) 26182
pretreatment Inside diameter 7 R' for Arrhenius fit 0.9601
AfterPd 29.3085

Hydrogen perrneance in micromoleJPa.mAZ.s (Za)

Selecitivity data for 2a

10.1 750
__ Nitrogen permeance
--
10.0
."
. •••
: .,
8c ..
~

------
- .... Hydrogenn-litrogen

----...".
selectivity

-
----
- ... ~>
n '.8 '. '. ~.~ ...
-- II
•
it • " , / ...,;lj
" E
zs. 9.7
_t
- --- .' ..,
".
. J5()
9.'
---
9.'
330 370 '10 ...,250
TI"C)

365
 Name: 2b (1.43 J.lm)

Mass (g) (rom) Hz permeance parameters

Clean 29.2217 Plated length 250 Po (mol.m!mz.Pa.s) 3.205E-I0
After - Permeable length 230 ED (J/mol) 17799
pretreatment Inside diameter 7 R Z for Arrhenius fit 0.9832
AfterPd 29.3085

Hydrogen permeance in micromoleJPa.mA 2.s (2b)

Selec:itivity data for 2b

13.5 1000
-'

13.1 "-- .,
, . -- -'
900

~ ,--- ,, -'
BOO
,
~

/~ 700 jj

, " - -- --
,-
..
- -
,, -
.-
-+- Nitrogen penneance ~ .eo
.!!
.ll

11.9
-- - __ . Hydrogen/Nitrogen 500
selectivity
11.5 400
330 370 410
T('C)

366
 Name: 3a (2.40 fll11)

Mass (g) (mm) Hz permeance parameters

Clean 29.2900 Plated length 250 Po (moLmlm2 .Pa.s) 1.752E-IO
After - Permeable length 230 ED (J/mol) 13223
pretreatment Inside diameter 7 R2 for Arrhenius fit 0.9643
AfterPd 29.4486

Hydrogen permeance in micromoleJPa.m A 2.s (3a)

Selecitivlty data for 3a

6.5 1500

,,
.
6.3
--- ,,
, 1400

penn~ ,, 1300
,,
I- ~
><
-+--Nitroge"
1200

l-
-.-, Hydrogen/Nitrogen
/' 1100 IJl
selectivity
., , ,
~
,

/
., " 1000
5.7

, ,
, ,-- , ,
, , ,
, ,
" 900

5.' 800
330 370 410 450
T('C)

367
 Name: 3b (4.43 11m)

Mass (g) (mm) H2 permeance parameters

Clean 29.2900 Plated length 250 Po (moLmlm".Pa.s) 1.691£-10
After - Permeable length 230 ED (J/mol) 12673
pretreatment Inside diameter 7 R" for Arrhenius fit 0.9424
AfterPd 29.5830

Hydrogen permeance in micromoleJPa.m A 2.s (3b)

6,5
5
11 4,5
i A
. ,
~ ,,5

SelecitiVity data for 3b

1.5 -----.. --.-----------.----------..------------------------.------------. ----... ----_. 4500

~ - . - Nitrogen permeanca ~

1.4 ~ -.- -1-/ydrogen/Nitrogen

f-----.--,.,./",--/--+' 4000
r'

selectMty
3. / /'
Ii ~ 1.3 ~ w" 3500 t
ll,.2 t--------'.=__=====--=-=o:==----t J 3000
z .s. ._-
~~~
_______
----
1.1 -I---:~"-'----------------t 2500

1.0 --- .-
+-_____----_-------+2000
330 370 410 450
T("C)

368
 Name: 5 (1.00 pm)

Mass (g) (mm) H2 permeance parameters

Clean 28.9540 Plated length 240 Po (moJ.m!m-.Pa.s) 7.504E-ll
After - Permeable length 230 ED (J/mo\) 9778
pretreatment Inside diameter 7 R- for Arrhenius fit 0.9676
AfterPd 29.0173

Hydrogen penneance in micromoleIPamA 2.s (5)

16
15
114
11:1
;a 1~
lIS 11
.,p
<1P
..., of>

',;;-'!P -,»
...""#

Selecltlvity data for 5

73.0 ----.-.----.. ----------.-- -..-.. --------.-._--.------_._------- ---.--.---------------.---... 240

65.0 +--=""""'-,'------1 1---1 '60

/ - .-. HydrogenlNltrogen
selectivity
63.0 1-_____-....'===~===--_l140
330 370 410 450
TC'C)

369
 Name: 6 (1.47 11m)

Mass (g) (mm) H2 permeance parameters

Clean 28.9106 Plated length 220 Po (mo1.m!m'.Pa.s) 8.178E-ll
After - Permeable!ength 220 ED (J/mo!) 9654
pretreatment Inside diameter 7 R" for Arrhenius fit 0.9969
AfterPd 28.9961

Hydrogen permeance in micromoleJPa.mA 2.s (6)

.

Selecltivity data for 6

40.0 .-.-.------.-----------.------.-.--------------------"---------------.-.--------------._-----------------.. 300

- • . Hydrogen/Nitrogen
selectMty
36.0 L-----_-===:::;:==='----1. 200
330 370 410 450
TI'C)

370
 Name: 8 (1.10 11m)

Mass (g) (mm) H2 permeance parameters

Clean 28.1700 Plated length 250. Po (mol.m1m'.Pa.s) 7.779E-ll
After 28.1730 Permeable length 230 ED (J/mol) 9435
pretreatment Inside diameter 7 R· for Arrhenius fit 0.9642
After Pd 28.2445

Hydrogen permeance in micromolelPa.mA 2.s (8)

SeleclUvlty data for 8

i5.!S 24.0 700
1A.5
23.'
050
~ 1,.5
~ 23.0

j 1~·6
~~ -----------_ 800 ~

~
11.5
10.6
=.
11
225

no
..

. - .---
~~

--- 550¥
~

j
4> z! 21.5
-- --
,4>. .
__ Nitrogen permellnce 500

...
4> 21.0 ,-
...
,;>

...~ ...4>
20.5

20.0
"0 370
..o-
-.- HydrogenJNl1fogen
r-
'10 ""
T('C)

% H2 permeated with N2 sweepgas and space % H2 permeated with N2 sweepgas and space
lime = 2.885 lime = 1.335
99.8 100.0
;j ---
- ..
99.6 .. ,_. ';:-'~~'''~--'';:- :--
-~
... .... ~:.>-.-
99.0
A>""-'
"",'-"
i
E
i
99.2
/~ .........
':/:' .- !
; IE 98.0

97.0
~""
J.'
I
.I
".0
../.~/' ! t I
£ 98.8 ; :£ 'B.O I
... ."F
. ; ... __ T= 45Q·C _ _ T=410°C
--+-T = 450°C _T",410·C
, I

1
98.S

96.4 / - ... T-370·C .... ·r:330·C

95.0 -.-T=370·C _·T 330'0

".2
0 , 10 15 20
94.0
0 5 10 15 20
N2 sweep rata as molar % of H2o feed N2 sweep rate as molar % of HI! fead

371
 Name: Nl (1.12 Jim)

Mass (g) (mm) Hz permeance parameters

Clean 29.3000 Plated length 230 Po (mo1.m!m".Pa.s) 1.238E-1O
After 29.3140 Permeable length 205 ED (J/mol) 12483
pretreatment Inside diameter 7 R' for Arrhenius fit 0.9872
AfterPd 29.3750

Hydrogen perrneance in micromoleIPa.mA 2.s (N1)

Selecitivity data for N1

3.11 -----------------------------------------------------_.. --.- --••• _------------
__ Nlrogen permSWII:e ""
~-------4~

'"' __'""----------------t""
u + - - - - - - - _ - -_ _ _ _ _-+""
330 370 410
Trel

% H, permeated with N, sweepgas and space % H, permeated with N, sweepgas and space
time =2.375 time 1.195 =
".. --.--------------------...--------- .--------------------------~=~=-~-;t 102.0

99.' +---------=";;!j=.....-;;;;;;;;~~~ 100.0

1I -~ ".0
99.' -I-----;--:y~<'"'.9""---------
I
~
96.0
: I

.'/1 ==F~~~
94.0
:'1
l ::1=4'.'/:='
:£
iJt. 98.8 -T-45Q"C _T_410·C
i!.
:£
...
92.0
90.0
• I

+-_-t--____--jI--T=450°C __ T-410·C [

...... t=="~"/=:=:=~1.~.=.T=-~37=O=.C~-=.T=.330::=.=C~
88.0
I· ". T=370·C - -T"m"C L
88.0
84.0
•
o 5 10 15 ~ ~ ~ ~ ~ 0 5 10 15 20
NI sweep rata as molar % of HI feed N2 sweep rate as molar % of HJ; feed

372
 Name: N2 (1.08 I'm)

Mass (g) (mm) Hz permeance parameters

Clean 28.5746 Plated length 205 Po (mol.m!m'.Pa.s) 5267E-ll
After 28.5905 Permeable length 205 ED (J/mol) 6984
pretreatment Inside diameter 7 R' for Arrhenius fit 0.9857
AfterPd 28.6412

Hydrogen permeance in micromoleJPa.m A 2.s (N2)

Selecltlvlty data for N2

22 700

370 410
T("C)

% H, permeated with N, 5weepga5 and space % H, permeated with No sweepgas and space
time =2.375 time -1.195

. . ~~~=======
100,0 -.------------- 100.0

J
99.'
99.4
j __ T = 450'C -+-T = 410'C l
,11-------
~•
".'
99.' f-- _T::z450"C -+-T-410'C
~

I 99.2 r~='.~.~T~.'~70~.C~....~.T~.:330~.~C~'~~~-~-~~~~~.-~~ ..... ·r =33O·C

99.4 ..... T_370'C
-~---
E 99,2
CL 99.0 !. ".0 ,7
-£ .... +-----;:-,-,;,~y;~,:.-.---------
'#. 98.9
:£ 98,8
,7
. - : / ...-
"-
98.4i----0;£..,.c.----------
.......,
98.6

98.2±===::!=:'======:::====,
".0
o 5 10 15
Hz swa.p rate as molar % of Hz feed
20 o
£
5 10 15
N z sweep ra.. as molar '% of Hz feed
20

373
 Name: N3 (1.47 Jim)

Mass (g) (mm) H2 permeance parameters

Clean 28.6470 Plated length 250 Po (mo1.m!m'.Pa.s) 1.118E-IO
After 28.6575 Permeable length 205 ED (J/mol) 11065
pretreatment Inside diameter 7 R' for Arrhenius fit 0.9989
AfterPd 28.7497

Hydrogen permeance in micromoleJPa.m A 2.s (N3)

Selecitlvity data for N3

6 ----_________________________________________________ ------------------------------------------------------ 220D

-+- Nitrogen permeance

2000

"""
5~ ______________ ~ ______________ ~1~

330 370 410

TIOC)

374
 Name: N4 (1.18 J'm)

Mass (g) (mm) Hz permeance parameters

I
Clean 28.9363 Plated length 230 Po (mol.mlmz.Pa.s) 8.l31£-11
After 28.9540 Permeable length 205 ED (J/mol) 9541
pretreatment Inside diameter 7 RZ for Arrhenius fit 0.9745
AfterPd 29.0170

Hydrogen permeance in micromole/Pa.m A 2.s (N4)

SeleclUvl1y data for N4

16

l'
13
22

21.5
,
'00

'"

Is 10
1~
\1 .
8
Ii .
11E
~
t;' 21

2{l5
\

.-
" -.'
'.
--"
"
--- ---,--_..---

---.- ./"/-'
,.,
"'~
i
~

...01'> £ So .........
20 ---. Nltrogeo permeance
""
4> - ...
-
"-'"
01>

...... .,;?
19.5

"
HI-
- .-. HydrtJgen'Nllrogen

.10 .....,
""
37<)

TrC)

% H,. permeated with N2 sweepgas and space % H2 permeated with N2 sweepgas and space
Ume = 2.375 time =1.19s
~ -:>I"';==~~~=:::::::===:'
100.0
100.0 __
98.0
~.

I I
96.0
99.0
¥
no'·
I: I
94.0

8.
98,5
&. fI:' .'
92.0 [--r:..so'c __ T 410'C
.,.
-£ •i
-£
"-
98.0

-j:(
// J-r 450'C - T 410 'C~ ~ 90.0 ·""·T-370·C "·T-330'C
97.5 t--f'-.-----il-..t..- T:370 "C -- -T=33O·C 88.0

97.0 88.0
0 5 10 15 W ~ ~ ~ ~ 0 5 10 15 20
Nl sweep rate as mohIr % of H, feed N, sweep rata as molar '% of H2 feed

375
 Name: N7 (1023 11m)

Mass (g) (mm) Hz permeance parameters

Clean 29.0120 Plated length 205 Po (mo\.m!m-.Pa.s) 9.211E-ll
After 29.0275 Permeable length 205 ED (J/mol) 13164
pretreatment Inside diameter 7 R· for Arrhenius fit 0.9926
AfterPd 29.0865

Hydrogen penneance in micromolelPa.m A 2.s (N7)

Selecilivily data for N7

o.
__ Nitrogen permuncil
" ..
I------i ,,,.

'2~~~------------~,oo

, .• + - - - - - - - _ - - - - - - - - 1 0 0 0
330 )70 .10
TI°C)

% H2 permeated with N. sweepgas and space % H2 permeated with N2 sweepgas and space
Ume= 2.37s Ume = 1.195
99,0 100.0

98.5 96.0

i ... ; ,
I
98.0 96.0

•~ 97.5
~~~ ..O.·:>-·-·- 94.0 .. .'
I 97.0
""'. .....- it 92.0
%:
'it 96.5
......
.' .-;
- ,..- l_r=450'C _1=410':f %:
~
90.0 , __ T-450'C

1.•.. Te370·c
__ T .. 410·C

_·T_330"C
96.0 +---~"'"------1I_·'IIo· T=370'C _·T .. m"e 88.0

95.5 86.0
0 5 10 15 20 0 5 10 15 20
Nt sw.ep rate as molar % of Hz feed Nt SWIHlP rate as molar % of HI: feed

376
 Name: N8 (lo19J1m)

Mass (g) (mm) Hz permeance parameters

Clean 29.1480 Plated length 205 Po (mol.m!m'.Pa.s) 5.815E-ll
After 29.1633 Permeable length 205 ED (llmol) 8993
pretreatment Inside diameter 7 R' for Arrhenius fit 0.9958
After Pd 29.2200

Hydrogen penneance in micromolelPa.mA 2.s (N8)

1~ Selecltlvlty data for N8

11 , 6000

I 10 2.'
2.' , ",
5500

I1~
0

I 8
tI ii' 2.4

22
-,
/
,
,

'/<"
5000~
• ~

..
2 ..00
i'-, /
~ 1.8 ,
.-------- 4000 JJ
..tP
..".
..<§>
z .s i.e
/
--I'a'ogen perm.... c:e

.., of1/> '.' - ... -Hyttogwv'Nllrogal "00

<:P~-6> .,;....
'.2 91111:1....1'1

"
"" 370
T(°C)
410 ... >DOD

% H2 permeated with N2 sweepgas and space % H, permeated with N, sweepgas and space
time = 2.375 time 1.195 =
....
99.7 102.0
100.0
_ T = 450 "C __ T .. 410·C /."/
:g 99.'
I
98.0
......;~/
II 99.5 - -,.,·T-370·C _"T=33Q'C
., 98.0

i 99.5
........."
"..
~
~
~
.;Y

./
./ ~
!.
94.0
•
'%,.. 99.4
;jt. 99.4

99.3
...
~
/' .
;£ 92.0
90.0

88.0
_ T _450·C __ T:410·C

• .,. 'T .. 370·C _'T;; 330·C f-----;,, . L - - - - - -

99.3 86.0
2 .. 6 8 10 0 1 2 3 ..
NJ sweep rate as molar %of HI feed Nz sweap rata as molar %of Ht feed

377
 APPENDIXD2

Hydrogen and nitrogen permeance

data for Pd-Ag films

378
 Alloyin2 procedures

Name Elements Alloying Heating Gas Temp (0C) Time (b)

system
8 Pd No
8b Pd+Ag No
8e Pd+Ag Yes Tube oven AI 550,600 15,25

NI Pd No
Nib Pd+Ag No
Nix NewPd+Ag Yes Reactor H2 590 10

N3 Pd No
N3b Pd+Ag No
N3e Pd+Ag Yes Tube oven AI 545 50
N3d Pd+Ag Yes Tube oven AI 545 100 (total)
N3e Pd+Ag Yes Tube oven AI 545 150 (total)

N4 Pd No
N4b Pd+Ag No
N4c Pd+Ag Yes Tube oven AI 530 30
N4x NewPd+Ag Yes Reactor H2 540 30

N6 Pd+Ag Yes Tube oven AI 550,600 15,25

N2x NewPd+Ag Yes Reactor AI,N2 500, 550 50,25

N8x NewPd+Ag Yes Reactor AI 550 10

When two temperatures and two times are given in a row, the ftrst heating time is for the ftrst temperature
and the second heating time is for the second temperature.

379
Membrane Thickness T Avg.H, Avg. N, Selectivity Arrhenius parameters for R" value for n I
(micron) (0C) permeance permeance hydrogen permeance in penneance
(fllIIoVm'.Pa.s) (nmoVm'.Pa.s) equation

8b 1.54 410 5.41 11.50 472 Po - 6.550E-11 1

370 4.69 11.90 394 Eo - 11739 0.9998
330 4.11 12.20 336 R = 0.9966 0.9996

N1b 1.40 410 5.20 57.10 91 Po - 9.752E-11 0.9999

370 4.49 45.30 99 Eo - 14707 1
330 3.69 48.00 77 R - 0.9978 1

N3b 2.14 410 3.21 3.20 999 Po - 8.064E-11 0.9994

370 2.82 3.70 758 ED 13938 0.9994
330 2.32 3.70 632 R - 0.9936 0.9991

N3c 2.14 410 6.59 17.60 374 Po - 8.762E-11 0.9999

370 5.99 17.90 335 ED - 10335 0.9999
330 5.18 20.30 255 R - 0.9931 0.9998

N3d 2.14 410 5.86 40.80 143 Po 1.348E-10 0.9999

370 5.25 41.90 125 ED - 13412 0.9994
330 4.29 43.10 99 R<~ 0.9825 0.9997

N3e 2.14 410 6.03 54.10 111 Po - 1.022E-07 1

370 5.41 56.00 97 Eo - 11714 0.9994
330 4.59 59.30 77 R' 0.9933 0.9999

N4b 1.43 410 2.87 12.00 239 Po - 1.236E-10 0.9994

370 2.37 10.60 226 ED 19295 1
330 1.83 10.20 180 R 0.9975 0.9996

380
¥+,~-

Membrane Thickness T Avg. H, Avg. N, Selectivity Arrhenius parameters for R< value for n I
(micron) (DC) permeance penneance hydrogen perrneance m penneance
(!illIoVm' .Pa.s) (nmoVm'.Pa.s) equation

N4c 1.43 410 3.02 14.60 208 Po - 2.403E-10 0.9997

370 2.48 14.80 167 Eo - 20393 0.9995
330 1.95 15.00 130 R - 0.9985 0.9994
290 1.49 15.30 95
250 1.02 15.50 65
210 0.69 15.70 44

N4x 2.13 410 4.62 24.50 188 Po - 7.195E-11 0.9998

370 4.10 25.20 163 Eo - 11286 0.9997
330 3.53 25.60 138 R - 0.9983 1
290 3.08 26.20 118
250 2.50 27.00 93

N8x 2.16 410 6.03 18.18 332 Po - 8.469E-11 0.9999

370 5.26· 18.71 289 Eo - 10654 0.9997
330 4.75 19.31 261 R<- 0.9976 0.9996
290 4.03 20.43 221
250 3.37 22.71 185

381
 Name: 8b (1.54 J,1m)

Mass (g) (mm) Hz permeance parameters

Clean 28.1700 Plated length 250 Po (mo1.mJm2 .Pa.s) 6.550E-ll
After 28.1730 Permeable length 230 ED (J/mol) 11739
pretreatment Inside diameter 7 R2 for Arrhenius fit 0.9966
AfterPd 28.2445
After Ag 28.2695 Oxidation Yes
Ag% 25 Temperature 320°C
Time l.5h

Hydrogen perrneance in micromoleJPa.m"2.s (Bb)

Selecitivlty data for 8b

1l.0 500
12.8 \ - - - __ NitrClgenpermeance \ -_ _ _ _ -<

11.3+-_ _ _ _ _ _ _ _ _ _ _ _---1

11.0+-------_--------1300
330 370 410

% H2 permeated with N2 sweepgas and space % H2 permeated with N2 sweepgas and space
Hme =2.865 Hme = 1.33s
100.0 100.0
.... ~~------- ...
~j 99.'
,,":--,
99.' .-. ,
.,.-..... --
....~:"' r-: ;:-.:;.::-- 96.0
.. I, 94.' .'
I 99.' .-;,..-:..
.... "; ...
,
i I
~
92.'
/

i 98.5
J' .17
i
!. 90.'
.
;£
980
• _T::410°C - •. T"370OC
i
H .. '"'.
;£ 88• •
;
,. ~T::410GC .•. T-370°C

~
... ·T-330·C
.... ·T_330°C 84.'
97.5

97.0
i
i
82.0
80.'
" ,
0 5 10 15 20 0 • 10 15 20
Nt sweep rate as molar % of H2 feed N2 sweep rate as molar % of H2 feed

382
 Name: NIb (1.40 11m)

Mass (g) (mm) Hz permeance parameters

Clean 29.3000 Plated length 250 Po (mo1.m1m'.Pa.s) 9.752E-ll
After 29.3140 Permeable length 205 ED (J/mol) 14707
pretreatment Inside diameter 7 R' for Arrhenius fit 0.9978
AfterPd 29.3750
After Ag 29.3965 Oxidation Yes
Ag% 22 Temperature 320°C
Time 1.5h

Hydrogen permeance in micromoleJPa.mA 2.s (N1b)

Selecltlvity data for N1b

B i i 520+---~"--------7"'/---foo
~t. / /.... f
i="'E:w 48.0
--.- ~
/
. ,/ 85 t;
•
; ! ~.oh-'-/-''------------------''''~-··-,.,,----r-j80 i
,/ - - NI\rogm ptnneanco

.00 + - - - - - - - - 1 _._ Hyd"ogt:nlNilrogen r-"

selectivity
".. 1-------_-'=======---t,.
3311 370 410
T("CI

'k H2 permeated with N2 sweepgas and space 'k H, permeated with N, sweepgas and space
time 2.375 = time -1.195
100.0
--':-~''::;i?-:;;~;;-~;;;;;;;';;~'~=:=;.o;-=--
100.0 -.--.----.-.. ------.-.--.-.--~-;:-;;-::-:::;;.:--.::-.::-::~~:~..:~~.-,:=-.;.'1

95.0
95.0 ,.
,r 90.0

I• 90.0
~ 85.0
80.0
~
i!.
65.0
l 75,0
/
r
..
:£ BO.O

75.0
•
1_.
_T-410'C

T ",330 C
0
-,.. T .. 370·C
..
-£
70.0
05.0
60.0
55.0
J __ -T:::330'C
+-------,.'"'----i
•
__ T .. 410·C - .... T=370'C
[

70,0 SO.O
0 5 10 15 ~ ~ 30 ~ ~ 5 10 H
Nt sweep rat. as molar % or HI teed Nz sweep"" as molar" of ~ feed

383
 Name: N3b (2.14 11m)

Mass (g) (mm) H2 permeance parameters

Clean 28.6470 Plated length 205 Po (mol.mJm·.Pa.s) 8.064E-ll
After 28.6575 Permeable length 205 ED (J/mol) 13938
pretreatment Inside diameter 7 R· for Arrhenius fit 0.9936
AfterPd 28.7355
After Ag 28.7640 Oxidation Yes
Ag% 24 Temperature 320°C
Time 1.5h

Hydrogen permeance in micromolelPa.m A 2.s (N3b)

,.6 Selecilivity data for N3b

4.0 ",,.
..., ,-Z
i ~.s
~ ~~
" .po
oF"
.... # '2r----~~~-------__1'oo
4.. ...<1> --'
~ ,,~ ,.l-------~------___I'"
lI30 :)1U 410
TI-t:)

% Hz permeated with Nz sweepgas and space % Hz permeated with Nz sweepgas and space
=
----
time 2.375 time = 1.195
100.0 100.0
95.0
"
..
.,' - .;'
96.0 90.0
~ f1".' • I.: f
~
•~
!.
92.0

88.0
:

i
1

I
!.
85.0
80.0
75.0
70.0
I:
/:
I: .'
.'
,
-£
~
I-T=410°C ..... T=370'C
-£
~
65.0
:
_ T 410'C .... T=370'C t-
84.0 +--------1,_ -T=3JO'C 60.0
55.0
+--ri------IJ -- -T = 330'C t
80.0 50.0
0 10 15 ~ ~ ~ ~ ~ 0 5 10 15 E ~ E ~ ~
Nt SWHP rate as malar % of Hz feed Nz swe.p nita as molar % of HI feed

384
 Name: N3c (2.14 11m)

Mass (g) (mm) H2 permeance parameters

Clean 28.6470 Plated length 205 Po (mo1.m1m".Pa.s) 8. 762E-11
After 28.6575 Permeable length 205 ED (J/mol) 10335
pretreatment Inside diameter 7 R" for Arrhenius fit 0.9931
AfterPd 28.7355
After Ag 28.7640 Oxidation Yes
Ag% 24 Temperature 320°C
Time l.Sh

Hydrogen permeance in micromoleIPa.mA 2.s (N3c)

6.&
G,A
Selecltivity data for N3c
..
i 5.:
~ 5~
".+----------------:=~_=J
..' '"

..~""
17.0+-----------------1
1 1 1 . 0 + - - - - - - - _ - - - - - - -.. 200
330 370 410
T("CI

% H, permeated with N, sweepgas and space % H, permeated with N, sweepga5 and space
time = 2.375 time = 1.195
100.0,.----------------- 100.0,.-----------------

911.0+----------------- 95.0

i• 96.0 I 90.0 _ :f
~
!. ".0
i 85.0
:'.'
:£ -£
"-
i· , J_T=410°C ••" T=370"C
"- 80.0 t - - ; i - - - - - - - j
92.0
• !__ 'T_330°C
90.0 75.0
0 5 10 15 ~ ~ E ~ ~ 0 5 10 15 ~ ~ E ~ ~
Nt sweep rata as molar % of Hi feed Nt sweep rata as molar % of Hi teed

385
 Name: N3d (2.14 Jim)

Mass (g) (rom) H2 permeance parameters

Clean 28.6470 Plated length 205 Po (mo1.m1m'.Pa.s) 1.348E-10
After 28.6575 Permeable length 205 ED (J/mol) 13412
pretreatment Inside diameter 7 R" for Arrhenius fit 0.9825
AfterPd 28.7355
After Ag 28.7640 IOxidation INo
Ag% 24

Hydrogen permeance in micromoie/Pa.mA2.s (N3d)

Selecitivity data for N3d

....0 --. ____ •____ ._----------._. ________ •__________ •__ .'_. ____ -----------.' _______ • __ •_____ ._ 1611

, ..

% H, permeated wtth N, sweepgas and space % H, permeated wtth N, sweepgas and space
time s2.37s time -1.195

--
100.0 100.0 _.• -----------. ---.--_._---------.'------------.. _----.'---.-.--------- ..--... _.---.------.-.
95.0 +-___.... -::-:'-="=~:_'-':"-"'-;':'-"'.;,;.~
.•",
.•;,:.-",.-",.-':..-:.=.
-::-::-.-:.-.~: ~:.:.-.-.
-'
.. -:':.·-"",-1
96.0

IE " ..
90.0

I~
92.0

• iI.
80.'
Go
-£
88.0
.....
.: ,,'
, -£ 75.0 __ T-410·C - .... ··T=370·C

;II
84.0
I
t--~-.T'------I! __ -r ..
-T;410°C

33O'C
.... T=370'C
;II 70.0 1 __ .T 330"C
65.0

80.0 60.0
0 5 10 1S ~ ~ ~ ~ ~ 0 5 10 15 ~ ~ ~ ~ ~
Nz sw.ep rate as molar % of Hz feed Nt sw..p rate as molar % or Hz feed

386
 Name: N3e (2.14 Jim)

Mass (g) (mm) H2 permeance parameters

Clean 28.6470 Plated length 205 Po (mol.mlm·.Pa.s) 1.022E-07
After 28.6575 Permeable length 205 ED (J/mol) 11714
pretreatment Inside diameter 7 R· for Arrhenius fit 0.9933
AfterPd 28.7355
After Ag 28.7640 IOxidation INo
Ag% 24

Hydrogen permeance in micromoleIPa.m A 2.s (N3e)

6:t Selecitivity data for N3e

...,
11
5.8
5·'
..,.,, '--,
""

\
, -~-' ''"
5 .......
I( 6rA
-"'''',
'"' ~
••6 ~"e
&. ~
:f!
6B.D
.
·~,:x::.a'
~
.I
... ..
"'" ,--- ---....--- .
__ Ntogen p.-me.'lCCl

,
. "'"
- .... HyctogaVtlltrogen
" - ...? seledl\o1ty
70
'" 4>
oS> ."."...
'"
T ("CI '"

% H. permeated with N. 5weepgas and space '10 H. permeated with N, sweepgas and space
time s 2.375 time =1.195
100.0 , - - - - - - - - - - - - - - - - 100,0

96.0 --- ... ".0

90.0

I
II.
92.0
III.
'M
80.0
...
.
-£
".0

64.0
..
~~"~.' J--T=410-C . '110- T=370·C
.
-£ 75.0

70.0

'5.0
I__ r
_T=410·C
33O"C
.... T=370·C

80.0 80.0
0 5 10 15 ~ ~ ~ u ~ 0 5 10 15 ~ ~ ~ H ~
Nz sweep rata as molar % of HI teed Nz sweep rata as molar 'I. of Hz faed

387
 Name: N4b (1.43 11m)

Mass (g) (mm) H2 permeance parameters

Clean 28.9363 Plated length 205 Po (mol.mlm2 .Pa.s) 1.236E-10
After 28.9540 Permeable length 205 ED (J/mol) 19295
pretreatment Inside diameter 7 R2 for Arrhenius fit 0.9975
AfterPd 29.0004
After Ag 29.0197 Oxidation Yes
Ag% 23 Temperature 320°C
Time 1.5h

Hydrogen penneance in micromoleJPa.m A 2.s (N4b)

Selecilivity data for N4b

»,
11.84------------==~~
/ ""
.-- --~-------;//
". 4---------,-""----~-'----__1 23'
~
II
i. "'4-----."7. .-"-/-'-
.. ------7','--/-----1 2tD
_~- //
~

i
10' 1

'"

% H, permeated with N, sweepgas and space % H, permeated with N, sweepgas and space
time 2.375 = time = 1.198
--··-K:;':~:·:::'::;;~:'=;=:::=-;':;':-==:';':---------~ 100.0 .•. ----.-.----.---------.----.---~:~~~?~!~:~:::-:::-::.-=':=:~~-""II

95.0 l---..,;'!C,."'.:...'----------- 90.0

11
'i
(:7
'" I ~~"-·~7···,:./-·J-·---------
80.0.j-_ _.....
~ :::l-_LJ":..!/~:.../____________
I

"'~.
i 70.0 l--~."':-'," " - - - - - - - - - - - - -
... " 1"

~
~ 80.0 l-~:['-----f~~~~~~~
... l--T."10~C T_370'C -'II._ ,;:
'if!. "" +~.'
-.0 I--...::,,:·.i/-----r::=:;=;::;:--:-::-;;~;;:]
l--r=410 c - 'IL- T=370'C
0

75.0 t--~"-----L....
~·T~-..:330~.C'__ _..._J 50,0 j--=------I..::....~.T.:.~~330~·::.C_ _ _..J
7O.0j......-_-~-~_-~-~-~-~ ...ol--_-~-_ _-~-_-_-~
o 10 15 ~ ~ m ~ ~ o 10 15 ~ ~ m ~ ~
N2 SWHP rata as molar % of HI feed Nz SWHP rata as molar % of H,: feed

388
 Name: N4c (1.43 11m)

Mass (g) (mm) Hz permeance parameters

Clean 28.9363 Plated length 205 Po (mo1.m!m'.Pa.s) 2.403E-1O
After 28.9540 Permeable length 205 ED (J/mol) 20393
pretreatment Inside diameter 7 R' for Arrhenius fit 0.9985
After Pd 29.0004
After Ag 29.0197 IOxidation INo
Ag% 23

Hydrogen penneance in micromolelPa.m A 2.s (N4c)

Selecltlvlty data for N4c

,<0.-----------------,""
15.6 ---....... 200

- .-. HydrogenlNltrogen
selectivity
14.0 L ___--.-'==~;::::===;='...----to
n "" _ ~ m m

389
 Name: N4x (2.13 Jim)

Mass (g) (mm) H2 permeance parameters

Clean 28.9365 Plated length 250 Po (mol.mlm'.Pa.s) 7. 195E-11
After 28.9428 Permeable length 205 ED (J/mol) 11286
pretreatment Inside diameter 7 R" for Arrhenius fit 0.9983
AfterPd 29.0476
AfterAg 29.0760 Oxidation Yes
Ag% 20 Temperature 350°C
Time 3h

Hydrogen penneance in micromoleJPa.mA 2.s (N4x)

Selecitivity data for N4x

".0 ,.,
__ f'l1rogIn perme.,ce
".
'80

Ii 2"
27.0
...... ",.

~.
- ... ' l-/yItogenlNlrogB'l
9IId\4ty

.. ...'
J'"

.., f
~ '"E
.'
2t.O ". Ii
a.'"!.
%' 2U
215.0
.", ..><:....
------ ""
li
..- ---~
" .• '00

".,., ,., ...

T(OC)
>7, <10
80

% H, permeated with N, sweepgas and space % H, permeated with N, sweepgas and space
arne = 2.375 time = 1.19 5

100.0
95.0
[==::::;;;;~.;_;._~~~~.~.~.~.~
~>.-'
. ".0

~ 90.0 . ,.,
fI":-" .rI
~ 90.0

~
II.
".0 , ~
3. ".0
.'
.'
,
..
,£ .<10

75.0
__ T:370'C _T_330"C

r · ..··T:290·C _'T=25Q"C
,£
;I-
80.0

75.0
+-------,"'-------1 _ T " 370·C
• .' J '6' T=290"C
__ T '" 33O'C
_'T=250'C

70.0 70.0
0 10 15 E ~ ~ ~ ~ 0 5 10 15 ~ ~ ~ ~ ~
N2 sweep rata as molar % 01 HI feed H2 SWHP nrte as molar % of HI feed

390
 Name: N8x (2.16 p.m)

Mass (g) (mm) H2 permeance parameters

Clean 29.1785 Plated length 220 Po (moLmlm".Pa.s) 8.469E-ll
After 29.1845 Permeable length 205 ED (J/mol) 10654
pretreatment Inside diameter 7 R" for Arrhenius fit 0.9976
AfterPd 29.2745
After Ag 29.3030 Oxidation Yes
Ag% 23 Temperature 310 ·C
Time Ih

Hydrogen permeance In micromoI/Pa.m A 2.s (N8x)

Selecitivity data for N8x
23.0 ----.. ----------------"-.-----..- •. -.------.-----. __________ .. -------- 340
,\,,_ __ Nitrogen permeance ./ / 320
/

22.0 +-'-,----1 f---~.L--I

'-.-H~~~ ~~ 300

§ ~D.
til 21.0+
I __ ··...'.\¢==H=-===~'"".,::::/-=--.-'-----t
,
280 ~
'S:
•

E :§
N'

'+, /.~ 260 ¥

Z. ~ 20.0 ". /' 240 ~
-£ .£
~ ~ ..' ,"
>"'- "',~",-_
220
19.0 +--'""7~-----"'-~-=-------1
/// --______ ___. 200

18.0 f---~---~---~---f 180

250 290 330 370 410
T (OC)

% H2 permeated with N2 sweepgas and % H2 permeated with N2 sweepgas and

space time = 2.37s space time = 1.19 s
100.0 100,0
-- .
98,0 ".0
//: ,
h-'"
/
/ -'
~ l•
I'

iE
90.0
".0
.-:.' i 85.0
f/;' .I
.
• 0

• I E
1 0

• ".0
~
,

I '(lO
: I
_ T = 370°C -...-T""330·C-
N
l: 92.0 _T-370·C ___ T _330°C I- -£ : I
o 0
__ ·T= 250·C -
~ 75.0 - .. ·T .. 290"C
~ I
-,.,- T:290·C ___ ·T"Z50·C i- ;'
90.0
• 70.0

65.0
~
88,0
0 5 10 15 20 25 30 35 40 0 5 10 15 20
Nt sweep rate as molar % of Hz feed
25 30 35 40
,
Nz sweep rate as molar % of ~ feed

391
 For some membranes a full characterisation was not done:

Name: 8e (1.54 flm)

Mass (g) (mm)
Clean 28.1700 Plated length 250
After pretreatment 28.1730 Permeable length 230

After Pd 28.2445 Inside diameter 7

After Ag 28.2695
Ag% 25 Oxidation No

Name: N6 (2.05 flm)

Mass (g) (mm)
Clean 28.0606 Plated length 250
After pretreatment 28.0702 Permeable length 205

AfterPd 28.1680 Inside diameter 7

After Ag 28.1966
Ag% 21 Oxidation No

Name: Nix (2.13 flm)

Mass (g) (mm)
Clean 29.3154 Plated length 250
After pretreatment 29.3175 Permeable length 205

After Pd 29.4225 Inside diameter 7

After Ag 29.4527
Ag% 22 Oxidation No

392
 Name: N2x (2.02 Jim)
Mass (g) (mm)
Clean 28.5978 Plated length 250
After pretreatment 28.6115 Permeable length 205

After Pd 28.7000 Inside diameter 7

After Ag 28.7335
Ag% 25 Oxidation No

393
 APPENDIXE

Literature data on Pd and Pd-alloy thin

films

394
 "'"
----------------------------------------------------------~.'
-'-:.-~ -

Preparation Support Coating Thickness Temperature Permeance Selectivity Reference

method description composition (11m ) (0C); IlP (kPa) Mollm2/s/Pa H2/gas

Non Palladium membranes:

Clean tube a-alumina 400 21 3 (N2) Kusakabe (1996)
150nm(NOK)
Clean tube air-alumina (a) 25 15.8 3.6 Casanave (1995)
3 nm (SCT)
Tube Silica silica 7.5 440;300 0.79 56 (CO 2) Weyten (1997)
Slip casting a -alumina (a) Si0 2 5.0 100 28 3.5 (N2) Vitulli (1995)
Tube 200nm (SCT)
Tube a -alumina (a) Zeolite 1.15 3.0 Casanave (1995)
200 nm (SCT) 0.6 nm pores i

CVD in pores Porous Vycor Si02 450, 100 0.0191 25 Tsapatsis (1991)
of membrane 4 nm (Coming) Ah 0 3 450, 100 0.0020 210
tube B2 0 3 450, 100 0.0068 20
Ti02 450, 100
-_. -
0.0041
'----. _.
63

Palladium and palladium alloy foils

Foil Foil Pd .
50 350 0.30 Jung (2000) .

Disc 500 0.60

Foil Foil Pd 24 350; 156 1.20 Hurlbert (1961)
350;444 0.87
Foil Foil Pd 20 540;340 0.23 Grashoff (1983)

Quenching and Foil Pd so Si2o 40 379; 100 0.016 - Itoh (1995b)

rolling
Foil Foil (TK) Pd77A~3 150 75; 100 0.029 Itoh (1997)
Foil Foil Pd77 Ag23 100 350; 1950 0.12 Jorgensen (1997)
Foil Foil Pd77Ag23 100 300;689 0.18 Ali (1994)
400;689 0.20
'--

395
 I

l
Preparation
method_ _
J
Support
description
Coating
composition
Thickness
(I'm)
Temperature
(0C); OP (kPa)
Permeance
Mol/m2/s/Pa
Selectivity
H2/gas
Reference

Foil Ta Foil Pd-Ta-Pd 1-13-1 300;48 l.32 Peachey (1996)

(GFM) 340;47 1.76
Foil Different foil Pd75 Ag25 25 on both 400;780 0.66 > 100 (N2) Edlund (1996)
substrates sides
Foil Foil Pd91 RIl6In3 200 300; 100 0.0000105 Hara (1999)
Foil Foil Pd-Ah03-V 700; 689 0.43 Edlund (1995b)
25-2S0-2} (!!m) . - - -

Electroless Pd plating on ceramics (discs, inside and onside oftubesl

Plating (electr) Modified u- Pd I 4S0; 101 9.75 23 (N2) Zhao (1998)
Disc alwnina
Plating (electr) u-alwnina (s) Pd 8.S See comments Ilias (1997)
disc ISOnm 12.0
(VCMC)

Plating (electr) Porous ceramic Pd 4.5 400 2.86 Kikuchi (199S)

Outside tube
Plating (electr) u-alwnina (a) Pd 4.S 400; 194 1.54 - Uemiya (1991 a)
Outside tube 200nm(TCC)
Plating (electr) u-alwnina (a) Pd 10.3 467; 102 2.68 970 (N2) Li A (1999)
Outside tube 160 nm (ECN)

396
I
Preparation
method
JSupport
description
Coating
composition
Thickness
(Jim)
Temperature
(0C); 1)P (kPa)
Permeance
Mol/mz/s/Pa
Selectivity
Hz/gas
Reference

Plating (electr) aly-alumina (a) Pd 2.0 300 0.63 Shu (1996b)

Inside tube + 5 run (SCT) 400 1.39
pores 500 2.34
Plating (electr) aly-alumina (a) Pd 2.1 450; 2.55 4.09 Shu (1997b)
Inside tube 3-4 run (SCT) 500; 2.55 5.27
Plating (electr) a-alumina (s) Pd 8 500;689 0.19 48 (N2) Paglieri (1999)
Inside tube 200 run (GTC) 16 450;689 0.22 200 (N2)
10 450;689 0.21 49 (N2)
Plating (electr) a-alwnina (a) Pd 11.4 550; 100 1.16 550 (N2) Collins (1993b)
Inside tube 200 run (USF) 17 550; 100 0.78 63 (N2)

?? I Ceramic tube 1 Pd 14 1 350; 1950 12.93 1 127 (N2) 1 Jorgensen (1997)

Electroless Pd plating on porous glass, porous SS and refractory metals

Plating (electr) Porous glass (s) Pd 20 400; 196 0.68 - Uemiya (199\b)
Outside tube 300nm (lSI)

Plating (electr) Porous SS Pd ?? 380;300 0.42 Li A (2000)

Outside tube 100nm (MM)
Plating (electr) Porous SS (s) Pd 6.0 400; 100 0.53 Jemaa (1996)
Disk 500 run (MM)
Plating (electr) Porous SS Pd ?? 480; 100 0.73 1420 (N2) Li A (1997)
Outside tube 100nm (MM)

Plating (electr) Niobium Pd 2-\50-2 ??; 100 1.44 Buxbaum (1996)

Tube \50 JlIIl thick
Plating (electr) Tantalum Pd 2-70-2 420; 100 1.93 Buxbaum (1996)
Tube 70 J.1m thick - ----

397
Preparation Support Coating Thickness Temperature Permeance Selectivity Reference
method description composition (JIm) (0C); liP (kPa) Mol/mz/s/Pa Hz/gas

CVD of Pd on alumina discs and tubes

CVD (In pores air-alumina (a) Pd 0.5-1.0 350 0.9 3.6 (N2) Xomeritakis (1996)
+ surface disk) 4nm
CVD (In pores air-alumina (a) Pd 1.0-1.5 300 0.21 200 (He) Xomeritakis (1997)
+ surface disk) 4.5nm

CVD (In pores a-alumina Pd 4 300; 100 4.0 10000 (N2) Van (1994)
+ outside tube) l50nm(NOK) 500; 100 4.0 6700 (N2)
CVD (In pores a-alumina Pd 4.4 400;200 0.38 3000 (N2) Aoki (1996)
+ outside tube) l50nm(NOK) 500;200 0.60 1900 (N2)
CVD (In pores a-alumina Pd 4.4 400 0.80 1330 (N2) Kusakabe (1996)
+ outside tube) 150 nm (NOK)
CVD (In pores a-alumina Pd 4.4 400 0.7 1160 (N2) MOTOoka (1995)
+ outside tube) 150nm(NOK)

Pd membranes prepared by electroplating and wet impregnation

Wet impregn. Ni/porous SS Pd 0.5 450 15.8 1600 Jun (1999)
Disc 50nm(MM)
Electroplating Nilporous SS Pd 0.8 550 17.8 4700 Nam (1999)
Disc 50 nm(MM)

398
Preparation Support Coating Thickness Temperature Permeance Selectivity Reference
method description composition (/lIn) (oq; OP (kPa) Mol/m2/s/Pa H2/gas

Palladium alloys prepared by electroless plating

Plating (electr) a-alumina (a) Pdg9 Ag 11 5.0 400; 194 1.97 - Uemiya (1991a)
Outside tube 200 nm (TCC) Pd77A~3 5.8 400; 194 2.24
Pd69 Ag31 6.4 400; 194 1.45
Plating (electr) Porous ceramic Pd77Ag23 5.8 400 4.09 Kikuchi (1995)
Outside tube
Plating air-alumina (a) Pd-Ag 7.5 440;300 1.60 15000 (CO2) Weyten (1997)
(chemical)
Plating air-alumina (a) AulPd-Ag 7.5 440;300 l.01 4000 (CD2) Weyten (1997)
(chemical)

Plating (electr) Porous Vycor Pdgg -Ag 12 1.2 500; 170 anneal 0.10 Cheng (1999)
Outside tube 5 nm (Coming) 400;170 0.074
Plating (electr) Porous glass (8) Pd93 -Ag7 21.6 400; 196 0.32 - Uemiya (1991b)
Outside tube 300 nm (lSI)
Plating (electr) Porous glass (s) Pd 94 -C U 6 18.9 400; 196 0.21 - Uemiya (1991b)
Outside tube
~ -
300 nm (lSI) ----

399
Preparation Support Coating Thickness Temperature Permeance Selectivity Reference
method descriptioD_ composition (11m ) (0C); OP (kPa) Mollm2/s/Pa H~gas

Palladium alloys prepared by sputtering

Sputtering air-alumina Pd7S Ag2S 0.35 250 0.28 5.7 (N2) Jayaraman (1995b)
(disc) 3nm

Magnetron air-alumina (a) Pd7SAgz s 0.4 300 0.15 80 (He) Xomeritakis (1997)
sputtering 4.5nm
(disc)
Dc sputtering air-alumina (a) PdnAgs 0.33 300 0.071 63 (He) McCool (1999)
(disc) 4nm
Spray pyrolysis air-alumina (a) Pd7~gz4 1.75 500 0.080 24 (N2) Li ZY (1993)
Outside tube 5 nm(NOK)
Sputtering MgOss-Ytrials Pd~lI6 10 700; 1000 1.23 Gryaznov (1993)
300 nm (s)

Sputtering Porous SS (s) Pd94R lI6 10 800,2000 0.68 Gryaznov (1993)

(disk) (MPV)
Sputtering Porous SS (s) Pd93.s In6R Uo.5 1.5 372; 100 0.45 - Gryaznov (1993)
(disk) (MPV)

Sputtering Porous glass Pdn Ag23 6.0 380; 125 0.056 Gobina (1994) i

(Outside tube) 4nm

Sputtering Poly(dimethy]si Pd76A g24 0.05 25 0.003] 100 (CO 2) Athayde (1994)
loxane) coated
Membrane (a)
Sputtering Polyarilyde (a) Pd94RlI6 0.4 200; 1000 0.16 ._-
Gryaznov (1993)

400
Comments
1. All permeances were changed from the original units reported to j.Ullolf(m2 .s.Pa)
2. Open cells in the data pages represent data not supplied in the specific reference.
3. Selectivities were only included if permeation data for the other gas was listed as a function of
pressure and/or temperature.
4. In one reference (!lias 1997), the graphs and values mentioned in the paper contradict each other
and was thus not included.

Abbreviations
ECN: Energy Corporation of The Netherlands
GFM: Goodfellow Metals, Berwyn, P A
GTC: Golden Technology Company, Colarado
lSI: Ise Chemical Industry Company
MM: Mott Metallurgical
MPV: Metallurgical plant, Yyksa, Russia
NOK: NOK Corporation, Japan
SCT: Societe des Ceramiques Techniques
SS: Stainless steel
TCC: Toshiba Ceramics Co.
TK: Tanaka Kikinzoku, Japan
USF: US Filter Corporation, Warrendale, Pennsylvania
YCMC: Velterop Ceramic Membrane Company of The Netherlands

(a): asymmetric
(s): symmetric

401
 APPENDIXFl

Plug flow reactor data and membrane

reactor data for ethanol dehydrogenation

402
Membrane name: N8x Catalyst: 3.00 g (14.4 wt% Cu on silica)
Reaction temperature: 250°C Catalyst particle size: 500-800 microns

Feed Sweep to Total ethanol Ethanol conversion Ethanol Ethanol to acetic Reaction selec-
rate feed molar conversion to acetaldehyde conversion to MEK acid ethyl ester tivity towards
mlfh ratio acetaldehyde
Average Average Average Average Average
5 0.00 35.7 27.8 0.7 2.8 78.0
27.1 26.2 0.5 0.4 96.7
34.5 32.4 30.3 28.1 0.7 0.6 1.4 1.5 87.7 87.4

1:2.5 42.1 34.5 0.8 2.8 81.9

31.9 29.0 0.7 0.8 90.9
43.4 39.1 36.4 33.3 0.9 0.8 2.4 2.0 84.0 85.6

1:1 41.9 34.1 0.9 2.4 81.4

38.8 32.6 0.8 2.1 84.0
47.1 42.6 36.0 34.2 1.2 1.0 3.2 2.5 76.5 80.6

2:1 43.6 37.3 0.7 2.0 85.4

45.4 37.1 0.8 2.3 81.6
46.0 45.0 38.5 37.6 0.9 0.8 2.0 2.1 83.7 83.6

4:1 49.1 41.3 0.7 2.4 84.2

50.0 42.3 0.7 2.2 84.7
39.4 46.2 33.6 39.1 0.8 0.8 1.4 2.0 85.4 84.7

8:1 55.3 46.0 0.8 2.5 83.1

48.6 41.7 0.9 1.9 85.7
. -
52.3 52.1 45.1 44.2 0.7 0.8 1.9 2.1 86.2 85.0

403
Membrane name: N8x Catalyst: 3.00 g (14.4 wt% Cu on silica)
Reaction temperature: 250°C Catalyst particle size: 500-800 microns

Feed Sweep to Total ethanol Ethanol conversion Ethanol Ethanol to acetic Reaction selec-
rate feed molar conversion to acetaldehyde conversion to MEK acid ethyl ester tivity towards
mIlh ratio acetaldehyde
Average Average Average Average Average
10 0 27.4 26.4 0.5 0.5 96.3
28.5 27.2 0.5 0.5 95.2
30.3 28.8 27.7 27.1 0.7 0.6 1.0 0.7 91.4 94.3

1:5 34.3 31.2 0.9 0.9 90.8

36.6 35.5 32.1 31.6 0.8 0.8 1.0 1.0 87.7 89.3

1:2 35.7 32.4 0.9 0.6 90.6

34.8 30.4 I 0.8 1.4 87.3
30.2 33.6 26.5 29.7 0.5 0.7 1.1 1.1 87.6 88.5

1:1 28.3 27.1 0.6 0.6 95.8

28.0 27.1 0.5 0.4 96.7
29.9 28.1 0.7 0.6 94.1
28.4 28.6 26.9 27.3 0.6 0.6 0.4 0.5 94.6 95.3

2:1 35.8 33.7 0.8 0.6 94.0

31.1 28.9 0.5 0.5 92.9
29.8 32.2 28.5 30.4 0.6 0.7 0.4 0.5 95.5 94.2

4:1 34.6 32.3 0.7 0.5 93.5

34.5 34.5 - 33.0
- -
32.7 -
0.7 0.7 0.4 0.5 95.7 94.6

404
Membrane name: N8x Catalyst: 3.00 g (14.4 wt% Cu on silica)
Reaction temperatnre: 275 DC Catalyst particle size: 500-800 microns

Feed Sweep to Total ethanol Ethanol conversion Ethanol Ethanol to acetic Reaction selec-
rate feed molar conversion to acetaldehyde conversion to MEK acid ethyl ester tivity towards
mlfh ratio acetaldehyde
Average Average Average Average Average
10 no sweep 47.0 38.2 0.9 2.7 81.3
45.2 46.1 38.2 38.2 0.9 0.9 2.0 2.4 84.5 82.9

1:5 43.8 39.4 0.9 1.3 90.0

42.4 43.1 39.2 39.3 0.8 0.8 0.8 1.1 92.5 91.3

1:2 48.2 43.8 0.9 1.1 91.0

49.1 48.6 45.1 44.5 0.8 0.9 1.0 1.0 91.9 91.5

1:1 48.2 43.9 0.9 1.1 91.0

58.9 53.6 53.0 48.4 1.1 1.0 1.3 1.2 90.0 90.5

2:1 62.3 53.7 0.6 1.9 86.2

61.8 62.1 55.5 54.6 1.1 0.8 1.8 1.8 89.8 88.0

4:1 61.8 58.1 1.8 0.6 94.1

58.1 59.9 54.7 56.4 1.8 1.8 0.5 0.6 94.2 94.2

405
 "
:;;;

Membrane name: N8x Catalyst: 3,00 g (14A wt% Cu on silica)

Reaction temperature: 275 DC Catalyst particle size: 500-800 microns

Feed Sweep to Total ethanol Ethanol conversion Ethanol Ethanol to acetic Reaction selec-
rate feed molar conversion to acetaldehyde conversion to MEK acid ethyl ester tivity towards
mllh ratio acetaldehyde
Average Average Average Average Average
20 0.00 41.8 33.3 0.6 2.5 79.8
45.9 43.8 37A 35A 0.8 0.7 2,1 2.3 81.5 80.6

1:5 47.7 40.9 0.8 1.9 85.8

48.6 48.1 43.6 42.2 0.8 0.8 1.6 1.7 89.6 87.7

1:2 43.8 38.9 0.8 1.3 88.8

50A 47.1 43.5 41.2 0.6 0.7 L7 1.5 86.2 87.5

1:1 45.9 42.8 0.9 0.8 933

55.3 50.6 48.7 45.8 0.9 0.9 1.2 1.0 88.1 90.7

2:1 46.0 44A 0.7 OA 96.5

48.1 47.1 44.3 44A 1.0 0.9 0.8 0.6 92.1 94.3

4:1 55.2 50.8 0.9 1.1 92.0

48.4 45.4 0.9 0.7 93.8 I

48.8 50.8 44.8

--
47.0 0.9 0.9 0.8 0.9 91.9 92.5 I

406
F.,!~,,",O-

Membrane name: N8x Catalyst: 3.00 g (14.4 wt% Cu on silica)

Reaction temperature: 275°C Catalyst particle size: 500-800 microns

Feed Sweep to Total ethanol Ethanol conversion Ethanol Ethanol to acetic Reaction selec-
rate feed molar conversion to acetaldehyde conversion to MEK acid ethyl ester tivity towards
mllh ratio acetaldehyde
Average Average Average Average Average
30 0.00 34.3 31.9 0.5 0.7 92.9
36.2 34.2 0.6 0.6 94.6
31.5 34.0 30.0 32.0 0.4 0.5 0.5 0.6 95.2 94.2

1:5 40.0 38.0 0.6 0.7 95.1

33.6 32.8 0.4 0.3 97.8
43.8 39.1 40.4 37.1 1.0 0.7 0.8 0.6 92.2 95.0

1:2 37.0 34.7 0.6 0.7 93.6

36.1 34.8 0.7 0.4 96.3
43.6 38.9 41.4 36.9 0.9 0.8 0.5 0.5 95.0 95.0

1:1 45.3 42.0 1.0 0.7 92.6

39.8 42.6 37.6 39.8 0.9 1.0 0.5 0.6 94.6 93.6

3:1 -
46.4 46.4 42.7 42.7 1.4 1.4 0.8 0.8 91.9 91.9

407
i';'<t'0-~-.-

Membrane name: N8x Catalyst: 3.00 g (14.4 wt% Cu on silica)

Reaction temperature: 300°C Catalyst particle size: 500-800 microns

Feed Sweep to Total ethanol Ethanol conversion Ethanol Ethanol to acetic Reaction selec-
rate feed molar conversion to acetaldehyde conversion to MEK acid ethyl ester tivity towards
mllh ratio acetaldehyde
Average Average Average Average Average
10 no sweep 59.4 50.4 2.9 2.3 84.8
58.2 50.3 2.2 2.1 86.4
I 54.5 51.2 1.4 0.3 94.1
53.7 56.5 50.0 50.5 1.7 2.1 0.4 1.3 93.2 89.6

1:5 58.5 52.0 2.2 1.4 88.9

60.4 59.5 49.5 50.8 3.7 3.0 1.9 1.6 81.9 85.4

1:2 67.7 58.6 3.8 1.4 86.6

66.9 56.0 3.4 1.6 83.7
56.3 63.6 51.9 55.5 2.0 3.1 0.6 1.2 92.2 87.5

1:1 60.7 54.4 2.1 0.8 89.6

66.6 63.7 59.7 57.1 3.0 2.5 1.0 0.9 89.6 89.6

2:1 62.4 56.3 1.8 1.0 90.3

63.5 57.9 2.0 0.6 91.1
60.8 62.2 56.6 56.9 1.9 1.9 0.5 0.7 93.0 91.5

4:1 65.4 60.7 1.8 0.5 92.8

63.7 64.5 60.1 60.4 1.5 _ L __ 0.4 0.5 94.4 93.6

408
 Membrane name: N8x Catalyst: 3.00 g (14.4 wt% Cu on silica)
Reaction temperature: 300°C Catalyst particle size: 500-800 microns

Feed Sweep to Total ethanol Ethanol conversion Ethanol Ethanol to acetic Reaction selec-
,
rate feed molar conversion to acetaldehyde conversion to MEK acid ethyl ester tivity towards
mllh ratio acetaldehyde
Average Average Average Average Average
20 0.00 47.8 45.1 1.2 0.3 94.3
47.8 45.8 1.3 0.2 95.8
46.2 47.2 44.7 45.2 1.2 1.2 0.2 0.2 96.9 95.7

1:5 48.3 46.4 1.3 0.2 96.2

47.7 45.6 1.3 0.2 95.6
55.7 50.5 53.0 48.4 1.4 1.3 0.2 0.2 95.3 95.7
i

1:2 58.7 52.6 1.3 1.8 89.8

57.0 51.9 1.2 1.5 91.1
53.8 56.5 49.3 51.3 1.0 1.2 1.5 1.6 91.6 90.8

4:1 57.4 52.9 1.1 1.3 92.2

54.7 49.9 1.0 1.3 91.2
52.8 54.9-
48.4 50.4
-------- -
'---- 0.9 1.0 1.3 1.3 91.8 .91.7

409
 APPENDIXF2

Plug flow reactor data and membrane

reactor data for 2-butanol dehydrogenation

410
$

Membrane name: NSx

Catalyst: 3.50 g (14.4 wt% Cu on silica)
Reaction temperature: 190°C
Catalyst particle size: 500-S00 microns

Feed Sweep:feed Total 2-butanol 2-Butanol to MEK 2-Butanol to SeleCtivity

mllh molar ratio conversion (%) conversion (%) 3-octanol towards MEK
conversion (%) (%)
Avg. Avg. Avg. Avg.
5 0.0 55.3 55.1 0.1 99.7
54.3 54.2 0.2 99.7
55.0 54.9 54.S 54.7 0.2 0.2 99.6 99.7

0.6 60.4 . 60.1 0.3 99.6

59.6 59.3 0.3 99.5
59.1 59.7 5S.7 . 59.4 0.3 0.3 99.4 99.5

1.6 64.4 64.0 0.3 99.5

64.9 64.5 0.4 99.4
63.S 64.4 63.3 63.9 0.6 0.4 99.1 99.3

3.1 70.5 70.0 0.5 99.3

70.9 70.2 0.7 98.9
67.3 69.6 66.8 69.0 0.5 0.6 99.3 99.2

6.3 76.6 76.1 0.6 99.3

74.4 73.7 0.7 99.1
73.6 73.1 0.5 99.3
73.7 73.7 0.0 100.0
76.7 75.0 76.2 74.5 0.5 0.4 99.4 99.4

12.6 S1.1 80.5 0.6 99.3

78.4 77.S 0.6 99.2
78.0 79.2 77.5 78.6 0.5 0.6 99.4 99.3

411
 Membrane name: N8x
Catalyst: 3.50 g (14.4 wt% Cu on silica)
Reaction temperature: 190 DC
Catalyst particle size: 500-800 microns

Feed Sweep:feed Total 2-butanol 2-Butanol to MEK 2-Butanol to Selectivity

ml/h molar ratio conversion (%) conversion (%) 3-octanol towards MEK
conversion (%) (%)
Avg. Avg. Avg. Avg.
10 0.0 55.5 55.3 0.2 99.7
55.2 54.9 0.2 99.6
53.8 54.8 53.6 54.6 0.2 0.2 99.5 99.6

0.8 60.3 . 60.0 0.2 99.6

59.7 59.4 0.3 99.5
57.6 59.2 57.1 . 58.8 0.5 0.4 99.1 99.4

1.6 58.9 58.6 0.4 99.4

59.4 59.2 0.2 99.6
59.3 59.2 59.1 58.9 0.3 0.3 99.6 99.5

3.1 61.2 61.2 0.0 100.0

59.3 59.3 0.0 100.0
63.3 61.3 63.3 61.3 0.0 0.0 100.0 100.0

6.3 62.7 62.7 0.0 100.0

63.6 63.6 0.0 100.0
64.6 63.6 64.6 63.6 0.0 0.0 100.0 100.0

412

,.
,
Membrane name: N8x
Catalyst: 3.50 g (14.4 wt% Cu on silica)
Reaction temperature: 190°C
Catalyst particle size: 500-800 microns

Feed Sweep:feed Total 2-butanol 2-Butanol to MEK 2-Butanol to Selectivity

mllh molar ratio conversion (%) conversion (%) 3-octanol towards MEK
conversion (%) (%)
Avg. Avg. Avg. Avg.
15 0.0 45.3 45.3 0.0 100.0
45.9 45.9 0.0 100.0
44.0 45.1 44.0 45.1 0.0 0.0 100.0 100.0

0.6 45.2 . 45.2 0.0 100.0

46.5 46.5 0.0 100.0
49.5 47.1 49.5 . 47.1 0.0 0.0 100.0 100.0

1.6 46.1 46.1 0.0 100.0

51.1 51.1 0.0 100.0
47.0 48.1 47.0 48.1 0.0 0.0 100.0 100.0

3.2 49.7 49.5 0.0 99.6

46.9 46.9 0.0 100.0
47.6 47.6 0.0 100.0
48.6 48.2 48.6 48.1 0.0 0.0 100.0 99.9

6.3 52.7 52.7 0.0 100.0

52.5 52.5 0.0 100.0
54.0 53.1 54.0 53.1 0.0 0.0 100.0 100.0

20 0.0 44.8 44.8 0.0 100.0

44.1 44.1 0.0 100.0
44.4 44.4 44.4 44.4 0.0 0.0 100.0 100.0

0.8 47.5 47.5 0.0 100.0

46.5 46.5 0.0 100.0
48.8 47.6 48.8 47.6 0.0 0.0 100.0 100.0

413
Membrane name: N8x
Catalyst: 3.50 g (14.4 wt% Cu on silica)
Reaction temperature: 215°C
Catalyst particle size: 500-800 microns

Feed Sweep:feed Total 2-butanol 2-Butanol to MEK 2-Butanol to Selectivity

mllh molar ratio conversion (%) conversion (%) 3-octanol towards MEK
conversion (%) (%)
Avg. Avg. Avg. Avg.
5 0.0 68.4 65.6 1.9 95.8
70.3 67.9 1.7 96.6
68.8 66.4 1.7 96.6
70.7 69.5 68.6 67.1 1.5 1.7 97.0 96.5

0.6 73.5 70.8 1.9 96.3

72.7 70.3 1.8 96.7
73.1 73.1 70.5 70.5 2.0 1.9 96.4 96.5

1.6 77.8 75.0 2.2 96.4

76.5 77.1 74.4 74.7 1.7 1.9 97.2 96.8

3.1 82.8 79.5 2.8 96.1

81.3 82.1 78.9 79.2 2.1 2.5 97.0 96.5

6.3 85.9 83.6 2.3 97.4

86.1 84.5 1.6 98.2
88.0 86.7 86.3 84.8 1.8 1.9 98.0 97.8

12.6 90.4 88.4 2.0 97.8

92.3 91.0 1.3 98.6
89.5 90.7 88.3 89.2 1.2 1.5 98.7 98.4

414
Membrane name: N8x
Catalyst: 3.50 g (14.4 wt% Cu on silica)
Reaction temperature: 215°C
Catalyst particle size: 500-800 microns

Feed Sweep:feed Total 2-butanol 2-Butanol to MEK 2-Butanol to Selectivity

mllh molar ratio conversion (%) conversion (%) 3-octanol towards MEK
conversion (%) (%)
Avg. Avg. Avg. Avg.
10 0.0 67.9 66.9 0.9 98.5
68.8 67.6 1.0 98.3
68.6 68.4 67.3 67.3 1.0 0.9 98.2 98.3

0.8 74.2 . 72.7 1.3 98.0

74.7 73.5 1.0 98.4
75.6 74.8 74.4 . 73.5 1.1 1.1 98.4 98.3

1.6 78.8 77.1 1.5 97.9

77.6 76.6 1.1 98.6
78.7 78.4 77.5 77.1 1.2 1.3 98.5 98.3

3.1 79.7 79.5 0.2 99.7

81.9 81.5 0.4 99.5
81.7 81.1 81.1 80.7 0.6 0.4 99.3 99.5

6.3 86.5 85.9 0.6 99.3

86.4 85.7 0.7 99.2
86.8 86.6 85.9 85.8 1.0 0.8 98.9 99.1

415
Membrane name: NSx
Catalyst: 3.50 g (14.4 wt% Cu on silica)
Reaction temperature: 215 DC
Catalyst particle size: 500-S00 microns

Feed Sweep:feed Total 2-butanol 2-Butanol to MEK 2-Butanol to Selectivity

mllh molar ratio conversion (%) conversion (%) 3-octanol towards MEK
conversion (%) (%)
Avg. Avg. Avg. Avg.
20 0.0 62.4 62.4 0.0 100.0
63.0 63.0 0.0 100.0
62.2 62.2 0.0 100.0
63.1 62.7 63.1 62.7 0.0 0.0 100.0 100.0

O.S 67.1 67.1 0.0 100.0

66.7 66.7 0.0 100.0
67.3 67.0 67.3 67.0 0.0 0.0 100.0 100.0

1.6 65.7 65.7 0.0 100.0

67.9 67.9 0.0 100.0
67.9 67.2 67.8 67.1 0.0 0.0 99.9 100.0

3.1 68.2 68.2 0.0 100.0

68.6 68.4 68.6 68.4 0.0 0.0 100.0 100.0

6.3 70.7 70.7 0.0 100.0

69.5 70.1 69.5 70.1 0.0 0.0 100.0 100.0

15 0.0 67.3 67.3 0.0 100.0

67.3 67.3 0.0 100.0
66.5 67.0 66.5 67.0 0.0 0.0 100.0 100.0

30 0.0 54.8 54.S 0.0 100.0

56.S 56.8 0.0 100.0
54.8 55.4 54.8 55.4 0.0 0.0 100.0 100.0

416
Membrane name: N8x
Catalyst: 3.50 g (14.4 wt% Cll on silica)
Reaction temperature: 240 DC
Catalyst particle size: 500-800 microns

Feed Sweep:feed Total 2-butanol 2-Butanol to MEK 2-Butanol to Selectivity

mllh molar ratio conversion (%) conversion (%) 3-octanol towards MEK
conversion (%) (%)
Avg. Avg. Avg. Avg.
5.0 0.0 78.3 76.7 1.1 98.0
78.3 77.2 1.1 98.6
79.2 78.6 77.6 77.2 1.1 1.1 98.0 98.2

0.6 83.4 81.6 1.3 97.9

85.9 84.7 83.9 82.8 1.6 1.4 97.6 97.7

1.6 85.8 84.8 1.0 98.9

86.1 85.9 85.1 85.0 0.9 1.0 98.9 98.9

3.1 88.8 87.9 0.9 99.0

88.8 88.0 0.8 99.1
90.4 89.4 89.4 88.4 1.1 0.9 98.8 99.0

6.3 93.1 91.7 1.4 98.5

92.0 90.8 1.2 98.7
92.8 92.6 91.7 91.4 1.1 1.2 98.8 98.7

12.6 95.0 93.5 1.4 98.5

95.4 94.4 1.0 98.9
95.9 95.4 94.4 94.1 1.5 1.3 98.5 98.6

417
Membrane name: N8x
Catalyst: 3.S0 g (14.4 wt% Cu on silica)
Reaction temperature: 240°C
Catalyst particle size: SOO-800 microns

Feed Sweep:feed Total2-butanol 2-Butanol to MEK 2-Butanol to Selectivity

mVh molar ratio conversion (%) conversion (%) 3-octanol towards MEK
conversion (%) (%)
Avg. Avg. Avg. Avg.
10 0.0 79.5 79.3 0.2 99.7
77.9 77.6 0.3 99.6
79.6 79.0 79.3 78.7 0.3 0.3 99.7 99.6

0.8 84.4 84.1 0.4 99.6

82.8 83.6 82.1 83.1 O.S 0.4 99.1 99.3

1.6 8S.8 8S.5 0.4 99.6

87.0 86.4 86.6 86.0 0.4 0.4 99.S 99.S
.

3.1 89.6 89.2 0.4 99.6

88.0 87.6 0.4 99.5
88.4 88.7 88.1 88.3 0.3 0.4 99.7 99.6

6.3 91.3 91.0 0.3 99.6

90.9 90.5 0.3 99.6
90.S 90.9 90.S 90.7 0.0 0.2 100.1 99.8

418
Membrane name: N8x
Catalyst: 3.50 g (14.4 wt% Cu on silica)
Reaction temperature: 240°C
Catalyst particle size: 500-800 microns

Feed Sweep:feed Total 2-butanol 2-Butanol to MEK 2-Butanol to Selectivity

mVh molar ratio conversion (%) conversion (%) 3-octanol towards MEK
conversion (%) (%)
Avg. Avg. Avg. Avg.
20 0.0 82.4 81.2 1.2 98.5
81.8 80.6 1.1 98.6
81.8 82.0 80.4 80.7 1.4 1.3 98.3 98.5

0.8 82.2· 81.3 0.9 98.9

86.1 85.3 0.7 99.2
85.6 84.6 84.8 83.8 0.7 0.8 99.1 99.1

1.6 88.4 87.9 0.5 99.4

87.4 87.0 0.4 99.5
88.9 88.2 88.2 87.7 0.7 0.5 99.2 99.4

3.1 87.9 87.4 0.5 99.5

88.8 88.5 0.3 99.6
87.1 87.9 86.7 87.6 0.3 0.4 99.6 99.6

6.3 89.2 88.5 0.7 99.2

91.2 90.2 90.9 89.7 0.3 0.5 99.6 99.4

30 0.0 77.4 77.1 0.3 99.6

78.2 77.8 0.3 99.6
79.1 78.2 78.7 77.9 0.4 0.3 99.5 99.6

40 0.0 78.8 78.7 0.0 100.0

77.0 77.9 77.0 77.9 0.0 0.0 100.0 100.0

419
 APPENDIXG

Sensitivity analysis for the catalytic

membrane reactor model

420
 The following conditions were studied:

1. The effect of reaction rate k'-values at standard conditions.

2. Effect of acetaldehyde adsorption coefficient at standard conditions.
3. Effect ofR2 to N2 selectivity at standard conditions (Pm,hydrogen is constant).
4. The effect of the effectiveness factor at standard conditions.
5. Effect of ethanol feed flow rate at standard conditions.
6. Effect ofN2 sweep gas to ethanol molar feed ratio at standard conditions.
7. Effect of hydrogen permeance at constant selectivity and standard conditions.
8. Effect of H2 permeance at constant permeances for N2, ethanol and acetaldehyde.
9. Effect of hydrogen permeance at varying selectivity and non-standard conditions.
10. Effect of hydrogen permeance at constant selectivity and non-standard conditions.

Note:
Each of these ten investigations consists of six different figures. For the first three figures, the
conditions at the membrane tube exit are used. For the last three figures, data is plotted along
the axis of the membrane.

The ethanol conversion in the first figure on each page is defined as follows:

X = Fet(O),t -F-F
et,t ct,!
(G.1)
ethanol F
et(O),t

421
 Standard conditions for ethanol dehydrogenation

Reaction T (0C) 275

Ethanol feed flow rate (ml/h) 10
(moVs) 4.771 *10·5

N2 sweep to ethanol feed molar ratio I

Shell pressure [Pa] 100000
Tube pressure [Pal 100000

Reactor length [m] 0.18

Rm[m] 0.0035
Ph [kg/mj] 430
m [kg] 0.0030

Eb 0.4

k [mol/kg cat.s.Pa] 1.27*10-0

.

Kethanol [Pa-'] 4.32*IO-v

K.cetaldchyde [Pa"'] 2.91 *1O"U

Khydrogcn [Pa"'] 6.57*10"0

Keq [Pal 24080

1] I

Pm,hydrogcn [lJ.IIIoVm< .Pa.s] 3.78

Pm,nitrogcn [nmoVm<.Pa.s] 21.29

Pm,ethanol [nmoVm .Pa.s] 5.32

Pm,acetaldehyde [nmoVm".Pa.s] 5.32

422
 Shell and tube flow profiles for standard conditions

Flow profiles on tube side for standard

membrane reactor
6.0E-05 --- -_._-_._._._----------,
i
5.0E-05
!
~

'"
i
~
-~
4.0E-05

3.0E-05
""'"----- Ethanol
II
...
I;:

/- ---
.!!! 2. DE-05
o Acetaldehyde
Hydrogen
== 1.0E-05 I
NHrogen I
O.DE+OO
o 0.2 0.4 0.6 0.8
Dimensionless length

Flow profiles on shell side for standard

membrane reactor
------ .---

-
1.0E-05
Hydrogen
_ 8.0E-06
~
o
§. 6.0E-OO
~
~
I;: 4.0E-06
/
So
== 2.DE-OO i =
Ethanol lop
=
Aclltaldehyde bottom
-
O.OE+OO
Il
o 0.2 0.4 0.6 0.8
Dimensionless length

423
1. The etTect of reaction rate k' -values at standard conditions.

Effect of reaction rate on ethanol conversion and Errect of reacUon rate on shell side exit
percentage of ethanol feed lost through membrane composlUon (excluding sweep gas)
--
.."
,,------.--.~-----------.------ 90

,
--- ..
• / ..
e-
7
l """ ,
'-- -
------ .. ~-- -------- - ------ --- ---
-.- Ethanol
·· .. ··AcetaId.~
--Hydrogen
.. J ~

T __ Converalon (1-'\ axl.)

" t
,. IJ I -.- Lost r.,d (right axil) I
o
", 2 3 4 5 8 7 • 2 3 4 8 7
Multlp" ohtand..-d k-value Multlpl. rI standard k-valu.

Errect of reacUon rate on tube side exit H. driving force (PrP.) vs. axial position

i"
so·

50
\
, -~-
composlUon

I--Ethanol
I·· ..·· Hydrogen
'> _.. ------------4--------------------------
-'-_'d_~1
-." Nitrogen

~
;;
12000

10000

8000
._--------------

0._
Reaction rate cCMffIcIenW;
O.6k, k, 2k, 4k, '"
for linn from bottom to top
----,----

~
8000
•-
30

i .......... ....... .......... ...... If h

..
~
Q. 20 ........ ..............
£ 4000

10
...: 2000 ~
Ir--·~
o
0
0 1 , 3 4
• 6 7 8 o 0.' 0.4 0.5 0.'
Multiple of standard k-value Dimen.ionle•• length

Normalised ethanol flow on tube side Percentage of produced hydrogen

permeated to shell side
0.9
~ eo r----------
0.8 .'\'-....------
"--.: :---------
0.7
---- 50

V
f
'\ 0.'
lL 0.4
0.6
" ----'''----. = ..__.,.-
- -~ f
l Reaction ,.. coeft'lclentl::
o.21k for bottom and 8k for tap line
I
0.3 20
Reaction rate coftflcl.nt.:
0.' 0.261e, 0.11e, k, 211, 4k, SIr;
for lin•• from top to bottom 10
0.1
o o
, 0.' 0.4 0.6 0.8 o 0.2 0.4 0.5 0.9
Dimensionless length Dlmen.~nles. length

424
2. Effect of acetaldehyde adsorption coefficient at standard conditions.

Effect of K.c.t on ethanol conversion and percentage E1'fect of K...t on shell side exit composilion
ethanol feed 100t through membrane (excluding sweep gas)
• -- - ---_._- u 11 - - - - - - - - - - - - - -.-_ _ _ _ _ _ _ _ _ _ _ _ _ 90

---- ~. - ---- --
- 2.3 ., ~
•,
10

--- .
...e
-_ .. -
r'~·-
2.1 li·
-~
!• ~
------ .. •• "l-

-.- --
-- -- -
~ 1.9
1~
i I ii 7
._.a--
--- -.- ElhanCII
--- .. ~ ~
-- -'- ~ ..
i ··.··AcNld~

Ii •
".-
~- ~
Ii
I --+-Converslon (left axis) 1
" ..................... ...............
__ Hydrogen
-.- Lost feMi (right axle)
1.7
5 "
40
o U U U U ~ ..
fAlltlpla of K....
u .. U
1.5 4
0.1 02 0.3 0.' 0.5
Multlpl. rif K.at
.. 0.7 ,. 0.' 1
eo

E1'fect of K...t on tube side exit compos Ilion Hz driving force (PrP.I vs_ axial position
6000 ,-------- ._-
4500
4000
t\
';" 3600
\~ ,----
···l\~ IEffwt of acetaldehyd. adIIorptk'ln:
I!:. 3000 O.UK, 0.1K, UK, D.I5K, G.70K, K -
~ 2500
\\\\ for linn from top to bottom
-
~ 2000 \\~
:£ 1500 \\.'\.
1000
""~
"--~

~
500
' ....
o
~1 U U U M MUM U o 0.2 0.4 0.6 0.'
Multiple of K"" Dimensionles. length

Normalised ethanol flow on tube side Percentage of produced hydrogen

~--- -- -, permeated to shell side

,,.
0.'
I\, 50

0.8
~ 45

0.7 ~ 40

~ 0.6
'~ -----.::: 35
J
"'-----:::::-- 30
' \ 0.5
lL 0.4
~-
.-
",. 20
I EffKt of ac.tlIldehyde adsorption:
O.OI5K, 0.1K, 0.2K, 0.6K, 0.75K, K
0.3 r--- Effect of ac.taldshyde ad.orptlon: I.
0.2 r--- Q,OOK, o.1K, O.2K, o.OK, 0.76K, K
for lin.. fro m bottom to top
ID
I
0.1 5
o o I
o 0.2 0.4 0.6 0.8 o 0.2 0.4 0.6 0.8
Dimenslonle.s length Dlmenstonless length

425
3. Effect of Hz to N z selectivity at standard conditions (Pm.hydrogen is constant).

Effect of membrane selectivity on ethanol conversion Effect of membrane selecUvlty on shell side
and percentage ethanol feed lost through membrane
exit composlUon (excluding sweep gas)

....• .. •
45 - - - - - -_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ .. 14
50 --.--- 100

\., I --+-COnversion (left Uls) I 80

1Jl.+U
c , -.- Lost feed (right axill)
" ,a ~
80

1!•
! .....6
,
•
10

• 1"
-J
2
l' }"
;30
,
/
../"
--- EthanDi
..• ". Acetaldehyde
~
f--- eo
70
..
" t
1.. . . II/
g /'\
h ~" -+-Hydrogen

ii I M J
4 ,
•, ,
~20
•li15 / . " I
• !III.
"
30
~
'" 44.2
.. --------------
---.
iii 10 --.........
~.,---.
------- ---. 20

.. '"
0 0
100 10>
. ................. 10
0
0 50 100 150
~N, ..Iec:tlvtty "" "" '" 50 200
H,:N2 ·elacttvity "" 300
'"

E1Tect of membrane selecUvlty on tube side Ii, driving force (PrP.1 vs. axial position
exit composlUon
45 - r - - - -, 3600 . , . - - - - - - - - - --
40 3000 A

&30
~,.
35 '.
/,----_.--
---
./' ----
_____ -----------------------a ~

~2000
2500
N
HycIrogan to nItrOgen MttctlVlties:
10. 20,. 40., ID, 1eo, 320 _I
1'-:'--- , .; 1500 (\ for On" from bottom to top

~ 20 I ............... ,...•'..:::... ..................... ................ ............. ...........

II. 15
!£
£ \,
'- 1000
10 ~ --+- Ethanol -.- .AoetaIdellyde
l__ ·
o
•• _ - •• ---. - •• ". ".0 • •

500
~
..•.. Hydrogen
~
5 -." Nitrogen
0 0
0
"" 100
'" 200

H,:N, seloctivily
250
"'" 350 0 0.2 0.'
Dimensionless length
0.' 0.' 1

Normalised ethanol flow on tube side Percentage of produced hydrogen

1\""- permeated to shell side
0.9
l 80 'O-
0.' ~
-~ 50
0.7

i 0.6
-...::. -
tt-=------
40
~----" .
---~,,-
lit 0.5
IL 0.4
0.3
'.-

Hydrogen to nitrogen ..IMrtlvltles:

.~~"~ "30 I
20
---------- --- -- --. -- "-_.-

~ I
10, 20, 40, 80, 110, 320 Hydrogen to nitrogen ..1.ct1¥1t...:
0.2
tor Inn from bottom to top 10 10, 20, 40, 10, 110, 120
f
0.1
o o
for Un.. from bottom to top
,
o 0.2 0.4 0.6 0.9 o 0.2 0.4 0.6 0.8
Dimensionless length Dlmansionless length

426
 --

4. The effect of the effectiveness factor at standard conditions.

Effect of effectivene•• factor on ethanol conversion and Effect of etrectlveness factor on shell side exit
. percentage ethanol feed lost through membrane

" "
composition (excluding sweep gas)
..
~ ,. •
------ ..
•" , ~
m i'20 "
""".,'/ 2
l'
'Ii • L -.- Ethanol
....... AcMaIdehyde
/

--V . . _-----.---___
"
82
i'

7 ---,-"'--- -., - .-. ••

,
-~

•tIi
" ,
2.8 1~ ~ . ---Hydrogen
---a _______ '" r
I
7r
,
Ii
/
-,
- '-, 78
I
__ COnversion (left Dis) 2.3 iti
~

4i .............. , ... ................ .

_-
5
-.- Loetfeed(rlghtuls)
1" 78

o ~ u u U U M
err.ctlYene•• factor
U U U
2 0
0 0.1 02 0.3
.........
0.4 .. 0.' 0.7 . O. 1 "

Effect of etrectlveness factor on tube side exit Hz driving force (P,.P.) vs. axial position
composition ,.----
I

.-
70
r__ Ethanol
-.- Ac.u.ldehyde I 3000
60 -I------~
~ T·.·· ......... -e-. Nitrogen r 11\\
ett.ctlvt'n'" hlctor:
8. " ~ 2000
11\ 0.2, 0.4, O.e. 0.8, 1.0
for lin.. frum bottom to top
J!40
c
• --------- ~ IA~
~ 1500
'\..~
30

20 £ 1000

10 500
~
0 o
0 0.2 0.4 0.6 0.8 o 0.2 0.4 0.6 0.'
Effectiveness factor Dimensionless length

Normalised ethanol flow on tube side Percentage of produced hydrogen

---- permeated to shell side
0.'
~ .
50 ,---

i
0.8
0.7
0,6
""". 40
35
--, I
#'

EffKtlYe"_ fador.
~ 0.5
... "
30
I 0.2 and 1.0 r
IL 0.4 l-- Eff.ctlvenen factor. 20
0.3 l-- 0.2, 0.4, 0.1, O.B, 1.0
15
for lin.. from top to bottOm
0.2 10
0.1 5
o o
o 0.2 0.4 0.6 0.8 o 0.2 0.4 0.6 0.6
Dimensionless length Dimensionless length

427
5. Effect of ethanol feed flow rate at standard conditions.

Effect of ethanol feed rate on ethanol conversion and Eflect of eflecllveness factor on shell side exit
perc:entage ethanol feed lost through membrane
eo .. _ _ _ _ _ _ _ _ _._____ • composilion (excluding sweep gas)
--.~--.'------

25 - - - - . - - - - - _ _ _ _ _ _ ____.______ ..

",. ---------- ..
J ~.:~.:;::I.hyd. ~,:.:-..",,~e:...----------- 82 "-
i ~Hyd J7 -------.------___ ._______ ~ t
15 ....n
~1O+_------~L-------------~~~=-~
/
L l---------'------,~==
;;
.. ...... =
= .......'F=
.... =
.... ---4
"
f

1'·"" 'l<=
....

"
~+----~-----~-----~----+
2 2.' ~ u u u u u u u
- ....-
,+------------------------------------+~
u

Eflect of eflectlveness factor on tube side exit H. driving force (PrP.) ys, axial position
composition
"'00 ._---_._._----
ro.-----------~==============-~-
1--- Ethanol -.- AHlaldehyOe. ~
oo+----~-----~
~ I·· .. ·· Hydrogen -." Nitrogen 3000+~~~~~
m f""
5Ot======~~
..... 2500 Mullp" of ethanol rat.:
J.I 0.1, 1, UI, 2,. 2.15 - for linn from
:l. ~ 2000 -II---'lA~"---I bClttcm to tap III r1ght side of graph
J!4O

~
5
~ ~+----------------------=~~----~~~~~
~----~----- --- ~ 1600.JI~--~~~~~======================~--~
e;:
£ 1000 -1-----'._--=>.,,::,...,;"'--.;;;:"---==-________________--1
20+-------~~~~----------------- . ..~
.•~..~~~c{
",,_. on::::' -:;.:.~ ... :::::..:: ~.'~ '_" '.'~ ".:.":
10+_----~~~~~~~~~~--~------~

o~----~----~----~----
o 0.2 0.' 0.6
__----~
0.8 o 0.2 ,.• 0.8 0.8
Effectivenes. factor Dimensionless length

Normalised ethanol flow on tube side Percentage of produced hydrogen

permeated to shell side
0.9 ~ 50 -
.....~ 45
0.8
----= 40 f /L'/:'-
0.7 1/// .

..,,, m
36
i O.S ~
"i. 0.5
LL OA f-
0.' f -
Multi" of ethanol feed rate:
0.5, 1, 1.0, 2, 2.11 ~ for II,," from
bottom to top at right .1_ of graph
.,. l Multiple of IItllanal fMd rate:
0.6, 1, 1.6,2, 2.11- for """ from
bClttam to top ill: left sdgs at graph

0.2 f---- '0

0.1 5
o o
o 0.2 0.4 0.6 0.8 o 0.2 0.4 0.6 0.8
Dimensiontess length Dimensionless length

428
6. Effect of N2 sweep gas to ethanol molar feed ratio at standard conditions.

Effect of sweep:feed molar ratio on ethanol conversion Effect of sweep 10 feed molar ratio on shell
and percentage ethanol feed lost through membrane
30
side exit composition (excluding sweep gas)
- - . -___ •_ _ _ _ _ _ _ _ •_ _ _ _. _ _ _ _ _ _ _ _ _ "
.. --------.-------.------------ 3

~
55 +----~===~_--------_..,.,
.. ---------
.~~.~-~~~ .... e~ ~~ ,.11'\"-----:/:::::---==~~~=1 ..
j ~ 2 ~ •
~. 20 +-"-.,L-----------...j .. ,. .. Acetmdehyde I---
-.- Eth.nol 85
....
i~
" WVL_____~~=~=.... :E"=='_____.: i
o ____
S!
-!

f-- "U•
80
i",
~,,_
,./
.---______---1 '1
15

"I-
11O~~~~----------------------~
1 - - COnwrwlon (left axlfl) \

t
i .. , 1
-.- ' iII~
m '....... '~~~-- .. ____________ .--______________________
..
Lost feed (rlgtd axle) ';It. 70
" +--------------------+ •. , .. ,...........................
M+--__--_--_----_--_--__--__---+. 'I •
O+---~ __~_____--~--~----~--__---480
, 2 , o 2345678
s-p g •• to.thanol f ..d molar ~o Sw..p gas fa ..... anolfHd molar ratio

Effect of sweep to feed molar ratio on tube H. driving force (PrP.) vs. axial position
side exit composition 5000,-----------------·--------·---
46 - - - - - . - - . - - - - - - - - - - - - - - - - -
~O~~Ar\----------------------------~
~~~~--------~~~--~--~--~--~-~=---~
---------- ~----
.... ~~~~ ..
., 30 ._ 35t:Z'-=:' 1-- Acetald.hyd't Ethanot ---
'j;' @o
e,.
3500 W\ \\\'=:r=:=====~
tf.:-,,:t=
,---i
#>Tri,cn~,\-
Nltrogens....pto.thenolhed-

f:t:S··~;.:=~============~·=··=..=~~~=~==n====-.==.N=.=Mg==.n==~~
lOOO molar nIHil); 0.26,0.5, 1,2, 4, a -

e::~ j=~'\~\"~~==~==~f~
..~H~n~n~f~~~~~.~m~w~~~p~==t-==j
'I- 2500 ~
. 2000 \
~ 15+---~·..~··~ __------------------~ \"-->."::''''~~''''''''''--------__1
.
~ :~!::"'~".:;:.:' .-_._+ '-'-' -" _._.-.-. _. _.-' _.-._._' _._.
:£ 1500 -t\-"....-->"

'~t~\,,~··~"~~~~~~~--~~--~~~~~~~~
10~~~~~~~~~~~~--------_1 ................. .
5+----------------~~~=t
0+-___- __---__--__--__----__--__---4
o 2 3 4 5 e 7 8 o 0.2 0.4 0.• 0.8
Sweep gas to ethanol feed molar ratio Dimensionless length

Normalised ethanol flow on tube side Percentage of produced hydrogen

permeated to shell side
100
0.' I
0.8
0.7 "'" ' - 90
80
70
/'
i O.S //
' \ 0.' ~ ;I!.
.0
50
/1/
IL 0.4 f.- Nltrog_n sweep to .thanol feed 40
[f,__ NltTogsn IlWeePta sthanal fHd
molar nitta; 0.28,0.1, 1, 2, 4, 8
0.' f . - mol... ratio: 0.20,0.6, 1, 2, 4, a
~ for In•• from bottom to tap
0.2
for IInM from top to bottoIfI
"20 ~.
.. ...- -- - _._- .----.--~

0.1 10
o 0
,
o 0.2 0.4 0.6 0.' o 0.2 0.4 0,6 0.8
Dimensionless length Dimensionl••• length

429
7. Effect of hydrogen permeance at constant selectivity and standard conditions.
For selectivity to be constant; N 2, ethanol and acetaldehyde permeance have to change with H2 perrneance

Effect of Pm,H2 (con,t. selectivity) on ethanol conversion Effect of Pm,H2 (at constant selectivity) on shell

. ------------_.
and percentage ethanol feed lost through membrane
-- 8
30 I"
side exit composition (excluding sweep gas)
-- 100

#- 44.8
-- -
,.,.-,,-- 7
-.- Ethanol
-- 00

1 ~ -' ,
,
--
i: 44.2
.- -- ---
..i•8 7 - ..- --
43,8

iT ~~~~'
, --
t
0( 43.4 - I --+- Converalon (I'" axis) !
,
.-•
~.- LCIat fHd (r1ght axla)
/-'

o
0.5 1.5 '_5
Multiple ofP• ..., .......
3.5 4 o 1.5 2
Multiple ~ P..".,.....
2.5 3 3.5
•
.

EITect of Pm,H2 (at constant selectivity) on tube H. driving force (PrP.) vs_ axial position
side exit composition 12000 , - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
--------------..:.--------~
50
Hydrogln perm•• nee WIIh cormant
.5+.~~~---------------~ 10000 r----.. seteutlvlty: D.2lPm, O.6Pm, 1Pm, 2Pm, 4Pm-
____
.o+--~~~~- - - - - - - - - - - - - i
I
"'" ~
for lin.. from tap to bottom

.
E~i::::~~~~~~~::::~~======;;~::===i
··a _____ _
30

825+----------------"'_-,~==-~~-
'ii'
I!,.

~
8000

8000
~
!£ ~ ~
1 : 1:=:--':-:.--:-':.=:-=. ':.:---~--~-:::--~.-'~-'~,.~-~-~--.~---;---:-:- .:- :- ~-.-;- ~- ~- -;- ~- ~-.-;,- ~- .;,-.~- -
10
5
t /
-I-_~.c::_____-I .. " .. Hydrogtn
o .....
1 --+- Ethanol -.- AcftI1del¥l;r
-.' Nitrogen r
:£ 4000

2000

o "'-
~
~ "'"
~"'( ____
---- - .
--------
o 0.5 1.5 2
Multiple of Pm,hrlll'Ogln
2.5 3 3.5
• 0 0.2 0.'
Dimensionless length
0.8 0_8 1

Normalised ethanol flow on tube side Percentage of produced hydrogen

1':.
0_" --.;;;;;:
--
. 48
r·
permeated to shell side

0.8
~ (77/' .--
i
0.7

0,6 "" .-<:..::..

.~.:.:-.
40
36
1//
I /"
--
--. - .- 30
!!;.0.5
I.L 0.4

0.3
0.2
B Hyd_n ..mo..,•• wHh
Mlectlvity. O.26Pm, O.I5Pm, 1Pm, 2Pm, 4Pm-
fOr Iinetl from top to bottom
""""ont
.- ;1<"
20
15
II
,-
7
/
/"

-I
/"

I Hydrogen permeance WIIh conetant .e.ectMty:

o.Z&Pm, O.I5Pm, 1Pm, 2Pm, 4Pm - for In. . from
bottom to top Itt ten .dge of graph
t
10
0.1 5
;/
o o I
o 0.2 0.4 0.6 0_8 o 0.2 0.4 0.6 0.8

Dimensionless length Dimensionless length

430
8. Effect of Hz permeance at constant permeances for Nz, ethanol and acetaldehyde.
This implies that the selectivities will increase with an increase in Hz permeance

Etfectof p .... H2 (conal other p ..·.lon ethanol conversion Effect of Pm,H2 (at constant other Pm's) on shell
and percentage ethanol feed lost through membrane side exit composition (excluding sweep gas)
45 r--------..- - - - - - . - - - - - - - - - . - -

".. 44.7 + - - - - - - - - - - - - - - - - - - - - \ 2 . 4
2.5

1m
.,
12,-------------------------------
~ 10~~~--~~------~~-----------------i~
00

~i --.---------- ----------- ~- •

iI i
... ---
~44.4·- 2.3.~
8 ( .iii
] 441 2.2 H 8 ..."'.=. =
.•.=
t{,.,", ....=
..."'•.=...=...=. =...=...=...=...='.•=========t:
...
l
t I -+- COnv.....on (left axle, I ~
... 4t----------------------r=_.=_=a.==~==m====,---~
00( 43.8 _._ Lo.,ueod (right axil;, f---+2.1 ~ 2 ~---------__I ..•.. AcftaIdehyde _ 82

<J.' -I-_ _-_--~-_-_----_--+2

0.' Hi 2.5 4 ,.• O~
I----r----r---__----r-~---~=~==....
~=·==~---l 3 4
~

Multiple of p........

Effect of Pm,H2 (at constant other Pm's) on tube H. driving force (PrP.) vs, axial position
side exit composition 12000
45 . - - - - - - - - - - - - - . - - - - - Hydrogen permeanee WIth conatant eU.nol,
acetaldehyde WId nitrogen permeIUIee:

.
40 10000

t: .-.--..------.-------------.--------------------------
35

I - t - Ethanol -.- Ac.t,ldehyde If----l

I!:.
SODO
O.28Pm. UPm, 1Pm, 2Pm, "Pm
for linn from top to bcIItom

~
-1--------11............. J 8000
~20 - ... NHl'OSIen
61 ........................... ...................................... __ •••••••••••••••••••• _
:. 1. £ 4000

10
2000
• -."-~--
0 0
0 D•• 1.' 2 2.' 3 3.' 4 0 0.2 0.4 0.8 0.8

MultipleofPm~ Dimensionless length

Normalised ethanol flow on tube side Percentage of produced hydrogen

- permeated to shell side
D."
,\. 80 ---
........
0.8
0.7
--...... 50
..
i 40 1// =I
0.6
'(I ----
\ 0.'
.. 0.4
0.3
liydrogln ~rm.anctl with constant ethanol,
acetaldehyde and n"rogen penneanc:e: 'I // '" ------------
Hydrogen permeanee with co....nt ethanol,
~
acetaldehyde and nHrogen permeMce:
o.28Pm, o.ePm, 1Pm, 2Pm, 4Pm
0.2 / o.26Prn, O.5Pm, 1Pm, 2Pm, "Pm
10 for 11M! fram bottOm to top -
0.1 V
o o
o 0.2 0.4 0.6 0.8 o 0.2 0.4 0.6 0.8

Dimensionless length Dimensionless length

431
9. Effect of hydrogen permeance at varying selectivity and non-standard conditions.
Four times standard sweep gas flow rate and four times standard k-value

Effed of P"",M2 (varying selactivity)on ethanol conver- Effect of P",H2 (varying selectivity) on shell
sion and lost feed through membrane (4k, 4 x sweep)
------ .. ,_._--------
side exit composillon (excl. N2 ; 4k, 4 x sweep)
-;==::===;--,"
. . •\
05 ~-.---
1.8

-.- Ethanol

~83
V",
L .
#.2t'·----------4
}
'
\
.. " .. Ac.tllldehyde
-+- Hydrogen
1---+"

-3 ,. l-I-\".=!=:::!:::===========l03 ~
: 62
c
~0 --- Y'

e
i
i Iii
61

t~.. ~
c
•
tOO
.
.t !l9
I . . . . . . COnv....lon (left axis)
/ -.- ~Ifeed(rlghtax")
~
ii 3.3
3.8 ...

'-•
It.• ....... ' It···

__ _____ __
i2

91

" 3+---~---- __- - - r - -______- r____r -______ ~~

57
0 0.5 1.5 2.5 3.5 , o 0.' 1.5 2 2.5 3
•
Multiple of P......-- Multiple ~ P""IIYdN9...

Effect of P",H2 (at varying selectlvlty) on tube H. driving force (P.-P.) vs. axial position
side exit composition (4 x k, 4 X sweep) ~oo --------;=~~~~-=========il
00,---- Hydrogen perme.nee wtth verytng ••Ieo-
tlv1ty (rate • 4k and 4 tim .. SWMP rats):

~t~~-=---~--~-~---~~========~1
o.2IPm, o.8Pm, 1Pm, 2Pm,04Pm
•
".-- .1 ........... Ethanol
for tines from top to battom
Ii'
& ".)-=--------I ..•.. Hydl'f;Jgln
!
-.- Acetaldehyde/

L-~~~_~~~
-"'Nltrogen r ~ '~IJO HP"-----"''''''-----------4

~30~~~==~======~==============~ ~
~ 10000 tf:-"'--",,------"""---:::--------t
~wt--------------~ £

o~--_-- __----_--_--_--_----_--~
o 0.5 1.5 2 2,5 3 3.5 4 o 0.2 0.4 0.6 0.'
Multiple of PIIl,hylhgen Dimensionless length

Normalised ethanol flow on tube side Percentage of produced hydrogen

permeated to shell side
0.9 Hydrogen permeance WIll varying ••IK- - 80 ,-- - ----,
0.' \
tlYlty (rate· 4k and 4 tllMll SWHp nne): - 80

----------
O,2IPm, O.ePm,1Pm, 2Pm, 4Pm
0.7 '\. for 11M. fram top to bottom - 70 //
......... 00 '// ~
; 0.6
~ '/ / ----~
~-----
' \ 0.'
#" /
LL 0.4

0.3
0.2
30
20
! ..
/"
/ ...---
Hydrogen permeanee wtth varylnll_.I.e-
tlvtty (..... 411 and" tim" sweep me):
o.2t5Pm, o.8Pm, 1Pm, 2Pm, o4Pm
0.' 10 / for IInH from bottom to top on ten: edge

o 0
o 0.2 0.4 0.6 0.' 0 0.2 0_4 0.6 0.'
Dimensionless length Dimensionless length

432
10. Effect of hydrogen permeance at constant selectivity and non-standard conditions.
Four times standard sweep gas flow rate and four times standard k-value

Effect of PIII,M2 (consl selectivity-)on ethanol conversion E1'Iect Or Pm,H2 (at const selectivity) on shell
and lost feed through membrane (4 x k, 4 x sweep) side exit composition (exel. N.; 4k, 4 x sweep)
" -.-.-.-----.---.----------,,-~
--100

"

o 1.5 2.5
MulUpi. rif p ....h.,.......
3
•

E1'Iect Or Pm,H' (at constant selectivity) on tube H2 driving force (PrP.l vs, axial position
side exit composition (4 x k, 4 x sweep) 2~00r-.------r_=-~~-=-==_=_=============~
60
Acatald.hyd~
Hydrogln perm•• nc. with COMtIInt ....c-
I--+-EthanOI -.- UYIy (rata. 4k and .. times swap rate):
50 • ,--.",--. ...... Hydrogen - ... Nitrogen 20000 +---,.--.~-----! O.2fPm, O.6Pm, 1Pm, 2Pm,"Pm
for linn from top to bottom
------------- ';'
~ lrooo+4~~--~~~--------~

e:.~ 10000 -ll-.....:"'---",,-------=""-~-----___<

-£
10 +-•.c....&.~.:::...::!.'..'"."
:::"."."..,,<!.t'...,,-::~.::""~..,,...".."...:::""'...",."""":::..,,..."":::,,."":::..."..".."",,..:::..,,..."..:::...".."•. ".. ~
k

o 0,5 1.. 2 2.' 3 3.' 4 0.2 0.4 0.6 0.8

Multiple of Pm~n Dimensiontesslength

Normalised ethanol flow on tube side Percentage of produced hydrogen

permeated to shell side
90
0.' Hydrogen permeance with content .... c- f--
1\ tlvlly (rata. 4k and 4 tim" .weep rate):
f-- 80

-------
0.8

"
G.2I5Pm, o."Pm,1Pm, 2Pm,"Pm
0.7 for liMa from top to bctI:om f--I 70
~ Ill/'
i 0,6 ..
80
17 / .' ~
"i. 0.' '
~.",-: . " ... ~

40 / ,/
,,
lL 0.4
0.3
-----:.•. .. -.. ". ." 30
1/ .' /,/
H)'dfDPn penn8anctl with con.tant ....c-
20
. /' tlvlty ,fIlM -
o4k and 4 tim.. - p rabJ;
0.2 D.21SPm, O.ISPm, 1Pm, 2Pm. o4Pm
0.1 10 / for liMa from bottOm totop on left eda-
V/
o 0
o 0.2 0.. 0.6 0.8 0 0.2 0.4 0.6 0.8

Dimensionless length Dlmensionles. length

433