University of Southern California Dissertations and Theses

A flow-through membrane reactor for destruction of a chemical warfare simulant

(USC Thesis Other)

A flow-through membrane reactor for destruction of a chemical warfare simulant

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
A FLOW-THROUGH MEMBRANE REACTOR FOR DESTRUCTION
OF A CHEMICAL WARFARE SIMULANT

by
Mirmohammadyousef Motamedhashemi

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(CHEMICAL ENGINEERING)

December 2012

Copyright 2012 Mirmohammadyousef Motamedhashemi

ii
Dedication
This thesis is dedicated to all those brave Iranian men and women who lost their lives
due to chemical weapon attacks during the eight-year war (1980-1988) imposed by Iraq.

iii
Acknowledgements
I would like to express my sincere regards to my advisors Professor Theodore T.
Tsotsis and Professor Fokion Egolfopolous for giving me a chance to do my PhD under
their valuable supervision. This dissertation is a result of their help, support, and patience.
I also wish to thank Dr. Malancha Gupta for serving on my qualifying and dissertation
committees.
This Thesis could not be finished without the help of Mr. Nitin Nair, who was with me
from the beginning of my career at USC and assisted me in many aspects of my research,
Dr. Ryan Mourhatch who taught me how to use GC/MS, Dr. Megha Dadwhal who
introduced me to the FTIR and BET techniques, Ms. Xiaojie Yan who took the TEM
images, and Ms. Basabdatta Roychaudhuri who carried out the He pycnometry
measurements.
I would also like to thank all my colleagues and friends in our research group who
helped me in different ways: Mr. Aydin Jalali, Ms. Sahar Soltani, Mr. Wangxue Deng,
Ms. Malak Khojasteh, and many others.
Many thanks go to Mrs. Tina Silva, instructional lab manager, and Mr. Shokry
Bastorous, instructional lab assistant, for helping me in every way possible throughout
these past four years.
Special gratitude is extended to the administrative staff of the Mork Family
Department of Chemical Engineering and Materials Science, Ms. Karen Woo, Mr. Martin
Olekszyk, Ms. Heather Alexander, Ms. Aimee Barnard, and Ms. Angeline Fugelso for all
their help and support throughout my graduate studies.
iv
Above all, I would like to thank my parents, Jafar Motamedhashemi and Fahimeh
Moayyedbakhtiari, my brother Javad Motamedhashemi, my uncle Kayvon Baktiar and
his lovely wife Antoinette Baktiar for their patience, encouragement, and support during
my Ph.D. studies.
This research was made possible by financial support from the Defense Threat
Reduction Agency (DTRA).
Finally, technical support from Media and Process Technology, Inc. is gratefully
acknowledged.

v
Table of Contents
Dedication ii

Acknowledgements iii

List of Tables vii

List of Figures ix

Nomenclature xvii

Abstract xxi

Chapter 1: Introduction 1
1.1. Background 1
1.2. Literature Survey 4
1.2.1. Complete-conversion integral FTCMR 5
1.2.2. Selective-integral FTCMR 10
1.2.3. Selective-differential FTCMR 20
1.3. Scope of the Present Work 26
1.4. Conclusions 69

Chapter 2: Catalytic Membrane Studies 71
2.1. Introduction 71
2.2. Membrane Studies 73
2.2.1. Estimation of the structural properties 74
2.2.2. SEM analysis 75
2.2.3. Using the Archimedes method to study
membrane properties 78
2.2.4. Helium pycnometry 79
2.2.5. BET analysis 80
2.2.6. Bubble-point (Coulter) perporometry analysis 83
2.2.7. Permeation studies 90
2.3. Catalyst Characterization 98
2.3.1. SEM/EDAX analysis 99
2.3.2. TEM study 101
2.3.3. Chemisorption study 102
2.3.4. Permeation studies 105
2.4. Conclusions 107

Chapter 3: Reaction Studies 108
3.1. Introduction 108
3.2. Experimental Set-up 108
3.3. Reaction Studies 113
3.4. Theoretical Studies 121
vi
3.5. Modeling Results and Analysis 131
3.6. Conclusions 140

Chapter 4: Catalytic Membrane Deactivation via Active Site
Coverage and Pore Blockage 142
4.1. Introduction 142
4.2. Background 144
4.3. Experiments 166
4.4. Membrane Recovery 179
4.5. Theoretical Studies 183
4.5.1. Model development 183
4.5.2. Model Validation 193
4.6. Conclusions 200

Chapter 5: Hybrid System 204
5.1. Introduction 204
5.2. The Hybrid SFMS-FTCMR Concept 205
5.3. Experiments 208
5.4. Conclusions 216

Chapter 6: Suggestions for Future Work 217

References 222

Appendix 242

vii
List of Tables
Table 1.1: Prior studies in complete-conversion integral FTCMR 7

Table 1.2: Flow-through catalytic membrane reactors for selective reactions 12

Table 1.3: Flow-through catalytic membrane reactors for selective reactions
with liquid recycling 21

Table 1.4: Previous studies on the catalytic decomposition/adsorption
of DMMP 32

Table 1.5: Destructive adsorption of DMMP with [M
x
O
y
] CP/AP-CaO
at 500
o
C 50

Table 2.1: Classes of ceramic membranes 72

Table 2.2: Various M&P ceramic membranes used in this research 73

Table 2.3: Some physical properties of M&P ceramic membranes
used in this research, as reported by M&P 74

Table 2.4: Structural properties estimated by simple pore structure
model of Figure 2.2 75

Table 2.5: The results obtained by utilization of Archimedes method 75

Table 2.6: N
2
adsorption/desorption results for S-type membrane samples 83

Table 2.7: Estimation of the porosity of the top-layer based on N
2

adsorption/desorption data 83

Table 2.8: Maximum, minimum and mean pore sizes obtained by the
application of the bubble-point method for a membrane support 87

Table 2.9: Maximum, and minimum and average pore sizes obtained by
the application of bubble-point method for a 500 Å membrane 89

Table 2.10: Summary of the single-gas permeation results for the
SN-1 and SN-20 symmetric support membranes 96

Table 2.11: Summary of the single-gas permeation results for the
A-02-2010 asymmetric membrane 98

Table 2.12: The results of carbon monoxide chemisorption studies for
a 500 Å membrane 103
viii
Table 2.13: The relation between the number of unit cells in a FCC
and available atoms for reaction 104

Table 2.14: Permeation characteristics of different membrane samples
before impregnation 106

Table 2.15: Permeation characteristics of different membrane samples
after impregnation 106

Table 3.1: Dimensionless parameters used for the theoretical model 130

Table 3.2: Parameters and their ranges used for the modeling 130

Table 4.1: Nitrogen adsorption/desorption results for fresh and aged
catalytic membranes 171

Table 4.2: Dimensionless parameters used for the theoretical
unsteady-state model 191

Table 4.3: Experimental data collected in the FTCMR experiments
with the NA03 membrane 196

Table 4.4: Fitted experimental parameters 198

Table 4.5: Prior studies reporting reaction rate constants for the
destruction of key OPC compounds 201

Table 4.6: Experimental data collected from the monolith reaction
studies with the NA10G membrane 202

Table 5.1: Parameters used for the simulation studies for the effect
of feed concentration on FTCMR performance 206

Table 5.2: Specifications of the catalytic and surface-flow membranes
used for the hybrid system experiments 212

Table 6.1: Characteristics of the catalytic alumina (NA09) and
titania (T3) membranes 219

ix
List of Figures
Figure 1.1: Catalysis and mass transport in pores, example for a
consecutive reaction in a membrane and in a catalyst pellet 3

Figure 1.2: Molecular structure of Sarin (GB) and its simulant DMMP 27

Figure 1.3: The adsorption of DMMP onto Al
2
O
3
as presented by
Templeton and Weinburg (1985a) 29

Figure1.4: Proposed sequence for the continuous decomposition of
DMMP on Fe
2
O
3
34

Figure1.5: Proposed models for the decomposition reaction of
DMMP on Pd (111) and Ni (111) 36

Figure1.6: Proposed models for the decomposition reaction of
DMMP on heat-treated MgO 38

Figure 1.7: Suggested reaction scheme for the production of
P-containing intermediate during the catalytic oxidation
of DMMP over Pt/Al
2
O
3
40

Figure 1.8: Major products from the TiO
2
-catalyzed photolytic
degradation of DMMP in aqueous solutions 44

Figure 1.9: Scheme showing the DMMP reaction mechanism on AMO 45

Figure 1.10: Results from the decomposition of DMMP on
γ-alumina at 25 °C 54

Figure 1.11: (Left) Product flow curves from the decomposition of
DMMP at 25 °C on Al
2
O
3
, 5 wt.% Fe on Al
2
O
3
,
7.5 wt.% Ce on Al
2
O
3
, and 5 wt.% Fe/7.5 wt.% Ce on Al
2
O
3
.
(Right) Accumulated product formation curves from the
adsorption of DMMP at 25 °C on Al
2
O
3
, 5 wt.% Fe on
Al
2
O
3
, 7.5 wt.% Ce on Al
2
O
3
, and 5 wt.% Fe/7.5 wt.% Ce
on Al
2
O
3
54

Figure 1.12: Proposed scheme for the production of DME during
reaction of DMMP on alumina supported cerium oxide 55

Figure 1.13: Proposed scheme for the production of MeOH during
reaction of DMMP on alumina supported cerium oxide 55

x
Figure 1.14: Surface-bound product concentration in static batch
reactor with lightly loaded DMMP 58

Figure 1.15: GC peak area and concentration versus time data
for gas phase products in bulk static reactor
degradation experiments 58

Figure 1.16: Proposed reaction pathway for the photodegradation
and adsorption of DMMP on TiO
2
59

Figure 1.17: Proposed reaction pathway for catalytic-degradation
and adsorption of DMMP on TiO
2
supported
Au-nanoparticles 62

Figure 1.18: Proposed mechanism for degradation of DMMP
on MnO
x
/Al
2
O
3
68

Figure 2.1: Schematic representation of an asymmetric
composite membrane 72

Figure 2.2: Simple representation of a membrane structural block 74

Figure 2.3: SEM pictures for single separation layer membranes
showing the top layer thicknesses for different membrane
batches supplied by M&P 77

Figure 2.4: The results obtained by SEM analysis 77

Figure 2.5: SEM picture for a double separation layer membrane
along with the variation of top layer thickness along
the length of the membrane 78

Figure 2.6: Pore size distribution obtained by N
2

adsorption/desorption on/from an A-type M&P membrane 82

Figure 2.7: Pore size distribution obtained by N
2

adsorption/desorption on/from a B-type M&P membrane 82

Figure 2.8: Theoretical flow–pressure curve for the bubble point
test/progressive displacement test 85

Figure 2.9: Experimental flow–pressure curves for the bubble point
test of an M&P membrane support 87

xi
Figure 2.10: Pore size distribution for an M&P membrane support
based on the assumption of a Bell-shaped pattern and
exponential fitting 88

Figure 2.11: Pore size distribution for an M&P membrane support
based on the assumption of Bell-shaped pattern and
polynomial fitting 88

Figure 2.12: Experimental flow–pressure curves for the bubble
point test of a 500 Å membrane 89

Figure 2.13: Pore size distribution for a 500 Å membrane assuming
a Bell-shaped pattern and using polynomial fitting 90

Figure 2.14: Ar permeance data for SN-1 membrane 94

Figure 2.15: He permeance data for SN-1 membrane 94

Figure 2.16: Ar permeance data for SN-20 membrane 95

Figure 2.17: He permeance data for SN-20 membrane 95

Figure 2.18: Ar permeance data for A-02-2010 membrane 97

Figure 2.19: He permeance data for A-02-2010 membrane 97

Figure.2.20: SEM picture of a single-layer membrane after impregnation 100

Figure 2.21: Pt distribution along the 1 mm radial direction of
a single-layer membrane 100

Figure 2.22: TEM micrographs of a single-layer Pt/Al
2
O
3
catalytic
membrane after calcination at 623 K 101

Figure 2.23: Simple representation of a face-centered cubic
crystal structure 104

Figure 3.1: Experimental set-up used for the catalytic
destruction of DMMP 110

Figure 3.2: Schematic of the stainless-steel reactor 113

Figure 3.3: DMMP conversion in the stainless-steel reactor
under different temperatures 114

xii
Figure 3.4: Destruction at different temperatures under low
DMMP load (300 ppm) using unimpregnated
single-layer membranes 115

Figure 3.5: Destruction at different temperatures under
high DMMP load (1000 ppm) using unimpregnated
single-layer membranes 116

Figure 3.6: The effect of reaction temperature on DMMP
destruction for impregnated single-layer membranes
(DMMP load: 300 ppm) 116

Figure 3.7: The effect of feed concentration on DMMP destruction
for impregnated single-layer membranes
(reactor temperature of 573 K) 118

Figure 3.8: The effect of membrane pore size on DMMP destruction
(reactor temperature of 573 K, DMMP load: 300 ppm) 118

Figure 3.9: DMMP destruction in plug-flow mode of operation
(single-layer membrane, DMMP load: 300 ppm) 119

Figure 3.10: DMMP destruction in plug-flow mode of operation
(single-layer membrane, DMMP load: 1000 ppm) 120

Figure 3.11a: DMMP conversion for the FTCMR and the monolith
reactor (50 Å, α=0.006) 132

Figure 3.11b: DMMP conversion for the FTCMR and the monolith
reactor (500 Å, α=0.5) 132

Figure 3.12a: FTCMR superiority region - effect of thickness
(50 Å, P
ratio
= 1.03) 134

Figure 3.12b: FTCMR superiority region - effect of thickness
(50 Å, P
ratio
= 1.14) 135

Figure 3.12c: FTCMR superiority region - effect of thickness
(50 Å, P
ratio
= 1.68) 135

Figure 3.13: FTCMR superiority region-effect of P
ratio
136

Figure 3.14: FTCMR superiority region-effect of pore diameter 138

Figure 3.15: Performance regions for the FTCMR and the
monolith reactor (50 Å, α=0.006) 138
xiii
Figure 3.16: FTCMR superiority region-effect of support thickness 139

Figure 3.17: FTCMR vs. monolith reactor performance 139

Figure 4.1: Permeance decrease during the DMMP destruction
at 673 K (DMMP load: 300 ppm) 144

Figure 4.2: Poisoning deactivation curves 145

Figure 4.3: Phosphorous concentration profiles along the radius
of a poisoned catalyst 149

Figure 4.4: Decrease of the BET-surface for different catalysts
with increasing poison content 152

Figure 4.5: Interaction of phosphoric acid with alumina
surface hydroxide groups 154

Figure 4.6: Acidic nature of alumina 156

Figure 4.7: Conjugated acidic nature of γ-alumina 156

Figure 4.8: Possible reactions between γ-alumina and phosphoric acid 157

Figure 4.9: Possible single bond formation during the reaction of
γ-alumina and phosphoric acid 158

Figure 4.10: Surface area (A) and pore volume (B) variation 159

Figure 4.11: Proposed mechanism of phosphorous adsorption
on alumina 160

Figure 4.12: Proposed mechanism for protonation of alumina
hydroxyls by phosphoric acid 162

Figure 4.13: The effect of membrane pore size on the rate of
pore blockage manifested by the reactor inlet pressure
build-up (reactor temperature: 573 K, DMMP load: 300 ppm,
feed flow: 1 sccs) 167

Figure 4.14: Reactor inlet pressure change during the DMMP oxidation
with different Pt-loaded FTCMRs (reactor temperature:
573 K, DMMP load: 300 ppm, feed flow: 1 sccs) 169

xiv
Figure 4.15: DMMP conversion during the reaction with different
Pt-loaded FTCMRs (reactor temperature: 573 K,
DMMP load: 300 ppm, feed flow: 1 sccs) 169

Figure 4.16: Reduction of membrane pore size due to the plugging
of the pores by phosphate poisons
(NB22, FTCMR operation) 172

Figure 4.17: Reduction of membrane pore size due to the plugging
of the pores by phosphate poisons
(NB26, monolith operation) 172

Figure 4.18: SEM images of a support membrane before the reaction
(left) and after the pore blockage (right) 173

Figure 4.19: Elemental profile for platinum across the aged
catalytic membranes 174

Figure 4.20: Elemental profile for phosphorous across the aged
catalytic membranes 175

Figure 4.21: Elemental profile for carbon across the aged NB26 membranes 175

Figure 4.22a: FTIR spectra for an α-alumina membrane before and after
exposure to phosphoric acid (wave number < 1000 cm
-1
) 177

Figure 4.22b: FTIR spectra for an α-alumina membrane before
and after exposure to phosphoric acid
(wave number > 1000 cm
-1
) 177

Figure 4.23a: FTIR spectra for an α-alumina catalytic membrane
after exposure to air/DMMP (wave number < 1000 cm
-1
) 178

Figure 4.23b: FTIR spectra for an α-alumina catalytic membrane
after exposure to air/DMMP (wave number > 1000 cm
-1
) 178

Figure 4.24: Change of the permeance for membrane support before
and after reaction with H
3
PO
4
180

Figure 4.25: Change of the permeance for an α-alumina membrane
before and after reaction with H
3
PO
4
180

Figure 4.26: Change of the permeance for a γ-alumina membrane before
and after reaction with H
3
PO
4
181

xv
Figure 4.27: The effect of acid washing on inlet pressure of a
500 Å membrane 182

Figure 4.28: The effect of acid washing on catalytic conversion
of a 500 Å membrane 183

Figure 4.29: FTCMR experiments with the NA03 membrane 197

Figure 4.30: Change of the membrane permeance vs. average
pressure during the time on stream experiments
for the NA03 membrane 197

Figure 4.31: Goodness of fit plot for conversion values 198

Figure 4.32: Goodness of fit plot for stage-cut values 198

Figure 4.33: Rapid deactivation of the NA10G catalytic membrane
compared with the model predictions (T=573 K) 202

Figure 4.34: Comparison between the model predictions and
experimentally collected values for the DMMP conversion 203

Figure 5.1: Effect of feed concentration on the protection time
predicted by the mathematical model for a typical FTCMR
( α = 0.05, Pé
ref
= 0.35, Φ
ref
= 1) 207

Figure 5.2: Effect of the feed concentration on the permeance predicted
by the mathematical model for a typical FTCMR
( α = 0.05, Pé
ref
= 0.35, Φ
ref
= 1.3) 207

Figure 5.3: Effect of the imposed pressure gradient across MP13 CMS
membrane on its DMMP removal rate and stage-cut
(T=373 K, DMMP concentration = 825 ppm, sweep ratio = 1.2) 210

Figure 5.4: Effect of the sweep gas ratio on DMMP removal rate for
MP13 CMS membrane (DMMP concentration = 625 ppm) 210

Figure 5.5: Hybrid system performance compared with the FTCMR
by itself for the high-activity, B-type catalytic membrane 213

Figure 5.6: The positive effect of the hybrid system in deferring
catalytic membrane pore blockage compared with the FTCMR
alone for the high-activity B-type catalytic membranes 214

Figure 5.7: Hybrid system performance compared with the single FTCMR
for the low-activity B-type catalytic membranes 214
xvi
Figure 5.8: The positive effect of the hybrid system in deferring
catalytic membrane pore blockage compared with the
FTCMR alone for the high-activity A-type catalytic membranes 215

Figure 6.1: Comparison between the protection times offered by the
Pt/Al
2
O
3
and Pt/TiO
2
catalytic membranes 219

Figure 6.2: Comparison between the appearances of different TiO
2
/Al
2
O
3

membranes (From right to left: fresh bare TiO
2
/Al
2
O
3
,
aged TiO
2
/Al
2
O
3
, and aged Pt/TiO
2
/Al
2
O
3
) 219

Figure 6.3: Comparison between the conventional method (left) and
hypothetical membrane fabricated based on FTCMR
concept (right) 221

xvii
Nomenclature
a Parameter in Antoine equation (Equation 3.1)
b Parameter in Antoine equation (Equation 3.1);
Dimensionless pore plugging factor defined in Table 4.2
b
/
Pore plugging factor, (m
3
mol
-1
)
B
e
Membrane geometrical parameter defined in Equation 4.5, (m
2
)
B
0

Viscous flow term defined in Equation 2.7
d
p
Pore or particle diameter, (m)
D Diameter of structural building blocks in Equation 2.1, (m)
D
D
Effective diffusion coefficient defined in Equation 4.3, (m
2
s
-1
)
D
ij
Binary molecular diffusion coefficient, (m
2
s
-1
)
D
i,K
Knudsen diffusion coefficient for component i, (m
2
s
-1
)
E Dimensionless porosity defined in Table 4.2;
Activation energy in Arrhenius equation (J mole
-1
)
J Liquid flux in Equation 2.3, (m s
-1
);
Permeance in Equations 2.6 and 2.7, (mol m
-2
s
-1
Pa
-1
)
k Reaction rate constant in Arrhenius equation, (s
-1
)
k
0
Pre-exponential factor in Arrhenius equation, (s
-1
)
k
'
Reaction rate constant based on the reactor volume, (s
-1
)
k" Reaction rate constant, (m s
-1
)
K
0
Knudsen flow parameter defined in Equation 2.8, (m)
Kn Knudsen number
L Length, (m)
xviii
M Molecular mass, (kg kmol
-1
)
n Molar flow rate, (mol s
-1
)
n
AB
Number of pores having diameters ranging between d
p,A
and d
p,B
in
Equation 2.3
N Avogadro's number (Equation 2.11);
Flux, (mol s
-1
m
-2
)
Pe Modified Péclet number defined in Tables 3.1 and 4.2
P Total pressure, (Pa)
p Partial pressure, (Pa)
r Radius/radial distance, (m)
Rxn Reaction rate
R Universal gas constant, (8.314 J gmol
-1
K
-1
)
s Dimensionless deactivation factor defined in Table 4.2
s' Surface area deactivation factor, (m
2
mol
-1
)
S Specific surface area, (m
2
g
-1
)
S
/
g
Catalyst specific surface area, (m
2
m
-3
)
t Time, (s)
T Temperature, (K)
V Volume, (m
3
)
Vp Vapor pressure, (Pa)
w Dimensionless radial distance defined in Table 4.2
X Dimensionless partial pressure defined in Table 4.2
y Mole fraction
xix
Y Dimensionless molar flow defined in Table 4.2
z Longitudinal distance, (m)
Z Dimensionless longitudinal coordinate

Greek letters
λ Mean free path of gas molecules, (m)
 density, (kg m
-3
)
ε Membrane porosity
ψ Function defined in Equations 3.14 and 4.26
ϕ Function defined in Equations 3.15 and 4.27
Φ Thiele module defined in Tables 3.1 and 4.2
α Dimensionless parameter defined in Table 4.2;
Parameter defined in Equation 2.6
β Parameter defined in Equation 2.6
φ Function defined in Tables 3.1 and 4.2
σ Dimensionless parameter defined in Tables 3.1 and 4.2;
Surface tension, (kg m
-2
);
Effective area occupied by the molecule in Equation 2.11
θ Dimensionless time defined in Table 4.2
Ω Dimensionless parameter defined in Table 4.2
γ Dimensionless parameter defined in Tables 3.1 and 4.2
τ Membrane tortuosity
ξ Dimensionless parameter defined in Tables 3.1 and 4.2
xx
δ Dimensionless parameter defined in Tables 3.1 and 4.2
μ Gas viscosity, (Pa.sec)

Subscripts
av Average
A Air
D DMMP
i Component i
in Inner
out Outer
Pt Platinum
ref Reference

Superscripts
e Effective
j The correlating parameter in Equation 4.21
0 Inlet/initial conditions

xxi
Abstract
The possibility of the use of chemical weapons has been increased in recent years both
as a result of potential terrorist attacks and of ongoing international conflicts. The focus
of our research is the development of a novel, hybrid catalytic membrane reactor system,
which consists of a flow-through catalytic membrane reactor (FTCMR) integrated with a
surface-flow based membrane separator (SFMS). The combined system has been shown
to achieve complete oxidation of chemical warfare agents (CWA) at trace levels, and is
appropriate for use in integrated individual protection (IP) systems as well as collective
protection (CP) systems for civil and military applications.
As a part of this research, a catalytic tubular alumina membrane is prepared via
impregnation with chloroplatinic acid solutions, and is utilized in a FTCMR for the
catalytic oxidation of dimethyl methylphosphonate (DMMP) in air. DMMP is known as a
chemical precursor for the more toxic gas Sarin (GB), and has been widely used to
simulate its characteristics.
In this Thesis experiments are reported for different DMMP feed concentrations (150-
1000 ppm) and reactor temperatures (373-573K), which demonstrate the potential
advantage of the FTCMR in the complete catalytic oxidation of this important CWA
simulant. Complete destruction of low and high concentrations of DMMP was achieved
at lower temperatures compared to the values obtained in this study for a wall-coated
plug-flow (monolith) reactor containing the same amount of catalytic metal.
A mathematical model has also been developed in order to provide a better
understanding of the fundamental transport phenomena underpinning the FTCMR
operation. It makes use of the Dusty Gas Model (DGM), which incorporates in an
xxii
appropriate fashion continuum and Knudsen diffusion, and viscous flow as the
mechanisms for gas transport through the porous membrane being utilized in the
FTCMR. In the first application, the model is used for identifying the advantages of the
FTCMR concept compared to the wall-coated catalytic monolith, and also for
investigating some of the limitations, which may exist in applying this concept for the
complete oxidation of chemical warfare simulants. The results of the model support the
superiority of the FTCMR concept over the more conventional (plug-flow) monolith
reactor.
During the FTCMR experiments it was found that one of the challenges associated
with the catalytic destruction of DMMP is catalytic membrane deactivation via active site
coverage and pore-blockage. Therefore, in the second application, the model is extended
to incorporate the effect of catalyst deactivation and pore-blockage, on the performance
of the FTCMR. The model is successfully applied to determine from experimental data
important parameters such as the reaction rate constants and the poisoning and pore
plugging factors. The simulation results also confirm the experimental observations in
that the protection time provided by the FTCMR is a function of the DMMP
concentration in the feed, pointing out that an appropriate role for the FTCMR to play is
as a second stage in a hybrid system, following a bulk-toxin removal unit in the first
stage. The main advantage of the proposed hybrid system, combining a surface-flow
membrane (SFM) separation unit (which is capable of continuously physically removing
a large portion of the CWA from contaminated air streams) with the FTCMR, is that it
completely destructs the CWA that remains with a lower rate of pore blockage, thus
xxiii
resulting in the continuous CWA destruction for extended time periods, which are
appropriate for both IP and CP applications.
1
Chapter 1: Introduction

1.1. Background
The increasing number of publications dealing with the concept of catalytic membrane
reactors in the last two decades underlines the potential of such reactors for applications
for a broad range of both gas and liquid phase reactions. Conventional catalytic
membrane reactors vary from other types of reactors in that they combine chemical
reactions and separation in a single unit and operation, whereby the membrane removes
one or more of the product species. As a result, for equilibrium-limited reactions
conversion can be increased beyond the equilibrium value [Dixon, 2003]. In other
instances, the membrane acts as an interface and as a means to deliver a reactant in a
controlled way. These reactors are known as catalytic membrane contactors.
A third class of membrane reactors are the so-called forced-flow (or flow-through)
catalytic membrane reactors (FTCMR) in which a porous membrane, either inherently
catalytic or rendered catalytic by impregnation with catalysts such as noble metals, is
applied in a dead-end configuration, thereby forcing all the reactants to flow through it.
The function of the membrane in this mode of operation is to provide a reaction
environment with short and controlled residence times and high catalytic activity
[Westermann and Melin, 2009]. The basic premise here is that, while in classical fixed-
bed reactors the desired conversion is limited by the pore diffusion, in FTCMR the
reactants flow convectively through the membrane pores where the catalyst is located,
thus resulting in intensive contact between reactants and the catalytic sites and in high
catalytic activity with negligible mass transport resistances.
2
The number of collisions between the reactants and catalytic sites inside the pores can
be noticeably amplified by decreasing the pore diameters to such small values so that
Knudsen diffusion prevails. The combination of Knudsen diffusion (KF) with forced-
flow catalytic membranes results in a unique and fundamental feature: All the reactant
molecules are guaranteed to have multiple contacts with the catalyst surface. A further
improvement can be achieved by the application of nanocatalysts in combination with the
catalytic membrane operating under Knudsen diffusion control. Nanocatalysts are highly
active, since due to the high (surface/volume) ratio of the particles most of the metal is
available for reaction. This synergy of combining nanocatalysis with operating under KF
control maximizes the benefits offered by the nanocatalysts, while overcoming their
intrinsic limitations.
Figure 1.1 demonstrates schematically a key application of the FTCMR concept, namely
to a consecutive, heterogeneously-catalyzed reaction whereby the reactants A and B react
to produce the desired intermediate product C, which unfortunately further reacts to
produce an undesired product D. Obviously, for optimal performance the consecutive
reaction to the product D should be inhibited. In the fixed-bed reactor the selectivity is
limited by mass transport in and out of the catalyst pores. This process is rather slow and
the large residence time of A, B and C in contact with the catalytically active sites
promotes the formation of D. In the FTCMR the ability to impose a pressure gradient
across the membrane significantly increases the transport rate through the pores and as a
result reduces the contact time of the intermediate product C with the catalytic sites. As a
consequence, significantly better selectivities for the intermediate product are obtained
[Schmidt et al., 2005].
3

Figure 1.1: Catalysis and mass transport in pores, example for a consecutive reaction in a membrane
and in a catalyst pellet [Schmidt et al., 2005]

In general, the potential advantages of FTCMR are as follows:
- Minimization of the intraparticle mass transfer resistance: This is due to the combined
effect of higher diffusive flow and effective multiple molecule-catalyst site collisions.
- Better intraparticle transport: In conventional catalysts, one cannot independently
adjust the pressure gradient along the pore length, thus resulting in slower transport. On the
other hand, in FTCMR, as noted in the example of Figure 1.1, by varying the trans-
membrane pressure one can control the flux through the membrane, and as a consequence
the contact time of the reactant and products with the catalyst; this, in turn, often allows for
an effective control of the reaction selectivity.
- Higher catalytic reaction rates: Since the molecules collide mostly with the catalyst
rather than with each other, gas phase reactions are avoided, and most of reactant molecules
are ultimately brought in contact with the catalyst sites.
- Operation under lower temperatures: The higher reaction rates may permit operation
under lower temperatures and pressures. Milder operating temperatures, in turn, minimize
4
nanocatalyst sintering, a limitation in the use of such materials in conventional reactors.
Another potential advantage of lower operating temperatures is that they may permit the
use of low-cost catalysts (e.g., CrO
x
), which are unsuitable for the conventional reactors,
because of the higher temperatures required, as opposed to the conventional more
expensive noble metal (e.g., Pt) catalysts. However, even when these expensive catalysts
are used, using the FTCMR may decrease the amount of catalyst required for a given
conversion.
- No by-pass or external mass transport limitations: Operating in the “dead-end flow-
through” mode minimizes by-pass, often encountered with short catalytic packed-beds, and
also eliminates external mass transport limitations, that typically plague granular
adsorbents.
- Better catalyst utilization: Because mass transfer limitations are eliminated, the
intrinsic catalyst reactivity is better exploited compared with conventional reactors in
which higher catalytic activity (or loading) does not necessarily translate into higher rates.
- Catalyst poisoning is minimized: The reduced mass transfer resistance allows certain
products, which can potentially inhibit catalytic activity, to quickly transport out of the
catalyst pores.

1.2. Literature Survey
In the technical literature, so far, the incentive for applying a FTCMR is either (i) to
reach complete conversion in a minimum of time or using the minimum of catalyst
weight or reactor volume, by taking advantage of the high catalytic efficiency of such
reactors; or (ii) to reach maximum selectivity for a given reaction as a result of the
5
narrow distribution of contact times, typically characterizing such reactors. Westermann
and Melin (2009) have classified FTCMR into the following three groups:
• Complete conversion integral FTCMR which are designed to reach complete
conversion in a single pass. They are mostly applied to gas phase reactions, with a few
exceptions of applications with multi-phase reactions.
• Selective integral FTCMR which try to achieve optimum selectivity in a single pass
through the catalytic membrane. They are also preferably used for fast gas phase
reactions, mainly partial hydrogenations and oxidations.
• Selective differential FTCMR which are operated in a semi-batch mode in a loop
with a membrane module and a saturation tank, causing only small conversions per pass,
with the aim of reaching high selectivities. They have been exclusively investigated for
partial hydrogenations in the liquid phase, which are limited by the low solubility of
hydrogen.

1.2.1. Complete-conversion integral FTCMR
In this mode of operation the premixed reactants flow through the catalytic membrane
in a single pass with the aim of reaching complete conversion, taking advantage of the
high catalytic efficiency caused by the intensive contact between the reactants and the
catalyst. Most applications of such FTCMR involve the decomposition of volatile organic
compounds (VOC), including photocatalytic oxidations. An overview of some of the key
studies, so far, is given in Table 1.1.
The first known application of such a FTCMR was reported by Yamada et al. (1988).
They prepared catalytic alumina membranes by the anodic oxidation of aluminum plates,
6
and tested them for isomerization of 1-butene as a model reaction. The FTCMR showed
under some conditions higher catalytic activity compared to the reactor packed with the
powder of the same anodized alumina film. Splinter et al. (2002) designed and tested a
micro-membrane flow-through reactor which converted CO into CO
2
as part of a gas
micro-analyzer. Their FTCMR utilized a porous silicon membrane whose average pore
diameter could be adjusted in the range of (4 to 8 µm). The membranes were rendered
catalytic by immersing them into Pd (acac)
2
solution in toluene. Though the membrane
geometric area was smaller than a square millimeter, the active surface area ranged
between 8.6 cm
2
and 39.4 cm
2
. The gas molecule residence times through the membrane
depended on its thickness, and were reported to be ~1 ms. By adjusting the active
membrane area and thickness, nearly 100 % conversions could be attained.
For potential application in flue-gas clean-up, Saracco and Specchia (1995) and Saracco
and Montanaro (1995) prepared a catalytically active membrane infiltrating the pores of
sintered α-Al
2
O
3
with γ-Al
2
O
3
. This so called “filter” could mechanically remove the
particles, and because of the acidic properties of γ-Al
2
O
3
was also capable of
decomposing chemical pollutants such as NO
x
and VOC that were flown through it. The
performance of the filters was assessed by carrying out the dehydration of isopropanol
which was catalytically promoted by the γ-alumina. The authors reported that nearly
complete conversion could be achieved for superficial velocities of industrial interest.

7
Table 1.1: Prior studies in complete-conversion integral FTCMR
Authors Reaction Membrane Catalyst Pore diameter, (nm)
Yamada et al., (1988)
1-Butene
isomerization
Anodized
alumina
None ~ 100
Splinter et al., (2002)
CO
oxidation
Porous silicon Pd 4000 - 8000
Saracco and Specchia, (1995)
VOC
decomposition
Porous ceramic γ-Al
2
O
3
1000 -15000
Pina et al., (1996)
VOC
decomposition
γ-Al
2
O
3
Pt 3.5
Zalamea et al., (1999)
VOC
decomposition
γ-Al
2
O
3
Pt 2000 - 10000
Maira et al., (2003)
VOC
decomposition
Zeolite TiO
2
~1
Tsuru et al., (2003)
VOC
decomposition
α-Al
2
O
3
TiO
2
6.5
Picasso et al., (2003)
VOC
decomposition
Alumina/
Stainless-steel
Fe
2
O
3

5 (Alumina)/
500 (Stainless-steel)
Chea et al., (2012)
Phenolic compounds
decomposition
α-Al
2
O
3
Laccase 200 -1400

Pina et al. (1996) utilized commercial microfiltration (MF) membranes coated with γ-
Al
2
O
3
separation layers on the top of the α-Al
2
O
3
support to prepare different FTCMRs.
Depending on the amount of γ-Al
2
O
3
the resulting membranes had different Knudsen
diffusion contributions, from a mixed Knudsen diffusion/Poiseuille flow regime to
completely Knudsen diffusion dominated regime. They investigated the performance of a
sample for which Knudsen diffusion was completely dominant for the complete
combustion of VOC. A VOC-containing air stream (100 to 5100 ppm toluene) was forced
to permeate through the Pt/Al
2
O
3
catalytic top layer operating in the Knudsen diffusion
regime leading in intimate contact of VOC molecules with the combustion sites, resulting
in highly efficient use of the catalyst and complete conversions for fairly low
temperatures. Gas flow rates in the range of 0.4 to 1.0 L/min required pressure drops of
0.18 to 0.35 bar. While Pina et al. (1996) studied VOC decomposition with a FTCMR
8
operating in the Knudsen regime, Zalamea et al. (1999) studied such reactors operating in
the mixed permeation regime, where both Knudsen and Poiseuille diffusions contributed
to the total diffusion, for the combustion of methyl ethyl ketone (MEK) and n-hexane.
The temperatures required for the complete conversion in the mixed-regime FTCMR
with higher Poiseuille flow contribution were compared with the performance of FTCMR
using asymmetric membranes operating in the Knudsen dominated regime. Zalamea et al.
(1999) reported that the temperatures required for complete conversion in the latter
FTCMR were lower at the expense, however, of higher pressure drops.
Picasso et al. (2003) also investigated a FTCMR operating in the Knudsen-flow
regime, using Fe
2
O
3
impregnated ceramic and stainless-steel (SS) membranes, for the
total combustion of MEK (used as a model of volatile organic compounds (VOCs)) in
diluted streams. Essentially, complete combustion of MEK for concentrations in the
range of 500–2000 ppm
V
was achieved at temperatures as low as 255
o
C (for the ceramic
membrane) and 300
o
C for SS membranes. The authors suggested that the higher activity
of the ceramic compared with the SS membrane was a consequence of (i) their higher
Knudsen diffusion contribution to transport, and (ii) the better distribution of the active
phase along the thin layer of the membrane.
Maira et al. (2003) evaluated the performance of a silicalite-1 membrane coated with
nanostructured anatase TiO
2
for the gas-phase photocatalytic oxidation of
trichloroethylene. The catalytic membrane outperformed a catalytic SS plate coated with
a similar amount of the nanostructured TiO
2
catalyst. While both the SS plate and the
membranes showed catalytic activity in the parallel flow mode, the conversion was
reported to significantly increase by switching into the flow-through mode or into a
9
mixed-flow mode. With respect to the selectivity (defined as the ratio of the fraction of
VOC that was completely mineralized to the fraction that was converted during the
heterogeneous photocatalytic oxidation), the membrane outperformed the porous plate.
This was explained potentially from the fact that during the reaction the membrane acted
as a sieve allowing only the smaller product molecules resulting from total mineralization
to pass through the membrane while retaining the larger pollutant molecules for further
conversion.
The gas-phase oxidation of methanol, as a model VOC, was studied by Tsuru et al.
(2003) in a photocatalytic membrane reactor. They utilized cylindrical α-alumina MF
membranes on which they deposited mesoporous TiO
2
membrane layers which were
active towards the photocatalytic decomposition of methanol. An air stream containing
1000 ppm of MeOH was fed into the shell side of the membrane, which was irradiated
with black-light lamps to catalyze the generation of OH radicals on the TiO
2
surface.
Two different flow modes were studied: In the first flow mode without permeation, the
product stream was also taken from the shell side, allowing the reactants to contact the
catalytic surface only by diffusion. In the second flow mode with permeation, the
reactants left the reactor on the tube side of the membrane, after being forced to flow
convectively through the membrane pores, thus providing for a more efficient contact
between the reactants and the catalytically active TiO
2
surface. A higher decomposition
rate was observed for the reactor mode with membrane permeation, which was attributed
to two factors: the enhanced transport to the catalyst surface by forced convection, in
addition to the transport by diffusion, as well as to the utilization of a larger catalytic
surface area.
10
In a recent publication, Chea et al. (2012) described how enzymatic membrane
reactors (EMR) were optimized and characterized to improve enzyme activity and
substrate conversion. They studied Al
2
O
3
membranes impregnated with the enzyme
laccase working in the flow-through mode of operation for the degradation of two
phenolic compounds, namely 2,6-dimethyloxyphenol (DMP) and guiaicol. Their results
showed that the optimized enzymatic membrane was very effective for removing DMP
from aqueous model solution. However, fouling and pore plugging was observed during
long-term experiments due to the formation of polymers resulting from the degradation of
DMP. In order to increase the mass transfer and, therefore, limit the degree of polymer
accumulation, the feed flux was increased, but this led to an increase in pressure which
induced structural changes in the gelatin layer. Based on these results, they concluded
that EMR in the flow-through configuration is not suited to the degradation of phenolic
compounds.

1.2.2. Selective-integral FTCMR
This mode of operation is similar to the complete-conversion integral FTCMR
described in the previous section, from the standpoint that during operation premixed
reactants flow through the catalytic membrane in a single pass. However, what
differentiates this class of FTCMR from the previous one is that they involve applications
of sequential reactions, whereby in addition to high catalytic efficiency one also seeks to
attain high selectivity; this results from the narrow residence time distributions, inherent
to FTCMR, which help to limit sequential reactions towards the undesired by-products.
For such FTCMR to be successful, high-quality, pin-hole free membranes are necessary
11
in order to prevent maldistribution effects, which are common for fixed-bed as well as
monolithic reactors (Kreutzer et al., 2005). The good performance in such FTCMR with
sequential reactions, as manifested by high selectivity towards the desired intermediate
products, is commonly considered as indirect proof of the narrow residence time
distribution attained with such systems as compared to the competing reactor concepts.
The selective-integral FTCMR has been applied to a number of sequential reactions
(mostly in the gas-phase, but in a few instances for liquid-phase reactions as well), such
as partial oxidation (including direct methane coupling), partial hydrogenations, and
oligomerizations. Some of the key studies are summarized in Table 1.2, which indicates
the reaction studied and the membranes and catalysts.
The first application of the selective-integral FTCMR concept dates back to the early
1990’s, when Zaspalis et al. (1991) used non-separative, catalytically active alumina
membranes for the dehydrogenation of methanol to formaldehyde. The flow-through
operation mode led to higher conversions but lower selectivities to formaldehyde due to
the comparatively high residence time of methanol in the catalytically active layer. The
observed activity per gram of γ-alumina catalyst was reported to be up to ten times higher
than what is observed in a tubular reactor with a catalytic bed of the same material. This
was attributed by the authors to the higher effective surface accessible to the reactants per
gram of catalytic material in the membrane than in the powder form of the catalyst in the
fixed-bed reactor.

12
Table 1.2: Flow-through catalytic membrane reactors for selective reactions
Authors Reaction Membrane Catalyst Pore diameter, (nm)
Zaspalis et al., (1991)
MeOH
dehyrogenation
α-Al
2
O
3
γ- Al
2
O
3
3-4
Kobayashi et al., (2003)
Propane
epoxidation
Macroporous
glass
Cs-Ag, Re-Ag,
Ag
2
O
220-480
Zhu et al., (2003)
Propane
oxidation
α-Al
2
O
3
AgBiVMo oxide 2-20
Pellin et al., (2005)
Cyclohexane
oxidation
Anodized
Al
2
O
3

V
2
O
5

(ALD)
10-38
Mucherie et al., (2007)
Propane
oxidation
Anodized
Al
2
O
3

V
2
O
5

(ALD)
40
Alfonso et al., (2001)
Butane
dehyrogenation
α-Al
2
O
3
V/MgO 200-12000
Schuessler et al., (2001)
MeOH
reforming
Cu-matrix Cu/ZnO ~10000
Lange et al., (1998)
Hexadiene
hydrogenation
α-Al
2
O
3

Pt
(Containing TiO
2
)/
γ- Al
2
O
3

5
Lambert and Gonzalez,
(1999)
Acetylene
/Butadiene
hydrogenation
Asymmetric
γ- Al
2
O
3

Pd/ γ- Al
2
O
3
5
Vincent and Gonzalez,
(2002)
Acetylene
hydrogenation
Asymmetric
γ- Al
2
O
3

Pd/ γ- Al
2
O
3
3.6
Groschel et al., (2005)
Propyne
hydrogenation
Porous
polymer
Pd particles 3600
Westermann and Melin,
(2007)
Acetylene
hydrogenation
Anodic
Al
2
O
3

Pd wet
impregnation
200
Ma et al., (1998)
Methane
oxidation coupling
Asymmetric
α - Al
2
O
3

Sm
2
O
3
20-5000
Torres et al., (2003)
Isobutene
dimerization
γ- Al
2
O
3
on

α - Al
2
O
3

Zeolite layer 1.4 – 5.0
Fritsch et al., (2004)
Isobutene
dimerization
Solid acid on
Porous polymer
Solid acid
copolymer
410 - 610
Khassin et al., (2005)
Fischer-Tropsch
synthesis
Composite
porous metal
Co-Al
co-precipitation
500 - 3000
Minyukova et al., (2012)
Triglyceride
hydrogenation
Composite
porous metal
CuZn/CuCr N/A

The heterogeneously catalyzed selective epoxidation of propene in a FTCMR was
studied by Kobayashi et al. (2003). Three different catalysts (Cs-Ag, Re-Ag, and Ag
2
O)
were immobilized in the pores of a macroporous glass membrane and were used in three
different reactor systems: A convection-flow reactor, a diffusion flow reactor, and a
13
packed-bed-flow reactor (using the crashed membrane as a catalyst). Comparing the three
immobilized catalysts, the selectivity towards propylene oxide increased with the amount
of intermediate formed on the catalyst, which strongly depended on the types of catalyst
and the reactor system. The highest selectivity towards propylene oxide was achieved with
the convection-flow reactor and the Re-Ag catalyst, demonstrating that convective flow
through the membrane pores effectively enhanced the selectivity towards propylene oxide.
Zhu et al. (2003) studied a FTCMR for the selective oxidation of propane to acrolein.
In their study they used a tubular mesoporous mixed oxide (Ag
0.01
Bi
0.85
V
0.54
Mo
0.45
O
4
)
catalytic membrane prepared by means of a modified sol-gel-method and α-Al
2
O
3
as the
support. The FTCMR, operated with reactants flowing from shell to tube side, and its
performance was compared with that of a fixed-bed reactor (FBR). For the FTCMR the
catalyst loading was 50 to 100 mg, whereas the FBR was loaded with 2 g of catalyst. In
the FTCMR selectivities towards acrolein higher than 50 % were attained, compared to
less than 10 % for the FBR, which produced mainly methanol and acetic acid as a liquid
phase product.
Stair and coworkers [Pellin et al., 2005; Stair et al., 2006] synthesized mesoporous
inorganic catalytic membranes by a combination of anodic oxidation and atomic layer
deposition (ALD). They reported that their technique had allowed them to control the
pore diameter and wall composition along the length of the pore. The pores of anodic
alumina membranes (thickness 70 μm and pore diameter of 40 nm) were coated by ALD
with a thin alumina layer of 1 nm or 15 nm respectively, resulting in membranes with
pore diameters of either 38 or 10 nm. The catalytic performance of these membranes was
compared to that of the original alumina membrane, after coating all membranes with one
14
monolayer of vanadium oxide (V
2
O
5
). Additionally, the membranes were compared to a
conventional high surface area alumina powder catalyst also impregnated with V
2
O
5
. As
a test reaction, they studied the oxidative dehydrogenation of cyclohexane, which at 450
o
C yielded higher selectivities towards the partial dehydrogenation product cyclohexene
for all three catalytic membranes compared to the catalytic powder. The authors
attributed the enhanced selectivity to a decrease in the contact time (1000 - 10000 times)
when compared to the conventional catalyst bed. For higher temperatures, the
improvement in selectivity decreased; however, in all studied cases the membranes
outperformed the supported alumina catalysts. While the conversion of O
2
and C
6
H
12
was
lower for the 10 nm membrane than for those with larger pores, the conversion per mole
of vanadate was significantly higher.
The same approach was also applied by the same group to the oxidative
dehydrogenation of propane [Mucherie et al., 2007]. They used vanadium oxide catalyst
supported on anodic aluminum oxides which were ALD-coated with three different metal
oxides (Al
2
O
3
, Nb
2
O
5
, and TiO
2
). Highest reactivity was observed for TiO
2
, which was
attributed to its highest dispersion, whereas the highest selectivity was observed for the
Al
2
O
3
based membrane.
Alfonso et al. (2001) proposed an interesting FTCMR which consisted of two catalytic
porous layers: An inside layer containing a V/MgO catalyst, which was active towards
the oxidative dehydrogenation of butane and an outside layer containing a Pt/Sn/Al
2
O
3

catalyst prepared by a sol-gel technique, which was thought to be active towards the
dehydrogenation of butene. The presence of both layers allowed them to couple the two
reactions both energetically, but also chemically since the product of the oxidative
15
dehydrogenation reaction was the reactant for the dehydrogenation reaction. An
additional beneficial effect of this reaction coupling appeared to be that the steam and
CO
2
formed in the oxidative dehydrogenation seemed to help the dehydrogenation
catalyst maintain its catalytic activity. The authors reported that the presence of the
Pt/Sn/Al
2
O
3
layer had increased the butane conversion by 2-13 %.
Schuessler et al. (2001) studied the autothermal reforming of methanol for mobile
applications through a stack of thin porous catalytic disks, prepared by mixing Cu/ZnO
catalyst particles (5 to 20 μm) with copper powder and sintering at moderate
temperatures. The resulting metal disks were catalytically active, providing an open
structure for mass transport, and efficiently transporting heat. Though Schuessler et al.
(2001) did not explicitly identify their reactor as a FTCMR, it featured many similar
aspects. By allowing for an operation at an optimum temperature, it resulted in high
reaction rates and selectivities and improved dynamic behavior.
Lange et al. (1998) used a FTCMR to study the selective hydrogenation of hexadiene.
The bulk support was a macroporous α-alumina membrane, covered by an intermediate
α-alumina with an average pore diameter of 60 nm, and a top γ-alumina layer (3 μm
thick) with a narrow pore size distribution and an average pore diameter of ~5 nm. On top
of this layer a thin Pt-containing TiO
2
layer (0.3–0.5 µm) was added by a dip-coating
procedure. The observed FTCMR hydrogenation activity was significantly higher than
that of comparable catalytic batch reactor using 100 mg of the same catalyst powder. The
authors noted that the low membrane permeability might represent a drawback for such
reactors, however. Gonzalez and coworkers [Lambert and Gonzalez, 1999; Vincent and
Gonzalez, 2002] investigated the hydrogenation of acetylene as well as 1,3-butadiene in a
16
catalytic membrane reactor. The active Pd/ γ-Al
2
O
3
membrane layer was prepared by the
sol-gel method on a macroporous α-Al
2
O
3
tube. Knudsen diffusion prevailed through the
top layer whose pore diameter was 3.6 nm. The reactor operated both with a separated
feed of reactants as well as in the FTCMR mode (which Gonzalez and coworkers termed
the "short contact time reactor"). The latter reactor resulted in the highest selectivity to
the partially hydrogenated products while still maintaining high conversion which the
authors attributed to the decreased contact time between the hydrocarbon and the catalyst.
FTCMR, based on porous polyacrylic acid (PAA) polymeric membranes containing
catalytically active Pd nanoparticles were studied by Schomäcker and coworkers
[Gröschel et al., 2005; Schomäcker et al., 2005] with the gas-phase partial hydrogenation
of propyne to propylene. They investigated the effects of membrane porosity, catalyst
loading and flow rate on the FTCMR performance with this model reaction. For the flow
rates that they applied, corresponding to residence times in the range of 1 to 4 s, they
found that the conversion decreased with increasing membrane porosity. Based on this
result, they concluded that there should be an optimal porosity for the membrane,
resulting in optimal catalyst distribution and in optimal residence times. The activity and
selectivity of the FTCMR was compared with that achieved in a fixed-bed reactor filled
with uniformly impregnated or “egg-shell” type catalysts. For equivalent catalyst
loadings and residence times the fixed-bed with the uniform catalysts showed low
conversion and poor selectivity, whereas both the FTCMR and “egg-shell” type catalyst
gave equally high conversions and selectivities, implying an optimized utilization of the
Pd.
17
Oxidative coupling of methane in a radial flow FTCMR was investigated by Ma et al.
(1998), feeding a mixture of oxygen, methane and helium in the tube-side of the
membrane, and forcing it to flow through the pores, which increased the linear velocity of
reactant gases over the catalyst, thus reducing external mass transfer limitations. Catalytic
membranes with different pore diameters (0.02 μm to 5 μm) were utilized, and the
performance of the FTCMR was seen to be superior to that of a packed-bed reactor using
the crushed membranes as catalysts. The membranes with the smallest pores showed the
best conversion and selectivity, confirming the benefits of operating in the Knudsen
regime. The location of the catalyst inside the membrane was also shown to have an
important impact on the reactor performance.
A composite consisting of a thin (a few μm) zeolite membrane film deposited on a
porous ceramic tubular support ( α- Al
2
O
3
with γ- Al
2
O
3
separation layer) was applied to the
catalytic dimerization of isobutene to isooctene by Torres et al. (2003). The reactants were
forced to permeate through the catalytic zeolitic layer with the FTCMR showing
comparable conversions with the fixed-bed reactor but with no apparent deactivation due to
coking from long-chain species. This was explained with the short and controlled residence
time in the FTCMR, which appeared to prevents further oligomerization. The dimerization
of isobutene in an FTCMR was also studied by Fritsch et al. (2004) applying polymer
composite catalytic membranes. The membranes were generated by deposition of a porous
reactive layer on top of a porous polymeric tubular membrane. Several catalysts, such as
silica supported Nafion
®
(Nafion
®
SAC-13), Nafion
®
NR50, Amberlyst
TM
15, and silica
supported phosphotungstic acid were mixed in solution with a polymeric binder. Teflon
®

AF, Hyflon
®
AD, polytrimethylsilylpropyne, and polydimethylsiloxane (PDMS) were
18
applied as binder to form a porous, reactive layer on top of a porous, polymeric, supporting
membrane. The experiments were performed in steady state by adjusting the reactant mass
flows and flowing against normal pressure at the outlet, resulting in an inlet pressure
around 4 bar, depending on the structure of and thus the pressure drop across the
membrane. The conversion of isobutene increased with the flow rate and thus with the
built-up pressure, up to values of 90 % without a significant change of selectivity to the
dimer, which reached around 14 %. It was assumed that the liquid products would fill the
membrane pores and required a certain flow to purge the pores from products and
oligomers. The authors discussed the difficulty of properly comparing the performance of
membrane reactors with that of conventional reactors regarding parameters such as space-
time yield.
Fischer-Tropsch (FT) synthesis (CO hydrogenation into liquid hydrocarbons) in
catalytic membrane reactors was investigated by Khassin and coworkers [Khassin, 2005;
Khassin et al., 2005]. In fact, fixed-bed reactors face difficulties for this reaction in terms
of overcoming internal diffusional limitations (e.g., through using smaller particles)
without undue increases in pressure drops, as well as heat management issues. Also
slurry and circulating fluidized-bed reactors used to overcome this problem suffer from
low catalyst loadings. Khassin and coworkers proposed the use of a FTCMR with heat-
conductive "plug-through contactor membranes" as a solution. They used permeable
composite monolith (PCM) membranes modified by a Co-Al co-precipitated catalyst.
During operation, the larger membrane pores (radius >2 to 3 µm) were wetted with liquid
but also filled with gas, but the smaller pores were completely flooded with liquid
products and are not permeable. The space-time-yield, based on the exterior volume of
19
the membrane, achieved with the FTCMR (200 kg of liquid hydrocarbons/m
3
h) was
higher than in traditional reactor designs for FT. The observed catalytic activity was also
up to three times higher than those in slurry reactor.
Westermann and Melin (2007) studied the use of commercial anodized alumina
membranes with very uniform parallel pores with diameters of 200 nm and a length of 60
µm the for selective gas-phase hydrogenation of acetylene. The pores were rendered
catalytically active by impregnation with Pd. Their experiments proved the high catalytic
activity of the FTCMR compared to the fixed bed reactors with the same amount of
catalyst. On the other hand, improved selectivity (ethene was the desired and ethane was
the undesired products) was only achieved in a small window of operating conditions.
Very recently, Minyukova et al. (2012) demonstrated the use of a permeable tubular
composite membrane (PCM) that acts as a catalytically active contactor in the 3-phase
hydrogenation reaction of liquid fatty acid triglycerides. Their PCM had Cu as the main
catalytic active component (and either Zn or Cr as the additional active metal ingredient),
and high permeability (300 mD, resulting in low pressure drop), high catalyst loading of
1 g/cm
3
(the same as a fixed-bed reactor used for comparison purposes), and high thermal
conductivity (1-4 W/m/K), making it possible to operate the reactor isothermally. High
conversion of 92 % – 99 % was achieved in the PCM experiments, with the rate of the
ester formation being higher for CuZn–PCM than for the CuCr–PCM. As the main
achievement of their research, they report a higher selectivity towards valuable ester
formation rather than the formation of alcohols and hydrocarbons as in case with
conventional reactors.

20
1.2.3. Selective-differential FTCMR
Liquid-phase partial hydrogenations are an important class of industrial reactions for
which the reaction rate is often limited by low hydrogen solubility in the liquid, which
argues against performing such reactions in a single-pass integral reactor. An alternative
is using a differential reactor combined with a recycle and re-saturation loop. This
concept is also ideally suited for use with FTCMR due to their feature of providing
effective contact between reactants and catalysts caused by the convective flow through
the pores. These so-called “Selective-differential FTCMR” have been applied to the
hydrogenation of nitrates in water and to the partial hydrogenation of several unsaturated
liquid hydrocarbons. Some of the key studies are summarized in Table 1.3, which also
indicates the reactions studied, and the types of membranes and catalysts utilized.
Ilinitch et al. (2000) performed, for example, the reduction of nitrates in water using
three different reactor configurations. In a fixed-bed reactor with supported Pd-Cu/ Al
2
O
3

catalysts pronounced internal diffusion limitations were observed. A catalytic membrane
reactor in the absence of forced-flow also showed low catalytic activity. The use of a
FTCMR with liquid recycling, on the other hand, succeeded in minimizing diffusion
limitations, and in significantly increasing the catalytic activity. FTCMR using catalytic
macroporous polymeric membranes were also successfully applied by the same group
(Ilinich et al., 2003) to water denitrification.
Reif and Dittmeyer (2003) compared the FTCMR concept for aqueous nitrate and
nitrite reduction with the catalytic diffuser concept, in which the reactants were delivered
from different sides of the membrane. Both reactors induced active contact between
reactants and catalysts, however, only the FTCMR allowed for the elimination of pore
21
diffusion and for the attainment of very short contact times. Due to high pressure drops in
the FTCMR, conventional asymmetric ceramic membranes with pore diameters of 100
nm in the active layer were found to be unsuitable. Consequently, symmetric supports
with pore diameters ~3 μm, coated with Pd, were tested. Reif and Dittmeyer (2003)
reported that flowing from the outside to the inside reduced membrane plugging. For
nitrite hydrogenation in the FTCMR, a significantly higher activity and a lower formation
of ammonium was observed.

Table 1.3: Flow-through catalytic membrane reactors for selective reactions with liquid recycling
Authors Reaction Membrane Catalyst Pore Diameter, (nm)
Ilinitch et al., (2000)
Water
denitrification
SiO
2
/Al
2
O
3
ceramic
Pd-Cu /
γ- Al
2
O
3

1000
Ilinitch et al., (2003)
Water
denitrification
Macroporous
PA MF
Pd/Cu/Pd-
Cu
400
Reif and Dittmeyer, (2003)
Water
denitrification
Symmetric ceramic Pd 3000
Ludtke et al., (1998)
Water
denitrification
Microporous
PEI
Pd-Cu /
γ- Al
2
O
3

~ 2
Schmidt et al., (2005)
Cyclooctadiene;
1-octyne; phenyl
Acetylene
hydrogenation
Porous PAA Pd 130 - 380
Purnama et al., (2006)
Methyl Styrene
hydrogenation
α-Al
2
O
3
Pd 1900
Schmidt et al., (2008)
Cyclooctadiene
hydrogenation
α-Al
2
O
3
Pd 600 - 3000
Fritsch and Bengston, (2006)
Sunflower oil
hydrogenation
Porous polymer
(PES,PAI)
Pt/Pd 3900 - 9000
Schmidt and Schomäcker,
(2007)
Sunflower oil
hydrogenation
Porous
α-Al
2
O
3

Pd/Pt 3000
Wehbe et al., (2010)
Water
denitrification
Porous
α-Al
2
O
3

Pd-Cu /
γ- Al
2
O
3

5, 10, and 25
Pera-Titus et al., (2012)
Water
denitrification
Porous
α-Al
2
O
3

Pd-Cu /
γ- Al
2
O
3

5, 10, and 25

22
A somewhat different FTCMR was tested by Lüdtke et al. (1998). In their study
nitrates in water were hydrogenated flowing through a catalytic membrane reactor
operating in the cross-flow mode. In contrast to the FTCMR above, the treated water was
continuously removed after only a single passage through the membrane, while the
remaining water returned though a recycle loop into a tank pressurized with hydrogen.
For trans-membrane fluxes of 5 to 25 L/m
2
h a fairly constant selectivity of 80 % towards
nitrogen was observed.
Schmidt et al. (2005) used a FTCMR for the partial hydrogenation of several
unsaturated hydrocarbons (cyclooctadiene, 1-octyne, phenyl acetylene and geraniol), and
compared its behavior to that of a slurry and of a fixed-bed reactor. The FTCMR utilized
a macroporous, cross-linked PAA membrane with imbedded Pd catalyst nanoparticles (1
wt %) connected through a loop to a saturation vessel. The liquid reaction mixture was
re-saturated with hydrogen up to 100 times. For the hydrogenation of octyne the desired
reaction was much faster than the undesired reaction, so that mass transfer enhancement
in the FTCMR (or in the slurry reactor) had little effect. For phenyl acetylene and
cyclooctadiene the desired reaction was still faster than the undesired one, but mass
transfer was more significant, as indicated by increased selectivity obtained in the
membrane and the slurry reactor when compared to the fixed-bed reactor. For geraniol
the reaction rates were nearly equal. For this reaction the FTCMR reached higher
selectivities than the slurry reactor; however, the activity was lower due to the decrease
of hydrogen concentration in each pass through the membrane, with the average
hydrogen concentration that the catalyst encountered being substantially lower than the
saturation concentration.
23
Purnama et al. (2006) studied the partial hydrogenation of α-methylstyrene to cumene
in a FTCMR using a macroporous α-Al
2
O
3
membrane impregnated with Pd. The
measured productivity in the FTCMR, defined as conversion per unit time and catalyst
mass, exceeded that in other types of membrane as well as in conventional (slurry,
trickle-bed, and bubble-column) reactors. For optimum performance, the flow rate had to
be adjusted, so that the hydrogen conversion reached 100 % exactly when the reactants
left the membrane. The same FTCMR was applied to the hydrogenation of
cyclooctadiene, reaching selectivities similar to those in slurry experiments, and
outperforming the comparable fixed bed reactor [Schmidt, 2007]. For the same Pd
amount, membranes with smaller pores (0.6 μm) were more active than the membranes
with bigger pores (1.9 or 3.0 μm). The obtained space time yields were much higher than
those reported for conventional fixed-bed or trickle-bed pellet catalysts or slurry catalysts
in bubble columns [Schmidt et al., 2008]. Due to encouraging scale-up results, this
FTCMR was singled out [Caro et al., 2007] as one of the most promising candidates for
the first industrial application of catalytic membrane reactors.
In an interesting computational study Albo et al. (2006) used detailed atomistic
molecular dynamics simulations to study transport and reaction phenomena in straight
pores with diameters between 10 and 150 nm, typical of those encountered in
nanostructured catalytic membranes prepared by anodic oxidation. They chose the
selective oxidation of hydrocarbons as a model reaction in order to assess the ability to
prevent over-oxidation by taking advantage of the short contact time in the FTCMR.
Knudsen diffusion was identified as the dominant mass transfer mechanism, particularly
24
for smaller pores and at lower pressures, whereas surface diffusion was only present at
temperatures below 700 K.
Fritsch and Bengtson (2006 a and b) used catalytically active porous membranes for
the selective hydrogenation of viscous liquids, such as sunflower oil in a FTCMR. The
high viscosity of the oil and its low hydrogen solubility induces severe mass transfer
limitations, so industrially vegetable oil hydrogenation takes place in stirred tank reactors
with finely dispersed catalysts at temperatures between 170
o
C and 200
o
C, and at 2 to 5
bar of H
2
pressure. In such reactors up to 50 % of the undesired trans-isomerized fatty
acids are generated. By applying the FTCMR, the content of trans-isomers was reduced,
while catalyst immobilization of the membrane pores avoided expensive filtration steps.
Schmidt and Schomäcker (2007) also used a FTCMR with a porous α-Al
2
O
3

membrane impregnated with Pd or Pt as active catalyst for partial hydrogenation of
sunflower oil in n-heptane as solvent, and compared its performance with that of a slurry
reactor with powdered catalyst. Their experimental FTCMR did not provide the desired
outcome reaching significantly lower contents of the fully hydrogenated stearic acid and
a higher content of the undesired trans-isomers.
Wehbe et al. (2010) used Pd-Cu catalysts supported on mesoporous membranes (pore-
size 5-25 nm) under the flow-through configuration to study the membrane performances
and the influence of pore size on the hydrogenation of nitrates at room temperature by
pumping a solution of NaNO
3
saturated with dissolved hydrogen through the catalytic
membrane. Since according to Henry’s law, the hydrogen solubility is proportional to the
hydrogen partial pressure in the gas phase, their study highlighted the issue of reaction
limitation by the low solubility of hydrogen in water when the FTCMR was used.
25
During their experiments, concentration polarization of nitrate ions was observed,
especially when using small pore size (5 nm and 10 nm) membranes, leading to the
improvement in the conversion of nitrate ions by increasing their concentration near the
catalytic zone. On the other hand, concentration polarization resulted in a detrimental
effect on the selectivity because it interfered with the control of residence time through
the catalytic zone, and as a result undesirable ammonium ions were formed. Also, they
observed that under the flow-through configuration, the rate of nitrate disappearance
increased with increasing flow through the membrane, an unusual finding based on the
fact that higher flow rates signify decreased contact time in the catalytic zone.
In a follow-up publication from the same group, Pera-Titus et al. (2012) investigated
the aforementioned unusual phenomenon by modeling nitrate reduction in FTCMR
taking into account concentration polarization. They verified their model with the
experimental data gathered by Wehbe et al. [Wehbe et al., 2010]. They showed that the
surprising increase of the nitrate conversion with increasing the trans-membrane flow rate
could be likely due to an enhancement of water viscosity in the membrane pores,
providing unexpectedly low ionic and H
2
diffusivities, especially for narrower-sized
membranes. Lower diffusivities, in turn, increase the retention of nitrates in the
membrane pores, and thus nitrate conversion increases. They attributed the increased
viscosity to the formation of an adsorbed water layer on the pore walls, showing a higher
viscosity than the bulk water in membrane pores.

26
1.3. Scope of the Present Work
As stated previously, complete decomposition of VOC is one of the major applications for
FTCMR. One class of harmful chemicals receiving increased attention today are chemical
warfare (CW) agents, due to the increased chance in recent years of the use of chemical
weapons both as a result of potential terrorist attacks and of ongoing international conflicts.
Activated carbon (AC) impregnated with various metals (e.g., Ag, Cu and Cr) has been used
commonly as adsorbent in individual (IP) and collective protection (CP) systems against CW
agents [Tsotsis et al., 2008]. The chemical agents are either physically adsorbed and/or
reacted with the metals, which often act as catalysts. Large persistent molecules (LPM), such
as nerve agents can be easily adsorbed by activated carbon. However, the CW agents are
typically not destroyed, but simply physisorbed on the AC. Though the AC bed has a large
theoretical saturation capacity, breakthrough often occurs well before such saturation occurs;
and since even at trace levels CW agents could have devastating consequences, the AC beds
are often greatly overdesigned, a situation exacerbated by the fact that it is often difficult to
predict “a priori” when AC breakthrough will occur, for a given agent. Since the CW agents
remain intact on the spent AC media, their handling and storage prior to ultimate destruction
presents obvious risks [Tsotsis et al., 2008].
Small, non-persistent molecules (SNPM), on the other hand, such as hydrogen cyanide
and cyanogen chloride (CK) undergo irreversible reaction with the metals impregnated
on the carbon, and as a result the activated carbons become spent (or poisoned).
Regeneration with common approaches, such as PSA, TSA, and chemical washing, is not
possible. Therefore, although activated carbon is generally effective for adsorbing a
broad range of compounds, it is only suited for one-time usage in IP and in CP systems
27
[Tsotsis et al., 2008].
In the present study the FTCMR concept is applied for the thermal oxidation of a CW
simulant, namely dimethyl methylphosphonate (DMMP). DMMP is known as a chemical
precursor for the more toxic gas Sarin (GB), and is usually used to simulate its
characteristics (Figure 1.2). It is a colorless and combustible liquid which emits a distinct
odor. DMMP has a vapor pressure of 111.1 Pa at the temperature of 298.2 K [Butrow et
al. 2009], and a liquid density of 1154.2 kg/m
3
at the temperature of 293.15 [Fan and
Wang, 2010].

Figure 1.2: Molecular structure of Sarin (GB) and its simulant DMMP [Li et al., 1992]

In our study tubular porous alumina membranes have been employed as the support
for a platinum oxidation catalyst, and radial forced-flow of reactants through the pores is
generated. As with all FTCMR, one can envision this reactor as a conventional radial-
flow, packed-bed reactor filled with catalytic pellets whereby the pellets are crushed to
such a small size that the interstitial volume is reduced to nanometer size.
The adsorption and reaction of DMMP on different substrates and catalyst have been
investigated by a number of research groups over the years. Some of the key previous
studies on this subject are summarized in Table 1.4. In an early study, Graven and
coworkers [Graven et al., 1966] studied the conversion of DMMP vapor in a stream of
air, or nitrogen, over platinum-alumina catalysts in a cylindrical fixed-bed reactor. In
28
their research a commercial catalyst and a number of laboratory-prepared catalysts were
investigated over a range of temperatures from 300
o
C to 500
o
C, residence times from
0.15 to 2.7 s, and average catalyst particle sizes from 0.31 to 2.4 mm. According to their
findings, the activity of fresh commercial catalyst was too high to permit kinetic studies,
but after some hours on stream, deactivation occurred to the point that measurable
quantities of DMMP appeared in the effluent. Pseudo-steady-state kinetics over
deactivated commercial catalyst was approximately first order with respect to DMMP,
with an activation energy of 7 to 8 kcal/mol. Though they found that DMMP conversion
proceeded almost as rapidly in N
2
as in air, they reported different gas phase reaction
products for air and N
2
: When air was used as the DMMP carrier, they found methanol
(MeOH) and carbon dioxide (CO
2
) were the major products along with a small amount of
dimethyl ether (DME). With N
2
no CO
2
was detected, but methane, formaldehyde, and
DME were found in small amounts. In both cases, MeOH was formed in amounts
comparable with the amount of DMMP converted. They suggested that the MeOH might
result either by a hydrolytic reaction between DMMP and residual catalyst “water,”
present as surface hydroxyl groups even on alumina dried at 500
o
C or by the reaction of
DMMP with water formed as an oxidation product from DMMP. Moreover with an air
carrier, a syrupy condensate largely H
3
PO
4
was obtained at the exit of the reactor.
Templeton and Weinberg (1985a) investigated the adsorption and decomposition of
DMMP on γ-alumina using inelastic electron tunneling spectroscopy. They found that
DMMP adsorbed molecularly at 200 K via hydrogen bonding of the phosphoryl oxygen
to a surface OH site producing saturation coverage. On the other hand, at higher
temperatures between 295 and 473 K, they found that DMMP adsorbs dissociatively. The
29
initial step involved the binding of the phosphoryl oxygen at a coordinately unsaturated
aluminum atom and hydrogen bonding of one of the methoxy oxygens to a surface OH
site. This was followed rapidly by nucleophilic substitution at the phosphorus by a
surface OH group, leading to loss of the hydrogen-bonded methoxy group as methanol,
resulting into a bridging phosphonate species. At temperatures above 573 K, the methyl
methylphosphonate adspecies decomposed to yield the tri-denate methyl phosphonate
adspecies. They estimated the total surface coverage was less than 20 % of a monolayer.
Based on the above findings, they proposed the following scheme (Figure 1.3) for the
chemisorption of DMMP on alumina surface:

Figure 1.3: The adsorption of DMMP onto Al
2
O
3
as presented by Templeton and Weinburg (1985a)
A) DMMP dissociates via cleavage of the phosphorus-oxygen bond. B) Nucleophilic attack by lattice
oxygen occurs at the phosphorus to form a penta-coordinate intermediate. C) A bridged phosphonate
product forms on the surface. Methoxy groups are present on the surface at lower temperatures

Templeton and Weinberg (1985b) also reported that once the bridging phosphonate
species was formed, it underwent nucleophilic substitution at the remaining methoxy
carbon by a surface O atom, with simultaneous protonation at the methoxy oxygen by a
surface hydroxyl group to generate a surface-bound unsubstituted phosphonate and a
surface methoxide. The unsubstituted phosphonate and methoxide were eliminated
rapidly to give gaseous methanol and a surface-bound methylphosphonate. The surface
bound hydroxyl-methylphosphonate was observed as an intermediate when DMMP was
30
adsorbed at 373 K and heated to 573 K, but not when DMMP was adsorbed at 573 K.
The final form of the adsorbed phosphonate in each case was methylphosphonate.
The surface chemistry of DMMP on Rh was studied by Hegde and coworkers [Hegde
et al., 1985]. They investigated the adsorption of DMMP on clean as well as carbon-
covered Rh (100) surfaces. At 100 K, two different adsorption states were identified by
temperature programmed desorption (TPD) and spectroscopic measurements, a
monolayer phase, and a multilayer phase. For the monolayer phase of DMMP on carbon-
free Rh (100), between 60 % and 70 % of it was decomposed upon heating, leaving
carbon, phosphorus, and oxygen on the surface. On carbon-covered Rh, the
decomposition of DMMP was strongly inhibited.
Dulcey and Lin (1985) analyzed the thermal interaction of DMMP with a platinum
wire (99.99 % pure) at temperatures above 1100 K using the laser-induced fluorescence
(LIF) technique. In the decomposition of DMMP, the PO radical was observed to desorb
from the freshly-cleaned Pt catalyst surface above 1100 K. No PO signals were observed,
however, when the Pt catalyst was not activated prior to each run indicating clearly,
according to the authors, that the production of PO in these decomposition reactions was
catalytic in nature. Moreover, they observed that for DMMP decomposition, slow surface
deactivation occurred at high temperatures. Dulcey and Lin (1985) also studied the effect
of pressure on DMMP interaction with Pt. They showed that the continued increase of the
DMMP pressure rapidly reduced the PO signal, reflecting the propensity for surface
deactivation at high pressures. Prolonged experiments at high DMMP pressures resulted
in the total disappearance of the PO signal because of surface deactivation. Interestingly,
the deactivated Pt surface could be readily cleaned and regenerated by flowing oxygen at
31
temperatures above 1300 K in the absence of DMMP. During the catalytic decomposition
of DMMP, addition of a small amount of O
2
to the flow was found to enhance the yield
of PO. Continued increase of O
2
however, rapidly diminished the PO yield, and
ultimately resulted in the deactivation of the surface. They suggested the deactivation of
the catalyst might be due to the deposition of less volatile phosphorus oxides generated
when O
2
and DMMP were both simultaneously present. The addition of O
2
to DMMP
reaction system was also found to generate OH radicals, strongly indicating the presence
of absorbed H atoms.

32
Table 1.4: Previous studies on the catalytic decomposition/adsorption of DMMP
Authors Metal
catalyst
Support/
Metal oxide
Operating
temperature, (K)
DMMP load Protection
time*, (hr)
Graven and Weller,
(1966)
Pt
Commercial/
γ-Al
2
O
3

573 - 773
0.2 – 3.5 mg/L of
Air/N
2
carrier gas
5 - 16
Templeton and
Weinberg, (1985a,b)
- γ-Al
2
O
3
200 - 673
Direct exposure
under vacuum
N/A
Hegde et al., (1985) Rh - 100 - 250 N/A N/A
Dulcy and Lin,
(1985)
Pt - 1100 - 1350
Direct exposure
under vacuum
N/A
Henderson et al.,
(1986)
Fe
2
O
3
,
SiO
2

- 170 - 700
Direct exposure
under vacuum
N/A
Hegde and White,
(1987)
Oxidized
iron
- 100 - 300
Direct exposure
under vacuum
N/A
Henderson and White,
(1988)
Pt - 100 - 800
Direct exposure
under vacuum
N/A
HSU et al., (1990) Pt TiO
2
623 - 796
1 % in He/O
2

carrier gas
16 for
monolith and
less than 6 for
Guo et al., (1990) Pd, Ni - 100 -900 N/A N/A
Li and Klabunde,
(1991)
- MgO 473 - 773
2 µL cumulative
in 50 mL/min He
N/A
Atteya and Klabunde,
(1991)
- MgO N/A N/A N/A
Li et al., (1991) - MgO
Ambient for
vacuum and 773
under He flow
1 µL cumulative
in 15 mL/min He
N/A
Li et al., (1992) - MgO 473 - 1173
2 µL cumulative
in 50 mL/min He
N/A
Tzou and Weller,
(1994)
Pt γ-Al
2
O
3
423 - 673 N/A 50 - 135
Lee et al., (1994) Cu Hydroxyapatite 646 - 1046
28.85 µmol/L of
N
2
/O
2
carrier gas
0.5 – 9
Ryu et al., (1995)
Pt, Pd,
Rh, Ru
Al
2
O
3
573
6.5 mg/L of
Air carrier gas
2.5 - 7
Mitchell et al.,
(1997)
-
γ-Al
2
O
3
, Fe
2
O
3
,
MgO, La
2
O
3

298 - 573 N/A N/A
O’Shea et al., (1997) - TiO
2
N/A
0.1 Molar– 2.72 Molar
solution
N/A
Tesafai et al., (1998) Fe γ-Al
2
O
3
298 1:1000 in He N/A
Obee and Satyapal,
(1998)
TiO
2
glass N/A
1 ppm
V
– 90 ppm
V
in
a N
2
/O
2
mixed carrier gas
N/A
Segal et al., (1999) -
Manganese
oxide
313 - 373
0.14 mol % of a 30
mL/min flow of
air carrier gas
2 -5

33
Table 1.4: Previous studies on the catalytic decomposition/adsorption of DMMP, (continued)
Authors
Metal
catalyst
Support/
Metal oxide
Operating
temperature, (K)
DMMP load
Protection
time*, (hr)
Cao et al., (2000)
Pt, Ni,
Fe, Cu, V
γ-Al
2
O
3
,
SiO
2
, TiO
2
573 - 723 1300 ppm in Air
1.5 – 12.5 for
alumina and up
to 100 for others
Rusu and Yates,
(2000a)
- TiO
2
200-300
Direct exposure
under vacuum
N/A
Rusu and Yates,
(2000b)
- TiO
2
160-486
Direct exposure
under vacuum
N/A
Segal et al., (2001) Mn Al
2
O
3
473 - 673
Saturated DMMP/Air
at 298 K
0.08 - 8
Cao et al., (2001) -
Activated
carbon
573 - 723 1300 ppm in Air 0 - 100
Decker et al., (2002)
Ni, Fe,
Cu, V
CaO 473 - 773
1µL injection
in He
0
Sheinker and
Mitchell, (2002)
Fe γ-Al
2
O
3
298 - 673
30.25 µmol/L of
He carrier gas
N/A
Mitchell et al.,
(2003)
Ce/Fe γ-Al
2
O
3
298
30.25 µmol/L of
He carrier gas
N/A
Mitchell et al.,
(2004)
Ce γ-Al
2
O
3
298
47.0 µmol/L of
He carrier gas
N/A
Trubinsyn and,
Vorotsov, (2005)
Fe TiO
2
300 N/A N/A
Moss et al., (2005) - TiO
2
Room Temperature
2e -5 Torr/Air
bubbling in DMMP
N/A
Mitchell et al., (2007) Fe η-Al
2
O
3
Room Temperature
47.0 µmol/L of
He carrier gas
N/A
Panayotov and
Morris, (2008)
Au TiO
2
295 N/A N/A
Han et al., (2008) -
Sulfated
TiO
2

348 210 ppm 17 -23
Ratliff et al., (2009) Pt/Au TiO
2
100 - 800 N/A N/A
Panayotov and
Morris, (2009a)
- TiO
2
295 - 600 0.6 Torr N/A
Panayotov and
Morris, (2009b)
- TiO
2
295 0.6 Torr N/A
Mattsson et al.,
(2009)
Zr TiO
2
N/A
6.5 µg/min in Air
carrier gas for 20 min
N/A
Chen et al., (2010) Ce Ru 100 -900 Direct exposure N/A
Mera et al., (2010) - TiO
2
300 - 373 28.5 µM - 34.9 µM N/A
Mitchell et al., (2011) Mn η-Al
2
O
3
Room Temperature
47.0 µmol/L of
He carrier gas
9
* The period during which the concentration of DMMP is so low that it cannot be detected by online measurements
34
Henderson and coworkers [Henderson et al., 1986] studied the interaction of DMMP
dosed at 170 K onto SiO
2
and Fe
2
O
3
by TPD and Auger electron spectroscopy (AES). On
dehydrated SiO
2
there was no DMMP decomposition and there were two DMMP TPD
peaks, a multilayer state at 200 K - 210 K and a monolayer state at 275 K. On hydrated
SiO
2
no more than 10 % of a monolayer of DMMP decomposed, and the only detectable
TPD products were methylphosphonate (MP) and methanol. On clean Fe
2
O
3
a multilayer
DMMP phase was observed, but no molecular peak corresponding to the monolayer
phase. Decomposition led to CO
2
, CO, MeOH, HCOOH, H
2
, H
2
O, and an adsorbed
phosphate species. They proposed that the migration of the phosphorus into an iron
oxide-phosphate phase below the surface could explain the experimental observations
during the DMMP decomposition according to the following scheme (Figure 1.4):

Figure1.4: Proposed sequence for the continuous decomposition of DMMP on Fe
2
O
3
[Henderson et al., 1986]
35
The adsorption/desorption behavior of DMMP, dosed on oxidized iron at 100 K, with
and without co-adsorbed water was investigated by Hegde and White (1987) using TPD
and AES. DMMP decomposition was found to be limited, and large exposures led to
molecular DMMP desorption which is characteristic of multilayers (200 K - 210 K).
Their experiment showed that pre-exposure to H
2
O increased the extent of DMMP
decomposition. At low coverage, DMMP adsorbed at 100 K on a clean FeO
x
surface
completely decomposed. The major low temperature desorption products were MeOH,
CO, H
2
, and H
2
O. The appearance of these desorption products at such low temperatures
(175 K-185 K) showed that some DMMP had decomposed even below 175 K. On the
other hand, they found that carbon, phosphorus, and oxygen were left on the surface after
heating to 500 K. Co-adsorption of water and DMMP increased the fraction of adsorbed
DMMP which decomposes, probably due to hydrolysis.
The interaction of DMMP with Pt (111) was studied with high-resolution electron
energy loss spectroscopy (HREELS), positive and negative static secondary ion mass
spectrometry (SSIMS), TPD, and AES by Henderson and White (1988). According to their
results, DMMP could bind strongly and molecularly to Pt (111) at 100 K. Moreover, their
HREEL results indicated that it had bound through the oxygen lone pairs on the ―P=O.
Decomposition occurred after heating above 300 K, yielding exclusively CO and H
2
in
TPD, with P and a small amount of C left on the surface. The decomposition mechanism
first involved PO-CH
3
bond cleavage (300 K-400 K), with some P-OCH
3
cleavage at high-
DMMP coverage, and then P-CH
3
bond cleavage (400 K-500 K) leaving predominantly PO
and an unidentified PO
x
species. Both species were stable up to at least 500 K. One
important conclusion from these surface science studies is that using Pt for catalytic
36
decomposition of organophosphorus compounds, like DMMP, in a non-oxidizing
environment may be limited by the accumulation of surface P.
The decomposition of DMMP on Pd (111) and Ni (111) surfaces was studied by Guo
et al. (1990) using AES and temperature-programmed reaction spectroscopy (TPRS). In
the absence of O
2
thermal decomposition of DMMP occurred for Pd below 300 K and for
Ni below 340 K based on the H
2
and CO evolution. Phosphorus was deposited on both
the Pd (111) and Ni (111) surfaces following the DMMP decomposition. They proposed
the following models for decomposition of DMMP on these metal surfaces (Figure 1.5):

Figure1.5: Proposed models for the decomposition reaction of DMMP on Pd (111) and Ni (111)
[Guo et al., 1990]

Guo et al. (1990) also investigated the effect of oxygen treatment for the recovery of
metals surfaces, after being coated with P during DMMP decomposition. Oxidation at
1075 K removed the surface P on Pd (111). On Ni (111); however, surface P could not be
removed by oxidation at 1075 K. Combined with similar experiments on Mo (110), they
concluded that the early transition metals might be more suitable for the catalytic
oxidation of organophosphonate compounds, like DMMP, on the basis of the lower
temperature for sustained removal of surface P by oxygen at the temperature of 900 K on
Mo (110) compared to the temperature of 1075 K for Pd (111).
37
Hsu and coworkers [Hsu et al., 1990] studied a low-temperature oxidation catalyst
(monolithic Pt-TiO
2
) for the oxidation of DMMP at 350 K-523 K under low-pressure
conditions in a quartz fixed-bed reactor. The concentrations of reagents (DMMP and O
2
)
and products (CO, CO
2
and H
2
O) were analyzed by a mass spectrometer during the
course of reaction. No P-containing species (other than unreacted DMMP) were
observed. The major oxidation products detected were CO
2
and H
2
O and no detectable
breakthrough of DMMP was noted over 16 h of operation at 523 K with O
2
/DMMP> 1.
Moreover, no apparent poisoning was observed during the course of this study. The
apparent activation energies of formation of CO
2
and H
2
O were found to be 3.6 and 4.2
kcal/mol, respectively. These authors also compared the performance of the monolithic
catalyst and a three-way automotive catalyst pellet. They found that the monolithic
catalyst showed performance characteristics superior to three-way automotive catalysts in
terms of lower operating temperature, more catalyst protection time (16 h for monolith
vs. less than 6 h for the pellets), and the types of reaction products generated (for
example, poisonous CO was one of the reaction products for the catalyst pellets). Results
of surface scanning Auger microscopy (SAM) and X-ray photoelectron spectroscopy
(XPS) revealed the formation of P
2
O
5
on the catalyst surfaces, and irreversible poisoning
was observed when insufficient O
2
was supplied to the feed stream at low temperatures.
Based on this information, they attributed the possible deactivation of the catalyst directly
to the formation P
2
O
5
on the Pt active sites.
In a series of publications, Klabunde and coworkers [Li and Klabunde, 1991; Li et al.,
1991; Atteya and Klabunde, 1991; Li et al., 1992] reported the results of their studies on
38
the interaction of DMMP with heat-treated MgO under vacuum and also under the flow
of an inert gas, i.e., He. Their studies concluded that:
(1) DMMP and organophosphates adsorbed strongly on heat-treated MgO, probably
on Lewis acid sites (Mg
2+
) through the oxygen atom of the P=O bond.
(2) Based on measurements of heats of adsorption, DMMP adsorbs on MgO mainly by
(sometimes dissociative) chemisorption.
(3) Upon heating to 500
o
C, DMMP completely decomposed evolving formic acid and
methanol, while leaving adsorbed on the MgO an organophosphate species, according to
the following scheme (Figure 1.6):

Figure1.6: Proposed models for the decomposition reaction of DMMP on heat-treated MgO
[Li and Klabunde, 1991]

(4) The decomposition of DMMP on nanoscale particles of MgO was a surface
stoichiometric process. Two surface MgO moieties were needed for destructively
adsorbing one DMMP molecule (Figure 1.6).
(5) The lower the temperature the lower the decomposition capacity of MgO.
(6) Addition of water to the DMMP compound enhanced dramatically the ability of
MgO to destroy the phosphorus compound.
39
(7) It was found that only about 30 % of the deactivated MgO capacity for DMMP
decomposition could be regenerated at 500
O
C using water. The surface-adsorbed species
that were released during this regeneration did not contain any phosphorus.
The catalytic decomposition of DMMP by alumina-supported precious metal catalysts
was studied in a fixed-bed flow reactor by Ryu et al. (1995). Pt showed the best
performance for the reaction compared to other metals (Ru, Pd, and Rh). DMMP vapor in
the air stream was decomposed completely in the start of the experiment but the
performances of the catalysts were reduced rapidly with the lapse of time. Ryu et al.
(1995) tried to elucidate the mechanism for the decomposition of DMMP and for the
catalyst deactivation by analyzing the reaction products and the surface of the catalysts.
According to them, the methoxy groups of DMMP were converted via the hydrolysis
reaction on the alumina support into methanol, which was subsequently oxidized to
carbon dioxide and water. P-containing decomposition products included mostly
nonvolatile monomethyl-methylphosphonate (MMMP) and methyphosphonic acid
(MPA). Some of these nonvolatile products adhered strongly onto the surface of the
catalysts, thus resulting in catalyst deactivation.
After 27 years from the early paper of Graven et al. (1966) on the catalytic oxidation
of DMMP in which he was one of the authors, Weller returned to the subject in 1994
[Tzou and Weller, 1994]. In the most recent study, the catalytic oxidation of DMMP in
air was investigated over Pt/Al
2
O
3
catalysts (Pt loading of 0.5 wt.% or 2.0 wt.%) at
various temperatures (150
o
C, 250
o
C, or 400
o
C) in a conventional fixed-bed reactor. At
250
o
C and a residence time of 0.69 s, carbon balances indicated that 90-95 % of the C
atoms in the DMMP feed were completely oxidized to CO
2
during the periods when
40
DMMP destruction was > 99 % (50 h for the 0.5 wt.% Pt catalyst and 76 h for the 2.0
wt.% Pt catalyst). Only after the end of the protection period, the period during which the
concentration of DMMP in the reactor outlet was so low that it could not be detected by
analytical measurements, was MeOH detected in the gaseous effluent. HPLC analysis of
a liquid product that was collected showed that it consisted of four intermediate P-
containing compounds, namely dimethyl phosphate (DMP), monomethyl phosphate
(MMP), MMMP, and MPA. They proposed the following reaction scheme to account for
the production of these intermediates by a series of oxidative and hydrolytic reactions
(Figure 1.7):

Figure 1.7: Suggested reaction scheme for the production of P-containing intermediate during the
catalytic oxidation of DMMP over Pt/Al
2
O
3
[Tzou and Weller, 1994]

They observed different results at 400
o
C. Although under this temperature the period
of complete DMMP destruction was very long (> 135 h) with the 0.5 % Pt catalyst,
oxidation as measured by CO
2
production was incomplete and the oxidation activity
decreased with time on stream. At 400
o
C the hydrolytic reactions apparently became
sufficiently efficient to effect complete DMMP conversion for extended periods despite
the drop in oxidation activity. Characterizing the deactivated catalyst provided
incremental evidence that the alumina had reacted with product phosphoric acid to give
41
AlPO
4
. So far, no other research group has reported such a long protection time in catalytic
oxidation of DMMP.
Lee and coworkers [Lee et al., 1994] studied the oxidative decomposition of DMMP
over Cu-substituted hydroxyapatite catalysts in a conventional packed-bed reactor in the
temperature range of 373 K-773 K. By substitution of a portion of the Ca
2+
by Cu
2+
in the
hydroxyapatite lattice, its catalytic activity was markedly increased. A catalyst with the
composition of Cu
2
Ca
8
(PO
4
)
6
(OH)
2
, (named Cu
2
-HA), had the highest activity among
these catalysts. All the catalysts showed 100 % conversion at the beginning of the
reaction, but became deactivated after a period of time. Only CO
2
and H
2
O were
produced during the "protection period" of the 100% conversion. With deactivation, CO
2

formation decreased and MeOH and DME appeared as products. No P-containing
product was detected in the gas phase. Some phosphoric acid condensate appeared in the
outlet of the reactor after deactivation. Lee et al. (1994) tried to regenerate the spent
catalysts by treating them with O
2
for 1 h at 573 K and 773 K. While only a small
fraction of the initial activity was recovered at 573 K, the initial activity was fully
recovered at 773 K. After the 773 K oxidative regeneration no carbonaceous compounds
remained on the surface, but the phosphorus compounds that had accumulated during the
reaction were not eliminated. The state of the phosphorous compounds was investigated
by solid-state P-NMR, indicating only the presence of methylphosphonate on the catalyst.
No peak for phosphoric acid/phosphorous pentoxide appeared during the P-NMR studies
indicating the transformation of PO
x
to a form of phosphates by reaction with water vapor
formed during the reaction.
42
Mitchel et al. (1997) examined the adsorption and decomposition of DMMP on four
different metal oxide surfaces, namely aluminum oxide, magnesium oxide, lanthanum
oxide, and iron oxide. Aluminum, magnesium, and lanthanum oxides were observed to
behave in much the same way, with initial binding of the P=O species to the surface at an
acid site, followed by stepwise elimination of the methoxy groups, beginning at
temperatures as low as 50 °C, which combined with surface hydrogen to yield methanol
that evolved from the surface. The final product observed for these oxides was a surface-
bound methylphosphonate, with the P-CH
3
bond intact, which was resistant to further
oxidation even in the presence of 70 Torr of oxygen at 300 °C-400 °C. According to their
results, adsorption on iron oxide yielded a different sequence of events, with an initial
adsorption occurring again with the P=O moiety binding to an acid site, although there
was some indication of the formation of a second type of surface complex. The primary
interaction on iron oxide appeared to be much stronger than with the other oxides, and
probably involved the unidentate coordination of the DMMP to a Lewis acid site on the
surface. The nonselective elimination of both the methoxy and the phosphorus-bound
methyl groups began only after heating above 200 °C, but occurred with total elimination
of the methyl and methoxy groups observed after heating above 300 °C in vacuum.
Mitchel et al. (1997) attributed the ease with which iron oxide cleaves the P-CH
3
bond to
the availability of multiple oxidation states to the iron atom. They also explained how
participation of the Fe (III)/Fe (II) redox couple in the reaction could provide a low-
energy path for oxidative cleavage of the P-CH
3
bond. The other oxide surfaces could not
provide a similar path, and on these surfaces the P-CH
3
bond was resistant to cleavage.
43
In a subsequent study, Mitchell and coworkers [Tesafai et al., 1998] examined the
decomposition of DMMP on alumina-supported iron oxide. DMMP could react upon
adsorption at room temperature, apparently through cleavage of the phosphorus-carbon
bond. This bond was observed to be extremely resistant to cleavage, however, when
DMMP was adsorbed on oxides such as alumina, magnesia, and lanthana. The
phosphorus-methoxy bonds, which were the most readily cleaved on the other oxides,
appeared, at least initially, to remain intact on the alumina-supported iron oxide. They
proposed an oxidation pathway involving the Fe (II)/Fe (III) redox couple to account for
the unusual activity of the iron oxide surface.
Using a photochemical reactor O’Shea et al. (1997) investigated the photocatalytic
decomposition of DMMP along with diethyl-methylphosphonate (DEMP) in a mixture of
TiO
2
and distilled water saturated with oxygen. The DMMP and DEMP could be readily
degraded by the TiO
2
photocatalyst over a range of concentrations and solution pH. At
low pH (pH <2), the decomposition was observed to be very slow, presumably as a result
of a highly protonated surface, which led to weak adsorption of the simulant onto the
surface of the TiO
2
particles. By increasing the pH, the rate of DMMP disappearance
increased. Final products, indicative of complete mineralization, included phosphoric
acid, formaldehyde, formic acid and carbon dioxide. Mechanistic considerations implied
that hydroxyl radicals were involved in the degradation of DMMP to methylphosphonic
acid. The involvement of additional oxidative pathways was suggested for the
transformation/oxidation of the C-P bond, ultimately leading to phosphoric acid. Based
on the above discussion, the following scheme (Figure 1.8) was suggested for the
photocatalytic oxidation of DMMP:
44

Figure 1.8: Major products from the TiO
2
-catalyzed photolytic degradation of DMMP in aqueous
solutions [O’shea et al., 1997]

Obee and Satyapal (1998) studied the photocatalytic decomposition of gas-phase
DMMP (1- 90 ppm
V
in

90 % N
2
/10 % O
2
) on TiO
2
-coated glass substrates. UV-light and
oxygen were both shown to be necessary for the photocatalytic decomposition of DMMP
on TiO
2
. A carbon balance indicated that only about 20 % of the feed carbon was being
released as gas phase CO and CO
2
. In addition to CO and CO
2
, other products were
methylphosphonic acid, phosphate (PO
4
3-
), methanol and methyl formate (from the
reaction of methanol and formic acid). Catalyst deactivation (after a few hours) was
observed and intensified by increasing the feed gas. One of the interesting results of this
study was that the catalyst activity was completely recovered by washing with water, and
only partially recovered by exposure to UV light.
Segal et al. (1999) studied the gas-phase decomposition of DMMP over amorphous
manganese oxide (AMO) catalysts in a photo reactor in the presence of light (~200 nm-
800 nm). The reaction was studied under oxidizing conditions using air at low
temperatures (40 °C-70 °C). DMMP was found to adsorb strongly onto the AMO surface
and to produce small amounts of MeOH even in the absence of light. When AMO was
irradiated with light of ~200 nm-800 nm, large amounts of MeOH and CO
2
were initially
formed. Like other previous reports on the catalytic oxidation of DMMP, following the
initial period of high activity, strong deactivation was observed. Ion chromatography (IC)
analyses on aqueous extracts from spent AMO indicated that several products accumulate
45
on the AMO surface. These products included MMP and MPA, with greater amounts
being produced after irradiation. IR investigations indicated that DMMP had bonded to
the Mn Lewis acid sites on the AMO surface via the phosphoryl oxygen. On the basis of
these results, the researches proposed the following mechanistic scheme (Figure 1.9) for
the adsorption and photo-assisted decomposition of DMMP over AMO:

Figure 1.9: Scheme showing the DMMP reaction mechanism on AMO surface [Segal et al., 1999]

Rusu and Yates (2000a) studied the photo-oxidation of DMMP on powdered TiO
2

under UV-irradiation (2.1 eV-5.0 eV) in a stainless-steel IR cell capable of operating at
temperatures from 100 K to 1500 K. At 200 K, only the photo-oxidation process took
place, whereas at room temperature the hydrolysis and photo-oxidation occurred together.
An adsorbed PO
3
species was observed to be the final phosphorus-containing photo-
46
oxidation product. Under high coverage conditions, the DMMP was not photo-oxidized
due to site blockage for O
2
adsorption on the TiO
2
surface. DMMP on TiO
2
was stable in
the presence of molecular oxygen, but during photo-oxidation both the methyl groups of
the P-CH
3
and P-O-CH
3
moieties were destroyed at equal rates. The photo-oxidation at
200 K produces mainly CO
2
, CO, and H
2
O, with bidentate formate groups adsorbed on
the surface also detected. The TiO
2
surface also covered with species containing a free
P=O moiety.
Rusu and Yates (2000b) used Fourier transform infrared spectroscopy (FTIR) in order
to gain a better understanding on adsorption and decomposition of DMMP on TiO
2
. The
DMMP was observed to condense on the outer surface of TiO
2
at temperatures lower
than 166 K. As the temperature increased from 166 K to 200 K, the DMMP diffused into
the TiO
2
interior. The DMMP chemisorbed on the TiO
2
through the phosphoryl oxygen to
the surface hydroxyl groups and to the surface Lewis acid sites. For temperatures >214 K,
the dissociation of chemisorbed DMMP was observed initiated at the methoxy groups of
the DMMP molecule and accompanied by the consumption of all of the isolated surface
hydroxyl groups. This produced adsorbed methoxy groups on the TiO
2
and phosphonate
groups exhibiting a lower P-O bond.
Cao and coworkers [CaO et al., 2000; Cao et al., 2001; Segal et al., 2001] investigated
the thermocatalytic decomposition of DMMP on activated carbon (AC) and also Ni, Fe,
Cu, and V catalysts supported on metal oxides including MgO, Al
2
O
3
, SiO
2
, and TiO
2
.
Activated carbon (BET surface area of 897 m
2
/g) was used for the catalytic
decomposition of DMMP at various temperatures ranging from 573 K to 723 K [Cao et
al., 2001] in a fixed-bed reactor (1300 ppm of DMMP a flow of 50 cc/min air). BET
47
measurements indicated that the majority of pores in the AC were micropores, which
transformed into macropores at 723 K in air. BET analysis of the fresh and spent samples
indicated that during the course of the reaction, large amounts of phosphorus species and
coke occupied the pores, resulting in a sharp decrease in pore volume and surface area.
The coke and an unidentified viscous material in the pores of the carbon were washed
with water, and the wash solution was analyzed using IC and shown to contain a large
amount of PO
3
4-
, suggesting that the viscous material was perhaps P
2
O
5
or phosphoric
acid. During the reaction a white vapor was observed to exit from the reactor which was
shown (by IC) to contain small amounts of PO
3
4-
(the authors theorized that most of the
phosphoric acid or P
2
O
5
formed in the reaction had been condensed within the reactor
because of its high sublimation temperature) and MPA. The decomposition of DMMP
was shown to take place in two stages. In the first stage, requiring oxygen supplied from
air, the activated carbon behaved as an initiator, converting DMMP into CO
2
, methanol,
and P
2
O
5
. In the second stage, which was autocatalytic, P
2
O
5
that had accumulated in the
reactor was able to catalyze the decomposition of DMMP. When comparing their catalyst
with 1 wt.% Pt/ γ-Al
2
O
3
under the same conditions, the AC-based material exhibited
better life than Pt-Al
2
O
3
catalyst (26 h protection time for the AC vs. 8.5 h for the 1 wt%
Pt/ γ-Al
2
O
3
sample)
As noted above, the same group [Cao et al., 2000] investigated the thermocatalytic
oxidation of DMMP on nickel, iron, copper, and vanadium oxide supported on γ-Al
2
O
3
.
The vanadium catalyst was found to exhibit exceptional catalytic activity, even better
than the platinum catalysts. Protection times for these catalysts were as follows: 10 %
V/Al
2
O
3
(12.5 h) >1 % Pt/Al
2
O
3
(8.5 h) >10 % Cu/Al
2
O
3
(7.5 h) >Al
2
O
3
(4.0 h) >10 %
48
Fe/Al
2
O
3
(3.5 h) >10 % Ni/Al
2
O
3
(1.5 h) at 673 K. In their studies they investigated
vanadium contents ranging from 1 to 15 wt.%, concluding that the 10 % vanadium
loading on Al
2
O
3
(corresponding to a monolayer dispersion of V
2
O
5
) was optimal. Of the
different supports studied (Al
2
O
3
, SiO
2
, and TiO
2
), SiO
2
was shown to be the optimal
support because of its large surface area and the ability to resist poisoning by P
2
O
5
. For
example, on 10 % V/SiO
2
catalysts, 100 % conversion of DMMP was reached for more
than 100 h at 723 K. The accumulation of the phosphorus species on the catalysts resulted
in substantial surface area (BET) loss for all samples. On the other hand, during the
protection period, an approximately 100 % carbon balance (estimated by the amounts of
MeOH and CO
2
) was obtained only for 1% Pt/Al
2
O
3
compared to all other catalysts
tested. IR, XRD, IC, and XPS indicated that the used vanadium catalysts contained
various phosphorus species and coke. The presence of methylphosphonic acid on the
catalyst surface and downstream of the packed-bed reactor demonstrated the difficulty of
P–CH
3
cleavage. The deposition of coke in the catalyst bed and along the reactor wall
was attributed to the dehydration of methanol and DMMP on P
2
O
5
. However, P
2
O
5
itself
was observed to catalyze the decomposition of DMMP. These authors proposed the
following reaction scheme to explain these experimental observations for vanadium
catalysts:
DMMP + O
2
→ (on catalyst) → MeOH + CO
2
+ P
2
O
5
P
2
O
5
+ Metal oxide → Phosphate
DMMP + O
2
→ (on P
2
O
5
) → MeOH + CO
2
+ Coke + P
2
O
5
MeOH → (on P
2
O
5
) → Coke + CO
2
49
The same group [Segal et al., 2001] also studied the thermal decomposition of DMMP
over amorphous manganese oxide (AMO) and Al
2
O
3
-supported AMO catalysts for
temperatures between 200
o
C and 400
o
C, the highest reaction rate occurring at 400
o
C.
DMMP was found to oxidatively decompose over AMO and AMO/Al
2
O
3
, the highest
activity observed with a catalyst prepared by the precipitation of AMO on Al
2
O
3
. During
the initial stages of reaction, DMMP was completely removed from the gas phase,
producing only CO
2
, with no other gas phase products. After a certain period of time
(ranging from 5 min–8 h), DMMP reappeared in the gas phase. Simultaneously, the CO
2

concentration decreased and MeOH began to form, indicative of DMMP hydrolysis.
Segal et al. (2001) indicated that catalyst deactivation was due to adsorbed P-species.
FTIR spectroscopy and IC were used to examine adsorbed products on the surface of the
catalysts. IC analysis indicated that several products accumulate on the surface, including
methyl-methylphosphonate, methylphosphonic acid, and phosphoric acid. FTIR analysis
showed that DMMP had bonded strongly to Mn Lewis acid sites on the AMO surface via
phosphoryl oxygen. The bare Al
2
O
3
support was also studied in DMMP decomposition
reactions and showed high activity, with 100% DMMP removal from the gas stream for
over 15 h. The major products observed over Al
2
O
3
were dimethyl ether and MeOH. No
CO
2
was observed, indicating that DMMP was not oxidized over Al
2
O
3
. GC, IC, and
FTIR studies suggest that DMMP was dissociatively adsorbed over Al
2
O
3
.
Deckers et al. (2002) studied the catalyzed destruction of various environmental toxins
(including DMMP) with solid nanocrystalline oxides (AP-CaO) and microcrystalline
CaO (CP-CaO) in the temperature range of 200 °C-500 °C. Approximately 50 mg of CaO
(8.92 x 10
-4
mol) was used per experiment. They studied DMMP decomposition over AP-
50
CaO and CP-CaO as well as CaO coated with iron ([Fe
2
O
3
]-AP-CaO) at three
temperatures (200, 400, and 500 °C) -- see Table 1.5 (top). It is clear that temperature
plays a very important role in the amount of DMMP decomposed, regardless of the type
of sample. The decomposition of DMMP rose from essentially 0 % at 400 °C to 70-78 %
at 500 °C, and there was a 6-fold increase in the breakthrough time. In addition to Fe
2
O
3

they also studied the effect of various other [M
x
O
y
] coatings -- see Table 1.5 (bottom).
Unfortunately, the presence of transition-metal oxide did not seem to have a significant
effect on the decomposition of DMMP. This was contrary to the results for [Fe
2
O
3
]-MgO
where there was a very definite activity enhancement due to the presence of [Fe
2
O
3
].

Table 1.5: Destructive adsorption of DMMP with [M
x
O
y
] CP/AP-CaO at 500
o
C [Decker et al., 2002]

51
The adsorption and decomposition reactions of DMMP (and of the products produced)
on a commercial -Al
2
O
3
, a -Al
2
O
3
-supported iron oxide, and a sol-gel-prepared alumina
have been studied at temperatures from 25 °C to 400 °C by Sheinker and Mitchell (2002).
They utilized a microreactor connected to a two-stage bubbler containing liquid DMMP
through which He gas was bubbled at a flow rate of 30 ml/min. IR was used for the
analysis of the products. For all three materials studied a decomposition pathway was
active at temperatures as low as 25 °C, whose rate increased with temperature. The
reaction was poisoned by DMMP fragments and/or molecular DMMP, making the
reaction stoichiometric, and not catalytic, at these temperatures. Only ~10 % of the
DMMP that was adsorbed could be decomposed via this pathway at 25 °C. Increasing
temperature increased the decomposition activity because of higher surface reactivity and
less poisoning by molecular DMMP. This decomposition pathway was augmented at
temperatures in excess of 200 °C by a reaction that involved the continuous
decomposition of DMMP. Sheinker and Mitchell (2002) suggested that this second
reaction might involve DMMP fragments, mainly methyl-methylphosphonate, directly in
the continued decomposition. At 200 °C, the sol-gel Al
2
O
3
surface decomposed every
DMMP molecule up to the break through point (BTP) to yield, on average, one methanol
molecule and one adsorbed fragment. At 300 °C, the BTP was 40 % higher than at 200
°C, and every DMMP molecule (up to the BTP) decomposed to yield two methanol
molecules, or the equivalent, and an adsorbed fragment. At temperatures in excess of 300
°C, additional products (CO, CO
2
, and CH
4
) were observed. Sheinker and Mitchell
(2002) believed that these species may be generated from the initial decomposition
products of DMMP (methanol or DME), rather than directly from DMMP. The aluminas
52
yielded higher amounts of decomposition products than the supported iron oxide
material, with the sol-gel alumina, in particular, showing very high activity presumably
because of the presence of transitional phases that yielded more reactive surface sites. At
25 °C, the commercial -Al
2
O
3
showed a total decomposition capacity of 117 μmol/g, the
alumina-supported iron oxide material a capacity of 93 μmol/g (when corrected for
surface area; however, the supported iron oxide material showed an activity equal to that
of the -Al
2
O
3
support), and the sol-gel alumina a capacity of 208 μmol/g. At 100 °C,
these capacities increased by about a factor of 3, and at 200 °C and above, all of the
materials showed some capacity for sustained DMMP decomposition.
Mitchell et al. (2003) also studied the adsorption and decomposition reactions of
DMMP on cerium and iron oxides supported on alumina at 25 °C using the same reactor
set-up. The co-impregnated oxide formulations were significantly more reactive than the
alumina alone, and were 2.5 times more reactive at room temperature than any other
metal oxide studied previously. The most active formulation was one containing 5 wt.%
iron and 7.5 wt.% cerium. At 25 °C, Al
2
O
3
showed a decomposition capacity of 317
μmol/g, while the alumina-supported iron and cerium oxide combinations showed a
decomposition capacity of more than 510 μmol/g. Iron or cerium oxides individually
were more active than the unmodified alumina but less active than the co-impregnated
materials. Their study showed that when cerium oxide was impregnated on alumina by
itself two CeO
2
phases form, a three-dimensional CeO
2
phase that was inactive for
decomposition and a two-dimensional cerium oxide phase that could be responsible for
the increased reactivity of these materials. The increased activity of materials that
53
included iron was thought to be due to iron either increasing the number of defect sites in
the ceria crystallites that form, or facilitating the formation of smaller crystallites.
The products from the study of the alumina substrate alone are shown in Figure 1.10.
The right chart shows the total accumulated flow of volatile carbon species produced by
the reaction (“All” is equal to two times the flow of DME plus the flow of MeOH). As
can be seen in the left chart, DME was the first product generated. Product evolution is
characterized by an induction period, a rapid rise in product flow up to a maximum,
followed by a slower decrease in flow rate. Product formation continues for a relatively
long period at a low rate until it is no longer possible to determine the product
concentration above the noise level. Mitchell et al. (2003) theorized that the induction
period was due, at least in part, to the adsorption of MeOH and DME onto the alumina
surface. The DMMP breakthrough point for the solids appears to also be predominantly a
function of the surface area. Figure 1.11 shows the rate of product formation and the
accumulated products formed for the alumina itself and for three impregnated alumina
formulations, one containing 5 wt.% Fe, one containing 7.5 wt.% Ce, and one containing
5 wt.% Fe and 7.5 wt.% Ce prepared using simultaneous impregnation. The mole ratio,
Fe:Ce, for the 5 wt.% Fe/7.5 wt.% Ce sample was 1.67:1. Overall, all three impregnated
materials show increased product flow and total product yield compared to the
unimpregnated alumina.

54

Figure 1.10: Results from the decomposition of DMMP on γ-alumina at 25 °C. The left figure shows
the flow rate of products as µmol C1/min. “µmol of C1” means the equivalent number of one-carbon
products, that is two times the number of µmol /min (or µmol) of DME plus the number of µmol /min (or
µmol) of MeOH. The right figure shows the accumulated reaction products that have been measured in the
gas phase. “All” represents the µmol of C1 products observed [Mitchell et al., 2003].

Figure 1.11: (Left) Product flow curves from the decomposition of DMMP at 25 °C on Al
2
O
3
, 5 wt.%
Fe on Al
2
O
3
, 7.5 wt.% Ce on Al
2
O
3
, and 5 wt.% Fe/7.5 wt.% Ce on Al
2
O
3
. The curves are referenced to the
same vertical scale, except that they have been offset to aid viewing. The initial baseline of each curve
corresponds to its zero on the vertical scale. (Right) Accumulated product formation curves from the
adsorption of DMMP at 25 °C on Al
2
O
3
, 5 wt.% Fe on Al
2
O
3
, 7.5 wt.% Ce on Al
2
O
3
, and 5 wt.% Fe/7.5
wt.% Ce on Al
2
O
3
[Mitchell et al., 2003].

55
Mitchell et al. (2004) also studied the MeOH and DME production during the
adsorption and decomposition reactions of DMMP on cerium oxide supported on
aluminum oxide at 25 °C. Experiments were carried out that involved dosing the reactive
adsorbent with small doses of DMMP followed by quantitative determination of the
decomposition products. The results suggested that the formation reactions of methanol
and dimethyl ether were competitive processes involving the same surface intermediate,
which was most likely a surface methoxy species. The formation of DME was proposed
to be due to the combination of two surface methoxy groups (Figure 1.12), while another
important reaction producing MeOH involved a surface methoxy group interacting with a
vapor phase or physisorbed DMMP molecule (Figure 1.13).

Figure 1.12: Proposed scheme for the production of DME during reaction of DMMP on alumina
supported cerium oxide [Mitchell et al., 2004]

Figure 1.13: Proposed scheme for the production of MeOH during reaction of DMMP on alumina
supported cerium oxide [Mitchell et al., 2004]
56
Diffuse reflectance spectroscopy indicated the presence of significant amounts of
methoxy fragments formed upon DMMP adsorption primarily associated with the cerium
oxide domains. The alumina-supported cerium oxide reactive adsorbents developed in
their study were the most active among all those reported in the literature for ambient
temperature applications, decomposing approximately 775 μmol of DMMP per gr of
adsorbent at 25 °C, and strongly or irreversibly adsorbing an additional 400 μmol, for a
total capacity at room temperature of 1.1-1.2 mmol of DMMP/gr.
Trubitsyn and Vorontsov (2005) studied DMMP decomposition in air over a high
surface area anatase at 300 K in a thermostated batch reactor by FTIR. Complete removal
of injected and evaporated DMMP was obtained for only low quantities. However, an
increase in humidity from 1 to 50 % significantly improved the conversion. Two types of
experiments were carried out. In the first experiment, DMMP underwent reactive
adsorption (no UV-irradiation) followed injection of different amounts of DMMP, which
upon completion was followed by photocatalytic oxidation under a mercury lamp
generated UV-irradiation. Photocatalytic oxidation after injection of low DMMP quantities
proceeded completely, and it was an order of magnitude faster than oxidation after injection
of large DMMP quantities. In the second experiment, the DMMP underwent adsorption
and photooxidation at the same time. The fastest removal of DMMP from the gas phase
and completion of oxidation in 30 min was observed during the simultaneous adsorption
and photocatalytic oxidation, suggesting that the process might be practical for air
decontamination. However, the rate of complete oxidation was in the second case lower
than that in the first one. This was attributed to insufficient quantities of adsorbed DMMP,
which was consumed in oxidative reactions before accumulation on the TiO
2
surface. For
57
the first type of experiment, DMMP was captured on the TiO
2
surface at a rate equal to that
of external diffusion to the surface. Hydrolysis of the adsorbed DMMP led to the formation
of MeOH and MMP. On the other hand, photocatalytic oxidation generated CO
2
as the only
carbonaceous gaseous product and bidentate formates as the intermediate surface product.
At high concentrations, DMMP was captured incompletely and hydrolyzed only partially,
with CH
3
OH being the only gas phase species. Photocatalytic oxidation generated gaseous
HCOOH, CO, and CO
2
and was incomplete due to catalyst deactivation by nonvolatile
products. From the highest quantity of produced methanol, it was estimated that 0.68 µmol
is the maximum quantity of DMMP that could be hydrolyzed over 1 mg of TiO
2
, which
corresponds to a 0.8 nm
2
surface area per each adsorbed DMMP molecule. Therefore, at
least 0.8 nm
2
of TiO
2
surface per each DMMP molecule should be available for complete
purification of air.
The adsorption and degradation of DMMP over UV-irradiated TiO
2
powders and thin
films was also investigated by Moss et al. (2005). Adsorption of vapor-phase DMMP on
TiO
2
powder was characterized by diffuse reflectance infrared Fourier transform
spectroscopy (DRIFTS). Photochemically-assisted oxidation of adsorbed DMMP was
carried out in situ by UV-irradiation of the samples in the DRIFTS cell with a xenon
lamp, in order to generate kinetic data and information on the specific site binding of
DMMP and on catalyst poisoning. Gas-phase intermediates from the vapor phase reaction
in the static cell were identified by GC-MS, and surface-bound intermediates and
products were analyzed by HPLC-MS, and IC of both aqueous and organic extractions
from the TiO
2
(Figures 1.14, and 1.15).

58

Figure 1.14: Surface-bound product concentration in static batch reactor with lightly loaded DMMP
[Moss et al., 2005]

Figure 1.15: GC peak area and concentration versus time data for gas phase products in bulk static
reactor degradation experiments: (a) heavily loaded (30 min exposure to DMMP aerosol from a Collision
nebulizer), and (b) lightly loaded (30 min exposure to flowing DMMP vapor at room temperature)
[Moss et al., 2005]

Adsorbed DMMP was photo-degraded in a stepwise fashion to give methylphosphonic
acid, PO
4
3-
, H
2
O, and CO
2
as products. Moss et al. (2005) proposed a reaction pathway
(Figure 1.16) which they claimed was consistent with the rapid degradation of DMMP as
59
well as with the extensive poisoning of the catalyst by surface-bound phosphonate
products.

Figure 1.16: Proposed reaction pathway for the photodegradation and adsorption of DMMP on TiO
2

[Moss et al., 2005]

Mitchell et al. (2007) studied the reaction of DMMP with ozone at room temperature
on an alumina-supported iron oxide reactive adsorbent placed in a U-tube reactor with
helium as the carrier gas. The surface oxidation reaction was found to involve a surface
adsorbed and activated DMMP molecule (or a fragment) on a Lewis acid site, and an
active oxygen surface species formed from ozone. Carbon dioxide was the primary
reaction product, and the amount formed was found to plateau as a function of ozone
concentration after a concentration of approximately 300-400 ppm, with higher
concentrations of ozone having no additional effect on the surface reaction. Their
findings indicate, that the surface reaction with ozone on the 10 wt.% Fe
2
O
3
/Al
2
O
3

60
adsorbent generated approximately 2.7 times as much gas-phase carbon (as CO
2
and CO),
as does the same surface reaction in the absence of ozone, in the form of methanol.
During the initial stages of DMMP adsorption and reaction, after exposure to 100 µmol
DMMP, the conversion was found to be 12 %, based on the assumption that one
molecule of CO
2
indicated the decomposition of one molecule of DMMP. After exposure
to 300 µmol of DMMP, the conversion dropped to 6 %. The ozone appeared to increase
the number of active sites for DMMP decomposition by liberating some of the reaction
products from the surface and recycling the surface sites. Supported iron oxide seemed to
be a more effective reactive surface than alumina for the reaction with ozone, primarily
because of its greater facility for ozone decomposition.
Han et al. (2008) prepared and studied an acid photocatalyst (TiO
2
modified with
H
2
SO
4
) for the decontamination of pure (without air) CWA simulants 2-CEES (2-
chloroethyl ethyl sulfide) and DMMP in the vapor phase in a fixed-bed. The initial
concentrations of 2-CEES and DMMP were 85 and 210 ppm, respectively. Compared
with the unmodified pure TiO
2
photocatalyst, the adsorption capacity on the sulfated
material increased by 50 %, and the conversion for 2-CEES increased more than 20 %. It
was also found that the decontaminating capacity for DMMP was increased by 200 min
with the modified photocatalyst. The surface characteristics of the photocatalysts were
analyzed by XPS and FTIR, showing more Lewis and Bronsted acid sites on the sulfated
TiO
2
, thought to be responsible for improving the performance of the photocatalyst. The
volatile products were mainly CO
2
, CH
3
OH, HCOOH, HCHO and non-volatile surface
product is MPA. Since no P-containing compounds were found in the gas phase, this
61
suggested that all P atoms were deposited on TiO
2
surface, leading to the catalyst
deactivation.
Panayotov and Morris (2008) used transmission IR to investigate the reaction
pathways of DMMP on nanoparticulate Au/TiO
2
and pure TiO
2
in a high-vacuum IR cell.
DMMP oxidatively decomposed over the Au nanoparticles supported on TiO
2
. Moreover,
unlike traditional photocatalytic approaches for decomposing CWA on titania, the
Au/TiO
2
system did not require UV radiation. Partial decomposition of DMMP occurred
on both samples through hydrolysis reactions involving trace surface OH groups of TiO
2
.
The primary products of the hydrolysis were methanol and/or surface methoxy species.
The infrared spectral signature of adsorbed CO, one major gas phase reaction product,
was used to track the oxidation state of Au and Ti during the reaction. As a result, it was
found that small (<5 nm) Au particles on the surface of TiO
2
could activate this material
toward the oxidative degradation of DMMP, in contrast to pure TiO
2
, which was an
effective sorbent, but showed no ability to decompose the DMMP. When the DMMP
adsorbed on Au/TiO
2
was exposed to molecular O
2
at 475 K, intense reaction took place
to form CO
2
and various intermediate products of partially oxidized organic species.
Panayotov and Morris (2008) proposed the following mechanistic scheme to explain the
reaction pathway for catalytic degradation of DMMP over TiO
2
-supported nanoparticles
(Figure 1.17):

62

Figure 1.17: Proposed reaction pathway for catalytic-degradation and adsorption of DMMP on TiO
2

supported Au-nanoparticles [Panayotov and Morris, 2008]

Panayotov and Morris (2009a) also studied the thermal decomposition of DMMP on
high surface area TiO
2
nanoparticles (50 m
2
/g). The molecular adsorption of DMMP
occurred through hydrogen bonding to trace amounts of isolated surface hydroxyl groups,
while reaction occurs on Ti
4+
Lewis acid sites. Highly reactive surface oxygen, that was
present at locations adjacent to the Ti
4+
adsorption sites, oxidized DMMP to produce
carbonyl-containing surface adsorbates, which were short-lived and rapidly converted to
carboxylate or formate products. The dominant reaction pathway in the low-temperature
regime (295 K to 400 K) is the nucleophilic attack of adsorbed DMMP by neighboring
oxygen to produce Ti-OCH
3
and a variety of P-O
x
surface bound groups. Above 400 K,
thermally activated lattice oxygen begins to play a dominant role in driving the oxidation
of surface Ti-OCH
3
groups by attacking them. In their most recent paper Panayotov and
Morris (2009b) continued their study of the reaction pathways of DMMP on a
commercial nanoparticulate ( ⊁20 nm) titania. Initial DMMP uptake occurred through
both molecular and reactive adsorption. The molecular adsorption was driven by
hydrogen-bond formation to isolated hydroxyl groups, while the reactive sorption
63
appeared to occur through interaction with both Lewis acid sites and active oxygen
species present on the initial TiO
2
surface to produce MeOH accompanied by the
emergence of new carbonyl-containing moieties. However, the reactive sites were
quickly poisoned by oxidation products, including a strongly bound, nonvolatile
phosphorus compound. The presence of this highly stable phosphorous compound was
confirmed by XPS and FTIR. Thermal reactivation of the TiO
2
in oxygen seemed to be
able to restore the physisorption capacity of the particles toward the DMMP, but the
reactive adsorption pathway was nearly completely eliminated. Based on the above
findings, the authors suggested that the catalytic activity and poisoning durability of TiO
2

could be further improved by strengthening the acid/base properties of the catalytic
material in order to ensure the dissociative adsorption of DMMP. To attain this goal, they
proposed the use of high surface area rutile nanoparticles as well as doping of the TiO
2

with metal nanoparticles to promote the Lewis acidity of the catalyst material.
Mattsson et al. (2009) investigated the photo-degradation of DMMP and CEES in air
on Zr-doped titania nanoparticles (Zr content varied from 0 to 15 wt.%) at room
temperature in a DRIFTS/FTIR cell. All samples, other than the non-crystalline TiZr
5

(amorphous sample) exhibited significant photoactivity for the decomposition of CEES
and DMMP. But, among these samples the most efficient for the photodegradation of
CEES and DMMP was TiZr
2
with a fairly low Zr concentration (6.8 wt.%). Intermediate
decomposition products such as sulphur, carbonate and phosphorous complexes,
carboxylates, aldehydes, and acetates could all be identified on the samples after UV
irradiation.
64
Ratliff et al. (2009) studied the decomposition of DMMP using TPD, XPS, and AES
on TiO
2
-supported Pt, Au, and Au-Pt clusters, as well as on the TiO
2
itself. Their results
revealed that despite the higher activity of the Pt and Au-Pt clusters for DMMP
decomposition, the TiO
2
support exhibited more promising catalytic activity because of
its ability to sustain activity over multiple cycles of adsorption and reaction. In fact, for
TiO
2
active sites for DMMP decomposition were blocked after a single cycle, but some
activity for methyl production was sustained even after five cycles. According to Ratliff
et al. (2009), C-H bond scission on the titania surface was less facile than on Pt, but C-O
bond scission occurred readily. Furthermore, phosphorus could be removed from TiO
2
by
heating at 965 K under UHV or at 800 K in the presence of O
2
, restoring its activity for
DMMP decomposition. The Pt clusters had higher initial activity for DMMP reaction
compared to TiO
2
, but the activity was suppressed after a single adsorption-reaction cycle
due to encapsulation of the Pt clusters by reduced titania upon heating, as well as
poisoning of the active sites by P-containing by-products. Phosphorus removal from the
Pt clusters by high temperature oxidation was complicated by the fact that heating in O
2

accelerates the encapsulation of the Pt clusters by TiO
2
and the encapsulated clusters
exhibited activity characteristic of TiO
2
and not of Pt. The addition of Au to the Pt
clusters was found to diminish the overall activity while allowing phosphorus to desorb
from the surface at 900 K. However, significant cluster sintering, Au clusters desorption,
and encapsulation of the Au-Pt clusters occurred at 900 K, and therefore the
morphologies of the clusters were dramatically changed after the phosphorus removal.
The activity of the Au-Pt bimetallic clusters (in terms of DMMP decomposition to CO
and H
2
) was found to be intermediate between that of pure Pt and pure Au. The
65
production of new products was not observed, but hydrogenation of methyl to methane is
promoted on Au-Pt sites.
Mera et al. (2010) studied the use of TiO
2
immobilized on a glass plate for the
effective decontamination of air contaminated by Sarin gas (they used DMMP as a
simulant for Sarin). In order to prevent DMMP adsorption during their experiments, they
employed silane to coat the interior part of their Pyrex-glass reactor. High activity of
TiO
2
(0.01 g) was observed under UV irradiation and DMMP (33.5 µM) was removed.
Mera et al. (2010) concluded that the photocatalytic treatment was very effective for the
removal of DMMP. CO, CO
2
, and H
2
O, as well as methanol, formaldehyde, formic acid,
and methyl formate were detected as the primary products in the gas phase, and no
poisonous substances were generated. They performed photocatalytic decomposition by
repeated-batch reactions in order to study the long-term performance. The experiments
were carried out by introducing vaporized DMMP repeatedly until photocatalytic activity
and/or adsorption of TiO
2
-glass plate against DMMP were lost. Though DMMP removal
stayed at a high level after three cycles, at the end of the seventh cycle, the TiO
2
plate lost
its capability for photocatalytic degradation completely. During the repeated-batch
reactions with 0.1 g active TiO
2
, the total amount of DMMP removed reached 968.5 µM
(from a total of 1131.5 µM fed to the reactor) by both photocatalytic decomposition and
strong adsorption of TiO
2
. Mera et al. (2010) suggested that the photocatalytic and
adsorptive performances of TiO
2
would make it an ideal material for the removal of large
amounts of Sarin on a practical scale.
Chen et al. (2010) reported experiments on the thermal decomposition of DMMP on
crystalline ceria thin films grown on Ru (0, 0, 0, 1). They used TPD, XPS, and the IRAS
66
methods to analyze their observations. Their results showed that DMMP had readily
decomposed on crystalline ceria thin films upon heating. The main gaseous products
were formaldehyde, methanol, water and CO, with PO
x
remaining on the surface after
heating to 900 K. Ceria films were capable of sustaining decomposition activity after
multiple cycles of adsorption and reaction, but the activity significantly diminishes with
each cycle. Product yields did not decrease substantially over the first three cycles, but
there was a sharp drop between cycles three and six. Yields did not decrease to zero after
the seventh cycle indicating that the surface still sustained some activity for DMMP
decomposition. The loss of activity was attributed to passivation of the active sites by
PO
x
, and to the reduction of Ce
+4
to Ce
+3
. Even after reoxidation of the ceria, the activity
was not restored, indicating that DMMP decomposition on ceria was not catalytic. At 100
K, DMMP initially adsorbed to the surface via the phosphoryl oxygen, but the P=O bond
converted to a bridging O-P-O species at 200 K. Upon further heating, DMMP
decomposed via P-OCH
3
bond scission, forming MMP and MP surface intermediates as
well as surface methoxy species. The more stable P-CH
3
bond was not cleaved until
temperatures above 700 K. The authors pointed out that the similarities between DMMP
and methanol chemistry on the ceria films suggesting that the methoxy species as a key
surface intermediate in both reactions.
In nature, bulk quantities of water will typically be present when chemical agents
come in contact with environmental substrates such as a-SiO
2
. Given that chemistry is
often facilitated in solution relative to vacuum, the presence of bulk water could
dramatically affect the fate of the agent by facilitating chemical reactions not captured
with a simple surface/vacuum-interface model. To examine this effect, Quenneville et al.
67
(2010) performed molecular dynamics (MD) computer simulations to investigate the
interaction of DMMP with hydrated amorphous silica surfaces both in the presence and
in the absence of water. The nature of the adsorption of DMMP on silica was found to
depend strongly on the hydroxyl coverage of the silica surface. At sufficiently high
coverage, ⊁37 % of the DMMP molecules hydrogen-bond to the surface. Also, their
results showed as the surface hydration decreased, the surface defect density increased,
and the reactivity of DMMP increased along with it. According to their findings,
although rare, when DMMP does react with the silica surface, DMMP can covalently
attach to the surface and, less frequently, fragment to yield adsorbed methyl species.
Their reactive MD simulations with the ReaxFF force field identified a number of
chemical reactions important in the interaction of DMMP with hydrated amorphous
silica. Also, simulations were performed to examine the fate of DMMP molecules which
had been pre-adsorbed on the silica surface to the addition of bulk quantities of water.
These simulations, predict that adsorbed DMMP can react with water to yield solvated
methanol species.
Finally, as a part of a long-term research on sustained room temperature
decomposition of DMMP on different metal oxides, Mitchell et al. (2011) examined
alumina-supported manganese oxide solids for adsorption/destruction of DMMP at room
temperature under the presence of ozone (O
3
) as a way of increasing the chemical energy
available to the reaction system and potentially increasing the yield of the decomposition
reaction. For the experimental run, they placed 63 mg of the solid sample placed in a U-
tube micro-reactor and after pretreatment in O
2
at 350
o
C the sample was cooled to 25
o
C
in flowing He. The gas mixture containing DMMP in helium was flowing at a rate of 30
68
sccm, with a DMMP concentration in the stream of 47.0 μmol/L, giving a DMMP flow
rate of 1.4 μmol/min. The DMMP/He flow was mixed with an appropriate amount of the
O
3
/O
2
mixture before sending to the micro-reactor. The total flow of gas through the
system was typically 36 sccm. After the U-tube reactor, the gas stream flowed to a long-
path gas cell for detection and the determination of gas-phase products in an infrared cell.
The results of the experiments showed in the absence of ozone, DMMP decomposes
on alumina-supported manganese oxide to yield methanol and dimethyl ether. With
ozone being present, the amount of CO produced is virtually identical to that of CO
2
and
trace amounts of methanol are observed even in the presence of high concentrations of
ozone. Also, DMMP breakthrough, representing saturation of the DMMP adsorption
sites, roughly coincides with the end of the methanol flow, as is seen in experiments
carried out with oxygen as the only oxidant. So, these researchers concluded that each
DMMP molecule appears to produce, on average, one molecule of CO
2
and one of CO,
and a likely phosphorous-containing product of the (DMMP + O
3
) reaction. They report
the observations of a surface species H
3
C-P(-O-surf)
3
, with two of the oxygen atoms
provided by the surface lattice, although this is not the only surface species present, and
inorganic phosphorus species may also be present. Based on their observations, they
suggested the following mechanism for the total decomposition of DMMP on the surface:

Figure 1.18: Proposed mechanism for degradation of DMMP on MnO
x
/Al
2
O
3
[Mitchell et al., 2011]
69
1.4. Conclusions
The above comprehensive literature review highlights some important issues in
regards to the decomposition reaction of DMMP. Almost in all studied cases addition of
metal catalysts has been shown to enhance the rate of DMMP decomposition, and some
precious metals like Pt and V have been reported to be more potent in this regard
compared to others. On the other hand, catalyst poisoning during the DMMP
decomposition has been noticed to be almost inevitable, and the decomposition capacity
of every catalyst has been found to be limited to the available active sites of that catalyst.
This may help one to understand why supported catalyst having more surface area have
shown better durability and protection period during the decomposition of DMMP.
Formation and deposition of phosphorous compounds on solid surface and/or production
of P
2
O
5
/H
3
PO
4
and their subsequent reactions with catalyst and/or support materials have
been found to be the main reason for this catalyst deactivation. Also, temperature has
been observed to have a positive effect on the decomposition rate of DMMP on active
metal sites, i.e., by increasing the temperature the rate of DMMP has been increased in
studied cases.
Also after this literature review it is clear that, other than our own preliminary
literature report [Tsotsis et al., 2008], there is no other published study of DMMP
destruction using the FTCMR concept. In our present research we will therefore focus on
the following topics:
1- Understanding the fundamental phenomena underpinning the FTCMR operation for
DMMP (and VOC in general) conversion. As the literature survey, so far indicates,
70
though a limited number of prior studies have appeared, a complete systematic study of
such reactors is currently lacking.
2- Identifying the advantages and disadvantages of the concept compared to its
competitors.
3- Investigating the major limitations, if any, which exist in applying this concept for
complete oxidation of chemical warfare simulants, and finding the optimum operating
conditions to achieve the best performance of this type of reactors for the proposed
application.
4- Offering a conclusive decision about the possibility of application of this concept as
an efficient catalytic reactor to achieve complete oxidation of CW agents at trace levels,
low amounts of catalyst, and at low operating temperatures.
As it can be concluded from the above discussion, there are two major aspects of our
research:
- Preparing an efficient FTCMR
- Investigating the reaction(s) occurring during the oxidation of DMMP inside the
pores of this FTCMR
Therefore, the next Chapter is devoted to preparation and characterization of a
FTCMR followed by another Chapter dedicated to the study of the chemical reactions
that occur in such reactors.

71
Chapter 2: Catalytic Membrane Studies

2.1. Introduction
In general, a ceramic membrane can be thought of as a permselective barrier or a fine
sieve. The permeability and separation factor are the two most important performance
indicators of a ceramic membrane. For a porous ceramic membrane, they are typically
governed by the thickness, pore size and surface porosity of the membrane, while for a
dense ceramic membrane, the principle for permeation and separation is more complex [Li,
2007]. Table 2.1 shows a classification of asymmetric ceramic membranes and their
various uses depending on their structure.
Ceramic membranes are usually composite materials consisting of several layers of
one or more different ceramic materials. They generally have a macroporous support, one
or two mesoporous intermediate layers and a microporous (or a dense) top layer. As
shown in Figure 2.1, the bottom layer provides mechanical support, while the middle
layers bridge the pore size differences between the support layer and the top layer where
the actual separation takes place [Li, 2007]. Commonly used materials for ceramic
membranes are Al
2
O
3
, TiO
2
, ZrO
2
, SiO
2
, etc. or a combination of these materials.
Ceramic membranes have an advantage over their polymeric counterparts for applications
at high temperature, pressure, and in aggressive environments. Ceramic membranes also
offer an additional advantage in that they can be cleaned at high temperatures or by the
use of steam (sterilization). The first approach enables their extended use in household
applications, while the latter is of importance in food processing and medical applications
[Biesheuvel, 2000].
72
Oxidation reactions in general require high temperatures. Furthermore, according to
the comprehensive literature review presented in Chapter 1, one may expect the presence
of one or more acids (e.g., phosphoric and methylphosphonic acids) produced during the
DMMP decomposition over catalytic materials. Therefore, for this research project we
use ceramic membranes which are potentially resistant to these conditions. It should be
noted once more that in a FTCMR there is no need for the membrane utilized to be
permselective. So, in this research we mostly utilize mesoporous and macroporous
ceramic membranes.

Table 2.1: Classes of ceramic membranes [Li, 2007]

a
:IUPAC classification

Figure 2.1: Schematic representation of an asymmetric composite membrane [Li, 2007]
73
2.2. Membrane Studies
The ceramic membranes for our research are provided by our industrial collaborator in
this project, namely Media and Process Technology, Inc. (M&P). There are four different
tubular membranes available for our research, all of which are prepared using an α-Al
2
O
3

support substrate. Some physical properties of these membranes are presented in Tables
2.2 and 2.3. Comparing the data provided by M&P with the membrane classification
given in Table 2.1, all the membranes (other than the support tubes themselves – referred
to as membrane S) can be categorized as “mesoporous” materials.
As stated above, in multilayer membrane configurations, the macroporous substrate
acts as a mechanical support for the intermediate layer in order to provide strength to the
final membrane structure. The intermediate layer, in turn, serves the purpose to mask the
surface irregularities of the substrate, so that an ultra-thin, high-flux surface layer can be
deposited on it. In the following sections the different methods we have used for the
characterization of these membranes and the data generated, will be described.

Table 2.2: Various M&P ceramic membranes used in this research
Membrane module Support material First separation layer Second separation layer
S-type α-Al
2
O
3
- -
A-type α-Al
2
O
3
α-Al
2
O
3
-
B-type α-Al
2
O
3
α-Al
2
O
3
γ-Al
2
O
3

C-type α-Al
2
O
3
α-Al
2
O
3
γ-Al
2
O
3

74
Table 2.3: Some physical properties of M&P ceramic membranes used in this research, as reported
by M&P

Membrane module Support
thickness (µm)/
Pore diameter (Å) / Porosity (%)
First separation layer
thickness (µm)/
Pore diameter (Å) / Porosity (%)
Second separation layer
thickness (µm)/
Pore diameter (Å) / Porosity (%)
S-type 1100 /2000-4000 /20-25 - -
A-type 1100 /2000-4000 /20-25 10 – 20 /500 /N.A -
B-type 1100 /2000-4000 /20-25 10 – 20 /500 /N.A 2 – 3 µm /100 Å/25-35
C-type 1100 /2000-4000 /20-25 10 – 20 /500 /N.A 2 – 3 µm /40 Å/N.A

2.2.1. Estimation of the structural properties
Assuming that the average pore diameter (d) of the membrane can be measured, some
of the other structural membrane properties can be estimated if one makes appropriate
simplifying assumptions about the membrane pore structure. For example, if one was to
assume that the membrane consists of spherical particles stacked regularly on the top of
each other or by cylindrical particles touching each other and aligned perpendicularly to
the membrane surface (Figure 2.2), then the size of these structural building blocks, i.e.,
the diameter (D) can be estimated by the following simple equation:
) 1 2 /(  
p
d D (2.1)

D/ 2
d p/ 2

Figure 2.2: Simple representation of a membrane structural block

75
Using the value obtained by the Equation 2.1, and the reported average diameters by
M&P, the porosity and surface area of M&P membranes have been calculated and the
results are shown in Table 2.4. A comparison between these results and M&P’s reported
data reveals that the attached, uniform size parallel cylinders model can predict the
porosity of the support material fairly well, while for separation layers its predictions are
rather poor. On the other hand, the uniformly-stacked spheres model can predict the
structural properties of the separation layers much better than the other model, but its
prediction for the porosity of the support material is rather poor.

Table 2.4: Structural properties estimated by simple pore structure model of Figure 2.2
D, (Å)
Porosity, (%)
[Sphere]
Porosity, (%)
[Cylinder]
Surface area, (m
2
/g)
[Sphere]
Surface area, (m
2
/g)
[Cylinder]
Support 4828 - 9657 48 21 1.55 – 3.14
a
1.03 – 2.07
a

First
separation layer
1207 48 21 12.45 8.30
Second
separation layer
241 48 21 62.28
b
41.52
b

a
BET value is 0.3 - 0.4 m
2
/g (refer to section 2.2.5)
b
reported value by the vendor is 50 – 100 m
2
/g

2.2.2. SEM analysis
In order to have a more in-depth look into the structure of the membrane samples the
scanning electron microscopy (SEM) method was utilized, using a JEOL JSM-7001F
SEM/EDX instrument which has the ability to magnify the samples by up to ~ 50,000
times. The attention was focused primarily on the top layers and their structural
uniformity as these are the places in the membrane where most of the catalyst crystallites
are deposited. Two types of characterization experiments were carried out:
- A small (~10 cm) piece was cut from a longer membrane tube (~ 100 cm in length)
that was supplied (in 2010) by M&P. This smaller piece of membrane was cut, in
76
turn, into 10 smaller sub-pieces, and a ~ 4 mm portion of these sub-pieces
(appropriate for SEM) was then analyzed. The goal of this test was to examine the
uniformity of thickness and other structural properties along the membrane length.
- Different tubes were studied from different membrane batches provided by M&P
from 2007 to 2010. Then, a small piece (~ 4 mm) from each membrane was used
for SEM analysis.
The data obtained from the SEM analysis are presented in Figures 2.3-2.5 indicating that
the thickness uniformity along the length of a single membrane is generally good. For
example, the standard deviation for the thickness of the top 500 Å separation layer is 0.4
µm (5.8 %) while it is 0.2 µm (9.3 %) for the 100 Å separation layer. These results are
important in the sense that, since most of the membranes we utilize are of similar length, in
the data analysis one can utilize an average thickness layer. However, membranes from
different batches prepared at different years show quite a bit of variation in membrane
thickness. Consultation with Dr. Ciora of M&P indicates that is likely to variations in the
parameters of the coating procedures, but M&P is currently not willing to share any
additional information on this issue. The “take-home” message here is that the thickness of
each membrane utilized in the experiments must be measured independently; however, one
or two measurements suffice, as layer uniformity for each membrane is quite good.
Interestingly enough (as it agrees with the previous conclusions based on the simple
geometric model of Figure 2.2), SEM pictures also suggest that the uniformly-stacked
spheres model is more realistic compared to the aligned cylinders model.

77

Figure 2.3: SEM pictures for single separation layer membranes showing the top layer thicknesses for
different membrane batches supplied by M&P

0
5
10
15
20
25
2007 2008 2009 2010
Membrane samples (year of batch)
Separation layer thickness, (µm)
6
6.2
6.4
6.6
6.8
7
7.2
7.4
7.6
123 45 6
Length of membrane, (cm)
Separation layer thickness, (µm)

Figure 2.4: The results obtained by SEM analysis: Variation of top layer thickness for different
membrane batches (left), Variation of top layer thickness along the length of a membrane module (right)

Batch # 2007
Batch # 2008
Batch # 2009
Batch # 2010
78

Figure 2.5: SEM picture for a double separation layer membrane along with the variation of top layer
thickness along the length of the membrane

2.2.3. Using the Archimedes method to study membrane properties
The Archimedes' method is based on the principle that a body immersed in a fluid is
buoyed up by a force equal to the weight of the displaced fluid. The method involves
measuring the weight of the sample in air as well as in a fluid which completely wets the
sample. In our measurements we utilized an analytical balance (Mettler Taledo AT-201)
in order to measure the skeletal (net) density of α-alumina support materials and to
estimate the porosity of membrane samples. The same method can be used to measure the
overall skeletal density and porosity of the one-layer and two-layer membranes, but
deducing the porosity of the individual layers is a difficult task as it becomes obvious
from Table 2.3 as there is a 2-3 orders of magnitude difference between the mass of the
support and that of the top layers. Three S-type samples and three different wetting fluids
Batch # 2007 Double separation layer
0
0.5
1
1.5
2
2.5
3
1 2 3 4 5
Length of membrane, (cm)
Separation layer thickness, (µm)
1
2
3
79
(Water, Acetone, and Ethanol) were used for these experiments, and we modified the
procedure by forcing the fluids through the support tubes by pumping them through the
samples prior to immersing them in liquid container.
The data collected are shown in Table 2.5. The experimental results for the porosity
are close to those reported by M&P of ~25 %.

Table 2.5: The results obtained by utilization of Archimedes method
Dry weight in air,
(g)
Apparent (bulk)
volume, (cc)
Apparent density,
(g/cc)
Measured porosity,
Ethanol/Acetone/Water, (%)
Skeletal density,
(g/cc)
First sample
1.65 0.58 2.85
21.78/20.15/20.42 3.44
Second sample
1.58 0.56 2.80
21.27/19.14/19.61 3.37
Third sample
1.69 0.62 2.73
21.53/19.94/20.11 3.29
Average 1.64 0.59 2.79 20.44 3.37

2.2.4. Helium pycnometry
The porosity of the support is a key factor in determining FTCMR performance as a
highly porous support with large pores is important in guaranteeing a large air throughput
with minimal pressure drop. Therefore, in order to verify previous measured porosity of
S-type samples, we utilized a TPI-219 instructional helium pycnometer (Coretest
systems, Inc.) to measure the porosity of ten S-type samples (totally 100 cm of support
materials). The helium pycnometry method replaces the traditional method of liquid
displacement by the use of a test gas and the volume-pressure relationship known as
Boyle's Law [Lowell et al., 2004]. Since helium, which can enter even the smallest voids
or pores, is used to measure the unknown volume of a material with a known weight, the
density and consequently the porosity of powders and porous materials can be determined
precisely.
80
Using this method, the average resulting value for porosity of samples was 21 (± 5 %),
which is in very good agreement with the average value obtained by Archimedes' method
for three different wetting liquids.

2.2.5. BET analysis
For membranes which are to be rendered catalytic for use in FTCMR, as with
conventional catalysts, the membrane surface area is one of most important parameters
affecting the ability of the membrane for enhancing chemical reactions. One key test for
characterizing such membranes, therefore, is to measure their surface area. BET analysis,
based on gas adsorption, is the most widely used technique for total surface area
measurements. In BET gas molecules of known size are adsorbed and condensed within a
given sample’s pore structure. By measuring the amount of gas that adsorbs by a surface
area analyzer one can deduce the surface area of the material. Though there are a number
of gases used in BET, the most common gas is nitrogen at its liquid temperature [Sing,
2001].
BET analysis was carried out for macroporous supports and mesoporous membranes
using a Micromeritics ASAP 2010 instrument. For the symmetric support material the
corresponding surface area was between 0.3–0.4 m
2
/g not so far from the estimated
values by simple structural models (see 2.2.1).
In Figure 2.6, the pore size distribution for an asymmetric A-type membrane with a
BET surface area of 0.5031 m
2
/g is presented, indicating a bimodal pore size distribution,
as expected. For the mesoporous top layer, the nitrogen adsorption/desorption analysis
using the BJH method demonstrates a broad pore size distribution with pores ranging
81
from 20 Å to 400 Å and an average pore diameter of 230 Å. The pore size distribution for
a B-type membrane with a BET surface area of 0.55011 m
2
/g is presented in Figure 2.7,
showing a trimodal pore size distribution. For the top layer ( γ-alumian), the nitrogen
adsorption/desorption analysis using the BJH method illustrates a sharp peak with a
summit at 100 Å almost equal to the nominal value reported by M&P; while for the
intermediate layer ( α-alumina), the same analysis shows a broader peak ranging from 300
Å to 500 Å and an average pore diameter of 350 Å.
Knowing the thickness of the separation layer (measured by SEM analysis of the
samples), and also assuming that the difference between the cumulative pore volume of a
S-type membranes with that of A-type membrane is only due to the deposition of a
separation layer on top of the support material, the results of nitrogen
adsorption/desorption analysis can be used to estimate the porosity of the top-layer for A-
type asymmetric membranes. Thus from the BET reports, the values of the BJH
adsorption/desorption cumulative pore volume of pores for two S-type and three A-type
samples were extracted. The average of pore volumes for the S-type samples (four data)
was considered as the pore volume of the S-type samples used to prepare A-type
membranes, assuming that the support material used for fabricating A-type membranes
have the same properties as the bare supports. In Table 2.6, the corresponding values for
these two S-type membranes are presented. Using and are presented in this table, the
average porosity of the top-layer for three A-type samples is estimated to be 0.28 which
is in good agreement with the available data by M&P. All data for A-type membranes are
gathered in Table 2.7.

82

Figure 2.6: Pore size distribution obtained by N
2
adsorption/desorption on/from an A-type M&P
membrane

Figure 2.7: Pore size distribution obtained by N
2
adsorption/desorption on/from a B-type M&P
membrane
83
Table 2.6: N
2
adsorption/desorption results for S-type membrane samples
Adsorption
pore volume, (cc/g)
Desorption
pore volume, (cc/g)
Average
pore volume, (cc/g)
First sample 0.000514 0.000665 0.000589
Second sample 0.000636 0.000664 0.000650
Average 0.000575 0.000664 0.000619

Table 2.7: Estimation of the porosity of the top-layer based on N
2
adsorption/desorption data
Top-layer bulk
volume, (cc)
Adsorption
pore volume,
(cc/g)
Desorption
pore volume,
(cc/g)
Average
pore volume,
(cc/g)
Top-layer average
pore volume,
(cc/g)
Porosity,
(%)
First sample 0.000730 0.001025 0.001206 0.001025 0.000405 25.40
Second sample 0.006286 0.001107 0.001023 0.001107 0.000487 30.55
Third sample 0.004200 0.000893 0.000880 0.000893 0.000273 28.12
Average 0.003738 0.001008 0.001036 0.001008 0.000388 28.02

2.2.6. Bubble-point (Coulter) perporometry analysis
As stated above, most analytical methods for characterizing multilayer membranes
methods are having difficulty to differentiate between the thin separation layer and the
porous support. Although the separation layer determines the performance of the
membranes in almost all cases its mass typically constitutes only a very small fraction of
the overall membrane's mass. Additional complications result from the large differences
in pore size and surface area between the active layer and the support, which results in
the characterization of the separation layer typically lacking detail and resolution [Jakobs
and Koros, 1997].
Nitrogen adsorption-desorption and mercury intrusion methods can detect a broad
range of pore sizes, but cannot differentiate and distinguish between flow-through and
dead-end pores. In addition, the current methods for interpreting the data are based on
artificial and simple models of the porous structure (e.g., straight cylindrical non-
84
intersecting pores of uniform radius) which makes the interpretation rather questionable
[Jakobs and Koros, 1997]. One technique capable of measuring only flow-through pores
is the bubble-point perporometry method that is capable of determining the pore size and
pore size distribution of the separation layers of multilayered membranes. It is based on
the principle that, for a given liquid and pore size with a constant wetting characteristics,
the pressure required to force gas through the pore is inversely proportion to the size of
the pore [Li, 2007].
The principle of the bubble-point perporometry method is illustrated in Figure 2.8. At
the start, the membrane is wetted by a liquid, which acts as barrier, and no flow can occur
until the applied pressure reaches the capillary pressure of the largest pores. After
increasing the pressure over this limit, the liquid is expelled from the largest pores,
allowing the gas to permeate through. By successively increasing the pressure, smaller
and smaller pores are opened for permeation of the gas. The ideal flow versus pressure
drop curve generated in this fashion is usually S-shaped, as shown in Figure 2.8. In this
figure the dashed and solid lines indicate the relationship between the applied pressure
and air flow for dry and wet membranes, respectively. The first gas bubble represents the
largest pores measured at a certain minimum pressure. When the gas flow measured
under wet conditions is equal to that in dry conditions, it means that all pores are emptied
and the corresponding pressure is a measure of the smallest pores. To interpret the data in
this figure one makes use of the Laplace equation, in order to determine the pore size and
the pore size distribution. Laplace equation is expressed as [Jakobs and Koros, 1997]:
p d
p
  / cos 4   (2.2)
where θ is the liquid-surface contact angle and σ is the liquid surface tension.
85
This method is better suitable for the characterization of macropores but can be also
applied for the characterization of mesoporous membranes [Li, 2007]. Using different
liquids of varying surface tension allows one to assess different regions of the pore space.

Figure 2.8: Theoretical flow–pressure curve for the bubble point test/progressive displacement test
[Li, 2007]

A main problem in the technique is that currently the method of data interpretation is
based on the model of the membrane pore structure being that of straight nonintersecting
pores. This introduces various uncertainties. For example, one finds that the rate of pressure
increase influences the result [Li, 2007]. The validity of the assumption of cylindrical pores
also comes into question in view of the SEM analysis of the materials that was previously
presented. Other experimental uncertainties result from solvent evaporation, particularly
when the pores are getting close to being completely empty of any liquid, whereby the
wetting phase might evaporate out of the membrane pores instead of being displaced,
leading to erroneous results [Jakobs and Koros, 1997].
86
For this series of tests isopropyl alcohol (IPA) was used as the wetting fluid while
helium was used as the flowing fluid. Our current experimental set-up is limited to a
maximum flow rate of 2500 sccm. As a result, the flow-pressure curve does not reach the
point where the wet and the dry curves meet. In order to utilize the data to interpret the
full range of pore sizes we have extrapolated the experimental flow-pressure data using
three different approaches i.e., polynomial, power law, and exponential forms having a R
2
value greater than 0.92. Also, as stated above, this technique will not provide the entire
range of pore distribution. So, in this research it was assumed that the pore size
distribution has a conventional Bell-shape. Therefore, finding the biggest pore size along
with the mean pore size is enough to draw the pore size distribution. The value of mean
pore size can be estimated by crossing the half-dry flow line with wet flow curve.
In Figure 2.9 the experimental flow–pressure curves for an M&P support material is
presented and corresponding results estimated based on it are collected in Table 2.8. Very
good agreement between the vendor’s data (2000 Å ~ 4000 Å) and calculated values are
observed for all three methods of extrapolation. However, it should be reminded that in
calculating these values the value of contact angle is assumed to be zero i.e. it is assumed
that IPA completely wets the support structure. McGuire and coworkers [McGuire et al.,
1995] have reported a value of 67
o
for the contact angle of Air/IPA system, which is not
necessarily valid for the He/IPA system. But, only for sensitivity analysis, by assuming
the same value for He/IPA system, all the values in Table 2.8 would be decreased by a
factor of 0.4 (cos(67
o
) = 0.4).
Pore size distribution of this sample was determined by rearranging the Hagen-
Poiseuille as follow [Capannelli et al., 1983]:
87
AB AB AB
J p
L
n  
3
4
2 

(2.3)
Where n
AB
is the number of pores having radii ranging between r
A
and r
B
which
correspond to p
A
and p
B
, L is the length of the pores, µ is effluent fluid viscosity, and J is
the effluent flux. The subsequent pore distribution charts are shown in Figure 2.10 for
exponential fitting and in Figure 2.11 for polynomial fitting. As it can be seen, unlike the
estimated values for mean pore size, the resulted pore size distributions are completely
different. In polynomial fitting 25 percent of the pores have the same diameter as the
mean pore size whereas in the exponential fitting only 14 percent of the pores show the
same behavior. These results reveal that the assumption of Bell-shape pattern is more
close to the results obtained by polynomial fitting..

Table 2.8: Maximum, minimum and mean pore sizes obtained by the application of the bubble-point
method for a membrane support

Fitting curve Polynomial (R
2
=0.99) Power law (R
2
=0.93) Exponential (R
2
=0.92)
Pore size, (Å)
[Max, Min, Mean]
5509,1686, 2908 5509, 2762, 3327 5509, 3069, 3367

0
20
40
60
80
100
120
140
0 1020 30 4050 60
Trans-membrane pressure, (psi)
Passing flow, (cc/s) 11
Dry
wet

Figure 2.9: Experimental flow–pressure curves for the bubble point test of an M&P membrane support
88
0
2
4
6
8
10
12
14
16
6017
5237
5142
4159
3928
3673
3449
3367
3177
2908
2631
2402
2210
2046
Pore diameter, (Angstrom)
Percent, (%)

Figure 2.10: Pore size distribution for an M&P membrane support based on the assumption of a Bell-
shaped pattern and exponential fitting

0
5
10
15
20
25
30
5438
5142
4159
3928
3673
3449
3327
3069
2762
2511
2302
2125
1973
1842
1709
1550
Pore diameter, (Angstrom)
Percent, (%)

Figure 2.11: Pore size distribution for an M&P membrane support based on the assumption of Bell-shaped
pattern and polynomial fitting

In Figure 2.14 the experimental flow–pressure curves for an asymmetric 500 Å
membrane is shown and corresponding results estimated based on them are collected in
Table 2.9. The pore size prediction is poor compared with the nominal (500 Å) average
pore diameter. One reason for this weakness can be attributed to the lack of enough
experimental points. In fact, for the support material there were enough experimental points
89
till the mean pore size, but that was not the case for the 500 Å membrane because of our
experimental set-up limitations. However, it should be noted that in calculating these values
not only the value of contact angle is assumed to be zero, but also the pressure difference
between both ends of the separation layer is assumed to be the same as measurable values
i.e. trans-membrane pressure. The latter assumption has been applied in almost all previous
published results on the same topic. In fact, the permeation studies revealed that the
pressure difference across the separation layer is ~ 55 % of the total pressure difference
(refer to section 2.2.7). For comparison purposes, the corresponding pore size values when
taking into account the correction of the contact angle and for the pressure difference are
also collected in Table 2.9. Finally, the pore size distribution for the 500 Å based on the
assumption of a Bell-shaped pattern and polynomial fitting is shown in Figure 2.13.

Table 2.9: Maximum, and minimum and average pore sizes obtained by
the application of bubble-point method for a 500 Å membrane

Calculation tip Without any correction or
adjustment
Including
contact angle for air/IPA
Pore size, (Å)
[Max, Mean, Min]
1871, 760, 548 730, 300, 213

0
100
200
300
400
500
0 25 50 75 100 125 150 175 200 225 250
Trans-membrane pressure, (psi)
Passing flow, (cc/s) 11
Dry
wet

Figure 2.12: Experimental flow–pressure curves for the bubble point test of a 500 Å membrane
90
0
4
8
12
16
20
1283
1169
1080
1003
936
854
747
664
598
543
498
460
435
417
400
385
370
357
345
Pore diameter, (Angstrom)
Percent, (%)

Figure 2.13: Pore size distribution for a 500 Å membrane assuming a Bell-shaped pattern and using
polynomial fitting

2.2.7. Permeation studies
As discussed in the first chapter of this report, the number of collisions between the
reactants and catalytic sites inside the catalyst pores can be noticeably amplified by
decreasing the pore diameters to such small values that it leads to the predominance of
Knudsen diffusion. The value of maximum pore size for which Knudsen diffusion
contributes 100 % to transport can be estimated based on the Knudsen number defined as:
Kn = λ /H (2.4)
where H is the characteristic lateral dimension of the pore and λ is the mean free path of
gas molecules, defined by:
RT
2
2 /      (2.5)
In the above equation µ is the absolute viscosity and R the universal gas constant. For
values of Kn ≥ 10, Knudsen diffusion is thought to completely dominate the gas flow
inside the membrane pores [Rostami et al., 2002]. According to this definition, for air, as
the carrier gas in our experiments, the minimum value of H necessary to achieve 100 %
91
Knudsen transport is calculated to be 56 Å, 71 Å, 91 Å, and 110 Å for temperatures of
293 K, 373 K, 473 K, and 573 K respectively under atmospheric pressure conditions. So,
it is expected that for the M&P porous supports and the single-layer membranes the gas
transport would not be in the completely Knudsen–dominated regime.
In actual situations, determining the contribution of Knudsen flow can be done by
plotting the permeance of a carrier gas (e.g., air), as a function of the average pressure
across the membrane according to the following equation [Zalamea et al., 1999]:
J
av av
P P
RT L
r
MRT L
r
 




   
2
125 . 0 06 . 1 (2.6)
where J is the permeance which is permeation flux normalized by the trans-membrane
pressure (mol/m
2
/s/bar), P
av
is the arithmetic average pressure across the membrane (bar),
and L, ε, τ, μ, and r are the membrane thickness, porosity, tortuosity, viscosity of the gas,
and pore radius respectively. In the above equation α and βP
av
indicate the Knudsen and
laminar contributions to the permeation flux, respectively. The axis intercept α (at P
av
=
0) represents the pressure-independent contribution of Knudsen diffusion, whereas any
additional flux can be attributed to the convective contribution. The above equation is
consistent with the fact that by increasing the temperature from ambient to a higher value
appropriate for reaction to take place, the contribution of Knudsen should increase.
Moreover, membranes with smaller pore radii should show a more Knudsen-dependent
permeation flux.
For a tubular membrane configuration, modification and rearrangement of the
Equation 2.6 will result [Saracco et al., 1995]:
J =







M
RT
K P
B
RT r r r
av
in out out
 
8
3
4
) / ( ln
1
0
0
(2.7)
92
0
2
0
2B
K



(2.8)
 
0
4K
d
p
 (2.9)
In the above equations r
out
is the support outside radius, r
in
is the support inside radius,
R is the universal gas constant, T is the temperature, ε is the porosity, τ is the tortuosity, μ
is the gas viscosity, P
av
is the average pressure across the support, M is the molecular
weight of the flowing gas, B
0
is the viscous flow parameter, and K
0
is the Knudsen flow
parameter. Equations 2.7-2.9 can be applied easily for estimation of membrane geometric
factor ( ε/ τ) and pore size for each membrane sample. However, implementing the above
equations for a membrane consisting of more than one layer necessitates estimating the
intermediate pressure at the interface between the two different layers. This can be done
by equating the flux equations for all the layers based on the resistance in series model
approach [Henis and Tripodi, 1981]. Implementing this model, Lin and Burggraaf (1991)
derived the following equation for a membrane consisting of a separation layer and a
support:
       / ] / 2 ) / ( 2 ) / [(
2 / 1 2
1 1
2
     S Q p p p
m
(2.10)
In the above equation, p
1
and p
m
are the membrane upstream and intermediate pressures
respectively, Q is the gas flow, and S is the membrane geometric surface area. Parameters
α and β are defined as before (Equation 2.6).
Once the values of intermediate pressure is known the contribution of Knudsen and
Poiseuille flows along with the other parameters like membrane geometric factor and
pore size can be found for support and separation layer(s) individually using Equations
2.7-2.9. However, this method needs to be utilized with care when estimating the pore
93
size and geometric factor of separation layer(s) for a multi-layer membrane, especially
when one of the layers contributes more significantly to the total permeance [Uchytil,
1996]. Under such conditions, the values of the estimated parameters using Equation 2.8,
and particularly the pore diameter using Equation 2.9 are rather sensitive to the
experimental values of pressure as well as the flux through the membrane.
The membrane permeation experiments were carried out in the dead-end mode,
meaning that there was no reject side flow and all the inlet gas, whose flow rate was set
constant using a mass flow controller (MFC, Cole Parmer SO-305035), was forced to
flow through the membrane. The upstream pressure of the membrane was measured by a
digital pressure gauge (Omega DPG 4000), while the trans-membrane pressure difference
was measured by a differential pressure transducer (Omega PX2300-25BDI). The
experimental data are analyzed using Equation 2.7.
The Ar permeation data for one of the support membranes (SN-1) are shown in Figure
2.14. Figure 2.15 shows the corresponding He permeation data for the same membrane.
Figure 2.16 show the Ar permeance for a different support membrane (SN-20), while
Figure 2.17 shows the corresponding data for He. Table 2.10 summarizes the calculated
transport data for SN-1 and SN-20 membrane. Other than one set of data (the Ar data at
473 K for SN-20) the estimated transport properties are fairly consistent. There is also
good agreement between the two S-type tubes.
94
R
2
= 0.9732
R
2
= 0.9818
0.0E+00
7.0E-07
1.4E-06
2.1E-06
2.8E-06
3.5E-06
100000 120000 140000 160000 180000 200000
Average transmembrane pressure, (Pa)
Permeance, (mol/m
2
/Pa/s)
T=373 K
T=473 K

Figure 2.14: Ar permeance data for SN-1 membrane

R
2
= 0.8683
R
2
= 0.9534
2.0E-06
2.8E-06
3.6E-06
4.4E-06
5.2E-06
6.0E-06
100000 110000 120000 130000 140000 150000
Average transmembrane pressure, (Pa)
Permeance, (mol/m
2
/Pa/s)
T=373 K
T=473 K

Figure 2.15: He permeance data for SN-1 membrane

95
R
2
= 0.9518
R
2
= 0.9830
0.0E+00
6.0E-07
1.2E-06
1.8E-06
2.4E-06
3.0E-06
100000 120000 140000 160000 180000 200000
Average transmembrane pressure, (Pa)
Permeance, (mol/m
2
/Pa/s)
T=373 K
T=473 K

Figure 2.16: Ar permeance data for SN-20 membrane

R
2
= 0.881
R
2
= 0.9793
2.0E-06
2.8E-06
3.6E-06
4.4E-06
5.2E-06
6.0E-06
100000 110000 120000 130000 140000 150000
Average transmembrane pressure, (Pa)
Permeance, (mol/m
2
/Pa/s)
T=373 K
T=473 K

Figure 2.17: He permeance data for SN-20 membrane

96
Table 2.10: Summary of the single-gas permeation results for SN-1 and SN-20
symmetric support membranes

Membrane B
0
, (m
2
) K
0
, (m) dp, (Å) ε/ τ
SN-1 (Argon, T=373 K) 1.16E-15 7.79E-09 11908 0.028
SN-1 (Argon, T=473 K) 1.08E-15 8.00E-09 10754 0.030
SN-1 (Helium, T=373 K)
1.25E-15 7.75E-09
12868 0.024
SN-1 (Helium, T=473 K)
1.12E-15 7.88E-09
11365 0.028
SN-20 (Argon, T=373 K)
1.01E-15 7.90E-09 10266
0.031
SN-20 (Argon, T=473 K)
8.51E-16 8.22E-09
8285
0.040
SN-20 (Helium, T=373 K)
1.01E-15 7.87E-09
10235
0.031
SN-20 (Helium, T=473 K)
1.14E-15 7.06E-09 12861 0.022
Average
1.08E-15 7.81E-09 11068 0.029

For asymmetric membranes, the Ar permeation data for one of the A-type membranes
(A-02-2010) are shown in Figure 2.18, and Figure 2.19 shows the corresponding He
permeation data for the same membrane. Table 2.11 shows the calculated K
0
and B
0
values. There is again good consistency between the four set of calculated parameter
values. However, when analyzing the experimental data by assuming that the support had
transport properties that were the same with those of supports SN-1 and SN-20, and that
the top layer had a thickness of 7 µm based on SEM observations (section 2.2.2) very
unrealistic transport characteristics of the top layer were calculated. Such difficulties in
calculating the transport characteristics of multilayer membranes were also previously
reported by other investigators [Edreva et. al., 2009, Uchytil, 1994, Zhang et. al. 2009]
and were attributed to the simplifying assumptions of the analysis technique (e.g., that the
membrane consists of cylindrical, non-intersecting pores with a constant diameter) as
well as the presence of micro-cracks and pinholes in the mesoporous layers.
97
R
2
= 0.9872
R
2
= 0.9877
0.0E+00
5.0E-07
1.0E-06
1.5E-06
2.0E-06
2.5E-06
100000 120000 140000 160000 180000 200000 220000
Average transmembrane pressure, (Pa)
Permeance, (mol/m
2
/Pa/s)
T=373 K
T=473 K

Figure 2.18: Ar permeance data for A-02-2010 membrane

R
2
= 0.9632
R
2
= 0.9871
2.0E-06
2.4E-06
2.8E-06
3.2E-06
3.6E-06
4.0E-06
4.4E-06
100000 110000 120000 130000 140000 150000 160000
Average transmembrane pressure, (Pa)
Permeance, (mol/m
2
/Pa/s)
T=373 K
T=473 K

Figure 2.19: He permeance data for A-02-2010 membrane

98
Table 2.11: Summary of the single-gas permeation results for the A-02-2010
asymmetric membrane

Membrane B
0
, (m
2
) K
0
, (m) dp, (Å) ε/ τ
A-02-2010 (Argon, T=373 K)
7.32E-18 1.04E-10 5623 0.001
A-02-2010 (Argon, T=473 K)
5.16E-18 1.18E-10 3487 0.001
A-02-2010 (Helium, T=373 K)
9.88E-18 1.03E-10 7654 0.001
A-02-2010 (Helium, T=473 K)
7.71E-18 1.07E-10 5773 0.001
Average
7.52E-18 1.08E-10 5634 0.001

2.3. Catalyst Characterization
In heterogeneous catalysis, catalytic activity typically scales with the solid catalyst’s
surface area, and practical catalysts are generally high surface-area materials. To
achieve high surface areas, it is necessary to either prepare catalysts in a very finely
divided state or to deposit them on highly porous supports. For a certain support with a
given surface area, the smaller the size of metal nano-particles deposited on it the
higher is the catalytically active surface area. In this research the alumina membranes
are rendered catalytic by the deposition of Pt. Three different techniques are utilized to
characterize these catalytic membranes, namely SEM/EDAX, CO chemisorption
analysis, and transmission electron microscopy (TEM).
The membranes are rendered catalytic by wet impregnation of their interior surface
with a hexa-chloroplatinic acid solution (8 wt.% Sigma Aldrich). To do so, the membrane
tubes were filled by this precursor solution (or by a 1 wt.% solution for some of the
experiments) and were uniformly shaken for 5 min followed by drying and calcination at
a temperature of 350
o
C for 5 hr. Before the actual reaction runs, the catalyst was
activated by reduction in a H
2
stream for 3 hr at 400
o
C. Before and after the 5 min
99
shaking, the volume of the precursor solution was measured via a graduated syringe;
while the Pt concentration in the solution before and after shaking was measured using an
ICP-MS instrument (Perkin-Elmer model ELAN-9000). Typically when using the above
impregnation procedure, the amount of deposited Pt was measured to be ~0.05 wt.% of
the total weight of the membrane tube or 5.21 wt.% based on the weight of the separation
layer.

2.3.1. SEM/EDAX analysis
In Figure 2.20 the cross section of a single layer membrane after impregnation with Pt is
shown. The white spots in this figure show the location of the platinum catalyst in the
separation layer of the A-type (500 Å) membrane. As it can be seen from the figure most of
the platinum is concentrated on the inner membrane layer. However, a small fraction of the
Pt has escaped from the membrane layer and has also deposited on the porous alumina
support. Since the very thin separation layer is the part where Knudsen diffusion
contributes most to transport, this then gives a lot of reactivity for a small amount of
catalyst. The distribution of the Pt atoms normalized per atoms of Al, along the first 1 mm
of tubular membrane in radial direction, measured using the EDAX technique, is shown in
Figure 2.21. The results in the figure are consistent with the SEM results indicating that the
Pt mostly accumulates in the membrane layer rather than in the porous support.
A better look at the main SEM picture and its magnified zone reveals that the Pt
nanocrystals are uniformly distributed on the alumina platelets constituting the top
membrane layer. From the EDAX analysis the Pt content is measured to be 8.93 wt.%
based on the weight of the separation layer, which given the uncertainties involved in the
100
various measurement techniques and in identifying the exact weight of the top layer, is in
fairly good agreement with the results of the ICP-MS measurements.

Figure.2.20: SEM picture of a single-layer membrane after impregnation

0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 0.2 0.4 0.6 0.8 1
Dimensionless radial direction
Ratio of Pt/Al

Figure 2.21: Pt distribution along the 1 mm radial direction of a single-layer membrane

101
2.3.2. TEM study
Although the shining spots in the back-scattered electron image shown in Figure 2.20
are a good indication of the presence and uniform distribution of Pt particles, the
technique itself is not reliable enough to accurately measure the actual size of the catalyst
particles. Though many methods are currently available for estimating the particle size
(e.g., X-ray diffraction (XRD), extended X-ray absorption fine structure (EXAFS), probe
molecule absorption- magnetic measurements, optical absorption, and small-angle
neutron scattering (SANS)), only transmission electron microscopy (TEM) currently
allows for the direct (real space) visualization of the size of the nano-particles [Pyrz and
Buttrey, 2008]. The technique has been applied therefore in this study (via the use of a
JEOL 1200 TEM instrument) in order to provide a more accurate estimate of the size of
the catalyst particles deposited on the membrane layer. The results are presented in
Figure 2.24. These observations confirm that majority of Pt particles are smaller than 10
nm, and the average of 11 counted particles of the first figure shows an average of 9 nm.

Figure 2.22: TEM micrographs of a single-layer Pt/Al
2
O
3
catalytic membrane after calcination at 623 K
0
4
8
12
16
20
1 234 567 89 1011
Counted Particles
Particle Size, (nm)
102
2.3.3. Chemisorption study
An important consideration in heterogeneous catalysis is the relationship between
activity and the state of dispersion of the catalytically active metal component [Yates and
Sinfelt, 1967], which requires that one determines the specific surface area of the metal
itself. The most direct way for doing that is via the selective chemisorption of gases
[Yates and Sinfelt, 1967], particularly CO [Wong and Hofmann, 1991]. To measure CO
chemisorption in this study we have used a flow technique named Pulse Chemisorption,
utilizing a Micromeritics ChemiSorb 2720 instrument. The technique involves exposure
(further details of the procedure can be found elsewhere [Fadoni and Lucarelli, 1999]) of
the catalyst to repeated CO pulses at room temperature until complete saturation of the
metal surface occurs [Brooks and Kehrer, 1969]. The results of the chemisorption studies
are summarized in Table 2.12 that shows that 40 % of the Pt atoms are exposed to CO,
assuming that one CO molecule chemisorbs per Pt atom, as widely accepted [Fadoni and
Lucarelli, 1999].
Knowing the amount of adsorbed CO, the specific surface area S (m
2
/g), and the size
of the platinum particles d (m) can be estimated by the following equations [Brooks and
Kehrer, 1969, Geyer et al. 2012]:
8
10 28 . 2 

VN
S
Pt

(2.11)

Pt Pt
S d  . / 5  (2.12)
In the above equations V is the volume of the adsorbed gas (cm
3
STP/g),  is the
density (m
3
/g), N is the Avogadro's number, and  is the effective area occupied by the
molecule in an adsorbed monomolecular layer. Applying the above equations we
calculate an average particle size of 2.35 nm.
103
There is also another method for estimating the catalyst particle size using the CO
chemisorption data. Pt has a face-centered cubic (FCC) crystal and a density of 21.5
g/cm
3
[Smith, 2004]. Its FCC unit cell, shown in Figure 2.23, contains the equivalent of
four atoms (the eight corner octants account for one atom (8 × 1/8 = 1), and the six half-
atoms on the cube faces contribute another three atoms). Assuming that a Pt particle
consists of n such unit cells, as indicated in Figure 2.23, one can show [Smith, 2004] that
there are totally 12n
2
+2 atoms on the particle surface together with 4n
3
-6n
2
+3n-1 interior
atoms.

Table 2.12: The results of carbon monoxide chemisorption studies for a 500 Å membrane

Average weight of support + Membrane, (g) 3.62E+00
Average weight of the catalyst + Support + Membrane (after regeneration), (g) 3.54E+00
Weight of Pt for regenerated catalyst, (dimensionless) 6.87E-03
Weight of Pt per gram of regenerated catalyst, (dimensionless) 1.94E-03
Average weight of sample used for chemisorption, (g) 1.95E-01
Weight of Pt deposited on the sample, (g) 3.79E-04
Number of Pt moles deposited on the sample 1.94E-06
TCD calibration signal per 100 microliter of CO 2.10E-01
Observed TCD signal for the sample 1.70E-01
Pressure of experiment, (Pa) 1.03E+05
Temperature of experiment, (K) 2.98E+02
Number of CO moles adsorbed on the sample, (mol) 7.93E-07
Fraction of Pt moles available for reaction, (dimensionless) 4.08E-01
104

Figure 2.23: Simple representation of a face-centered cubic crystal structure

Table 2.13: The relation between the number of unit cells in a FCC and available atoms for reaction

Number of unit cells Number of
surface atoms
Number of
interior atoms
Ratio of surface
to interior atoms
1 50 13 3.85
2 110 62 1.78
3 194 171 1.13
4 302 364 0.83
5 434 665 0.655
6 590 1098 0.535
7 770 1687 0.455
8 974 2456 0.395
9 1202 3429 0.35
10 1454 4630 0.314
11 1730 6083 0.284
12 120,002 3,940,299 0.0304
100 12,000,002 3,994,002,999 0.003

A straightforward calculation (Table 2.13) then indicates that the experimental surface
availability value of ~40 % corresponds to a particle which contains on the average 9 unit
cells and having a corresponding particle size of ~3.6 nm. Note that the predicted values
105
for the catalyst particle size by either method based on the chemisorption data are lower
than the values obtained by TEM analysis.

2.3.4. Permeation studies
The effect of catalytic impregnation on the permeation characteristics of both A-type
and B-type membranes were also investigated following the same experimental method
described before (section 2.2.7) by correlating the flow of air with average pressure
across the membrane and the results are shown in Table 2.14 before impregnation, and in
Table 2.15 after impregnation with Pt precursor. The data presented here for individual
layers are obtained by assuming that the characteristics of the support layers does not
change when another alumina layer ( α-Al
2
O
3
for A01 and γ-Al
2
O
3
for B01 samples) is
deposited on top of them. These results indicate the higher the pore size the less decrease
in Knudsen contribution after impregnation. Overall, the impregnation with a 1 wt.% Pt
solution did not affect the fractional Knudsen contribution to the transport of membrane
samples considerably; though based on the TEM and chemisorption analysis results,
which are discussed earlier (the size of Pt crystals being < 10 nm), this may be expected
for the S- and A-type membranes, and the observation is somewhat surprising for the B-
type membranes for which the average pore size is comparable with the Pt crystallite
size.

106
Table 2.14: Permeation characteristics of different membrane samples before impregnation

Sample Type Nominal pore
diameter, (Å)
Temperature,
(K)
Support material
Knudsen contribution,
(%)
First Separation layer
Knudsen contribution,
(%)
Second Separation
layer Knudsen
contribution, (%)
A01
α-alumina over
support
500 293 51 55 -
A01
α-alumina over
support
500 373 50 65 -
A01
α-alumina over
support
500 473 53 73 -
A01
α-alumina over
support
500 573 53 73 -
B01
γ-alumina over
α-alumina/support
100 293 51 55 89
B01
γ-alumina over
α-alumina/support
100 373 50 65 83
B01
γ-alumina over
α-alumina/support
100 473 53 73 79
B01
γ-alumina over
α-alumina/support
100 573 53 73 87

Table 2.15: Permeation characteristics of different membrane samples after impregnation

Sample Type Nominal pore
diameter, (Å)
Temperature,
(K)
Support material•
Knudsen contribution,
(%)
First Separation layer
Knudsen contribution,
(%)
Second Separation
layer Knudsen
contribution, (%)
A01-Pt Impregnated A01 500 293 51 55 -
A01-Pt Impregnated A01 500 373 50 64 -
A01-Pt Impregnated A01 500 473 53 71 -
A01-Pt Impregnated A01 500 573 53 73 -
B01-Pt Impregnated B01 100 293 51 55 87
B01-Pt Impregnated B01 100 373 50 64 82
B01-Pt Impregnated B01 100 473 53 71 79
B01-Pt Impregnated B01 100 573 53 73 90

 Assuming no catalyst is deposited on the porous support

107
2.4. Conclusions
The performance of catalytic ceramic membranes is not only associated with their
overall morphology, but also it directly depends on the distribution and average size of
catalyst particles within the porous structure of the membrane. Therefore, information
about the pore size, porosity, particle size, and membrane surface area are of importance.
In this Chapter, we have characterized the catalytic ceramic membranes fabricated for
application under FTCMR mode of operation using different available methods including
SEM, TEM, BET, CO chemisorption, permporometry, porosimetry, and single gas
permeation. The results of these characterization study can be used in explaining the
experimental behavior of FTCMR under different operating conditions reported in this
Thesis, and also in simulation studies in Chapter 4.

108
Chapter 3: Reaction Studies

3.1. Introduction
In this Chapter experimental observations and results obtained during the
thermocatalytic decomposition of DMMP using catalytic membrane reactors will be
discussed. The catalytic membranes used in these experiments were characterized using
different methods, and the results were presented in Chapter 2. Reactor experiments were
performed under various reaction temperatures (373 K - 673 K), different feed DMMP
concentrations (120 ppm - 1000 ppm) as well as different feed flow rates. In order to
demonstrate the probable superiority of FTCMR over other conventional reactor systems,
experiments were also carried out under the plug-flow mode of operation, i.e., with the
reject side of the reactor being left open and the membrane external surface being
completely coated with an impermeable glaze. A mathematical model has also been
developed in order to provide a better understanding of the fundamental transport
phenomena underpinning the FTCMR operation. The model is used for identifying the
advantages of the FTCMR concept in comparison with the plug-flow reactor, and also for
investigating some of the limitations, which may exist in applying this concept for the
complete destruction of chemical warfare simulants.

3.2. Experimental Set-up
The schematic of the experimental set-up utilized is shown in Figure 3.1. It comprises
of three different sections:
109
- Feed generation section
- Reactor section
- Analytical section
Among the various methods for producing a vapor stream of a liquid chemical warfare
simulant [Sun and Ong, 2005], we have chosen the bubbler method because of its
widespread use in continuous vapor generation in previously published articles (see
Chapter 1). The reactor feed stream was generated and introduced into the reactor system
by bubbling the air, as the carrier gas, through one (and for some studies two successive)
glass-bubblers filled with DMMP. The bubbler(s) was loaded with glass beads to increase
the contact surface. High purity (>99.5 %) DMMP was purchased from Pfaltz & Bauer
and it was used without any further treatment or purification. Standard dry compressed
air was used as the carrier gas while, ultra-high purity hydrogen and helium gases
(supplied by Gilmore Liquid Air Co.) were used for the catalyst activation and the system
purge, respectively. Mass flow controllers (Brooks 5850E and Cole Parmer EW-32907-
81) were used to set the flow rate of the gases through the system.
In order to make certain that constant DMMP concentrations were generated, the
bubbler was immersed completely in a large insulated vessel filled with circulating water
from a temperature-controlled thermal bath (Lauda RE-220). By varying the bath
temperature it was possible to generate a broad range of DMMP concentrations. The
concentration (partial pressure) of DMMP in air (in complete equilibrium with liquid
DMMP) can be calculated by the Antoine equation [Butrow et al., 2009]:
) /( ) ( ln T c b a p    (3.1)
where p is the DMMP vapor pressure (Pa) and T is the temperature (K) of the system.
110

Figure 3.1: Experimental set-up used for the catalytic destruction of DMMP

Butrow et al. (2009) reported a = 22.319, b = 4340.0, and c = -51.7 for the constants
of the Antoine equation for the air/DMMP systems. They verified the accuracy of these
coefficients for a temperature range of 258.2 K – 454.4 K. Fan and Wang (2010) reported
different coefficients for Equation 3.1, namely a = 4.887, b = 771.0 and c = -185.1,
whereby ln (p, Pa) in Equation 3.1 should also be replaced by log (p, kPa)); however,
their correlation does not apply for temperatures below 358.2 K and above 453.5 K. As a
comparison, the values of DMMP vapor pressures predicted using the coefficients of
Butrow et al. (2009) for two temperatures of 400 K and 450 K are 19.12 kPa and 91.37
kPa respectively; while the corresponding values using the coefficients of Fan and Wang
(2010) are 19.91 kPa and 94.72 kPa.
111
For the on-line measurement of DMMP an HP-6890 gas chromatograph coupled to a
HP-5793 mass-spectrometric detector was used. It is equipped with a capillary column
(J&W HP-5 ms) which was able to provide good separation of DMMP from both the feed
and product gas streams. Samples were injected into the GC column via a pneumatically-
driven six-port valve. A 1 ml sample loop was utilized heated at 323 K in order to
minimize the potential for condensation/adsorption of DMMP.
In our experiments, the DMMP concentrations exiting the bubbler were always
measured using this GC/MS rather than being estimated from the Antoine equation, as
one cannot always guarantee that under all conditions complete saturation is achieved.
The GC/MS was calibrated for different concentrations of DMMP using standard
samples of DMMP in MeOH and DMMP in H
2
O. The difference between the results
obtained from MeOH and those of H
2
O was found to be ~ 5 %. On the other hand, since
the real feed gas supplied to the reactor was in the gas phase, the GC/MS was also
calibrated for gas phase air/DMMP mixtures using the bubbler outlet stream diluted with
different flows of helium or carbon dioxide. Assuming the validity of ideal gas law and
atmospheric pressure inside the sample loop, since the volume and the temperature of the
sample loop were known one could estimate the total number of moles injected into the
GC/MS. So, the concentration of DMMP in the gas phase could easily be evaluated by
comparing the area of the resulting peaks from gas phase injections with the area of the
peaks obtained by injection of the standard DMMP/water or DMMP/MeOH liquid
samples.
Tubular alumina membranes provided by M&P (9 cm in length) were utilized in our
experiments. 1 cm from each end was glazed with a special gloss glaze (Duncan GL612),
112
so for all membranes the active permeation length was ~7 cm long. Two different types
of membranes were studied (in addition to the alumina support). They include A-type,
which are single-layer membranes that have an α-alumina layer on the top of the α-
alumina support with a nominal pore size of 500 Å, and B-type and C-type membranes,
with a γ-alumina layer deposited on the A-type membrane, with a nominal pore size of
100 Å (for the B-type) and 40 Å (for the C-type) membranes. For all these membranes
the selective layers were coated on the inside surface of the tubular support.
The reactor was a cylindrical stainless-steel vessel having three openings and two
thermocouple insertion ports (Figure 3.2). The reactor was placed inside a furnace (Gow-
Mac 580 series), whose temperature could be controlled with adequate precision from
room temperature to 673
o
K. The difference between the oven and the reactor inside
temperature was recorded to be less than 10
o
K. The feed-side and reject-side pressures
were measured by local pressure gauges (Omega DPG-4000); while the exact trans-
membrane pressure between the feed-side and the permeate-side was being monitored
using a pressure differential transducer (Omega PX2300-25BDI). During the
experiments, the reactor was fed from the tube side. All tubing from the bubbler to the
reactor and from the reactor to the analytical section was made of Polytetrafluoroethylene
(PTFE) in order to minimize potential DMMP adsorption. Moreover, the temperature of
all tubing from bubbler to the reactor was kept 30 K above the ambient (55
o
C) using
heating tapes (HTS/Amptek AWH-101-020DM).

113

Figure 3.2: Schematic of the stainless-steel reactor

3.3. Reaction Studies
For the reaction studies a key metric in all experiments was the so-called “protection
time”, which is defined as the period of time during which the conversion of DMMP is
complete. Our attention was first focused on whether the stainless-steel reactor as well as
the alumina membranes (without any Pt) by themselves exhibited any activity towards
the destruction of DMMP. The results of experiments at three different temperatures (373
K, 473 K, and 573 K) for an air stream containing 300 ppm DMMP and flowing through
the empty stainless steel reactor are shown in Figure 3.3. Initially, it appears that the
reactor is reactive towards the DMMP, however, the DMMP conversion rapidly declines
(within a few min) to very low values. Conversion at later times shows some scatter, but
Feed entry
Product exit for
plug-flow/kinetics
experiments
Product exit for
FTCMR/kinetics
experiments
Thermocouple
insertion port # 1 Thermocouple
insertion port # 2
Reactor main body
Ceramic membrane
5.7 mm
Ceramic membrane
5.7 mm
114
generally temperature does not have a substantial effect on DMMP conversion which is
consistent with the idea that the initial disappearance of DMMP is due to adsorption on
the SS reactor wall which saturates after a short time on stream.
The results for interaction of DMMP with unimpregnated alumina membranes (in this
case an A-type membrane) are shown in Figures 3.4 and 3.5 for two different DMMP
concentrations of 300 ppm and 1000 ppm respectively. Overall, the Pt-free alumina
membrane, even in the FTCMR mode of operation, does not appear capable of
catalytically destructing substantial amounts of DMMP. In fact, under all conditions
studied the protection time was zero even at a high temperature of 573 K.

0
20
40
60
80
100
0 50 100 150 200 250
Time on stream, (min)
DMMP conversion, (%)
373 K
473 K
573 K

Figure 3.3: DMMP conversion in the stainless-steel reactor under different temperatures

The results of the catalytic conversion of DMMP on a catalytically-impregnated A-
type membrane are presented in Figure 3.6 for the same three temperatures (373 K, 473
K, and 573 K) as the unimpregnated membranes. Note, that unlike the case with the
115
unimpregnated membranes (Figure 3.4), there is an extended period of time over which
no DMMP escapes the reactor. As the temperature increases the protection time increases
(in this case ~ 30 min for every 100 K). The higher protection times can be attributed to
increased catalytic reaction rates associated with the higher temperatures, but also to the
increase in the contribution of the Knudsen flow at elevated temperatures.

0
20
40
60
80
100
0 100 200 300 400 500
Time on stream, (min)
DMMP conversion, (%)
373 K
573 K
473 K

Figure 3.4: Destruction at different temperatures under low DMMP load (300 ppm) using
unimpregnated single-layer membranes

116
0
20
40
60
80
100
0 100 200 300 400 500
Time on stream, (min)
DMMP conversion, (%)
373 K
473 K
573 K

Figure 3.5: Destruction at different temperatures under high DMMP load (1000 ppm) using
unimpregnated single-layer membranes

0
20
40
60
80
100
0 100 200 300 400 500
Time on stream, (min)
DMMP conversion, (%)
373 K
473 K
573 K

Figure 3.6: The effect of reaction temperature on DMMP destruction for impregnated single-layer
membranes (DMMP load: 300 ppm)

117
The effect of feed concentration on the complete combustion of DMMP in the FTCMR
was also investigated using three different feed concentrations of 150 ppm, 300 ppm, and
1000 ppm (reactor temperature of 573 K). As Figure 3.7 shows, the protection time is a
strong function of the DMMP concentration with long protection times observed for the
lower concentration (150 ppm) and shorter protection times associated with the higher
concentrations. The observations in Figure 3.7 indicate that a more appropriate role for the
FTCMR device would be that of a 2
nd
stage following a bulk-toxin removal stage, e.g., via
condensation or physical adsorption. Such a process will be discussed in Chapter 5.
In order to investigate the impact of increasing the Knudsen flow contribution in the
performance of the FTCMR we carried out experiments for the same temperature of 573
K and DMMP concentration of 300 ppm using three different membranes, namely a
tubular support (nominal pore diameter of 4000 Å), a single-layer membrane (nominal
pore diameter of 500 Å) and a dual-layer membrane (nominal pore diameter of 40 Å).
The results are shown in Figure 3.8, and they clearly indicate that the smaller pore size
membranes provide a much larger protection times than the larger pore size membranes.
These observations are consistent with the idea that for the smaller pore size membranes
Knudsen diffusion makes a more significant contribution to transport (e.g., for the two-
layer 40 Å membrane the Knudsen contribution is more than 90 %).

118
0
20
40
60
80
100
0 100 200 300 400 500
Time on stream, (min)
DMMP conversion, (%)
150 ppm
300 ppm
1000 ppm

Figure 3.7: The effect of feed concentration on DMMP destruction for impregnated single-layer
membranes (reactor temperature of 573 K)

0
20
40
60
80
100
0 100 200 300 400 500
Time on stream, (min)
DMMP conversion, (%)
4000 Angstrom
500 Angstrom
40 Angstrom

Figure 3.8: The effect of membrane pore size on DMMP destruction
(reactor temperature of 573 K, DMMP load: 300 ppm)

119
In order to compare the behavior of the FTCMR with that of the competitive monolith
reactor, we carried out experiments in which the outside surface of the membrane was
completely glazed in order to become impermeable to gas transport, and in addition the exit
from the permeate side was also closed (refer to Figure 3.2). During the operation with the
glazed membranes the system operated as a truly wall-coated, catalytic monolith reactor,
with the contaminated air-stream flowing parallel to the interior surface of membrane and
diffusing into and reacting in its pores. Experiments were carried out for three different
temperatures (373 K, 473 K, and 573 K) and two different feed concentrations (300 ppm
and 1000 ppm), and the results are shown in Figures 3.9 and 3.10. They indicate that under
the same operating conditions (temperature, inlet pressure, and the amount of catalyst), the
FTCMR is noticeably superior than the monolith reactor. In fact the performance of the
monolith reactor is so inferior that it cannot reach complete DMMP conversion under any
of the conditions investigated.

0
20
40
60
80
100
0 100 200 300 400 500
Time on stream, (min)
DMMP conversion, (%)
T = 473 K
T = 373 K
T = 573 K

Figure 3.9: DMMP destruction in plug-flow mode of operation
(single-layer membrane, DMMP load: 300 ppm)
120
0
20
40
60
80
100
0 100 200 300 400 500
Time on stream, (min)
DMMP conversion, (%)
T = 473 K
T = 373 K
T = 573 K

Figure 3.10: DMMP destruction in plug-flow mode of operation
(single-layer membrane, DMMP load: 1000 ppm)

Great care was taken in these experiments to try to run the two sets of experiments under
very similar conditions. For example, the monolith reactor was operated under the same
pressure conditions with the FTCMR. The same method was also applied for the
impregnation of these membranes. In order to prevent the Pt crystal growth, and therefore
possible pore blockage at the high temperatures required for the glazing, the impregnation
was performed after the glazing. However, it is very difficult to directly compare the two
systems experimentally as the optimal conditions for the FTCMR may not be the same with
those of the catalytic monolith reactor. In the remainder of this Chapter, we will utilize a
mathematical model to compare FTCMR behavior, on as much as an equitable basis as
possible, with the other reactors. For packed-bed reactors, for example, one may
significantly increase the catalyst loading without causing significant operational
difficulties. On the other hand, the same is not true for the catalytic monolith and the
FTCMR. It is important also not to fall in the “trap” of comparing one reactor under its
121
optimal operating conditions with the other competitive reactors operating under sub-
optimal circumstances.
The data presented, so far, demonstrate the occurrence of deactivation phenomena
during the catalytic destruction of DMMP over the Pt/Al
2
O
3
membranes depending on
the catalyst activity and reaction conditions. At the same time, absence of any
phosphorous-containing product in the permeate flow coming from the reactor indicates
the accumulation of these compounds inside the porous structure of the catalytic
membrane, leading to gradual plugging of the pores. In fact, during the experiments
where the air flow through the membrane was kept constant by the mass flow controller
and the permeate side pressure was kept atmospheric accumulation of phosphorous-
containing products inside the membrane manifested itself as a pressure build-up at the
reactor inlet, a phenomenon which cannot be observed in conventional fixed-bed and
monolith reactors. Deactivation and pore plugging of the catalytic membrane during the
catalytic destruction of DMMP will be discussed comprehensively in Chapter 4.

3.4. Theoretical Studies
Here, a theoretical steady-state model is developed and described, in order to study the
FTCMR, and to allow one to compare its behavior with that of the monolith reactor. The
model is continuous and is, therefore, unable to predict the phenomenon that under
Knudsen flow conditions one may potentially attain more effective catalyst utilization
[Albo et al., 2006]. The model, on the other hand, accounts properly for the transport
phenomena that take place in the FTCMR, and validates the positive impact that enhanced
mass transport has on the performance of such reactors (albeit at an increased cost of
122
needing to maintain a trans-membrane pressure). It should be emphasized here, however,
that the enhanced mass transport characteristics of the FTCMR are likely to be only one of
several reasons why such reactors perform better than the more conventional counterparts.
More effective catalyst utilization, under the Knudsen flow conditions, and the apparent
ability of the FTCMR to better accommodate and tolerate poisons and undesirable
intermediates are likely to be more key contributors to the improved performance of the
FTCMR for such applications. A more complicated model incorporating the effect of
catalyst deactivation and pore plugging on the performance of the reactor during the
reaction will be described in the next Chapter
In the model, in the description of transport and reaction in each membrane layer in
the FTCMR, the Dusty-Gas Model (DGM) [Mason and Malinauskas, 1983] is utilized.
This is thought to describe properly the physics of molecular transport in mesoporous and
macroporous membranes and catalysts that occurs by a combination of molecule-
molecule and molecule-pore-wall collisions, augmenting the convective flow of the
whole mixture (since the oxidation reaction occurs at relatively high temperatures, the
transport of molecules by surface diffusion is ignored).
In the DGM context, transport and reaction in each FTCMR layer is described by the
following equations:












 

  
n
j i
j
e
K i
e
i
e
K i
i
e
ij
j i i j
i
D
P B
RT
p
D
N
D
N y N y
p
RT
1 , ,
1

(3.2)
i i
Rxn N   (3.3)
Where:
123
i
p
K i
e
K i
M
RT
d
D D
 


 8
3
, ,
  (3.4)
32
2
p e
d
B


 (3.5)
In the above equations, p
i
is the partial pressure, M
i
the molecular mass, y
i
the mole
fraction, N
i
the flux of component i, R the universal gas constant, T the temperature, and
D
ij
the binary diffusion, whose value is estimated using Fuller’s method [Poling et al.,
2000]; D
i
,
k
is the Knudsen diffusion coefficient, P the total gas pressure, μ the gas
viscosity, d
p
the average pore size of the porous layer, ε is the porosity, and τ is the
tortuosity. Inherent in the above equations is the assumption that the membrane layer
pore structure consists of straight nonintersecting cylindrical pores (with an average pore
diameter d
p
). For describing and comparing the initial rates (which is the key goal of the
model presented here), this is not necessarily a substantially limiting assumption.
With a few notable exceptions (e.g., the blood agent cyanogen chloride), CWA and
their simulants have low volatilities, so their vapor phase concentrations in contaminated
air steams are also low [Munro et al., 1999], typically in the ppm-level range. For
DMMP, in particular, under ambient conditions (25
o
C), Equation 3.1 predicts a vapor
pressure of ~111 Pa, corresponding to a concentration in atmospheric air of 1100 ppm.
This fact significantly simplifies the modeling of the FTCMR, and justifies the following
simplifying assumptions:
- The transport of air molecules through the porous membrane is not affected by
DMMP or other gas phase reaction products (e.g., CO
2
or water). Air behaves as a single
gas, and not as a mixture of N
2
and O
2
, which is a good assumption for mesoporous
membranes and for reactions

in significant excess air.
124
- The change in the number of oxygen (air) molecules during the reactions is
negligible. Therefore, the mole fraction of air stays constant ~1.0.
- Reactant and product molecules are so dilute in the carrier gas (air) that they
effectively do not “see” each other. In the DGM equation, therefore, frictional forces
between such molecules are ignored; individual reactant or product molecules effectively
behave as if they were in a binary mixture with air.
In addition the mathematical model assumes:
- Ideal gas conditions (a good assumption for dilute air mixtures at elevated
temperature conditions).
- Isothermal conditions to match the laboratory experimental conditions and the
expected mode of operation for CWA IP and CP devices.
- Steady-state operation, since deactivation and potential pore blockage are ignored
(see Chapter 4 for a model describing catalyst deactivation and potential pore blockage).
Based on the above assumptions, the DGM equation for the carrier gas (air) reduces to
the following form (hereafter, subscript “A” stands for air, while subscript “D” stands for
DMMP)
dr
dp B
RT
p
RT
D
P
B
RT
p
RT
D
N
A
e
A
e
K A
e
A
e
K A
A








   








  
 
, ,
. (3.6)
which must be combined with the mass conservation equation (in cylindrical coordinates
for the tubular membranes), i.e., Equation (3.7) below.
0 ) (
1
 
A
rN
dr
d
r
(3.7)
For DMMP, the DGM equation becomes:
125
e
K D
e
D
e
K D
D
e
DA
A D D
D
D
P B
RT
p
D
N
D
N y N
p
RT
, ,
1 








 

  

(3.8)
Combining Equations (3.6) and (3.8), one obtains:





























    
A
e
e
K A
e
K D
e
DA
e
DA A
A D
D e
D
D
p B
D D
D
D p
N p
p
RT D
N

, ,
1
1
1
1
(3.9)
where
e
D
D , the effective diffusivity of DMMP is given by the following equation:
1
,
1 1









 
e
K D
e
DA
e
D
D D
D (3.10)
Mass conservation for DMMP, in the membrane layers that are catalytic, becomes:
D D
Rxn rN
dr
d
r
  ) (
1
(3.11)
where Rxn
D
is the reaction rate for the oxidation of DMMP (mol/s-m
3
). In this study the
catalytic oxidation of DMMP is assumed to follow a first order global reaction rate
expression, and the temperature dependence of the pseudo-rate constant is taken to be the
same with that reported by Graven et al. (1966). In the non-catalytic layers the right side
of Equation (3.11) is set equal to zero. Following some algebraic manipulations,
Equations. (3.7) and (3.11) become:
0
) ( 1
2
2
2







 
dr
dp
p
r
dr
dp
r dr
p d
A
A
A A

(3.12)
D D
e
D
D e
D
D
e
D
Rxn p
r
r
dr
r d
D
dr
dp
r
RT r
D
dr
p d
RT
D







 






 
) ( ) (
) (
) (
1
2
2
 
 (3.13)
where
126










A
e
e
K A
p B
D
r


,
1
1
)( (3.14)
















  ) ( 1 ) (
,
r
D
D
D p
N
r
e
K D
e
DA
e
DA A
A
  (3.15)
In the above equations
e
K D
e
DA
D D
,
is the ratio of the continuum to the Knudsen
diffusional flux of DMMP, and
A
e
e
K A
p B
D 
,
is the ratio of the mass of air that transports via
Knudsen to that by convective flow. In membrane layers where Knudsen flow prevails,
e.g., when the average pore size is substantially less than the mean free path of air
molecules, 0 ) (  r  and the above equations simplify considerably. The same is true
also when convective flow prevails (very large pores), whereby 1 ) (  r  .
For the case of the monolith the flux of air (N
A
) becomes zero (e.g., in the laboratory
experiments the permeate side of the membrane is closed, and the air flow in the radial
direction is further precluded by glazing the outer membrane surface), and the transport
and reaction of DMMP molecules inside the membrane pores can be expressed by the
well-known diffusion equation:
D
D D
e
D
Rxn
dr
dp
r dr
p d
RT
D










1
2
2
(3.16)
The reactor cross-sections for the FTCMR and the monolith reactor modeling are
shown schematically in Figure 3.2. In modeling the reactors, the transport equations for
the FTCMR and the monolith must be coupled with the proper mass balance equations
for the tube-side (FTCMR and monolith) and the shell-side (FTCMR). Here, it is
assumed that the tube-side and shell-side show plug-flow behavior, the flows are
127
cocurrent, and that the pressure drops are negligible (all these assumptions apply well to
the laboratory reactors). The governing equations for tube-side and shell-side (for DMMP
-- similar equations also apply for the transport of air) are as follows:
in
r r
D
D
rN
dz
dn

  ) ( 2  , for the tube-side (3.17)
out
r r
D
D
rN
dz
dn

 ) ( 2  , for the shell-side (3.18)
For the FTCMR the above differential equations must be solved together with the
appropriate boundary conditions, as follows (for the DMMP -- similar equations are also
used for air):
tube tube
D D
P y p  at
in
r r  (3.19a)
shell shell
D D
P y p  at
out
r r  (3.19b)
0 , tube
D
tube
D
p p  at 0  z (3.19c)
0 , shell
D
shell
D
p p  at 0  z (3.19d)
For the monolith reactor, Equations (3.19a) and (3.19c) together with the following
boundary condition apply:
0 
dr
dp
D
at
out
r r  (3.20)
The above equations can be made dimensionless by defining appropriate dimensionless
variables and groups, as listed in Table 3.1, and the resulting dimensionless equations are
presented below:
For DMMP:
128
D D ref
D A
A
D
A D D
X w Φ
X w X dw
dX
w X
dw
dX
w
dw
dX
dw
X d
 
 

 
/ ) 2 exp(
) (
1
) 1 (
) ( ) (
2
2
2
2
2

































(3.21)
0
) (
Pe







 
w
A
D
D
ref
ratio D
tube
D
dw
dX
w X
dw
dX P
dZ
dY



(3.22)

1
) (
Pe







  
w
A
D
D
ref
ratio D
shell
D
dw
dX
w X
dw
dX P
dZ
dY



(3.23)

And correspondingly for air:
0
) 1 (
2
2
2









dw
dX
X dw
X d
A
A
A


(3.24)
0
) 1 (
Pe







 
w
A
A
ref
ratio A
tube
A
dw
dX
X
P
dZ
dY



(3.25)

1
) 1 (
Pe







  
w
A
A
ref
ratio A
shell
A
dw
dX
X
P
dZ
dY



(3.26)
With the boundary conditions:
tube
i i
y X  at 0  w (3.27a)
ratio
shell
i i
P y X /  at 1  w (3.27b)
0 , tube
i
tube
i
y Y  at 0  Z (3.27c)
0 , 0 shell
i ratio
shell
i
y n Y  at 0  Z (3.27d)
For the monolith reactor there is no equation for the shell-side, and the governing
equations for the reactive layer and tube-side are as follow:
D D
D
X w Φ
dw
X d
  / ) 2 exp(
2
ref 2
2
 (3.28)
129
0 ref
Pe








w
D ratio D
tube
D
dw
dX P
dZ
dY


(3.29)
With the boundary conditions:
tube
D D
y X  at 0  w (3.30a)
0 
dw
dX
D
at 1  w (3.30b)
0 , tube
D
tube
D
y Y  at 0  Z (3.30c)
The above systems of equations for the FTCMR and the monolith reactor were solved
using the finite difference method in a Matlab
®
program code developed for this purpose
(Appendix). All the parameter values used during modeling are shown in Table 3.2.

130
Table 3.1: Dimensionless parameters used for the theoretical model
Dimensionless group
or parameter
Mathematical
expression
Physical definition
α ln (r
out
/r
in
) -
w ln (r/r
in
)/ α Dimensionless radial coordinate
Z z/L Dimensionless longitudinal coordinate
X
i

0 , tube
i
P p -
Y
i

0 , tube
i
n n -
Pe
ref

e
K A
tube
D LP
RT n
ref , , ref
0 ,
2 

Modified reference Pe ćlet number:
The ratio of longitudinal inlet flow to the radial
diffusion at the standard pressure normalized by the
length of the membrane for a 50 Å base membrane
Φ
ref

e
K A
in
D
k
r
ref , ,
/


Modified reference Thiele module for a 50 Å base
membrane
γ
e
DA
e
K A
D D
,
-
φ 1/X
A
+ ξ + ξσ -
δ
e
K A
e
i
D D
ref , ,
-
ξ

e
K A
A
e
D
p B
,

The ratio of the mass of air that transports by
convective flow to that via Knudsen
σ
e
K D
e
DA
D D
, The ratio of the mass of DMMP that transports by
continuum diffusion to that via Knudsen diffusion
P
ratio
P
tube
/P
shell
Tube-side to the shell-side pressure ratio
n
0
ratio
n
shell
,0
/n
tube
,0
Sweep gas to the feed gas molar flow ratio

Table 3.2: Parameters and their ranges used for the modeling
Porosity/
Tortuoasity
Pore
diameter,
(Å)
Thiele
module
(reference)
Pe ćlet
number
(reference)
DMMP
Load,
(ppm)
Sweep
ratio
Shell-side
pressure,
(Pa)
Tube-side
pressure,
(Pa)
α Reaction
temperature,
(K)
0.35/1/porosity
50,
100,
500
0.1-10 1-100 1000 0.05 101325
170250,
115110,
104770
0.006,
0.05,
0.5

573

131
3.5. Modeling Results and Analysis
Two different types of membranes are investigated here: (i) symmetric single-layer
catalytically active membranes, which represent the case where the macroporous non-
catalytic support layer offers little resistance to transport; and (ii) two-layer membranes in
order to investigate the impact on the FTCMR behavior resulting from the additional
resistance to transport offered by the support layer(s). In all instances, the behavior of the
FTCMR is compared to that of the monolith reactor with the same catalytic layer under
the same operating conditions.
In Figure 3.11a, the conversion of the FTCMR is compared to that of the monolith
reactor for a 50 Å single-layer membrane ( α = 0.006) as a function of (1/Pe
ref
) till the
point where the FTCMR operates in a total flow-through (100 % stage-cut) mode (the
exact stage-cut corresponding to each value of Pe
ref
is also shown on the figure). For large
values of (1/Pe
ref
) the conversion of both reactors goes to 100 %, each reactor then
offering complete protection. In Figure 3.11a for the monolith reactor two different cases
are studied. In the first case, it is assumed (as was done by Pina et al. (1997)) when
comparing the monolith reactor with the FTCMR) that the feed pressure is close to
atmospheric (101,760 Pa). In the other case, the monolith reactor is assumed to operate at
the same feed pressure as the FTCMR. The FTCMR, under the conditions in Figure
3.11a, is superior to the catalytic monolith for the whole range of (1/Pe
ref
) values studied
irrespective of the feed pressure applied. Note that the FTCMR provides complete
protection (> 99.5%) for a broad region of 1/Pe
ref
values, while this is not true for the
monolith reactor. The effect of feed pressure on reactor performance is shown in Figure
3.11b for a 500 Å single layer membrane ( α = 0.5). Increasing the feed pressure, as it is
132
depicted in Figure 3.11b, improves the performance for both types of reactors under the
conditions studied. Note again, that for these conditions the FTCMR outperforms the
monolith reactor.
70
80
90
100
0.001 0.003 0.005 0.007 0.009 0.011
1/Pe
ref
DMMP conversion, (%)
20 40 60 80 100
Stage-cut, (%)
FTCMR
Monolith
Monolith (atmospheric)

Figure 3.11a: DMMP conversion for the FTCMR and the monolith reactor (50 Å, α=0.006)

0
20
40
60
80
1.01.3 1.51.8 2.02.3 2.52.8 3.0
P
ratio
DMMP conversion, (%)
025 50 75 100
Stage-cut, (%)
Monolith
Monolith (atmospheric)
FTCMR

Figure 3.11b: DMMP conversion for the FTCMR and the monolith reactor (500 Å, α=0.5)
133
There are conditions, of course, under which the monolith reactor outperforms the
FTCMR reactor. As described by the equations in the Theoretical Model section, the key
advantage of the FTCMR (see previous discussions about other potential advantages that
are not accounted for by these equations) is its ability to deliver the reactants at enhanced
rate to the catalytic sites in the active layer. However, this only presents an advantage if
the increased flow of reactants during the operation under FTCMR conditions is
complemented by providing sufficient catalytic activity inside the pores; otherwise,
reactant slip (loss) to the shell-side will occur which compromises the FTCMR
performance. This, of course, is a common problem that other membrane reactors also
face [Sanchez-Marcano and Tsotsis, 2002], necessitating the need to optimize membrane
characteristics and, therefore, reactor performance. The above discussion suggests that
there should be a certain region in the parameter space where the FTCMR shows superior
performance to the monolith (packed-bed) reactor, whereas outside this region either both
reactors will have the same complete conversion, or the FTCMR will not demonstrate
any substantial improvement. This region in the Péclet number (proportional to the ratio
of a characteristic time for transport through the membrane to the characteristic time for
flow through the membrane lumen) vs. Thiele modulus (proportional to the ratio of a
characteristic time for transport through the membrane to the characteristic time for
reaction) parameter space for a 50 Å membrane is shown in Figure 3.12 for various
dimensionless thicknesses of the catalytic layer (parameter α). Figure 3.12a corresponds
to a dimensionless pressure ratio of P
ratio
= 1.03, Figure 3.12b to P
ratio
= 1.14, and Figure
3.12c to P
ratio
= 1.68 (in these figures, as noted in Table 3.2, the FTCMR operates with
the permeate side at atmospheric conditions, and the monolith reactor pressure is set
134
equal to the feed pressure of the FTCMR). Figure 3.13 summarizes the P
ratio
effect for a
membrane with a fixed thickness. Both figures manifest the detrimental effect that
increased reactant loss (due to the decrease in membrane thickness or increase in P
ratio
)
has on FTCMR performance. It is worth noting that in these figures only the regions in
which FTCMR is absolutely superior to he monolith are shown, and the areas where both
configurations show complete conversion, e.g., high Φ
ref
and 1/Pe
ref
, are not indicated
(see Figure 3.15 below). Under such conditions the FTCMR is likely to exhibit superior
performance, e.g., in terms of enhanced resistance to deactivation, as previously
discussed.

0.0
0.2
0.4
0.6
0.8
1.0
02 46 8 10
Φ ref
1/Pe ref
Alpha = 0.006
Alpha = 0.05
Alpha = 0.5

Figure 3.12a: FTCMR superiority region - effect of thickness (50 Å, P
ratio
= 1.03)

135
0.0
0.2
0.4
0.6
0.8
1.0
02 46 8 10
Φ
ref
1/Pe ref
Alpha = 0.006
Alpha = 0.05
Alpha = 0.5

Figure 3.12b: FTCMR superiority region - effect of thickness (50 Å, P
ratio
= 1.14)

0.0
0.2
0.4
0.6
0.8
1.0
02 46 8 10
Φ ref
1/Pe ref
Alpha = 0.006
Alpha = 0.05
Alpha = 0.5

Figure 3.12c: FTCMR superiority region - effect of thickness (50 Å, P
ratio
= 1.68)

136
0.01
0.05
0.09
0.13
0.17
0.21
0.25
0.0 2.0 4.0 6.0 8.0 10.0
Φ
ref
1/Peref
Pressure ratio = 1.03
Pressure ratio = 1.14
Pressure ratio = 1.68
α =0.05

Figure 3.13: FTCMR superiority region-effect of P
ratio

Figure 3.14 depicts the effect of pore diameter on FTCMR reactor performance for
three membranes, each having the same thickness ( α = 0.006) but different pore
diameters, namely 50 Å, 100 Å and 500 Å, respectively. For the same feed pressure
(P
ratio
= 1.03), as expected due to the enhanced reactant loss, increasing the pore diameter
shrinks the region in the (1/Pe
ref
vs. Φ
ref
) where the FTCMR shows superior performance.
Figure 3.15 provides additional insight by showing, in detail, the three regions of the
parameter space namely the region where the FTCMR shows superior behavior, the
region where the monolith reactor prevails, and the region where both reactors provide
complete protection.
The impact on FTCMR behavior of the transport limitations imposed by the
membrane support is shown in Figure 3.16. In this Figure the “superiority region” for the
FTCMR in the ( Φ vs. 1/Pe
ref
) parameter space is plotted for three different two-layer
membranes. The top layer for each of these membranes is identical to a 50 Ǻ single-layer
137
membrane with a dimensionless thickness of α = 0.006; this layer sits on a support layer
that has an average pore size of 1.5 µm. The difference in the three membranes is in the
thickness of the support layer (0.5 mm, 1.0 mm, and 1.6 mm or α = 0.25, 0.45, and 0.65
respectively) corresponding to a contribution of the support layer to the total transport of
the membrane of 13 %, 22 %, and 30 %, respectively. Note that the increased support
layer resistance has a negative impact on performance and shrinks the region of FTCMR
superiority over the monolith reactor; this then indicates that for good performance the
resistance of the support should be minimized by either choosing thinner supports, or
supports having bigger average pore diameters.
Figure 3.17 depicts a unique feature of the FTCMR which is absent from the catalytic
monolith, namely the ability of the FTCMR to operate at less than 100 % stage cut, with
a permeate stream with a substantially lower CWA concentration than either the exit
stream from the monolith reactor or the reject stream from the FTCMR. Changing the
stage-cut during FTCMR operation is accomplished by either changing the feed flow-rate
(Pe
ref
) under a constant pressure (like in Figure 3.11a), or by altering the dimensionless
pressure ratio. Since in this study the permeate pressure is kept constant, the latter is
equivalent to changing the feed pressure – in practice (like in Figure 3.11b), alternately,
one can maintain the feed pressure constant and vary the permeate pressure, and the
conclusions do not change qualitatively.

138
0.01
0.03
0.05
0.07
0.09
0.11
0.13
02 4 6 8 10
Φ ref
1/Pe ref
50 Angstrom
100 Angstrom
500 Angstrom

Figure 3.14: FTCMR superiority region-effect of pore diameter

0.01
0.03
0.05
0.07
0.09
0.11
0.13
02 46 8 10
Φ ref
1/Pe ref
FTCMR
superiority
complete
conversion

Figure 3.15: Performance regions for the FTCMR and the monolith reactor (50 Å, α=0.006)
139
0.01
0.02
0.03
0.04
0.05
0.06
024 68 10
Φ ref
1/Pe ref
Alpha = 0.65
Alpha = 0.45
Alpha = 0.25

Figure 3.16: FTCMR superiority region-effect of support thickness

0
100
200
300
400
500
1.0 1.4 1.8 2.2 2.6 3.0
P
ratio
FTCMR shell-side DMMP, (ppm V)
0
10
20
30
40
50
0 2040 6080 100
Stage-cut, (%)
Monolith exit DMMP, (ppm V) 1233
Monolith
FTCMR

Figure 3.17: FTCMR vs. monolith reactor performance

In Figure 3.17, the dimensionless partial concentration of the DMMP (ppm
v
) in the
permeate-side stream from the FTCMR and the corresponding value for the exit stream
of the monolith are plotted as a function of the pressure ratio. The conditions chosen in
140
Figure 3.17 are such so that the FTCMR operating in a total flow-through mode (as well
as the monolith operating under the same conditions) offers no complete protection, the
FTCMR attaining 97.5 % conversion with the monolith attaining 81.0 % conversion
(under such conditions either reactor would have to be accompanied by an additional
polishing stage). As it can be observed from Figure 3.17, under all operating conditions
the permeate-side DMMP concentration (ppm
v
) from the FTCMR is always less than
2.6 % of the DMMP load to the reactor, while for the monolith reactor the exit DMMP
concentration never falls below 19 % of the initial concentration. This points out a
distinct advantage of the FTCMR as it, potentially, provides an additional degree of
functionality (over the catalytic monolith and packed-bed reactors) in terms of screening
and excluding possible reaction by-products, which are often encountered during CWA
catalytic oxidation.

3.6. Conclusions
In this Chapter, the application of the FTCMR concept to the catalytic oxidation, in an
air stream, of dimethyl methylphosphonate which is a chemical warfare simulant for
Sarin was investigated. Preliminary experiments are reported for different DMMP feed
concentrations and reactor temperatures, which demonstrate the potential advantage of
the FTCMR in the complete catalytic oxidation of this important CWA simulant.
A mathematical model has also been developed in order to theoretically support the
experimental findings. The results of the model support the superiority of the FTCMR
concept over the more conventional (plug-flow) monolith reactor.
141
These experimental and theoretical results indicate that the FTCMR concept shows
promise for application in the design of individual-protection systems (e.g., gas masks)
for soldiers and other military personnel, as well as in the design of collective-protection
systems for military armored vehicles and buildings.

142
Chapter 4: Catalytic Membrane Deactivation via Active Site
Coverage and Pore Blockage

4.1. Introduction
During the FTCMR experiments (see Chapter 3) it was found that one of the
challenges associated with the catalytic destruction of DMMP is catalytic membrane
deactivation via active site coverage and pore-blockage. These phenomena appear to be
(see further discussion in this Chapter) the result of phosphorous-containing reaction by-
products interacting with the alumina structure and leading to a gradual decrease, with
increasing time on stream, of the catalytic activity and the average active membrane layer
pore size. As it will be discussed further in this Chapter, these phenomena place
restrictions on the maximum operating temperature that can be utilized and on the
amount of catalyst that can be loaded onto the membranes, as higher reaction
temperatures and higher loadings of Pt seem to lead into more substantial pore blockage.
In a series of experiments during which the air flow through the membrane was kept
constant (via the use of a MFC) and the permeate side pressure was maintained
atmospheric, pore blockage manifested itself as a pressure build-up at the reactor inlet,
thus providing a sensitive indicator of the phenomenon. Since the use of the glass bubbler
(see Chapter 3 for the description of the experimental system) dictated that the maximum
operating pressure is 26 psig, all experiments were terminated when the reactor pressure
exceeded that threshold; this then is referred to as the blockage time, which is the time
when the reactor pressure in these experiments exceeded its safe operating limit of 26
143
psig. Interestingly enough, under conditions for which DMMP conversion was 100 % at
the start of the experiment it often stayed at that level up to the point of pore-blockage
(by the aforementioned definition). As expected, the time for pore-blockage depends on
the average initial pore diameter, with the dual-layer membranes (B- and C-type), with a
smaller average pore size, having a shorter time for pore blockage than the single-layer
membranes (A-type).
A typical “pressure-time” curve, under conditions leading to pore blockage (DMMP
concentration 300 ppm and operating temperature of 623 K), is shown in Figure 4.1 for a
B-type (100 Å) membrane. The phenomenon of pore blockage results in a decrease of the
membrane permeance (as shown in Figure 4.1), which is the reason why, in order to keep
the air-flow through the membrane constant, the inlet pressure in the FTCMR (operating
at 100 % stage-cut or total flow-through) must be proportionally increased. One
interesting side effect of this phenomenon is that as a result of the pore blockage the
Knudsen flow contribution to overall membrane transport increases and the FTCMR
destruction activity is potentially enhanced. This, likely, is one of the main reasons that,
despite the fact that the catalytic surface area decreases, the DMMP conversion still
remains quite high.

144
0.0
0.8
1.6
2.4
3.2
4.0
0 100 200 300 400 500 600
Time on stream, (min)
Normalized inlet pressure
(P /P 0, dimensionless)
0.E+00
2.E-07
4.E-07
6.E-07
8.E-07
1.E-06
Permeance, (mol/Pa/m
2
/sec)
Normalized pressure
Permeance

Figure 4.1: Permeance decrease during the DMMP destruction at 673 K (DMMP load: 300 ppm)

This Chapter presents experimental observations relating to catalytic membrane
deactivation due to active site coverage and pore blockage during the thermocatalytic
decomposition of DMMP. In order to gain knowledge about these phenomena analytical
methods like SEM/EDAX, FTIR, and nitrogen adsorption (BET analysis) are applied.
Also, the theoretical model developed in Chapter 3 for the steady-state conditions, is
modified to incorporate the impact of catalyst deactivation and pore blockage on the
performance of the reactor.

4.2. Background
In the literature review presented in Chapter 1 of this report, catalyst pore-blockages
was noted during the catalytic destruction of DMMP by Tzou and Weller (1994) and by
Cao et al., (2000, 2001). These researchers had attributed this phenomenon to the
formation of P
2
O
5
/H
3
PO
4
and the subsequent reaction of phosphoric acid with the
145
catalyst’s alumina support to form AlPO
4
or by the deposition of P
2
O
5
inside the pores of
catalysts because of its high sublimation temperature of 623 K. Catalyst deactivation (i.e.,
the decline in destruction activity with time on stream) was also reported in virtually all
the previously published articles on the catalytic destruction of DMMP, reviewed in
Chapter 1. Among these research groups, several [Graven et al., 1966; Lee et al., 1994;
O’Shea et al., 1997; and Segal et al., 2001] attributed the decline in catalytic activity to
the formation of H
3
PO
4
.

Figure 4.2: Poisoning deactivation curves [Acres et al., 1975]

In fact, the pore-blockage and deactivation phenomena observed during the catalytic
destruction of DMMP are typical of what one observes during the catalytic destruction
and decomposition of many other P-containing compounds. As a result, the generation of
P
2
O
5
/phosphoric acid or other phosphorous by-products during the catalytic reactions of
146
P-containing compounds over Pt/Al
2
O
3
catalysts (and their reactions with the alumina
support) has been the topic of numerous studies during the past four decades. Acres et al.
(1975) studied poisoning of platinum catalysts by phosphorus and lead compounds
(tributyl o-phosphate and tetraethyl-lead) present in a simulated internal combustion
engine exhaust gas stream. They observed that the rate of catalyst deactivation was
directly proportional to the poisoning compound concentration. Moreover, they found out
that the poisoning process followed a first order kinetics, its rate increased with higher
catalyst bed temperatures, and the accumulation of poison on the catalyst was directly
proportional to the local concentration of the poison precursor in the gas phase. Using an
electron probe microanalysis technique, they determined that phosphorus had
accumulated on the periphery of the wash-coat at the gas-solid interface. This observation
was also consistent with the deactivation curve due to phosphorus (Figure 4.2) which is
typical to what is observed with selective pore mouth poisoning. Acres et al. (1975)
attributed the loss in catalyst activity to the progressive poisoning of active sites at the
pore mouths, which meant that the reactant had to diffuse deeper into the pore structure to
find the progressively fewer active sites to react.
McArthur (1975) studied the degradation of exhaust NO
x
catalysts by poisons
comprising of phosphorous-containing compounds originating in the engine oil and fuel.
He used electron probe and XRD techniques in order to characterize contaminant
concentration profiles in both particulate as well as monolithic NO
x
destruction catalysts.
The major contaminants observed were calcium, lead, phosphorus, and zinc. His studies
showed that most (>90 %) of the poisons (Ca, Zn, Pb, P) present on contaminated
catalysts were confined, principally, to the wash-coat. However, contaminant poisons
147
were also present, to some extent, in the cordierite substrate. The high phosphorus
concentration in the external surface region of the wash-coat indicated that chemical
interaction between phosphorus and the wash-coat material had occurred in order to form
hydrothermally-stable phosphates (e.g., Al(PO
3
)
3
and Al(PO
4
)). XRD analysis failed to
confirm the presence of crystalline aluminum phosphates, so he concluded that they were
present as amorphous compounds.
McArthur (1975) also noted a significant difference in the distribution of phosphorus
in the catalyst wash-coat between the front and rear (outlet) monolith core sections. In the
front section, the phosphorus was distributed uniformly throughout the wash-coat, the
concentration at the external surface of the wash-coat being somewhat higher. In the
outlet core section, the phosphorus was concentrated very heavily at the external surface
of the wash-coat. There was little phosphorus in the interior of the wash-coat, and its
concentration at the interface between wash-coat and cordierite substrate was greater than
in the wash-coat interior. This pattern indicated that phosphorus compound reactivity
toward the catalysts varied considerably between the inlet and outlet ends, perhaps
because of changes in exhaust gas composition or differences in the contaminant
concentrations on the catalyst between these two points. Alternatively, there could be
several gas phase phosphorus compounds present with different reactivities, the most
reactive being removed in the inlet portion of the catalyst. McArthur (1975) suggested a
mechanism for the interaction between phosphorus and the catalyst that involved
adsorption of P
4
O
10
and/or P
4
O
6
, followed by their reaction on the surface with adsorbed
H
2
O to form H
3
PO
4
, and then followed by reaction between H
3
PO
4
and the catalyst to
form hydrothermally-stable phosphate compounds.
148
Hegedus and Summers (1977) studied the resistance of Pt/Al
2
O
3
catalysts to
phosphorous- and lead-containing poison compounds present in commercial fuels which
could react both with the active component as well as the catalyst support. In the first set
of their experiments, four Pt/Al
2
O
3
catalysts of varying macropore structure (and
effective diffusivity) were exposed to the poisons. Electron microprobe analysis indicated
that the poisons had penetrated the catalyst pellets in the form of sharp progressive shells.
Moreover, their studies revealed that the activity of the poisoned catalysts was higher if
their initial diffusivity was larger, which meant that a larger pore could show a better
catalyst performance. In another series of experiments, they investigated the effect of the
catalyst impregnation profile (depth) on the poison resistance of Pt/Al
2
O
3
catalysts by
studying porous alumina pellets impregnated to a depth of about 42 μm below their
surface. They observed a sudden drop in activity after ~30 hr of exposure time. Electron
microprobe analysis of the poisoned catalyst samples revealed that, indeed, the poison
front had reached the noble metal impregnation depth at the time the activity drop was
observed. Due to the high space velocity (110,000 hr
-1
) and the corresponding flat poison
profiles along the reactor’s length, the activity drop was reasonably sharp so that the
poison reached the impregnated depth approximately at the same time for most of the
catalyst pellets in the reactor. Though using catalysts with bigger pores seemed to
enhance catalyst performance, large pores could also increase the rate of poison
penetration. Increasing the surface area of the support could also have a beneficial effect,
as it would increase the saturation concentration of the poisons in the catalyst, and since
the support could react with the poison precursors it could also act as an effective “trap”
for them. Hegedus and Summers (1977) reported that by modifying the pore structure,
149
support surface area, and noble metal impregnation depth of these catalysts,
improvements in activity and poison resistance could be attained for automobile exhaust
emission-control applications.
Hegedus and Baron (1978) investigated steady-state phosphorus accumulation in
automotive catalysts in a multistage integral reactor mounted on an engine dynamometer
system. Electron microprobe studies showed that phosphorus accumulation was
controlled by pore diffusion and that the phosphorus tended to form a monolayer over the
pore surfaces, which was consistent with results of diffusivity measurements that showed
little change in the effective diffusivity upon poisoning. Phosphorus concentration
profiles along the pellet radius (Figure 4.3) resembled a step function. Hegedus and
Baron (1978) believed that the poison precursor (presumably H
3
PO
4
) upon entering the
pores of the catalyst had preferred to diffuse past the already poisoned layer to react with
the support (and also with the active component). Phosphorus accumulation was shown
to be irreversible and independent of the presence of an active metal component over the
alumina support.

Figure 4.3: Phosphorous concentration profiles along the radius of a poisoned catalyst
[Hegedus and Baron, 1978]
150
Angele and Kirchner (1980a) studied poisoning of Pt gauzes, Pt-black, as well as Pt
supported on alumina and silica by a triethyl-phosphate compound in a differential fixed-
bed reactor. Platinum gauzes were completely deactivated by even very small quantities
of poison, though their appearance remained unchanged, metal-bright. Complete
regeneration was achieved by washing with water and also by annealing at 1100 K.
Poisoning and regeneration were repeated several times without the platinum gauzes
changing their appearance. The Pt-black catalyst behaved similar to the platinum wire, in
that it was also very sensitive to poisoning. The quantities of poison in this case were
however sufficient to cause readily-detectable acidification of the wash water. Annealing
was not possible, as this could cause sintering of the Pt-black. After the catalyst had been
poisoned and washed a number of times, it had become mechanically brittle. The
supported catalysts were found far less sensitive to poisoning, with considerable
quantities of phosphorus needed to achieve complete deactivation. Initially, the rate by
which the catalyst took up phosphorus was very high, but decreased as its content
increased. Attempts to wash the phosphorus out of the catalysts produced varying results,
depending on the temperature at which poisoning had taken place. When the catalysts
were heated to over 800 K, the phosphorous could not be washed out. When poisoned at
lower temperatures (<700 K), it was possible to wash out a considerable fraction of the
poison (phosphate ions were found in the wash water). Angele and Kirchner (1980a)
concluded that the poisoning of platinum gauze and Pt-black catalysts was due to the
coverage of the surface by cross-linked, glass-like coatings of P
2
O
5
or polymeric
phosphoric acids which could be hydrolyzed and washed off with water (or vaporized by
151
heating-- the boiling point of an azeotropic system of 94 % P
2
O
5
and 6 % H
2
O is 1140
K).
The alumina supports, on the other hand, could form various acidic and basic
phosphates. Angele and Kirchner (1980a) reported that these phosphate coatings were a
number of monolayers thick, and for Al
2
O
3
, if the temperature was sufficiently high and
the reaction time long these coatings were AlPO
4
. Though AlPO
4
was sparingly soluble
in water, it did dissolve in acids. As a result, poisoned catalysts prepared on α-Al
2
O
3

supports could be regenerated by washing with acid, but those on γ-Al
2
O
3
or amorphous
SiO
2
supports were destroyed. Phosphorus reduced the porosity of the catalyst as well as
the surface area (Figure 4. 4). In fact, the mean pore radius increased as the phosphorus
content increased as the smallest pores (which contributed the most to the internal surface
area but only little to total porosity) were filled first. Angele and Kirchner (1980a)
reported that the poisoning of noble metal supported catalysts by phosphorus compounds
under oxidizing conditions consisted of two stages: a “reversible” process due to physical
coverage of the surface of the metal and the support by P
2
O
5
glasses or condensed higher
phosphoric acids, and another, “irreversible” process due to chemical reaction of the
carrier (Al
2
O
3
, SiO
2
) with the poison to form phosphates. The deposits from the
reversible stage could be removed by washing with water or by heating and sublimation.

152

Figure 4.4: Decrease of the BET-surface for different catalysts with increasing poison content (left),
Apparent increase of pore radii with increasing for two different catalysts (right)
[Angele and Kirchner, 1980a]

In a companion publication, Angele and Kirchner (1980b) determined the poisoning
kinetics of a Pt/Al
2
O
3
catalyst by triethyl-phosphate during the total oxidation of ethane
by air in a differential fixed-bed reactor. Their experimental results indicated a highly
pronounced reaction inhibition by diffusion during catalyst poisoning. A mathematical
model was also developed, which was shown to properly describe the experimental data.
The mathematical model assumed (i) that the poisoning reaction was first-order with
respect to both the poison and the adsorption sites; (ii) that the main reaction was first
order with respect to the concentration of C
2
H
6
and the active sites; (iii) that there was a
linear relationship between the quantity of poison deposited and the number of active
sites destroyed (for the case of phosphorus poisoning for which, in addition to active site
coverage, pore constriction also occurs, this assumption may not be necessarily valid);
153
(iv) that the poisoning reaction was irreversible and so slow that the system could be
considered quasi-stationary.
Fitz and Rase (1983) studied P-containing Ni-Mo/Al
2
O
3
hydrodesulphurization (HDS)
catalysts with varying amounts of Mo and P, and a constant Ni/Mo ratio. The catalysts
without phosphorus had greater surface areas than those of Al
2
O
3
, while those with
phosphorus had either the same or smaller area than the Al
2
O
3
(the high-Mo catalyst with
phosphorus having an area that was significantly less). The pore volumes for the catalysts
without phosphorus were the same with that of Al
2
O
3
while for the P-containing samples
they were 10 % to 20 % below those for the Al
2
O
3
samples alone. Fitz and Rase (1983)
suggested that phosphorus addition had caused pore plugging. Phosphorus-containing
catalysts were less susceptible to coking and produced a more hydrogen-rich coke, a
possible explanation lying in the fate of the phosphoric acid on the catalyst. In fact, γ-
alumina is an acidic support containing exposed aluminum atoms, which may act as
Lewis acid sites, and hydroxide groups bonded to tetrahedral Al ions, which act as
Brønsted acid sites. The phosphoric acid molecule is also tetrahedral in structure,
containing three hydroxide groups and one lone oxygen atom. As shown in Figure 4.5, an
acid-base reaction may occur between a hydroxide group bonded to the alumina surface
and one of the hydroxide groups of the phosphoric acid. Either one of these groups may
donate a hydrogen ion to the other, split out a water molecule, and result in the formation
of an oxygen bond between the alumina support and the phosphoric acid. If the
phosphoric acid bonds to the alumina at only one point, surface acidity will increase,
since one of the available acidic hydrogen from the hydroxide group has been replaced by
two available acidic hydrogens from the phosphoric acid. However, the phosphoric acid
154
molecule should be capable of multiple bonding, as shown in Figure 4.5, where as many
as three bonds may be formed with three different surface hydroxide groups. The
formation of three bonds would result in the loss of three acidic hydrogens, and the
formation of two bonds would result in the loss of acidic hydrogen.

Figure 4.5: Interaction of phosphoric acid with alumina surface hydroxide groups: single group (top),
multiple groups [Fitz and Rase, 1983]

Fitz and Rase (1983) theorized that when phosphoric acid was first added to the catalyst,
most likely a large number of multiple bonds were formed with the alumina support,
which rapidly decreased the availability of unreacted surface hydroxide groups. The
addition of more phosphoric acid could then result in the formation of a higher fraction of
single bonds. A decrease in surface acidity with increasing phosphorus content was
expected until a minimum surface acidity was reached. At concentrations above this
155
phosphorus level the surface acidity began to rise because only single bonds would be
formed.
Mundschau and Vanselow (1986) studied phosphorous adsorption on Pt surfaces using
field electron emission microscopy (FEM). During their experiments, thermally stable Pt-
P layers formed and could not be removed at temperatures below the sintering
temperature of most Pt catalysts. Considering the known chemical stability of PtP
2
(e.g.,
it was unchanged even after 24 h in boiling aqua regia) the P-containing layer could also
be expected to be extremely difficult to remove chemically. The very stable phosphorus
adsorption also caused a change in the crystal form and activity. Some potentially very
important catalytic sites were, thus, eliminated or were reduced in number.
Stanislaus et al. (1988) studied the effect of phosphorus on the acidity of γ-alumina
and on the thermal stability of γ-alumina-supported Ni-Mo HDS catalysts. Ammonia
TPD studies showed that the addition of phosphorous had significantly altered the acid
strength distribution of γ-alumina. With increasing phosphorous level the number of
strong acid sites was reduced, while the concentration of medium-strength acid sites
increased. In fact, as previously noted, it is now well established that γ-alumina is an
acidic material containing both Lewis and Brønsted acid sites. The Lewis acid sites are
associated with the exposed aluminum atoms while the Brønsted acid character is
exhibited by the hydroxyl groups attached to tetrahedral aluminum atoms (Figure 4.6).

156

Figure 4.6: Acidic nature of alumina [Stanislaus et al., 1988]

Some authors have also proposed that there are strained Al-O-Al groups on γ-alumina
which have conjugated Lewis acid-base character, as shown in Figure 4.7 [Stanislaus et
al., 1988].

Figure 4.7: Conjugated acidic nature of γ-alumina [Stanislaus et al., 1988]

Adsorption of traces of moisture by the surface of alumina can also lead to the generation
of surface hydroxyl groups (Brønsted acidity). Stanislaus et al. (1988) suggested that
during the incorporation of phosphate anion by the addition of an aqueous solution of
phosphoric acid or ammonium di-hydrogen phosphate followed by drying and calcination,
the reactions shown in Figure 4.8 might take place.

157

Figure 4.8: Possible reactions between γ-alumina and phosphoric acid [Stanislaus et al., 1988]

The formation of multiple bonds and dimeric surface structures depends on the degree
of hydration of the support and the concentration of phosphoric acid. Multiple bonds
between the alumina support and phosphoric acid decreases the availability of surface
hydroxyl groups and the number of exposed aluminum atoms and as a result, the acidity as
well as the strength of the acid sites decrease. When phosphoric acid was added in low
concentrations, the multiple bond formation between phosphoric acid and the alumina
158
surface was high due to the availability of high concentrations of surface hydroxyl groups
on the alumina surface. It was also possible that initially (at low phosphorus levels) the
strong acid sites of the alumina support were affected by the interaction with the
phosphoric acid. With increased phosphorus loading, the reactions with individual
(isolated) surface hydroxyl groups become predominant leading to the formation of
single bonds, as shown in Figure 4.9. Single bond formation increases the surface acidity
of alumina, since one acidic hydrogen atom from the hydroxyl group on the alumina
surface is replaced by two acidic hydrogen atoms from the phosphoric acid (Figure 4.9).

Figure 4.9: Possible single bond formation during the reaction of γ-alumina and phosphoric acid
[Stanislaus et al., 1988]

According to Stanislaus et al. (1988), addition of phosphates progressively poisoned
strong acid sites but higher phosphate levels created medium-strength acid sites, and the
total acidity of the supports increased. Phosphate ions had a very significant effect in
improving the thermal stability of γ-alumina with respect to sintering and phase transition
into α-alumina (the presence of Mo, on the other hand, impaired this stabilizing function).
Morales et al. (1988) studied the effect of phosphorus addition on the acidity and
physical properties of γ-alumina. The change in the specific surface area and pore volume
of alumina upon phosphorus addition is shown in Figure 4.10. A gradual decrease in the
area (~14 %) and pore volume (7 %) was observed up to a phosphorus content of ~2
wt.% beyond which these properties remained constant. Significant modification of the
159
pore texture took place when the phosphorus was adsorbed. Morales et al. (1988)
indicated that phosphorus had two effects when adsorbed on alumina. First, it behaved as
a corrosive agent destroying some of the micropores of the support with a consequent
increase in macroporosity (e.g., the average pore diameter of the original alumina was
95.2 Å, while an alumina containing 8.3 wt.% of P
2
O
5
had a pore diameter of 112 Å).
Second, the phosphorus blocked the micropores, as a result of its interaction with the
hydroxyl groups of alumina.

Figure 4.10: Surface area (A) and pore volume (B) variation [Morales et al., 1988]
160
Figure 4.11 describes the proposed reaction mechanism. Phosphoric acid interacts first
with the basic groups of alumina, generating a water molecule. For each basic hydroxyl
group eliminated by the phosphoric acid, two new acid sites are created. Further addition
of acid progressively titrates the basic sites up to a point where most of the hydroxyl
groups are titrated and a monolayer forms. Once all the basic hydroxyl groups, (OH), are
titrated, the acid hydroxyl groups, (OH), begin to be titrated and interactions between
neighboring phosphorus molecules begin to occur leading to the formation of linear di-
and tri-phosphate chains. From this point on, the number of acid sites cannot increase any
further, because the substitution of two acidic hydroxyl groups on alumina would yield
two acidic sites associated with the phosphorus.

Figure 4.11: Proposed mechanism of phosphorous adsorption on alumina [Morales et al., 1988]
161
Busca et al. (1989) investigated the structure of the species arising from phosphoric
acid impregnation on silica, alumina, and titania by TG/DTA analysis, XRD, and FTIR
spectroscopy (with CO as a probe molecule). The interaction between phosphoric acid
and silica was weak, producing mainly liquid-like species, together with covalently
bonded phosphate and hydrogen phosphate species. In the case of H
3
PO
4
on alumina and
titania the persistence of Lewis sites, and the almost total poisoning of the basic sites
pointed to a “selective” interaction of the phosphoric acid with the support surface. Also,
the absence of H-bonded phosphate species even after hydration was interpreted as
evidence of the formation of “localized” monomeric species bonded ionically onto the
support surface containing POH groups. The authors attributed the stronger interaction of
titania and alumina towards H
3
PO
4
when compared to silica, to their basicity and much
lower tendency of H
3
PO
4
to condense on these surfaces and to form H bonds.
A series of P-modified γ-Al
2
O
3
catalysts, prepared by impregnation with phosphoric
acid, were characterized by XRD, SEM, and electrophoretic migration measurements by
Lopez Cordero et al. (1989). At P
2
O
5
contents below 2.5 wt.%,

phosphorus was
homogeneously distributed on the alumina surface, but at higher contents inhomogeneity
increased. The surface areas of samples decreased almost linearly with increasing
phosphorus content. BET analysis indicated that phosphorus incorporation in alumina
caused a decrease in the micropores and an increase in the mesopores, reflected by an
increase of the average pore radius. Lopez Cordero et al. (1989) attributed the surface
area loss to the loss of alumina microporosity, caused primarily by fusing of pore walls
through phosphate links, and also by erosion of the carrier material in those regions, the
resulting fragments also blocking the pores. P-containing species were found to form thin
162
films on the alumina particles, giving a strong SEM contrast like gold films. A further
increase in the phosphorus content turned the surface smoother with most of the
interparticle cavities becoming filled with phosphorus species, and with many particles
appearing to be glued.

Figure 4.12: Proposed mechanism for protonation of alumina hydroxyls by phosphoric acid
[Lewis and Kydd, 1991]

Lewis and Kydd (1991) investigated a series of P/Al
2
O
3
catalysts with phosphorus
loadings in the range [0-7 wt.%] using IR spectroscopy. Addition of phosphoric acid
resulted in a reduction of all hydroxyls other than the most basic ones, which increased to
a maximum intensity at ~1 wt.% P. Lewis and Kydd (1991) believe that during the
preparation procedure, the phosphoric acid protonates the alumina hydroxyls by an acid-
base type reaction, see Figure 4.12. Water was lost either before or during the calcination
process, which resulted in the formation of Al-O-P bonds and in the elimination of the
bonds associated with the O-H stretching vibration. The mechanism they proposed for the
adsorption of phosphoric acid on alumina involved the different alumina hydroxyls
reacting with the H
3
PO
4
preferentially (depending on their basicity) up to a critical
surface concentration of ~10 × 10
13
H
3
PO
4
molecules/cm
-2
. Above this concentration,
H
3
PO
4
could react not only with the Al-OH groups but also with the P-OH groups to form
polyphosphate species. The intensity of the band associated with the most basic alumina
163
hydroxyls showed a maximum, since at low phosphorus loadings, these sites might be
generated by the reaction of H
3
PO
4
with bridging and triply-bridging alumina hydroxyls,
but were not formed by this process at higher loadings.
Nooney et al. (1998) examined the kinetics of phosphate uptake by hematite, titania,
and alumina by exposing freshly prepared thin films to phosphate solutions and analyzing
their surfaces by Auger Electron Spectroscopy (AES), TPD, and AFM. Thin hematite
films exposed to a sodium phosphate solution demonstrated initially rapid phosphate
chemisorption (for the first 10 min), followed by island growth. For titania and alumina
exposed to a calcium phosphate solution, the initially rapid reaction was completed after
1 and 3 h, respectively. Subsequent rapid three-dimensional growth occurred after 3-4 h
for titania and 25-30 h for alumina. AFM data at the 10-20 µm scale did not show distinct
changes between the clean and phosphate-covered alumina and hematite surfaces;
however, structures 200-500 nm in diameter could be observed on these surfaces after
exposure to phosphate solution.
Araujo et al. (2006) investigated Al
2
O
3
and H
3
PO
4
/Al
2
O
3
catalysts during the
conversion of oleic acid to biofuels and biolubricants at 1 atm and at 623 K. Two
different alumina supports were used: one commercial and one laboratory-made. BET,
XRD, 31-P NMR and FTIR spectroscopies were employed to evaluate the textural,
structural and acidic properties of the catalysts. BET analysis indicated that impregnation
of phosphoric acid on alumina samples had reduced their surface areas considerably, both
supports clearly undergoing partial pore blockage. For the commercial Al
2
O
3
catalyst,
impregnation with acid markedly reduced its mesoporous volume, indicating that
phosphoric acid might be located mainly on the inner surface of the Al
2
O
3
mesopores. An
164
increase in the average pore diameter of Al
2
O
3
was also observed as a result of
phosphoric acid impregnation. A comparison of the laboratory made Al
2
O
3
before and
after impregnation revealed a reduction of the total pore volume, but an increase in the
mesoporous volume after the addition of the acid. Similarly to the commercial catalyst,
the average pore diameter had also increased after H
3
PO
4
impregnation. The IR spectra
showed no qualitative change after the incorporation of phosphoric acid on the supports.
Nevertheless, compared to the pure alumina, most of the band intensities decreased, with
a slight band shift towards the higher frequencies occurring at 1615 cm
-1
. The decrease in
the peaks' intensities may be ascribed to the lower acid density of these impregnated
samples. With regard to the band at 1615 cm
-1
, the higher this vibrational frequency is the
stronger are the Lewis acid sites in the sample, indicating that the Lewis acid sites of
impregnated samples were stronger than those of pure alumina. This could be attributed
to the higher electronegative feature of the P-alumina bonding and also by the decrease in
the Al electronic density, resulting in an increase of the acidity of the Lewis sites.
According to the P-NMR data, no signal was detected at 0 ppm, indicating the absence of
free phosphoric acid in these samples. The spectra signals (- 18 and at - 28 ppm in the
commercial catalysts) were attributed to tetra-coordinated phosphorus species in bridging
structures, as depicted below:

The peak at - 12 ppm in the laboratory-made catalyst was possibly related to linear
phosphorous and aluminum bonded species, as depicted below:
165

Kroger et al. (2007) studied the effect of phosphorous poisoning on the catalytic
activity of diesel exhaust gas catalysts containing different oxides including ZrO
2
, Al
2
O
3
,
CeO
2
, and TiO
2
and a precious metal, i.e., Pt. They tested the activity of the catalysts by
measuring the light-off temperature of propene before and after exposure to a
phosphorous compound, namely (NH
4
)
2
HPO
4
. Their results showed that phosphorous has
a significant deactivating effect on most samples: the least increase in light-off
temperature was for Pt/CeO
2
(40
o
C) and the most was for Pt/ZrO
2
(76
o
C). For alumina
and titania the change in light-off temperatures were 64
o
C and 65
o
C respectively.
Kroger et al. (2007) applied XRD to analyze their aged catalysts, and they found that in
the case of Pt/Al
2
O
3
, phosphorous appeared as AlPO
4
, and in Pt/TiO
2
as TiP
2
O
7
(this is
the only study that reports that XRD is capable of detecting the P-containing deposits;
studies by other groups, including ours, have found them to be XRD-amorphous). The
FTIR studies confirmed their XRD result by indicating the presence of AlPO
4
in the aged
catalyst
To summarize the previous discussion:
- Organic phosphorus compounds show, in general, relative low thermal stability
and decompose relatively quickly, especially under oxidation conditions; they are
either converted into P
2
O
5
or (if water is present either in the flowing reactants or
on the catalytic material) into condensed phosphoric acid.
- Poisoning reaction is of first order and pore diffusion-controlled.
166
- The activity of poisoned Pt sites is recoverable by oxidizing the poisoned material
at high temperatures under the presence of oxygen, or by washing it with water;
however, the reaction of phosphorous compounds with some of the support
materials (e.g., γ-Al
2
O
3
) is irreversible.
The remainder of this Chapter is devoted to the description of our studies performed to
investigate pore blockage during the thermocatalytic destruction of DMMP in the
FTCMR. Since the FTCMR operates in the forced-flow mode pore blockage directly
affects the flow field of the reactor and potentially the effectiveness of the application.
This is an important issue which has not received significant attention in previous
FTCMR investigations.

4.3. Experiments
In Figure 4.13 experimental data describing the effect of membrane pore size on the
rate of membrane pore blockage are presented for two different membranes, a type A
(500 Å) and a type C (40 Å) membrane each having the same amount of catalyst (the
corresponding DMMP conversion data are shown in Figure 3.8 in Chapter 3). Figure 4.13
shows the inlet pressure into the FTCMR, operating under total flow-through (100 %
stage-cut) conditions under a constant feed flow rate. Note that for the 500 Å membrane
there is little change in the inlet pressure indicating no impact on the membrane transport
properties (permeance) during DMMP catalytic decomposition. Note, however, that as
Figure 3.8 indicates from a certain time on stream and beyond this membrane no longer
provides complete protection, indicative of a diminishing catalytic activity.
167
On the other hand, for the 40 Å membrane significant changes in the membrane
transport properties are observed, as manifested by the gradual increase right from the
start of the feed pressure (since in this experiment the permeate pressure is kept constant
the average membrane permeance is inversely proportional to the feed pressure). This is
probably due to the accumulation and build-up of phosphates within the porous structure
of the catalytic membrane. Interestingly, as Figure 3.8 indicates, this membrane continues
to provide complete protection despite the significant decline in the average membrane
permeance (more than a factor of 2). Another interesting observation is that the change in
membrane permeance slows down significantly after an initial period with an almost
linear rate of decrease (increase in feed pressure).

0
1
1
2
2
3
3
0 100 200 300 400 500
Time on stream, (min)
P /P 0, (dimensionless)
500 Angstrom
40 Angstrom

Figure 4.13: The effect of membrane pore size on the rate of pore blockage manifested by the reactor
inlet pressure build-up (reactor temperature: 573 K, DMMP load: 300 ppm, feed flow: 1 sccs)

168
In order to further investigate these phenomena, three A-type catalytic membranes
were prepared containing different amounts of Pt. To accomplish this, each membrane
was impregnated following the impregnation procedure described in Chapter 2, but using
a different Pt precursor solution for each membrane with a different Pt precursor
concentration, namely 0.5 wt.%, 1 wt.%, and 5 wt.%, resulting in membranes containing
0.0019 g, 0.0031 g, and 0.0113 g of Pt, respectively. The results of the DMMP catalytic
decomposition experiments at a temperature of 573 K and a DMMP feed concentration of
300 ppm are presented in Figure 4.14, which shows the reactor inlet pressure (normalized
by initial inlet pressure at the start of the experiment) and Figure 4.15 which shows the
conversion with respect to time on stream. These results indicate that increasing the
amount of Pt deposited inside the membrane pores has a beneficial effect resulting (for
the membrane prepared with the 5 wt.% solution) in very long protection times. On the
other hand, increasing the amount of Pt in the pores results in the faster blockage of the
pores. For example, while for the membranes impregnated with the 5 wt.% solution Pt
solution pore blockage begins even after than one hour on stream, for the membranes
impregnated with the less concentrated solution the apparent pore blockage was much
less pronounced. As is the case with the behavior shown in Figure 4.13 (and Figure 3.8),
for the membrane prepared with the 5 wt.% Pt precursor solution, while its permeance
declines as the pores are gradually being filled with phosphorous containing deposits, the
membrane continues to provide complete protection, with conversion remaining at 100 %
throughout the experiment.

169
0.3
0.7
1.1
1.5
1.9
2.3
0 300 600 900 1200
Time on stream, (min)
P /P o, (dimensionless)
0.5 wt % Pt Precurser
1.0 wt % Pt precurser
5.0 wt % Pt precurser
Bare membrane

Figure 4.14: Reactor inlet pressure change during the DMMP oxidation with different Pt-loaded
FTCMRs (reactor temperature: 573 K, DMMP load: 300 ppm, feed flow: 1 sccs)

0
25
50
75
100
0 300 600 900 1200
Time on stream, (min)
DMMP conversion, (%)
Bare Membrane
0.5 % Wt Pt Solution
1.0 % Wt Pt Solution
5.0 % Wt Pt Solution

Figure 4.15: DMMP conversion during the reaction with different Pt-loaded FTCMRs
(reactor temperature: 573 K, DMMP load: 300 ppm, feed flow: 1 sccs)

170
In another set of experiments, two B-type (100 Å) catalytic membranes were prepared
following the impregnation procedure described in Chapter 2 using a 1 wt.% Pt precursor
solution. After preparation the weight of metal deposited in each membrane was
measured and was found to be ~0.034 g. The first membrane (referred to as NB22) was
used in experiments carried out in the FTCMR mode operating at total flow-through or 100
% stage-cut; other operating conditions included a reactor temperature of 623 K, a feed
flow rate of 0.5 sccs, and a DMMP feed concentration of 500 ppm, all chosen as to
accelerate pore plugging. Under such conditions the membrane operated for a period of 6
hr before the pressure rose to the 26 psig level that dictated shutting the system down
because of safety concerns about the bubbler as previously outlined. The second membrane
(referred to as NB26) was used in experiments carried out in the “monolith mode”, which
in this case means a 0 % stage-cut; the other operating conditions were kept exactly the
same as in the FTCMR (100% stage-cut) experiments with membrane NB22. After the
same time on stream (6 hr) with the FTCMR experiments, the monolith mode (0 % stage-
cut) in the NB26 experiments was switched into the FTCMR mode of operation (100 %
stage-cut) by opening the permeate side exit while simultaneously closing the reject-side
exit, to soon discover that the inlet pressure to the reactor was 26.3 psig, forcing the shut-
down of the reactor out of safety concerns for the bubbler operation. It should be noted that
throughout the operation both membranes provided complete protection (~100 %
conversion).
Upon completion of the experiments the membranes were removed from the reactor and
their weight were measured to discover that it had changed by 0.0345 g and 0.0299 g for
NB22 and NB26, respectively. The membranes were then were analyzed by the nitrogen
171
adsorption/desorption technique (using a Micromeritics ASAP-2010 instrument). The same
membranes were also analyzed by the same technique prior to their exposure to DMMP. A
summary of the results is presented in Table 4.1. There are small but noticeable differences
in both the measured surface area as well as the pore volumes. Specifically, for membrane
NB22 the surface area changes little but the pore volume changes by 5.4 %. For membrane
NB26, the surface area changes by 2.1 % and the pore volume changes by 1.7 %.

Table 4.1: Nitrogen adsorption/desorption results for fresh and aged catalytic membranes
Catalyst BET
Surface area, (m
2
/g)
BET
Average pore diameter, (Å)
BJH
Cumulative volume of
the pores*, (cm
3
/g)
NB22 (fresh) 0.4190 83 0.001193
NB22 (aged) 0.4184 74 0.001129
NB26 (fresh) 0.3939 75 0.000990
NB26 (aged) 0.3855 73 0.000973

* Average of the adsorption/desorption data

That such small changes in pore volume are associated with such significant changes in
permeance is consistent with a pore-mouth plugging mechanism under these conditions of
relatively high temperatures and small average pore sizes.
The changes in pore size distribution (PSD), calculated using the BJH method from the
adsorption branch, for both the fresh and the aged catalytic membranes are shown in Figure
4.16 (for the NB22 membrane) and Figure 4.17 (for the NB26 membrane).

172
0.E+00
2.E-06
4.E-06
6.E-06
8.E-06
1.E-05
1.E-05
1.E-05
0 50 100 150 200 250 300 350 400 450 500
Pore size, (Å)
Pore volume, (cm³/g-Å)
Deactivated catalyst (FTCMR)
Fresh catalyst (FTCMR)

Figure 4.16: Reduction of membrane pore size due to the plugging of the pores by phosphate poisons
(NB22, FTCMR operation)

0.E+00
1.E-06
2.E-06
3.E-06
4.E-06
5.E-06
6.E-06
7.E-06
8.E-06
0 50 100 150 200 250 300 350 400 450 500
Pore size, (Å)
Pore volume, (cm
3
/g-Å)
Deactivated catalyst (monolith)
Fresh catalyst (monolith)

Figure 4.17: Reduction of membrane pore size due to the plugging of the pores by phosphate poisons
(NB26, monolith operation)

Both Figures indicate the substantial impact phosphorous deposition has on the pore
network structure of the top mesoporous catalytic layer, with the PSD shifting towards the
smaller pore sizes (for example, the minimum pore sizes estimated by BJH method for the
173
fresh NB22 and BN26 membranes were 52 Å and 44 Å, respectively; for the aged
membranes the corresponding values were 18 Å and 19 Å, respectively).
Upon the completion of the BET analysis, the NB22 and NB26 membrane samples were
analyzed by SEM/EDAX (JEOL JSM-6610, equipped with an EDAX analyzer). Due to the
technical limitations of the SEM instrument, accumulated phosphates could not be detected
on these small pore size samples (100 Å). However, in a previous experiment involving a
catalytically impregnated α-Al
2
O
3
support membrane (using 8 wt.% Pt precursor solution)
SEM proved capable of distinguishing the presence of P-containing deposits (this
membrane was utilized in DMMP catalytic decomposition experiments at 623 K, with a
DMMP feed concentration of 1000 ppm). Figure 4.18 compares the SEM images of the
latter membrane before and after it had been exposed for 48 hr to the DMMP/air mixture.
One can clearly distinguish a film coating the alumina granules.

Figure 4.18: SEM images of a support membrane before the reaction (left) and after the pore blockage
(right)

174
Figures 4.19-4.21 show the EDAX elemental maps for Pt, P, and C in the top surface
(30 µm thick) of both the NB22 and NB26 catalytic membranes. Note the sharp
distribution of Pt across the membrane thickness which validates the effectiveness of the
catalyst preparation method. Most of the Pt is deposited within the top 8 µm or of the
membrane where the 100 Å and 500 Å layers are located, see further data and discussion
in Chapter 2. Phosphorous, though its distribution is broader than that of Pt seems to also
accumulate, for the most part, in the top 10 - 15 µm region, indicative of its good affinity
to the underlying membrane surface.
Carbon could be due to various reasons. It could associate with the P-containing
deposits (as previously discussed, the technical literature reports that methylphosphonic
acid together with H
3
PO
4
are key end-product during catalytic decomposition of organo-
phosphorous compounds) or signify the presence of coke, or the presence of chemisorbed
and unreacted DMMP. The P and C profiles of NB22 and NB26 are qualitatively similar.

0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 5 10 15 20 25 30
Radial distance, (micron)
Pt/Al net ROI counts my
Monolith
FTCMR

Figure 4.19: Elemental profile for platinum across the aged catalytic membranes
175
0.00
0.08
0.16
0.24
0.32
0.40
0 5 10 15 20 25 30
Radial distance, (micron)
P/Al net ROI counts my
Monolith
FTCMR

Figure 4.20: Elemental profile for phosphorous across the aged catalytic membranes

0.00
0.04
0.08
0.12
0.16
0.20
0 5 10 15 20 25 30
Radial distance, (micron)
C/Al net ROI counts my

Figure 4.21: Elemental profile for carbon across the aged NB26 membranes

XRD analysis of the S-type membranes after they had been exposed to the reaction
conditions failed to confirm the presence of crystalline phosphates, meaning that any bulk
176
or surface phosphorous-containing deposits formed are amorphous to X-ray diffraction.
Such observation is consistent with the technical reports discussed.
FTIR analysis (using a Nicolet-4700 instrument from Fischer-Scientific) was utilized to
further probe the nature of the deposits formed. We studied the aged catalytic support tube
whose SEM image is shown in Figure 4.18. We also examined a fresh support membrane
tube not exposed to DMMP as well as a fresh support tube after it was boiled for 10 min
in a 85 wt % phosphoric acid solution. In Figure 4.22, a comparison is made between the
fresh and acid-boiled membranes revealing that treatment in H
3
PO
4
gives rise to new
peaks in three different regions: between 1000 cm
-1
and 1200 cm
-1
,

between 2000 cm
-1
and 2400 cm
-1
., and finally between 3500 cm
-1
and 3900 cm
-1
. Bautista et al. (1993)
reported that the peak around ~ 1100 cm
-1
corresponds to AlPO
4
. Cao et al. (2000)
assigned the three peaks in the second region to a (P)–O–H stretch, but indicated that the
bands with wave numbers in the third region are difficult to assign, and may be
combination of bands.
In Figure 4.23, a comparison is made between the acid-boiled and the catalytic
membrane sample exposed to DMMP. The overlap between most of the peaks for the two
samples in this figure indicates to the production of H
3
PO
4
during the decomposition of
DMMP over the catalytic membranes. However, in the membrane used during DMMP
decomposition we fail to observe peaks in the region assigned to AlPO
4
, so FTIR will not
provide support for the formation of this species.
177
87
89
91
93
95
97
99
101
200 400 600 800 1000 1200
Wave number, (cm
-1
)
Reflectance, (%)
Acid-boiled sample
Fresh sample

Figure 4.22a: FTIR spectra for an α-alumina membrane before and after exposure to phosphoric acid
(wave number < 1000 cm
-1
)

98
99
100
101
102
900 1300 1700 2100 2500 2900 3300 3700 4100
Wave number, (cm
-1
)
Reflectance, (%)
Acid-boiled sample
Fresh sample

Figure 4.22b: FTIR spectra for an α-alumina membrane before and after exposure to phosphoric acid
(wave number > 1000 cm
-1
)

178
86
89
92
95
98
101
104
200 400 600 800 1000 1200
Wave number, (cm-1)
Reflectance, (%)
Aged catalyst
Acid-boiled sample

Figure 4.23a: FTIR spectra for an α-alumina catalytic membrane after exposure to air/DMMP
(wave number < 1000 cm
-1
)

98
100
102
104
900 1300 1700 2100 2500 2900 3300 3700 4100
Wave number, (cm-1)
Reflectance, (%)
Aged catalyst
Acid-boiled sample

Figure 4.23b: FTIR spectra for an α-alumina catalytic membrane after exposure to air/DMMP
(wave number > 1000 cm
-1
)
179
4.4. Membrane Recovery
Since, as shown before, catalyst deactivation along with membrane pore blockage
phenomena restricts the application of FTCMR concept for the higher temperatures and
catalyst loads, it is important to explore if the spent catalytic membrane can be recovered to
its initial activity (or close to it) for further applications.
Three different membranes (an α-alumina support tube, a 500 Å α-alumina single-
layer membrane, and a 100 Å γ-alumina dual-layer membrane) were exposed to
phosphoric acid. In the experiments, 10 ml of phosphoric acid (85 % wt) was dissolved in
140 ml of de-ionized water. All membranes were then boiled in this solution for 1 hr
followed by washing with water and drying at 523 K overnight. Then they were boiled in
500 cc water for 4 hrs and dried in oven overnight. After that, their transport
characteristics were measured and compared with those of fresh membranes. As can be
seen (see Figures 4.24-4.26) exposure to H
3
PO
4
decreases the permeance of all samples,
the reduction being dependent on the type of membrane used, ranging from 32 % for the
α-alumina support, 52 % for the 500 Å α-alumina membrane and 74 % for the γ-alumina
membrane.
The membranes were subsequently boiled in 200 cc of a dilute (3 wt.%) hydrochloric
acid solution and their transport characteristics were measured again. As can be seen in
Figures 4.24-4.26, for the α-alumina support tube and the 500 Å α-alumina membranes
the permeance after the dilute acid wash was almost the same as that of the fresh
membrane. For the 100 Å membrane the permeance of the acid-washed membrane was
higher than the permeance of the fresh membranes indicating that some damage may have
occurred to the γ-alumina layer during the acid wash.
180
3.E-06
5.E-06
7.E-06
9.E-06
1.E-05
1.0E+05 1.5E+05 2.0E+05 2.5E+05 3.0E+05
Average trans-membrane pressure, (Pa)
Membrane permeance, (mol/Pa/m
2
/s)
After glazing
After reaction with H3PO4
After Acid washing

Figure 4.24: Change of the permeance for membrane support before and after reaction with H
3
PO
4

0.E+00
1.E-06
2.E-06
3.E-06
4.E-06
5.E-06
1.0E+05 1.5E+05 2.0E+05 2.5E+05 3.0E+05
Average trans-membrane pressure, (Pa)
Membrane permeance, (mol/Pa/m
2
/s)
After glazing
After reaction with H3PO4
After Acid washing

Figure 4.25: Change of the permeance for an α-alumina membrane before and after reaction with H
3
PO
4

181
0.E+00
1.E-06
2.E-06
3.E-06
4.E-06
5.E-06
6.E-06
1.0E+05 1.5E+05 2.0E+05 2.5E+05 3.0E+05
Average trans-membrane pressure, (Pa)
Membrane permeance, (mol/Pa/m
2
/s)
After glazing
After reaction with H3PO4
After Acid washing

Figure 4.26: Change of the permeance for a γ-alumina membrane before and after reaction with H
3
PO
4

In a second series of experiments a 500 Å alumina membrane (impregnated with 1.0
wt.% Pt precursor solution) which had been utilized in a FTCMR experiment and which
had been subjected to pore blockage was chosen for regeneration. Three different
methods were applied:
- Heating in an oven at 1000 K for 4 hr
- Passing hot air flow at 573 K for 4 hr
- Boiling in an acid solution (prepared by dissolving 20 ml of 37 wt.% hydrochloric
acid in 720 ml of water) for 5 hr followed by 2 hr of boiling in pure water, and
then drying for 3 hr at 573 K, the sample being reactivated with hydrogen
overnight.
The first two methods could not open the pores, while the last treatment was able to
regenerate the membrane’s pore structure. Figures 4.27 and 4.28 compare the inlet
pressure profiles and DMMP conversion for both the fresh and regenerated membranes.
182
Clearly, the regeneration procedure must be further optimized (e.g., reducing the strength
of the acid solution and/or the treatment time) in order to avoid inadvertent loss of the
catalyst and support material. This is the focus of our current research as the ability to
regenerate the FTCMR by a simple and effective technique is critical in terms of the
eventual practical application of these devices.

0.90
0.95
1.00
1.05
1.10
1.15
0 100 200 300 400 500 600 700
Time on stream, (min)
P /P 0, (dimensionless)
Fresh catalyst
Washed catalyst

Figure 4.27: The effect of acid washing on inlet pressure of a 500 Å membrane
183

40
60
80
100
0 100 200 300 400 500 600 700 800
Time on stream, (min)
DMMP conversion, (%)
Fresh catalyst
Washed catalyst

Figure 4.28: The effect of acid washing on catalytic conversion of a 500 Å membrane

4.5. Theoretical Studies
4.5.1. Model development
When a product of the catalytic decomposition of DMMP (e.g., methylphosphonic
acid or H
3
PO
4
) gets trapped within the membrane structure it may cause two problems:
first coat the active surface sites (which will then prevent other DMMP molecules from
reacting on such sites – in the catalysis literature this is commonly referred to as
poisoning) and also occupy (clog) the membrane pore structure, which then causes a
decrease in the membrane permeance, as the experiments described in this Chapter
indicate. The gradual plugging of the pores, in addition, also impacts the rate of DMMP
decomposition and the performance of the FTCMR as the molecules encounter more
difficulty to diffuse to the active sites and products encounter more resistance in escaping
from the pore structure.
184
The steady state theoretical model discussed in Chapter 3 is not capable to account for
the phenomena of site coverage and pore blockage, and thus its application, as noted in
Chapter 3, is limited to describing the early stages of FTCMR operation. In this Chapter
the focus is on developing a model that is capable of describing the performance of the
FTCMR decomposing DMMP in the presence of trapping of P-containing products in the
membrane structure. The following assumptions were made in deriving this transient
FTCMR model:
(I) The membrane layer pore structure consists of straight, parallel nonintersecting
cylindrical pores, each with the same initial average pore diameter of
0
p
d . The model,
therefore, does not account for pore network interconnectivity effects or with phenomena
associated with pore size distribution. In Chapter 6, accounting for such effects is
recommended as future modeling work by the Group, as they are likely to have a major
impact on the behavior of the FTCMR.
(II) The catalytically active sites are uniformly distributed over the surface of the pore.
(III) Transport and reaction within the membrane porous layer is described by the
Dusty-Gas model (DGM), similarly to the FTCMR model in Chapter 3.
(IV) During the course of the catalytic decomposition reaction, one or more of the
reaction products (e.g., methylphosphonic acid or H
3
PO
4
) deposit on the pore walls –
however, the model assumes that their impacts are similar, but it can be easily modified
to account for such differences, if experimental data were ever to become available that
differentiate among the different types of deposits in the pores. The model assumes that
the decomposition of a single DMMP molecule, at the locality of the pore where it
occurs, coats a certain fraction of the active surface area as well as it clogs a certain
185
portion of the pore volume. The real phenomena are, of course, very complex. For
example, the P-containing species may form on the Pt metal sites or the alumina surface
(which in itself has a certain catalytic activity, as the studies in Chapter 3 indicate) and
they may also react with the alumina surface or deposit on the phosphorus-coated
surfaces. Since, as the review of the technical literature above indicates, the details are
still rather sketchy and little definite is known today about the mechanism; the model
ignores the detailed mechanism and only accounts for overall macroscopic effects.
(V) The phenomena of site coverage and pore-plugging and the decomposition
reaction of DMMP itself are irreversible.
Other simplifying assumptions implemented for the modeling are the same as those
described in Chapter 3.
Mass conservation for the gas component i in a tubular membrane layer can be
expressed as:
 
t
p
RT
Rxn
r
rN
r
i
i
i


 



 1 1
(4.1)
Where N
i
, the gas flux for component i in radial coordinate (r), can be found from the
Dusty Gas Model (DGM) for air (A) and DMMP (D) [Chapter 3]:
dr
dp B
RT
p
RT
D
N
A
e
A
e
K , A
A








  

(4.2)





























    
A
e
e
K A
e
K D
e
DA
e
DA A
A D
D e
D
D
p B
D D
D
D p
N p
p
RT D
N

, ,
1
1
1
1
(4.3)
Where:
186
i
p
K i
e
K i
M
RT
d
D D
 


 8
3
, ,
  , (i =A or D) (4.4)
32
2
p e
d
B


 (4.5)
In the above equation p
i
is the partial pressure,

M
i
the molecular mass, R the universal
gas constant, T the temperature, and
e
DA
D (=
DA
D


) is the effective binary diffusion, D
i
,
k

is the Knudsen diffusion coefficient,
e
D
D (=
D
D


) is the overall diffusivity of DMMP
whose value can be approximated by Equation 3.10, d
p
the average pore size of the
porous layer, ε is the porosity, and τ is the tortuosity, μ the gas viscosity, and Rxn
i

(mol/m
3
.s) is the reaction rate for component i (oxidation reaction for DMMP and zero
for air). The value of binary diffusion is estimated using Fuller’s method (Poling et al.,
2000). Due to the low concentration of DMMP in the air, the catalytic oxidation of
DMMP is assumed to follow a first order global reaction rate expression, and based on
the internal surface area of the membrane active layer; the rate of reaction is expressed
as:
RT
p
S k Rxn
D
g D
    (4.6)
Where S'g is the catalytic surface area per unit volume of membrane (m
2
/m
3
) and k'' is the
rate constant (m/s) (these need to be interpreted as effective parameters as both the metal
and the alumina surface appear to have activity towards DMMP decomposition).
Two overall effective parameters are used to describe the change in the local active
surface area and membrane porosity as a result of DMMP decomposition. They include:
(i) the poisoning factor s' (m
2
/mol) which is interpreted as the active surface area (m
2
)

187
which is inactivated by the P-containing deposits resulting from the decomposition of one
mol of DMMP, and (ii) the pore volume plugging factor b' (m
3
/mol) which is the
corresponding pore volume (m
3
) that such deposits end-up occupying. Though, in
principle, these two factors are interrelated the phenomena that take place are so complex
that it is not entirely clear of how to calculate that relationship, so in this model they are
viewed as two independent adjustable parameters.
The change in the active surface area and membrane pore volume are then described
by the following two equations:

RT
p
S k s Rxn s
dt
S d
D
g D
g
        

(4.7)

RT
p
S k b Rxn b
dt
d
D
g D
        

(4.8)
In modeling the reactors, the transport equations for the FTCMR must be coupled with
the proper mass balance equations for the tube-side and the shell-side. Here, it is assumed
that the tube-side and shell-side show plug-flow behavior, the flows are co-current, and
that the pressure drops are negligible (all these assumptions apply well to the laboratory
reactors). The governing equations for tube-side and shell-side (for DMMP -- similar
equations also apply for the transport of air) are as follows:

t
p
RT
r
rN
z
n
D in
r r D
tube
D
in










  



2
2

 , for the tube-side (4.9)

 
t
p
RT
r r
rN
z
n
D out shell
r r D
shell
D
out









 
 



2 2
2

 , for the shell-side (4.10)
188
The above differential equations must be solved together with the appropriate initial
and boundary conditions, as follows (for the DMMP -- similar equations are also used for
air):
Initial conditions at t=t
0
=0:
0 ,
g g
S S    , at any r (4.11)
0
   , at any r (4.12)
state steady
D D
p p


0
, at any r (4.13)
state steady tube
D
tube
D
n n


, 0 ,
, at any z (4.14)
state steady s
D
shell
D
n n


, 0 ,
, at any z (4.15)
One should note that the above initial conditions are valid only when the required time
to reach an initial steady-state is much smaller than the rate of change for the porosity and
catalyst activity, which was actually demonstrated to be the case during the time-on-
stream experiments. The initial steady-states values required to solve the equations can be
readily obtained by solving the model developed in previous Chapter for simulating
steady-state performance of the reactor. Boundary conditions at any t> t
0
:
tube tube
D D
P y p  , at
in
r r  (4.16)
shell shell
D D
P y p  , at
out
r r  (4.17)
inlet tube
D
tube
D
n n
,
 , at 0  z (4.18)
inlet shell
D
shell
D
n n
,
 , at 0  z (4.19)
Since the pore size of the membrane (and its porosity as well) changes with time, in
order to simplify the equations, two further assumptions are made for the relationship
189
between the local membrane porosity and pore diameter, and for relationship between the
local membrane porosity and tortuosity as follows:
2
0 0









p
p
d
d


(4.20)
j









1
(4.21)
Equation 4.20 assumes that the clogged membrane maintains locally a tubular 1-D
structure, while Equation 4.21 is an empirical relationship commonly utilized, and the
value of the exponent j needs to be estimated experimentally.
The above equations can be made dimensionless by defining appropriate
dimensionless variables and groups, which are listed in Table 4.2, as follows:
D
ratio
ratio
ΩX
T
P b
d
dE
2
0

 








  (4.22)
D
ratio
ratio
ΩX
T
P
s
d
d
2










 

(4.23)
For the carrier gas (air) within the membrane:
 






































    


 E
X
X
E
X E
w
w
X
X w
X
A
A
A
A
A
A
1
2 exp
1
2 5 0 0
2
2
2
(4.24)
And, for DMMP within the membrane:
 























 
















































  




 
E
X
X
E
X Φ
w
E
E
X
w
X
X w
X
w
w
X
w
X
w
w
X
D
D D
D
A
A
A D A D
0
2
2 5 0
0 2 1
2
2 2
2
2
2
2 exp
(4.25)
Where
190










A
e
e
K A
p B
D
w


,
1
1
)( (4.26)
















  ) ( 1 ) (
,
w
D
D
D p
N
w
e
K D
e
DA
e
DA A
A
  (4.27)
Shell/Tube side equations for air:


0
2
0
2 5
0
2
0
Pe 2
1
1
Pe
 

















 

w
A
ref w
A
A
ref ratio
ratio
tube
A
X
w
X
X E
T
P
d
dY
 




(4.28)


1
2
1
2 5
0
2
0
Pe 2
1
1
Pe
 

















 

w
A
ref w
A
A
ref ratio
ratio
shell
A
X
w
X
X E
T
P
d
dY
 




(4.29)
And for DMMP:


0
2
0
0 2 1
2 5
2
0
Pe 2
1
) ( Pe
1

































w
D
ref
w
A
D
D
ref ratio
ratio
tube
D
X
w
X
X w
w
X
E
E
T
P
d
dY
 

 

(4.30)


1
2
1
0 2 1
2 5
2
0
Pe 2
1
) ( Pe
1

































w
D
ref
w
A
D
D
ref ratio
ratio
shell
D
X
w
X
X w
w
X
E
E
T
P
d
dY
 

 

(4.31)
With the initial conditions at 0
0
    :
E= E
0
= 1 (4.32)
Ω = Ω
0
=1 (4.33)
state steady
D D D
X X X

 
0
(4.34)
state steady
A A A
X X X

 
0
(4.35)
tube state steady
D
tube
D
tube
D
Y Y Y
, , 0 
  (4.36)
tube state steady
A
tube
A
tube
A
Y Y Y
, , 0 
  (4.37)
191
shell state steady
D
shell
D
shell
D
Y Y Y
, , 0 
  (4.38)
shell state steady
A
shell
A
shell
A
Y Y Y
, , 0 
  (4.39)

Table 4.2: Dimensionless parameters used for the theoretical unsteady-state model
Dimensionless group
or parameter
Mathematical
expression
Physical definition
α ln (r
out
/r
in
)
-
w ln (r/r
in
)/ α
Dimensionless radial coordinate
Z z/L
Dimensionless longitudinal coordinate
X
i

0 , tube
i
P p
-
Y
i

0 , tube
i
n n
-
Pe
ref

0
0
2
DA ref
ref A
D LP
RT n

Modified Pe ćlet number

Φ
0
/
DA
in
D
k
r 

Modified Thiele module
γ
e
DA
e
K A
D D
,
-
φ 1/X
A
+ ξ + ξσ
-
δ
e
K A
e
i
D D
,
-
ξ

e
K A
A
e
D
p B
,
0

The ratio of the mass of air that transports by convective
flow to that via Knudsen
σ
e
K D
e
DA
D D
,
The ratio of the mass of DMMP that transports by
continuum diffusion to that via Knudsen diffusion
P
ratio

P
tube
/P
shell

Tube-side to the shell-side pressure ratio
(shell-side pressure is assumed to be 101325 Pa = P
ref
)
θ
2 2
in
DA
R
tD


Dimensionless time
b
ref
ref
RT
P b
/

Modified pore plugging factor
s
ref g
ref
RT S
P s
0
/

Modified catalyst deactivation factor
Ω
0 /
/
g g
S S
-
E ε/ ε
0

-
T
ratio
T/T
ref

Dimensionless temperature (T
ref
= 298.15 K)
n
0
ratio

n
shell
,0
/n
tube
,0

Sweep gas to the feed gas molar flow ratio
192
For the boundary conditions at any θ:
At w=0:
tube
D
tube
A
tube
D tube
D
Y Y
Y
X

 , any Z (4.40)
inlet tube
D
inlet tube
D
tube
D
y Y Y   , Z=0 (4.41)
tube
D
tube
A
tube
A tube
A
Y Y
Y
X

 , any Z (4.42)
inlet inlet
A
inlet tube
A
tube
A
y Y Y   , Z=0 (4.43)
At w=1:
shell
D
shell
A
shell
D shell
D
Y Y
Y
X

 , any Z (4.44)
inlet shell
D
ratio inlet shell
D
shell
D
y n Y Y   , Z=0 (4.45)
shell
D
shell
A
shell
A shell
A
Y Y
Y
X

 , any Z (4.46)
inlet shell
A
ratio inlet shell
A
shell
A
y n Y Y   , Z=0 (4.47)
The above set of equations can be applied to model the transport (DMMP and Air) and
reaction (DMMP) within the active layer of the membranes studied. For the membrane
support, it is assumed that no chemical reaction is possible due to the absence of catalytic
metal (the latter assumption is validated by SEM/EDAX analysis, and the results of
which are reported in this Chapter, and Chapter 2 as well) but also because of the
relatively low surface area per unit volume of the support; therefore, the same set of
equations as above can be used to model the transport phenomena within the support
layer of the membrane by setting the values of the deactivation and pore plugging factors
193
(b and s in Equations 4.22 and 4.23) along with Thiele module ( Φ in Equation 4.25) to
zero. At the interface between two layers:
port sup
A
layer active
A
X X  , (4.48)
port sup
D
layer active
D
X X  , (4.49)
port sup
A
layer active
A
N N  , (4.50)
port sup
D
layer active
D
N N  , (4.51)
The above equations form a system of PDEs in radial (membrane) and longitudinal
(tube/shell) coordinates, and time. They are solved in a MATLAB® program developed
for this purpose.

4.5.2. Model Validation
The above model involves a number of experimental parameters, including the
transport properties of the membrane, the reaction rate constant, the poisoning factor, and
the pore volume plugging factor. An important use of the model is to utilize it to find the
values of these parameters by fitting experimental data. These experimental data were
generated by using a catalytic membrane (hereinafter referred to as membrane NA03)
which was prepared using a A-type sample (single-layer membrane with a nominal
average pore size of 500 Å) via impregnation with a 8 wt.% Pt precursor solution using
the impregnation method described in Chapter 2. The impregnated membrane contains
0.0178 g of Pt (equivalent to 0.41 wt% with respect to the total membrane weight).
The reactor in these experiments was a modified version of the one used for reaction
studies in Chapter 3, redesigned with inlet and outlet ports in the shell-side for passing a
sweep gas. It was operated as a FTCMR with an air sweep in the permeate side of 60
194
sccm and with less than 100 % stage-cut. The feed DMMP concentration was kept
constant at 120 ppm and the feed flow rate at 260 sccm. During operation the inlet feed
reactor inlet pressure was kept constant at a value of 4.6 psig while the trans-membrane
pressure was kept constant as well at a value of 4.5 psig. To accomplish this under
reactive conditions during which the permeance of the membrane was slowly decreasing,
we utilized a needle valve installed downstream of the reject exit of the reactor. With the
aid of this valve the air flow through the membrane was changed gradually in order to
adjust to the changes in the membrane permeance due to the reaction.
As the flow of air through the membrane into the shell-side of the reactor declines
gradually, so does the reactor stage-cut (defined as the ratio of the permeate flow rate to
the feed flow rate). The FTCMR experiments were carried at three different temperatures,
i.e., 508 K, 562 K, and 606 K. During the experiments, the DMMP conversion and the
membrane stage-cut were measured and reported in Table 4.3, and are also plotted in
Figure 4.29 (conversion vs. time plot). In Figure 4.30 the change of the permeance vs. the
average pressure for this membrane is plotted for different times on stream during the
experiment.
We have assumed that the temperature dependence of the reaction rate constant can be
described by the Arrhenius equation [Levenspiel, 1999]:
) R / exp(
0
T E k k   (4.52)
The model is then used to fit these data, using an in-built Matlab
®
function called
“nlinfit” in order to estimate the reaction rate constant (k
0
and E), the poisoning factor (s),
and pore volume plugging factor (b). The confidence interval calculation (95 %) for the
parameters was then performed by applying another in-built Matlab
®
function called
195
“nlparci”. The results of data fitting are presented in Table 4.4, and Figures 4.31 and 4.32
are iso-conversion and iso-(stage-cut) plots indicating the goodness of fit.
For the initial transport properties through the membrane, the actual thickness (1080
µm measured by SEM photos) and the experimental porosity (21 % measured by Helium
pycnometry and the Archimedes method -- see sections 2.2.3 and 2.2.4) were used to fit
the single-gas permeation data (Ar and He – see section 2.2.7) for a membrane support
tube, resulting in the value of 11,000 Å for the average pore diameter and 0.17 for the
geometrical factor (  / ) of the support. Using the latter value with Equation 4.21
estimates the value of j to be 1.2. For the active-layer of the membrane, it was assumed
that the support material on which this active layer was deposited was the same as an
individual support material with known structural properties, as measured above. Also, it
was postulated that the geometrical factor of this α-alumina layer was the same as the α-
alumina used in the support layer, and consequently fitting the air permeation data before
exposure to DMMP (t=0 in Figure 4.30) resulted in a pore size of 300 Å for the active
layer.

196
Table 4.3: Experimental data collected in the FTCMR experiments with the NA03 membrane
Time on
stream,
(s)
DMMP Feed
Concentration,
(ppm)
Feed
Flow-rate,
(sccm)
Reactor
Temperature, (K)
Trans-
membrane
Pressure, (psi)
Stage-cut,
(%)

Conversion,
(%)
0 120 260 508 4.5 26.9 N/A
3600 120 260 508 4.4 26.9 26.1
5400 120 260 508 4.4 25.8 21.1
10800 120 260 508 4.5 25.9 20.1
15000 120 260 508 4.4 24.8 19.5
18000 120 260 508 4.5 24.3 19.9
25200 120 260 508 4.4 22.9 19.2
25200 120 260 562 4.5 21.2 N/A
28800 120 260 562 4.5 20.5 38.9
30600 120 260 562 4.4 20.5 38.7
34200 120 260 562 4.5 20.1 40.1
37800 120 260 562 4.5 19.7 37.7
41400 120 260 562 4.6 19.7 37.2
46800 120 260 562 4.5 19.1 34.1
50400 120 260 562 4.6 18.8 32.7
50400 120 260 606 4.5 16.9 N/A
54000 120 260 606 4.5 16.9 54.1
57600 120 260 606 4.5 16.6 54.0
61200 120 260 606 4.5 16.5 53.0
64800 120 260 606 4.5 16.3 52.0
68400 120 260 606 4.5 16.1 50.0
72000 120 260 606 4.6 15.9 43.7

197
0
10
20
30
40
50
60
0 16000 32000 48000 64000 80000
Time on stream, (sec)
DMMP conversion, (%)
508 K
562 K
606 K

Figure 4.29: FTCMR experiments with the NA03 membrane

0.0E+00
3.0E-07
6.0E-07
9.0E-07
1.2E-06
1.5E-06
1.8E-06
1.0E+05 1.3E+05 1.6E+05 1.9E+05 2.2E+05
Average pressure, (Pa)
Permeance, (mol/Pa/m
2
/sec)
508 K (t=0 sec)
508 K (t=25200 sec)
562 K (t=25200 sec)
562 K (t=41400 sec)
562 K (t=50400 sec)
606 K (t=50400 sec)
606 K (t=64800 sec)
606 K (t=72000 sec)

Figure 4.30: Change of the membrane permeance vs. average pressure during the time on stream
experiments for the NA03 membrane

198
Table 4.4: Fitted experimental parameters
Parameter b

s E
(kJ/mol)
k
0

(1/sec)
Fitted value 0.000218 0.0000768 99 6.0  10^ (+12)
95 % confidence interval ± 0.000037 ± 0.0000035 ± 6 ± 0.4  10^ (+12)

0
10
20
30
40
50
60
0 102030 4050 60
DMMP conversion, (experimental values, %)
DMMP conversion
(model predictions, %)
Model predictions
Experimental values

Figure 4.31: Goodness of fit plot for conversion values

0
5
10
15
20
25
30
35
0 5 10 15 20 25 30 35
Mermbrane stage-cut, (experimental values, %)
Membrane stage-cut
(model predictions, %)
Model predictions
Experimental values

Figure 4.32: Goodness of fit plot for stage-cut values
199
Table 4.5 summarizes prior kinetic studies and the reported reaction rate constants
(pre-exponential factor and activation energy) for destruction of DMMP and similar
organophosphorus compounds (OPCs). The rate data in Table 4.5 include information
gleamed from two studies on the heterogeneous destruction of DMMP (the first on
concrete and the other on Pt/Al
2
O
3
), while the rest of the data are from studies of the
homogeneous destruction of OPCs in plug-flow reactors and flames.
In terms of activation energy, the value calculated in this study (99 kJ/mol, see Table
4.4) is greater than the ones reported for the heterogeneous destruction on concrete and
Pt/Al
2
O
3
, while smaller than those reported for the homogeneous destruction in plug-flow
reactors (> 154 kJ/mol). For comparison purposes, for example, the value of activation
energy reported by Graven et al. (1966) for a commercial 0.50 wt.% Pt/Al
2
O
3
catalyst 2.4
mm in diameter is 31 kJ/mol. The smaller activation energy reported by Graven et al.
(1966) may be due to the fact that diffusional limitations and catalyst deactivation effects
were not taken into account when studying the reactor performance. In fact, in terms of
the reaction rate constant for the destruction of DMMP, by considering a simple
deactivation correlation for their laboratory-prepared catalysts, at a temperature of 573 K
they found a value of 543 s
-1
for a batch of catalysts with an average diameter of 0.31 mm
compared to the value of 141 s
-1
for the similar catalyst but with an average diameter of
1.24 mm, revealing that their reaction experiments were highly diffusion-limited.
In order to verify the performance of the model in terms of reliably estimating the
reaction rate constants and the deactivation factors, a separate set of experiments was
carried out under the monolith mode of operation with a different feed concentration (250
ppm vs. 120 ppm) and a different reactor inlet pressure (0.5 psig vs. 4.6 psig) using a
200
different catalytic membranes (NA10G). This membrane was prepared using the same
method as the NA03 catalytic membrane. Its catalyst loading was determined to be
0.0156 g of Pt corresponding to 0.38 wt % of the membrane, fairly close to that of NA03.
Higher feed concentration for this case led to a rapid deactivation of the catalytic
monolith, i.e., the conversion dropped by 30 % in a 5-hr time on stream experiment. The
rate data gathered during this experiment are presented in Table 4.6, and the conversion
vs. time plot is shown in Figure 4.33. In Figure 4.34, a comparison is made between the
model predictions with the experimental values in terms of DMMP conversion, showing
a reasonable agreement between these values. However, the model predicts a slower
deactivation rate compared to the actual experimental values, see Fig. 4.33.

4.6. Conclusions
Catalyst deactivation and pore plugging play major roles during catalytic destruction
of DMMP. In this Chapter the focus was on gaining more fundamental knowledge about
these phenomena. For that purpose we utilized analytical methods like SEM/EDAX,
FTIR, and N
2
adsorption/desorption methods in order to study the performance and
characteristics of the catalytic membranes before and after exposure to DMMP under
various reaction conditions. The results of these studies show that the adverse effects of
catalyst deactivation and pore plugging increase by increasing the activity of catalytic;
membranes loaded with more catalytic metal and with smaller pore sizes show more
propensities towards deactivation and pore blockage.
A mathematical model, developed in Chapter 3, was also extended to incorporate the
effect of catalyst deactivation and pore-blockage on the performance of the FTCMR. The
201
model was successfully applied to determine from experimental data important
parameters such as the reaction rate constants and the poisoning and pore plugging
factors. The same model is used to investigate the effect of important operating
conditions like the feed concentration on the effective life of the catalytic membrane, as
is discussed in the next Chapter of this report.

Table 4.5: Prior studies reporting reaction rate constants for the destruction of key OPC compounds
Authors Organophosphorus
compound (OPC)
Reactor Reaction conditions
k
0
,
(s
-
1
)
E,
(kJ/mole)
Graven et al.,
(1966)
Dimethyl
methylphosphonate
Packed-bed
Catalytic decomposition on
Pt/Al
2
O
3

3.0  10^(+5) 31
Zegers and Fisher,
(1996)
Diethyl
methylphosphonate
Plug-flow
Homogeneous pyrolysis in
nitrogen at
2.3  10^(+16) 247
Tsang, (1996) Triethyl phosphate Shock-tube N/A 1.0  10^(+13) 187
Zegers and Fisher,
(1998)
Triethyl phosphate Plug-flow
Homogeneous pyrolysis in
nitrogen at
1.0  10^(+12) 198
Zegers and Fisher,
(1998)
Diisopropyl
methylphosphonate
Plug-flow
Homogeneous pyrolysis in
nitrogen at
1.0  10^(+12) 154
Werner and Cool,
(1999)
Dimethyl
methylphosphonate
Premixed
flat-flame
Combustion in H
2
/O
2

flame
1.0  10^(+11) ~
3.7  10^(+14)
0 ~ 365
Glaude et al., (2000)
Diethyl
methylphosphonate
N/A
(Theoretical
study)
Combustion in H
2
/O
2

flame
5.0  10^(+12) ~
5.0  10^(+13)
163 ~
189
Glaude et al., (2000)
Diisopropyl
methylphosphonate
N/A
(Theoretical
study)
Combustion in H
2
/O
2

flame
1.0  10^(+13) ~
3.5  10^(+13)
172 ~
176
Glaude et al., (2000) Triethyl phosphate
N/A
(Theoretical
study)
Combustion in H
2
/O
2

flame
1.9  10^(+12) ~
1.5  10^(+14)
184 ~
192
Korobeinichev et al.,
(2000a)
Dimethyl
methylphosphonate
Premixed
flat-flame
Combustion in Ar/H
2
/O
2

flame
1.0  10^(+11) ~
3.7  10^(+14)
0 ~ 209
Korobeinichev et al.,
(2000a)
Trimethyl phosphate
Premixed
flat-flame
Combustion in Ar/H
2
/O
2

flame
5.0  10^(+11) ~
2.0  10^(+14)
0 ~ 146
Korobeinichev et al.,
(2000b)
Diisopropyl
methylphosphonate
Premixed
flat-flame
Combustion in Ar/H
2
/O
2

flame
5.0  10^(+11) ~
1.0  10^(+14)
13 ~ 153
Sullivan et al., (2004)
Methylphosphonic
Acid
Plug-flow
Oxidation in
supercritical water
1.0  10^(+14) 186
Brevett and Wagner,
(2005)
Dimethyl
methylphosphonate
N/A
Hydrolysis decomposition
on wet concrete
3.6  10^(+08) ~
1.6  10^(+09) *
68 ~ 78

* s
-1
mol
-1

202
Table 4.6: Experimental data collected from the monolith reaction studies with the NA10G
membrane

Time on stream,
(sec)
DMMP Concentration
in the Feed, (ppm)
Feed
Flow-rate, (SCCM)
Reactor
Temperature, (K)
Conversion, (%)
0 250 260 573 N/A
1200 250 260 573 63.8
4800 250 260 573 60.3
9000 250 260 573 55.1
10800 250 260 573 50.7
12600 250 260 573 42.0
14400 250 260 573 38.0
16200 250 260 573 34.8

0
20
40
60
80
100
0 50 100 150 200 250 300
Time on stream, (min)
Conversion, (%)
Model predictions
Experimental values

Figure 4.33: Rapid deactivation of the NA10G catalytic membrane compared with the model
predictions (T=573 K)

203
0
15
30
45
60
75
90
0 1530456075 90
DMMP conversion, (experimental values, %)
DMMP conversion
(model predictions, %)
Experimental values
Model predictions

Figure 4.34: Comparison between the model predictions and experimentally collected values for the
DMMP conversion

204
Chapter 5: Hybrid System

5.1. Introduction
In the studies, so far, a FTCMR has been used to carry out the destruction of DMMP, a
nerve agent simulant. The results indicated that higher conversions could be obtained by
this reactor when compared to a monolith reactor operating under the same temperature and
residence time. However, as discussed in Chapter 4, trapping of the P-containing products
of DMMP decomposition in the membrane pore structure presents challenges for the
FTCMR operation, particularly for high concentrations of DMMP. In Chapter 4 we
discussed methods for removing these phosphate deposits and for recovering the
properties of the membranes used in the FTCMR for further reuse. The most effective
method involves washing the used membranes with a dilute acidic solution. The
concentration of the acidic solution required depends on the type of the membrane
material ( γ- or α-alumina) and the amount of catalytic metal impregnated on the
membrane. The key challenge with this technique for recovery of the activity and
permeance of the membrane is that the FTCMR module needs to be put “off-line” for the
wash step. This may be feasible for collective protection (CP) operations, whereby one
may be able to operate similarly to a pressure-swing adsorption (PSA) system whose
parts are being regenerated while others are in operation. For individual protection (IP)
applications, however, the recovery step is likely to be a cumbersome task for the user,
and definitely entails the risk of damaging the membrane and the FTCMR itself.
In this Chapter a new concept is investigated which aims to prolong the effective
catalytic membrane life during FTCMR operation. It involves the use of a hybrid system
205
which combines a physical separation step which removes the bulk of DMMP via the use
of a surface-flow membranes system (SFMS) followed by the FTCMR as a second stage
polishing step. In this Chapter preliminary modeling and experimental results are presented
which we believe validate the use of this hybrid device as an effective tool to prolong the
effective life of catalytic membranes in situ and under the reaction conditions.

5.2. The Hybrid SFMS-FTCMR Concept
As part of this study, the effect of the feed concentration of DMMP on its complete
destruction in a FTCMR was studied in Chapter 3. As Figure 3.7, which shows data for
three different feed concentrations (150 ppm, 300 ppm, and 1000 ppm) at 573 K,
indicates the protection (complete conversion) time is a function of the DMMP
concentration; longer protection times are observed for the lower concentrations, and
shorter protection times are associated with the higher concentrations. These
experimental findings are further substantiated here by using the mathematical model that
was developed in Chapter 4. It is applied here to a typical A-type membrane that treats
DMMP-air mixtures of four different feed concentrations 100 ppm, 250 ppm, 500 ppm,
and 1000 ppm respectively (the membrane and operating parameters are shown in Table
5.1 – for the definition of the various parameters, see Chapter 3). The simulation results
are shown in Figure 5.1 and confirm the experimental observations in that the protection
time provided by the FTCMR is a function of the DMMP concentration in the feed. In
addition the correlation between protection time and the feed concentration is not linear.
Figure 5.2 shows the simulation results for a different catalytic membrane in a
FTCMR (operating as in the case of Figure 5.1 in a total flow-through or 100 % stage-cut
206
mode) treating again the same DMMP-air mixtures. The membrane and operating
parameters for the simulations of Figure 5.2 are again shown in Table 5.1. They are the
same for those of Figure 5.1 other than the Thiele modulus Φ
ref
, which in this case is
equal to 1.30 resulting in a deposition of P-containing products of DMMP decomposition
in a more narrow zone of the catalytic active layer, and thus having a more significant
impact on the membrane permeation characteristics. Figure 5.2 shows the fractional
change in the membrane permeance, due to the pore blockage, with respect to the time on
stream; clearly the rate of pore blockage is also a function of DMMP concentration in the
feed-stream, with the lower decrease of the permeance corresponding to the lower feed
concentrations.
These experimental and modeling observations point out that an appropriate role for
the FTCMR to play is as a second stage in a hybrid system, following a bulk-toxin
removal unit in the first stage. The main advantage of the proposed hybrid system,
combining a surface-flow membrane (SFM) separation unit (which is capable of
continuously physically removing a large portion of the CWA from contaminated air
streams) with the FTCMR, is that it completely destructs the CWA amount that remains
with a lower rate of pore blockage resulting in the continuous CWA destruction for
extended time periods, which are appropriate for both IP and CP applications.

Table 5.1: Parameters used for the simulation studies for the effect of feed
concentration on FTCMR performance

Reference
Thiele modulus
Reference
Péclet number
Pressure
ratio
Temperature
ratio
Low catalyst activity case 1.0 0.35 2 0.85
High catalyst activity case 1.3 0.35 2 0.85

207
70
80
90
100
0 40 80 120 160 200 240 280 320
Time on stream, (min)
DMMP conversion, (%)
100 ppm
250 ppm
500 ppm
1000 ppm

Figure 5.1: Effect of feed concentration on the protection time predicted by the mathematical model for
a typical FTCMR ( α = 0.05, Pé
ref
= 0.35, Φ
ref
= 1)

-50
-40
-30
-20
-10
0
0 50 100 150 200 250 300 350
Time on stream, (min)
Change in the permeance, (%) bb
100 ppm
250 ppm
500 ppm
1000 ppm

Figure 5.2: Effect of the feed concentration on the permeance predicted by the mathematical model for a
typical FTCMR ( α = 0.05, Pé
ref
= 0.35, Φ
ref
= 1.3)

As noted previously, the stage I of the proposed hybrid system makes use of SFM
membranes which remove DMMP via a physical separation mechanism. In the
experiments reported here, we have made use of carbon molecular sieve (CMS)
208
membranes with a high DMMP removal rate capability, available prepared by our
industrial collaborators in this project, M&P.
Separation of condensable compounds (e.g., CO
2
, lower M.W. hydrocarbons, water
and methanol, etc.) has previously been investigated by other researchers as well (e.g.,
Sperry et al., 1991; Sircar et al., 1999; Ahmad et al., 2007). Under optimal conditions,
when separating condensable vapors from their mixtures with non-condensable fixed
gases (e.g., air or nitrogen), one observes a high separation selectivity of the membrane
towards the condensable (and heavier) component, a phenomenon also known as “reverse
selectivity” (e.g., Yoo et al., 2009; Merkel et al., 2002). This interesting type of behavior
observed with microporous membranes is due to condensation that occurs within the pore
structure as a result of Van der Waals forces (physical adsorption) between the surface of
the membrane and the condensable species (equivalent to liquefaction).

5.3. Experiments
For the hybrid system experiments, in stage I, a 10-in CMS membrane (hereinafter,
referred to as MP13) was utilized. Its transport properties were measured via single-gas
permeation experiments. Its air permeance was 1.54x10
-8
mol/Pa/m
2
/s at 323 K, while its
He permeance at the same temperature was 8.27 x10
-8
mol/Pa/m
2
/s. These are typical
values our Group (and M&P) are reporting for these membranes, but as expected
considerably lower than the permeance of the typical mesoporous membranes used in the
FTCMR experiments (see Chapter 2).
For the CMS membranes to be effective in removing the DMMP, they must have a
narrow pore size distribution with an average pore size in the range of a few Å somewhat
209
larger than the size of the DMMP (5.7 Å [Jung, 2010; Lu , 2008]) molecule itself, but not
substantially larger as to avoid significant air by-pass. In addition, the expectation is that
DMMP removal will be enhanced (i) by an increase in the trans-membrane pressure,
which aids the condensation through the membrane and speeds-up the rate of DMMP
flow within the pores; and (ii) by an increase in the sweep gas flow, which improves the
rate of the external mass transfer and also decreases the vapor pressure of DMMP in the
membrane permeate side.
Tests were carried with the MP13 CMS membrane in order to investigate its
performance in terms of being able to separate DMMP from its mixtures with air. In the
first set of the experiments, results of which are shown in Figure 5.3, the effect of the
trans-membrane pressure on the rate of the DMMP removal was investigated. Figure 5.3
shows that, as expected, the DMMP removal rate increases with increasing pressure
gradient across the membrane.
However, as Figure 5.3 also indicates, raising the trans-membrane pressure difference
results into an increase in the air-leak (represented in terms of the stage-cut of the
membrane, which however remains quite reasonable below 10 %). This may be due to
opening of some of pores filled by DMMP molecules at the lower pressure differences.
In another experiment, the effect of the sweep gas ratio (sweep gas flow rate to the
feed flow rate) on the rate of DMMP removal was studied, and the data are plotted in
Figure 5.4 for two different temperatures. These results demonstrate an increase in the
DMMP removal rate with respect to the sweep gas ratio, as expected.

210
0
10
20
30
40
50
60
70
80
0 5 10 15 20 25 30
Transmembrane pressure, (psi)
Removal rate, (%)
0
2
4
6
8
10
Stage cut, (%)
Removal rate
Stage cut

Figure 5.3: Effect of the imposed pressure gradient across MP13 CMS membrane on its DMMP
removal rate and stage-cut (T=373 K, DMMP concentration = 825 ppm, sweep ratio = 1.2)

15
25
35
45
55
65
01 23 4 5 6
Sweep ratio, (dimensionless)
Removale rate (%)
T=323 K
T=303 K

Figure 5.4: Effect of the sweep gas ratio on DMMP removal rate for MP13 CMS membrane
(DMMP concentration = 625 ppm)

211
The data in Figure 5.4 also show that increasing the temperature slightly improves the
efficiency of the CMS membrane in removing DMMP. While the lower temperatures
favor more condensation inside the pores on the membrane feedside, higher temperatures
favor better movement of the condensates inside the pores and higher vaporization rates
on the permeate-side of the membrane, highlighting the fact that there should be an
optimum value for the operating temperature. During the hybrid system experiments,
however, the maximum operating temperature of stage I was kept below 323 K to avoid
the potential for decomposition of DMMP over the alumina substrates over which the
CMS layers are deposited [Mitchell et al., 1997]. Determining the optimum values for the
trans-membrane pressure, sweep gas ratio, and temperature requires a comprehensive
theoretical model that is able to describe the mass transfer inside the porous structure of
the CMS membrane; such a model is currently under development in our group, and will
be followed by the development of a more detailed molecular dynamics model in the near
future.
Following its characterization studies, the MP13 CMS membrane was installed
upstream of the FTCMR for the hybrid system experiments. In these experiments the
reject-side air stream of the SFM stage I (employing the MP13 CMS membrane) was
utilized as the feed for FTCMR stage II. In the FTCMR we utilized a number of different
high- and low-active catalytic membranes prepared by the wet impregnation of A- and B-
type alumina membranes with a 8 wt.% hexa-chloroplatinic acid solution for high-
activity samples and with a 1 wt.% hexa-chloroplatinic acid solution for low-activity
samples, followed by the post treatment method previously described in Chapter 2. The
characteristics of all the catalytic membranes are shown in Table 5.2. All the hybrid
212
experiments were conducted with the FTCMR operating in the total flow-through (100 %
stage-cut) mode. The temperature of the FTCMR was kept constant at 573 K, while the
temperature of the CMS membrane was kept constant at 323 K. The feed flow rate to the
overall system was kept constant at 60 sccm with a DMMP concentration of 300 ppm,
and the sweep ratio (for stage II, no sweep was utilized for the FTCMR) was kept
constant at the value of 4. No attempt was made to control the pressure drop across the
CMS membrane, and as a result the trans-membrane pressure for the 1
st
stage was
dictated by the "natural" pressure drop built across the catalytic membrane in the
FTCMR.

Table 5.2: Specifications of the catalytic and surface-flow membranes used for the hybrid system
experiments

Catalytic membrane
(FTCMR)
Pt weight (g) Nominal pore size
(active layer, Å)
CMS
NA06 (high-active) 0.01242 500 (inner-coated) MP13
NA07 (high-active) 0.01228 500 (inner-coated) MP13
NB05 (low-active) 0.00221 100/500 (inner-coated) MP13
NB07 (low-active) 0.00192 100/500 (inner-coated) MP13
NB10 (high-active) 0.03362 100/500 (inner-coated) MP13
NB27 (high-active) 0.03053 100/500 (inner-coated) MP13

In Figure 5.5, the performance of the hybrid system (consisting of the MP13 CMS
membrane followed by the NB27, B-type catalytic membrane) is compared with that of a
FTCMR (using the NB10 B-type membrane having approximately equal amount of
catalyst weight with the NB27 membrane, see Table 5.1), showing the superiority of the
hybrid system in terms of offering higher destruction rates for DMMP (the hybrid system
213
offers full protection while the FTCMR does not). The benefits in using the hybrid
system becomes more obvious by noting that the data gathering for the NB10 catalytic
membrane stopped after 150 min of time on stream; this is due to the high rate of the pore
blockage this membrane experienced (see Figure 5.6) which dictated that the experiment
be terminated as to not create unsafe conditions in the operation of the glass-bubblers, as
noted in Chapter 4.

85
90
95
100
0 50 100 150 200 250 300 350
Time on stream, (min)
Conversion, (%)
FTCMR
HYBRID

Figure 5.5: Hybrid system performance compared with the FTCMR by itself for the high-activity, B-
type catalytic membrane

However, as shown in Figure 5.6 for the hybrid system, even after 300 min of
operation, the permeance of the 2
nd
stage did not match that of NB10 catalytic membrane,
consistent with the modeling predictions.
Figure 5.7 plots the results of experiments when using lower activity B-type
membranes. The hybrid system in this case consists of MP13 (stage I) followed by the
FTCMR containing catalytic membrane NB05. Its behavior is compared with that of the
214
FTCMR (containing membrane NB07) by itself. The behavior in Figure 5.7 is very
similar to that shown in Figure 5.6 for the more active B-type membranes.

-90
-75
-60
-45
-30
-15
0
0 50 100 150 200 250 300 350
Time on stream, (min)
Change in permeance, (%)
FTCMR
HYBRID

Figure 5.6: The positive effect of the hybrid system in deferring catalytic membrane pore blockage
compared with the FTCMR alone for the high-activity B-type catalytic membranes

40
55
70
85
100
0 100 200 300 400 500 600
Time on stream, (min)
Conversion, (%)
FTCMR
HYBRID

Figure 5.7: Hybrid system performance compared with the single FTCMR for the low-activity B-type
catalytic membranes
215
When using A-type membranes, while the conversion for both the FTCMR (containg
membrane NA06) and hybrid system (MP13 membrane followed by the NA07
membrane) remained above 99 % (data not shown here) for a time on stream of more
than 20 hr, the drop in the permeance of the catalytic membrane during the hybrid mode
of operation was significantly lower than in the case of the FTCMR alone. These results
are shown in Figure 5.8, demonstrating that the A-type membranes in the hybrid
configuration may be considered as the best choice in terms of the long-term destruction
of DMMP.

-80
-70
-60
-50
-40
-30
-20
-10
0
0 300 600 900 1200 1500
Time on stream, (min)
Change in permeance, (%)
FTCMR
HYBRID

Figure 5.8: The positive effect of the hybrid system in deferring catalytic membrane pore blockage
compared with the FTCMR alone for the high-activity A-type catalytic membranes

216
5.4. Conclusions
Because of their ability to decrease the required amount of catalyst and also decrease the
reactor temperature needed to achieve a certain desired conversion (see Chapter 1),
FTCMR are promising for application in both gas- and liquid-phase reactions. However,
when applied for the destruction of DMMP, they face the challenge of activity decline and
increase in the pressure drop due to the deposition in the membrane structure of P-containg
species (during DMMP decomposition). In this Chapter in order to overcome this challenge
we proposed the use of a hybrid system in which the FTCMR functions as a second stage
following a bulk-toxin removal first stage using surface-flow CMS membranes. Our
findings demonstrate that this hybrid system achieves long-term and complete destruction
of DMMP in an efficient way.

217
Chapter 6: Suggestions for Future Work
So far, we have found out that the application of a forced-flow through catalytic
reactor using Pt/Alumina for the oxidation of DMMP is a promising approach, but the
observed pore blockage limits its application to:
- Low catalyst loadings
- Membranes having large pores
- Low reaction temperatures
In order to deepen the knowledge into the pore-blockage phenomenon and to find
possible methods to overcome it (or, at least, lessen its effects) without the need for
application of a surface-flow CMS membrane, it is worth to investigate the pore-blockage
phenomenon using other support materials (in addition to alumina), including titania,
zirconia, and stainless-steel.
As a preliminary effort, here it was tried to compare the performance of a Pt/TiO
2

membrane with a Pt/Al
2
O
3
under the FTCMR mode of operation. Like all alumina-based
membranes, the TiO
2
membranes with the nominal pore size of 500 Å used for this
experiment were also supplied by our industrial collaborator in this project, namely
Media and Process Technology, Inc.; however, unlike the alumina membranes the latter
membranes were coated from the outside with a thin layer of TiO
2
(nominal thickness of
15 µm) on a similar symmetric alumina support with a nominal pore size of 4000 Å. The
characteristics of the catalytic membranes prepared for these experiments are shown in
Table 6.1.
All experiments were carried out at a temperature of 573 K, a feed flow rate of 1 sccs,
and DMMP concentration of 460 ppm. The results, which are plotted in Figure 6.1, show
218
that the Pt/TiO
2
catalytic membrane also exhibits deactivation behavior shortly after
exposure to a feed stream of DMMP in air. However, the performance offered by the
Pt/TiO
2
membrane is somewhat better than that of Pt/Al
2
O
3
with the same amount of
platinum. This improvement can be attributed either to the thicker top layer of the former
membrane, which provides for a longer residence time for the reactants inside the active
layer of the membrane, or to the higher activity of the titania itself towards destruction of
DMMP. In fact, even for the case of a bare titania membrane, since the titania layer was
coated on the outer-side of the alumina support, formation of a yellowish deposit could
be observed with the naked eye, as it is shown in Figure 6.2. In this Figure, a fresh TiO
2

bare membrane, an aged TiO
2
bare membrane after exposure to DMMP reaction at 573 K
for 5 hrs, and a catalytic TiO
2
membrane are shown alongside to each other. (For the
catalytic membrane, due to the dark grey color of the Pt, the yellowish deposits could not
be detected via the naked eye).
Another important aspect of this experiment is that even in the case of outer-coated
membranes, where the residence time for products and by-products are shorter than for
the inner-coated membranes (when operating, as in these experiments, with the
DMMP/air mixture flowing in the inside of the membrane tube), no phosphorous-
containing products were detected by GC/MS, showing that the products of the reaction
are rapidly trapped within the porous structure of the catalytic membrane, consistent with
the previously discussed EDAX results.

219
0
20
40
60
80
100
0 100 200 300 400 500
Time on stream, (min)
DMMP conversion, (%)
Catalytic titania
Catalytic alumina

Figure 6.1: Comparison between the protection times offered by the Pt/Al
2
O
3
and Pt/TiO
2
catalytic
membranes

Table 6.1: Characteristics of the catalytic alumina (NA09) and titania (T3) membranes
Membrane Pt weight, (g) Active layer thickness*, (µm) Nominal pore size (active layer, Å)
T3 0.00404 13 500 (outer-coated)
NA09 0.00380 7 500 (inner-coated)
* The actual values obtained by SEM analysis

Figure 6.2: Comparison between the appearances of different TiO
2
/Al
2
O
3
membranes
(From right to left: fresh bare TiO
2
/Al
2
O
3
, aged TiO
2
/Al
2
O
3
, and aged Pt/TiO
2
/Al
2
O
3
)
220
For theoretical studies, our current model does not account for pore network
interconnectivity effects or for phenomena associated with pore size distribution. Sahimi
and Tsotsis (1985) studied catalyst deactivation by active site poisoning and pore
blockage under globally kinetic control, and showed that the interconnectivity of the
pores plays a fundamental role in the overall catalytic behavior. So, it is suggested here
that the current model be improved, so that it can account for the effect of pore
interconnectivity and pore-size distribution on FTCMR performance during the course of
pore-plugging and catalyst deactivation.
Furthermore, the current continuum-type model cannot be used to study the effect of
the possible enhanced number of collisions between the reactant(s) and catalytic pore
wall, the main advantage of FTCMR working totally under a Knudsen-flow mode, on the
rate of the reaction. So, we suggest here that a molecular dynamics-based model be
developed which is able to predict the number of collisions within each pore, and its
effect on the overall rate of reaction inside these pores [Albo et. al, 2006].
Finally, the FTCMR concept may be used for fabrication of new types of membranes.
In the proposed preparation technique, a mixture of the precursor and another agent
(oxidizing or reducing based on the application, as needed) is forced through the
membrane support under well-controlled temperature, pressure, and concentration
conditions. The desired product (separation layer) of the possible reaction gradually fills
up the existing big support tube pores, resulting into a membrane with a smaller pore
size. By controlling the operating parameters (like the Péclet number), the depth of
deposition may be adjusted. This concept will work only when there is an affinity
between the support and the material used as the separation layer, or the support is
221
reactive towards the precursor. Figure 6.3 shows, schematically, a comparison between a
membrane made by the conventional process and one made based on the proposed new
concept.

Figure 6.3: Comparison between the conventional method (left) and hypothetical membrane fabricated
based on FTCMR concept (right)

222
References
Acres G.J.K., Cooper B.J., Shutt E., and Malerbi B.W., “Platinum catalysts for
exhaust emission control: the mechanism of catalyst poisoning by lead and phosphorus
compounds”, Advances in Chemistry Series, ACS publications (1975), 54-71

Ahmad F., Mukhtar H., Man Z., and Dutta B.K., “Predicting separation of lower
drocarbon from natural gas by a nano-porous membrane using capillary condensation”,
Chemical Engineering Technology, 30 (2007), 1266–1273

Albo S.E., Broadbelt L.J., and Snurr R.Q., “Multiscale modeling of transport and
residence times in nanostructured membranes”, AIChE Journal, 52 (2006), 3679–3687

Alfonso M., Menendez M., and Santamaria J., “Coupling of consecutive reactions in a
two-layer, flow-through catalytic membrane”, Industrial & Engineering Chemistry
Research, 40 (2001), 1058–1064

Angele B.and Kirchner K., “The poisoning of noble metal catalysts by phosphorus
compounds- Part I”, Chemical Engineering Science, 35 (1980a), 2089-2091

Angele B. and Kirchner K., “The poisoning of noble metal catalysts by phosphorus
compounds- Part II”, Chemical Engineering Science, 35 (1980b), 2093-2099

Araujoa L.R.R, Scofielda C.F., Pasturaa N.M.R, and De Araujo Gonzalez W.,
“H
3
PO
4
/Al
2
O
3
Catalysts: characterization and catalytic evaluation of oleic acid
conversion to biofuels and biolubricant”, Materials Research, 9 (2006), 181-184

Atteya M., and Klabunde K.J., “Nanoscale metal oxide particles as chemical reagents:
Heats of adsorption of heteroatom-containing organics on heat-treated magnesium oxide
samples of varying surface areas”, Chemistry of Materials, 3 (1991), 182-187

223
Bautista F.M., Campelo J.M., Garcia A., Luna D., Marinas J.M. and Romero A.A.,
“AlPO
4
-A1
2
O
3
catalysts with low alumina content: I. Structural and textural
characterization of catalysts obtained with aqueous ammonia”, Applied Catalysis A:
General, 96 (1993), 175-199
Bermudez V.M., “Adsorption on carbon nanotubes studied using polarization-
modulated infrared reflection-absorption spectroscopy”, Journal of Physical Chemistry B,
2005, 109 (2005), 9970-9979

Biesheuvel P.M., “Porous ceramic membranes”, PhD thesis, University of Twente,
Netherland (2000)

Biswas J. and Do D.D., “A unified theory of coking deactivation in a catalyst pellet”,
The Chemical Engineering Journal, 36 (1987), 175-191

Brevett C.A. and Wagner G.W., “Fate of nerve agent simulants on concrete”, Geo-
Centers, Inc. (2005)

Brooks C.S. and Kehrer V.J., “Chemisorption of carbon monoxide on metal surfaces
by pulse chromatography”, Analytical Chemistry, 41 (1969), 103-106

Busca G., Ramis G., Lorenzelli V, Rossi P. F., Ginestra A. L., and Patrono P.,
“Phosphoric acid on oxide carriers: 1. Characterization of silica, alumina, and titania
impregnated by phosphoric acid”, Langmuir, 5 (1989), 911-916

Butrow A.B., Buchanan J.H., and Tevault D.E., “Vapor pressure of organophosphorus
nerve agent simulant compounds”, Journal of Chemical Engineering Data, 54 (2009),
1876–1883

Cao L., Segal S. R., Suib S. L., Tang X., and Satyapaly S., “Thermocatalytic oxidation
of dimethyl methylphosphonate on supported metal oxides’, Journal of Catalysis, 194
(2000), 61–70
224
Cao L., Suib S. L., Tang X., and Satyapaly S., “Thermocatalytic decomposition of
dimethyl methylphosphonate on activated carbon”, Journal of Catalysis”, 197 (2001),
236–243
Capannelli G., Vigo F. and Munari S., “Ultrafiltration membranes- characterization
methods”, Journal of Membrane Science, 15 (1983), 289-313

Caro J., Caspary K., Hamel C., Hoting B., Kölsch P., Langanke B., Nassauer K.,
Schiestel T., Schmidt A., Schomäcker R.,, Seidel-Morgenstern A., Tsotsas E., Voigt I.,
Wang H., WarsitzR., Werth S., and Wolf A., “Catalytic membrane reactors for partial
oxidation using perovskite hollow fiber membranes and for partial hydrogenation using a
catalytic membrane contactor”, Industrial & Engineering Chemistry Research, 46 (2007),
2286–2294

Chea V., Paolucci-Jeanjean D., Belleville M.P., and Sanchez J., “Optimization and
characterization of an enzymatic membrane for the degradation of phenolic compounds”,
Catalysis Today, Under Press (2012)

Chen D.A, Ratliff J.S., Hua X., Gordon W.O., Senanayake S.D., and Mullins D.R.,
“Dimethyl methylphosphonate decomposition on fully oxidized and partially reduced
ceria thin films”, Surface Science, 604 (2010), 574–587

Decker S.P., Klabunde J.S., Khaleel A., and Klabunde K.J., “Catalyzed destructive
adsorption of environmental toxins with nanocrystalline metal oxides. Fluoro-, chloro-,
bromocarbons, sulfur, and organophosophorus compounds”, Environmental Science and
Technology, 36 (2002), 762-768

Dixon A.G., “Recent research in catalytic inorganic membrane reactors”, International
Journal of Chemical Reaction Engineering, 1 (2003), review R6

Dulcey C.S. and Lin M.C., “Thermal desorption of the PO radical from polycrystalline
Pt surfaces”, Chemical Physics Letter, 115 (1985), 481 - 485
225
Edreva V., Zhang F., Mangold M., Tsotsas E., “Mass transport in multilayer porous
metallic membranes – diagnosis, identification and validation”, Chemical Engineering
and Technology, 32 (2009), 632–640

Fadoni M. and Lucarelli L., “Temperature programmed desorption, reduction,
oxidation and flow chemisorption for the characterization of heterogeneous catalysts:
theoretical aspects, instrumentation and applications”, Studies in Surface Science and
Catalysis, 120 (1999), 177–225

Fan C.L. and Wang L.S., “Vapor pressure of dimethyl phosphite and dimethyl
methylphosphonate”, Journal of Chemical Engineering Data, 55 (2010), 479–481

Fitz J.C.W. and Rase H.F., “Effects of phosphorus on nickel-molybdenum
hydrodesulphurization catalysts of varying metals”, Industrial and Engineering
Chemistry Product research and Development, 22 (1983), 40-44

Fritsch D., and Bengtson G., “Catalytic polymer membranes for high temperature
hydrogenation of viscous liquids”, Advanced Engineering Materials, 8 (2006a), 386–389

Fritsch D., and Bengtson G., “Development of catalytically reactive porous
membranes for the selective hydrogenation of sunflower oil’, Catalysis Today, 118
(2006b), 21–127

Fritsch D., Randjelovic I., and Keil F., “Application of a forced-flow catalytic
membrane reactor for the dimerisation of isobutene”, Catalysis Today, 98 (2004), 295–
308
Geyer R., Hunold J., Keck M., Kraak P., Pachulski, A., Schödel R., “Methods for
determining the metal crystallite size of Ni supported catalysts”, Chemie Ingenieur
Technik, 84 (2012), 160–164

226
Glaude P.A., Curran H.J., Pitz W.J. and Westbrook C.K., “Kinetic study of the
combustion of organophosphorus Compounds”, Proceedings of the Combustion Institute,
28 (2000), 1749–1756

Graven W.M., Weller S.W., and Peters D.L., “Catalytic conversion of an
organophosphate vapor over platinum-alumina”, Industrial and Engineering Chemistry
Research, 5 (1966), 183-189

Gröschel L., Haidar R., Beyer A., Colfen H., Frank B., and Schomäcker R.,
“Hydrogenation of propyne in palladium-containing polyacrylic acid membranes and its
characterization”, Industrial & Engineering Chemistry Research, 44 (2005), 9064–9070

Guo X., Yoshinobu J., and Yates J. T., “Decomposition of an organophosphonate
compound (Dimethyl Methylphosphonate) on the Ni (111) and Pd (111) Surfaces”
Journal of Physical Chemistry, 94 (1990), 6839-6842

Han S.T., Zhang G.Y., Xi H., Xu D.N., Fu X.Z., and Wang X.X., “Sulfated TiO
2

decontaminate 2-CEES and DMMP in vapor phase”, Catalysis Letter, 122 (2008),106–
110

Haüy R.J., “Essai d'une théorie sur la structure des cristaux appliqué à plusieurs genres
de substances cristallines”, Goguée & Née de la Rochelle, Paris, (1784)

Hedge R.I. and White J.M., “Interaction of dimethyl methylphosphonate with oxidized
iron and the effect of coadsorbed water”, Applied Surface Science, 28 (1987) 1-10

Hegde R.I., Greenlief C.M., and White J.M., “Surface chemistry of dimethyl
methylphosphonate on Rh (100)”, Journal of Physical Chemistry, 89 (1985), 2886-2891

Hegedus L. and Baron K., “Phosphorus accumulation in automotive catalysts”,
Journal of Catalysis, 54 (1978), 115-119 (1978)
227
Hegedus L. and Summers J., “Improving the poison resistance of supported catalysts”,
Journal of Catalysis, 48 (1977), 345-353

Henderson M.A., Jin T., and White J.M., “A TPD/AES study of the interaction of
dimethyl methylphosphonate with α-Fe
2
O
3
and SiO
2
”, Journal of Physical Chemistry, 90
(1986), 4607-4611

Henderson M.A., Jin T., and White J.M., “Adsorption and decomposition of dimethyl
methylphosphonate on platinum (111)”, Journal of the American Chemical Society, 110
(1988), 6939 -6947

Henis J.M.S, and Tripodi M.K., “Composite hollow fiber membranes for gas
separation: the resistance model approach”, Journal of Membrane Science, 8 (1981), 233
- 246

Hsu C.C., Dulcey C.S., Horwitz J.S., Horwitz J.S., and Lint M.C., “Mass
spectrometric characterization of performance of a low-temperature oxidation catalyst”,
Journal of Molecular Catalyst, 60 (1990), 389-398

Ilinich O.M., Cuperus F.P., Nosova L.V., and Gribov E.N., “Catalytic membrane in
reduction of aqueous nitrates: operational principles and catalytic performance”,
Catalysis Today, 56 (2000),137–145

Ilinich O.M., Gribov E.N., and Simonov P.A., “Water denitrification over catalytic
membranes: Hydrogen spillover and catalytic activity of macroporous membranes loaded
with Pd and Cu”, Catalysis Today, 82 (2003), 49–56

Jakobs E. and Koros W.J., “Ceramic membrane characterization via the bubble point”,
Journal of Membrane Science, 124 (1997) 149-159

228
Jung K.H., “Nonwovens containing novel polymer fillers”, PhD Thesis, North
Carolina State University (2010)

Kang Li, “Ceramic membranes for separation and reaction”, John Wiley & Sons, West
Sussex, UK (2007), ISBN: 9-7804-7001-4400

Khassin A.A., “Catalytic membrane reactor for conversion of syngas to liquid
hydrocarbons”, Energeia, 16 (2005), 1–3

Khassin A.A., Sipatrov A., Yurieva T., Chermashentseva G., Rudina N., and Parmon
V., “Performance of a catalytic membrane reactor for the Fischer-Tropsch synthesis”,
Catalysis Today, 105 (2005), 362–366

Kobayashi M., Togawa J., Kanno T., Horiuchi J., and Tada K., “Dramatic innovation
of propene epoxidation efficiency derived from a forced flow membrane reactor”,
Journal of Chemical Technology and Biotechnology, 78 (2003), 303–307

Korobeinichev O.P., Ilyin S.B., Bolshova T.A., Shvartsberg V.M. and Chernov A.A.,
“The chemistry of the destruction of organophosphorus compounds in Flames-III: The
destruction of DMMP and TMP in a flame of hydrogen and oxygen”, Combustion and
Flame, 121 (2000a), 593–609

Korobeinichev O.P., Ilyin S.B., Bolshova T.A., Shvartsberg V.M. and Chernov A.A.,
“The chemistry of the destruction of organophosphorus compounds in Flames-IV:
destruction of DIMP in a flame of H
2
/O
2
/Ar”, Combustion and Flame, 123 (2000b),
412–420

Kreutzer M., Bakker J., Kapteijn F., Moulijn J., and Verheijen P., “Scaling-up
multiphase monolith reactors: Linking residence time distribution and feed
maldistribution”, Industrial & Engineering Chemistry Research, 44 (2005), 4898–4913

229
Kröger V., Hietikko M., Angove D., French D., Lassi U., Suopanki A., Laitinen R.
and Keiski R.L., “Effect of phosphorus poisoning on catalytic activity of diesel exhaust
gas catalyst components containing oxide and Pt”, Topics in Catalysis, 42-43 (2007),
409-413

Lambert C. and Gonzalez R., “Activity and selectivity of a Pd/gamma-Al
2
O
3
catalytic
membrane in the partial hydrogenation reactions of acetylene and 1,3-butadiene”,
Catalysis Letters, 57 (1999), 1–7

Lange C., Storck S., Tesche B., and Maier W.F., “Selective hydrogenation reactions
with a microporous membrane catalyst, prepared by sol-gel dip coating”, Journal of
Catalysis, 175 (1998), 280–293

Lee K. Y., Houalla M., Hercules D.M., and Hall W.K., “Catalytic oxidative
decomposition of dimethyl methylphosphonate over cu-substituted hydroxyapatite”,
Journal of Catalysis, 145 (1994), 223–231

Leggett D.C., “Solvent/water partitioning of dimethyl methylphosphonate (DMMP) as
a probe of solvent acidity”, Journal of Solution Chemistry, 22 (1993), 289-296

Levenspiel O, “Chemical Reaction Engineering”, John Wiley & Sons, Inc, New
Jersey, USA (1999), ISBN 0-471-25424-X

Lewis J.M. and Kydd R.A., “Adsorption mechanism of phosphoric acid on γ-
alumina”, Journal of Catalysis, 132 (1991), 465-471

Li Y.X. and Klabunde K.J., “Nanoscale metal oxide particles as chemical reagents
destructive adsorption of a chemical agent Simulant”, Langmuir, 7 (1991), 1388-1393

Li Y.X., Koper O., Atteya M., and Klabunde K.J., “Adsorption and decomposition of
organophosphorus compounds on nanoscale metal oxide particles: In situ GC-MS studies
230
of pulsed microreactions over magnesium oxide”, Chemistry of Materials, 4 (1992), 323-
330

Li Y.X., Schlup J.R., and Klabunde K.J., “Fourier transform infrared photoacoustic
spectroscopy study of the adsorption of organophosphorous compounds on heat-treated
magneseium oxide”, Langmuir, 7 (1991), 1394-1399

Lin Y.S., and Burggraaf A.J., “Preparation and characterization of high-temperature
thermally stable alumina composite membrane”, Journal of the American Ceramic
Society, 74 (1991), 219 - 224

López Corderoa R., Gil Llambíasb F.J., Palaciosc J.M., Fierrod J.L.G. and López
Agudod A., “Surface changes of alumina induced by phosphoric acid impregnation”,
Applied Catalysis, 56 (1989), 197-206

Lowell S., Shields J.E., Thomas M.A., and M. Thommes, "Characterization of porous
solids and powders: surface area, pore size and density", Springer (originally by Kluwer
Academic Publishers), Netherlands (2004), ISBN: 978-1-4020-2302-6

Lu X., “Design and study of lyotropic liquid crystal-butyl rubber nanocomposites for
chemical agent vapor barrier applications”, PhD Thesis, University of Colorado at
Boulder, Publication No. 3284451(2008)
Lüdtke K., Peinemann K.V., Kasche V., and Behling R.D., “Nitrate removal of
drinking water by means of catalytically active membranes”, Journal of Membrane
Science, 151 (1998), 3–11

Ma Y., Moser W., Dixon A., Ramachandra A., Lu Y., and Binkerd C., “Oxidative
coupling of methane using inorganic membrane reactors. Technical report, Worcester
Polytechnic Institute, Worcester, MA, United States (1998)

231
Maira A.J., Lau W.N., Lee C.Y., Yue P.L., Chan C.K., and Yeung K.L., “Performance
of a membrane-catalyst for photocatalytic oxidation of volatile organic compounds”,
Chemical Engineering Science, 58 (2003), 959–962

Mason E.A. and Malinauskas A.P., “Gas transport in porous media: the dusty-gas
model”, Chemical Engineering Monograph 17, Elsevier, New York, USA (1983), ISBN:
0-44442-190-4

Mattsson A., Lejon C., Stengl V., Bakardjieva S., Oplustil F., Andersson P.O., and
Osterlund L., “Photodegradation of DMMP and CEES on zirconium doped titania
nanoparticles”, Applied Catalysis B: Environmental, 92 (2009), 401–410

McArthur D. P., “Deposition and distribution of lead, phosphorus, calcium, zinc, and
sulfur poisons on automobile exhaust NO
x
catalysts”, Advances in Chemistry Series,
ACS publications (1975), 85-102

McGuire K.S., Lawson K.W, Lloyd D.R., “Pore size distribution determination from
liquid permeation through microporous membranes”, Journal of Membrane Science, 99
(1995), 127-137

Mera N., Hirakawaa T., Sanoa T., Takeuchia K., Setob Y., Negishia N., “Removal of
high concentration dimethyl methylphosphonate in the gas phase by repeated-batch
reactions using TiO
2
”, Journal of Hazardous Materials, 177 (2010) 274–280
Merkel T.C., Freeman B.D., Spontak R.J., He Z., Pinnau I., Meakin P., and Hill A. J.,
“Ultrapermeable, reverse-selective nanocomposite membranes”, Science, 296 (2002),
519-522

Minyukovaa T.P., Shtertsera, N.V., Khassina A.A., and Yurievaa T.M., “Permeable
composite membrane as a catalytically active contactor for hydrogenation reactions”,
Catalysis Today, In Press (2012)

232
Mitchell M.B., Sheinker V.N., and Cox W.W., “Room temperature reaction of ozone
and dimethyl methylphosphonate on alumina supported iron oxide”, Journal of Physical
Chemistry C, 111 (2007), 9417-9426

Mitchell M.B., Sheinker V.N., and Mintz E.A., “Adsorption and decomposition of
dimethyl methylphosphonate on metal oxides”, Journal of Physical Chemistry, B, 101
(1997), 11192-11203

Mitchell M.B., Sheinker V.N., Gatimu E.N., Tesfamichael A.B., and Nunley M.,
“Decomposition of dimethyl methylphosphonate (DMMP) on supported cerium and iron
co-impregnated oxides at room temperature”, Journal of Physical Chemistry, B, 107
(2003), 580-586

Mitchell M.B., Sheinker V.N., Gatimu E.N., Cox W.W., Tesfamichael A.B., and
Nunley M., Tesfamichael A.B., “The room temperature decomposition mechanism of
DMMP decomposition on alumina supported cerioum oxide- Participation of nano-sized
cerium oxide domains”, Journal of Physical Chemistry, B, 108 (2004), 1634-1645

Mitchell M.B., Sheinker N.V, Cox W.W., and Hardcastle K., “Sustained Room
Temperature Decomposition of Dimethyl Methylphosphonate (DMMP) by O
3
on
Alumina-Supported MnOx”, Journal of Physical Chemistry, C, 115 (2011), 11514–
11524
Morales A., Ramireze de Agudelo M.M., Hernandez F., “Adsorption mechanism of
phosphorus on alumina”, Applied Catalysis, 41 (1988), 261-271

Moss J.A., Szczepankiewicz S.H., Park E., and Hoffmann M.R., “Adsorption and
photodegradation of dimethyl methylphosphonate vapor at TiO2 surfaces”, Journal of
Physical Chemistry, 109 (2005), 19779-19785

233
Mucherie S., Kim H.S., Marshall C.L., Libera J.A., Elam J.W., and Stair P.C.,
“Oxidative dehydrogenation of propane over nanostructured membrane catalyst”, In
North American Catalysis Society 20
th
North American Meeting (2007)

Mundschau M. and Vanselow R., “Phosphorus on platinum — field electron emission
microscopy”, Surface Science, 166 (1986), L131-L135

Munro N.B., Talmage S.S., Griffin G.D., Waters L.C., Watson A.P., King J.F, and
Hauschild V., “The sources fate and toxicity of chemical warfare agent degradation
products”, Environmental Health Perspectives, 107 (1999), 933-974

Nooney M.G., Campbell A., Murrell T.S., Lin X.F., Hossner L.R., Chusuei C.C., and
Goodman D.W., “Nucleation and growth of phosphate on metal oxide thin films”,
Langmuir, 14 (1998), 2750-2755

Obee T.N., and Satyapal S., “Photocatalytic decomposition of DMMP on titania”,
Journal of Photochemistry and Photobiology A: Chemistry”, 118 (1998), 45-51

O'Shea K.E., Beightol S., Garcia I., Aguilar M., Kalen D.V., Cooper W.J.,
“Photocatalytic decomposition of organophosphonates in irradiated TiO
2
suspensions”,
Journal of Photochemistry and Photobiology A: Chemistry, 107 (1997), 221-226

Panayotov D.A. and Morris J.R., “Catalytic degradation of a chemical warfare agent
simulant: Reaction mechanisms on TiO
2
-supported Au nanoparticles”, Journal of
Physical Chemistry C, 112 (2008), 7496–7502

Panayotov D.A. and Morris J.R., “Thermal decomposition of a chemical warfare agent
simulant (DMMP) on TiO
2
: Adsorbate reactions with lattice oxygen as studied by
infrared spectroscopy”, Journal of Physical Chemistry, C, 113 (2009a), 15684–15691

234
Panayotov D.A. and Morris J.R.,” Uptake of a chemical warfare agent simulant
(DMMP) on TiO
2
: Reactive adsorption and active site poisoning”, Langmuir, 25 (2009b),
3652-3658

Pellin M., Stair P., Xiong G., Elam J., Birrell J., Curtiss L., George S., Han C., Iton L,
Kung H., Kung M., and Wang H.H., “Mesoporous catalytic membranes: Synthetic
control of pore size and wall composition;, Catalysis Letters, 102 (2005), 127–130

Pera-Titusa M., Fridmanna M., Guilhaumea N., and Fiatyb K., “Modelling nitrate
reduction in a flow-through catalytic membrane contactor: Role of pore confining effects
on water viscosity”, Journal of Membrane Science, 401– 402 (2012), 204– 216

Picasso G., Quintilla A., Pina M.P., and Herguido J., “Total combustion of methyl-
ethyl ketone over Fe
2
O
3
based catalytic membrane reactors”, Applied Catalysis B:
Environmental, 46 (2003), 133–143

Pina M.P., Irusta S., Menendez M., and Santamaria J., “Combustion of volatile
organic compounds over platinum-based catalytic membranes”, Industrial and
Engineering Chemistry Research, 36 (1997), 4557-4566

Pina M.P., Menendez M., and Santamaria J., “The Knudsen-diffusion catalytic
membrane reactor- An efficient contactor for the combustion of volatile organic
compounds”, Applied Catalysis B: Environmental, 11 (1996), L19-L27
Poling B.E., Prausnitz J.M., O’Connell J.P., “The properties of gases and liquids”,
McGraw-Hill Professional, USA (2000), ISBN: 0-07011-682-2

Purnama H., Kurr P., Schmidt A., Schomäcker R., Voigt I., Wolf A., and Warsitz R.,
[alpha]-methylstyrene hydrogenation in a flow-through membrane reactor”, AIChE
Journal, 52 (2006), 2805–2811

235
Pyrz W.D. and Buttrey D.J., "Particle size determination using TEM: A discussion of
image acquisition and analysis for the novice microscopist", Langmuir, 24 (2008),
11350-11360

Ratliff J.S., Tenney S.A., Hu X., Conner S.F., Ma S., and Chen D.A., “Decomposition
of dimethyl methylphosphonate on Pt, Au, and Au-Pt clusters supported on TiO
2
(110)”,
Langmuir, 25 (2009), 216-225

Reif M. and Dittmeyer R., “Porous, catalytically active ceramic membranes for gas-
liquid reactions: A comparison between catalytic diffuser and forced through flow
concept”, Catalysis Today, 82 (2003), 3–14

Rostami A.A., Mujumdar A.S., and Saniei N., “Flow and heat transfer for gas flowing
in microchannels: a review”, Heat and Mass Transfer, 38 (2002), 359-367

Rusu C.N. and Yates J.T., “Photooxidation of dimethyl methylphosphonate on TiO
2

powder”, Journal of Physical Chemistry, B, 104 (2000a), 12299-12305

Rusu C.N. and Yates J.T., “Adsorption and decomposition of DMMP on TiO
2

powder”, Journal of Physical Chemistry, B, 104 (2000b), 12292-12293

Ryu S.G., Yang J.K., Lee H.W., and Yang Y.S., “Decomposition of dimethyl
methylphosphonate over the alumina-supported precious metal catalysts”, HWAHAK
KONGHAM, 33 (1995), 462-470
Sahimi, M. and Tsotsis, T.T. “A percolation model of catalyst deactivation by site
coverage and pore blockage”, Journal of Catalysis, 96 (1985), 552–562

Sanchez Marcano J.G. and Tsotsis T.T., Catalytic membranes and membrane reactors,
Wiley-VCH, Weinheim (2002), ISBN: 3-527-30277-8

236
Saracco G. and Montanaro L., “Catalytic ceramic filters for flue gas cleaning. 1.
Preparation and characterization”, Industrial & Engineering Chemistry Research, 34
(1995), 1471-1479

Saracco G. and Montanaro L., “Catalytic ceramic filters for flue gas cleaning. 1.
Preparation and characterization”, Industrial & Engineering Chemistry Research, 34
(1995), 1471–1479

Saracco G. and Specchia V., “Catalytic ceramic filters for flue gas cleaning. 2.
Catalytic performance and modeling thereof”, Industrial & Engineering Chemistry
Research, 34 (1995), 1480–1487

Saracco G. and Specchia V., “Catalytic inorganic-membrane reactors – Present
experience and future opportunities”, Catalysis Reviews, Science and Engineering, 36
(1994), 305–384

Schmidt A. and Schomäcker R., “Partial hydrogenation of sunflower oil in a
membrane reactor”, Journal of Molecular Catalysis A: Chemical, 271 (2007), 192–199

Schmidt A., “A pore-flow-through membrane reactor for selective hydrogenation
reactions”, PhD thesis, Technische Universität Berlin (2007)

Schmidt A., Haidar R., and Reinhard Schomäcker R., “Selectivity of partial
hydrogenation reactions performed in a pore-through-flow catalytic membrane reactor”,
Catalysis Today, 104 (2005), 305–312
Schmidt A., Wolf A., Warsitz R., Dittmeyer R., Urbanczyk D., Voigt I., Fischer G.,
and Schomäcker R., “A pore-flow-through membrane reactor for partial hydrogenation of
1,5-cyclooctadiene”, AIChE Journal, 54 (2008), 258–268

Schomäcker R., Schmidt A., Frank B., Haidar R., and Seidel-Morgenstern A.,
“Membranen als Katalysatorträger”, Chemie Ingenieur Technik, 77 (2005), 549–558
237
Schuessler M., Lamla O., Stefanovski T., Klein C., and zur Megede D., “Autothermal
reforming of methanol in an isothermal reactor - concept and realization”, Chemical
Engineering & Technology, 24 (2001), 1141–1145

Segal S.R. and Suib S.L., “Photoassisted decomposition of dimethyl methylphosphonate
over amorphous manganese oxide catalysts”, Chemistry of Materials, 11 (1999), 1687-
1695

Segal S.R., Cao L., Suib S.L., Tang X., and Satyapal S., “Thermal decomposition of
dimethyl methylphosphonate over manganese oxide catalysts”, Journal of Catalysis, 198
(2001), 66–76

Sheinker V.N. and Mitchell M.B., “Quantitative study of the decomposition of
dimethyl methylphosphonate (DMMP) on metal oxides at room temperature and above”,
Chemistry of Materials, 14 (2002), 1257-1268

Sing K., “The use of nitrogen adsorption for the characterization of porous materials”,
Colloids and Surfaces A: Physicochemical and Engineering Aspects, 187-188 (2001), 3-9

Sircar S, Rao M.B, and Thaeron C.M.A., “Selective surface flow membrane for Gas
separation”, Separation Science and Technology, 34 (1999), 2081–2093

Smith W.F., “Foundations of Materials Science and Engineering”, McGraw-Hill
Science Engineering, USA (2005), ISBN: 0-07-295358-6
Sperry D.P., Falconer J.L., and Noble R.D., “Methanol-hydrogen separation by
capillary condensation in inorganic membranes”, Journal of Membrane Science, 60
(1991), 185-193

Splinter A., Sturmann J., Bartels O., and Benecke W., “Micro membrane reactor: A
flow-through membrane for gas pre-combustion”, Sensors and Actuators B: Chemical, 83
(2002), 169–174
238
Stair P., Marshall C., Xiong G., Feng H., Pellin M., Elam J., Curtiss L., Iton L., Kung
H., Kung M., and Wang H.H., “Novel, uniform nanostructured catalytic membranes”,
Topics in Catalysis, 39 (2006), 181–186

Stanislaus A., Absi-Halabi M., and Al-Dolama K., “Effect of phosphorus on the
Acidity of γ-Alumina and on the thermal stability of γ-Alumina supported nickel-
molybdenum hydrotreating catalysts”, Applied Catalysis, 39, 239-253 (1988)

Sullivan P.A. and Tester J.W., “Methylphosphonic acid oxidation kinetics in
supercritical water”, AIChE Journal, 50 (2004), 673 - 683

Templeton M.K., Weinberg W.H., “Adsorption and decomposition of dimethyl
methylphosphonate on an aluminum oxide surface”, Journal of the American Chemical
Society, 107 (1985a), 97-108

Templeton M.K., Weinberg W.H., “Decomposition of phosphate esters adsorbed on
aluminum oxide”, Journal of the American Chemical Society, 107 (1985b), 774-779
Tesfai T.M., Sheinker V.N, and Mitchell M.B.,”Decomposition of dimethyl
methylphosphonate (DMMP) on alumina-supported iron oxide”, Journal of Physical
Chemistry, B, 102 (1998), 7299–7302

Torres M., Lopez L., Dominguez J. M., Mantilla A., Ferrat G., Gutierrez M., and
Maubert M. “Olefins catalytic oligomerization on new composites of beta-zeolite films
supported on α-Al
2
O
3
membranes”, Chemical Engineering Journal, 92 (2003), 1–6

Trubitsyn D.A. and Vorontsov A.V., “Experimental study of dimethyl
methylphosphonate decomposition over Anatase TiO
2
”, Journal of Physical Chemistry,
B, 109 (2005), 21884-21892

Tsang, W., National Institute of Standards and Technology, Gaithersburg (1996)

239
Tsotsis T.T., Motamedhashemi M.M.Y., Nair N., Snyder M., and Navarro F., Hwang
H., “A Knudsen flow-through membrane-reactor-based integrated individual protection
system,” Proceedings Chemical and Biological Defense Physical Science and
Technology Conference (2008)

Tsuru T., Kan-no T., Yoshioka T., and Asaeda M., “A photocatalytic membrane
reactor for gas-phase reactions using porous titanium oxide membranes”, Catalysis
Today, 82 (2003), 41–48

Tzou T.Z. and Weller S.W., “Catalytic oxidation of dimethyl methylphosphonate”,
Journal of Catalysis, 146 (1994), 370-374

Uchytil P., “Gas permeation in ceramic membranes Part I. Theory and testing of
ceramic membranes”, Journal of Membrane Science, 97 (1994), 139- 144

Uchytil P., “Pore-size determination in the separation layer of a ceramic membrane
using the permeation method”, Journal of Material Science, 31 (1996), 6293-6298

Vincent M.J. and R. D. Gonzalez R.D., “Selective hydrogenation of acetylene through
a short contact time reactor”, AIChE Journal, 48 (2002), 1257–1267

Wehbea N., Guilhaumea N., Fiatyb K., Miachona S., and Dalmona J.A.,
“Hydrogenation of nitrates in water using mesoporous membranes operated in a flow-
through catalytic contactor”, Catalysis Today, 156 (2010) 208–215

Werner J.H. and Cool T.A., “Kinetic model for the decomposition of DMMP in a
Hydrogen/Oxygen flame”, Combustion and Flame, 117 (1999), 78–98

Westermann T. and Melin T., “Flow-through catalytic membrane reactors—Principles
and applications”, Chemical Engineering and Processing, 48 (2009), 17–28

240
Westermann T., Melin T., “Single flow-through catalytic membrane microchannel
reactor for intensified heterogeneous catalysis: characterisation and application to
hydrogenation of ethyne”, in: European Congress of Chemical Engineering (ECCE-6),
Copenhagen, Denmark (2007), 807–808

Wong Y.T. and Hoffmann R., “Chemisorption of carbon monoxide on three metal
surfaces: Ni (111), Pd (111), and Pt (111). A comparative study”, Journal of Physical
Chemistry, 95 (1991), 859-867

Yamada M., Fugii K., Haru H., and Itabashi K., “Preparation and catalytic properties
of special alumina membrane formed by anodic oxidation of aluminum”, Technical
report, The Light Metal Educational Foundation, Inc. Research Group for Functionalizing
of Aluminum and Its Surface Films (1988)

Yates D.J.C. and Sinfelt J.H., “The catalytic activity of Rhodium in relation to its state
of dispersion”, Journal of Catalysis, 8 (1967), 348-358

Sun Y. and Ong K.W., “Detection technologies for chemical warfare agents and toxic
vapors”, CRC press, Florida, USA (2005), ISBN 1-56670-668-8

Yoo S., Yeua S., Shermanb R.L., Simanek E.E., Shantza D.F., and Ford D.M.,
“Reverse-selective membranes formed by dendrimers on mesoporous ceramic supports”,
Journal of Membrane Science, 334 (2009), 16–22

Zalamea S., Pina M.P., Villellas A., Menendez M., and Santamaria J., “Combustion of
volatile organic compounds over mixed-regime catalytic membranes”, Reaction Kinetics
& Catalysis Letter, 67 (1999), 13-19

Zaspalis V.T., Van Praag W., Keizer K., Van Ommen J.G., Ross J.R.H., and
Burggraaf A.J., “Reaction of methanol over catalytically active alumina membranes”,
Applied Catalysis, 74 (1991), 205–222
241
Zegers E.J.P. and Fisher E.M., “Gas-phase pyrolysis of diethyl methylphosphonate”,
Combustion Science and Technology, 116-117(1996), 69-89

Zegers E.J.P. and Fisher E.M., “Gas-phase pyrolysis of diisopropyl
methylphosphonate”, Combustion and Flame, 115(1998), 230–240

Zegers E.J.P. and Fisher E.M., “Pyrolysis of triethyl phosphate”, Combustion Science
and Technology, 138 (1998), 85-103

Zhang F., Mangold M., Kienle A., “Optimal design of experiments for parameter
identification of ceramic porous membranes”, Chemical Engineering and Technology, 32
(2009), 641–649

Zhu B., Li H., and Yang W., “AgBiVMo oxide catalytic membrane for selective
oxidation of propane to acrolein”, Catalysis Today, 82 (2003), 91–98

242
Appendix
MATLAB
®
Program for simulation of FTCMR
The following MATLAB
®
program solves the transport-reaction equations developed
in Chapters 3 and 4 of this report for an isothermal tubular catalytic membrane in a
shell/tube reactor configuration.
Since the governing equations are derived based on the simplified dusty gas model
(DGM) for the case of very low content (ppm) of DMMP in a stream of Air, this program
can be used for similar cases only where one of the two components exists in such a low
value that the molar change of the other component in the reactive gas mixture can be
effectively neglected.
% ****************************************************************************************
function
STAGE_CUT,DMMP_CONVERSION,TUBE_MOLAR_FLOW_AIR,TUBE_MOLAR_FLOW_DMMP,SHELL_MOLAR_FLOW_AIR,SH
ELL_MOLAR_FLOW_DMMP]= SOLVER_L1_DMMP_C_PECLET
% ****************************************************************************************
FEED_DMMP = 0.001 ; % Mole fraction
XD_IN = FEED_DMMP ;
XD_OUT = 0.000 ;
%
MEMBRANE_LENGTH = 0.07 ; % m
%
REACTOR_TEMPERATURE = 300 ; % Centigrade Celsius (Isothermal reactor)
TK = REACTOR_TEMPERATURE + 273.15 ; % Kelvin
%
STANDARD_PRESSURE = 101325 ;
TUBE_PRESSURE = 101325+(0.5/14.7)*101325 ;% Pascal
SHELL_PRESSURE = 101325+(0.00/14.7)*101325 ;
SHELL_TO_TUBE_PRESSURE = SHELL_PRESSURE/TUBE_PRESSURE ;
%
[VISC]= GAS_VISCOSITY_L(TK,(0.5/101325)*(TUBE_PRESSURE+SHELL_PRESSURE)) ;
VISC_A= VISC(1,1) ;
R = 8.314 ; % [J/mole.K]
%
PORO1 = 0.3500 ;
%
TORT1 = 1/PORO1 ;
%
%
PORD1 = 50 * 10^(-10) ; % Pore Diameter of the top layer
STANDARD_PORD = 50 * 10^(-10) ; % Pore Diameter of basis
%
%
BE1 = (PORO1/TORT1)*(PORD1^2)/32 ; % Effective "B" parameter for Top layer
%
SWEEP_RATIO = 0.05 ; % Dimensionless [Shell-side volumetric flow to Tube-
side flow]
%**************************************************************************
FEED_FLOW_STD = 1.2* 10^(-6) ; % m3/sec [STD @ inlet]
SWEEP_FLOW_STD = SWEEP_RATIO * FEED_FLOW_STD ; % m3/sec [STD @ inlet]
SHELL_FLOW_STD = SWEEP_FLOW_STD ; % m3/sec [STD @ inlet]
243
TUBE_MOLAR_FLOW = FEED_FLOW_STD * 1000 * 273/298/22.4 ; % mole/sec [m3 => Liter,
STD => Normal]
SHELL_MOLAR_FLOW = SHELL_FLOW_STD * 1000 * 273/298/22.4 ; % mole/sec [m3 => Liter,
STD => Normal]
%
SWEEP_MOLAR_RATIO = SHELL_MOLAR_FLOW/TUBE_MOLAR_FLOW ;
DMMP_INLET_MOLAR_FLOW = TUBE_MOLAR_FLOW*XD_IN ;
% REMEMBER: Tube and shell flows are not equal to sweep and feed flows, because they are
changing along the tube and shell
% *************************************************************************
[KNUD_MOLD_EFFD]=DIFFUSION_L([1-XD_IN XD_IN 0 0 0 0
0],TK,0.5*(TUBE_PRESSURE+SHELL_PRESSURE),TORT1,PORO1,PORD1,2) ; % L: local means here

KNUD_D1= KNUD_MOLD_EFFD(2,1);
MOLD_D1= KNUD_MOLD_EFFD(2,2);
EFFD_D1= KNUD_MOLD_EFFD(2,3);
KNUD_A1= KNUD_MOLD_EFFD(1,1);
MOLD_A1= KNUD_MOLD_EFFD(1,2);
EFFD_A1= KNUD_MOLD_EFFD(1,3);
%
% STANDARD PECLET NUMBER
[KNUD_MOLD_EFFD]=DIFFUSION_L([1-XD_IN XD_IN 0 0 0 0
0],TK,0.5*(TUBE_PRESSURE+SHELL_PRESSURE),TORT1,PORO1,STANDARD_PORD,2); % L: local means
here
STANDARD_KNUD_A1= KNUD_MOLD_EFFD(1,1);
%
% For future remember you have to use permeation studies results so that
% you don't need to know the value of PORO/TORT
% DIMENTIONLESS GROUPS
SAY1 = BE1*TUBE_PRESSURE/VISC_A/KNUD_A1 ;
%
BETA1 = KNUD_A1/MOLD_D1 ;
%
DELTA1= EFFD_D1/KNUD_A1 ;
%
SIGMA1= MOLD_D1/KNUD_D1 ;
% *************************************************************************
% ALPHA1 = log(R_OUT/R_IN)
ALPHA1 = 0.006 ;
STANDARD_PHI = 30 ;
PHI = sqrt(STANDARD_KNUD_A1)*STANDARD_PHI/sqrt(KNUD_A1) ;
% return
%
% PECLET = TUBE_MOLAR_FLOW*R*TK/2/3.1415/MEMBRANE_LENGTH/TUBE_PRESSURE/KNUD_A1
;
STANDARD_PECLET =
TUBE_MOLAR_FLOW*R*TK/2/3.1415/MEMBRANE_LENGTH/STANDARD_PRESSURE/STANDARD_KNUD_A1
%
% return
% STANDARD_PECLET = 110
% return
% PECLET = 1 ;
% STANDARD_PECLET = 39.73 ;
%
[TUBE_FLUX_DMMP,SHELL_FLUX_DMMP,TUBE_FLUX_AIR,SHELL_FLUX_AIR,PRESSURE_PROFILE_DMMP,PRESSUR
E_PROFILE_AIR,DMMP_MOL_PERCENT,VISCOUS_RATIO,DMMP_NET_FLUX] =
TTT_DMMP_1_C(XD_IN,XD_OUT,TUBE_PRESSURE,SHELL_TO_TUBE_PRESSURE,PHI,ALPHA1,SAY1,DELTA1,SIGM
A1,BETA1)
% return
% *************************************************************************
% Now we attack to the main problem
%
%
% STANDARD_PECLETT = [9.5 9 8.5 8] ;
% PHII = [0.1 0.5 1.0 1.5 2]/sqrt(1) ;
% STANDARD_PECLETT = 40
% PHII = 1.85
% DMMP_CONVERSION = zeros(5,5) ;
% STAGE_CUT = zeros(5,5) ;
% TUBE_MOLAR_FLOW_AIR = zeros(5,5) ;
% TUBE_MOLAR_FLOW_DMMP = zeros(5,5) ;
244
% SHELL_MOLAR_FLOW_AIR = zeros(5,5) ;
% SHELL_MOLAR_FLOW_DMMP = zeros(5,5) ;
%
% for i=1:1:4
% for j=1:1:5
% STANDARD_PECLET =STANDARD_PECLETT(i)
% PHI =PHII(j)
%
%
% FEED_DMMP = 0.001 ; % Mole
fraction
% XD_IN = FEED_DMMP ;
% XD_OUT = 0.000 ;
%
%
SEGMENTS_NUMBER_Z = 10001 ; % The number of longitudinal segments
SEGMENTS_NUMBER_R = 100 ; % The number of radial segments
SNZ = SEGMENTS_NUMBER_Z ;
SNR = SEGMENTS_NUMBER_R ;
LL = linspace(0,1,SNZ) ;
SL = LL(2)-LL(1) ; % Length of
each segments
kk = 10 ;
% %
% TUBE_AIR_MOLAR_FLOW = zeros(1,kk) ;
% SHELL_AIR_MOLAR_FLOW = zeros(1,kk) ;
% TUBE_DMMP_MOLAR_FLOW = zeros(1,kk) ;
% SHELL_DMMP_MOLAR_FLOW = zeros(1,kk) ;
% %
% TUBE_MOLE_FRACTION_DMMP = zeros(1,kk) ;
% SHELL_MOLE_FRACTION_DMMP = zeros(1,kk) ;
%
DMMP_TUBE_FLUX = zeros(1,kk) ;
DMMP_SHELL_FLUX = zeros(1,kk) ;
AIR_TUBE_FLUX = zeros(1,kk) ;
AIR_SHELL_FLUX = zeros(1,kk) ;
%
% ZOOBIN_TUBE_AIR = zeros(1,kk) ;
% ZOOBIN_TUBE_DMMP = zeros(1,kk) ;
% ZOOBIN_SHELL_AIR = zeros(1,kk) ;
% ZOOBIN_SHELL_DMMP = zeros(1,kk) ;
% *************************************************************************
% Molar
n = 1 ; % Counter
% TUBE_MOLAR_FLOW_DMMP = TUBE_MOLAR_FLOW*(XD_IN) ;
% TUBE_MOLAR_FLOW_AIR = TUBE_MOLAR_FLOW*(1-XD_IN) ;
% SHELL_MOLAR_FLOW_DMMP = SHELL_MOLAR_FLOW*(XD_OUT) ;
% SHELL_MOLAR_FLOW_AIR = SHELL_MOLAR_FLOW*(1-XD_OUT) ;
% %
% TUBE_MOLAR_FLOW_AIR0 = TUBE_MOLAR_FLOW_AIR ;
% TUBE_MOLAR_FLOW_DMMP0 = TUBE_MOLAR_FLOW_DMMP ;
% %
% TUBE_AIR_MOLAR_FLOW(1,1) = TUBE_MOLAR_FLOW_AIR ;
% SHELL_AIR_MOLAR_FLOW(1,1) = SHELL_MOLAR_FLOW_AIR ;
% TUBE_DMMP_MOLAR_FLOW(1,1) = TUBE_MOLAR_FLOW_DMMP ;
% SHELL_DMMP_MOLAR_FLOW(1,1) = SHELL_MOLAR_FLOW_DMMP ;
% %
% TUBE_MOLE_FRACTION_DMMP(1,1) = XD_IN ;
% SHELL_MOLE_FRACTION_DMMP(1,1) = XD_OUT ;
% %
%
DMMP_TUBE_FLUX(1,1) = 0 ;
DMMP_SHELL_FLUX(1,1) = 0 ;
AIR_TUBE_FLUX(1,1) = 0 ;
AIR_SHELL_FLUX(1,1) = 0 ;
%
TUBE_ZOOBIN_AIR = (1-XD_IN) ;
TUBE_ZOOBIN_DMMP = (XD_IN) ;
SHELL_ZOOBIN_AIR = SWEEP_MOLAR_RATIO*(1-XD_OUT) ;
SHELL_ZOOBIN_DMMP = SWEEP_MOLAR_RATIO*(XD_OUT) ;
%
245
TUBE_ZOOBIN_AIR0 = TUBE_ZOOBIN_AIR ;
TUBE_ZOOBIN_DMMP0 = TUBE_ZOOBIN_DMMP ;
%
SHELL_ZOOBIN_AIR0 = SHELL_ZOOBIN_AIR ;
% SHELL_ZOOBIN_DMMP0 = SHELL_ZOOBIN_DMMP ;
% %
% ZOOBIN_TUBE_AIR(1,1) = TUBE_ZOOBIN_AIR ;
% ZOOBIN_TUBE_DMMP(1,1) = TUBE_ZOOBIN_DMMP ;
% ZOOBIN_SHELL_AIR(1,1) = SHELL_ZOOBIN_AIR ;
% ZOOBIN_SHELL_DMMP(1,1) = SHELL_ZOOBIN_DMMP ;
%
for L=0:SL:1
[TUBE_FLUX_DMMP,SHELL_FLUX_DMMP,TUBE_FLUX_AIR,SHELL_FLUX_AIR,PRESSURE_PROFILE_DMMP,PRESSUR
E_PROFILE_AIR,DMMP_MOL_PERCENT,VISCOUS_RATIO,DMMP_NET_FLUX] =
TTT_DMMP_L1_C_PECLET(XD_IN,XD_OUT,TUBE_PRESSURE,SHELL_TO_TUBE_PRESSURE,PHI,ALPHA1,SAY1,DEL
TA1,SIGMA1,BETA1);
% TUBE_ZOOBIN_AIR = TUBE_ZOOBIN_AIR + (SL) * ( (1/PECLET/ALPHA1 * TUBE_FLUX_AIR)
) ;
% SHELL_ZOOBIN_AIR = SHELL_ZOOBIN_AIR - (SL) * ( (1/PECLET/ALPHA1 *
SHELL_FLUX_AIR) ) ;
% TUBE_ZOOBIN_DMMP = TUBE_ZOOBIN_DMMP + (SL) * ( (DELTA1/PECLET/ALPHA1 *
TUBE_FLUX_DMMP) ) ;
% SHELL_ZOOBIN_DMMP = SHELL_ZOOBIN_DMMP - (SL) * ( (DELTA1/PECLET/ALPHA1 *
SHELL_FLUX_DMMP)) ;
%
TUBE_ZOOBIN_AIR = TUBE_ZOOBIN_AIR + (SL) * (
(1/(STANDARD_PECLET*STANDARD_KNUD_A1*STANDARD_PRESSURE/KNUD_A1/TUBE_PRESSURE)/ALPHA1
* TUBE_FLUX_AIR) ) ;
SHELL_ZOOBIN_AIR = SHELL_ZOOBIN_AIR - (SL) * (
(1/(STANDARD_PECLET*STANDARD_KNUD_A1*STANDARD_PRESSURE/KNUD_A1/TUBE_PRESSURE)/ALPHA1
* SHELL_FLUX_AIR) ) ;
TUBE_ZOOBIN_DMMP = TUBE_ZOOBIN_DMMP + (SL) * (
(DELTA1/(STANDARD_PECLET*STANDARD_KNUD_A1*STANDARD_PRESSURE/KNUD_A1/TUBE_PRESSURE)/ALPHA1
* TUBE_FLUX_DMMP) ) ;
SHELL_ZOOBIN_DMMP = SHELL_ZOOBIN_DMMP - (SL) * (
(DELTA1/(STANDARD_PECLET*STANDARD_KNUD_A1*STANDARD_PRESSURE/KNUD_A1/TUBE_PRESSURE)/ALPHA1
* SHELL_FLUX_DMMP)) ;
%
if TUBE_ZOOBIN_AIR<0
TUBE_ZOOBIN_AIR=0 ;
end
if TUBE_ZOOBIN_DMMP<0
TUBE_ZOOBIN_DMMP=0 ;
end
if SHELL_ZOOBIN_DMMP<0
SHELL_ZOOBIN_DMMP=0 ;
end
if SHELL_ZOOBIN_AIR<0
SHELL_ZOOBIN_AIR=0 ;
end
%
XD_IN = ( TUBE_ZOOBIN_DMMP /(TUBE_ZOOBIN_DMMP + TUBE_ZOOBIN_AIR) ) ;
%
% Only for Cross-flow
% XD_OUT = 0 ;
%
XD_OUT = ( SHELL_ZOOBIN_DMMP/(SHELL_ZOOBIN_DMMP + SHELL_ZOOBIN_AIR) ) ;
%
if L==0
DMMP_TUBE_FLUX(1,1) = TUBE_FLUX_DMMP ;
DMMP_SHELL_FLUX(1,1) = SHELL_FLUX_DMMP ;
end
%
if (L==0.1) || (L==0.2) || (L==0.3) ||(L==0.4) || (L==0.5) || (L==0.6) ||(L==0.7)
||(L==0.8) ||(L==0.9) ||(L==1.0)
n=n+1
end
%
if TUBE_ZOOBIN_AIR <0.001*TUBE_ZOOBIN_AIR0
break
end
246
end
%
%
DMMP_CONVERSION = 100*(1 - ( ( TUBE_ZOOBIN_DMMP+SHELL_ZOOBIN_DMMP ) /
TUBE_ZOOBIN_DMMP0) ) ;
STAGE_CUT = 100*(SHELL_ZOOBIN_AIR-SHELL_ZOOBIN_AIR0)/TUBE_ZOOBIN_AIR0
;
TUBE_MOLAR_FLOW_AIR = TUBE_ZOOBIN_AIR * TUBE_MOLAR_FLOW ;
TUBE_MOLAR_FLOW_DMMP = TUBE_ZOOBIN_DMMP * TUBE_MOLAR_FLOW ;
SHELL_MOLAR_FLOW_AIR = SHELL_ZOOBIN_AIR * TUBE_MOLAR_FLOW ;
SHELL_MOLAR_FLOW_DMMP = SHELL_ZOOBIN_DMMP * TUBE_MOLAR_FLOW ;
%
%
% DMMP_CONVERSION = 100*(1 - ( ( TUBE_ZOOBIN_DMMP+SHELL_ZOOBIN_DMMP ) /
TUBE_ZOOBIN_DMMP0) )
% STAGE_CUT = 100*(SHELL_ZOOBIN_AIR-SHELL_ZOOBIN_AIR0)/TUBE_ZOOBIN_AIR0
% % SHELL_MOLE_FRACTION_DMMP = XD_OUT
;
% % TUBE_MOLE_FRACTION_DMMP = XD_IN
;
% TUBE_MOLAR_FLOW_AIR = TUBE_ZOOBIN_AIR * TUBE_MOLAR_FLOW
% TUBE_MOLAR_FLOW_DMMP = TUBE_ZOOBIN_DMMP * TUBE_MOLAR_FLOW
% SHELL_MOLAR_FLOW_AIR = SHELL_ZOOBIN_AIR * TUBE_MOLAR_FLOW
% SHELL_MOLAR_FLOW_DMMP = SHELL_ZOOBIN_DMMP * TUBE_MOLAR_FLOW
% %
%
% end
% end
%
%
% *************************************************************************
function [VISC]=GAS_VISCOSITY_L(TK,PBar)
% This function estimates the viscosity of gas components using the method
% of corresponding states (The Properties of Gases and Liquids, 1966,pp405)
% PPa = OPERATING PRESSURE, Pascal
% PAVG = AVERAGE PRESSURE, Pascal
% X = GAS COMPONENT MOLE FRACTION MATRIX
% TK = OPERATING TEMPERATURE, K
% 1 = Air
% 2 = C3H9O3P (DMMP)
% 3 = CO2
% 4 = H2O
% 5 = CH4O (MeOH)
% 6 = C2H6O (DME)
% 7 = H3PO4
% 8 = P2O5
% 9 = P4O10
% 10 = CO
% 11 = H2
% 12 = C2H7PO4 (Dimethyl Phosphate)
% 13 = CH5PO4 (Monomethyl Phosphate)
% 14 = C2H7PO3 (Monomethyl Methylphosphonate)
% 15 = CH5PO3 (Methylphosphonic Acid)
% 16 = He (Helium)
% *************************************************************************
% Critical Properties from "The Properties of Gases and Liquids, 1966"
% TC = [K]
% PC = [Bar]~[Atm]
TC(1) = 132.5 ;
TC(2) = 700.6 ;
TC(3) = 304.2 ;
TC(4) = 647.0 ;
TC(5) = 513.2 ;
TC(6) = 400.1 ;
% TC(7) = ?
% TC(8) = ?
% TC(9) = ?
% TC(10) = 133.0;
% TC(11) = 33.3 ;
% TC(12) = ?
% TC(13) = ?
247
% TC(14) = ?
% TC(15) = ?
TC(16) = 5.19 ;
% *************************************************************************
PC(1) = 37.2 ;
PC(2) = 49.7 ;
PC(3) = 72.9 ;
PC(4) = 218.3 ;
PC(5) = 78.5 ;
PC(6) = 52.6 ;
% PC(7) = ?
% PC(8) = ?
% PC(9) = ?
% PC(10) = 34.5 ;
% PC(11) = 12.8 ;
% PC(12) = ?
% PC(13) = ?
% PC(14) = ?
% PC(15) = ?
PC(16) = 2.28 ;
% *************************************************************************
MW(1) = 29 ;
MW(2) = 124 ;
MW(3) = 44 ;
MW(4) = 18 ;
MW(5) = 32 ;
MW(6) = 46 ;
% MW(7) = 98 ;
% MW(8) = 142 ;
% MW(9) = 284 ;
% MW(10) = 28 ;
% MW(11) = 2 ;
% MW(12) = 126 ;
% MW(13) = 112 ;
% MW(14) = 110 ;
% MW(15) = 96 ;
MW(16) = 4 ;
% *************************************************************************
PC = [PC(1) PC(3) PC(16)] ;
TC = [TC(1) TC(3) TC(16)] ;
MW = [MW(1) MW(3) MW(16)] ;
VISCOSITY = zeros(1,length(PC)) ;
PR = PBar./PC ;
TR = TK./TC ;
for i=1:1:length(PC)
if TR(1,i)<1.5
VISCOSITY(1,i)= (34.00)*(0.00001)*((TR(1,i))^0.94) * ((PC(1,i))^(2/3))
* sqrt(MW(1,i)) / ((TC(1,i))^(1/6)) ;
else
VISCOSITY(1,i)= (17.78)*(0.00001)*((4.58*TR(1,i)-1.67)^(5/8)) * ((PC(1,i))^(2/3))
* sqrt(MW(1,i)) / ((TC(1,i))^(1/6)) ;
end
end
VISC = 0.001*VISCOSITY;
% *************************************************************************
function [KNUD_MOLD_EFFD]=DIFFUSION_L(X,TK,PPa,TORT,PORO,PORD,NC)
% This sub-function estimates the molecular, bulk, and Knudsen diffusion
% for all gases present in the gas mixture. The reference is
% "The Properties of Gases and Liquids", 5th edition
% PPa = OPERATING PRESSURE, Pa
% X = GAS COMPONENT MOLE FRACTION MATRIX
% TK = OPERATING TEMPERATURE, K
% NC = Number of component
% 1 = Air
% 2 = C3H9O3P (DMMP)
% 3 = CO2
% 4 = H2O
% 5 = CH4O (MeOH)
% 6 = C2H6O (DME)
% 7 = H3PO4
% 8 = P2O5
248
% 9 = P4O10
% 10 = CO
% 11 = H2
% 12 = C2H7PO4 (Dimethyl Phosphate)
% 13 = CH5PO4 (Monomethyl Phosphate)
% 14 = C2H7PO3 (Monomethyl Methylphosphonate)
% 15 = CH5PO3 (Methylphosphonic Acid)
% *************************************************************************
% ADV = Atomic Diffusion Volume (page:11.11, table 11.1)
% ADV_C = Carbon = 15.9
% ADV_O = Oxygen = 6.11
% ADV_H = Hydrogen = 2.31
% ADV_P = Phosphorous (To be assumed as that of "S") = 22.9
% ADV_CO2 = 26.9
% ADV_CO = 18.0
% ADV_H2O = 13.1
% ADV_H2 = 6.12
% ADV_Air = 19.7
% NC = Number of Components
% *************************************************************************
% I use the Fuller formula in this reference [section 11.10,
% equation(11-4.4)] for binary gases; while for multicomponet-diffusion
% coefficient I use Blanc?s law [page: 11.20, equation(11-7.4)]
% Assumptions:
% 1- Pressure is constant and is equal to 20 psig = 1.378 Bar
% 2- D_AB = D_BA
% 3- D = (1/D_KNUDSEN + 1/D_MOLECULAR)^(-1) + D_Poiseuille
% PPa = 200000
% X=[0.25 0.25 0.25 0.25]
% TK = 500
% PORO = 0.50 ; % POROSITY OF MEMBRANE/CATALYST
% PORE_DIAMETER_A = 500*10^(-10) ; % A-TYPE MEMBRANE
% PORE_DIAMETER_B = 100*10^(-10) ; % B-TYPE MEMBRANE
% PORE_DIAMETER_C = 40*10^(-10) ; % C-TYPE MEMBRANE
% PORE_DIAMETER = PORE_DIAMETER_C ;
% PORD = PORE_DIAMETER ;
% NC = 4 ; % Number_of_Components
PBar = PPa/(101325) ;
% ******************************;
% For Champagnie: Ethane, Ethylene, Hydrogen, and Argon
% ******************************;
% TORT = 1/PORO ; % TORTOUSITY OF MEMBRANE/CATALYST
TOPO = PORO/TORT ; % Effective Factor
% ******************************;
% For Champagnie:
MW = zeros(NC,1) ;
ADV = zeros(NC,1) ;
KN_DIFF = zeros(NC,1) ;
ML_DIFF = zeros(NC,1) ;
% *******************************
MW(1) = 29 ;
MW(2) = 124 ;
MW(3) = 44 ;
MW(4) = 18 ;
MW(5) = 32 ;
MW(6) = 46 ;
MW(7) = 98 ;
% MW(8) = 142 ;
% MW(9) = 284 ;
% MW(10) = 28 ;
% MW(11) = 2 ;
% MW(12) = 126 ;
% MW(13) = 112 ;
% MW(14) = 110 ;
% MW(15) = 96 ;
% *******************************
% Champagnie
% MW(1) = 30 ;% Ethane C2H6
% MW(2) = 28 ;% Ethylene C2H4
% MW(3) = 2 ;% Hydrogen H2
% MW(4) = 40 ;% Argon Ar
249
% *******************************
% % Champagnie
% ADV_C = 15.9 ;
% ADV_O = 6.11 ;
% ADV_H = 2.31 ;
% ADV_H2 = 6.12 ;
% ADV_Ar = 16.2 ;
% *******************************
ADV_C = 15.9 ;
ADV_O = 6.11 ;
ADV_H = 2.31 ;
ADV_P = 22.9 ;
% *******************************
ADV_CO2 = 26.9 ;
ADV_CO = 18.0 ;
ADV_H2O = 13.1 ;
ADV_H2 = 6.12 ;
ADV_Air = 19.7 ;
% *******************************
ADV(1) = ADV_Air ;% ADV_Air
ADV(2) = 9*ADV_H + 3*ADV_C + 3*ADV_O + ADV_P ;% ADV_DMMP
ADV(3) = ADV_CO2 ;% ADV_CO2
ADV(4) = ADV_H2O ;% ADV_H2O
ADV(5) = ADV_C + ADV_O + 4*ADV_H ;% ADV_MeOH
ADV(6) = 2*ADV_C + 6*ADV_H + ADV_O ;% ADV_DME
ADV(7) = 3*ADV_H + ADV_P + 4*ADV_O ;% ADV_DME
% *******************************
% Champagnie
% ADV_C2H6 = 2 * ADV_C + 6 * ADV_H ;
% ADV_C2H4 = 2 * ADV_C + 4 * ADV_H ;
% ADV(1) = ADV_C2H6 ;
% ADV(2) = ADV_C2H4 ;
% ADV(3) = ADV_H2 ;
% ADV(4) = ADV_Ar ;
% *******************************
% CALCULATION OF KNUDSEN DIFFUSION
for k=1:1:NC
KN_DIFF(k)= PORD * 48.5 *((TK/MW(k))^0.5) ;% m2/sec
end
% CALCULATION OF MOLECULAR DIFFUSION BASED ON FULLER'S FORMULA
for i=1:NC
for j=1:NC
% % if j==i
% % ML_DIFF(i,j) =1 ;
% else
MW_AVG = 2*((1/MW(i) + 1/MW(j)).^-1) ;
ML_DIFF(i,j)= 10^(-4) * (0.00143*TK^(1.75))/PBar/(MW_AVG^0.5)/((ADV(i)^(0.334)
+ ADV(j)^(0.334))^2) ;% m2/sec
%ML_DIFF(j,i)= ML_DIFF(i,j) ;
% end
end
end
% *************************************************************************
% % CALCULATION OF MULTICOMPONENT DIFFUSION ACCORDING TO Blanc's law
% MIX_DIFF=zeros(NC,1) ;
% for i=1:NC
% for j=1:NC
% if j==i
% q=0 ;
% else
% q=1 ;
% end
% MIX_DIFF(i) = q * ( X(j)/ML_DIFF(i,j) )+ MIX_DIFF(i) ;
% end
% MIX_DIFF(i) = (MIX_DIFF(i))^-1 ;
% end
% *************************************************************************
% CALCULATION OF MULTICOMPONENT DIFFUSION
MIX_DIFF=zeros(NC,1) ;
for i=1:NC
for j=1:NC
250
if j==i
q=0 ;
else
q=1 ;
end
MIX_DIFF(i) = q * ( X(j)/ML_DIFF(i,j) )+ MIX_DIFF(i) ;
end
MIX_DIFF(i) = (1-X(i))/(MIX_DIFF(i)) ;
end
% **************
% if X(i)==1
% MIX_DIFF(i)=0 ;
% else
% MIX_DIFF(i) = (1-X(i))/(MIX_DIFF(i)) ;
% end
% [GAS_DIFFUSION] = TOPO * [1./MIX_DIFF + 1./KN_DIFF].^-1 ;
% b = ML_DIFF
% x = MIX_DIFF
EFFD = (1./MIX_DIFF + 1./KN_DIFF).^-1 ;
[KNUD_MOLD_EFFD] = TOPO*[KN_DIFF,MIX_DIFF,EFFD] ;
function
[TUBE_FLUX_DMMP,SHELL_FLUX_DMMP,TUBE_FLUX_AIR,SHELL_FLUX_AIR,PRESSURE_PROFILE_DMMP,PRESSUR
E_PROFILE_AIR,DMMP_MOL_PERCENT,VISCOUS_RATIO,DMMP_NET_FLUX] =
TTT_DMMP_L1_C_PECLET(XD_IN,XD_OUT,TUBE_PRESSURE,SHELL_TO_TUBE_PRESSURE,PHI,ALPHA1,SAY1,DEL
TA1,SIGMA1,BETA1)
% This program calculates the DMMP pressure profile inside membrane
% *************************************************************************
% Loop starting point
% *************************************************************************
GUESS_A = 0.5*1.000*(1+SHELL_TO_TUBE_PRESSURE) ;
GUESS_D = 0.5*(XD_IN+XD_OUT)*(1+SHELL_TO_TUBE_PRESSURE) ;
solinit = bvpinit(linspace(0,1,10),[GUESS_D GUESS_D GUESS_A GUESS_A
]) ;
options = bvpset('RelTol', 1e-3,'AbsTol',1e-8,'NMax',10000000)
;
sol =
bvp4c(@MEMBRANE_dydx,@MEMBRANE_bc,solinit,options,XD_IN,XD_OUT,TUBE_PRESSURE,SHELL_TO_TUBE
_PRESSURE,PHI,ALPHA1,SAY1,DELTA1,SIGMA1,BETA1) ;
xx = linspace(0,1,10) ;
yy = deval(sol,xx) ;
%
TUBE_FLUX_DMMP = ( yy(2,1) + yy(4,1) *yy(1,1 )*BETA1*(1/yy(3,1) + SAY1
+ SAY1*SIGMA1) ) ;
SHELL_FLUX_DMMP = ( yy(2,10) + yy(4,10) *yy(1,10)*BETA1*(1/yy(3,10) + SAY1
+ SAY1*SIGMA1) ) ;
TUBE_FLUX_AIR = ( yy(4,1) ) * (1+SAY1*yy(3,1) ) ;
SHELL_FLUX_AIR = ( yy(4,10) ) * (1+SAY1*yy(3,10)) ;
%
PRESSURE_PROFILE_DMMP = TUBE_PRESSURE*yy(1,:)' ;
PRESSURE_PROFILE_AIR = TUBE_PRESSURE*yy(3,:)' ;
oo = length(yy(1,:)) ;
DMMP_MOL_PERCENT = 100*[yy(1,:)./(yy(3,:)+(yy(1,:)))] ;
VISCOUS_RATIO = (BETA1*yy(4,:).*yy(1,:).*(1./yy(3,:) + SAY1 +
SAY1*SIGMA1)./yy(2,:))' ;
DMMP_NET_FLUX = TUBE_FLUX_DMMP-SHELL_FLUX_DMMP
;
% *************************************************************************
function dydx =
MEMBRANE_dydx(x,y,XD_IN,XD_OUT,TUBE_PRESSURE,SHELL_TO_TUBE_PRESSURE,PHI,ALPHA1,SAY1,DELTA1
,SIGMA1,BETA1)
%
dydx = [ y(2)
1*y(1)*exp(2*x*ALPHA1)*(PHI^2)/DELTA1 - ( y(2)* y(4) * BETA1 * ( 1/y(3) +
SAY1 + SAY1*SIGMA1 ) - y(1)* BETA1*( y(4)^2 )*( ( SAY1/(1+SAY1*y(3)) ) * (1/y(3) + SAY1
+ SAY1*SIGMA1) + (1/y(3)^2) ) )
y(4)
- y(4)^2* ( SAY1/(1+SAY1*y(3)) )];
% *************************************************************************
% Continuity for conentration and flux at the interfaces
251
function res=
MEMBRANE_bc(ya,yb,XD_IN,XD_OUT,TUBE_PRESSURE,SHELL_TO_TUBE_PRESSURE,PHI,ALPHA1,SAY1,DELTA1
,SIGMA1,BETA1)
%
D = ( yb(2) + yb(4) * yb(1)*BETA1*(1/yb(3) + SAY1 + SAY1*SIGMA1) ) ;
A = (1/DELTA1)* yb(4) * (1+SAY1 * yb(3) ) ;
% VERY IMPORTANT ! Terms(1-XD_IN) and (1-XD_OUT) are eliminated from Air
res=[ ya(1) - (XD_IN)
yb(1) - (XD_OUT)*SHELL_TO_TUBE_PRESSURE
% yb(1) - (D/(A+D))*SHELL_TO_TUBE_PRESSURE
ya(3) - (1-XD_IN )
% yb(3)-(SHELL_TO_TUBE_PRESSURE-yb(1)) ];
yb(3) - (1-XD_OUT)*SHELL_TO_TUBE_PRESSURE ] ;
% yb(3) - (A/(A+D))*SHELL_TO_TUBE_PRESSURE ] ;
% *************************************************************************

Abstract (if available)

Abstract The possibility of the use of chemical weapons has been increased in recent years both as a result of potential terrorist attacks and of ongoing international conflicts. The focus of our research is the development of a novel, hybrid catalytic membrane reactor system, which consists of a flow-through catalytic membrane reactor (FTCMR) integrated with a surface-flow based membrane separator (SFMS). The combined system has been shown to achieve complete oxidation of chemical warfare agents (CWA) at trace levels, and is appropriate for use in integrated individual protection (IP) systems as well as collective protection (CP) systems for civil and military applications. ❧ As a part of this research, a catalytic tubular alumina membrane is prepared via impregnation with chloroplatinic acid solutions, and is utilized in a FTCMR for the catalytic oxidation of dimethyl methylphosphonate (DMMP) in air. DMMP is known as a chemical precursor for the more toxic gas Sarin (GB), and has been widely used to simulate its characteristics. ❧ In this Thesis experiments are reported for different DMMP feed concentrations (150-1000 ppm) and reactor temperatures (373-573K), which demonstrate the potential advantage of the FTCMR in the complete catalytic oxidation of this important CWA simulant. Complete destruction of low and high concentrations of DMMP was achieved at lower temperatures compared to the values obtained in this study for a wall-coated plug-flow (monolith) reactor containing the same amount of catalytic metal. ❧ A mathematical model has also been developed in order to provide a better understanding of the fundamental transport phenomena underpinning the FTCMR operation. It makes use of the Dusty Gas Model (DGM), which incorporates in an appropriate fashion continuum and Knudsen diffusion, and viscous flow as the mechanisms for gas transport through the porous membrane being utilized in the FTCMR. In the first application, the model is used for identifying the advantages of the FTCMR concept compared to the wall-coated catalytic monolith, and also for investigating some of the limitations, which may exist in applying this concept for the complete oxidation of chemical warfare simulants. The results of the model support the superiority of the FTCMR concept over the more conventional (plug-flow) monolith reactor. ❧ During the FTCMR experiments it was found that one of the challenges associated with the catalytic destruction of DMMP is catalytic membrane deactivation via active site coverage and pore-blockage. Therefore, in the second application, the model is extended to incorporate the effect of catalyst deactivation and pore-blockage, on the performance of the FTCMR. The model is successfully applied to determine from experimental data important parameters such as the reaction rate constants and the poisoning and pore plugging factors. The simulation results also confirm the experimental observations in that the protection time provided by the FTCMR is a function of the DMMP concentration in the feed, pointing out that an appropriate role for the FTCMR to play is as a second stage in a hybrid system, following a bulk-toxin removal unit in the first stage. The main advantage of the proposed hybrid system, combining a surface-flow membrane (SFM) separation unit (which is capable of continuously physically removing a large portion of the CWA from contaminated air streams) with the FTCMR, is that it completely destructs the CWA that remains with a lower rate of pore blockage, thus resulting in the continuous CWA destruction for extended time periods, which are appropriate for both IP and CP applications.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The purification of contaminated air streams via recative separation techniques

PDF

Methanol synthesis in a membrane reactor

PDF

Methanol synthesis in the membrane reactor

PDF

Novel methods for landfill gas and biogas clean-up

PDF

A hybrid adsorbent-membrane reactor (HAMR) system for hydrogen production

PDF

Process intensification in hydrogen production via membrane-based reactive separations

PDF

The use of carbon molecule sieve and Pd membranes for conventional and reactive applications

PDF

Mesoscale SOFC-based power generator system: modeling and experiments

PDF

A study of the application of membrane-based reactive separation to the carbon dioxide methanation

PDF

Methanol synthesis in a contactor-type membrane reactor

PDF

On the use of membrane reactors in biomass utilization

PDF

Lab-scale and field-scale study of siloxane contaminants removal from landfill gas

PDF

Chemical and mechanical deformation of porous media and materials during adsorption and fluid flow

PDF

A process-based molecular model of nano-porous silicon carbide membranes

PDF

A high efficiency, ultra-compact process for pre-combustion CO₂ capture

PDF

Continuum and pore netwok modeling of preparation of silicon-carbide membranes by chemical-vapor deposition and chemical-vapor infiltration

PDF

The roles of surface and pore properties in wetting resistance for membrane distillation membranes

PDF

Controlling polymer film patterning, morphologies, and chemistry using vapor deposition

PDF

Silicon carbide ceramic membranes

PDF

Experimental studies and computer simulation of the preparation of nanoporous silicon-carbide membranes by chemical-vapor infiltration/chemical-vapor deposition techniques

Asset Metadata

Creator Motamedhashemi, Mirmohammadyousef (author)

Core Title A flow-through membrane reactor for destruction of a chemical warfare simulant

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Chemical Engineering

Publication Date 11/21/2012

Defense Date 10/05/2012

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag catalytic membrane reactors,chemical warfare agent,CWA,DMMP,flow-through Membrane Reactors,FTCMR,OAI-PMH Harvest

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Tsotsis, Theodore T. (committee chair), Egolfopoulos, Fokion N. (committee member), Gupta, Malancha (committee member)

Creator Email motamedh@usc.edu,myhashemi@hotmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-113772

Unique identifier UC11290435

Identifier usctheses-c3-113772 (legacy record id)

Legacy Identifier etd-Motamedhas-1320.pdf

Dmrecord 113772

Document Type Dissertation

Rights Motamedhashemi, Mirmohammadyousef

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA