UNIVERSITY OF CINCINNATI

Date:______

I, ______, hereby submit this work as part of the requirements for the degree of: in:

It is entitled:

This work and its defense approved by:

Chair: ______

SYNTHESIS OF ORDERED MESOPOROUS SILICA AND ALUMINA

WITH CONTROLLED MACROSCOPIC MORPHOLOGIES

A dissertation submitted to the

Division of Graduate Studies and Research of the University of Cincinnati

in partial fulfillment of the requirements of the degree of

DOCTOR OF PHILOSOPHY

in the Department of Chemical Engineering of the College of Engineering

November, 2004

by

Hatem Mohammad Sadi Alsyouri

B.Sc. Chemical Engineering, University of Jordan, 1998

Dissertation Advisor and Committee Chair: Dr. Jerry Y. S. Lin

ABSTRACT

The ability to synthesize nanostructured inorganic materials with controlled microstructural and morphological features will provide materials with unique characteristics in unprecedented ways. This thesis investigates the synthesis of porous silica and alumina materials with controlled microstructures and desirable shapes using novel approaches based on template-assisted synthesis and chemical vapor deposition

(CVD) techniques. It primarily focuses on fabricating mesoporous materials with unique microstructures and different morphologies (particles and membranes) and exploring the potential of the particle morphology in a polymer reaction application.

The template-assisted growth of mesoporous silica under acidic and quiescent conditions at an oil-water interface can generate mesostructured silica at the interface with fibrous, gyroidal, spherical, and film morphologies. Synthesis conditions can be used to alter the growth environment and control the product morphology. Fiber morphology is obtained at narrow range of experimental conditions due to slow and one- dimensional diffusion of silicon alkoxide through the interface. Variation in these conditions can alter the axial growth of silica and yield non-fibrous shapes. The fibers grow from their base attached to the interface and coalesce to form fibers with larger diameters. Gas transport in the mesoporous silica fibers is governed by combination of

Knudsen and surface diffusion mechanisms. Surface diffusion contributes to 40% of the net flow reflecting a highly smooth pore surfaces. Real Knudsen and surface diffusivities are in the order of 10-3 and 10-5 cm2/s respectively. The one-dimensional mesopores are

45 time longer than the macroscopic fiber length and align helically around the fiber axis, confirming the literature observations, with a pitch value of 1.05 micron. For preparation of mesoporous silica materials as membranes, a novel counter diffusion self assembly (CDSA) approach is demonstrated. This approach, adopted from growth of silica fibers, introduces the precursors from the opposite sides of a ceramic supports and yields silica plugs grown within the support pores. Growth of defect-free silica membrane by this approach requires a hydrophobic support to enhance diffusion of the silicon alkoxide. According to gas permeation properties, silica plugs grow with thickness of ~ 0.5 mm and have a mesoporous structure. Such mechanically strong membrane offers high potential in protein separation and polymer reaction applications.

Mesoporous membranes with controlled pore microstructure can be also obtained using CVD technique. Cyclic CVD modification of straight 20 nm pore alumina membranes demonstrates that variation of the precursor introduction scheme can affect the microstructure of alumina deposition within the support pores. Leaving residual of precursors in the support pore after introduction of each precursor causes deposition of alumina in a fractal structure suitable for gas separation applications. Purging the pore after each precursor, on the other hand, causes alumina to deposit in an atomic layer fashion to give cylindrical mesopores suitable for membrane reaction applications.

The use of ordered mesoporous materials supported with titanocene catalyst as nano-reactors for the extrusion of high quality polyethylene (PE) fibers is demonstrated.

The steric effect of the straight 3 nm pores of mesoporous silica template the growth of

60 nm diameter PE fibers with extended-chain crystalline structures. The nascent polymer fibers aggregate into 1−30 µm microfibers which further aggregate into PE fiber bundles. Mechanical properties, measured for the first time, demonstrate that these fibers exhibit improved tensile strength compared to commercial PE fibers.

To my parents

To my wife

And To my son

iv

“And when your Lord said to the angels, I am going to place in the earth a khalif, they said: What! wilt Thou place in it such as shall make mischief in it and shed blood, and we celebrate Thy praise and extol Thy holiness? He said: Surely I know what you do not know. And He taught Adam all the names, then presented them to the angels; then

He said: Tell me the names of those if you are right. They said: Glory be to Thee! we have no knowledge but that which Thou hast taught us; surely Thou art the Knowing, the Wise. He said: O Adam! inform them of their names. Then when he had informed them of their names, He said: Did I not tell you that I know the secrets of heaven and earth, and I know what ye reveal and what ye conceal?”

From Holy Qur’an 2:30-32

v ACKNOWLEDGEMENTS

First of all I would like to thank our creator the Lord (Allah) for His blessings of brain and senses. Thank you Allah for supporting me with faith, strength, patience, and motivation without which I would not have achieved this. Thanks for blessing me with a righteous wife and decent friends who were good companions through this long journey.

O Allah, advance my knowledge and help me utilizing it for your sake.

I would like to express my sincere gratitude and respect to my dissertation advisor,

Dr. Jerry Y. S. Lin for his support, kindness, patience and commitment to excellence throughout the course of this research. His knowledge, intelligence, efficiency and organization have always been a source of inspiration for me to become a distinguished researcher. Without him, this dissertation would not have finished. I would also like to thank the committee members; Dr. William Krantz, Dr. Gregory Beaucage, and Dr. Chia-

Chi Ho for their scientific support and valuable times in reviewing this dissertation.

I am grateful to Dr. Shiping Zhu and Dr. Zhibin Ye from McMaster University at

Canada for their distinguished efforts in our collaborative polymerization work. I would like also to acknowledge Dr. Vadim Guliants who gave me the opportunity to expand my knowledge and experience in the membrane area by participating in his OCDO membrane project for few months. Thanks are also due to Dr. Sun-Tak Hwang and Dr.

Carlos Co for their valuable articles and times for useful discussions.

I would like to greet all the former and current members of our group for their kindness, help and friendship during my stay at the University of Cincinnati, especially

Dr. M. Pan, Dr. X. Qi, Dr. G. Buelna, Dr. T. Akin, Dr. Z. Yang, Dr. C. Cooper, Dr. W.

Yuan, Dr. H. Zou, Dr. J. Ida, C. Langheinrich, S. Cheng, R. Xiong, N. Rane, M.

vi Skrobanek , A. Daumichen, Q. Yang, Q. Yin, S. Gladding, A. Chung, Dr. J. Park, V.

Gupta, D. Li, D. Singh and Z. Zheng. I owe thanks to Srinivas Subramaniam and Dr.

Reghvendra Tewari from the materials characterization center at UC and Dr. Mohamed

Hassan from the department of Chemistry at UC for their dedicated help in material

characterization of my samples.

Outside the academic world, my sincere gratitude and love goes to my parents,

brothers, sisters and friends in Jordan for their emotional support and encouragement that

made my study more comfortable all these years away from home. I am also indebted to my parents in law who sacrificed lots of their efforts and time to help me complete my dissertation. I wish also to extend my gratitude to all my friends who made my stay in

Cincinnati full of happiness and excitement.

Finally, I owe a great deal to my wife Malyuba Abu Daabes for her love, patience and understanding. Since we have met in our undergraduate study in Jordan, she has been my real source of encouragement and motivation for excellence in my academic and personal life. I realize how hard was her life as a wife and mother while being a graduate student. Thanks for all your sacrifice and patience. Above of all, thanks for gifting me with the dearest person to my heart, my son, Ibrahim. This little guy whose natural instincts and smile has reduced much of my burden as a father and provided me inspiration as a researcher in performing my work.

vii TABLE OF CONTENTS

List of Tables xii List of Figures xiv

CHAPTER 1. INTRODUCTION 1 1.1. Porous Materials, 1 1.2. Ordered Mesoporous Materials, 3 1.2.1. Overview, 3 1.2.2. Mechanism of formation, 7 1.2.3. Synthesis pathways, 9 1.3. Morphological Control of Mesoporous Materials, 12 1.4. Applications, 14 1.5. Chemical Vapor Deposition, 15 1.6. Research Objectives, 16 1.6.1. Motivation of this research, 16 1.6.2. Synthesis and characterization of ordered mesoporous silica fibers, 20 1.6.3. Ordered mesoporous inorganic membranes with controlled microstructure, 21 1.6.4. Polyethylene extrusion polymerization in ordered mesoporous silica materials, 23 1.7. Structure of Dissertation, 24 1.8. References, 25

CHAPTER 2. QUIESCENT INTERFACIAL SYNTHESIS OF MORPHOLOGICAL NANOSTRUCTURED SILICA 35 2.1. Introduction, 35 2.2. Experimental, 37

viii 2.2.1. Quiescent interfacial growth and characterization of mesoporous silica, 37 2.2.2. Ethylene extrusion polymerization, 40 2.3. Results and Discussion of Interfacial Growth, 42 2.3.1. Results of synthesis, 42 2.3.2. Macroscopic properties of silica fibers, 45 2.3.3. Microscopic properties of silica fibers, 51 2.3.4. Non-fibrous morphologies, 59 2.3.5. Mechanism of fiber formation, 69 2.4. Results and Discussion of Ethylene extrusion polymerization, 71 2.5. Conclusions, 78 2.6. Reference, 79

CHAPTER 3. GAS DIFFUSION AND MICROSTRUCTURAL PROPERTIES OF MESOPOROUS SILICA FIBERS 82 3.1. Introduction, 82 3.2. Experimental, 85 3.3. Results and Discussion, 87 3.3.1. Characteristics of mesoporous silica fibers, 87 3.3.2. Gas uptake on silica fibers, 89 3.3.3. Evaluation of transport diffusivity, 94 3.3.4. Fiber internal microstructure, 105 3.4. Conclusions, 109 3.5. Reference, 111

CHAPTER 4. COUNTER DIFFUSION SELF ASSEMBLY OF ORDERED NANOPORUS SILICA MEMBRANES 115 4.1. Introduction, 115 4.2. Principle of counter diffusion self assembly, 116 4.3. Experimental, 117

ix 4.3.1. Ceramic supports, 117 Preparation of α-alumina supports, 119 Preparation of γ-alumina membranes, 120 Hydrophobic modification of supports, 121 4.3.2. CDSA growth of silica membranes, 123 4.3.3. Characterization, 124 4.4. Results and Discussion, 127 4.4.1. Characterization of supports, 127 4.4.2. CDSA growth on un-modified supports, 132 4.4.3. Characterization of hydrophobic modified supports, 138 4.4.4. CDSA growth on hydrophobic-modified supports, 142 4.4.5. Permeation and microstructural properties of CDSA- grown silica membranes, 145 4.4.6. Effect of hydrophobic chain length, 152 4.4.7. Expected mechanism of formation, 154 4.5. Conclusions, 157 4.6. Reference, 158

CHAPTER 5. CYCLIC CVD MODIFICATION OF STRAIGHT PORE ALUMINA MEMBRANES 161 5.1. Introduction, 161 5.2. Experimental, 164 5.2.1. Anopore membrane modification, 164 5.2.2. Characterization, 167 5.3. Results and Discussion, 170 5.3.1. Single gas permeation properties of CVD-modified Anopore membranes, 170 5.3.2. Multi-gas permeance and separation properties of CVD-modified Anopore membranes, 176 5.3.3. Pore structure and mechanism of CVD modification, 182 5.4. Conclusions, 186

x 5.5. Reference, 187

CHAPTER 6. SUMMARY 189

CHAPTER 7. AREAS OF FUTRE RESEARCH 194

APPENDIX A. EXPERIMENTAL PROCEDURES 198

APPENDIX B. SAMPLE OF CALCULATIONS 218

xi LIST OF TABLES

Table 1-1. Major Mesoporous Silicates ...... 6

Table 2-1. Summary of silica samples and preparation conditions ...... 39 Table 2-2. Fiber diameter, mass production and yield of mesoporous silica fibers at different HCl acid ...... 47 Table 2-3. Fiber diameter, mass production and yield of mesoporous silica fibers prepared with different growth times ...... 49

Table 2-4. Interplanar spacing and pore wall Thickness, and N2 adsorption- desorption results for silica fibers prepared at different growth times ...... 56

Table 2-5. Interplanar spacing and pore wall Thickness, and N2 adsorption- desorption results for silica fibers prepared at different growth times ...... 57 Table 2-6. Physical properties of mesoporous silica morphologies synthesized at low HCl acid content ...... 63 Table 2-7. Physical properties of mesoporous silica morphologies

synthesized with HNO3 and H2SO4 acids ...... 67

Table 2-8. Polymerization conditions, PE fiber MW data and DSC characterization results ...... 73

Table 3-1. Experimental parameters of diffusing gases in mesoporous silica fibers ...... 99 Table 3-2. Knudsen and surface diffusivities and contribution to net gas diffusion in MSF ...... 99 Table 3-3. Effective and real gas diffusion coefficients in silica fibers . . . . . 102

Table 3-4. Real Surface Diffusivities of CO2 in ordered mesoporous silica fibers versus other non-ordered porous solids ...... 103

xii Table 4-1. Regression parameters and average pore sizes of the supports obtained from steady state helium permeation data ...... 132 Table 4-2. Nitrogen permeance of CDSA silica membranes on unmodified ceramic supports ...... 136 Table 4-3. Nitrogen permeance of CDSA silica membranes on hydrophobic modified alumina supports ...... 143 Table 4-4. Comparison of nitrogen permeance of mesoporous silica membranes by CDSA approach and other methods ...... 151

Table 5-1. Helium permeation properties for the Anopore membranes after CVD modification ...... 174 Table 5-2. Percentage reduction in average microstructural values due to CVD modification ...... 175

xiii LIST OF FIGURES

Figure 1-1. Major members of M41S mesoporous silicate family ...... 5 Figure 1-2. Schematic model of formation of mesoporous structures ...... 8

Figure 2-1. Silica fiber growth after 4, 7 and 10 days ...... 43 Figure 2-2. SEM image of sample grown for 10 days showing the fiber morphology ...... 46 Figure 2-3. Enlarged view of cross-section of mesoporous silica fibers ...... 46 Figure 2-4. Accumulative fiber diameter distribution curves at different acid concentrations ...... 47 Figure 2-5. Accumulative fiber diameter distribution curves at different growth times ...... 49 Figure 2-6. Effect of aging time on mesoporous silica fiber mass production . . 51 Figure 2-7. TEM images for mesoporous silica fibers ...... 52 Figure 2-8. XRD diffraction pattern for mesoporous silica fibers ...... 54 Figure 2-9. Nitrogen sorption isotherm for silica fibers ...... 54 Figure 2-10. XRD diffraction patterns for samples under different growth times 56 Figure 2-11. XRD diffraction patterns for samples under different acid concentrations ...... 57 Figure 2-12. Morphologies of silica synthesized under low HCl acid contents . . 61 Figure 2-13. XRD patterns of non-fibrous silica under low HCl acid contents . . 62 Figure 2-14. Nitrogen sorption isotherms for non-fibrous silica morphologies under low HCl acid content ...... 62 Figure 2-15. Morphologies of silica synthesized under different acids ...... 66 Figure 2-16. XRD patterns of non-fibrous silica morphologies prepared with

HNO3 and H2SO4 acids ...... 67 Figure 2-17. Schematic describtion of the mecahnism of fiber formation ...... 70 Figure 2-18. SEM micrograph of MCM-41 particles used in extrusion polymerization ...... 72 Figure 2-19. XRD spectra of the extrusion polymerized polyethylene fibers . . . . 74

xiv Figure 2-20. SEM micrographs of the polyethylene fibers ...... 75 Figure 2-21. Illustration of the proposed scheme of polyethylene fiber formation 76

Figure 3-1. Schematic of Cahn microbalance system ...... 86 Figure 3-2. Normalized fiber length distribution ...... 88

Figure 3-3. Absolute and fractional uptakes of CO2 and C2H4 on silica fibers . . 90 Figure 3-4. Comparison of data fitting between the fiber length-corrected diffusion model and the uncorrected model ...... 93

Figure 3-5. Equilibrium adsorption isotherms of CO2 and C2H4 on silica fibers 98 Figure 3-6. Schematic representation of the internal pore structure of silica fibers ...... 108

Figure 4-1. Schematic illustration of the interfacial growth of ordered mesoporous silica fibers and the concept of counter diffusion self assembly (CDSA) approach to grow silica membranes ...... 118 Figure 4-2. Possible morphologies of CDSA growth ...... 118 Figure 4-3. Schematic illustration of surface modification of alumina supports with hydrophobic chains ...... 121 Figure 4-4. Experimental setup for CDSA growth of silica membranes ...... 123 Figure 4-5. Steady state single gas permeation setup ...... 126 Figure 4-6. Unsteady state single gas permeation setup ...... 126 Figure 4-7. Steady state helium gas permeance of supports used for CDSA approach ...... 128 Figure 4-8. SEM and XRD of the water-phase sided surface of 4 µm pore- defected alumina support after silica deposition by CDSA ...... 134 Figure 4-9. FTIR spectra of hydrophobic modified alumina supports ...... 139 Figure 4-10. TGA results of hydrophobic modified alumina supports ...... 141 Figure 4-11. SEM and XRD of C18-modified alumina support after silica deposition by CDSA ...... 144 Figure 4-12. Single gas permeance data of CDSA silica membranes ...... 146

xv Figure 4-13. Nitrogen permeance contributions of the composite CDSA silica membrane, deposited silica and support ...... 148 Figure 4-14. Temperature and molecular weight dependencies of single gas permeance data of CDSA silica membrane ...... 149 Figure 4-15. Schematic illustration of the side gap created between silica plugs and support pore wall ...... 153

Figure 5-1. SEM images of straight pore alumina membrane (Anopore) ...... 165 Figure 5-2. Schematic illustration of the atomic layer CVD ...... 166 Figure 5-3. Water vapor separation system ...... 169 Figure 5-4. gas helium permeance of unmodified and 6 time CVD modified Anopore membranes under two CVD schemes ...... 171 Figure 5-5. Permeance of oxygen and water vapor through unmodified Anopore and γ-alumina membranes ...... 177 Figure 5-6. Oxygen and water permeance for Anopore membranes modified under the residual pressure scheme ...... 179 Figure 5-7. Oxygen and water permeance for Anopore membranes modified under the purge scheme ...... 181 Figure 5-8. Water/oxygen separation factor for unmodified and CVD modified Anopore membranes under the purge and residual gas schemes . . . 181 Figure 5-9. Schematic representation of the pore microstructures obtained under CVD modification ...... 182 Figure 5-10. ESEM images of Anopore modified by two CVD cycles under residual pressure and purge schemes ...... 185

xvi The following chapters are modified versions of the papers:

Chapter 2 H. Alsyouri and Y.S. Lin, “Effects of synthesis Conditions on Macroscopic and microscopic properties of ordered mesoporous silica fibers”, Chem. Mater., 15 (2003) 2033.

Z. Ye, S. Zhu, W.J. Wang, H. Alsyouri, and Y.S. Lin, “Morphological and mechanical properties of nascent polyethylene fibers produced via ethylene extrusion polymerization with a metallocene catalyst supported on MCM-41 particles”, J. Polymer Science Part B: Polymer Physics, 41 (2003) 2433.

Z. Ye, H. Alsyouri, S. Zhu, and Y.S. Lin, “Catalyst impregnation and ethylene polymerization with mesoporous particle supported Ni-diimine catalysts”, Polymer, 44 (2003) 969.

Chapter 3 H. Alsyouri, Y.S. Lin, “Diffusion and microstructural properties of ordered mesoporous silica fibers, to be submitted to J. Phys. Chem. B., 2004.

Chapter 4 H. Alsyouri, Y.S. Lin, “Counter Diffusion Self Assembly of Nanostructured Silica Membranes”, to be submitted to J. Membr. Sci., 2004.

Chapter 5 H. Alsyouri, C. Langheinrich, Y.S. Lin, S. Zhu, and Z. Ye “Cyclic CVD modification of straight pore alumina membranes”, Langmuir, 19 (2003) 7307.

xvii CHAPTER 1

INTRODUCTION

1.1. Porous Materials

Porous materials are solids that contain empty pores dispersed within their

framework. The fraction of voids, or porosity, can vary between 0.2-0.95. The pores can

be open pores connecting to the outside of the material or closed pores isolated from the

outside. Porosity provides materials with lower density and higher surface area properties

compared to dense materials. Porous metals, ceramics and glasses are particularly important for industrial applications in chemistry, mechanical engineering, biotechnology

and electronics. For most industrial applications of porous materials, open pores play a

crucial rule especially the penetrating pores which are permeable to fluid and connect

both sides of the material. They are desired in various applications including separation,

catalysis and bioreactors. Closed porous materials are used mainly for thermal insulation

and low density structural components [Ishizaki et al. 98].

Porous materials are classified by different criteria such as pore size, pore shape,

materials and production method. On the basis of average pore diameter, porous materials

are divided into microporous (< 2nm), mesoporous (2-50 nm) and macroporous (> 50

nm). Well-known members of the microporous class are the crystalline aluminosilicate

zeolite molecular sieves [Flanigen 98] that contain well-defined geometries of one or two dimensional pores with diameters of 4 to 7 Å dispersed within the crystal lattice network.

The high surface area and uniform pore size distribution properties of zeolites made them ideal for many applications including catalysis, reaction processes and size- and shape-

1 selective separation [Flanigen 98, Cuny and Cox 03, Suib 93]. However, the narrow pore

size range of zeolite materials limits their use in applications involving large molecules

such as polymers and proteins.

Silica gels, alumina and various types of etched membranes are examples of

mesoporous materials. The walls of these materials are amorphous showing no long-

range order and the pores exist as a disordered network within the solid. These materials

have larger pores than zeolites, providing better supports to host large species. However,

the wide range of pore size distribution and irregularity of pore shapes greatly reduce

their applicability in processes where selectivity is important. This has lead to a growing

interest in developing porous materials that combine the large pore size property of

mesopores silica gels and the uniform pore structure properties of zeolites. Such materials would be effective in applications for selective processing of large molecules.

Considerable efforts have been focused on expanding the pore size of the microporous zeolite materials to mesoporous range. Success has been achieved in creating a new generation of larger pore zeotype molecular sieves based on aluminum and gallium phosphate structures. The major products are AlPO4-8 that contains 14

member rings [Dessau et al. 90], VPI-5 [Daviset al. 88], Colverite [Estermann et al. 94]

and JDF-20 [Jones et al. 93]. These materials exhibit either poor stability or weak acidity

which limits their use in applications. Besides, they have pore diameters in the range 0.8-

1.3 nm which is believed to be the upper limit of pore size that can be synthesized

without large organic molecules [Davis and Lobo 92].

2 1.2. Ordered Mesoporous Materials

1.2.1. Overview

The major challenge was to develop new methods or modify the current ones for fabrication of mesoporous solids with ordered pore channels of pore diameters over 2 nm.

It was in 1992 when the researchers of Mobil Corporation have first reported the direct synthesis of a new family of mesoporous silicate molecular sieves (M41S) based on liquid crystal templating (LCT) method [Kresge et al. 92, Beck et al. 92]. The well- known Mobil Composition of Matter (MCM) materials belong to this novel family. The

LCT approach has directed the synthesis of aluminosilicates and metal oxide materials with well-defined pore sizes up to 33 nm break past the pore size constraint (< 1.5 nm) of microporous zeolites. The extremely high surface area (>1000 m2/g) and the precise tuning of the pore sizes are among the desirable properties of these materials.

The LCT synthesis approach of the M41S materials employs self assembled arrays of surfactant molecules as structure directing templates compared to single surfactant molecules as in case of zeolites. These surfactants create ordered liquid crystal phases (or micelles) in the solvent used (usually water) around which inorganic species polymerize to form continuous solid network encapsulating the micelles. Ordering in these systems present only on the mesoscale arising from the liquid crystalline arrangements of the surfactant micelles. The inorganic phase framework is locally amorphous and the porosity is created by removal of the organic surfactant micelles through liquid extraction, calcination or ozonation.

3 Several ordered pore structures corresponding to various surfactant liquid crystal

phases has been synthesized. The most common structure is the two-dimensional

hexagonal with the P6mm symmetry, consisting of close-packed hexagonal arrays of

cylindrical surfactant micelles. Lamellar silicate-surfactant phases were also obtained.

Normally, the lamellar phases are not stable to removal of the surfactant and lead to

collapse of the silica layers unless the layers are folded. Various cubic phases have also

been reported. The bicontinuous cubic gyroid phase with the Ia3d symmetry is found in alkaline-catalyzed syntheses. This phase with its network of interconnected pores is attractive for diffusion applications. Another cubic phase is the Pm3n composed of spherical micelles in a cubic close-packed arrangement.

The major products reported in the pioneering work of the Mobil researchers are referred to as MCM-41 (P6mm 2D hexagonal phase), MCM-48 (Ia3d cubic phase) and

MCM-50 (lamellar phase) as illustrated in Figure 1-1. The unique pore structure and high

surface area of these mesophases identify them as exceptional porous solids. MCM-41

has attracted considerable attention to elucidate its potential industrial applicability as

well as synthetic conditions (precursor type, molar ratios, temperature, pH, counter ions,

etc) and characterization methods. Several groups have reported alternative synthesis

methods based on the LCT mechanism for the preparation of ordered mesoporous

materials with similar MCM symmetries and distinct wall properties desirable in

applications [Beck et al. 92, Inagaki et al. 93, Huo et al. 94, Bagshaw et al. 95, Huo et al.

95, Fukushima et al. 96, Mokaya and Jones 96, Ryoo et al. 96, Tenav et al. 97, Zhang et

al. 97, Zhao et al. 98a-b, Zhao et al. 99, Lukens et al. 99, Inagaki et al. 99, Huang et al.

00,].

4

MCM-41 MCM-48 MCM-50 (Hexagonal P6mm) (Cubic Ia3d) (Lamellar)

Figure 1-1. Major members of M41S mesoporous silicate family and their corresponding microstructural symmetries

Several surfactant systems have been used to vary the pore size and properties of the ordered mesoporous products. Pore dimension of the surfactant-inorganic composite is related to the chain length of the hydrophobic tail of the surfactant. Surfactants used in the LCT preparation of ordered mesoporous materials can be anionic (Sodium dodecylsulphate) [Huo et al. 94, Huang et al. 00], cationic (alkyltrimethyl ammonium halides) [Beck et al. 92] or neutral (e.g., amines [Mokaya and Jones 96], polyethylene oxides (EOx) [Bagshaw et. Al. 95, Tanev and Pinnavaia 95], alkyl (polyethylene oxide)

[Zhao et al. 98a]). Water soluble triblock co-polymers have been also used to create materials with much larger pore sizes (5-30 nm depending on the chain length of the polymer) compared to the conventional micelles of smaller surfactant molecules (< 10

nm) [Zhao et al. 98b].

The M41S molecular sieves were originally prepared under alkaline-catalyzed conditions. KIT-1 is a disordered mesoporous product with wormlike channel network prepared at alkaline conditions [Mokaya and Jones. 96]. Preparations have been subsequently extended to acidic conditions and have led to ordered mesoporous materials with various symmetries and distinct wall properties. They include the SBA-n (n= 1-3, 8,

5 11, 12, 14, 16) molecular sieve series with variable hexagonal and cubic symmetries

[Huo et al. 94, Huo et al. 95, Zhao et al .98b, Zhao et al. 98a, Lukens et al. 99, Zhao et al.

99]. Neutral conditions were employed to prepare ordered mesoporous material using non-ionic primary amines and alkyl-EOx-templated synthesis, e.g., disordered hexagonal

MSU [Bagshaw et al. 95] and HMS [Tanev et al. 97, Zhang et. al. 97]. Major mesoporous silicates and corresponding symmetries are summarized in Table 1-1.

Table 1-1. Major Mesoporous Silicates Sample Dimensionality, order Surfactant Medium mean Reference code and space group type pore size (nm) MCM-41 2D hexagonal (P6mm) cationic basic 3.70 Beck, 92 MCM-48 cubic (Ia3d) basic 3.49 FSM-16 2D hexagonal (P6mm) cationic basic 2.80 Inagaki, 93 SBA-1 cubic (Pm3n) cat/anionic acidic 2.00 Huo, 94

SBA-2 3D hexagonal (P63/mmc) gemini acidic 2.22 Huo, 95 SBA-3 2D hexagonal (P6mm) cat/anionic acidic 2.77 Huo, 94 SBA-8 2D Rectangular (cmm) bolaform acidic 1.87 Zhao, 99 SBA-11 cubic (Pm3m) copolymer acidic 2.50 Zhao, 98b

SBA-12 3D hexagonal (P63/mmc) acidic 3.10 SBA-14 cubic (Pm3n) acidic 2.40 SBA-15 2D hexagonal (P6mm) copolymer acidic 7.80 Lukens, 99 SBA-16 cubic (Im3m) copolymer acidic 5.40 Zhao, 98b

HMM 3D hexagonal (P63/mmc) cationic basic 2.70 Inagaki, 99 MSU hexagonal (disordered) neutral neutral 3.10-5.80 Bagshaw, 95 MSU-G lamellar cationic neutral 3.20 Ryoo, 96 HMS hexagonal (disordered) neutral neutral 2.80 Zhang, 97 KIT-1 hexagonal (disordered) neutral basic 3.52 Mokaya, 96

6 1.2.2. Mechanism of formation

Several mechanisms have been proposed to describe the formation of highly ordered arrays of pore openings from the self assembly between the inorganic species and the organic alkyl-halide surfactants including liquid crystal templating mechanism (LCT), silica-driven self assembly and layered silica-driven ordering. These mechanisms are summarized in Figure 1-2. The LCT mechanism was originally proposed by Mobil researchers based on the similarity between the ordered mesopores and the mesophases of the liquid crystalline surfactants in aqueous solutions. They suggest two pathways to form the ordered pores either by condensation of silica species with pre-existing ordered liquid crystal (LC) mesophases (path a) or from ordering of the LC micellar rods into ordered mesophases enhanced by addition of silica species followed by silica condensation.

The first pathway is not likely the right mechanism because MCM-41 materials were readily prepared at surfactant concentration below the critical micelle concentration

(CMC) required to from the hexagonal mesophase [Vartuli et al. 94]. The second pathway suggested by Mobil researchers has been further confirmed by other studies based on NMR [Chen et al. 93], neutron [Auvray et al. 95] and X-ray scattering [Rregev

96] and electron microscopy [Cheng et al. 95]. These studies propose that once added to the solution, silica species start to encapsulate the rod micelles and pack into disordered arrays and eventually into hexagonal mesostructured (path c) implying the structure- directing role of inorganic species in these systems.

7

Figure 1-2. Schematic model for formation of mesoporous structures from liquid crystal templating mechanism: (a) via pre-existing surfactant liquid crystal phases, (b) via formation of silica-coated surfactant species that either (c) pack into ordered and disordered arrays or (d) form a layered phases that undergo phase transition into hexagonal phase. Removal of surfactant (e.g., by calcination) ends up with the hexagonal array of open pores. (Obtained from Edler and Roser 01)

8 Several groups have observed the formation of lamellar silica-surfactant phase in

early preparation stages which subsequently formed the hexagonal phase materials. As a

refinement of the silica-driven self assembly mechanism, Stucky and coworkers

[Monnier et al. 93, Stucky et al. 94] proposed a charge density matching model to

describe the derivation of a hexagonal structure from a lamellar phase. In this scheme, a

layered system is initially formed due to attraction between the silica oligomers and the

surfactant head groups. Condensation of silica reduces its charge density at a relatively higher charge density of surfactant. This induces re-construction and curvature of the relatively flexible silica layers to balance the charge density. This causes a phase change from lamellar to hexagonal structure (path d). To date, the mechanism of ordered phase

formation by the self assembly is still a matter of debate. However, all the suggested

mechanistic models proposed so far share the basic idea of LCT mechanism that the

silicate species promoted the liquid crystal phase formation below the CMC.

1.2.3. Synthesis pathways

Six different synthesis pathways have been readily employed to prepare ordered

mesoporous materials under wide range of pHs, temperatures and surfactants [Huo et al.

94, Tanev et al. 95, Behrens 96]. These pathways reflect the molecular interaction

between inorganic material and the surfactant polar head groups. The pathways are S+ I−,

S−I+ , S+ X− I+, S−X+ I−, S°I° and S−I, where S is the surfactant, I is the inorganic phase

and X is the mediating counter-ion. The first four pathways take place by electrostatic covalent interaction between the ionic surfactant and inorganic molecules of opposite or

similar charge (in presence of mediated ions like halides). With neutral or non-ionic

9 surfactant, interaction proceeds via hydrogen bonding based on the S°I° pathway. S−I

pathway applies to transition-metal oxides systems where inorganic precursor is

covalently bonded to the surfactant [Antonelli et al. 96].

For pathways involving electrostatic interaction, charge of the inorganic source is

controlled by the pH and isoelectric point, i.e., the pH at which the charge of molecules is

zero. Silica species (isoelectric point =2) has a positive charge at pH < 2 (acidic

conditions), neutral charge at pH=2 and a negative charge at pH > 2 (alkaline conditions).

M41S silica molecular sieves were originally fabricated at alkaline conditions (I−) with cationic surfactant (S+) through the S+ I− strong interaction. Likewise, mesostructures can

be formed with anionic surfactant (S−) at acidic or basic conditions with pH < 2 (I+) using the S− I+ approach. The synthetic pathways involving a mediating ion (e.g., S+ X− I+) have a weaker electrostatic interaction between the cationic silica precursor and the S+ X− active sites which leads to products with rich morphologies [Yang et al. 97].

1.2.4. Acidic Synthesis

The acidic synthesis of ordered mesoporous silica at pH < 2 represents an important advance in material synthesis with versatile morphologies [Zhao et al. 98c]. In the acidic media, with pH< 2, the positive silica species (e.g., silicon alkoxides) combine with the cationic surfactant through a mediating or bridging counterion using the pathway (S+ X−

I+). The counterion is usually provided by acids or mineralization additives (such as

salts). For acids (HX), the protons (H+) act to protonate silica species in the aqueous

phase, and the counterions (X−) serve as the bridging ions. The main reactions involved in the acidic synthesis are summarized in steps 1-4 [Lin et al. 01]. In the first reaction, the

10 micellar surfactants S+ adsorb the counter X− by electrostatic interaction to form S+X−

micelles. The silica precursor, e.g., silicon alkoxide Si(OR)4, undergo a fast hydrolysis

and protonation reactions under the acidic conditions to give positively charged siloxane

+ + + − species ≡SiOH2 (denoted as I ) as shown in steps 2 and 3. The S X micelles act as

catalytic sites and promote the condensation of the positively charged siloxanes near the

micelle surface (step 4). The condensation step proceeds increasing the molecular weight

of the polysilicate product. The long polysilicate would then act as a bridge to bring the

cylindrical micelles together to form the mesophase separation. Silica condensation will

eventually cause neutralization of the surface charges and finally produce the gel of the

S+ X− Iº combination.

In the alkaline route the surfactant and silicates organize by the strong S+I− electrostatic interaction. On the other hand, in the acidic route the silicates/surfactant interaction in S+ X−I+ is weaker and can lead to many topological constructions. In acidic

synthesis, the silica is usually less condensed and the structure order is softer (allowing

more micellar defects) and often leads to rich morphologies.

11 1.3. Morphological Control of Mesoporous Materials

The discovery of the MCM-41 silicate with structures analogous to the organic

surfactant lyotropic symmetries has initiated tremendous interest to understand the origin

and shape over all length scales [Mann and Ozin 96, Zhao et al. 98]. The self

organization properties and the chemistry of inorganic liquid crystals appear to be the key

to understanding the morphological control of inorganic shapes. The classical synthesis

of mesoporous materials by Mobil was carried out at basic conditions with anionic silica

species and led to loose agglomerates of hexagonal mesoporous silica particles smaller

than 10 µm. Synthesis at acidic conditions [Huo et al. 94] was a major advance that

extended the synthesis to cationic surfactants at pH<2 and has led to ordered mesoporous materials with remarkable morphological features. Copolymer templating [Yang et al.

98a], mesoporous silica films [Yang et al. 96, Ryoo et al. 97], gyroids [Yang et al. 97], fibers [Schacht et al. 96, Huo et al. 97] and ropes [Schmidt et al. 99, Lin et al. 99] are examples of achievements based on the acidic synthesis.

The formation of rich morphological patterns was explained in terms of topological defects in liquid crystals and colloidal interactions [Ozin 99]. Individual or combinational rotations of the point, linear, or planar defects in the liquid crystal along their transverse or longitudinal axes determine the shape of the product. For example, curved morphologies may originate from longitudinal disclinations with a rotation vector along the longitudinal axes. Similarly, transverse dislocations along rotation axes produce closed shapes such as discoids or toroids [Yang et al. 98b].

Precise control over the experimental conditions of materials synthesis has demonstrated to offer the possibility of simultaneously controlling the shape on the µm to

12 cm length and the periodic mesostructure at the molecular mesoscale. Variation of

counterion, salt additives, pH, contact mode of precursors and hydrodynamics of the

medium are among many different conditions that have been employed to vary the final

product morphology. For example, self assembly under shear fluid flow generated rope

and fiber shapes from flat films [Lin and Mou 96, Lin et al. 99]. Controlled stirring speed,

on the other hand, induced a systematic tuning from curved shapes to hollow porous spheres [Walshet al. 95]. Emulsion biphase chemistry has been employed in the self assembly of mesoporous materials [Schacht et al. 96, Huo et al. 97a, Huo et al. 97b]. In this approach the silica precursor is dissolved in oil phase (e.g., organic solvent) and dispersed as emulsion in an aqueous solution containing the surfactant. Slow diffusion of silica precursor through the water-oil interface can create mesoporous hard and hollow spheres (with slow stirring), intertwined rope morphologies (with no stirring), thin films and mesoporous silica fibers at the oil-water interface (with no stirring).

Due to their importance in technological applications, tremendous efforts were focused on synthesis of the mesoporous materials as membranes. These membranes have pores arranged in an ordered fashion compared to the tortuous and interconnected pore structures of conventional mesoporous ceramic membranes. Several strategies have been used to grow oriented mesoporous silica as free standing or supported thin films on several dense and porous substrates on which a multitude of journal papers were published. Most of techniques used to prepare ordered mesoporous films (OMF) are based on the sol-gel science utilizing a colloidal sol containing the surfactant and silica precursors. Free standing, submicron thick OMF’s were prepared at the air-water [Yang et al. 96] and oil water interface [Schachtet al. 96] from a homogeneous solution of the

13 reactants after hydrolysis in the water phase. Supported mesoporous films, on the other

hand, were prepared by the dip-or spin- coating [Ogawa 94, Martin et al. 97], or solution

growth directly on the support. The later technique involves immersion [Nishiyama et al.

97] or floatation [Miyata and Kuroda 99] of substrate in the solution, or by flowing the

solution over the support surface [Hillhouse et al.99]. These techniques generated OMF’s

with various pore symmetries, pore orientation and macroscopic properties such as films

thickness continuity and surface roughness. The extensive work on synthesis of ordered

mesoporous films (OMF) and membranes is summarized in several review papers

[Brinker 98, Pevzner et al. 00, Elder et al. 01, Guliants et al. 04].

1.4. Applications

The outstanding tunable microscopic and macroscopic properties of ordered

mesoporous materials have generated commercial interest for their ability to overcome

common application problems associated with microporous zeolites and conventional

mesoporous materials. The high surface area, uniform and tunable pore sizes and

accessibility of pores are among many desirable features that made these materials the

focus of great attention. The surfactant-templated chemistry has led to novel catalysts,

sorbents, sensors and host materials for large guest molecules [Brinker 96, Davis 02].

Advances in the synthesis of ordered mesoporous membranes and incorporation of new

metals [Ying et al. 99] and functional groups [Stein et al. 00] into their frameworks have

motivated pronounced progress in separation, heterogeneous catalysis [Yang et al. 97b],

drug delivery [Ramila 03], sensors [Yang et al. 97b], microelectronics devices [Braun 99]

and modulated nano-reactors [Kageyama et al. 99].

14 1.5. Chemical Vapor Deposition

Chemical vapor deposition (CVD) is one of the commonly used techniques in the

inorganic membrane field. CVD is an activated chemical process in which one or more gaseous reactants are allowed to diffuse into the pores of a substrate and react to form a solid deposition on the surface or inside the pores of the substrate. This process has been applied for fabrication of advanced materials such as ceramic matrix composites, super/semiconductors and diamond coatings [Lin 92]. Many researchers employed CVD process for fabrication of composite inorganic membranes consisting of thin, dense permselective film supported on top or inside a thick porous inorganic support

[Xomeritakis 97]. Atomic layer chemical vapor deposition (ALCVD) is a useful method for the modification of porous inorganic membranes developed by George and Co- workers [George 96]. Compared to the conventional CVD, ALCVD can precisely tailor the membrane pores by the controlled growth of the oxide deposition layer inside the pores. The reagents are introduced separately in a step-wise manner and the thickness of deposit layer can be precisely controlled by the number of cyclic additions of the reagents causing one atomic layer to be formed at a time.

15 1.6. Research Objectives

1.6.1. Motivation of this research

The ability to selectively synthesize ordered mesoporous material with a particular shape over a range of length scales represents a major challenge in the field of materials science. It represents a big step towards the design of property-engineered materials and devices for a range of perceived applications in unprecedented ways. The surfactant supramolecular templating approach, created by Mobil researchers, was a major advance in the materials field that led to ordered mesoporous materials with hierarchical constructions over all microscopic and macroscopic length scales. This approach was employed to obtain mesoporous silica with rich morphologies including fibers, solid and hollow spheres, particulate constructs of rich shapes, and free and supported thin films.

Mesoporous inorganic membranes with pore sizes in the range of 2-5 nm have been widely studied in the past two decades [Burggraf and Cot 96]. They have been prepared from wide range of materials (alumina, zirconia and titania) and have found many applications in liquid filtration processes and as supports for microporous and dense membranes [Lin 01b, Lin et al. 02]. These membranes have disordered and tortuous microstructures derived from compacting their ceramic particles. In contrast, there are essentially no mesoporous inorganic membranes with ordered, non-connecting, and straight mesopores. The Anodic alumina membranes commercially available have a 1

µm thick top-layer containing straight, non-connecting pores of mono-dispersed pore diameters between 20 to 250 nm. The relatively larger pore diameter of these membranes is far beyond the desired 2-5 nm mesoporous size and this has limited their use in molecular reaction and gas separation applications.

16 The surfactant templating approach has provided new avenues for fabrication of

inorganic mesoporous membranes with ordered microstructures. This approach was

employed to preparation of hexagonal mesostructured membranes using variety of sol-

gel-based techniques (see section 1.3 for more details). In all cases the hexagonal structure was achieved but the pores were randomly oriented or aligned parallel to the

support surface. Some trials have applied external field such as continuous shear flow

[Hillhouse et al. 99] or magnetic field [Tolbert et al. 97] to orient the pores into desired directions. These fields caused local alignment of the mesopores but not over the whole

support. Moreover, such experiments require special apparatus and restricted support

shapes, and above of all cannot align the pores vertical to the support surface. Ordered

nanoporous membranes with 2-5 nm diameter pores oriented normal to support surface

allow easy transfer through the membrane and will offer unprecedented opportunities in

separation and modulated nano-reactor applications, e.g., to template growth inorganic

and polymeric materials with controlled microstructures and morphologies.

Mesoporous silica fiber (MSF) is one of the remarkable morphologies obtained by

the surfactant-templating approach. It has unique microstructural features compared to

conventional types of carbon, glass and polymer fibers. MSF can be created in a simple

experimental procedure by hydrolysis and condensation of silicon alkoxide in presence of

organic surfactant. This morphology has straight and non-connected channels of 2-5 nm

diameters running parallel to the fiber axis. The fibers are ~ 15 µm in diameter and can

grow as large as 5 cm [Huo et al. 97a]. Besides, they are at least 2-5 orders of magnitude

longer than the MCM-41-type particulate shapes which can generate interesting results in applications requiring long pores such as molecular reaction applications.

17 MSF has been obtained using the emulsion chemistry coupled with the surfactant- templating at an aqueous-oil interface under quiescent acidic conditions [Huo et al. 97a].

The counter diffusion of precursors through interface coupled with self assembly yields mesoporous silica with remarkable morphologies attached to the interface towards the

water. Fiber morphology is achieved when the diffusion rate of silica through the

interface is slow. This work has stimulated interests in the synthesis, microstructure and

applications of the fiber shape [Yang et al. 98c, Marlow et al. 00, Kleitz et al 01ab,

Marlow and Kleitz 01, Chu et al. 02]. The straight pore alignment across the fibers was also reported on fiber morphologies obtained using homogeneously mixed systems [Yang et al. 99, Yamaguchi et al. 04]. However, some studies showed that the pore alignment in fibers obtained by the oil-water interface approach is not exactly straight but rather whirl helically around the fiber axis [Marlow et al. 00].

Most of these studies were focused on exploring the product shape and inner microstructure (pore size, surface area and pore alignment) of the silica fiber mainly under at different surfactant, silica sources and oil (organic solvent) precursors. Special attention was given to study the pore orientation (straight vs. helical) and identify its mechanism of formation [Marlow et al. 00, Marlow et al. 01a-b]. Some efforts aimed at improving the fiber properties to increase their potential use in applications. For example, fiber spinning method was utilized to obtain long mesoporous silica fibers with macroscopic length scales (few centimeters) for consideration in VOC removal [Chu et al. 02]. Others were able to incorporate dyes in the structure of MSF generating new interesting types of laser materials [Marlow et al. 00].

18 None of the studies on MSF have tried to investigate the effect of synthesis conditions on the macroscopic properties of the fibers (e.g., fiber diameter and length, and pore length within the pore) and the possibility of extending their novel microstructure into more useful morphologies such as membranes. Equilibrium sorption of gases and vapors on ordered nanoporous silicates has been extensively studied for material characterization or exploring the use of these materials as sorbents [Selvam et al.

01]. The properties of gas diffusion in this group of materials are important for their applications in a number of areas such as catalysis and separation. However, essentially no studies were reported on gas diffusion in the ordered nanoporous materials. This was probably due to the anticipation of fast gas diffusion rate in nanoporous particles which makes it difficult to experimentally measure gas diffusion in these materials unless the diffusion study involves large gas molecules or much larger ordered mesoporous particles

(e.g., fibers) are available.

Ordered nanoporous particles such as MCM-41 with 2-5 nm pore diameter offer many potential applications in the areas of biological and polymer reaction engineering.

It has have been applied as a support for metallocene and other catalysts for olefin polymerization [Ko et al. 96, Van Looveren et al. 98, Kaminsky et al. 00, Ko and Woo

01]. Different from the conventional catalyst supports, the straight nanopores of these particles appears to serve as a template to control polymer microstructure and crystal morphology in olefin polymerization. Aida and coworkers [Kageyama et al. 99] synthesized polyethylene fibers on metalocene catalyst supported on ordered mesoporous silica particles. They found that the polyethylene chains could grow out of the nanopores unfolded and assembled into long-ranged ordered polyethylene (PE) nanofibers featuring

19 the desirable extended chain crystalline structure. A useful model is still needed to

explain formation of the extended-chain-crystalline PE nanofibers of Aida’s work.

Moreover, mechanical properties of the PE nano fibers need to be measured for

exploitation in potential commercial uses. These subjects will be part of this exploratory

research.

1.6.2. Synthesis and characterization of ordered mesoporous silica fibers

The short review presented above suggests that there still a need to develop new techniques for producing ordered mesoporous materials with controlled morphology and pore orientation. In principle, understanding the formation process of mesoporous silica fibers and their synthesis-shape-microstructure relationships can be used to develop such macroscopic morphologies. Previous studies on silica fibers have mostly focused on the microscopic features of the fibers such as surface area properties and pore orientation under effect of variable surfactant and silica source precursors. There is a significant lack of work on the macroscopic features (fiber diameters and length, pore length, etc) and gas

diffusional properties of mesoporous silica fibers. The macroscopic features will improve

our understanding of the fiber formation process and provide opportunities for fabrication

of more useful shapes. Gas diffusional properties can be used for analysis of the inner

microstructure and also for exploitation of these materials for potential applications.

In this work, mesoporous silica materials will be prepared using the quiescent water-oil biphasic approach under the same silica/surfactant precursors and conditions

reported in literature that will lead to fiber morphology. The growth conditions will be

broadened to involve more variables such as growth time, temperature, acid (or

counterion) type and acid concentration. The counter diffusion self assembly of

20 mesoporous silica products at the silica source-water interface will be evaluated by

studying the shape-microstructure-macroscopic properties of the products with emphasis

on the fiber morphology. These results will be utilized to propose a suitable mechanism of formation of these products. Gas diffusion properties of silica fibers will be measured using the transient weight uptake method. The results will be used to identify the gas diffusion coefficients in the long mesopores of the fibers, prevailing mechanisms of diffusion and the inner microstructure of this morphology, i.e., pore length and alignment.

1.6.3. Ordered mesoporous inorganic membranes with controlled microstructure

The past decade has seen significant advances in the fabrication of ordered mesoporous membranes. However, orientation of the mesopore normal to the support surface remains one of the major challenges in this area. In this work two approaches will be used to fabricate mesoporous membranes with controlled microstructures including pores normal to the membrane surface. The first approach is based on novel principle called the counter diffusion self assembly adopted from the quiescent interfacial formation of mesoporous silica fibers with pores aligned across the fiber axis. The second approach is based on CVD modification of straight pore alumina membranes (Anopore or

Anodisc) with 20 nm pore diameter.

In the counter diffusion self assembly (CDSA) approach, an inorganic support (e.g.,

α-alumina) will be placed at the interface between the silica source and the water phases.

Growth will be carried out under the optimized conditions obtained from the quiescent interfacial synthesis work (e.g., temperature, counterion, acid concentration, etc) that will lead to the fiber morphology. Precursors will be allowed to counter diffuse through the

21 support and self assemble into mesoporous silica membrane grown either as a distinct

film on the support surface or as fiber plugs within the support pores. If growth proceeds in the same fashion as the unsupported fibers, then the membrane will have pores normal to the support surface. The CDSA will be carried out under variable support pore size

(mesoporous to macroporous) and surface chemistry (hydrophilic and hydrophobic). Gas permeation and surface characterization of the CDSA silica membranes will be used to analyze quality of silica membrane, location of growth (film or plugs) and the microstructure of the silica membrane (pore size and possible orientation).

The second approach is based on narrowing the pore diameter of a 20 nm straight pore alumina membrane down to 2-5 nm using the cyclic CVD technique. The alumina membrane, commercially known as Anopore or Anodisc, has 60 µm thickness and is available with pore diameters between 20 to 250 nm. The CVD modification conditions will be controlled so that alumina deposition will take place either on the pore wall as atomic layers, or on the pore wall and the gas phase of the pore as fractal structure. The former microstructure will represent an alternative for obtaining ordered mesoporous membrane with pores vertical to the support which can be potential in membrane reaction applications. The later deposition scheme can give pore with fractal microstructure which can be potential for gas separation applications. Gas permeation characteristics will be used to analyze the microstructure of the CVD modification.

22

1.6.4. Polyethylene extrusion polymerization in ordered mesoporous silica materials

The review provided in section 1.6.1 shows that mesoporous silica materials were

used to template the growth of extended-chain crystalline polyethylene fibers by the

extrusion polymerization under the steric effect of the tubular mesopores. Studies on this

field have covered wide range of variables including catalyst type and polymerization

conditions and have analyzed the polymer product morphology and the possible models

of formation. However, there is essentially no suitable model that describes the formation

microscopic scale polymer fibers with extended-chain crystalline properties from the

meso-sized channels of the mesoporous support. Moreover, the mechanical properties of

these polymer fibers should be measured for application purposes.

In this work, polyethylene fiber production via the extrusion polymerization will be

demonstrated on MCM-41 mesoporous silica particles supported with metalocene

catalyst. Polymer structural and morphological characterizations will be used to provide a

suitable model for formation of the extended-chain crystalline product. Mechanical

properties will be measured and compared to those of polymer products obtained by

conventional techniques. The overall results will be crucial for future extension of this

concept into continuous process utilizing mesoporous membranes with pores vertical to

support surface.

23 1.7. Structure of Dissertation

This dissertation deals in general with synthesis of mesoporous silica and alumina

materials with controlled macroscopic morphologies and their potential use in polymer

reaction applications. Chapter 2 presents experimental findings on quiescent acidic

interfacial growth of mesoporous silica morphologies with emphasis on the fiber shape and its mechanism of formation, and demonstrates the use of the order structure of straight nanotubes of mesoporous silica particles as molecular extruders for production of high quality polymer fibers and their mechanical properties. In Chapter 3, the gas diffusional properties and the prevailing mechanisms of diffusion in mesoporous silica fibers are presented and used to evaluate the pore microstructure. Chapter 4 presents a novel counter diffusion self assembly (CDSA) technique for fabrication of supported

mesoporous silica membranes adopted from growth of mesoporous silica fibers. Finally,

Chapter 5 presents an alternative approach for preparation of mesoporous inorganic membranes with controlled microstructures based on cyclic chemical vapor deposition modification of straight pore alumina membranes.

24 1.8. References

Antonelli D.M., Nakahira A., Ying J.Y., Ligand-Assisted Liquid Crystal Templating in

Mesoporous Niobium Oxide Molecular Sieves, Inorg. Chem., 35 (1996) 3126.

Auvray L., Ayral A., Dabadie T., Cot L., Guizard C., Ramsay J., Sol-gel systems with

controlled structure formed in surfactant media, Faraday Discuss., 101 (1995) 235.

Bagshaw S.A., Prouzet E., Pinnavaia T.J., Templating Mesoporous Molecular Sieves by

Nonionic Polyethylene Oxide Surfactants, Science 269 (1995) 1242.

Beck J.S., Vartuli J.C., Roth W.J., Leonowizc M.E., Kresge C.T., Shmitt K.D., Chu C.T.,

Olson D.H., Shepard E.W., McCullen S.B., Higgins J.B., Schlenker J.L., A New

Family of Mesoporous Molecular Sieves Prepared with Liquid Crystal Templates, J.

Am. Chem. Soc., 114 (1992) 10834.

Behrens P., Voids In Variable Chemical Surroundings: Mesoporous Metal Oxides,

Angew. Chem. Int. Edn. Engl., 35 (1996) 515.

Braun A.E., Thermal Processing Options: Focus and Specialize, Semiconductor Int., 22

(1999) 56.

Brinker C., Porous Inorganic Materials, Curr. Opin. Colloid. Interface Sci., 1(1996) 798.

Brinker C., Oriented Inorganic Films, Curr. Opin. Colloid. Interface Sci., 3(1998) 166.

Burggraaf A.J., Cot L., Fundamentals of Inorganic Membrane Science and Technology;

Elsevier: Amsterdam, 1996, Cahpter 7.

Chen C.Y., Burkett S.L., Li H.X., Davis M.E., Studies on Mesoporous Materials. II.

Synthesis Mechanism of MCM-41, Microporous Mater., 2 (1993) 27.

25 Cheng C.F., He H.Y., Zhou W.Z., Klinowski J., Crystal Morphology Supports: The

Liquid-Crystal Formation Mechanism for the Mesoporous Molecular-Sieve MCM-

41, Chem. Phys. Lett., 244 (1995) 117.

Chu Y.H., Kim H.J., Song K.Y., Shul Y.G., Jung K.T., Lee K., Han M.H., Preparation of

Mesoporous Silica Fiber Matrix for VOC Removal, Catal. Today, 74 (2002) 249.

Cundy C., Cox P., The Hydrothermal Synthesis of Zeolites: History and Development

from the Earliest Days to the Present Time, Chem. Rev., 103 (2003) 663.

Davis M., Lobo R., Zeolite and Molecular Sieve Synthesis, Chem. Mater., 4 (1992) 756.

Davis M.E., Sladdarriage C., Montes C., Graces J., Crowder C., A Molecular Sieve with

Eighteen-Membered-Rings, Nature, 331 (1988) 698.

Davis M., Ordered Porous Materials for Emerging Applications, Nature, 417 (2002) 813.

Dessau R.M., Schlenker J.L., Higgens J.B., Framework Topology of AlPO4-8: The First

14-Ring Molecular Sieve, Zoelites, 10 (1990) 522.

Edler K.J., Roser S.J., Growth and Characterization of Mesoporous Silica Films, Int. Rev.

Phys. Chem., 20 (2001) 387.

Estermann M.L., McCusker B., Baerlocher C., Merrouche A., Kessler H., A Synthetic

Gallophosphate Molecular Sieve with a 20-Tetrahedral-Atom Pore Opening, Nature,

352 (1994) 323.

Flanigen E.M., Molecular-Sieve Zeolite Technology-The 1st 25 Years, Pure Applied

Chem., 52 (1980) 2191.

Fukushima Y., Inagaki S., Nano-Scale Structure Control of Mesoporous Silica, Mater.

Sci. Eng. A 217–218 (1996) 116.

26 George S.M., Ott A.W., Klaus J.W., Surface Chemistry for Atomic Layer Growth, J.

Phys. Chem., 100 (1996) 13121.

Gulinats V., Carreon M., Lin Y.S., Ordered Mesoporous and Macroporous Inorganic

Films and Membranes, J. Membr. Sci, 235 (2004) 53.

Hillhouse H.W., Egmond J.W., Tsapatsis M., Highly Ordered Mesostructured Thin

Films: Shear-induced Deposition of Optically Anisotropic Coatings of Tungsten

Oxide/Surfactant Composites, Langmuir, 15 (1999) 4544.

Huang, M.H., Dunn, B.S., and Zink, J.I., In Situ Luminescence Probing of the Chemical

and Structural Changes During Formation of Dip-Coated Lamellar Phase Sodium

Dodecyl Sulfate Sol-Gel Thin Films, J. Am. chem. Soc., 122 (2000) 3739.

Huo Q., Margolese D.I., Ciesla U., Feng P., Gier T.E., Sieger P., Leon R., Petroff P.M.,

Schuth F., Stucky G.D., Generalized Synthesis of Periodic Surfactant/Inorganic

Composite Materials, Nature, 368 (1994) 317.

Huo Q., Leon R., Petroff P. M., Stucky G.D., Mesostructure Design with Gemini

Surfactants: Supercage Formation in a Three-Dimensional Hexagonal Array, Science

268 (1995) 1324.

Huo Q.S., Zhao D.Y., Feng J.L., Weston K., Buratto S.K., Stucky G.D., Schacht S.,

Schuth F., Room Temperature Growth of Mesoporous Silica Fibers: A New High-

Surface-Area Optical Waveguide, Adv. Mater., 9 (1997a) 974.

Huo Q.S., Feng J.L., Schuth F., Stucky G.D., Preparation of Hard Mesoporous Silica

Spheres, , Chem. Mater., 9 (1997b) 14.

Inagaki S., Fukushima Y., Kuroda K., Synthesis of Highly Ordered Mesoporous

Materials from a Layered Polysilicate, J. Chem. Soc., Chem. Commun. (1993) 680.

27 Inagaki S., Guan S., Fukushima Y., Ohsuna T., Terasaki O., Novel Mesoporous Materials

with a Uniform Distribution of Organic Groups and Inorganic Oxides in their

Framework, J. Am. Chem. Soc., 121 (1999) 9611.

Ishizaki K., Komarneni S., Nanko M., Porous Materials: Process Technology and

Applications, Ch.1, Boston, Kluwer Academic Publishers, 1998.

Jones R.H., Thomas J.M., Chen J., Xu R., Huo Q., Li S., Ma Z., Chippindale A.M.,

2- Structure of an Unusual Aluminum Phosphate([Al5P6O24H] 2[N(C2H5)3H]+2H2O)

JDF-20 with Large Elliptical Apertures, J. Solid State Chem., 102 (1993) 204.

Kageyama K., Tamazawa J., Aida T., Extrusion polymerization: Catalyzed Synthesis of

Crystalline Linear Polyethylene Nanofibers within a Mesoporous Silica, Science, 285

(1999) 2113.

Kaminsky W., Strubel C., Lechert H., Genske D., and Woo S.I., Syndiotactic

Polypropene with MCM-41 Supported Metallocene [Me2C(Cp)(Flu)]ZrCl2,

Macromol. Rapid Commun., 21 (2000) 909.

Kleitz F., Marlow F., Stucky G.D., Schuth, F., Mesoporous Silica Fibers: Synthesis,

Internal Structure and Growth Kinetics, Chem. Mater., 13 (2001a) 3587.

Kleitz F., Wilczok U., Schuth F., Marlow F., Hollow Mesoporous Silica Fibers: Tubules

by Coils of Tubules, Phys. Chem. Chem. Phys., 3 (2001b) 3486.

Ko Y.S., Han T.K., Park J.W., and Woo S.I., Propene Polymerization Catalyzed over

MCM-41 and VPI-5-supported Et(Ind)2ZrCl2 Catalysts, Macromol. Rapid Commun.,

17 (1996) 749.

28 Ko Y.S., Woo S.I., Copolymerization of Ethylene and α-Olefin using Et(Ind)2ZrCl2

Entrapped inside the Regular and Small Pores of MCM-41, Macromol. Chem. Phys.,

202 (2001) 739.

Kresge C.T., M.E. Leonowicz, W.J. Roth, J.C. Vartuli, J.S. Beck, Ordered Mesoporous

Molecular Sieves Synthesized by a Liquid-Crystal Template Mechanism, Nature,

359 (1992) 710.

Lin H.P., Mou C.Y., ‘Tubules-within-a-Tubule’ Hierarchical Order of Mesoporous

Molecular Sieves in MCM-41, Science, 273 (1996) 765.

Lin H.P., Liu S.B., Mou C.Y., Tang C.Y., Hierarchical Organization of Mesoporous

MCM-41 Ropes, Chem. Commun. (1999) 583.

Lin H.P., Kao C.P., Mou C.Y., Counterion and Alcohol Effect in the Formation of

Mesoporous Silica, Micropor. Mesopor. Mat., 48 (2001a) 135.

Lin Y.S., Chemical and Electrochemical Vapor Deposition of Zirconia-Yttria Solid

Solutions in Porous Ceramic Media, Sc.D. Thesis, University of Twente, 1992.

Lin Y.S., Kumakiri I., Nair B.N., Alsyouri H., Microporous Inorganic Membranes,

Separ. Purif. Methods, 32 (2002) 229.

Lin Y.S., Microporous and Dense Inorganic Membranes: Current Status and

Prospective", Separ. Purif. Technol., 25 (2001b) 39.

Lukens W.W., Schmidt-Winkel P., Zhao D., Feng J., Stucky G.D., Evaluating Pore Sizes

in Mesoporous Materials: A Simplified Standard Adsorption Method and a

Simplified Broekhoff-De Boer Methodology, Langmuir, 15 (1999) 5403.

Mann S., Ozin G.A., Synthesis of Inorganic Materials with Complex Forms, Nature, 382

(1996) 313.

29 Marlow F., Slpliethof B., Tesche B., Zhao D., The internal architecture of mesoporous

silica fibers, Adv. Mater., 12 (2000a) 961.

Marlow F., Zhao D., Stucky G.D., Doped Mesoporous Silica Fibers: The Internal

Structure, Microp. Mesop. Mater., 39 (2000b) 37.

Marlow F., Kleitz F., Mesoporous Silica Fibers: Internal Structure and Formation,

Micropor. Mesopor. Mater., 44 (2001a) 671.

Marlow, F., Leike I., Weidenthaler C., Lehmann C.W., Wilczok U., Mesostructured

Silica Fibers: Ring Structures in Reciprocal Space, Adv. Mater., 13 (2001b) 307.

Martin J., Anderson M., Odinek J., Newcomer P., Synthesis of Periodic Mesoporous

Silica Thin Films, Langmuir, 13 (1997) 4133.

Miyata H., Kuroda K., Preferred alignment of Mesochannels in a Mesoporous Silica Film

Grown on a Silicon (110) Surface, J. Am. Chem. Soc., 121 (1999) 7618.

Mokaya R., Jones W., Synthesis Of Acidic Aluminosilicate Mesoporous Molecular

Sieves using Primary Amines, Chem. Commun., 8 (1996) 981.

Monnier A., Schüth F., Huo Q., Kumar D., Margolese D., Maxwell R.S., Stucky G.D.,

Krishnamurty M., Petroff P., Firouzi A., Janicke M., Chmelka B.F., Cooperative

Formation of Inorganic–Organic Interfaces in the Synthesis of Silicate

Mesostructures, Science 261 (1993) 1299.

Nishiyama N., Koide A., Egashira Y., Ueyama K., Mesoporous MCM-48 Membrane

Synthesized on a Porous Stainless Steel Support, Chem. Commun., (1998) 2147.

Ogawa M., Formation of Novel Oriented Transparent Films of Layered Silica-Surfactant

Nanocomposites, J. Am. Chem. Soc., 116 (1994) 7941.

30 Ozin G.A., Curves in Chemistry: Supramolecular Materials Taking Shape, Can. J. Chem.,

77 (1999) 2001.

Pevzner S., Regev O., Yerushalmi-Rozen R., Thin Films of Mesopoorus Silica:

Preparation and Characterization, Curr. Opin. Colloid. Interface Sci., 4 (2000) 420.

Ramila A., Munoz B., Perez-Pariente J., Vallet-Regi M., Mesoporous MCM-41 as Drug

Host System, J.Sol-Gel Sci. Tech., 26 (2003) 1199.

Regev O., Nucleation Events during the Synthesis of Mesoporous Materials using

Liquid Crystalline Templating, Langmuir, 12 (1996) 4940.

Ryoo R., Kim J.M., Ko C.H., Shin C.H., Disordered Molecular Sieves with Branched

Mesoporous Channel Network, J. Phys. Chem. 100 (1996) 17718.

Ryoo R., Ko C.H., Cho S.J., J. M. Kim, Optically Transparent, Single-Crystal-Like

Oriented Mesoporous Silica Films and Plates, J. Phys. Chem. B, 101 (1997) 10610.

Schacht S., Huo Q., Voigt-Martin I.G., Stucky G.D., F. Schuth, Oil-water Interface

Templating of Mesoporous Macroscale Structures, Science, 273 (1996) 768.

Schmidt-Winkel P., Yang P.D., Margolese D.I., Chmelka B.F., Stucky G.D., Fluoride-

Induced Hierarchical Ordering of Mesoporous Silica in Aqueous Acid-Syntheses,

Adv. Mater., 11 (1999) 303.

Stucky G.D., Monnier A., Schüth F., Huo Q., Margolese D., Kumar D., Krishnamurty

M., Petroff P., Firouzi A., Janicke M., Chmelka B., Molecular and Atomic Arrays in

Nano-and Mesoporous Materials Synthesis, Mol. Cryst. Liq. Cryst. 240 (1994) 187.

Suib S.L., Zeolitic and Layered Materials, Chem. Rev., 93 (1993) 803.

Tanev P. T., T. J., Pinnavaia, A Neutral Templating Route to Mesoporous Molecular-

Sieves, Science, 267 (1995) 865.

31 Tanev P.T., Liang Y., Pinnavaia T.J., Assembly of Mesoporous Lamellar Silicas with

Hierarchical Particle Architectures, J. Am. Chem. Soc., 119 (1997) 8616.

Tolbert S.H., Firouzi A., Stucky G., Chmelka B., Magnetic Field Alignment of Ordered

Silicate-Surfactant Composites and Mesoporous Silica, Science, 278 (1997) 264.

Van Looveren L.K., Geysen D.F., Vercruysse K.A., Wouters B.H., Grobet P.J., and

Jacobs P.A., Methylaluminoxane MCM-41 as Support in the Co-oligomerization of

Ethylene with Propene with [C2H4(1-Ind)2]Zr(CH3)2, Angew. Chem. Int. Ed., 37

(1998) 517

Vartuli J.C., Kresge C.T., Leonowicz M.E., Chu A.S., Mc-Cullen S.B., Johnson I.D.,

Sheppard E.W., Synthesis of Mesoporous Materials: Liquid-Crystal Templating

versus Intercalation of Layered Silicates, Chem. Mater., 6 (1994) 2070.

Walsh D., Hopwood J.D., Mann S., Crystal Tectonics-Construction of Reticulated

Calcium-Phosphate Frameworks in Bicontinuous Reverse Microemulsions, Science,

264 (1994) 1576.

Xomeritakis G., Synthesis of Inorganic Membranes by Vapor Deposition Processes:

Theory and Experiments, Ph.D. Thesis, University of Cincinnati, 1997.

Yang H., Coombs N., Sokolov I., Ozin G.A., Free-Standing and Oriented Mesoporous

Silica Films Grown at the Air-Water Interface, Nature, 381 (1996) 589

Yang H., Coombs N., Ozin G.A., Morphogenesis of Shapes and Surface Patterns in

Mesoporous Silica, Nature, 386 (1997a) 692.

Yang H., Coombs N., Sokolov I., Ozin G.A., Registered Growth of Mesoporous Silica

Films on Graphite, J. Mater. Chem., 7 (1997b) 1285.

32 Yang P., Zhao D.Y., Chmelka B.F., Stucky G.D., Triblock-Copolymer-Directed

Syntheses of Large-Pore Mesoporous Silica Fibers, Chem. Mater., 10 (1998a) 2033.

Yang. H., Ozin G.A., Kresge C.T., The Role of Defects in the Formation of Mesoporous

Silica Fibers, Films, and Curved Shapes, Adv. Mater., 10 (1998b) 883.

Yang P., Zhao D., Chmelka B., Stucky G.D., Triblock-Copolymer-Directed Syntheses of

Large-Pore Mesoporous Silica Fibers, Chem. Mater., 10 (1998c) 2033.

Yang S.M., Yang H., Coombs N., Sokolov I., Kresge C.T., Ozin G.A., Morphokinetics:

Growth of Mesoporous Silica Curved Shapes, , Adv. Mater., 11 (1999a) 52.

Ying J.Y., Mehnert C.P., Wong M.S., Synthesis and Applications of Supramolecular-

Templated Mesoporous Materials, Angew. Chem. Int. Ed., 38 (1999b) 56.

Zhang W., Pauly T.R., Pinnavaia T.J., Tailoring The Framework and Textural Mesopores

of HMS Molecular Sieves through an Electrically Neutral (SoIo) Assembly Pathway,

Chem. Mater., 9 (1997) 2491.

Zhao D., Feng J., Huo Q., Melosh N., Fredrickson G.H., Chmelka B.F., Stucky G.D.,

Triblock Copolymer Syntheses of Mesoporous Silica with Periodic 50 to 300

Angstrom Pores, Science, 279 (1998a) 548.

Zhao D., Huo Q., Feng J., Chmelka B.F., Stucky G.D., Nonionic Triblock and Start

Diblock Copolymer and Oligomeric Surfactant Synthesis of Highly Ordered,

Hydrothermally Stable Mesoporous Silica Structures, J. Am. Chem. Soc., 120

(1998b) 6024.

Zhao D., Yang P.D., Huo Q.S., Chmelka B.F., Stucky G.D., Topological Construction of

Mesoporous Materials, Curr. Opin. Solid St. Mater. Sci., 3 (1998c) 111.

33 Zhao D., Huo Q., Feng J., Kim J., Han Y., Stucky G.D., Novel Mesoporous Silicates with

Two-Dimensional Mesostructure Direction using Rigid Bolaform Surfactants, Chem.

Mater., 11 (1999) 2668.

34 CHAPTER 2

QUISCENT INTERFACIAL SYNTHESIS OF MORPHOLOGICAL

NANOSTRUCTURED SILICA

2.1. Introduction

The discovery of ordered mesoporous silicates by the template-assisted approach [Kresge et al. 92, Beck et al. 92] with variety of phase structures has led to development of several techniques for controlling the product shape. Spheres [Grun et al. 97], thin films [Yang et al. 96, Hillihouse et al. 97], curved shapes [Ozin et al. 97], and rods [Schmidt-Winkle et al. 99] are examples of morphologies that has been created using this approach. As introduced in Chapter 1, the oil-water interfacial growth approach under acid quiescent conditions can generate mesoporous silica with rich morphologies [Schacht et al. 96, Huo et al. 97]. Mesoporous silica fiber (MSF) is one of the product morphologies that were created by this approach. The fibers are

100 µm - 5 cm long and have a circular cross-section of 1-15 µm diameter.

Early microscopy investigations (TEM and optical microscopy) reveal that the pores align parallel to the fiber axis with high fidelity [Huo et al. 97]. However, recent studies [Marlow et al. 01, Marlow and Kleitz 01, Kleitz et al. 01] show that the pores run in a circular direction around the fiber axis. A mechanism was accordingly proposed to describe this novel architecture. In the proposed mechanism the long rod surfactant micelles available in the synthesis media undergo aggregation and restructuring processes that lead to formation of nanoscopic coil seeds. The simultaneous growth of seed coils by further aggregation of long rod micelles around the coils and silica condensation around the coil seeds causes the formation of the inner circular architecture from hexagonally packed pores.

35 Several studies have utilized the interfacial approach to prepare MSF under various conditions of silica precursors, solvents, and acid concentrations [Huo et al.

97, Marlow et al. 00, Kleitz et al. 01]. Most of these studies have focused on exploring the microscopic properties and formation of the internal structure of the fibers. Little attention was given to the effect of the synthesis conditions such as acid type, acid concentration and growth time on the fiber kinetics and the macroscopic properties such as fiber diameter and length and product morphology, which are key factors for some applications. These macroscopic results can be potential for utilizing the interfacial approach to expand the novel structure of MSF into more application- desirable shapes such as membranes.

Ordered mesoporous silica materials have been utilized to improve catalysis and separation applications that suffered from common limitations imposed by conventional microporous and mesoporous materials. Mesoporous silica with hexagonal pore structures has been recently used as supports for several types of catalysts for olefin polymerization [Tudor and O’Hare 97, Van Looveren et al. 98,

Kageyama et al. 99, Ko and Woo 01]. The straight nanotubes of these particles serve as polymerization reactors to template the growth of polymer fibers with extended- chain crystalline structures and properties unattainable by conventional techniques.

This application is based on a batch process utilizing mesoporous silica particles.

Commercialization of the process requires the development and use of nanostructured silica as membrane with controlled microstructures as a support.

The above brief review indicates that the successful application of ordered mesoporous materials requires the fabrication of these materials into desirable macroscopic shapes with controlled morphologies. In this chapter, first the quiescent interfacial growth of morphological silica products is systematically investigated at a

36 wide range of precursor type, growth times, acid type and concentration, and temperatures conditions. The influence of synthesis conditions on silica morphologies, microscopic and macroscopic properties, and mechanism of formation is reported with emphasis on the fiber morphology. The second part of this chapter demonstrates the use of MCM41-type hexagonal mesoporous silica particles for extrusion polymerization of polyethylene crystalline fibers. The overall results will be a key point for fabrication of nanostructured membranes in the following chapters.

2.2. Experimental

2.2.1. Quiescent interfacial growth and characterization of mesoporous silica

Mesoporous silica samples were prepared using cetyltrimethylammonium bromide (CTAB) (Aldrich) as surfactant and tetrabutylorthosilicate (TBOS) (Aldrich) as silica source in acidic synthesis based on procedures described in the literature

[Huo et al. 97, Marlow and Kleitz 01]. Silica samples were prepared under four different variables namely: growth time, acid concentration, acid type and temperature. The standard preparation conditions are 1 TBOS: 0.5 CTAB: 58.4 HCl:

2000 H2O molar ratio at room temperature and aged for 7 days. For preparation at different growth times, the samples were aged for 4, 7, and 10 days with molar ratio and temperature conditions similar to the standard. Different acid concentration samples were prepared with HCl/TBOS molar ratio ranging from 4 to 98. Nine samples were accordingly prepared with HCl volume percentage of 0.4-12% with other conditions same as the standard one. Effect of acid type and counter ion involved in synthesis of the mesoporous silica product was performed using HCl and

HNO3 with acid/TBOS molar ratios of 40, and H2SO4 with acid/TBOS molar ratios of

20 and 40. Finally samples at different temperatures were prepared with HCl acid at

37 27, 50 and 65 ºC under the standard conditions. Sample preparation conditions are summarized in Table 2-1.

In the synthesis, CTAB was dissolved in water under mild mixing. Subsequently

HCl (6M) was added slowly and the solution was mixed until it became clear. This solution (H2O, CTAB, HCl) is referred to as the water phase. Mixing was stopped and

TBOS (referred to as silica phase) was slowly added to form a thin layer above the aqueous solution. After almost two days a thin film can be observed at the interface between the water and silica phases. Silica fiber started to grow from the bottom of the thin film towards the water phase approximately after 2-3 days. After aging for the required growth time, silica fibers were removed carefully, washed with deionized water, and dried in air. The calcination was carried out at 813 K for 6 hours to remove the surfactant. MSF synthesis was also tried using a mixture of tetraethylorthosilica

(TEOS) (Aldrich) with organic solvent (hexane) as the silica phase.

Powder X-ray diffraction (XRD) patterns were obtained on Siemens D-50 using

CuKα radiation. XRD spectra were taken between 2θ of 1.5-8 and used to identify the phase structure and interplaner (d) spacings. Nitrogen adsorption-desorption measurements were performed on Micromeritics ASAP-2000. The samples were degassed at 200 ºC for 2 hours and analyzed at 77 K. The adsorption isotherm was used to calculate the BET surface area and pore volume. The desorption isotherm was used to obtain the pore size distribution from which the average pore diameter was calculated. Morphology and fiber pore diameter and length distributions were studied by scanning electron microscopy (Cambridge S-90B) and light microscopy

(Optiphot100-Nikkon). The fiber diameter and length distributions and aspect ratio were estimated by taking a representative section from the fiber microscope image and measuring the diameter and length of 100 randomly selected fibers. From the

38

Table 2-1. Summary of mesoporous silica samples and preparation conditions Sample Silica Growth Acid Acid Concentration Temp. Source time type (ºC) Acid/Si % vol. (day) Molar ratio 1 TBOS 4 HCl 58 5.1 27 2 7 3 10

4 TBOS 7 HCl 4 0.3 27 5 10 0.8 6 15 1.2 7 20 2.0 8 40 3.8 9 58 5.1 10 74 6.4 11 98 8.4

12 TBOS > 14 HCl 40 3.8 27

13 HNO3 4 0.4 14 10 1.0 15 20 2.0 16 40 3.9

17 H2SO4 20 3.0 40 5.9

18 TBOS 7 HCl 58 5.1 27 19 50 20 65 21 22 TEOS 7 HCl 58 5.1 27

Standard preparation conditions: 1 TBOS: 0.5 CTAB: 58 acid: 2000 H2O molar ratio, growth time: 7 days, temp.: 27 ºC.

39 measured values the fiber length and diameter, accumulative curves were constructed.

The fiber diameter and length distributions can be calculated from the cumulative curves by taking slopes at different points.

The silica yield is defined as the mass ratio of the silica formed to the total mass of silica present in the TBOS input. It is calculated for the fiber morphology using the calcined weight, TBOS mass input, and the theoretical silica (SiO2) production per gram of TBOS input which is 0.2 g SiO2/g TBOS. (Yield= calcined MSF mass/TBOS mass×0.2). The consumption of the surfactant (CTAB) was calculated from the difference in the silica fiber weight before and after calcination. The yield and consumption of the surfactant values give indication about the amount of TBOS and surfactant respectively utilized to produce mesoporous silica.

2.2.2. Ethylene extrusion polymerization

The polymerization experiments were performed in a collaboration work with

Professor Shiping Zhu at the University of McMaster/Canada. MCM-41 nanostructured silicate particles supported with titanocene dichloride catalyst

(Cp2TiCl2) were used as support for extrusion polymerization. For synthesis of MCM-

41 particles, ammonium hydroxide (NH4OH, 30 wt%), distilled water and CTAB were mixed under heating to obtain a homogeneous solution. Tetraethylorthosilicate

(TEOS, Aldrich) was added to the mixture under mixing. The resulting solution (1

TEOS: 0.123 CTAB: 29.38 NH4OH: 506 H2O) was left to age for 6 h at room temperature. The solid was finally obtained by filtration, washing with DI water and drying at room temperature. The surfactant was removed by calcination in air at 550

°C for 6 h [Cai et al. 99]. MCM-41 particles were characterized by X-ray diffraction, nitrogen adsorption–desorption and SEM techniques.

40 For catalyst supporting, the calcined MCM-41 particles (1.0 g) were mixed with

1.0 mmol of the Cp2TiCl2 catalyst (Strem Chemicals) in 60 ml of toluene. After stirring for 20 h at room temperature, the solid was filtered, collected and washed first with a large amount of toluene and then with 100 ml of anhydrous pentane. The supported catalyst was dried in vacuum at room temperature overnight. The extrusion polymerization of ethylene with the MCM-41-supported Cp2TiCl2 catalyst were conducted in a 1-L autoclave stainless steel reactor equipped with an air-activated mechanical stirrer. The reactor was carefully cleaned and purged before the reaction.

Toluene (400 ml) and the required amount of modified methylaluminoxane aluminum

(MMAO) solution (Akzo-Nobel) were added to the reactor under nitrogen protection.

The mixture was stirred for 10 min, and the reactor was heated to establish the polymerization temperature. The catalyst suspension was then added, and the system was stirred for 10 min and pressurized to the desired ethylene pressure to start the polymerization under stirring speed of 1000 rpm. The polymerization was stopped by venting of the reactor. The fibrous polymer product was washed with a large amount of methanol acidified with 0.2% HCl and then dried overnight.

A differential scanning calorimetry (DSC) analysis was carried out with a TA

Instruments Thermal Analysis 2910 MDSC instrument under 30 ml/min purge of

UHP N2 gas. The polyethylene sample (5 mg) was first heated to 180 °C at a rate of

10 °C/min. It was then cooled to 20 °C at 10 °C/min. Subsequently, a second heating cycle was conducted at a heating rate of 10 °C/min. The polymer molecular weights and molecular weight distributions were measured at 140 °C in 1,2,4-trichlorobenzene with a Waters Alliance GPCV 2000 with a differential refractive index (DRI) detector coupled with an inline capillary viscometer. The polymer molecular weight data were calculated with a polystyrene-based universal calibration curve.

41 The XRD spectra of the polymers were recorded on a Bruker D8 Advance diffractometer. A morphological study of the PE fibers was conducted on an

Electroscan ESEM 2020. The mechanical properties of the fibers were measured with a calibrated TA Instrument DMA 2980 dynamic mechanical analyzer with tensile clamps in the controlled force mode. The instrument required a minimum fiber length of about 5 mm and allowed for minimum and maximum tensile forces of 0.001 and 18

N, respectively. All the tensile measurements were conducted at 35 °C. The microfiber samples were carefully mounted onto the tensile clamps. A preload force of 0.005 N and a ramping rate of 0.05 N/min were chosen for the tests.

2.3. Results and Discussion of Interfacial Growth

2.3.1. Results of Synthesis

In the acidic stagnant synthesis media silica source diffuses slowly through the silica-water interface to the water phase and condense with the surfactant micelles to form three different silica forms: amorphous thin film or membrane at the silica-water interface, silica fibers grown from the bottom of the thin film in the water phase, and small fine particles that usually precipitate in the bottom of the water phase. The amount of each form and the fiber growth rate are dependent on the precursor type and growth conditions. Samples prepared at different growth times show that after two days the thin film started to grow at the interface between the silica and water phases in addition to some traces of fibers appearing in the water phase with its base attaching to the bottom of the thin film. After 4 days of growth a white complicated network of interwoven fibers can be seen in the water phase. Typical appearance of the MSF growth for the three different growth times for a sample prepared with 8.4 vol% HCl is shown in Figure 2-1. Some of the fibers grow longitudinally with maximum length exceeding 5 cm.

42

(a) (b) (c)

Figure 2-1. Fiber growth for MSF sample 11 prepared with 8.4 vol% HCl at 27 ºC

(HCl/Si molar ratio of 98): a) after 2 days, b) after 4 days, and 3) after 7 days.

Fine particles can be observed almost from the third day of growth. When fine particles start to precipitate their relative amount to the fibers is small. In general the amount of the three different forms increase with growth time. However, the growth of fiber form is the most significant. The amount of fibers increases almost linearly with time under the studied range of growth times. The ratios of fibers to fine particles and to the film become larger as growth time proceeds from 4 to 10 days.

Silica source used can affect the type of the grown silica form during synthesis. For example, growth using TEOS silica source dissolved in hexane using the above given molar ratio has lead to the formation of only relatively thick amorphous film without any fibers or particles. Other molar ratios of TEOS can produce fibers as demonstrated by other studies [Huo et al. 97, Kleitz et al. 01]. On the other hand, when TBOS was used a thin film, fibers, and fine particles were formed during synthesis with fiber amount relatively larger than the fine particles. For this reason TBOS was used as the silica source to study the effect of other growth conditions on the type, composition, and properties of the silica forms.

43 MSF preparation under different HCl acid concentration has lead to the three different forms of silica with relatively small amount of finer particles. In general three growth trends have been observed for MSF synthesis for 7 days under various acid concentrations; 1) no growth of any silica form at acid contents ≤0.4 or ≥12% vol%, 2) slow fiber growth at 1.0-2.0 vol% and 8.4-12.0 vol%, and 3) fast fiber growth at 2.0-8.4 vlo% range of acid content. In the fast growth range, fibers start to grow earlier compared to the slow growth where fiber growth becomes very slow and may take long time (few weeks) to form. This behavior gives a maximum in the fiber amount at 6.4 vol% HCl concentration. At concentrations lower than 0.4 or higher than 12.0 vol% no fibers were formed within 7 days of growth. Growth at high acid concentration may produce only a thick film at the silica-water interface.

Mesoporous silica growth using HCl acid was fast and has lead to fiber morphology within short time (~ 4 days). In contrary, HNO3 and H2SO4 acids have significantly slowed down the growth process to 14-30 days and lead to small yield of distorted structure of mesoporous silica product as film at the interface and fine particle precipitates. No fiber growth was observed under HNO3 and H2SO4 acids.

Preparations at high temperatures (50, 65 ºC) did not show the production of any fibers. The synthesis ended up with only particles grown at the interface and in the bulk of the water phase. The particles have hexagonally ordered mesoporous pores typical to MCM-41. This behavior at high temperature could be attributed to two factors: enhanced diffusivity of silica source in the water phase and the presence of convection streamlines inside the water phase. These factors can act together to supply silica source to the surfactant micelles from 3 directions both at the interface and inside the water phase. This 3-dimensional growth causes the formation of particle morphology rather than fibrous.

44 Among the various growth variables of mesoporous silica by the acidic interfacial method (summarized in Table 2-1), fiber morphology was observed as the major silica product only with HCl acid of HCl/TBOS molar ratio of 10 to 74 at ambient conditions. The counter ion involved with the use of other acids and the acid concentration appear to have some interruption on the mechanism of silica fiber growth and therefore yield non-fibrous products. In the following sections, the macroscopic and microscopic properties of the fibrous morphology will be evaluated and compared at different synthetic conditions. Then the non-fibrous morphologies and the conditions leading to them, such as the counter ion and acid concentrations, will be analyzed and used to improve our understanding of the mechanism of formation of mesoporous silica fibers.

2.3.2. Macroscopic properties of silica fibers

Figure 2-2 shows macroscopic view of the silica fibers. The diameter of the fiber, as shown in Figure 2-3, is around 10 µm. The fibers observed under microscope appear to be uniform in length, about 200-300 µm, which, however, does not represent the original length of the fibers up to a few centimeters. The process of collecting fibers from the synthesis solution caused breakage of the fibers along the length direction (not diameter). Therefore the diameter of the fibers observed under the microscope could be a close approximate of the fibers in the synthesis solution.

Diameter distributions for the silica fibers prepared were estimated from the micrographs. Figure 2-3 shows fiber diameter distribution for the samples prepared with different acid concentration (samples 7 to 11). The results of the fiber size and silica yields for these samples are summarized in Table 2-2. As shown, the fiber diameter increases with acid concentration reaching a maximum at 6.4 vol% and then

45

Figure 2-2. SEM image of sample grown for 10 days showing the fiber morphology of sample 3 prepared at 27 ºC with 5.1 vol% HCl acid (HCl/Si molar ratio 58).

5 µ

a b

10µ 4 µ

c d

Figure 2-3. Enlarged view of cross-section of mesoporous silica fibers prepared with HCl acid at room temperature of samples: (a) 3, (b) 7, and (c) and (d) 8.

46

120

100

80

60 No. 2.0 vol% HCl 3.8 % 40 5.1 % 6.4 % 20 8.4 %

0 0 5 10 15 20 25 30 35 Fiber Diameter (µm)

Figure 2-4. Accumulative fiber diameter distributions curves at different acid concentrations

Table 2-2. Fiber diameter, mass production and yield of mesoporous silica fibers prepared at different acid concentrations (samples 6-11) at 27 ºC for 7 days

Fiber diameter (µm) HCl % Silica % CTAB average distribution vol.% Yield consumption

2.0 6 4-12 5.62 0.98

3.8 9 4-18 6.81 5.60

5.1 9 4-18 19.94 9.62

6.4 10 4-24 20.71 6.47

8.4 7 4-12 8.64 0.78

47 starts to decrease at higher concentrations. The acid concentration of 6.4 vol%

(corresponding to HCl/Si molar ratio of 73) gives the silica fiber with largest average diameter and broadest fiber diameter distribution and highest silica yield. The surfactant consumption rate and silica yield show a similar behavior as the fiber diameter with increasing acid concentration. The highest silica yield and surfactant consumption occur at 5.0-6.4 HCl vol%.

Figure 2-5 shows the accumulative fiber distribution for samples prepared under different growth times (samples 1 to 3). The average fiber diameter and diameter range for the samples with three different growth times are summarized in

Table 2-3. As shown in Figure 2-5, the average fiber diameter is respectively about 6,

9 and 12 µm for samples with growth times of 4, 7 and 10 days. The distribution of fiber diameter becomes broader as growth time increases. The silica yield and surfactant consumption of fibers prepared with different growth times are also summarized in Table 2-3. As shown, the low values of yield and CTAB consumption for production of MSF reflect that the formation process is slow and encounters high resistances. Main resistances included are resistance to silica source diffusion from the strong hydrophobic silica phase to the hydrophilic water phase through the film at the interface and resistance to silica diffusion to the aggregated surfactant inside the water phase. Surfactant may face some resistance to diffusion towards the silica source inside the water phase but it is in general small since the mixing of water phase before adding silica source causes the surfactant to be well distributed and available in the molecular scale.

48

120

100

80

60 No.

40 4 days 7 20 10

0 0 5 10 15 20 25 30 35 Fiber Diameter (µm)

Figure 2-5. Accumulative fiber diameter distribution curves at different growth times

Table 2-3. Fiber diameter, mass production and yield of mesoporous silica fibers prepared with different growth times (samples 1-3) at 27 ºC and 5.1 vol% HCl

Fiber diameter (µm) Growth time % Silica % CTAB average distribution (day) Yield consumption

4 6 3-11 7.65 2.08

7 9 4-18 19.94 9.62

10 12 7-22 34.69 16.89

49 Figure 2-6 plots fiber yield versus growth time. The first two days appear to be the induction period during which no fiber is formed. It is interesting to note that fiber mass production is linearly dependent on the growth time as shown in Figure 2-6.

Silica source mass transfer and MSF mass production are inversely proportional to the diffused length inside the water phase, i.e. MMSF = A DSW (∆C/L)t where MMSF is the fiber mass production, A is the diffusing area, DSW is the diffusion coefficient for silica source in the water phase, ∆C is the concentration difference of silica source between the silica and water phase, L is the diffusion length, and t is the growth time.

Silica fibers grow in length with time (L is increasing) and their heads go deep in the water phase. If the silica fibers are assumed to grow from their heads this implies that silica source must diffuse long distance, L, to condense at the silica head so the fiber mass production should decrease with the growth time. However, the linear dependency of fiber mass production with time may indirectly imply that the fibers grow from their bottoms rather than the heads.

The average fiber diameter increases from 6 to 12 micron when the growth time increases from 4 to 10 days. If one assumes the increase in the silica yield is due to increase in the fiber diameter with the fiber length and fiber number remains constant, then the silica yield should be proportional to squares of the fiber diameter. The experimental data listed in Table 2-3 appear to follow this relationship. However, we clearly observed that fiber length increases substantially with the growth time, as indirectly shown in Figure 2-1. It is more likely that the increase in the silica yield is due to growth of the fiber in the length direction, not the diameter. It is possible that with increasing growth time fibers coalesce to form large fibers, thus decreasing total number of fibers present in the synthesis vessels.

50

0.6

0.4

0.2 Mass ofmesoporous silica fibers (g) 0 024681012 Time (day)

Figure 2-6. Effect of aging time on mesoporous silica fiber mass production for samples 1 to 3 prepared at 27 ºC with 5.1 vol% acid for 4-10 days.

Fiber aspect ratio, defined as fiber length to diameter ratio, was estimated from micrograph images. It was found that the aspect ratio for MSF samples prepared under both aging time and acid concentration variables lies in the range of 3-30 with average value close to 12. Very few fibers (<5%) show an aspect ratio larger than 50.

These values correspond to broken fibers as a result of sample collection process.

Fibers with original length could have aspect ratio values of at least 5000.

2.3.3. Microscopic Properties of silica fibers

Figure 2-7 shows TEM images of mesoporous silica fibers prepared for 14 days at 27 ºC with 5.1 vol% acid. TEM images revealed the existence of well-ordered mesopores within the silica fibers. As can be seen in Figure 2-7a, the pores are arranged in an ordered hexagonal fashion as confirmed by the FFT diffraction pattern

(inset) and appear to be non-connected and unidirectional aligning linearly along the c-axis (Fig. 2-7b).

51

(a)

100 nm

(b)

50 nm

Figure 2-7. TEM images for mesoporous silica fiber showing (a) cross section of silica fiber with the hexagonal packing of pores and its corresponding FFT pattern and (b) straight channel domains along the c-axis. Sample was grown for 14 days at 27 ºC with 5.1 vol% HCl acid).

52 XRD pattern for a mesoporous silica fiber sample is shown in Figure 2-8. The sample show the presence of three reflection peaks indexed as (100), (200), and (110) planes and interplanar d100 spacing= 3.9 nm. The peaks are indexed as (hk0) and indicate the presence of long-range order of long mesopores arranged inside the fiber in a structure typical to the p6mm hexagonal order observed in MCM-41 materials

[Beck et al. 92]. The hexagonal lattice parameter (ao) which represents the center-to- center spacing between mesopores calculated from the d100 value was 4.5 nm as schematically shown in Figure 2-8a. The wall thickness (wp) was evaluated from the lattice parameter (ao= 4.5 nm) and the pore diameter (dp= 2.67 nm evaluated from N2 ads-des isotherms as will be presented shortly) as 18Å.

The nitrogen sorption isotherms of calcined silica fibers are of type-IV (Figure

2-9b) characteristic of mesoporous materials with well-aligned channels. The first step in the adsorption isotherm is associated with the coverage of the inner surface with monolayer of nitrogen molecules and the second step is due to capillary condensation of nitrogen molecules in the cylindrical pores. The adsorption and desorption isotherms almost coincide with no hysteresis which is a characteristic of mesoporous silica materials with pore diameters less than 3.8 nm [Inoue et al. 98]. A narrow pore size distribution with mean BJH diameter value of 2.7 nm was obtained from the desorption branch (inset of Fig. 2-9b). This value is similar to that obtained from the adsorption branch indicative of a uniform cylindrical pore shapes. The fibers have a BET surface area close to 1000 m2/g, pore volume of 0.62 cm3/g and 57 vol% porosity. These microscopic results are typical to all silica fiber samples obtained through this study. Variation of the synthetic conditions can, however, induce differences in the fiber microscopic properties as will be discussed in the following paragraphs.

53

(a) 100 (b) pore diameter (dp)

wall thickness (wp) ao d100

60º 110 200

1234567 Lattice parameter (a ) = d × 2/ 3 2θ (degrees) o 100 Wall thickness (w ) = a − d p o p

Figure 2-8. XRD diffraction pattern for mesoporous silica fiber prepared at conditions of figure 2-7 (a), and evaluation of lattice parameter and wall thickness from the hexagonal pore arrangement (b).

450 Adsorption 400 Desorption /g STP) /g

3 350

300 /g) 3 250

200 Pore Vol. (cm Quantity Adsorbed (cm Adsorbed Quantity 150 22.533.544.55 Pore Diam. (nm) 100 0 0.2 0.4 0.6 0.8 1 Relative Pressure (P/Po)

Figure 2-9. Nitrogen sorption isotherm for silica fiber sample 9 grown at conditions of figure 2-7. Inset is the pore size distribution obtained using desorption branch by BJH method.

54 XRD patterns for three samples prepared at different growth times are shown in

Figure 2-10. The samples show the presence of three reflection peaks indexed as

(100), (200), and (110) planes typical to hexagonal pore symmetry. It can be noticed that longer growth time not only results in longer and bigger fibers (as previously discussed in the macroscopic properties) but also improves the uniformity of the fiber hexagonal pore structure as indicated by increasing the intensity and the number of

XRD peaks in Figure 2-10. The pore structure of the silica fibers with different growth times is summarized in Table 2-4. Within the accuracy of the data, the interplanar spacing, lattice parameter, pore size and pore wall thickness of all three silica samples are similar. All samples under different aging times are mesoporous with average pore diameter of ca. 2.5 nm. However, the surface area and pore volume increase with increasing growth time.

Figure 2-11 shows the XRD patterns for samples prepared at different acid concentrations. All samples, except at 2.0 vol%, exhibit clear XRD peaks indicative of ordered mesopore structure. In case of low acid content, 2.0 vol%, low signal-to- noise ratio suggests the presence of amorphous product. The increase in acid content is accompanied by narrow and higher intensity (100) peak indicating a substantial improvement in structural ordering. Best quality hexagonal structure was observed for sample obtained with 6.4 vol%. Further increase in the acid concentration, ≥ 8.4%, leads to reduction in pore ordering as indicated by reduction in the pore intensity. The pore structure of these samples is summarized in Table 2-5. As shown, increasing acid concentration causes slight increase in the pore center-to-center spacing, ao, between

4.04-4.55 nm. The change in pore center-to-center pacing can be caused by change in pore wall thickness (wp) and /or pore size (dp) since ao=wp+dp. As shown in Table 2-5, increasing acid content causes a very slight decrease in the pore size (~1 Å) compared

55

100 110 200

x 4 10 days

7 days x 4 4 days

1234567 2θ (degrees)

Figure 2-10. XRD diffraction patterns for samples under different growth times

Table 2-4. Interplanar spacing and pore wall Thickness, and N2 adsorption- desorption results for the samples prepared at different growth times (T=27 ºC at 5.1 vol%HCl) Growth Interplanar Lattice BJH Pore wall Pore BET

time d100 parameter pore thickness volume surface

spacing ao diameter area (day) (nm) (nm) (nm) (nm) (m3/g) (m2/g) 4 3.90 4.50 2.5 2.0 0.32 610 7 3.71 4.28 2.48 1.8 0.59 1032 10 3.77 4.35 2.42 1.93 0.72 1107 ao=d100×2/√3, pore wall thickness=ao−pore diameter.

56 6.4 vol% HCl

5.1 %

8.4 %

3.8 %

2.0 %

1234567 2θ (degrees)

Figure 2-11. XRD diffraction patterns for samples under different acid concentrations

Table 2-5. Interplanar spacing and pore wall thickness, and N2 adsorption- desorption results for the samples prepared at different acid concentrations (T=27 ºC, growth time=7days) HCl Interplanar Lattice BJH Pore wall Pore BET

vol.% d100 parameter pore thickness volume surface

spacing ao diameter area (nm) (nm) (nm) (nm) (m3/g) (m2/g) 2.0 3.5 4.04 2.52 1.52 0.44 865 3.8 3.69 4.26 2.48 1.78 0.74 1525 5.1 3.71 4.28 2.48 1.8 0.59 1032 6.4 3.94 4.55 2.40 2. 15 0.35 812 8.4 3.81 4.40 2.31 2.09 0.49 906 ao=d100×2/√3, pore wall thickness=ao−pore diameter

57 to ~3−6 Å increase in pore wall. Therefore, the net effect of acid content increase on increasing the ao is due to the increase in the pore wall thickness at relatively constant pore size. Change in surface area and pore volume was not uniform and can be sensitively affected by the presence of some amorphous particles in the fiber sample.

Acid is usually added to the synthesis mixture as a catalyst for silica hydrolysis and condensation. The increase in the pore wall thickness with increasing acid content is due to formation of more silica framework possibly due to enhanced hydrolysis and condensation of TBOS in presence of larger amount of acid. Acid can also affect the aggregation of surfactant rod micelles through a reduction in the surfactant molecules head area by means of re-distribution of charge at the surfactant molecule head and causes the rod micelle diameter to become smaller [Sprokel 80]. This could cause the slight decrease the pore size of pore with increasing acid content. Non-uniformity in surface area and pore volume properties could be attributed to the presence of some denser, nonfiber particles collected together with the silica fiber sample. As described earlier, the silica fiber growth is accompanied by the formation of amorphous thin film and fine particles. In bottom-left and upper-right corners of the SEM images shown in Figure 2-2 one can find such silica particles. Several samples of non-fibrous silica forms (e.g., thin film) were carefully collected from the synthesis beakers prepared at different growth times and acid concentrations and tested by XRD and N2 adsorption. These samples exhibit an amorphous XRD pattern without reflection peaks and have smaller surface area (about 145 m2/g) and pore volume (0.20 cm3/g).

These amorphous silica particles are possibly formed by hydrolysis and condensation of silica source without involvement of the template.

The silica fibers grow longer with the growth time. Therefore, the portion of the silica fibers with larger surface area and pore volume in the sample collected

58 increases with the growth time. Thus, the samples with longer growth time exhibit higher surface area and large pore volume per unit gram of the sample including both ordered mesoporous silica fiber and denser silica particles. Acid concentration may also affect the relative quantity of the denser silica particles contained in the samples collected by affecting the rate of hydrolysis and condensation of silica source in presence of template. If the rate of hydrolysis and condensation of silica is balanced by presence of flexible template, well-developed macroscopic fibers will be formed

[Kleitz et al. 01]. At higher acid content, silica formation becomes rapid and may proceed to form non-fibrous irregular forms besides the ordered fibers. This can reduce the overall uniformity of the sample as apparent for sample with 8.4 vol% acid which has less resolved diffraction pattern (Figure 2-11) and microstructural properties (Table 2-5). Thus, the non-uniformity of different pore volume and surface area of the samples under different acid contents could be due to different percentage of the denser particles in the samples.

2.3.4. Non-fibrous morphologies

According to results of synthesis discussed in section 2.3.1, some synthesis conditions have lead to non-fibrous silica morphologies for example in preparations at high temperatures (samples 19 and 20), with HCl acid contents < 20 HCl/Si molar ratios (samples 4−7) and with counterions from HNO3 and H2SO4 acids (samples

13−17). The non-fibrous morphologies of high temperature formation was attributed to altering the linear growth of silica fiber by enhancing silica source diffusion and introducing of mixing streamlines that led to growth of silica in 3 dimensions into particle shapes. In the following context the effect of low HCl acid contents and counterions on the morphological and microstructural properties of silica product will

59 be discussed. The results will be used to clarify the role of these variables in altering the fiber formation process and to improve our understanding of the mechanism of formation of ordered mesoporous silica under the quiescent, interfacial acidic growth conditions.

Growth of silica at low acid content (< 2 vol %) required 14 to 30 days to saturate. No product was observed in the first week except for traces of a film at the silica-water interface. The products will therefore have significant differences from those obtained at high acid contents (grown for 7 days) due to effects of growth time on microstructural and morphological properties. As shown in Figure 2-12, low HCl acid content with HCl/Si ratios of 4−15 (corresponding to 0.3−1.2 vol% HCl) resulted in a mixture of morphologies including dense aggregates (Fig. 2-12a), twisted ropes with textured surfaces (Fig. 2-12b), fibers (Fig. 2-12c), gyroids (Fig. 2-12d), and free

(Fig. 2-12e) or fiber-seeded (Fig. 2-12f) regular rope-derived shapes (Fig. 2-12g).

Although some long silica fibers can be seen, the low HCl acid content affects the silica-surfactant growth mechanism to co-produce silica product of rich morphologies of micrometer scales.

Powder XRD patterns of these products are shown in Figure 2-13. All patterns exhibit presence of a broad peak of low intensity at 4 to 5º. It not likely that any of these product has a hexagonal structure, but the presence of the broad peak is an indicator of a mesoporous structure probably composed of a mixture of ordered phases (including lamellar) and amorphous silica. The mesoporous structure of these products was confirmed by the nitrogen sorption isotherms shown in Figure 2-14 which depict type IV isotherms with surface area properties summarized in Table 2-6.

60

10 µ 4 µ

(a) (b)

10 µ 5 µ

(c) (d)

20 µ 20 µ 10020 µ µ 100 µ

(e) (f) (g)

Figure 2-12. Morphologies of silica synthesized under low HCl acid contents: (a) dense silica aggregates at 0.3 HCl vol%, (b) twisted ropes, (c) fibers and (d) gyroids at 0.8%, and (e), (f) and (g) mixture of fibers and particulate silica gyroids at 1.2 %.

61 1.2 vol% HCl

0.8 %

0.3 %

2345678 2θ (degrees)

Figure 2-13. XRD patterns of non-fibrous silica under low HCl acid contents

550 1.2 vol % HCl 500

450

/g STP) /g 0.8 % 3 400

350 0.3 % 300

250

200 Quantity Adsorbed (cm 150

100 00.20.40.60.81 Relative Pressure (P/Po)

Figure 2-14. Nitrogen sorption isotherms for non-fibrous silica morphologies under low HCl acid content. Filled and open and markers refer respectively to adsorption and desorption.

62 Table 2-6. Physical properties of mesoporous silica morphologies synthesized at low HCl acid content. HCl vol% BJH Pore volume BET pore diameter surface area (nm) (cm3/g) (m2/g) 0.3 4.0 0.66 775 0.8 2.70 0.80 1070 1.2 2.54 0.94 1200

The isotherm of silica prepared at 0.3 HCl vol % exhibits several adsorption steps indicating the presence of a pore size distribution (PSD). The PSD of this product has three wide peaks centered at 2.3 (wide), 4 (sharp) and 35 nm (wide) with average BJH pore diameter of 4.0 nm. Silica products at 0.8 and 1.2 vol% show type-

IV isotherms typical to mesoporous materials with average pore diameters of 2.7 and

2.5 nm respectively. As shown in Table 2-6, surface area and pore volume properties increase and the average pore diameter decreases with increasing the acid content.

The silica product at 0.3% acid has the lowest quality possibly due to presence of some amorphous dense silica particles.

In acidic conditions, the growth of mesoporous silica-surfactant assemblies by hydrolysis and condensation of the silica precursor around the surfactant micelles proceeds via the S+ X− I+ pathway where S+ is the surfactant, I+ is the inorganic silica and X− counterion provided by the acid (Cl− from HCl). The acid (HX) plays two main roles: it protonates the silica precursor (alkoxides) to facilitate condensation process and provides counterions which adsorb on the surfactant to form the S+ X− micelles that catalyze or template the hydrolysis of silica species into the silica- surfactant product. Low acid concentration means low content of H+ and X−. This will

63 reduce the extent of silicon alkoxide protonation and counterion adsorption on the surfactant which will slow down the silica condensation and the self assembly process. This process took long times (> 14 days) for samples under low acid contents and will eventually grow mesoporous silica product at the interface and as precipitates. At low acid conditions, silica precursor species can condense without involvement of the surfactant because of their low attraction to the surfactant to form dense silica that reduce the over all quality of the product (e.g., at 0.2 vol%). Higher acid concentrations will speed up the process to few days as discussed on samples in the previous section.

The exact effect of acid on controlling the product morphology is not well understood. The acid content can affect the kinetic growth aspect of the silica, but from SEM image in Figure 2-12, it is clear that it has influence on the morphology as well. In studying the effect of acid content on morphology of mesoporous silica, Lin and co-workers [Lin et al. 00] noticed that morphology transformed from ropes to gyroidal spheres upon increasing the concentration of HNO3 acid due to enhanced condensation process. In our experiments fiber and rope shape can be still observed at high acid concentrations because of the interfacial growth approach which reduces the speed of silica penetration and the rate of condensation. Under these conditions, rich morphologies were observed. Various shapes are proposed to generate through close packing of secondary building units under various accretion types and degrees of curvature induced by the defects in crystal seeds [Park et al. 01]. It is apparent from

Figures 2-12 b−d that rope, fiber and gyroid shapes emerge from submicron short disc units. Fast axial growth relative to the radial direction will lead to ropes. The opposite will lead to gyroids and spherical-like shapes. Growth of ropes in diameter with minimal curvatures gives fibers. If the growth is accompanied by bending stresses,

64 twisted shapes will be observed [Coombs et al. 97]. The presence of a mixture of morphologies at these conditions is an indicator of the complex growth process.

Variation of the counterion (supplied by the acid) has also exhibited a remarkable effect on patterning the morphology of silica product. Almost no fiber morphology was observed in samples prepared with HNO3 and H2SO4 acids (which

− 2− provide NO3 and SO4 counterions) at various concentrations. SEM micrographs of

− these samples in Figure 2-15 show that NO3 counterion has altered the fiber growth conditions and led to mainly non-fibrous shapes including corrugated and smooth spheres (Fig. 2-15a), gyroids and twisted ropes (Fig. 2-15 b-e), hollow spheres (Fig.

−2 2-15f). SO4 counterion, on the other hand, gave loose agglomerates (Fig. 2-15g) and dense silica membranes (Fig. 2-15h). Fibers have been rarely observed, e.g., Figure 2-

15c.

Powder XRD patterns of these samples are shown in Figure 2-16. All samples show the (100) plane peak at 2θ between 1.5-2.5º. XRD pattern with only one broad peak indicates the presence of a mesoporous structure with distorted pore structure.

Such product can be a mixture of ordered phases (especially lamellar and hexagonal).

Additional peaks corresponding to the hexagonal pore structure was only observed for

HNO3 at 0.4 and 1.0 vol% (with small intensities). At higher HNO3 content, the small peaks disappear and the (100) peak becomes broad and less intense reflecting a lower pore order. Silica products with H2SO4 acid have less pore structural order than those with HNO3 as indicated from their low intensity (100) plane broad peaks.

The mesoporous structure of these samples was confirmed by the Nitrogen sorption studies. All samples exhibited type-IV sorption isotherms with a typical capillary condensation step. Surface area properties are summarized in Table 2-7 and show that the average pore sizes are 3.0 nm for HNO3 and 2.5 nm for H2SO4 products.

65

(a) (b) 2 µ 2 µ

10 µ 5 µ 2 µ (c) (d) (e)

2 µ 2 µ 20 µ (f) (g) (h)

Figure 2-15. Morphologies of silica synthesized under different acids: (a) smooth and corrugated spheres at 0.4 vol% HNO3, (b) mixture of ropes and gyroids at

1% HNO3, (c), (d) and (e) mixture of ropes, gyroids, fibers and regular shapes at

2% HNO3, (f) solid and hollow spheres at 3.9% HNO3, (g) irregular loose deposits at 3% H2SO4, and (h) thin films at 5.9% H2SO4.

66

1.0 vol % HNO3

0.4% HNO3

2.0% HNO3

3.9% HNO3

3.0% H2SO4

5.9% H2SO4

1.5 2.5 3.5 4.5 5.5 6.5 7.5 2θ (degrees)

Figure 2-16. XRD patterns of non-fibrous silica morphologies prepared with

HNO3 and H2SO4 acids.

Table 2-7. Physical properties of mesoporous silica morphologies synthesized with HNO3 and H2SO4 acids Acid vol. % Interplanar BJH pore Pore wall Pore BET

d100 spacing diameter thickness volume surface (nm) (nm) (nm) (cm3/g) area (m2/g)

0.4% HNO3 4.61 3.0 2.32 0.52 623 1.0% 4.78 3.0 2.51 0.48 560 2.0% 3.40 3.0 0.93 1.38 1454 3.9% 3.19 2.9 0.78 0.61 800

3.0% H2SO4 3.20 2.5 1.20 1.15 1500 8.4% 3.46 2.6 1.40 1.13 1500

67 The spherical and gyroid-rope structures, shown in Fig. 2-15 a-b, prepared at 0.4

2 and 1 vol % HNO3 have a hexagonal structure with surface areas of 500 to 600 m /g.

Increasing the HNO3 content also give a gyroid-rope mixture with significantly increased surface area of 1450 m2/g. Increasing the acid normally enhances the silica condensation and could lead to amorphous structures with low surface areas as observed in the previous section. However, the high surface area observed implies the absence of any amorphous structures in these morphological products. The hexagonal structure products (at 0.4 and 1% HNO3) have pore size similar to those with less pore order prepared at higher HNO3 content, however, they have a thicker pore wall. This behavior and the inconsistent change in the physical surface area properties with

HNO3 acid concentration are not well understood. The mesoporous silica products

(films and loose aggregates) with H2SO4 acid have a smaller pore size and thicker pore wall compared to case of HNO3 under similar conditions. The concentration of

H2SO4 acid does not seem to significantly affect the product microstructural properties as noticed on Table 2-7.

− − The rate of the morphological silica product growth was in the order NO3 > Cl

2− > SO4 . This order is consistent with the Hofmeister series of binding affinities of anions (counterion: X−) to the cationic surfactants (S+) [Lin et al. 00, Leontidis 02].

Therefore, the self assembly process between the protonated silica species (I+) and the

+ − − − counterion-surfactant micelles (S X ) has increased in the same order NO3 > Cl >

2− SO4 as was observed experimentally. Counterion type certainly has effects on the silica growth to produce rich morphologies as shown in Figure 2-16. Based on this study, HCl acid favors the production of ordered mesoporous fibers at concentrations

> 2.0 vol%. Other acid types favor the production of ordered non-fibrous morphologies. Ozin and coworkers have related the type of morphology of

68 mesoporous materials to topological defects present in liquid crystal (LQ) [Yang et al.

98b] such as points, lines and planar defects. For example, curved morphologies such as fibers, emerge from longitudinal defects (e.g., disclination) rotating around the LQ micellar longitudinal axes. Similarly, transverse defects rotating around the transverse liquid axes can produce disc-like shapes including curved ribbons, discoids and toroids. It was demonstrated that the counterion can affect growth rate of mesoporous silica by changing the rates of hydrolysis and condensation reactions. Variation of counter ion has transformed product morphology from fibers (case of HCl) to non- fibrous shapes (with HNO3 and H2SO4). Although the exact reason is not well understood yet, it is clear, however, that the relative difference in counterion association with the cationic surfactant and the subsequent condensation of silica precursor at various rates enhance LQ defects of different types and affect the directing fields that generate the final patterns.

2.3.5. Mechanism of fiber formation

Early studies on mesoporous silica fibers [Huo et al. 97] assumed that the fibers have hexagonally packed pores that lie straight across the fiber length. However, it was found later [Marlow et al. 01, Kleitz et al. 01] that the pores run in a circular direction about the fiber axis and a mechanism of formation was accordingly suggested to explain the formation of this microscopic structure. These later studies did not highlight the importance of the macroscopic properties which may provide some useful insight into the fiber morphology formation mechanism. Here we propose a simple mechanism, schematically shown in Figure 2-17, based on the results obtained in this study.

69 Silica Phase (1) (2)

wormlike structure of micelle Water Phase Silica source Surfactant aggregate

Figure 2-17. A schematic describing the mecahnism of fiber formation. (1): Slow diffusion of silica source throught the interface towards the aggregated surfactant, (2) Growth of silica fiber from the base rather than the heads, (3) Aggregation of two neighboring fibers to from one larger diameter fiber.

As illustrated in Figure 2-17, the synthesis vessel at the beginning consists of hydrophilic water phase containing a stable solution of cylindrical micelles. After the addition of silica phase, TBOS starts to diffuse to the water phase. Since the silica phase is strongly hydrophobic it will diffuse very slowly in a one directional fashion.

At silica-water interface region where the relative amount of surfactant is small compared to the body of water phase, diffused TBOS will undergo hydrolysis and condensation without involvement of surfactant. This will result in formation of amorphous film at the interface through which TBOS will continue to diffuse to the water phase. The cylindrical micelles can undergo restructuring and aggregation process under the effect of dissolved silica species by means of van der Wall attraction leading to the formation of long wormlike micelles [Marlow and Kleitz 01,

Atkins 94]. The wormlike micelles will then undergo more restructuring and

70 aggregation process to from the internal structure with rods running circularly across the aggregate axis [Sprokel 80].

Large amount of the long aggregated rod micelles will be available at the interface and will start to self assemble with the uni-directionally diffused silica to form fibers with structure controlled by the way of aggregation of the wormlike micelles. The worm-like micelles will be, by somehow, continuously supplied at the base of the fiber and condense with the incoming diffused silica to grow the fiber from its bottom. From here it is clear why the percentage of silica source consumption is higher than the surfactant consumption. This is most probably due to the high resistance encountered by the surfactant micelles through diffusing in the water phase toward the base of the fiber. At the early stages of fiber growth most of the grown fibers are small in diameters and length. The presence of some wormlike micelle aggregate between two small neighboring fibers could lead to combining these two fibers by means of silica condensation with micelle and the unsaturated silica groups at the surface of the other fibers ending up with a fiber with larger diameter. If the diffused silica source was not accompanied by the presence of flexible aggregated micelles, the silica hydrolysis/condensation reactions will proceed without involvement of micelle and thus producing amorphous particle. These simultaneous processes continue and lead to the observed macroscopic properties.

2.4. Results and Discussion of Ethylene Extrusion Polymerization

This part demonstrates the use of the honeycomb ordered nanotubes of mesoporous silica materials in polymerization applications for fabrication of polyethylene (PE) fibers with improved mechanical and structural properties. The support used for polymerization is MCM-41-type silica particles supported with titanocene dichloride catalyst (Cp2TiCl2). Silica particles exhibit XRD diffraction

71 pattern typical hexagonal symmetry of pore arrangement. Nitrogen sorption porosimetry characterization revealed a silica mesopore diameter of 2.3 nm, a pore wall thickness of 2.2 nm, specific surface area of 744 m2/g and pore volume of 0.63 cm3/g. Mesoporous silica particles have tubular shape structure of 10-50 µm length and 10 µm diameter as shown in Figure 2-18. Each particle consists of huge number of nanotubes with length equivalent to the particle length. Titanocene catalyst was supported on silica nanotubes by direct impregnation and had a surface coverage of

0.11 mmol Ti/g silica support.

Figure 2-18. SEM micrograph of MCM-41 particles used as support for ethylene

extrusion polymerization

72 Polymerization was performed with MCM-41 supported Cp2TiCl2/MMAO catalyst system at 20 atm ethylene pressure for 1 hour at 40 ºC. The polymerization conditions and Polyethylene (PE) product molecular weight and DSC properties are summarized in Table 2-8. The polymer product was obtained as aggregates of nano- sized PE fibers confirming the template effect of the nanotubes on replication of the product morphology. The PE fibers exhibit unimodal melting point of 140 ºC in comparison with the commercial high density crystalline PE (HDPE; Tm=133-135

ºC). This indicates the presence of extended-chain crystals. Melting and re- crystallization of PE fibers have lowered the melting point (Tm) and melting enthalpy

(∆Hm) as a result of formation of folded-chain lamellar structures. The ratio of ∆Hm data to that of a perfect PE crystal (∆Hm=290 g/mol) suggest a crystallinity of 83%.

The XRD spectra (Figure 2-19) of the PE samples show a typical orthorhombic crystal structure with (110) and (200) diffraction peaks. The absence of amorphous halo at 19.6 indicates the high crystallinity of the polymer samples.

Table 2-8. Ethylene extrusion polymerizations with MCM-41 supported Cp2TiCl2 catalyst and PE fiber MW data and DSC characterization results

b Tm (ºC) ∆Hm (J/g) st nd st nd Pressure Temp. Mw 1 2 1 2 (atm) (ºC) Activity a (×104) scan scan scan scan

20 40 918 202 140 134 240 172

a Activity in kg PE/(molTi hr). c Peak temperature in the thermogram.

73

[110]

[200]

10 15 20 25 30 2 Theta (degree)

Figure 2-19. XRD spectra of the polyethylene fibers produced

Morphological features of the PE fibers were tested by SEM as shown in Figure

2-20. Three levels of fiber morphology can be identified; nanofibrils, microfibers, and fiber aggregates and bundles. The PE aggregates have porous structures consisting of large number of 1-30 µm diameter microfibers arranged in random orientations (Fig.

2-20a). The bundle morphologies can be of loosely (Fig. 2-20b) or compactly packed microfibers. The fiber aggregates and bundles represent the major morphological features of the nascent PE samples. Magnified images of individual cracked microfibers (Fig. 2-20 c and d) clearly show uniform and parallel nanofibril striations with nanofibril diameters of about 60 nm. The nanofibrils are the primary morphological units in the PE polymer samples. The morphological observations suggest that the PE fibers are produced by extrusion polymerization in honeycomb nanotubes according to the scheme proposed in Figure 2-21. Ethylene molecules polymerize under the template effect of the straight nanotubes into polymer chains.

The polymer chains grow out of the nanotubes and crystallize to form extended-chain- crystal nascent nanofibrils. Nanofibrils close pack to form individual microfibers which can further agglomerate to form fiber aggregates and bundles.

74

(a) (b)

(c) (d)

Figure 2-20. SEM micrographs of polyethylene fibers: (a) an aggregate of PE microfibers, (b) a loosely packed bundle of PE microfibers, (c) and (d) magnified images of microfibers showing nanofibrils. Scale bars for a, b, c, d are respectively 250, 150, 50 and 10 µm.

75 Figure 2-21. Proposed scheme of the PE fibers formation.

Growth of PE fiber samples into 8−25 mm length allowed the measurement of their mechanical properties for the first time. Tensile stress-strain profiles for PE fibers were curved unlike the linear behavior of ultra-high strength PE fibers made with post-treatment steps. The nascent PE microfibers exhibit a tensile modulus of

3.0−7.0 GPa (5% strain), a tensile strength at break of 0.3−1.0 GPa and elongation at break of 8.5−20%. The microfibers could sustain high tensile loading because of the extended –chain crystal structure in the nanofibrils. The low tensile modulus is caused by the high ductility of the PE fibers which is a result of the imperfect alignment of the nanofibrils in the microfibers. The nanofibrils were in a rather relaxed state instead of a fully extended state. Tensile drawing of at 80 ºC and drawing ratio of 1.6 caused re-orientation of the nanofibrils and improved the fiber mechanical properties.

Fiber drawing increased the tensile modulus from 4.4 to 14.4 GPa and the tensile strength from 0.55 to 1.6 GPa.

These results demonstrate the effective use of the hexagonally ordered nanotubes as molecular extruders for fabrication of PE fibers of extended-chain crystals. The PE fibers can be obtained in one step and have structural and mechanical properties far better than those obtained by the costly post-treatment conventional techniques. However, the process demonstrated here is a batch one that is based on a particulate silica support containing the honeycomb structure. The process can be

76 potentially scaled up and commercialized if the support is formulated as a membrane for continuous production of PE fibers. For successful use as molecular extruder, the membrane should have straight nanopores of 2-5 nm diameters aligned normal to the membrane surface. Two approaches will be demonstrated in the following chapters for the preparation of such nano-membrane reactors with controlled microstructural features that can be potential candidates for molecular extrusion applications.

77 2.5. Conclusions

Quiescent oil-water interfacial growth can be used to obtain variety of nanostructured silica morphologies. Fibers can be obtained with TBOS as the silica source under narrow range of the experimental conditions (temperature at 27 ºC,

TBOS, CTAB, HCl, and H2O molar ratio of 1, 0.5, 10−98, and 2000. Amorphous silica films or particles are formed with TEOS as the silica source dissolved in hexane. Non-fibrous geometry (films, gyroids, ropes, hollow and solid spheres and regular shapes) can be obtained at higher synthesis temperatures (above 27 ºC), low

HCl acid concentration (< 2 vol%) and with HNO3 and H2SO4 acids. Under low acid contents the growth rate becomes very slow and yields a mixture of morphologies.

The increase in HCl acid concentration speeds up the formation of silica and produce amorphous non-fibrous particles as a minor product besides the ordered fibers. Use of different acid sources (HNO3 and H2SO4) have alters the fiber growth conditions and leads to mixture of non-fibrous mesoporous morphologies. Growth time increases in the order HNO3 < HCl < H2SO4. Under controlled conditions the fibers diameter becomes larger with a broader distribution for samples with longer growth time. The experimental results suggest that the fibers grow from the bottom and coalesce to form fibers with larger diameters. Extrusion polymerization of ethylene on titanocene- supported MCM-41 type silica demonstrated the template effect of the straight nanopores on fabrication of extended-chain crystalline polyethylene fibers. Nascent nanofibrils of 60 nm diameter with extended chains grow out of the nanotubes and aggregate into 1−30 µm microfibers which further aggregate into PE fiber bundles.

Mechanical properties demonstrate that these fibers exhibit improved tensile strength compared to commercial PE fibers.

78 2.6. References

Atkins P.W., Physical Chemistry, Oxfrod, 1994.

Beck J.S., Vartuli J.C., Roth W.J., M.E. Leonowizc, Kresge C.T., Shmitt K.D., Chu

C.T., Olson D.H., Shepard E.W., McCullen S.B., Higgins J.B., Schlenker J.L.,

New Family Of Mesoporous Molecular-Sieves Prepared With Liquid-Crystal

Templates, J. Am. Chem. Soc., 114 (1992) 10834.

Cai Q., Lin W.Y., Xiao F.S., Pang W.Q., Chen X.H., Zou B.S., The Preparation of

Highly Ordered MCM-41 with Extremely Low Surfactant Concentration,

Micropor. Mesopor. Mat., 32 (1999) 1.

Coombs N., Khushalani D., Oliver S., Ozin G., Shen G.C., Sokolov I., Ynag H., Blue

Prints for Inorganic Materials with Natural Form: Inorganic Liquid Crystals and a

Language of Inorganic Shape, J. Chem. Soc., Dalton Trans., (1997) 3941.

Grun M., Lauer I., Unger K.K., The Synthesis of Micrometer- and Submicrometer-

Size Spheres of Ordered Mesoporous Oxide MCM-41, Adv Mater., 9 (1997) 254.

Hillihouse H., Okubo T., van Egmond J.W., Tsapatsis M., Preparation of Supported

Mesoporous Silica Layers in a Continuous Flow Cell, Chem. Mater., 9 (1997)

1505.

Huo Q., Zhao D., Feng J., Weston K., Buratto S.K., Stucky G.D., Schacht S., Schuth

F., Room Temperature Growth of Mesoporous Silica Fibers: a New High-Surface-

Area Optical Waveguide, Adv. Mater., 9 (1997a) 974.

Inoue S., Hanzawa Y., Kaneko K., Prediction of Hysteresis Disappearance in the

Adsorption Isotherm of N2 on Regular Mesoporous Silica, Langmuir, 14 (1998)

3079.

79 Kageyama K., Tamazawa J.I., Aida T., Extrusion Polymerization: Catalyzed

Synthesis of Crystalline Linear Polyethylene Nanofibers within a Mesoporous

Silica, Science, 285 (1999) 2113.

Kleitz F., Marlow F., Stucky G.D., Schuth F., Mesoporous Silica Fibers: Synthesis,

Internal Structure and Growth Kinetics, Chem. Mater., 13 (2001) 3587.

Kresge C.T., Leonowicz M.E., Roth W.J., Vartuli J.C., Beck J.S., Ordered

Mesoporous Molecular-Sieves Synthesized by a Liquid-Crystal Template

Mechanism, Nature, 359 (1992) 710.

Leontidis E., Hofmeister Anion Effects on Surfactant Self-Assembly and Formation

of Mesoporous Solids, Curr. Opin. Colloid Interface Sci., 7 (2002) 81.

Lin, H. P., Kao C.P., Mou C.Y., Liu S.B., Counterion Effect in Acid Synthesis of

Mesoporous Silica Materials, J. Phys. Chem. B., 104 (2000) 7885.

Marlow F., Kleitz F., Mesoporous Silica Fibers: Internal Structure and Formation,

Micropor. Mesopor. Mater., 671 (2001) 44.

Marlow F., Zhao D., Stucky G.D., Doped Mesoporous Silica Fibers: The Internal

Structure, Microp. Mesop. Mater., 39 (2000) 37.

Marlow, F., Leike I., Weidenthaler C., Lehmann C.W., Wilczok U., Mesostructured

Silica Fibers: Ring Structures in Reciprocal Space, Adv. Mater., 13 (2001) 307.

Ozin G.A., Yang H., Sokolov I., Coombs N., Shell Mimetics, Adv. Mater., 9 (1997)

662.

Park S.S., Lee C.H., Cheon J.H., Park D.H., Formation Mechanism of PMO with

rope- and gyroid-based morphologies via close packing of secondary building

units, J. Mater. Chem., 11 (2001) 3397.

Schacht S., Huo Q., Voigt-Martin I.G., Stucky G.D., Oil-Water Interface Templating

of Mesoporous Macroscale Structures, Science, 273 (1996) 768.

80 Schmidt-Winkle P., Yang P., Margolese D.I., Chmelka B.F., Stcuky G.D., Fluoride-

Induced Hierarchical Ordering of Mesoporous Silica in Aqueous Acid-Syntheses,

Adv. Mater., 11 (1999) 303.

Sprokel G.J., The physics and chemistry of liquid crystal devices, Plenum Press: New

York, 1980.

Tudor J., O’Hare D., Stereospecific Propene Polymerisation Catalysis using an

Organometallic Modified Mesoporous Silicate, Chem. Commun., (1997) 603.

Yang H., Kuperman A., Coombs N., Mamiche-Afara S., Ozin G.A., Synthesis of

Oriented Films of Mesoporous Silica on Mica, Nature, 379 (1996) 703.

Yang H., Ozin G.A., Kresge C., The Role of Defects in the Formation of Mesoporous

Silica Fibers, Films and Curved Shapes, Adv. Mater. , 10 (1998) 883.

Van Looveren L.K., Geysen D.F., Vercruysse K.A., Wouters B.H., Grobet P.J., and

Jacobs P.A., Methylaluminoxane MCM-41 as Support in the Co-oligomerization

of Ethylene with Propene with [C2H4(1-Ind)2]Zr(CH3)2, Angew. Chem. Int. Ed.,

37 (1998) 517.

Ko Y.S., Woo S.I., Copolymerization of Ethylene and α-Olefin using Et(Ind)2ZrCl2

Entrapped inside the Regular and Small Pores of MCM-41, Macromol. Chem.

Phys., 202 (2001) 739.

81 CHAPTER 3

GAS DIFFUSION AND MICROSTRUCTURAL PROPERTIES OF

ORDERED MESOPOROUS SILICA FIBERS

3.1. Introduction

Synthesis and characterization of the hexagonally ordered mesoporous silica fibers

(MSF) by the quiescent interfacial growth has been covered in details in Chapter 2. As outlined before, the original report on the preparation of silica fibers [Huo et al, 97] states that the pores align parallel to the fiber axis as suggested for other types of silica-based fibers [Yang et al. 98, Han et al. 00, Miyata and Kuroda 01]. However, subsequent studies on silica fibers prepared with tetrabutylorthosilicate (TBOS) revealed a circular arrangement of the pores according to transmission electron microscopy (TEM) investigations [Marlow et al. 00, Marlow et al. 01]. The hexagonal array of channels in

TEM overviews appear to be perpendicular to the fiber axis on both fiber edges suggesting that the channels have a helical structure with very small pitch angle of the helix. The helical pore orientation was found to be a common structure for silica fibers prepared by this approach based on various types of silica sources [Kleitz et al. 01].

No study has been reported so far on the measurement of the inner microscopic dimensions of the helical channels such as the pore length, pitch size and tortuosity. This could be useful in a number of practical applications like modulated channel-based nano- reactors and polymer molecular extrusion where channel length can affects the yield, conversion, and structure of the product. Representative measurement of the pore length and pitch size seems to be impossible with direct techniques like TEM due to sensitivity

82 of results to imaging location and sampling conditions. Besides, the fiber thickness would

not allow viewing the helical pores across the length direction perpendicular to the fiber

axis. This work presents an indirect measurement of the pore internal microstructural

dimensions by utilizing gas diffusion parameters of MSF and a suitable diffusion model.

In addition to the microscopic dimensions of the pores, this study will also provide

information about the diffusion properties and mechanism of gas transport through the

1D ordered mesoporous materials.

Most studies on the ordered nanoporous materials have focused on the synthesis and microstructure of this group of materials. Equilibrium sorption of gases and vapors

on ordered nanoporous silicates such as MCM-41 has also been extensively studied for

material characterization or exploring the use of this group of materials as sorbents

[Selvam et al. 01]. However, essentially no studies were reported on gas diffusion in the ordered nanoporous materials. Available investigations on gas transport in these materials are limited to steady state gas permeance of 3-D pore MCM-48 prepared as supported membranes to study their quality [Nishiyama et al. 01, McCool et al. 03, Xomeritakis et al. 03]. Gas permeance of 1-D ordered mesoporous silica (like MCM-41 and MSF) membranes has not been studied yet because no body was able to orient the pores preferentially normal to the support surface.

Unsteady state techniques were also used to measure diffusion properties of porous solids of powder and membrane geometries [Karger and Ruthven 92]. Transient weight uptake is one example of this approach in which the rate of change in sample weight due to gas adsorption is measured and used to evaluate diffusion properties. Up to our knowledge, no one has measured the transient uptake properties in ordered mesoporous

83 materials probably due to the anticipation of fast gas diffusion rate in nanoporous particles. The diffusion time constant (L2/D) for 5 µm MCM-41 particles (assuming

Knudsen diffusion, with D~10-2 cm2/s) could be several orders of magnitude smaller than that for 5 µm zeolite silicalite crystallites (zeolitic diffusion, D ~ 10-8 cm2/s). This makes it difficult to experimentally measure gas diffusion in the ordered mesoporous particles unless the diffusion study involves large gas molecules [Berenguer-Murcia et al. 03] or much larger ordered mesoporous particles are available.

Gas transport through ordered mesoporous materials is governed by Knudsen diffusion mechanism with negligible contribution from viscous flow. If the pore surface tends to adsorb the gas molecules, then additional flux contribution comes from diffusion of the adsorbed phase on the surface. Surface diffusion contributes significantly to transport in microporous and mesoporous and in some cases it accounts for over 50% of the total flow rate [Choi et al. 01]. Gas transport mechanisms were identified for various porous mesoporous materials [Karger and Ruthven 92], however, little attention was given to study the significance of the structural order of the pores on the transport mechanisms and their contributions. Such knowledge will guide to significant improvement in the design and applications of these materials as catalysts, membranes and sorbents in unprecedented ways.

In this chapter, the transient gas weight uptake of mesoporous silica fibers, prepared in Chapter 2, will be measured by the gravimetric method. Such measurement will be possible because the large size of the fibers will introduce high resistance to transport and response time large enough for experimental measurement. The uptake data will be treated with a suitable diffusion model to study the gas diffusion properties in the

84 ordered structure and its effect on the relative contribution of each diffusion mechanism.

Gas diffusion properties will be also used to elucidate the inner microstructural dimensions of the one-dimensional pores.

3.2. Experimental

MSF was prepared using tetrabutylortyhosilicate (TBOS) (Aldrich) as silica source and cetyltrimethylammonium bromide (CTAB) (Aldrich) as structure directing agent according to the same procedure described in Chapter 2 under conditions of sample 8.

The morphology was studied using SEM (Hitachi-S4000). Powder X-ray diffraction

(XRD) patterns were obtained on Siemens D-50 using CuKα radiation. XRD spectra were taken between 2θ of 1.5−8 and used to identify the phase structure. Nitrogen adsorption-desorption measurements were performed on Micromeritics Tristar to obtain pore surface area, volume, and size properties.

Sorption experiments were conducted gravimetrically on a Cahn microbalance system (CAHN C-1000) as shown in Figure 3-1 [Chandak et al. 97]. One arm of the microbalance had an aluminum pan suspended at its end to hold the sample. The pan is attached to the microbalance arm by a light platinum wire (Gauge 36, Fisher Scientific).

The pan was maintained at the desired temperature with an Omega CN 76000 temperature controller connected to a tubular furnace mounted on the outside of a 2 in i.d.

Pyrex balance tube. The furnace was mounted so that the sample pan was approximately at its center. The temperature was measured by a thermocouple (K-type Omega KMTSS-

032G-12). The gas delivery system consists of a helium purge gas and several sorption gases of various molecular weights (CO2, C2H4, N2, H2, O2 and CH4). The purge and

85 sorption gases were alternatively delivered to the system by switching a crossover valve

(Swagelock). The transient and equilibrium weight changes were recorded using a computer-aided data acquisition system.

Microbalance

To fume hood

Furnace

Thermocouple

Temp. Data Acquisition Controller

Sample Pan Mass Flow Controller

Crossover valve

He Sorption gas On-off valve Purge CO2, C2H4

Figure 3-1. Schematic of Cahn C-1000 microbalance system apparatus for

measurement of gas sorption-diffusion in mesoporous silica fibers.

In experiments, 200-300 mg sample of MSF was loaded to a light weight (200 mg)

Aluminum pan and degassed at 120 oC for few hours under helium purge flowing at 100

ml/min. After sample weight becomes constant, the temperature was cooled down to the

desired sorption temperature (room temperature). Then adsorption process was started by

switching from the purge gas to the sorption gas at 100 ml/min using the crossover valve.

After adsorption process reached equilibrium, gas desorption was conducted at the same

86 temperature by switching back to purge gas at 100 ml/min. The 100 ml/min flow rate is

an optimum value where systematic problems of gas dispersion (caused by switching

between gases) and weight instability (taking place at high flow rates) are minimized.

Data were reduced to get uptake (mg gas adsorbed/g adsorbent or mmol/g) and fractional

uptake rate (uptake/equilibrium uptake) curves. Gas adsorption isotherms were measured

at room temperature by adsorption of the gas from a binary mixture with helium at partial

pressures between 0 to 1 atm.

3.3. Results and Discussion

3.3.1. Characteristics of mesoporous silica fibers

Figure 3-2 shows the XRD pattern of MSF product with two reflection peaks

indexed as 100 and 110 typical to hexagonal pore arrangement. Surface area properties

measured by nitrogen adsorption-desorption indicate that MSF has a BJH pore diameter

of 2.74 nm, BET surface area of 799 m2/g, and pore volume of 0.62 cm3/g. From the pore

volume and silica density (2.15 g/ml) the porosity was calculated as 0.571

(=0.62/(0.62+2.15-1)). MSF prepared under the given conditions consists mainly of

fibrous form. The fibers are almost 15 µm in diameter and have a length distribution varying between 50 to 500 µm. This distribution can be due to fiber collection process which breaks the fibers in the length direction or to some growth-related mechanism that affects the growth of fibers in the length direction. The fiber length distribution was obtained from SEM micrographs by measuring the length of a large number of fibers.

The normalized length distribution E(L) is shown in Figure 3-2. The average fiber length

87 was calculated from E(L) as 196 µm and was subsequently used with the gas diffusion properties to characterize the pore structure of MSF.

0.004

0.003

0.002

0.001 Normalized Length Distribution E(L) Distribution Length Normalized 0 0 200 400 600 800 Fiber Length (µm)

Figure 3-2. Normalized fiber length distribution

88 3.3.2. Gas uptake on silica fiber

Kinetic gas uptake measurements on MSF were performed with gases of various

molecular weights. Due to the nature of physical adsorption of gases on MSF, no

adsorption was observed at temperatures over 60 oC for all gases. Therefore, all

experiments were performed at room temperature to get uptake values with reasonable

accuracy. Experiments show that N2, H2, O2 and CH4 gases are non-adsorbing on MSF

and exhibit small uptakes within the error limit of the microbalance. CO2 and C2H4 exhibited better uptake capacities; therefore sorption experiments were focused on these two gases.

o Absolute and fractional rate of weight uptake of CO2 and C2H4 on MSF at 28 C

and 1 atm are shown in Figure 3-3, where the fractional uptake is the ratio of the uptake

at time t (Mt) to the equilibrium uptake (M∞). Uptake capacities of CO2 and C2H4 are respectively 0.68 and 0.78 mmol/g which are equivalent to 29.92 and 21.84 mg/g. Gas uptake at these conditions is far below the monolayer coverage of MSF. The percentage of monolayer coverage is estimated as 4.6 and 7.0 % for CO2 and C2H4 using the

equation Xmonolayer= q×Ao×NA/SA where q (mol/g) is the uptake capacity, SA is the BET

2 surface area (799 m /g), Ao is the sorbate cross sectional area estimated from the

2 2 molecular kinetic diameter assuming a spherical shape (9 Å for CO2 and 12 Å for C2H4)

23 and NA is the Avogadro’s number (6.02×10 molecule/mol).

The fractional uptake of C2H4 is faster than CO2 indicating a slightly larger

diffusion coefficient. Kinetic sorption curves of these gases in MSF consist of a steep

uptake step up to 90% in less than 3 min followed by a slowly increasing uptake

achieving M∞ plateau within 10 min. Contrary to the anticipated fast gas diffusion rate in

89 0.8

0.6 (A)

0.4 Uptake (mmol/g) Uptake 0.2 C2H4 CO2

0 0 100 200 300 400 500 time (s)

1

0.8 (B) 0.6

C2H4 Exp 0.4 C2H4 Fitting Fractional Uptake Uptake Fractional 0.2 CO2-Exp CO2-Fitting 0 0 100 200 300 400 500 time (sec)

Figure 3-3. Absolute (A) and fractional (B) uptake of CO2 and C2H4 on mesoporous silica fibers. Lines represent model fitting on experimental data

90 the ordered mesoporous materials (governed by Knudsen type diffusion), the response

time in MSF was slow enough to be experimentally measured with acceptable accuracy.

This is probably due to high resistance to gas diffusion induced by the long straight pores

aligned within the fibers.

The transient weight uptake data can be utilized to investigate the gas diffusion

properties and to elucidate the fiber microstructural properties. Microscopic

investigations revealed that MSF consists of fibers with average length of 196 µm. The

fibers contain huge number of one dimensional nano-sized cylindrical pores aligning helically around fiber axis [Marlow et al. 00, Marlow et al. 01]. Diffusion of gases in the pores can be considered as one dimensional bounded by the two sides of the fiber. Such diffusion can be appropriately described by the plain sheet model of uni-dimensional diffusion [Crank 75] shown in Eq. (3-1), where Deff is the effective diffusion coefficient

and L is the fiber length.

M ∞  8  − (2n +1) 2 π 2 D  t = 1− × exp eff t  (3-1) ∑  2 2  2  M ∞ n=0 (2n +1) π  4 L 

2 The time constant (Deff/L ) was obtained by regression of the fractional uptake data

with Eq. (3-1) using the least square method on Mathematica software. Deff can be subsequently calculated and used to analyze the diffusion properties in ordered mesoporous fibers. Uptake curves of CO2 and C2H4 were calculated from Eq. (3-1) and plotted in Figure 3-3B (solid lines) with the experimental data. Figure 3-3B shows that the diffusion model deviates slightly from the experimental data due to one of two possible reasons: (1) presence of distribution in the fiber length (L) which affects the pore length, or (2) gas dispersion problem upon switching between the purge and sorbate gas.

91 If it is due to the first reason, then the diffusion model can be modified to account for the distribution in fiber length as follows:

Y (D ,t) = Y(L,D ,t)× E(L)dL≈ Y(L, D ,t)× E(L)× ∆L (3-2) ave eff ∫ eff ∑ eff

where Y is the fractional uptake and E(L) is the normalized fiber length distribution. It should be noticed that E(L) represents the fiber length and not the pore length. E(L) and

∆L data were obtained from Figure 3-2 and used with the fractional gas uptake data through Eq. (3-2) to estimate the length-corrected fractional uptake (Yave). After regression with experimental data it was noticed that the modified diffusion model deviates similarly to the unmodified model from the data as shown in Figure 3-4. This indicates that the main reason for deviation is the gas dispersion.

Gas dispersion is a major experimental problem that appears in systems for measuring the kinetic change in properties. In our system, it is caused by non-ideal flow of the sorbate gas through the purge gas prior to sorption into the sample. The non-ideal flow involves recycling of the gas and existence of stagnant regions of gas pockets in the system tubing. These factors increase the residence time of the gas inside in the tubes and reduce the effective concentration of sorbate gas upon reaching the sample. Therefore, when introducing the step input of the sorbate gas, it takes longer time to reach equilibrium pressure (1 atm) and causes slight delay in the weight uptake process. This is the main reason for deviation of the experimental data from the plain sheet diffusion model where, ideally, uptake curve should be steep at the initial stage and not slightly curved (see Fig. 3-3B).

92 1.00

0.80 2 Def f (cm /s) 2.00×10-6 0.60 1.57×10-6 1.00×10-6 0.40 CO2-Exp Fractional Uptake 0.20 Uncorrected model

Corrected model 0.00 0 100 200 300 400 500 time (sec)

Figure 3-4. Comparison of data fitting between the fiber length-

corrected diffusion model and the uncorrected model.

The effect of dispersion problem can be reduced by increasing the flow rate (e.g., to

400 ml/min) to minimize gas residence time inside the tubes. However, high flow rates have significantly affected the weight stability and data accuracy. For more accurate data, the delay time can be reduced by performing the uptake experiments at moderate flow rate (100 ml/min) and small incremental pressure steps from 0 to 1 atm. Due to the nature of fast uptake process of gases on the ordered pores of MSF, it was not possible to avoid the problem of gas dispersion without involving of additional experimental error.

Therefore, the slight deviation of model from experimental data was considered reasonably acceptable and these data were used for subsequent microstructural analysis of MSF. From Figure 3-3B, best fitting of the diffusion model to experimental data was

2 -3 -1 -3 -1 obtained with (Deff/L ) values of 4.09×10 s for CO2 and 5.71×10 s for C2H4 from

93 -6 -6 2 which Deff can be calculated respectively as 1.57×10 and 2.19x10 cm /s (where L is the average fiber length of 196 µm).

3.3.3. Evaluation of transport diffusivities

Single gas diffusion through porous materials can be generally described by Ficks law Ji=Deff×dCi/dl where Ji is the molar flux, dCi/dl is the concentration gradient and Deff is the effective diffusivity of the component i. Deff represents a combination of all mechanisms of transport which can be in the gaseous phase of the pore by viscous or

Knudsen diffusion, or on the pore surface by surface diffusion. Results of gas uptake show that MSF has a tendency to physically adsorb CO2 and C2H4 molecules with low concentrations. The adsorbed molecular phase can exhibit mobility on the pore surface by hopping between the adsorption sites under the effect of surface concentration gradient.

This mobility can considerably increase the net flux by an additional surface diffusion term. C2H4 is expected to have more surface diffusion flow than CO2 because it has higher uptake capacity on MSF.

The viscous flow in the 3 nm pores of MSF can be neglected; therefore, the gas phase diffusion in MSF pores is mainly caused by Knudsen mechanism. Ratio of effective pure gas diffusivities can be used to roughly identify the presence of additional surface diffusion contribution. The ideal diffusivity ratio of C2H4 to CO2 based on the

Knudsen flow is equal to the square root of the molecular weight of CO2 to C2H4

0.5 =(44/28) =1.25. The real diffusivity ratio of C2H4 to CO2 is 1.39 which is higher than the Knudsen-based ratio. This confirms that CO2 and C2H4 diffuse by a combined

Knudsen and surface flow in the ordered pores of MSF. The surface contribution of non-

94 adsorbing gases will be negligible. Since both mechanisms take place in parallel, the total effective diffusion can be represented as a linear combination of the two mechanisms as expressed in Eq. (3-3) [Karger and Ruthven 92] under ideal gas phase and equilibrium adsorption-desorption isotherm conditions. DK, eff and DS, eff are the effective Knudsen and surface diffusivities respectively, ε is the porosity of MSF (0.571) and K is the dimensionless adsorption equilibrium constant.

1− ε D = D + K D (3-3) eff K ,eff ε S ,eff

Direct calculation of DK, eff and DS, eff for each gas from one uptake curve at room

temperature is impossible. One strategy is to measure the diffusion of a non-adsorbing

calibrating gas for which the diffusion is governed by Knudsen mechanism (DK) then to

measure the combined surface and Knudsen diffusion for a desired gas. The Knudsen

0.5 contribution of the gas can be found with the relation: DKi=DKnon-ads×(Mnon-ads/Mi) then

the surface contribution can be evaluated from the linear combination of diffusion

coefficients using Eq. (3-3). Another way is to run the uptake experiment at high

temperature (e.g., 200 ºC) and obtain Deff which will be equivalent to DK,eff because

contribution of surface diffusion becomes negligible (with very small K value). Both

cases, however, could not be performed on MSF because of the low uptake capacity of

the non adsorbing gases (H2 and N2) and the negligible sorption uptakes at high

temperature [Rivarola and Smith 64].

In this work, an alternative analytical approach was used to obtain the diffusion

coefficients of gases in ordered MSF. Knudsen and surface diffusion contributions were

predicted utilizing a simple model based on theoretical diffusion relations,

2 experimentally-obtained time constant (Deff/L ) of Eq. (3-1) and adsorption equilibrium

95 constant (K) of adsorbing gases at room temperature. Knudsen [Karger and Ruthven 92] and surface [Hwang and Kammermeyer 66] diffusivities can be expressed as:

ε 8RT D = d (3-4) K ,eff κ p 9π M

BT T exp(−E / RT) D = (3-5) S ,eff M where κ, ε, and dp are respectively the tortuosity factor, porosity and pore diameter of

MSF. T and M are the temperature and molecular weight of diffusing gas. E is the surface diffusion activation energy and B is constant. Both diffusivities have similar dependency on the square root of the gas molecular weight. It should be pointed out that the tortuosity factor (κ) and the tortuosity (τ) are different values. They are wrongly used as interchangeable values in literature [Epstein]. Tortuosity factor (κ) is the square of tortusosity which is defined as the ratio of diffusion length to macroscopic length, i.e.,

κ= τ2= (diffusion length/macroscopic length)2. Substitution of Eqs. (3-4) and (3-5) in (3-

3) and simplification gives the following form:

D eff M = α + β K (3-6) L2 with the temperature constants α and β given by:

ε d 8RT α = p (3-7) κ L2 9π

(1− ε ) BT T exp(−E / RT) β = (3-8) ε L2

96 Dealing with Eq. (3-6) is much easier than with Eq. (3-3). DK and DS coefficients vary with the gas used and the experimental conditions. Obtaining these values for two gases using Eq. (3-3) is impossible because there will be two equations (yi=DKi+DSi xi,

2 where yi=Deff i /L and xi=Ki (1-ε)/ε for each gas) and four unknowns (DK1, DS1, DK2 and

DS2). Equation (3-6) is an extension of Eq. (3-3) but the constants, α and β, will be the same for any gas used. Therefore, these constants can be evaluated easily for two gases

0.5 2 using two equations (yi=α + βxi, where yi= Deff i (Mi) /L and xi=Ki for each gas) and two unknowns (α and β) and then be used to evaluate DK and DS with equations (3-4 and 3-7) and (3-5 and 3-8) respectively. Moreover, using Eq. (3-7), the tortuosity factor

(κ) can be evaluated and used to analyze the pore structure of the ordered silica fibers such as pore length and orientation.

Figure 3-5 shows the adsorption isotherms of CO2 and C2H4 on MSF measured at room temperature and 0 to 1 atm partial pressures. Gas adsorption capacity of MSF at room temperature is close to 25 mg/g which is almost 5-7% of the monolayer capacity based on its surface area. This value is consistent with the gas adsorption capacity of the unmodified ordered mcm-41 silica [Xu et al. 03]. C2H4 has higher molar adsorption capacity than CO2 and this implies a higher surface flow contribution. The isotherms are reasonably linear with low uptake capacities low below the monolayer surface coverage so K can be evaluated directly from the slope. The slope has a unit of (mg gas/g sorbent.atm), while K is dimensionless and is normally expressed in terms of total volume of the porous solid (moles adsorbed per unit solid volume/moles per unit volume in gas phase). Therefore, K was evaluated from the slope using Eq. (3-9):

97 R T ρ K = slope(g / g.atm)× solid (3-9) M where R is the ideal gas constant (82.1 atm.ml/mol.K), T is temperature (300 K), ρsolid is the silica density (2.15 g/ml) and M is the gas molecular weight (g/mol). Adsorption

-3 -3 isotherms have slopes of 26.54×10 and 22.03×10 g/g.atm respectively for CO2 and

C2H4 from which K were calculated as 31.94 and 41.66.

30

CO2 25 C2H4

20

15

10

Equilibrium UptakeEquilibrium (mg/g) 5

0 0 0.2 0.4 0.6 0.8 1 Gas Partial Pressure (atm)

Figure 3-5. Equilibrium adsorption isotherms of CO2 and C2H4 on silica fibers

at 28 ºC and partial pressure 0-1 atm.

98 Table 3-1 Experimental parameters of diffusing gases in mesoporous silica fibers

2 2 0.5 Gas M K Deff/L (Deff/L )(M)

(mol/g) (dimensionless) (s-1)

-3 CO2 44 31.94 4.09×10 0.027

-3 C2H4 28 41.66 5.71×10 0.030

Table 3-1 summarizes the experimental values that were used to estimate the

constant parameters of Eq. (3-6). The constants α and β were estimated respectively as

1.7×10-2 and 3.17×10-4 (g/mol)0.5/s. The intercept (α) is a temperature constant value that

is related to the Knudsen diffusivity while the slope (β) is related to the surface

diffusivity. The effective Knudsen diffusivity (DK, eff) was calculated from Eqs. (3-4)

and (3.7) then the surface contribution (K.DS, eff) was evaluated by subtracting the

Knudsen contribution from the total effective diffusion coefficient using Eq. (3-3).

Diffusivities and contribution of diffusion mechanisms are summarized in Table 3-2.

Table 3-2 Knudsen and surface diffusivities and contribution to net gas diffusion in MSF

1−ε p ( ) K.DS Gas Deff DK, eff ε p DS, eff 2 2 2 (cm /s) (cm /s) (cm2/s) (cm /s) (%) (%)

-7 -7 -7 -8 CO2 15.70×10 9.85×10 5.87×10 2.48×10

(63%) (37%)

-7 -7 -7 -8 C2H4 21.90×10 12.30×10 9.59×10 3.07×10

(56%) (44%)

99 According to the gas uptake results and data fitting with the plane sheet model, the overall effective diffusivity of gases under study is in the range of (15−22)×10-7 cm2/s.

Gas phase diffusion by Knudsen mechanism contributes to almost 60% of the transport in the straight pores of mesoporous silica fibers, while the remaining is contributed by the surface mechanism. The slight difference in the diffusion contributions observed by

CO2 and C2H4 gases is accounted to the difference in their diffusion and adsorption behaviors. According to Eqs. (3-4) and (3-5), C2H4 will have a higher Knudsen and surface diffusivities than CO2 because it has smaller molecular weight. Moreover, it exhibited higher molar adsorption capacity and adsorption equilibrium constant (K) and therefore had a higher diffusivity than CO2. A non-adsorbing gas like N2 or He, on the other hand, will have a negligible surface diffusion contribution and the net diffusion will be solely governed by Knudsen mechanism. It is worth mentioning, however, that diffusion properties of such gases could not be evaluated by the gravimetric approach as a result of their negligible weight uptakes.

Sholl and coworkers reported exceptionally high transport diffusivities of light gases in microporous single walled carbon nanotubes (SWCN) up to 10 cm2/s using atomistic simulation [Skoulidas et al. 02]. These values are at least two orders of magnitude larger than those exhibited by silicalite and ZSM-12 microporous zeolites of similar pore sizes. The fast diffusivities were attributed to the surface smoothness that can strongly controls the solid-sorbate interaction potentials. Given the similarity in pore dimensionality (1D), pore length and pore curvature features of the MSF pores and the carbon nanotubes; we can assume that the relatively high contribution of surface flow in

MSF can be attributed to the smoothness of the pore surface.

100 Diffusivities given in Table 3-2 are the effective values. They involve the effect of pore geometry parameters of the solid (e.g., porosity, pore length, connectivity, orientation, size distribution, etc) which contribute to reducing the diffusion by lengthening the diffusion path and reducing the concentration gradient. The real diffusion behavior is normally expressed in terms of diffusivity in straight cylindrical pore of equivalent diameter. This is called the real diffusivity and it is dependent on the properties of molecules, such as their velocity in the fluid phase, binding energy and their mobility on the surface. The net effective diffusivity is expressed in terms of the real gas (Knudsen) and surface diffusivities and pore geometry by [Zalc et al. 03]:

ε (1− ε )K DS ,real Deff = DK ,real + (3-10) κ g ε κ s

From Eq. (3-10) we notice that DK, eff=(ε/κg)×DK, real and DS, eff=(1/κs) ×DS, real where κg and κs are the gas (void) and surface tortuosity factors respectively. Tortuosity is a lumped structural factor defined as the length of the equivalent capillary required to describe the effective diffusivity in the void or surface phases. For the case of MSF, κg was estimated from Eq. (3-7) as 2010 which reflects a highly tortuous structure. The value of κs could not be evaluated using Eq. (3-5) because the constants B and E are unknown. κg and κs are not equivalent because the distance traveled by molecules in the pore gas phase is different from that on the pore surface especially with systems of irregular pore structure. Given that MSF has a regular cylindrical pore structure with uniform pore size distribution in the low limit of the mesoporous size (2.78 nm), the difference in diffusion path between gas phase and surface is not significant. Therefore, we can reasonably assume that κg and κs of MSF are equal.

101 Real Knudsen and surface diffusivities were accordingly calculated using Eq. (3-

10) and summarized in Table 3-3. Real Knudsen diffusivity of ordered MSF is in the range of ~ 10-3 cm2/s which is consistent with values of mesoporous materials. For most conventional mesoporous materials, the effective diffusivity is typically 1 order of magnitude smaller than the real diffusivity, given a typical porosity and tortuosity values of 0.3 and 3 respectively [Karger and Ruthven 92]. In case of MSF, the porosity is typical to other mesoporous materials, but the large tortuosity factor (2010) is the main reason for the significant reduction in effective transport properties by 3 to 4 orders of magnitude. Surface diffusivity values are low in the range of 10-5 cm2/s indicating that surface transport is an activated process which requires high activation energy

(Eact=R×T×ln(Do/DS)). The value of Eact could not be directly evaluated from weight uptake at various temperatures due to errors involved in measurement of low uptake capacities at temperatures over 30−40 ºC. Do value for the diffusion of CO2 in

-2 2 mesoporous Vycor glass (> 90% SiO2) is 1.6x10 cm /s [Gilliland et al. 74]. Assuming a similar Do value for gas diffusion in MSF (SiO2), then Eact for CO2 surface diffusion at room temperature can be estimated as 14.4 kJ.mol.

Table 3-3. Effective and real gas diffusion coefficients in MSF

2 2 DK (cm /s) DS (cm /s) Gas Effective Real Effective Real

-7 -3 -8 -5 CO2 9.85×10 3.47×10 2.48×10 5.00×10

-7 -3 -8 -5 C2H4 12.30×10 4.33 ×10 3.07×10 6.17×10

102 Table 3-4. Real Surface Diffusivities of CO2 in ordered mesoporous silica fibers versus

other non-ordered porous solids

Sorbent Temp. Measurement Surface flow DS Ref.

(Pore Size) (ºC) method contribution (cm2/s)

MSF 28 Unsteady State 37 % 5.0×10-5 This (2.7 nm) gravimetric work uptake

γ-alumina 20 Steady state 30 % 2.0−5.0×10-5 Kiezer et

(2.70 nm) Wicke- al. 88 Kallenbach

Boehmite 25 Steady state 54 % 9.2×10-5 Rivarola (Bimodal) a Wicke- and Kallenbach Smith 64

Vycor -50 Steady State 53 % 6.0×10-5 Gilliland Glass Permeability et al. 74

(4.6 nm)

a Bimodal pore size distribution of 4 and 60 nm

Although gas diffusion in mesoporous materials has been extensively studies over the past decades, only few papers have focused on CO2 and C2H4 diffusion at room temperature conditions. Table 3-4 gives a summary of the major CO2 real surface diffusivities reported on common non-ordered mesoporous alumina and Vycor glass in comparison with results on ordered MSF. In contrast to the hexagonally-ordered cylindrical pores of MSF, γ-alumina has irregular slit shaped pores [Lin 93], while Vycor glass has 3-D irregular pore structure made by random packing of spheroidal glass particles and [Levitz et al. 91]. As shown in Table 3-4, surface diffusivity of ordered

103 MSF and non-ordered materials are in the same order of magnitude (10-5 cm2/s) with values differing from MSF by a factor of 0.5 to 2. Surface diffusivity of Vycor glass was measured at -50 ºC at which adsorption affinity of CO2 is high. This increases the surface concentration of CO2 (driving force) and leads to higher surface flow. Using Arrhenuis relation, surface diffusivity of Vycor glass at room temperature is expected to be at least

30×10-5 cm2/s.

Diffusion results discussed in the previous paragraph suggest that the regular pore structure does not seem to have a major effect on the surface diffusion properties. To have rational conclusions, diffusion coefficients must be obtained or at least treated on the same experimental basis. DS in MSF was measured by unsteady method where the change in concentration is measured with time, while those reported on alumina and

Vycor glass were all measured by the steady state approach where the rate of molecular diffusion through the sample membrane due to a pressure gradient is measured. Reyes and coworkers [Reyes et al. 00] discussed the discrepancies in diffusion parameters obtained by steady and unsteady state techniques in terms of the adsorption capacity factor (ACF)= 1+K×Sv, where K is the adsorption equilibrium constant (at sufficiently low concentration) and Sv is the average surface to volume ratio (Sv=4/pore diameter for cylindrical pores). The effective diffusion coefficient obtained in unsteady state measurement differs from that obtained in a steady state experiment by the ACF factor

(i.e., Dunstaedy=Dsteady/ACF). For a non-adsorbing gas, for example, the ACF value will reduce to unity (since K=0) and give similar steady and unsteady diffusivities. For other situations, the ACF must be taken into account to give adequate comparisons between diffusion coefficients measured by these techniques.

104 The ACF value for CO2 diffusion MSF was calculated as ~ 30. Using the equation

Dunstaedy=Dsteady/ACF, the effective diffusivity of CO2 in MSF will be 3 fold higher if measured by a steady state approach. According to this argument, back calculation of surface diffusivity using Eqs. (3-6) and (3-10), will give a real surface diffusivity of

1.5×10-3 cm2/s by steady state compared to 5×10-5 cm2/s by unsteady state approach. If this manipulation is valid, then the surface diffusivity in ordered MSF is two orders of magnitude higher than diffusion in non-ordered mesoporous materials. Then this can strongly support that the ordered structure can significantly enhance the diffusion properties possibly due to lower transport resistances associated with the regular packing of the pores. To confirm the aforementioned argument, diffusion properties should be measured under steady state condition by membrane permeation method. This will allow the measurement of diffusion coefficients for adsorbing and non-adsorbing gases at wide range of temperature.

3.3.4. Fiber internal microstructure

The gas phase tortuosity factor (κg) of MSF was calculated in the previous section from Eq. (3-6) as 2010 and was assumed to be equal to the surface tortuosity (κs). This value is very high compared to typical tortuosities of porous solids [Karger and Ruthven

92]. If accurately measured, tortuosity factor can be used to provide more information about internal microstructure of the porous material. It is generally given by the ratio

2 (lp/L) [Epstein 89] where lp is the length of tortuous diffusion path and L is the mean thickness of diffusion path. This ratio defines the commonly used tortuosity (τ) value.

MSF has an internal structure composed of cylindrical pores that align helically across

105 the fiber axis as confirmed in literature by TEM images. In this case, lp will be the length of the helical pores inside the fiber and L is the average length of fibers (196 µm). From

κ= 2010, the ratio lp/L (=τ) was found to be 45. This means that the pores are 45 times longer than the fiber macroscopic length. From this ratio we can imagine that a 196 µm long fiber will contain pores of ~ 9 mm length. Given that the pores are cylindrical and non-connected as revealed by TEM images, and are much longer than the fiber length, the only possible orientation of these pores is to whirl helically across the fiber as schematically illustrated in Figure 3-6a.

Although the helical orientation was previously reported by other investigators using direct imaging techniques, this study confirms the helical pore structure and provides information on the microscopic dimensions of these pores. The degree of orientation or pitch length (z) can be estimated from the fiber diameter (df =15 µm), fiber length (L) and pore length (lp) using the equation lp=π×df×L/z (where lp/L =45). This corresponds to a pitch value of 1.05 µm. The degree of pore orientation and pitch measurement is illustrated in Figure 3-6b. Such measurements are very hard if not impossible by direct techniques like TEM images.

It should be noticed that these microscopic dimensions, i.e., pore length and pitch size, represent the average values. In fact, the pores have a distribution in their length. By imagining the hexagonal packing of the cylindrical pores, the formation of the internal structure by cylindrical rod micelles proceeds as follows: the first micelle starts to whirl around an imaginary axis of the fiber, and then additional rod micelles whirl helically around each other with hexagonal packing to grow the fiber in length and diameter. This process continues until it ends up with the final observed macroscopic dimensions of the

106 fiber. The outer pore will therefore be longer than the internal pores. The pitch on the other hand is not expected to change significantly due to the long range order of the hexagonal packing of the pores. The pore microstructure and dimensions of MSF obtained in this study provide new fundamental information about the formation process.

Technically, such information together with the diffusion properties have a significant impact on the design and application of this group of ordered materials in membrane reactors, catalysis and separation to name a few. In reactor application, pore length and diffusion properties can be used to understand the production process and to improve product yield and quality.

107

(a)

D= 15 µm

Pitch = 1.05 µm Lave= 196 µm

(b)

Figure 3-6. Schematic representation of the internal pore structure of silica fibers, (a) helical pore orientation of the hexagonally arranged pores, and (b) approximate internal pore dimensions (length, diameter and pitch) with tortusosity factor of 45.

108 3.4. Conclusions

Gas diffusion and microstructural properties of ordered mesoporous silica fibers were studied for the first time using the transient uptake of various gases at room temperature. Only CO2 and C2H4 gases tend to physically adsorb on MSF with capacities of close to 25 mg/g at room temperature and negligible capacities at higher temperatures.

Adsorption isotherms at room temperature revealed that the surface concentrations of these gases are very low with surface coverage values close to 5% relative to the monolayer capacity. Diffusion in the plain sheet model was used to fit the weight uptake data to obtain the overall effective diffusivities in MSF. Ratio of the effective diffusivities of the two gases was different from Knudsen-based ratio which indicates a presence of surface diffusion contribution to the net flow.

A simple model was derived and utilized to separate the contribution of each mechanism. The surface diffusion mechanism contributes to ~ 45% of the net flow. This contribution would be higher for gases with higher molar adsorption capacities (C2H4>

-3 -5 2 CO2). Real Knudsen and surface diffusivities were in the order of 10 and 10 cm /s respectively. The significantly high tortuosity factor for MSF, calculated as 2010, tends to reduce the effective diffusivities by 3 orders of magnitude. Surface diffusivity of ordered MSF (measured by unsteady state approach) was very close to the values

(measured by steady state techniques) of non-ordered mesoporous materials. This may suggest that the pore order does not have a significant impact on diffusion properties. It is, however, believed that the steady state measurement of surface diffusivity in MSF would lead to value 2 orders of magnitude higher than those of conventional non-ordered mesoporous materials which indirectly emphasize the effect of pore order.

109 The tortuosity factor defined as the square of the pore length to the fiber average length was used to identify the pore length. Pores were found to be 45 times longer than the fiber length. For a 1-D and non-connected pores, this implies that the pores align helically across the fiber axis confirming the literature investigations based on TEM images. Using the fiber macroscopic dimensions and tortusosity factor, it was revealed that the pore whirl helically around the fiber axis with pitch value of 1.05 µm.

110 3.5 Reference

Beck J.S., Vartuli J.C., Roth W.J., Leonowizc M.E., Kresge C.T., Shmitt K.D., Chu C.T.,

Olson D.H., Shepard E.W., McCullen S.B., Higgins J.B., Schlenker J.L., A new

family of mesoporous molecular sieves prepared with liquid crystal templates, J. Am.

Chem. Soc. 114 (1992) 10834.

Berenguer-Murcia A., Fletcher A.J., Garcia-Martinez J., Cazorla-Amoros D., Linares-

Solano A., Thomas K.M., Probe molecules kinetic study of adsorption on MCM-41,

J. Phys. Chem. B., 107 (2003) 1012.

Chandak M.V., Lin Y.S., Ji W., Higgins R.J., Sorption and diffusion of VOCs in DAY

zeolite and silicalite-filled PDMS membranes, J. Membrane Sci., 133 (1997) 231.

Choi J.G., Do D.D., Do H.D., Surface Diffusion of Adsorbed Molecules in Porous

Media: Monolayer, Multilayer, and Capillary Condensation Regimes, Ind. Eng.

Chem. Res., 40 (2001) 4005.

Crank J., The Mathematics of Diffusion, Clarsndon Press, Oxford, UK, 1975, Ch. 4, p.

44.

Epstein N., On Tortuosity and Tortuosity Factor in Flow and Diffusion through Porous

Media, Chem. Eng. Sci., 44 (1989) 777.

Gilliland E.R., Baddour R.F., Perkinson G.P., Sladek K.J., Diffusion on Surfaces I. Effect

of Concentration on the Diffusivity of Physically adsorbed Gases, Ind. Eng. Chem.

Fundam., 13 (1974) 95.

Han Y.J., Kim J.M., Stucky G.D., Preparation of noble metal nanowires using hexagonal

mesoporous silica SBA-15, Chem. Mater., 12 (2000) 2068.

111 Huo Q., Zhao D., Feng J., Weston K., Buratto S.K., Stucky G.D., Schacht S., Schuth F.,

Room temperature growth of mesoporous silica fibers: A new high-surface-area

optical waveguide, Adv. Mater., 9 (1997) 974.

Hwang S.T., Kammermeyer K., Surface diffusion in Microporous Media, Canadian J.

Chem. Eng., (1966) 82.

Karger A., Ruthven D.M., Diffusion in Zeolites and other Microporous Solids, J. Wiley

& Sons. Inc., 1992, Ch. 11, p.342.

Keizer K., Ulhorn R.J.R., Van Vuren R.J., Burggraaf A.J., Gas Separation Mechanism in

Microporous Modified γ-alumina Membranes, J. Membr. Sci., 39 (1988) 285.

Kleitz F., Marlow F., Stucky G.D., Schuth F., Mesoporous silica fibers: Synthesis,

internal structure and growth kinetics, Chem. Mater. 13 (2001) 3587.

Levitz P., Ehret G., Sinha S.K., Drake J.M., Porous Vycor Glass: The Microstructure as

Probed by Electron Microscopy, Direct Energy Transfer, Small-Angle Scattering, and

Molecular Adsorption, J. Chem. Phys., 95 (1991) 6151.

Lin Y.S., A Theoretical Analysis on Pore Size Change of Ceramic Membranes after

Modification, J. Membr. Sci., 9 (1993) 55.

Marlow F., Kleitz F., Mesoporous silica fibers: internal structure and formation, Microp.

Mesop. Mater., 44-45 (2001) 671.

Marlow F., Slpliethof B., Tesche B., Zhao D., The internal architecture of mesoporous

silica fibers, Adv. Mater., 12 (2000) 961.

McCool B.A., Hill N., DiCarlo J., DeSisto W.J., Synthesis and Characterization of

Mesoporous Silica Membranes via Dip-Coating and Hydrothermal Deposition

Techniques, J. Membr. Sci., 218 (2003) 55.

112 Miyata H., Kuroda K., Film to fiber transformation of preferentially aligned

mesostructured silica, Adv. Mater., 13 (2001) 558.

Nishiyama N., Park D.H., Koide A., Egashira Y., Ueyama K., A Mesoporous Silica

(MCM-48) Membrane: Preparation and Characterization, J. Membr. Sci., 182 (2001)

235.

Reyes S.C., Sinfelt J.H., DeMartin G.J., Diffusion in Porous Solids: The Parallel

Contribution of Gas and Surface Diffusion Processes in Pores Extending from the

Mesoporous Region into the Microporous Region, J. Phys. Chem. B, 104 (2000)

5750.

Rivarola J.B., Smith J.M., Surface Diffusion of Carbon Dioxide on Alumina, Ind. Eng.

Chem. Fundam., 3 (1964) 308.

Selvam P., Bhatia S.K., Sonwane C.G., Recent advances in processing and

characterization of periodic mesoporous MCM-41 silicate molecular sieves, Ind. Eng.

Chem. Res., 40 (2001) 3237.

Skoulidas A., Ackerman D., Johnson J.K., Sholl D., Rapid Transport of Gases in Carbon

Nanotubes, Phys. Rev. Lett., 899 (2002) 185901.

Xomeritakis G., Naik S., Braunbarth C.M., Cornelius C.J., Pardey R., Brinker C.J.,

Organic templated Silica Membranes- I. Gas and Vapor Transport Properties, J.

Membr. Sci., 215 (2003) 225.

Xu X.C., Song C.S., Andresen J.M., Miller B.G., Scaroni A.W., Preparation and

Characterization of Novel CO2 Molecular Basket Adsorbents Based on Polymer-

Modified Mesoporous Molecular Sieve MCM-41, Microp. Mesop. Mater., 1-2,

(2003) 29.

113 Yang P., Zhao D., Chmelka B.F., Stucky G.D., Triblock-copolymer-directed syntheses of

large-pore mesoporous silica fibers”, Chem. Mater. 10 (1998) 2033.

Zalc J.M., Reyes S.C., Iglesia E., Monte-Carlo Simulations of Solid and Gas Phase

Diffusion in Complex Porous Structures, Chem. Eng. Sci., 58 (2003) 4605.

114 CHAPTER 4

COUNTER DIFFUSION SELF ASSEMBLY OF ORDERED

NANOPOROUS SILICA MEMBRANES

4.1. Introduction

According to literature review provided in Chapter 1, ordered mesoporous materials offer a variety of attractive microstructural properties that made them the focus of great attention in areas of separation and reaction applications. From technological point of view, it is important to prepare these materials in the form of a membrane.

Ordered mesoporous silica materials were fabricated as membranes on several kinds of supports using various techniques including sol gel dip and spin coating [Ogawa et al.

94, Martin el 97] and growth from solution directly on support [Yang 96, Tolbert 97,

Nishiyama 97]. Depending on the preparation procedure, nanostructured membranes can grow as a distinct thin film on the surface of support or infiltrate and grow as particles inside the support pores.

The self-assembly methods reported so far were able to create nanostructured membranes with ordered pores, however, pores were randomly oriented or aligned parallel to the support surface. They could not yield a vertically oriented ordered nanoporous which is desirable to improve pore accessibility for practical applications.

All these methods are operated with one surface of the porous support in contact with synthesis solution. In chapter 2, the template-assisted quiescent interfacial approach was used to fabricate mesoporous silica fibers with pores oriented along the fiber axis. The interfacial principle of this approach will be used in this chapter to develop a novel

115 counter-diffusion self-assembly (CDSA) in-situ growth method for preparation of ordered nanoporous silica membranes. The CDSA approach has the potential to fabricate nanoporous membranes with one-dimensional pores normal to the support surface which remains a major challenge for membrane scientists. Such membrane will represent a potential candidate for the continuous production of extended-chain polymer fibers via extrusion polymerization application demonstrated in Chapter 2.

This chapter demonstrates the use of the novel CDSA approach for fabrication of nanoporous silica membranes on ceramic supports of variable pore sizes and surface chemistry. Microstructural and morphological properties of the produced membrane are investigated to evaluate the membrane quality and feasibility of this approach.

4.2. Principle of Counter Diffusion Self Assembly

The in-situ CDSA approach is an extension of our work on the synthesis of ordered mesoporous silica fiber at the interface of a hydrophilic water-surfactant phase and hydrophobic silica precursor phase as shown in Figure 4-1a. The counter diffusion of precursors through the interface leads to hydrolysis and condensation of silica precursor at the interface in presence of surfactant liquid crystals into fiber morphology with ordered pores aligning relatively straight across the fiber axis. The majority of the product grows at the interface towards the water phase. In the CDSA approach, a ceramic porous support will be placed at the interface and will separate the two phases as schematically illustrated in Figure 4-1b. Precursors will start to counter diffuse through the support and form interfaces of interaction where the product will form.

116 Depending on the surface chemistry of the porous support and interaction of the diffusing materials with the pore wall, the main interfaces of interaction can be on the either side of the support or somewhere inside the pores (Fig. 4-1c). If the hydrolysis and condensation proceed in the same fashion as the free interfacial method, then an ordered silica product will grow at the interface of interaction with similar microstructure of the silica fiber. If the interaction takes place at the support surface(s) then a films will grow with ordered pores vertical to support surface, while if the interface of interaction is within the support pores, then silica fiber plugs will grow inside the pores. The possible morphologies of CDSA growth approach are shown in Figure 4-2. For the later case, chances of obtaining a silica membrane plugs with pores vertical to support surface are better with a support of preferentially straight and vertical pores like the track-itched

Alumina membranes (Anopore).

4.3. Experimental

4.3.1. Ceramic supports

CDSA growth of silica membranes was performed on ceramic supports of two variables: 1) pore size (4 nm to 20 µm) and 2) surface chemistry (hydrophilic vs. hydrophobic). Variable pore size supports used were three types: mesoporous γ-alumina of 4 nm pore diameter prepared by the sol-gel method, macroporous α-alumina of 0.1 to

0.35 µm pore diameter prepared by pressing and sintering method, and macroporous borosilicate and quartz supports of 4-20 µm pore sizes obtained commercially

117

Silicon Source Silicon Source

Water + Water + Surfactant Surfactant

Mesoporous silica fibers Support (a) (b) (c)

Figure 4-1. Schematic illustration of the interfacial growth of ordered mesoporous silica fibers (a), counter diffusion self assembly growth of silica membrane on porous support (b), and an enlarged view of the interfacial interaction locations (c) on either sides of the support or inside the pores.

Silica film with straight Fiber plug inside pores on the support the pore

(a) (b) Support

Figure 4-2. Possible morphologies of CDSA growth; silica film with straight vertical pores on top of the support (a), or silica plugs (or fiber) with ordered nanopores inside the support pores (b)

118 (Chemglass, NJ). Surface chemistry study was performed the α-alumina supports. This support is relatively hydrophilic due to presence of some hydroxyl groups. It is converted to hydrophobic by modification with hydrophobic agents. All supports have disc geometry of 2 cm diameter and 2 mm thickness.

Preparation of α-alumina supports

The α-alumina supports were prepared by pressing and sintering of calcined alumina powder (ALCOA, TX). Two alumina formulas of different particles sizes have been used to obtain supports of various pore sizes. They are coded by the provider as

A15-SG and A16-SG and have a typical d50 Sedigraph particle size of 1.7 and 0.4 µm respectively. In experiments, 10 g of the powder is ground thoroughly with 8 wt % DI water (as a binder) with a quartz mortar and pestle until the mixture becomes homogeneous. This batch yields approximately 5 supports. If more supports are required, then this process should be repeated accordingly. For each support, 2.1 grams of the mixture was put in a mold and pressed at 1000 psi (corresponding to 5000 Ib) for one minute on one side and then is flipped upside down and pressed for 3 minutes at 3800 psi

(20,000 Ib). Before sintering, the supports were dried for two days in oven at 40°C and

40 % relative humidity to remove the water binder.

The supports were sintered at high temperature to consolidate the structure and increase the density using the following program. They were first ramped to 600 °C at 60

°C/hr, then ramped to 1260 °C at 96°C/hr, then ramped to 75 °C at 96 °C/min, then ramped to 1150 °C at 60 °C/hr and held for 30 hrs, and finally ramped to room temperature at 60°C/hr. After sintering, the supports were polished using a Metaserv

2000 grinder/polisher from Buehler. The side of support to be used, e.g., for coating, was

119 polished with 500, 800 and then 1200 grit polishing paper. The supports were polished by hand by holding on the polisher for 1 minute and then rotating by 90° and polishing for another minute. The supports were rotated 4 times so that each support was polished for 4 minutes at each different grit of polishing paper. The supports were then washed with deionized water and dried for two days at 40°C.

Preparation of γ-alumina membrane

The γ-alumina membranes are generally prepared by dip-coating of α-Alumina support with a boehmite sol. Stable boehmite sols of 1 M aluminum concentration were prepared from hydrolysis and condensation of aluminium tri-sec-butoxide (ALTSB)

(Acros) [Lin YS et al. 91]. The alkoxide precursor was slowly hydrolyzed in water at 80-

85°C, and after 1 h of stirring, the resulting slurry with AlOOH precipitates was peptized with nitric acid at a HNO3/AlOOH molar ratio of 0.07. The sol was refluxed overnight for more than 12 h at 90-100°C. The remnant alcohol was evaporated by refluxing the sol open to air for 2 h at 90°C to obtain a stable boehmite sol.

In preparing the supported membranes, 20 ml of 1M boehmite sol was mixed with

13 ml of 3 wt % poly (vinyl alcohol) (PVA) solution. The α-alumina supports were dip- coated on the polished side with the boehmite/PVA mixture for 5 sec. The support is removed from the dipping sol and the excess sol is dried from the edges with a wipe.

Unsupported γ-Al2O3 membranes were also prepared by drying a small amount of the same boehmite/PVA mixture in petri-dishes. The resulting unsupported membranes were irregularly shaped thin sheets of about 100 µm thickness. All the samples were dried at

40 ºC for 2 days and then calcined at 450 ºC for 1 h at 1 ºC /min heating and cooling rates. The supported γ-alumina membranes were approximately 6 µm thick and had an

120 average pore diameter of 3.6 nm. Additional layers can be applied by repeating the dipping and calcination process to cover up any pinhole defects in the previous layer.

Hydrophobic modification of supports

The α-alumina supports were modified with organic alkylsilane agents of various chain lengths to vary the degree of surface hydrophobicity. Alkylsilane agents used in this study were octadecyltricholrorsilane with 18-carbons chain (C18) and hexyltrichlorosilane with 6-carbons chain (C6) obtained from Gelest, PA. Surface modification is performed by two steps: 1) hydroxylation of the support pore surface followed by 2) grafting of the silane chains. Hydroxyl (OH) groups serve as the active sites for the attachment of silane groups. Reaction sequence is illustrated in Figure 4-3.

R R

OSi OH OH Cl3Si R OSi Hydroxylation O Grafting + HCl (g) Al Al Al Al Al Al

Alumina support Modified Alumina

Figure 4-3. Schematic illustration of surface modification of alumina supports with hydrophobic chains

Surface hydroxylation was carried out by boiling α-alumina supports in 100 ml of

30% hydrogen peroxide (H2O2) for 30 min followed by boiling in 100 ml DI water for another 30 min and finally drying at 90 ºC for 30 min [Lin HP et al. 02]. Prior to grafting of the organic silanes, the hydroxylated supports were preheated at 150-200 ºC under vacuum for few hours to remove any physically adsorbed water and to increase the surface concentration of the isolated OH groups by breaking up the hydrogen bonded ones. Up to 4 supports (of 8 g total weight) were dispersed in 150 ml toluene in a round

121 3-neck flask immersed in an oil bath at 100-120 ºC. Upon heating, the 3-neck bottle is attached to a condenser to reflux back the evaporated solvent. Then 0.1 ml of the silane precursor (C18 or C6) per gram support was added to the mixture under N2 environment where a 20 ml/min N2 stream was passed over the condenser to reduce introduction of

Air to the air-sensitive reactants. Precursors amounts were calculated assuming a support surface area of 10 m2/g and hydroxyl surface coverage of 2 groups/nm2 and were multiplied by an excess factor of at least 5. The pyridine was used as a catalyst to promote the reaction of the chlorosilanes with the surface hydroxyl groups and to trap excess hydrogen chloride generated during the reaction. The suspension was refluxed for approximately 24 h in presence of 2 ml pyridine under mechanical shaft mixing (not a stirring bar) to avoid breaking the supports. Surface functionalization procedure was adopted from several procedures reported in the literature [Leger et al. 96, McCarley et al. 01, Shigeno et al. 02, Cossement et al. 03].

Supports were removed carefully and washed according to the following sequential steps to remove excess precursors. The supports were first washed with toluene for 1-2 hours to remove excess organic silanes. Then they were washed with ethanol to remove the excess toluene, and finally washed with DI water to remove excess pyridine. The washing process can be performed in a vacuum filtration funnel, sonicator, or Soxhlet extraction apparatus. Finally, the supports were vacuum dried at 90 ºC to cure the covalent bonds of grafting. These supports were then used to grow silica membranes by the CDSA approach. The efficiency of surface activation steps on α-alumina supports was evaluated on α-alumina powder sintered and activated under the same conditions.

122 4.3.2. CDSA growth of silica membranes

CDSA method was performed using the optimum conditions reported in Chapter 2 that lead to silica fiber morphology. Typical silica source was tetrabutylorthosilicate

(TBOS) and the water-surfactant phase consisted of water, cetyltrimethylammonium bromide (CTAB) and HCl (in typical molar ratio of 100:0.0255:2.29 relative to 1 mol of

TBOS). Experimentally, the support disk was fixed at the edge of a rubber tube which was immersed in the aqueous solution of (H2O: CTAB: HCl) with one surface of the support facing the solution as shown in Figure 4-4. TBOS was then added inside the tube on the top of the other surface of the support. Precursors were allowed to inter-diffuse through the support for 7-14 days, then the membrane was dried in air at room temperature overnight followed by surfactant removal by calcination at 550 ºC for 6 h.

When hydrophobic-modified supports were used, the surfactant was removed by extraction with acidified ethanol (1 vol% HCl) at 120 ºC using Soxhlet extraction apparatus to avoid desorption of the hydrophobic groups.

H2O + HCl + surfactant

Figure 4-4. Experimental setup for CDSA growth of silica membranes

123 4.3.3. Characterization

Nitrogen adsorption/desorption isotherms of unsupported γ-alumina and unsupported mesoporous silica fibers were measured at liquid nitrogen temperature (77

K) using Micromeritics TriStar 3000 porosimeter. Surface areas were calculated using the

BET model, and pore volumes and pore diameters were calculated by the BJH (Barrett-

Joyner-Halenda) model [Gregg and Sing, 82]. Hydrophobic modification of the α- alumina supports and powder was evaluated qualitatively by Fourier Transform Infra-Red

(FTIR) spectroscopy in the range 4000-400 cm-1 (Perkin-Elmer, MA). Themogravimetric

Analysis (TGA) (SDT-2960, TA Instruments) was used to quantify the amount of surface coverage of hydrophobic groups on the α-alumina powder where a 30 mg sample of the powder was heated to 900 ºC and then cooled down to room temperature at a rate of 5

ºC/min under air flowing at 100 cm3/min.

The deposited CDSA silica membranes were characterized by XRD (Siemens D-

500, CuKα) to study the pore structure and SEM (Cambridge S-90) to study the morphology. Membrane quality and gas permeation properties were tested using single gas steady state and unsteady state time lag methods utilizing helium, nitrogen and carbon dioxide gases. The steady state method was utilized to evaluate the average pore size of the starting macroporous supports and the final CDSA-grown mesoporous silica membrane (if pressure drop of the later case is within the systems limit). In most cases, the pressure drop of the steady state permeation measurement of the CDSA silica membrane exceeds the limit of the pressure sensor (17.2 kPa= 2.5 psi). Therefore, permeation properties of these membranes were mostly tested by the unsteady state

124 approach which is usually used for tight membranes. The relative change in permeance properties can be used to evaluate growth quality of the silica membrane.

In the steady state experiments, helium gas permeation data were measured at different average pressures using the setup shown in Figure 4-5. The membrane was sealed in stainless steel membrane holder with silicon Tygon rubber O-rings. A constant flow of helium gas is continuously introduced into through the membrane face with the aid of mass flow controller. The upstream and transmembrane differential pressures were controlled using a downstream metering valve and monitored with a pressure meter

(omega). Permeation of mesoporous γ-alumina supported films and α-alumina supports were measured at 5 cm3/min feed flow rate and 3.8 to 13 kPa transmembrane pressures.

Large pore supports like borosilicate and quartz supports required much higher flow rates to induce reasonable transmembrane pressures and flow through supports. They were tested at 100-300 cm3/min gas flow rates and transmembrane pressures of 0.4 to 5.9 kPa. The permeance data at various average pressures were subsequently utilized to estimate the average pore diameter of the supports.

Unsteady state single gas permeation was used to study the change in permeation properties due to silica growth by the CDSA approach. Unsteady state setup is shown in

Figure 4-6. In this approach, the gas was allowed to permeate through the support under high transmembrane pressure (170 to 308 kPa) induced by vacuuming the downstream side. The rate of increase in pressure at the downstream side was measured by means of data acquisition system (Lab view) and utilized to evaluate the permeance assuming ideal gas low. Experiments were performed using helium, nitrogen and carbon dioxide at several temperatures (27 to 90 ºC) and gas feed pressures (170 to 308 kPa) to study the

125 microstructure of the membranes and the mechanisms of gas diffusion. The permeation tests were conducted for the bare support, deposited membrane with surfactant, and deposited membrane after surfactant removal to analyze membrane quality and formation of silica.

Differential pressure meter Pressure gauge ∆P High resolution P metering valve

Bubble flow meter Membrane cell Bubble flow He meter

Bypass needle valve Gas Cylinder

Figure 4-5. Steady state single gas permeation setup

Upstream To data acquisition pressure gauge Oven Downstream P P pressure gauge

Gas tank Membrane cell

Gas Cylinder (CO2, N2, He) Vacuum pump

Figure 4-6. Unsteady state single gas permeation setup

126 4.4. Results and Discussion

4.4.1. Characterization of Supports

Helium permeation results of the starting supports are shown in Figure 4-7 plotted in the logarithmic scale to display the relative magnitudes of data for all supports. All the supports exhibit a linear relationship between the permeance and the average pressure with a relatively different pressure dependency proportional to their average pore size.

Helium permeances of α-alumina and γ-alumina supports are in the order of magnitude of 10-6 mol/s.m2.Pa. The permeance of the α-A15 support prepared from the large particle alumina powder (A15) is 5×10-6 mol/s.m2.Pa at 125 kPa which is 5 times larger than that of the α-A16 support made from the small particle powder (A16). Resistance to permeation is proportional to thickness/pore diameter ratio of the support. With both supports having the same thickness (2 mm), the larger permeance of α-A15 indicates a smaller resistance and therefore a larger pore diameter.

The permeance of the mesoporous γ-alumina (8×10-6 mol/s.m2.Pa) is slightly higher than that of α-A15 although it has much smaller pore diameter. This is because of the small thickness of γ-alumina (6 microns) which gives much lower resistance to permeation than α-A-15. Borosilicate and quartz supports have much higher permeance than alumina supports. Their permeance is in the range of 10-4 to 10-2 mol/s.m2.Pa and have a stronger dependence on pressure (larger slopes) compared to the alumina supports as will be discussed later.

127

1.0E-02 Quartz Borosilicate Medium 1.0E-03

Borosilicate Fine .pa) 2 1.0E-04

Gamma alumina 1.0E-05

Alpha-A15 Permeance (mol/s.m 1.0E-06 Alpha-A16

1.0E-07 75 125 175 225 275 325 Average Pressure (kPa)

Figure 4-7. Steady state single gas helium permeance at room temperature of the supports used to grow silica membranes. Straight lines are the linear regressions.

128 Gas diffusion in porous materials can be in the gas phase of the pore by viscous and Knudsen mechanisms or on the pore surface by surface diffusion mechanism. For non-adsorbing gases, the surface diffusion contribution will be negligible and the diffusion will be essentially in the gas phase of the pore. The relative contribution of the viscous and Knudsen mechanisms to the net flow depends on the microstructural properties (pore diameter) and the measurement conditions (gas molecular size, pressure and temperature). These conditions identify the mean free path of the diffusing molecules and the pores diameter and therefore the extent of gas transport by collision between gas molecules (viscous) or by collision of molecules with pore wall (Knudsen)

[Karger and Ruthven, 92]. Diffusion in mesoporous materials is generally governed by

Knudsen mechanism. Viscous flow is negligible in the lower size range of the mesoporous materials. It can, however, appear to some extent in the upper mesoporous range (50 nm) and then starts to dominate in the macroporous size range (> 100 nm).

Macroporous materials with pore size diameters over 1 µm have viscous-based transport with negligible contribution from Knudsen flow.

Helium permeance data (F/L) at different average pressures (Pav) were regressed by the straight line relationship

(F/L)=α+β Pav (4-1)

where the permeance is defined as F/L=Q/A(Ph-Pl) with Q being the molar gas flow rate,

A the membrane area and (Ph-Pl) is the transmembrane pressure difference. The regression coefficients α and β represent contributions from Knudsen and Viscous flows respectively as [Lin and Burrgraaf 93]:

129 ε r α = 1.06( ) p (4-2) κ L M w RT

ε r 2 β = 0.125( ) p (4-3) κ LηRT

where ε, κ, rp, L are the porosity, tortuosity factor, average pore radius, and thickness of the membrane and η and Mw are the viscosity and molecular weight of the permeating gas. It is clear from Eqs (4-1) and (4-2) that the Knudsen contribution, expressed by α, is independent of pressure while the viscous part (βPav) is a pressure-dependent. In addition to identifying the transport mechanisms, these coefficients can be employed to measure the average pore size of the support from the ratio β/α and to examine the change in the microstructure of the support as a result of any modification process. The average pore diameter can be written in terms of the coefficients α and β as follows (where all units are in m, Pa, s, mol, K and g)

RT  β  d p =16.964η   (4-4) M  α 

Regression coefficients and average pore sizes of the supports used in this study are summarized in Table 4-1. The constant helium permeance of γ-alumina support over the given range of average pressure in Figure 4-7 and the coefficients in Table 4-1 confirm that the flow in this support is solely governed by Knudsen mechanism. The viscous flow coefficient (β) was very small negative value (~ -10-13 mol/s/m2.Pa2) due to failure of the viscous theory to apply in the pore size range of this support. This indicates that γ- alumina has a pore size in the lower range of mesoporous size. Average pore size of unsupported γ-alumina obtained by nitrogen porosimetry was 3.93 nm. It is worth

130 mentioning that the permeance data and regression coefficients of γ-alumina membrane

(which is originally a composite membrane of thin layer coated on α-alumina support) represent the top layer contribution. The contribution of the top layer was calculated from the permeance data of the composite membrane and the α support using the resistance in series model [Kiezer et al. 88] and Eqs (4-1 to 4-4).

The macroporous α-alumina supports (α-A16 and α-A15) have a respective pore diameters of 0.16 and 0.35 µm. They both exhibit a combination of Knudsen and viscous flows with respective viscous contributions of 7 and 14 % to the total permeance at 125 kPa calculated from the ratio of βPav to the total flow. The viscous contribution increases with the average pressure. The support α-A15 has a higher viscous flow than α-A16 obviously because of its larger pore diameter. The commercially-obtained Borosilicate fine (BSF), Borosilicate medium (BSM) and Quartz supports exhibit much higher viscous flow than α-alumina supports because of their large pore sizes. This is apparent from their large permeance data and the regression parameters. The average pore diameters of

BSF, BSM and Quartz supports were calculated from their permeance parameters respectively as 3.08, 5.83 and 21.29 µm. Their percentage contributions of the viscous flow in these supports were 60, 74 and 91% at 125 kPa.

131 Table 4-1. Regression parameters and average pore sizes of the supports obtained from

steady state helium permeation data

Average Support α β pore diameter (mol/s.m2.Pa) (mol/s.m2.Pa2) (µm)

γ-alumina 8.00×10-6 negative 3.93 nm a

α-alumina (A16 powder) 9.33×10-7 5.90×10-13 0.16

α-alumina (A15 powder) 4.55×10-6 6.13×10-12 0.35

Borosilicate-Fine 3.97×10-5 4.71×10-10 3.08 b

Borosilicate-Medium 3.17×10-4 7.13×10-9 5.83 b

Quartz 2.06×10-4 1.69×10-8 21.29 b

a From N2 ads-des experiment on unsupported γ-alumina sample

b Pore sizes in µm provided by vendor are 4-5.5 for BSF, 10-26 for BSM and 15-40 for Quartz

4.4.2. CDSA growth on un-modified supports

Growth of mesoporous silica product without a support (Fig. 4-1a) results in formation of silica fiber morphology at the interface towards the water phase. According to nitrogen adsorption-desorption study, unsupported mesoporous silica fibers have pore diameter and surface area of respectively 2.7 nm and 1100 m2/g. The nanopores arrange inside the fibers in a hexagonal symmetry as revealed by XRD pattern. Macroscopically, the fibers have a diameter of 15-25 µm and a distribution in the length with average value of 196 µm. Results of unsupported silica fiber growth were thoroughly discussed

132 in the previous two chapter. The first step in this work was to apply the concept of

CDSA growth of nanostructured silica membranes on the alumina and the large pore supports. These supports were used for CDSA without any surface modification. After

CDSA growth, the supports were subject to characterization before and after removal of surfactant to analyze the formation of ordered mesoporous silica and to verify the quality of the membrane.

For CDSA growth on the alumina supports (γ-alumina, α-A16 and α-A15), no silica films were observed on either side of the support as was confirmed by SEM images which displayed deposition-free surfaces. XRD patterns on both sides of these supports gave an amorphous response without any peak confirming the absence of any ordered silica films. The best cases exhibited a growth of a non-continuous flake-like silica deposition on the TBOS side (the upper surface) of the support. However, this growth was random, non reproducible and have an amorphous structure based on XDR tests.

Presence of defects in the support proved to play a role in the CDSA growth of silica membranes. A 0.16 µm α-alumina support containing 4.16 µm pinhole defects displayed a non-continuous silica film with hexagonal pore structure near the surface in contact with the water phase as shown in Figure 4-8. The ordered pores of silica have a diameter of 3 nm according to nitrogen adsorption/desorption measurement on a sample collected from the surface. The large size macroporous defect is expected to facilitate the transport of precursors through the support and lead to formation of silica film. The silica film was non-continuous as a result of the non-homogeneous distribution of the defects.

A support with a homogeneous distribution of large pores (> 3 µm) can potentially form a continuous silica films by the CDSA approach. When performed on the 3-20 µm pore

133 size BSF, BSM and Quartz supports, the CDSA growth has not displayed the expected morphology. It exhibited non-continuous growth of silica spots on the rough surface of the large pore supports with amorphous XRD patterns similar to those obtained on the alumina supports.

5µ

1357 2 Theta

Figure 4-8. SEM and XRD of the water-phase sided surface of 4 µm pore-defected alumina support after silica deposition by CDSA

134 Up to these observations, it is noticed that the CDSA approach does not create distinct silica film on the support surface. It is most likely that silica is depositing within the pores of the ceramic support as will be verified by the gas permeation tests. Gas permeation was used to examine growth of silica by the CDSA approach and identify the location of deposition. Unsteady state approach was selected because it can measure low permeance for tight membranes such as those obtained after continuous growth of silica- surfactant depositions on the support. Permeance data of the starting supports, as- synthesized and the surfactant-free silica membranes were measured and compared to examine the quality of the membrane. Gas permeation can also provide information on the microstructure and order of the silica by analyzing the controlling mechanisms of diffusion and the relative change of permeance upon removal of surfactant. Unsteady state nitrogen permeance of CDSA silica membranes on the unmodified supports are summarized in Table 4-2.

In terms of permeation, a good quality membrane should exhibit high reduction in permeance if silica-surfactant deposition grows continuously. As shown in Table 4-2, the as-synthesized CDSA silica membranes on alumina supports (γ-alumina, α-A16 and α-

A15) exhibit small reduction in permeance between 1.3 to 10 times relative to the original support. This implies that silica-surfactant product is depositing inside the support pores. However, silica growth was not sufficient to plug the pores of the support and suppress the permeance. Removal of surfactant opens up spaces of the mesopores in the silica and increases the permeance little bit. The permeance of the surfactant-free silica membrane was close to the original support value verifying the insufficient growth of silica to and the low quality of the final membrane. CDSA growth on the large pore

135 supports was slightly better. The as-synthesized silica membrane on the defected alumina, BSF, BSM and Quartz supports exhibited 1 to 3 folds reduction in permeance which reflects deposition of more silica inside the support pores. However, the growth was insufficient to completely seal the larger pores. The final surfactant-free silica membrane will have high permeance values (>> 10-6 mol/s.m2.Pa) as a results of the free or incompletely plugged macropores.

Table 4-2. Nitrogen permeance of CDSA silica membranes on unmodified ceramic

supports at 170 kPa feed pressure and 25 ºC

Sample Pore size Membrane Permeance (mol/s.m2.Pa) (µm) support As-synthesized Surfactant-free

γ-alumina a 4 (nm) 4.82×10-7 5.10×10-8 2.79×10-7

α-A16 0.16 3.85×10-7 3.00×10-7 3.18×10-7

α-A15 0.35 8.04×10-7 2.01×10-7 7.76×10-7

α-A16 (defected) 4.16 1.98×10-6 2.03×10-7 4.29×10-7

BS-fine b 3.08 9.44×10-5 9.36×10-6 2.19×10-5

BS-medium b 5.83 1.78×10-4 5.19×10-6 -

Quartz b 21.29 1.96×10-4 1.14×10-4 -

a Permeance data are for the composite membrane (γ layer on α-A15 support)

b Helium gas permeance, BS= Borosilicate

136 The low quality of CDSA growth of mesoporous silica membranes on unmodified alumina and large pore supports can be generally attributed to hindrance of the hydrolysis and condensation reactions. This could be due to the inhibited transport of some precursors through the support. It is likely that the hydrophilic surface of the support pores (containing OH groups) reduces the penetration of the hydrophobic silica precursor and facilitates the transport of only the hydrophilic water through the pores. In small-pore supports, it is expected that transport of the liquid crystalline surfactant micelles will be also inhibited. Therefore, the reaction of water and silica precursor will proceed at the upper side of the support without involvement of surfactant to give amorphous silica deposition as noticed for some cases on γ- and α-alumina supports.

For large-pore supports, transfer of the silica through the pores is easier and leads to hydrolysis and condensation near the lower surface of the support in presence of surfactant to produce ordered product. The co-transfer of water and surfactant through the pores causes reaction interfaces to be within the pores and leads to growth of silica inside the pores. However, silica growth was insufficient to completely seal the large pores and get good quality membrane as was observed from permeance data of silica membranes on large pore supports. Accordingly, it is demonstrated that the facilitated transport of silica precursor and the support pore size are key factors in the growth of silica membranes by the CDSA approach. A hydrophobic support would, therefore, be promising for growth of good quality ordered mesoporous silica membranes by the

CDSA approach.

137 4.4.3. Characterization of hydrophobic modified supports

To facilitate transfer of the hydrophobic silica precursor through the support, the pores were functionalized by grafting of hydrophobic groups. The functionalization was done on α-alumina (α-A15) supports using alkyl silanes of different chain lengths (C18 and C6). Alumina base samples were hydroxylated prior to grafting to improve the functionalization process. The hydroxylated and functionalized samples were tested by

FTIR and TGA to evaluate the quality and quantity of surface coating.

FTIR results on the base, hydroxylated, and hydrophobic-modified alumina samples are shown in Figure 4-9. Hydroxylation of the alumina base sample added some hydroxyl (OH) groups on the surface as confirmed by the broad band in the region 3398 cm-1 in the FTIR spectra of the hydroxylated sample (Figure 4-9b). Isolated (free) OH groups normally give sharp peak in the 3500-3600 cm-1 range [Smith, 96]. However, the broad band implies that most of the OH groups are hydrogen bonded with each other or physically bonded to adsorbed water molecules. Grafting of the hydrophobic chains requires isolated OH groups, therefore, the hydroxylated sample was heated at 200 ºC under vacuum, prior to grafting, to increase the concentration of isolated OH groups by removing adsorbed water and breaking the hydrogen bonds. The strength of the OH band decreased after grafting the C6 and C18 due to consumption of some OH groups as shown in Figure 4-9 c and d. Presence of the hydrophobic chains is verified by

-1 appearance of a band at 2892 and 2895 cm corresponding to methyl (−CH3) of the terminal C6 and C18 groups attached to the surface.

138

(d) C6-modified

2892 cm-1

(c) C18-modified

2895 cm-1 (b) Hydroxylated alumina

3398 cm-1 (a) Alumina

4000 3500 3000 2500 2000 1500 1000 500 Wavenumber (cm-1)

Figure 4-9. FTIR transmittance spectra of hydrophobic modified alumina supports, (a) alumina A15 base, (b) hydroxylated alumina, (c) C18-modified and (d) C6- modified alumina.

139 Surface coverage of the hydrophobic groups was qualitatively studied by the thermo gravimetric analysis (TGA) approach. TGA results and the corresponding derivative thermogravimetry (DTG) curves of the hydroxylated and silane-modified alumina samples are shown in Figure 4-10. The DTG is the derivative weight of the TGA results with respect to temperature. It provides precise information on desorption of different species from the alumina surface with temperature. DTG of the hydroxylated alumina sample (Fig. 4-10 a) shows two distinct desorption peaks at approximately 280-

350 and 450-530 ºC. The first peak corresponds to release of the water molecules hydrogen-bonded to the hydroxyl groups. The second peak is due to release of water molecules from the condensation of hydroxyl (Al−OH) groups to form Al−O−Al bonds.

Total weight loss from this sample due to water release was 0.227 % (Fig. 4-10b).

DTG curves of the C6- and C18-modified samples (Fig. 4-10a) also display two peaks similar to the unmodified sample. The first peak is, however, broader than that of the unmodified sample and appears in the range 230-350 ºC while the second peaks are almost identical. The broadness in the first DTG peak of modified samples is due to desorption of the grafted hydrophobic chains in that temperature range. The total weight loss from the C6- and C18-modified samples (0.424 and 0.340 % respectively) compared to the unmodified value (0.227 %) was used to evaluate the surface concentration of the hydrophobic chains. With 10 m2/g specific surface area of the alumina support, surface coverage of the terminal C6 (C6H13, M=85 g/mol) and C18 (C18H37, M=253 g/mol) was

2 calculated as 2.32 and 0.45 µmol/m respectively [= (%wt lossi g-0.227 g)/(100g

2 2 alumina×10 m /g×Mi g/mol)]. This corresponds to 1.40 and 0.27 and molecules/nm for

C6 and C18.

140

0.005 (a)

C) 0.004 o

0.003

C6-modified 0.002

C18-modified 0.001 Derivative Weight (%wt/

OH-alumina 0 100 200 300 400 500 600 700 Temperature (oC)

0.45

0.40 (b) C6-modified 0.35

0.30 C18-modified

0.25

0.20 OH-alumina

% Weight Loss Weight % 0.15

0.10

0.05

0.00 100 200 300 400 500 600 700 Temperature (oC)

Figure 4-10. TGA results of hydrophobic modified alumina supports, (a) differential thermogravimetry (DTG) curves and (b) percentage weight loss curves of hydroxylated, C18-modified and C6-modified alumina.

141 Surface coverage values indicate that C6 was grafted to the alumina surface more efficiently than C18. This is possibly due to effects imposed by the long hydrophobic

C18 chains that hinder the orientation of chains in directions favorable for grafting.

Orientation of shorter chains is easier and their grafting process is more efficient. If water molecules are present in the reaction, grafting process could lead to non-favorable blockage of the support pores by hydrolysis and condensation of the silane agent to microporous silica. This effect was tested using gas permeation for the alumina support before and after hydrophobic modification. The reduction in permeance of the support

-7 -7 2 due to grafting was insignificant (8.04×10 to 7.20×10 mol N2 /s.m .Pa) verifying that the grafting process did not block the pore but rather provided a monolayer coverage of the hydrophobic groups.

4.4.4. CDSA growth on hydrophobic modified supports

The use of hydrophobic modified supports in the CDSA approach has considerably affected the growth of silica membrane compared to unmodified supports. Nitrogen permeance decreased by 100-180 times (ratio of permeance of as-synthesized to support) after growth of the silica membrane on both C6- and C18-modifed supports as shown in

Table 4-3. This reduction is much higher compared to 4 times reduction in case of the unmodified supporter reflecting growth of more silica. Permeance of the as-synthesized membrane is almost tight (~ 10-9 mol/s.m2.Pa) indicating a good quality membrane absent of any pinhole defects as a results of the continuous growth of the silica-surfactant composite up to sealing most of the pores of the support.

142 The surfactant was removed from the composite membrane by extraction rather than by the conventional calcination to avoid burning off the hydrophobic chains. These chains thermally decompose at 250-350 ºC (see TGA results) and therefore will not be stable at the conditions of calcination at 500 ºC. Removal of the surfactant will opens up the ordered pores within the silica membrane and increases the gas permeance. As seen in Table 4-3, the surfactant-free membrane of both C6- and C18- modified supports exhibit a nitrogen permeance of 4.1-4.5×10-8 mol/s.m2.Pa which is about 2 folds smaller than that of the support. Hydrophobic modification of the large pore supports (BSF,

BSM, and quartz) has also improved the formation of silica membranes by CDSA. They displayed much lower permeance, particularly the BSF support, after deposition of silica compared to the unmodified case discussed in Table 4-2. However, analogues to the unmodified supports, the growth of silica was non-continuous to seal all the macroporous defects and give permeation properties representative of the mesoporous silica.

Table 4-3. Nitrogen permeance of CDSA silica membranes on hydrophobic modified

alumina supports at 170 kPa feed pressure and 298 K

Sample Pore size Membrane Permeance (×10-7 mol/s.m2.Pa) (µm) support As-synthesized Surfactant-free

C6-αA15 0.35 7.20 0.04 0.41

C18-αA15 0.35 7.20 0.07 0.45

143 XRD and SEM, shown in Figure 4-11, were performed on the hydrophobic supports to identify the location of silica deposition. No ordered structure peaks were observed in the XRD patterns on both sides of the support and the SEM image confirming the absence of any thin film on the surface of the support. These results together with the pronounced reduction in support permeance of the CDSA silica membrane confirm the deposition of silica-surfactant product within the support pores.

This is similar to the growth obtained on the unmodified supports demonstrating that

CDSA approach always yields deposition of silica membrane within the support pores and not as distinct film on the surface.

5µ

1357 2 Theta

Figure 4-11. SEM and XRD of C18-modified alumina support after silica deposition by CDSA

144 4.4.5. Permeation and microstructural properties of CDSA grown silica membranes

Due to deposition of silica within the support pores, it was not possible to analyze the microstructure directly by XRD. Gas permeation and diffusion characteristics of the silica membranes can, however, provide useful information for the analysis of the inner silica microstructure. The original support is macroporous and its permeance is governed by viscous mechanism which is pressure dependent as depicted from Eqs. (4-1) and (4-3)

(viscous permeance = β Pave). The silica deposited within the macropores of support by the CDSA approach is mesoporous. If the membrane is free of pinhole defects, then the gas permeance behavior of the membrane will be governed by Knudsen diffusion described by Eq. (4-2) (Knudsen permeance=α).

Figure 4-12 shows single gas permeance data of He, N2 and CO2 gases on the

CDSA silica membranes as function of the feed pressure measured by unsteady state permeation approach. Gas permeance of the silica membrane on the C18-modified support is almost constant over the range of applied pressure for all gases used which is characteristic of Knudsen diffusion. The presence of small slope is due to defects or pinholes in the membrane which contributes to the slight observed viscous flow. The

-7 -8 -8 2 permeances of He, N2 and CO2 were 2.20×10 , 6.07×10 and 5.02×10 mol/s.m .Pa (at

240 kPa) respectively. Permeance values were reproducible with 10-14% relative accuracy. These values are for composite membrane but they are characteristic of the deposited mesoporous silica as will be discussed in the following paragraph. Silica membrane on C6-modified alumina exhibited a similar behavior with slightly different permeance values as shown in Figure 4-12b.

145

1.0E-06 (a) C18-modifed .Pa) 2 He

1.0E-07

N2

CO2 Permeance (mol/s.m

1.0E-08 150 200 250 300 350 Feed Pressure (kPa)

1.0E-07 (b) C6-modifed 7.5E-08 .Pa) 2

5.0E-08 N2

CO2

2.5E-08 Permeance (mol/s.m

0.0E+00 150 200 250 300 350 Feed Pressure (kPa)

Figure 4-12. Single gas permeance data of CDSA silica membranes grown on (a) C18-modified and (b) C6-modified α-A15 alumina supports at room temperature. Solid lines are to lead the eye.

146 It should be pointed out that the gas permeation data given in Figures 4-12 is for the composite membrane of deposited mesoporous silica and the macroporous support. It is important to calculate the gas permeance of the deposited silica membrane not only for characterization of the membrane but also to study improvement of the composite membrane permeance by using other supports of higher permeance values. The CDSA silica membrane is slightly different from other supported membranes. The membrane is grown as plugs within the support pores and not as a distinct film on the surface.

Therefore, the exact thickness and location of silica deposition can not be directly determined. Permeance of the deposited silica membrane can be obtained using the resistance in series analysis. Resistance in the composite silica-support membrane is the summation of resistances in the silica membrane and the support according to the following equation:

−1 −1 −1 (F / L)composite = (F / L)silica + (F / L)sup port (4-5) where (F/L) is the gas permeance and the inverse of this value, i.e., (F/L)-1, is the

-1 resistance to gas transport according to Fick’s law, Flux=Resistance ×(Ph-Pl) =

(F/L)×(Ph-Pl). Permeance contribution of the silica membrane was evaluated using the nitrogen gas permeance data of the composite membrane with Eq. (4-5) and is plotted in

Figure 4-13. Permeance of the silica membrane is only 8% higher than corresponding composite membrane. This implies that the CDSA silica deposition was thick and induced high resistance to permeation compared to the support. If the deposited silica membrane thickness was small (e.g., few microns like γ-alumina layer), then the permeance of the silica membrane would be orders of magnitude higher than the composite.

147

1.00E-05 Support Composite Silica .Pa) 2 1.00E-06

1.00E-07 Permeance (mol/s.m

1.00E-08 150 200 250 300 350 Feed Pressure (kPa)

Figure 4-13 Nitrogen permeance contributions of the composite CDSA silica membrane, deposited silica and support at 25ºC.

The temperature and molecular weight dependencies of gas permeance of the

CDSA silica membranes are plotted in Figure 4-14. Permeance of He, N2 and CO2 gases is linearly proportional to the inverse square root of both temperature and molecular weight with good fit between the experimental data and the linear regression line passing through the origin. This behavior is a characteristic feature of the gas transport by

Knudsen mechanism where the permeance is inversely proportional to the square root of the temperature and molecular weight product according to Eq. (4-2), F/L= 1.06(ε/τ) r

/L/(MRT)1/2. From this relation it can be seen that every diffusing gas should have the same value of (F/L)(MT)1/2=1.06(ε/τ)r/(LR1/2) independent of temperature or molecular weight if gas transport in the membrane involved only Knudsen diffusion. This constant

148

2.5E-07 (a) He 2.0E-07 .Pa) 2

1.5E-07

1.0E-07 N2

CO2 Permeance (mol/s.m Permeance 5.0E-08

0.0E+00 0.051 0.054 0.056 0.059 0.061 Temperature -1/2 (K-1/2)

2.5E-07

(b) He 2.0E-07 .Pa) 2

1.5E-07

1.0E-07 N2

CO2

Permeance (mol/s.m 5.0E-08

0.0E+00 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Molecular Weight -1/2 (mol/g)-1/2

Figure 4-14. Temperature (a) and molecular weight (b) dependencies of single gas permeance data of CDSA silica membrane on C18-modified alumina support. Solid lines are the linear regression passing through the origin.

149 is directly related to the structural properties of the membrane and can be used to provide precise quantification of microstructure if membrane thickness and porosity properties are available. (F/L)(MT)1/2 evaluated from Figure 4-14 was found to be a relatively

-6 constant value of approximately 7.36×10 ± 2.4% for He, N2 and CO2 confirming that transport of these gases through the CDSA silica membranes is governed by Knudsen diffusion mechanism.

The Knudsen dominance of gas transport properties reflects that silica membrane obtained by CDSA approach has a mesoporous structure. The permeation behavior with temperature and gas molecular weight also implies that deposited silica is not microporous. If the silica has few angstrom micropores, the permeance would be size selective for which the diffusion of molecules will decrease with increasing the molecular diameter , i.e., permeance will be in the order He (2.6 Å) > CO2 (3.3 Å) > N2 (3.7 Å).

Also the permeance will be activated and increase with temperature [Lee ad Oyama 02].

For CDSA membranes, permeance is not size selective, i.e., permeance of N2 > CO2, and it is not temperature activated confirming that the silica is not microporous. Gas permeance of the CDSA mesoporous silica membranes is some how smaller than those obtained by sol-gel dip or spin coating as show in Table 4-4. This is mainly because it is much thicker compared to the micron thick films obtained by other approaches. Silica deposition by CDSA approach is comparable to the membrane obtained by pumping of precursors through the support [Saito et al. 03]. CDSA approach, however, provides precursors from both sides of the support and yields a thicker mesoporous silica deposition as depicted from its lower nitrogen permeance.

150 Table 4-4. Comparison of nitrogen permeance of mesoporous silica membranes by

CDSA approach and other preparation methods

Preparation method Permeance Ref.

Counter diffusion self assembly (CDSA) 0.45×10-7 This work

Sol-gel dip coating 7.4×10-6 Zhong et al. 03

0.9-1.0×10-7 Yang et al. 02

Klotz et al. 03,

2.25×10-6 McCool et al. 03

Hydrothermal synthesis 1.8×10-6 McCool et al. 03

Pumping precursors through the support 2.72×10-7 Saito et al. 03

The thick deposition of mesopores silica was the main reason for not using the steady state approach to measure gas permeance in CDSA silica membranes. It induces very high resistance to gas permeation and substantially increases the equilibration time up to several hours. Although unsteady state approach can provide useful information on gas permeation, the steady state approach is more accurate and can be additionally used to estimate of the average pore diameter of the membrane [Lin and Burggraaf 93].

Helium permeance of CDSA silica membrane on C18-modified alumina was measured by the steady state approach and used to evaluate the average pore diameter of the deposited mesoporous silica according to Eqs. (4-1) to (4-4). The pore diameter was estimated to be 8 nm. This value is higher than the observed 2.8 nm pore diameter of the

151 unsupported silica most probably due to pinhole or unfilled defects that contribute to increasing the average pore diameter of the silica membrane.

A rough estimate of the silica membrane deposition thickness can be found from the permeance data and Knudsen relation. Transport in the mesopores silica and the alumina support is Knudsen dependent for which the permeance (F/L) is proportional to d/L where d is the pore diameter and L is the membrane thickness. The (F/L) ratio of support to silica membrane is then proportional to d/L of support to silica as:

(F/L)support/(F/L)silica=(dsupport/dsilica)×(Lsilica/Lsupport). Given that the permeance ratio of support to silica is 12.78, dsupport=0.35 µm and dsilica=8 nm, the silica membrane thickness can be estimated to be 0.58 mm ± 3%.

4.4.6. Effect of hydrophobic chain length

Hydrophobic modification of the support proved to have a significant impact on the growth of silica membranes by the CDSA approach. Hydrophobic chain length, on the other hand, does not seem to have much effect on the of silica growth. Both C6- and C18- modified alumina supports facilitate the growth of good quality silica membrane with almost similar permeation and microstructural properties. From Figure 4-12 it can be noticed, however, that gas permeance of the silica membrane on the C18-modified is almost 30% higher than that on the C6-modified over the applied pressure range. Both supports (C6 and C18) exhibit slight increase in permeance with pressure due to possible pinhole defects in the membrane. Part of these defects can be from the empty spaces or volumes generated between the hydrophobic modification chains and the external surface of silica plug grown inside the support pore (side gaps) as illustrated in Figure 4-15.

152

Modification chain (C6 or C18)

Mesoporous silica plug Support pore wall

Side defect

Figure 4-15. Schematic illustration of the side gap created between silica plugs and support pore wall.

The size of these gaps depends on the surface coverage and the length of the hydrophobic chains. Surface coverage of C6 and C18 chains was estimated respectively as 1.45 and 0.45 group/nm2. The distance between two adjacent chains will be ~ 2.2 nm for C18 and ~ 0.7 nm for C6. The effective hydrophobic chain length can be estimated from the relation l=1.5+1.265n Å where n is the number of carbon atoms in the chain

[Israelchvilli, 85] from which C6 and C18 are respectively 1 nm and 2.4 nm long. The mechanism of silica growth on both supports is similar; therefore, the major difference in their silica membrane microstructure will be the side defects. Side gaps of the C18- modified supports is larger than those in the C6-modified supports (~ 2.2×2.4 nm vs.

0.7×1.0 nm) and this is the possibly why it exhibits slightly larger permeance value than

C6 silica membrane.

153 4.4.7. Expected mechanism of formation

Continuous deposition of silica within the hydrophobic support pores demonstrates the importance of the hydrophobic chains on the efficiency of precursor transport and reaction through the pores. Water phase can easily diffuse through the pores towards the side of silica precursor. The slow or inhibited transport of silica precursor was the reason for the insufficient deposition of silica on unmodified supports which yielded a low quality membrane full of defects. Presence of hydrophobic chains on the pore surface facilitates the transport and adsorption of hydrophobic silica precursor molecules on the inner pore surface. The hydrophobic surface will also enhance the adsorption of the surfactant molecules (which contain hydrophobic tails) on the wall and initiate the hydrolysis and condensation reaction to produce surfactant-templated silica. Production of alcohol by-product will speed up the diffusion of water and surfactant inside the pore and increase the extent of the reaction. Silica-surfactant deposition will continue to grow and pack together up to plugging the support pores until diffusion of more precursors to the pore is not possible.

If precursors are provided only from same side of the support, then there would be a concentration gradient of precursors which gradually decrease from the pore mouth towards the pore. Therefore the thickness of the deposited silica will be small as observed by Azuma and Coworkers [Saito et al. 03] who obtained a 10 µm silica deposition by introducing the precursors from one side. Counter diffusion, on the other hand, provides precursors from both sides and over a bigger range of the support pore and therefore yields a thicker (~ 0.55 mm) deposition of mesoporous silica. Order of the silica pore structure was not directly studied but the increase of gas permeance upon

154 removal of surfactant was used as indicator for involvement of surfactant in the reaction and presence of ordered structure in the deposited silica.

The orientation of the pores within silica is a major issue particularly for practical separation and reaction application. The pores should be aligned in a way to allow accessibility of gas molecules. Growth of mesoporous silica films on the surface of a support by sol-gel coating approach demonstrates the effect of the support to direct the pores orientation parallel to the support surface. With this effect in mind, if the surfactant is introduced inside the support then they will be aligned parallel to the pore wall and lead to mesopores oriented in way analogues to the shape of pores. The CDSA approach was applied in this work on alumina support of 3-dimensional tortuous pore structure.

Although the silica have oriented intraparticle mesopores, the over all orientation of the pores is 3 dimensional analogous to the support pore structure. The use of a support of well-defined and normally aligned macropores (e.g., Anopore) will be an ideal environment for the growth of ordered mesoporous silica with pores vertical to the support surface by the CDSA approach.

The CDSA was applied on Anopore alumina support but the support was unstable under the highly acidic conditions of growth. This 60 µm support dissolved few days after immersion in the acidic water phase. A thicker membrane of similar pore structure will be more potential for use as a support in the CDSA growth. In a recent paper published in Nature materials [Yamaguchi et al. 04], a Japanese group was able to demonstrate the preparation of ordered mesopores silica infiltrated in the pores of

Anopore support by dropping precursor solution on top of the support. A 10 µm long fiber-like mesoporous silica product was grown in the pores within the pores near to

155 support surface. The mesoporous silica has the required orientation of pores vertical to the support surface. From their SEM images, it appears that the silica product is not completely sealing the pores, but rather leaving some empty gapes between the silica and pore walls. These gaps would serve as defects that inversely affect the selectivity and practical use of the membrane. Fundamentally, this work represents a major advance in the membrane synthesis field, but practically the synthesis procedure needs modification to obtain fully plugged pores by the silica. This can be done by the CDSA approach using a mechanically strong support of ordered structure, e.g., micro channel plate.

Mesoporous silica membranes prepared by CDSA has an improved mechanical properties compared to those prepared as thin films by conventional techniques. Such membrane is very desirable for separation applications, e.g., size selective separation of enzymes and proteins. It can also be potential in various areas such as membrane reactor, catalysis, sensors, microelectronics and drug delivery with material properties and device characteristics in unprecedented ways [Davis 03]

156 4.5. Conclusions

This chapter reports a novel method for the fabrication mechanically strong nanostructured silica membrane with controlled pores orientation relative to the support surface. It is based on the counter diffusion self assembly (CDSA) where the silica and surfactant precursors are introduced in a counter diffusion fashion from both sides of the support and self assemble inside the pores to grow ordered mesoporous silica. Experiments were run under effects of variable support pore size and surface chemistry. CDSA approach generally leads to growth of mesoporous silica as plugs within the support pores.

Facilitated transfer of precursors through the support is the key requirement for the growth of defect-free mesoporous silica membrane by this approach. CDSA silica growth on alumina supports of 4 nm–0.35 µm pore sizes was inefficient and the final membranes were full of defects due to inhibited transfer of the hydrophobic silica precursor and surfactant through the pores. Modification of the support with hydrophobic groups improved the diffusion of the hydrophobic silica precursor and the amphiphilic surfactant and caused sealing of the pores by growth of mesoporous silica plugs. Gas permeation conducted on the good quality silica membranes exhibits a Knudsen type permeance behavior confirming the mesoporous structure of the silica membrane. Thickness of the silica deposition within the pores was estimated to be 0.55 mm. Such membrane offers high mechanical strength and can find many applications in separation and molecular membrane reaction areas.

157 4.6. References

Cossement D., Plumier F., Delhalle J., Hevesi L., Mekhalif Z., Electrochemical

Deposition of Polypyrrol Films on Organosilane-modified ITO Substrates, Synthetic

Metals, 138 (2003) 529.

Davis M.E., Ordered porous materials for emerging applications, Nature, 417 (2002) 813

Gregg S.J., Sing K.S.W., Adsorption, Surface Area, and Porosity, 2nd edition, London;

New York: Academic Press, 1982.

Israelachvili J.N., Surfactants in Solution, New York; Plenum, 1987

Karger A., Ruthven D.M., Diffusion in Zeolites and other Microporous Solids, J. Wiley

& Sons. Inc., 1992, Ch. 4, p. 98.

Keizer K., Ulhorn R.J.R., Van Vuren R.J., Burggraaf A.J., Gas Separation Mechanism in

Microporous Modified γ-alumina Membranes, J. Membr. Sci., 39 (1988) 285.

Klotz M, Ayral A., Guizard C., Cot L., Synthesis and Characterization of Silica

Membranes Exhibiting an Ordered Mesoporosity: Control of the Porous Texture and

Effect on the Membrane Permeability, Sep. Purif. Technol., 25 (2001) 71.

Lee D., Oyama S.T., Gas Permeation Characteristics of Hydrogen Selective Supported

Silica Membrane, J. Membr. Sci., 210 (2002) 291.

Leger C., Lira H., Paterson R., Preparation and Properties of Surface Modified Ceramic

Membranes. Part III: Gas Permeation of 5 nm Alumina Membranes Modified by

Trichloro-Octadecylsilane, J. Membr. Sci., 120 (1996) 187.

Lin H.P., Mou C.Y., Structural and Morphological Control of Cationic Surfactant-

Templated Mesoporous Silica, Accounts of Chem. Research., 35 (2002) 927.

158 Lin Y.S., deVries K.J., Burggraaf A.J., Thermal Stability and its Improvement of the

Alumina Membrane Top-Layers Prepared by Sol-Gel Methods, J. Mater. Sci., 26

(1991) 715.

Lin Y.S., Burggraaf A.J., Experimental Studies on Pore-Size Change of Porous Ceramic

Membranes after Modification, J. Membr. Sci., 79 (1993) 65.

Martin J., Anderson M., Odinek J., Newcomer P., Synthesis of Periodic Mesoporous

Silica Thin Films, Langmuir, 13 (1997) 4133.

McCarley K., Way J.D., Development of a Model Surface Flow Membrane by

Modification of Porous γ-Alumina with Octadecyltrichlorosilane, Sep. Purif. Tech.,

25 (2001) 195.

McCool B.A., Hill N., DiCarlo J., DeSisto W.J., Synthesis and Characterization of

Mesoporous Silica Membranes via Dip-coating and Hydrothermal Deposition

Techniques, J. Membr. Sci., 218 (2003) 55.

Nishiyama N., Koide A., Egashira Y., Ueyama K., Mesoporous MCM-48 Membrane

Synthesized on a Porous Stainless Steel Support, Chem. Commun., (1998) 2147.

Ogawa M., Formation of Novel Oriented Transparent Films of Layered Silica-Surfactant

Nanocomposites, J. Am. Chem. Soc., 116 (1994) 7941.

Saito H., Higuchi M., Katayama K., Azuma Y., Gas Permeation through a Porous

Alumina Filled with Mesoporous Silica, J. Mater. Sci. Lett., 22 (2003) 395.

Shigeno T., Nagao M., Kimura T., Kuroda K., Direct Silylation of a Mesostructured

Precursor for Novel Mesoporous Silica KSW-2, Langmuir, 18 (2002) 8102

Smith B.C., Fundamentals of Fourier Transform Infrared Spectroscopy, Boca Raton:

CRC Press, 1996, Ch. 3., p. 81.

159 Tolbert S., Schaffer T., Feng J., Hansma P., Stucky G.D., A New Phase of Oriented

Mesoporous Silicate Thin Films, Chem. Mater., 9 (1997) 1962.

Yamaguchi A., Uejo F., Yoda T., Uchida T., Tanamura Y., Yamashita T., Teramae N.,

Self-Assembly of a Silica–Surfactant Nanocomposite in a Porous Alumina

Membrane, Nature Mater., 3 (2004) 337

Yang S.M., Lee Y.E., Hyun S.H., Lee C.H., Organic Templating approach to Synthesis

of nanoporous Silica Composite Membrane (1): TPA-templating ad CO2 separation,

J. Mater. Sci., 37 (2002) 2519.

Yang H., Kuperman A., Coombs N., Afara S.M., Ozin G.A., Synthesis of Oriented Films

of Mesoporous Silica on Mica. Nature, 379 (1996) 703.

Zhong S.H., Li C.F., Li Q., Xiao X.F., Supported Mesopoorus SiO2 membrane

synthesized by sol-gel-template Technology, Sep. Purif. Technol., 32 (2003) 17.

160 CHAPTER 5

CYCLIC CVD MODIFICATION OF STRAIGHT PORE ALUMINA

MEMBRANES

5.1. Introduction

The work presented in this chapter summarizes a CVD-based technique for fabrication of inorganic mesoporous membranes with controlled microstructures. It represents an additional effort to the counter diffusion self assembly (CDSA) synthesis approach presented in Chapter 4 for fabrication of ordered nanostructures membrane with improved pore microstructures. One advantage of the CVD approach is its ability to generate membranes with rich pore microstructures, such as cylindrical mesopores normal to the support surface and pores with fractal microstructures. The successful synthesis of mesoporous membrane with such pore structures by the CVD would be a significant contribution in the membrane area. It would also represent a suitable alternative for the template-assisted membrane synthesis techniques that have not achieved the vertical pore membrane microstructure yet.

The CVD work presented in this chapter focuses on modification of the internal pores of alumina membranes with straight pore structure commercially known as

Anopore. Anopore membranes, also called Anodisc, are alumina films with well defined cylindrical, straight, and hexagonally packed pores running in the direction normal to the membrane surface [Furneaux et al. 89, Crawford et al. 92]. They are made by electrochemical anodic oxidation of aluminum and are available in 60 µm thickness. Anopore membranes with smallest pore size available commercially have a pore diameter of 20 nm. For proper performance in membrane reactor applications, pore size of Anopore membranes is required to be further narrowed to the low limit of

161 mesoporous size (2-4 nm). The modified 2-4 nm pore Anopore membrane may offer applications for gas separation and polymer reactions.

Atomic layer chemical vapor deposition (ALCVD) is one of the CVD techniques used for the modification of porous inorganic membranes [George et al.

96]. Compared to the conventional CVD [Lin and Burggraaf 92], ALCVD can precisely tailor the membrane pores by the controlled growth of the oxide deposition layer inside the pores. For the ALCVD of alumina (Al2O3), the binary reaction

2Al(CH3)3 + 3H2O Æ Al2O3 + 6CH4 was separated into two half reactions: AlOH* +

Al(CH3)3 Æ Al—O—Al (CH3)2* +CH4 and Al CH3* + H2O Æ AlOH* + CH4, where asterisks indicate surface species. The reagents, Al(CH3)3 vapor and H2O vapor are introduced separately in a step-wise manner. The thickness of deposit layer can be precisely controlled by the number of subsequent additions of the reagents causing one atomic layer to be formed at a time.

The ALCVD method have been used to reduce the pore size of a 22 nm

Anopore alumina membranes by the alternative introduction of trimethylaluminum

(TMA) and water vapor using variable reaction cycles [Ott et al. 97]. The nitrogen permeance value for the unmodified Anopore consistently decreased from 6.7×10-4 mol/s.m2.Pa (at of 1.5 atm feed pressure and differential pressure of about 0.7 atm) to

1.0×10-4 mol/s.m2.Pa after many cycles of ALCVD. They reported that the pore size was progressively reduced from 22 to 14 nm after 120 reaction cycles with Al2O3 reaction growth rate of 0.37Å/cycle. This study, which was focused on chemistry of

ALCVD on Anopore membrane pores, shows that the ALCVD method can precisely control pore size. However, reducing the pore size of a membrane from about 20 nm to 3-4 nm by this ALCVD method would require several hundred CVD cycles.

162 A modified ALCVD method was reported to effectively narrow the pore size of a sol-gel derived γ-alumina membrane for condensable vapor separation purposes

[Pan et al. 99, Cooper and Lin 02]. In the modified ALCVD, some residual reactant molecules were allowed to present in the gas phase between half-reaction cycles. This may have caused the deposition to take place both on the pore wall by heterogeneous reaction and in the pore gas phase by homogeneous reaction of residual reactants. It was proposed that such heterogeneous/homogeneous CVD might result in deposition of alumina of a fractal structure desirable for separation applications.

This chapter presents modification of a 20 nm straight pore Anopore membrane by the modified ALCVD method using two schemes adopted from the homogenous/heterogeneous deposition mechanisms proposed above. In one scheme some precursor residual will be allowed to present inside the pore. The other scheme is based on minimizing the presence of any residual precursor from the pore gas phase. This will give rise to deposition of different microstructures. Gas permeation and separation data are measured for the original and modified membranes and used to interpret the CVD modification mechanisms and the resulting deposition microstructure.

163 5.2. Experimental

5.2.1. Anopore membrane modification

The Anopore alumina membranes were obtained commercially (Whatman Co.,

England). These membranes are of 60 µm thickness and have an asymmetric structure. The majority of the membrane is comprised of straight, cylindrical, and non-connected pores of 200-250 nm diameter laying over 58 µm of the membrane thickness. The top layer of the membrane consists of 20 nm straight pores and has a thickness of 2 µm. A schematic of the Anopore membrane structure and SEM image

(Hitachi-S4000) of the composite pores, are shown in Figure 5-1. The Anopore membrane has a diameter of 25 mm and is supported with a polypropylene ring heat glued to the outer edge of the membrane. Anopore membranes were initially heat treated in oven at 480 oC to remove the polypropylene support ring.

The CVD modifications were conducted in a simple hot wall CVD reactor used for CVD of metallic/ceramic composite membranes [Xomeritakis and Lin 98]. The modified CVD for the present work is schematically shown in Figure 5-2. The CVD system consisted of a reactor chamber made of quartz and a central tube made of dense α-Al2O3. The two sides of the CVD reactor were respectively connected to a vacuum pump and reactant containers. Prior to deposition, the Anopore membrane was carefully placed on one end of the dense alumina tube with the 20 nm pore side facing outward since we needed to modify the small pore side. The other end of the α-

Al2O3 tube was connected to the vacuum pump. Liquid trimethylaluminum (TMA)

(97% purity, Aldrich) in a stainless steel cylinder and distilled water in a Pyrex flask were used as precursors for the deposition of Al2O3 in the CVD process. These precursors and a He gas cylinder were connected directly into the CVD reaction chamber as shown in Figure 5-2.

164

(B)

(A) 20 nm Top layer 1-2 µm

Support 58 µm

500 nm 200-250 nm

(C) (D)

Figure 5-1. Images of straight pore alumina membrane. (A) Schematic of the straight pore structure and dimensions, (B) SEM top view showing the top layer of 20 nm pore size, (C) and (D) SEM cross sectional view showing the support side with 200 nm pore size.

165 Pressure Sensor Tube Furnace Valve

Anopore

Vacuum pump H2O TMA He

Figure 5-2. Schematic illustration of the atomic layer CVD

In this work CVD modification of Anopore membranes was conducted under two different schemes: cyclic CVD with residual pressure and cyclic CVD with purge. The specific experimental conditions in this work were adapted from conditions used to narrow the pore size of mesoporous γ-alumina membranes by CVD modification conducted in our lab. In the first scheme (with residual pressure), the

CVD system was evacuated to 1 mbar and slowly heated to the desired deposition temperature (180 oC). Then the system was isolated from the vacuum pump and water vapor was introduced into the reaction chamber. The pressure in the reaction chamber quickly reached about 32 mbar. After the Anopore membrane was exposed to water vapor for about 5 min the system was evacuated to 1 mbar for at least 10 min. TMA vapor was introduced into the reaction chamber whose pressure reached 27 mbar.

After 5 min exposure to TMA and system was evacuated to 1 mbar again. This completed one cycle of CVD of alumina. The cycle was repeated to grow additional layers of alumina. In this scheme, evacuation of system to 1 mbar after introduction

166 of each precursor allowed presence of the precursor molecules in an amount equivalent to 1 mbar in the membrane pores when the second precursor is introduced into the CVD system.

The second scheme (with purge) followed the same procedure as the first scheme except the use of a purge gas to lower or eliminate the presence of the precursor in the CVD system prior to introduction of the second precursor. The reaction chamber was evacuated to 1 mbar for 10 min after the introduction of each precursor. Helium, as a diluting gas, was introduced to the chamber to 1500 mbar and was left for 2 min. Then the system was evacuated to 1 mbar for at least 10 min before introducing the second precursor. The diluting step was done after the introduction of each precursor to the reaction chamber. The main difference between the two schemes is the controlled removal of residual precursor between the half reaction cycles.

5.2.2. Characterization

Helium and nitrogen permeance at different average pressures were measured by a steady state single gas permeation system to examine the microstructure of the

Anopore membranes before and after CVD modification. See appendix for detailed experimental setup and procedure. The permeation system is normally used for strong

2 mm thick ceramic membranes. As the Anopore membranes were thin and very fragile, extra care was taken during the permeation measurements in order to ensure integrity of the membrane. The 60 µm thick Anopore membrane was carefully fixed in a stainless steel permeation cell, with a stack of two O-rings (viton) put on each side of the membrane and the permeation cell tightened softly. Permeation data were

167 measured at small transmembrane pressure (6.3-11.2 kPa) by a careful control of the gas flow rate (240-360 ml/min) using a downstream needle valve.

The separation of water vapor from air by the membrane was conducted using the apparatus shown in Figure 5-3. Unmodified or CVD modified Anopore membranes were carefully fixed in a stainless steel permeation cell. A 100 ml/min mixture of dry air and wet air were introduced to the upstream side of the sample. Dry nitrogen sweep gas from a cylinder was passed, at a flow rate of 75 ml/min, in a cross flow fashion over the downstream side of the sample. Different upstream humidity was obtained by changing the flow rate of dry air and wet air, with a total flow rate kept about 100 ml/min. The temperature was maintained at about 25 oC. The relative humidity and oxygen content of the streams were respectively detected by a humidity sensor (Thermohygrometer, Cole-Parmer 37950-10) and an oxygen sensor (6000

Oxygen Analyzer, Illinois Instruments). The permeance of each species was calculated from the permeation flux of that species divided by its transmembrane partial pressure. Then the water to oxygen separation factor was calculated from the ratio of water to oxygen molar fraction in the outlet of downstream to that in the outlet of the upstream.

Change in pore microstructure due to CVD modification wase studied by environmental scanning electron microscopy (FEI XL-30 ESEM-FEG). Membrane samples were mounted on carbon double-sided adhesive tape with the 20 nm modified membrane surface facing upward. To improve specimen conductivity and reduce image shift, samples were painted with silver adhesive paste along the perimeter while retaining an uncoated part of the sample in the center for imaging in the gaseous mode. Images with magnifications up to 400K times were taken at 10 KV in pressure range 2-20 Torr.

168

Flow meter

To oxygen and moisture sensors Valve

3-way Oven Valve Membrane cell Water Sparger N Air 2

Figure 5-3. Water vapor separation system

169 5.3. Results and Discussion

5.3.1. Single gas permeation properties of CVD modified Anopore membranes

CVD modification was conducted on the 20 nm pore side of the asymmetric

Anopore membrane. The membranes were modified by different number of cycles (2 and 6) under the two different schemes as described in the experimental section.

Unmodified Anopore samples are referred to as zero time CVD modified membranes and samples modified with 2 and 6 CVD cycles are referred to as 2 and 6 time CVD modified membranes. Samples with CVD modification under the residual pressure scheme are referred to as CVD-residual and those modified under the He purge scheme are referred to as CVD-purge.

Helium gas permeation results of unmodified Anopore membrane and 6 time modified under residual pressure and gas purge schemes are shown in Figure 5-4. The permeation data presented in Figure 5-4 exhibit a linear relationship. Unmodified

Anopore membrane has a He permeance of about 1.6×10-4 mol/s.m2.Pa at 1 atm transmembrane average pressure. This value is almost 2 orders of magnitude higher than the He permeance values of 0.2 µm pore α-alumina and supported 4 nm pore γ- alumina membranes [Lin and Burggraaf 93]. This is attributed mainly to the low mass transfer resistance to He permeance exhibited by the thin (60 µm) asymmetric

Anopore composite membrane and also to its ordered structure with low tortuosity factor. The resistances of the Anopore membrane to gas permeance due to the 59 µm of 250 nm and the 1 µm top layer of 20 nm diameter pores are respectively much less than those exhibited by the 2 mm thick 200 nm pore diameter of α-alumina and the 5

µm thick 4 nm pore diameter of the γ-alumina top layer.

170

2.20

0 mod 2.10 6 CVD-Residual

.Pa) 6 CVD-Purge 2 2.00 mol/s.m -4 1.90

1.80

Permeance (10 Permeance 1.70

1.60 1.50 2.00 2.50 3.00 3.50 5 Average Pressure Pav (10 Pa)

Figure 5-4. Single gas helium permeance of unmodified and 6 time CVD modified Anopore membrane under the two different schemes of CVD modification.

As presented in Figure 5-4, the permeance values for unmodified, 6 time CVD modified with residual pressure, and 6 time CVD modified with purge are respectively 1.92×10-4, 1.79×10-4, and 1.70×10-4 mol/s.m2.Pa at ~ 2.5 atm average transmembrane pressure. The helium gas permeance decreases as a result of modification under the two schemes of CVD modification. This is obviously due to reduction of pore size of the membrane by the deposition of solid alumina inside the pores. As shown in Figure 5-4 the CVD modification under the purge scheme resulted in a more reduction of He permeance (11%) than that by the residual pressure scheme

(7%). Same trend has been observed by the 2 time CVD modified samples. But in the

2 time CVD modified membrane the He permeance trend lines are closer to each

171 other and the reduction in He permeance values after modification by the purge and residual pressure schemes are respectively 7% and 5% at 2.5 atm average pressure.

By comparing the reduction in permeance for both 2 and 6 time CVD modification under the same scheme, it can be generally concluded that the reduction in the pore size by the six CVD cycles is more than that by the two CVD cycles. The difference in reduction of permeance between the two CVD modification schemes for each sample reveals that the two schemes end up with deposits of different microstructure, as shown next with further analysis of the permeation data.

Helium permeance data (F/L) at different average pressures (Pav) shown in

Figure 5-4 can be regressed by a straight line (F/L)=α+βPav. The coefficients (α) and

(β) represent contributions from Knudsen and viscous flows respectively as:

ε r α = 1.06( ) p (5-1) κ L M w RT

ε r 2 β = 0.125( ) p (5-2) κ LηRT

where ε, κ, rp, L are the porosity, tortuosity factor (about 1 for the straight pore

Anopore membrane), average pore radius, and thickness of the membrane and η and

Mw are the viscosity and molecular weight of the permeating gas. The change in α and β due to CVD modification can be theoretically used to examine the change in the average pore radius (rp) (from the ratio of β/α) and porosity (ε) (from the value of α).

It should be noted that α and β values obtained from the permeance data represent the average values of the whole Anopore membrane (thick large pore support and thin 20 nm pore top layer). For the unmodified Anopore membrane, the resistance of permeation in the top layer is around 3 times higher than that of the support assuming

2 a viscous flow mechanism (resistance ~ thickness/dp ). After CVD modification,

172 which causes deposition in the top-layer, the ratio of mass transfer resistance for the top layer increases due to reduction in pore size of the top layer. Therefore the reduction in α and β after CVD modification is mainly due to the change in the pore structure of the membrane top layer.

The presence of pinholes in the tested membrane can be easily checked from the value of the slope (β) and the calculated average pore size. Pinholes are normally macroporous and have gas permeance governed by viscous flow mechanism.

Therefore, the presence of pinholes would increase the slope value (β) and normally end up with an average pore size larger than 1 µm. In this work, the absence of pinholes in the tested unmodified and modified membranes was ensured by the trend line of gas permeance and the obtained average pore size values which were all less than 250 nm.

The permeance of the unmodified sample shown in Figure 5-4 was regressed by

-4 2 -10 the straight line (F/L)=α+βPav where α=1.62×10 mol/s.m .Pa and β=1.25×10 mol/s.m2.Pa2. From the ratio of β/α=7.76×10-7 Pa-1 the average pore size of the unmodified membrane was estimated to be 205 nm. Table 5-1 summarizes α, β,

(β/α), and average pore size (dp) of the Anopore membranes before and after modification under the two schemes. As shown in Table 5-1, the average pore sizes of the starting unmodified membranes obtained from the (β/α) ratio are slightly different. For membranes modified under the same scheme the increase in the number of CVD cycles causes more reduction in the (β/α) and hence more reduction in the average pore size (dp). This demonstrates that the cyclic CVD is narrowing the pores in the membrane 20 nm top-layer by deposition of solid alumina oxide inside the pores.

173

Table 5-1. Helium permeation properties for the Anopore membranes after CVD modification

Residual pressure With purge 2 CVD 6 CVD 2 CVD 6 CVD before after before after before after before after α (10-4 mol/m2 s Pa) 1.54 1.53 1.55 1.54 1.64 1.49 1.62 1.47 β (10-10 mol/m2 s Pa2) 1.45 1.21 1.45 1.17 1.43 1.21 1.25 0.97 β/ α (10-7 Pa-1) 9.40 7.88 9.35 7.59 8.73 8.14 7.76 6.59 dp (nm) 248 208 246 200 230 215 205 174

174 The percentage reduction in the porosity and average pore size by cyclic CVD modification under the two schemes are listed in Table 5-2. For 2 time CVD modified membranes under the residual pressure scheme, the percentage reduction in porosity

(0.62%) is much smaller than reduction in the average pore size (16%). This indicates that the presence of some residual pressure between the successive additions of precursor causes high reduction in the average pore size of the membrane top-layer while almost maintaining its porosity. Higher number of CVD cycles under the residual scheme causes the same trend but with slightly more reduction in porosity

(0.98%) and further more reduction in the average pore size (19%).

Modification under the purge scheme, on the other hand, resulted in a different microstructure. As shown in Table 5-2, two cycles of CVD with purge resulted in higher reduction in the porosity (9.23%) compared to the residual scheme (0.62%) as well as less reduction in the average pore size (7%) compared to that under residual scheme (16%). Similarly, 6 cycles of CVD cause additional reduction in the pore size

(15%) and a porosity value (8.5%) that are respectively lower and higher than the corresponding values obtained under the residual pressure scheme. This reveals that purging the pore opening after each addition of precursors using helium allows the alumina to deposit in such a way that reduces the porosity more effectively than the pore size.

Table 5-2. Percentage reduction in average microstructural values due to CVD modification residual pressure with purge 2 CVD 6 CVD 2 CVD 6 CVD ε 0.62 0.98 9.23 8.51

dp 16 19 7 15

175 5.3.2. Multi-gas permeance and separation properties of CVD modified Anopore membranes

The data of permeation and separation of water vapor from oxygen (or air) were measured for two purposes. First, these data are needed for selection of industrially important inorganic membranes with high permeance and good selectivity for removal of water from air [Pan et al. 99]. For example, inorganic membranes with controlled pore microstructure could provide unprecedented water vapor separation properties with high selectivity at significantly lower cost compared to conventional energy-intensive technologies. Second, these data resemble those from a permporosimeter useful for characterization of the pore structure of the top layer of composite membranes [Cooper and Lin 02]. Figure 5-5 shows the oxygen and water permeance values for the unmodified Anopore membrane and the well known sol-gel derived 4 nm pore γ-alumina membrane prepared in a previous work [Pan et al. 99].

For the γ-alumina membrane, the water vapor permeance decreases slightly with increasing relative humidity (RH). Oxygen permeance, on the other hand, remains almost constant at 3×10-7 mol/s.m2.Pa with RH values less than 60%, then it decreases sharply to 0.7×10-7 mol/s.m2.Pa at 90% RH. This sharp decrease in oxygen permeance is typically due to capillary condensation of water vapor inside the mesoporous 4 nm pores of γ-alumina [Cooper and Lin 02, Ulhorn et al. 92].

For a mixture of condensable vapor and non-condensable gas permeating inside mesopores, vapor may condense inside the mesopores at pressures lower than the saturated vapor pressure. As the RH of condensable vapor increases its condensate starts to block the mesopore and hence reduces the flow of the non-condensable gas.

Here water vapor was used as the condensable vapor and oxygen (in the air) as the non-condensable gas. Utilizing Kelvin equation, the gradual decrease of oxygen

176 permeance from 60 to 90% RH indicates the presence of a distribution in the pore size of γ-alumina membrane ranging almost between 2-5 nm.

The oxygen and water vapor permeances for unmodified Anopore membrane are respectively 7 and 4 times higher than the corresponding values for the γ-alumina membrane. The water permeance value decreases from 6.9×10-6 to 2×10-6 mol/s.m2.Pa from 10 to 90% RH. Unlike the case of γ-alumina, oxygen permeance for the Anopore remained constant at almost 1.7×10-6 mol/s.m2.Pa over the studied range of RH. This obviously is due to the large pores (20 nm) of the Anopore membranes where the amount of water vapor even at 90%RH was not sufficient to block the pores for the passage of oxygen. This indicates that capillary condensation using water vapor is much effective for pores in the low range of mesoporous size (2-4 nm).

1000 .Pa) 2 100 mol/s.m -8 -8 0 CVD Anopore, H2O 10 0 CVD Anopore, O2

Gamma-alumina, H2O

Permeance (10 Permeance Gamma-alumina, O2 1 0 20406080100 Relative Humidity

Figure 5-5. Permeance of oxygen and water vapor through unmodified Anopore

and γ-alumina membranes.

177 Oxygen permeance value obtained here (1.7×10-6 mol/s.m2.Pa) is 30 times smaller than pure oxygen permeance (56×10-6 mol/s.m2.Pa) estimated from the pure

He permeance values for the unmodified Anopore membrane given in Table 5-1 using

Knudsen permeation relation (Eq. 5-1). Helium permeance was obtained through single gas permeation while oxygen permeance was measured under multiple gas permeation. The driving forces for the single gas (total pressure difference) and multiple gas (partial pressure difference) permeation were almost equal (0.11 atm).

Therefore the driving force has no effect on the low value of oxygen permeance. The presence of other permeating components, like water vapor, is expected to interact with oxygen and thus lowers its permeance. Moreover, the presence of back flow of

N2 from the N2 sweep gas flow at the membrane downstream is also expected to lower the oxygen permeance [Burggraaf et al. 98, van de Graaf et al. 98 ].

The binary oxygen and water permeance values for the Anopore modified by two and six CVD cycles under the residual pressure scheme are presented in Figure 5-

6. Water permeance for both samples decreased slightly due to the modification, which means that some water is being adsorbed on the pore walls. Oxygen permeance for each modified membrane was constant (1.5−1.6×10-6 mol/s.m2.Pa) over the studied range of RH. The percentage reduction in this value for the two and six CVD cycles were respectively 8 and 10% compared to the unmodified oxygen permeance.

These observations reveal that the pore size reduction and pores structure properties for CVD modification under residual pressure are not favorable for capillary condensation of water. For that reason the permeance curves for the unmodified and modified membranes almost coincide. A possible pore structure for which such observations can be obtained is a pore channel with high porosity.

178 1000 0 CVD, H2O 2 CVD, H2O .Pa)

2 6 CVD, H2O 0 CVD, O2 2 CVD, O2 mol/s.m

-8 -8 6 CVD, O2 Permeance (10 Permeance

100 0 20406080100 Relative Humidity

Figure 5-6. Oxygen and water permeance for Anopore membranes modified under

the residual pressure scheme.

Figure 5-7 shows the oxygen and water permeance for the Anopore membranes modified by two and six cycles of CVD under the He purge scheme. Modification by two CVD cycles with purge slightly reduced the oxygen and water permeance values compared to the unmodified membrane. On contrary, the 6 time CVD modified membrane exhibits a significant improvement in the water/oxygen separation properties. Modification by 6 cycles under purge reduced the oxygen permeance by

40% and increased the water permeance by at least seven times. This means that the final pore structure after six CVD cycles under purge is favorable to water permeance.

A possible explanation is that the pore size has been narrowed such that the condensation of water vapor is enhanced but not to the limit of blocking the pores.

179 This is confirmed by the increase in the water permeance and the decrease of oxygen permeance to a lower constant value without being dropped sharply at high RH. These observations reveal that the structure of the modified pores of the 20 nm pore

Anopore membrane has cylindrical shape with low porosity in which capillary condensation is favorable.

Separation factor of water/oxygen for the unmodified and modified samples under both schemes are shown in Figure 5-8. The unmodified Anopore membrane has a separation factor of 2-4 that is constant over the studied range of RH. Modification by two and six CVD cycles under the residual pressure had no effect on separation factor of the membrane. As mentioned previously, this may be attributed to the final highly porous structure, which is not favorable for condensation of water vapor.

Another explanation is that the CVD modification under the residual pressure did not take place uniformly on the whole membrane surface. There may still some large pores (20 nm) that give separation properties close to the unmodified membrane. Two time CVD modification under the purge scheme gives separation factor in the vicinity of the unmodified sample. The separation factor for the six CVD cycles under the purge scheme was 20-25 in the range of 20 to 70% RH, which is almost 1 fold higher than the unmodified samples. The separation factor for six CVD modified sample is about 5 times higher than the 4 nm γ-alumina membrane in the same range of RH

[Coopoer et al. 02].

180 10000 .Pa) 2 1000 mol/s.m -8 -8

100

0 CVD, O2 0 CVD, H2O

Permeance (10 Permeance 2 CVD, O2 2 CVD, H2O 6 CVD, O2 6 CVD, H2O 10 0 20406080100 Relative Humidity

Figure 5-7. Oxygen and water permeance for Anopore membranes modified under the purge scheme.

100.0

10.0

1.0

0 CVD 2 CVD-Residual 2 CVD-Purge 6 CVD-Residual

Separation Factor (Water/Oxygen) Factor Separation 6 CVD-Purge 0.1 020406080100 Relative Humidity

Figure 5-8. Water/oxygen separation factor for unmodified and CVD modified

Anopore membranes under the two schemes

181 5.3.3. Pore structure and mechanisms of CVD modification

The above permeation and separation data have indicated two different microstructures of Anopore membrane modified by CVD with the residual pressure and purge schemes. The residual pressure scheme is more likely to narrow the pore diameter of the membrane top-layer while maintaining its high porosity. This suggests that alumina was deposited inside the pores of the modified layer in a cluster-like or fractal structure. For such structure, the porosity is high and the average pore size measured by single gas permeation is reduced due to random deposition of alumina inside the pore. Modification under the purge scheme exhibited a greater reduction in the porosity than the pore size. This can be explained by a cylindrical pore narrowed by deposition of alumina on the pore wall in an atomic layer fashion. Schematic representation of the final pore microstructures obtained via the residual pressure and purge schemes are shown in Figure 5-9.

(A) (B)

R R Ro

20 nm

Figure 5-9. Schematic representation of radial and sectional axial views of different pore microstructures obtained under CVD modification: (A) Fractal microstructure obtained via residual pressure scheme, (B) Atomic layer deposition cylindrical microstructure obtained via purge scheme.

182 The fractal structure obtained by the residual pressure scheme would be caused by combined homogeneous and heterogeneous reactions. Allowing a residual pressure

(1 mbar) of precursor to exist in the gas phase of pore after each addition of precursors causes the precursors to exist both on the pore wall and in the pore gas phase. This would cause the deposition of alumina on the pore wall due to heterogeneous solid/gas phase reactions and randomly inside the pore volume due to homogeneous vapor phase reactions. This microstructure is consistent with CVD modification on γ-alumina membranes under the same conditions [Cooper and Lin

02]. The fractal structure is desired in separation applications such as separation of gas molecules based on size and shape selectivities.

Purging the CVD reactor after each addition of precursors causes the removal of excess precursors from the gas phase in the pores. Therefore only heterogeneous reactions take place on the pore wall, which causes to narrow the cylindrically shaped pores by deposition of alumina oxide in an atomic layer fashion. The cyclic CVD modification under purge gas represents a simple and cost effective alternative for modification of pore size in a fashion similar to the atomic layer CVD. This simple

CVD method does not require very low vacuum pressure thus greatly simplifying the

CVD system. Modification of Anopore membrane by the purge scheme has resulted in a narrowing of the 20 nm cylindrically shaped pores. Such structure represents an alternative for obtaining a supported hexagonally ordered mesoporous film with pore vertical to the support that is un-attainable up to this moment. The narrowed cylindrical pore structure obtained by the purge scheme is desirable and may find new applications in membrane reactor applications as molecular extruder. Cylindrical pore structure was more effective in separation of water vapor from oxygen compared to the fractal structure, which is more desirable for separation of gas mixtures rather than

183 vapor/gas mixtures. This is because the separation of water vapor is based on the capillary condensation, which is favored in cylindrical pores like those obtained by the purge scheme.

It should be noted that direct verification of the pore structures of the Anopore membrane top layers discussed above is very difficult, if not impossible. Nitrogen or mercury porosimetry is not suitable for characterization of the top layer pore structure of the Anopore membranes for the following reasons. Anopore membranes have a low surface area (<1 m2/membrane). This requires a large number of identically modified membranes (at least 10) in each experiment for reliable measurement of its pore structure. Such a large number of identically modified membranes were not available in our study. Hg porosimetry is often used to characterize macroporous materials (pore size larger than 100 nm). For Anopore membranes with pores much smaller than 100 nm, Hg porosimetry should be operated at extremely high pressure, which may destroy the pore structure of the fragile Anopore membrane. Furthermore, the pore structure data obtained by either nitrogen or mercury porosimetry for the whole Anopore membrane would represent the pores for the support, not the much thinner top layer. High magnification (400K times) SEM images of the top layers of the 2 time CVD modified membranes with and without purge are compared in Figure

5-10. As shown, SEM is also not sufficient to verify the microstructure difference of these two membrane samples.

184

50 nm

50 nm

Figure 5-10. ESEM images of Anopore modified by two CVD cycles under (a) residual pressure scheme (top) and (b) purge scheme (bottom) at 400K magnification.

185 5.4. Conclusions

A simple cyclic CVD method can be used to modify 20 nm straight-pore

Anopore alumina membranes for obtaining smaller pore membranes with different microstructures desirable for several applications. Conducting CVD with residual pressure (1mbar) after each precursor application caused greater reduction in the pore size than porosity suggesting the deposition of alumina in a fractal structure. Cyclic

CVD modification using a purge after each addition of precursors exhibited higher reduction in the porosity than the average pore size suggesting that deposition of alumina on the pore wall in the atomic layer fashion. The fractal structure is caused by combined homogeneous and heterogeneous reactions taking place respectively at the pore wall and randomly inside the pore wall. The narrowed cylindrical structure is caused by homogeneous reactions taking place at the pore wall. Membranes modified under the purge scheme exhibited the best water vapor/oxygen separation properties.

The six time CVD modified membrane had water permeance of 2.2×10-5 mol/s.m2.Pa with water to oxygen separation factor of 25, about 10 times higher than the well studied 4 nm pore γ-alumina membrane.

186 5.5. Reference

Burggraaf A.J., Vroon Z.A.E.P., Keizer K., Verweij H., Permeation of Single Gases

in Thin Zeolite MFI Membranes, J. Membr. Sci., 144 (1998) 77.

Cooper C.A., Lin Y.S., Microstructural and Gas Separation Properties of CVD

Modified Mesoporous Gamma-Alumina Membranes, J. Mem. Sci., 195 (2002)

35.

Crawford G.P., Steele L.M., Ondris R., Iannacchione G.S., Yeager C.J., Doane J.W.,

Finotello D., Characterization of the Cylindrical Cavities of Anopore and

Nucleopore Membranes, J. Chem. Phys., 96 (1992) 7788.

Furneaux R.C., Rigby W.R., Davidson A.P., The Formation of Controlled-Porosity

Membranes from Anodically Oxidized Aluminum, Nature, 337 (1989) 147.

George S.M., Ott A.W., Klaus J.W., Surface Chemistry for Atomic Layer Growth, J.

Phys. Chem., 100 (1996) 13121.

Lin Y.S., Burggraaf A.J., Experimental Studies on Pore-Size Change of Porous

Ceramic Membranes after Modification, J. Membr. Sci., 79 (1993) 65.

Ott A.W., Klaus J.W., Johnson J.M., George S.M., Modification of Porous Alumina

Membranes using Al2O3 Atomic Layer Controlled Deposition, Chem. Mater., 9

(1997) 707.

Pan M., Cooper C., Lin Y.S., Meng G.Y., CVD Modification and Vapor Gas

Separation Properties of Nanoporous Alumina Membranes, J. Membr. Sci., 158

(1999) 235.

Ulhorn R.J.R., Keizer K., Burggraaf A.J., J. Membr. Sci., 66 (1992) 259. van de Graaf J.M., Kapteijn F., Moulijn J.A., Methodological and Operational

Aspects of Permeation Measurements on Silicalite-1 Membranes, J. Membr. Sci.,

144 (1998) 87.

187 Xomeritakis G., Lin Y.S., CVD Synthesis and Gas Permeation Properties of Thin

Palladium/Alumina Membranes, AICHE J., 44 (1998) 174.

188 CHAPTER 6

SUMMARY

A brief review on template-assisted synthesis of mesoporous inorganic materials with ordered microstructures in Chapter 1 suggests that macroscopic shape plays a significant role in the properties and applications of this group of materials. Hexagonal packing of uni-dimensional pores of MCM-41 type is one of the symmetries that has been widely prepared into several macroscopic morphologies including membranes. Failure to perfectly align the hexagonal pores normal to the membrane surface has limited the use of this unique microstructure in separation and molecular reaction applications. The major objective of this thesis is to fabricate hexagonally ordered materials with preferred pore orientation and macroscopic shapes. It primarily focuses on understanding the microstructure of a template-assisted grown hexagonal mesoporous silica material with perfect pore alignment and the possible extension of this microstructure into large scale membrane shape. It also investigates the use of chemical vapor deposition (CVD) as an alternative approach for obtaining inorganic membranes with controlled microstructures attractive in separation and reaction applications.

In Chapter 2, synthesis of ordered mesoporous silica material with the surfactant- templated assisted quiescent two-phase acidic approach is presented and the application of hexagonally ordered mesoporous silicate particles in olefin polymerization is demonstrated. For the first part, silica source, temperature, growth time, acid type and content were carefully varied to study their effect on the shape and microstructure of the product. Hexagonal mesoporous silica fibers, twisted ropes, hollow and solid spheres,

189 gyroids, regular and irregular constructs, ordered mesoporous films and amorphous films are obtained at the interface between water-silica source and as precipitates under different growth conditions. Silica fiber morphology grow as the major product at room temperature using TBOS as silica source and HCl acid with 2−8.4 vol%. At these conditions, the slow and uni-dimensional diffusion of silica source through the interface leads to linear growth of silica as fibers. Silica yield increases linearly with time and the fibers diameter becomes larger with broader distribution suggesting that fibers grow from the bottom and coalesce to form larger fibers. Growth at higher temperature (> RT), low acid content (< 2%), high acid content (> 8.3%) or with other acid counterions (HNO3 and H2SO4) alters the linear diffusion of silica source, inhibits silica condensation, significantly speeds up the silica condensation or interrupt the axial growth of silica to end up with non-fibrous shapes.

In the second part, extrusion polymerization of highly-ordered polyethylene (PE) fibers is demonstrated on MCM-41 type ordered mesoporous silica-supported titanocene catalyst. The steric effect of the 3 nm straight pores of the support prevent polymer chains from folding into amorphous structure but rather template the growth of nascent PE fibers with extended-chain and crystalline structure. SEM investigation reveals three step-model for PE fiber formation: (a) 60 nm diameter PE nanofibrils with extended-chain crystalline are formed by extrusion polymerization, (b) the fibrils aggregate into PE microfibers with

1−30 µm diameters and (c) the microfibers entangle and aggregate to form the final PE fiber bundles. The long polymer fibers allowed the measurement of their mechanical properties for the first time. They exhibit strong mechanical properties, compared to commercial PE fibers, with high tensile strength (0.3−1 GPa), low tensile modulus

190 (3.0−7.0 GPa) and high elongation at break. Additional tensile drawing could increase the strength and modulus of the microfibers. Polymerization results demonstrate the need for a hexagonally ordered mesoporous membrane with vertical pores as a support for this application to be operated in continuous mode and commercialized.

In Chapter 3, gas diffusion and microstructural properties of mesoporous silica fibers (MSF) prepared at optimized conditions of Chapter 1 are presented. This work is the first research effort on measurement of gas diffusivity of ordered mesoporous silica using transient gravimetric approach and the macroscopic evaluation of the pore length and orientation. Weight uptake of CO2 and C2H4 gases on MSF achieves equilibrium (~

25 mg/g) in almost 9 min at relatively different rates revealing a high diffusion resistance induced by the long straight nanopores. Effective gas diffusion coefficient is approximately 2×10-6 cm2/s obtained from regression of gas uptake data with diffusion in plain sheet model. Gas diffuses in the straight pores of MSF by a combination of

Knudsen and surface mechanisms with respective contributions of 60 and 40%. The high surface contribution suggests a smooth pore surface. Real gas surface diffusivity in MSF is in the order of 10-5 cm2/s. This value is comparable to values of conventional mesoporous membranes with disordered structures measured by steady state approaches.

The tendency of MSF to adsorb gases leads to discrepancy in the surface diffusivity values measured by steady and unsteady approaches. Real surface diffusivity in ordered mesoporous materials is believed to be at least 100 times larger than those of disordered mesoporous materials. The silica fibers exhibit a high tortuosity factor (κ) of 2010 estimated from Knudsen diffusivity. The tortuosity estimated from tortuosity factor

191 2 2 [κ=2010=τ =(lp/Lf) ] indicate that the mesopores are 45 times longer than the fiber length and align in a helical orientation with 1.05 µm pitch.

In Chapter 4, fabrication of mesoporous silica membranes by a novel counter diffusion self assembly (CDSA) approach is presented. The CDSA approach is an extension to the synthesis of mesoporous silica fibers by the interfacial approach but with a ceramic support placed at the interface. In this approach silica and water/surfactant precursors counter diffuse from their phases through the support and self assemble to deposit mesoporous silica plugs within the support pores. The quality and continuity of silica growth depends strongly on the surface chemistry and pore size of the support.

Hydrophilic supports of 0.2−5 µm pore sizes inhibit the diffusion of the hydrophobic silica source and yield a low quality silica membrane full of defects. Modification of the support pore walls with a hydrophobic agent strongly enhances the transfer of silica source and growth of continuous and defect-free silica-surfactant plugs. Gas permeation properties exhibit a Knudsen type permeance behavior confirming the mesoporous structure of the silica membrane. Counter diffusion of precursors yields a ~ 0.55 mm thick deposition of silica within the support pores. Such membrane displays strong mechanical properties and can find applications in olefin polymerization and protein separation applications.

In Chapter 5, a simple cyclic CVD modification of 20 nm straight-pore alumina membranes for obtaining smaller pore membranes with different microstructures is presented. The microstructure of alumina deposition within the membrane pores can be controlled by careful management of precursor introduction schemes. Conducting CVD with residual pressure after introduction of each precursor caused greater reduction in the

192 pore size than porosity suggesting the deposition of alumina in a fractal structure. Cyclic

CVD modification using a purge after each addition of each precursors exhibited higher reduction in the porosity than the average pore size suggesting that deposition of alumina on the pore wall in the atomic layer fashion. The fractal structure is caused by combined homogeneous and heterogeneous reactions taking place respectively at the pore wall and randomly inside the pore wall. The narrowed cylindrical structure is caused by homogeneous reactions taking place at the pore wall. Membranes modified under the purge scheme displayed the best water vapor/oxygen separation properties. It exhibits a water permeance of 2.2×10-5 mol/s.m2.Pa with water to oxygen separation factor of 25 which is 10 times higher than the well studied 4 nm pore γ-alumina membrane.

193 CHAPTER 7

AREAS OF FUTURE RESEARCH

In view of results obtained in this work, a number of recommendations can be made with objective to further improve our current state understanding of controlled- morphological synthesis of ordered mesoporous materials and to bring them into level of commercial exploitation. Suggested areas of future research involve topics related to microstructural analysis of ordered mesoporous silica fibers and the corresponding mesoporous silica membrane obtained by the novel counter diffusion self assembly approach discussed in this thesis and their potential use in applications.

Mesoporous silica fibers produced in this work by the quiescent interfacial approach was observed at narrow synthetic conditions only with HCl acid. Silica fiber growth is always accompanied by growth of additional morphologies mostly gyroids.

Three surface textures of silica fibers were observed in our work: smooth surface fibers, fiber with folded-sheets surface and fiber aggregates. The smooth surface texture has been reported in literature with homogenously mixed systems and is suggested to have straight pores across the fiber axis. The helical pore orientation reported in literature and used in this work most probably belongs to the folded-sheet fiber texture. Our suggested mechanism of fiber growth corresponds to the aggregated texture of fibers. It is apparent from these observations that several mechanisms are involved in the growth of silica fibers. Further analysis of the pore alignment in these different surface textures by TEM is recommended to advance our understanding of fiber formation mechanism.

194 Additional work in this area is to identify the effect of acid type and the counterion on the microstructure and morphology of silica product. Counterion study on silica morphologies performed in this work was based on samples grown at relatively different times. Therefore, it was not possible to give a representative comparison of their properties. It is recommended to prepare mesoporous silica material at different acid types, e.g., HCl, H2SO4, HNO3, and HBr (if possible) at comparable concentrations and growth times close to a month and to study their microstructure by XRD, nitrogen porosimetry and electron microscopy (SEM and TEM). The effect of concentration for acids other than HCl should be also addressed. Current view on acid is limited to its effect on the rate of hydrolysis and condensation reactions. This study could identify the role of counterions in controlling the growth of specific macroscopic morphologies.

The diffusion analysis and macroscopic evaluation of the pore length and alignment of mesoporous silica fibers in this work was based on unsteady state weight uptake approach. Due to the nature of physical adsorption of gases, it was not possible to measure the weight uptake at temperatures above room temperature with this approach.

Therefore we were unable to study diffusion behavior at various temperatures, directly identify the surface diffusion and Knudsen contribution without a need for a model, or able to estimate the activation energy of diffusion. Moreover, the transient approaches normally exhibit diffusion coefficient smaller than those by steady state approaches due to time elapsed for adsorption of gas molecules on the internal surface of the pores. It is recommended to study the diffusion properties of mesoporous silica fibers by the steady state membrane technique. To do so, a permeation cell with silica fiber properly oriented and sealed, probably with epoxy resin, should be prepared. Subsequently, permeation of

195 several gases (N2, CO2, He, H2, and CH4) can be conducted at temperature range of 27-

300 ºC utilizing a steady state gas permeation setup. Adsorption equilibrium isotherms of these gases on MSF samples in the same temperature range are measured by the gravimetric method. Gas diffusivities and their contributions can be directly measured and utilized to analyze the pore microstructure in the same manner of this work.

In the area of CDSA growth of silica membranes, three major research aspects are recommended. First is to study the macro- and micro-structure of silica plugs grown within the support pores. Macroscopic features include deposition location and thickness.

This can be done by SEM and EDS techniques. Microscopic features include the pore size and distribution, and the internal pore orientation. Gas permeation in this thesis confirms the mesoporous structure of silica plugs. Precise evaluation of the pores size and distribution can be studied by permporometry technique using a binary mixture of condensable and non-condensable gases. Pore alignment can be measured by TEM technique. This part is important for the complete evaluation of CDSA approach for synthesis of nanostructured membranes.

The second part involves application of CDSA on supports of straight pores. This will increase the potential to obtain mesoporous silica membranes with vertical pores if silica fiber grow as plugs in the straight pores of the support. Anopore alumina support is a good candidate but it is unstable under the acidic conditions of the CDSA in-situ synthesis. It cannot be used in the CDSA approach unless the growth conditions are modified to fabricate silica fibers at pH between 5−9. Other possibility is to use acid- stable ceramic or metallic support with straight pore orientation such as chemical itched silicon membranes commercially available for MEMs applications.

196 The final part is to investigate the potential use of CDSA silica membrane for applications after proper characterization of its internal microstructure. Two main applications are proposed: olefin polymerization of polymer nanofibers via extrusion polymerization demonstrated in Chapter 6, and protein separation. Both applications need mechanically strong membrane for efficient performance. Extrusion polymerization within mesoporous silica particles is a batch process and causes silica particle fragmentation. CDSA silica membrane is mechanically strong due to growth of ~ 0.5 mm thick mesoporous silica plugs within the support pores. The use of such strong membrane in extrusion polymerization will allow continuous production of polymer fibers with high yield and reduce the fragmentation problem.

197 APPENDIX A

EXPERIMENTAL PROCEDURES

A1. Preparation of Mesoporous Silica Fibers (Chapter 2)

Molar ratio: 1 TBOS: 0.5 CTAB: 58.4 HCl: 2000 H2O at 27 ºC for 4−14 days

1. Dissolve 3.99 g of cetyltrimethylammonium bromide surfactant (CTAB, Aldrich) in

water under 150 rpm magnetic stirring for 15 min to form a white solution.

Subsequently add 238 ml of HCl (6M) slowly under mixing until the solution

becomes clear.

2. Stop mixing and add 7.92 ml TBOS (Aldrich) on the top of the CTAB-H2O-HCl

mixture as thin liquid layer.

3. Age the mixture for more than 4 days without stirring. The fibers usually start to

grow after two days from the organic/water interface towards the water phase. More

aging time will give more fibers with better structure. Growth for 14 days will give

fibers with a typical XRD pattern.

4. After aging, fibers will be clearly grown with very thin (hardly seen) film at the

organic/water interface and some precipitated solids at the bottom of the beaker.

The fibers are carefully collected either by using pipette or direct pouring into

another beaker.

5. Wash the product several times by distilled water using vacuum filter. The product

is then dried in air at room temperature for 1-2 days.

6. Calcine the product at 540 oC for 6 hours using 1 oC/min heating and cooling rates.

198 A2. Preparation of MCM-41 (Chapter 2)

Molar ratio: 1 TEOS: 0.123 CTAB: 29.38 NH4OH: 506 H2O at 27 ºC

1. Mix 646 ml NH4OH (30 wt %) with 274 ml distilled water.

2. Add 3.29 ml of CTAB (Aldrich) to the solution under mixing. Slowly heat the

mixture to 50 ºC to obtain a clear mixture.

3. Cool down the solution to room temperature then add 16 ml of TEOS.

4. Age the slurry for 6 h. The solution becomes white after almost 2 h.

5. Wash the product with distilled water several times and dry at room temperature for

1 day then calcine at 550 ºC for 6 h at 1 oC/min heating and cooling rates to remove

the surfactant.

A3. Transient Weight Uptake Measurement (Chapter 3)

On silica fibers using CO2 and C2H4 gases at 1 atm and 27 ºC

1. Load 200-300 mg of the silica fibers to a light weight 200 mg Aluminum pan then

attach it to the arm of the microbalance as shown in Figure 3-1 using a light

platinum wire (Gauge 36, Fisher Scientific).

2. Replace the Pyrex balance tube and make sure the sample pan is not touching the

inner surface of the tube. Then mount the tubular furnace outside the Pyrex tube in

a way the sample pan is at its center.

3. Connect the Pyrex tube to the gas feed system. Heat the sample pan at 120 oC by

means of heating controller under 100 ml/min flow of helium purge for few hours

to desorb physically adsorbed water. Start recording the sample weight and keep the

process until the sample weight becomes constant.

199 4. When the weight becomes constant, stop recording the weight and cool down the

sample to the desired temperature (e.g., room temperature) by re-setting the

temperature controller under the same flow of the purge gas.

5. Introduce the adsorption gas (CO2 or C2H4) at flow rate of 100 ml/min by switching

from the helium gas using a crossover valve and start recording the weight uptake

data on the data acquisition system. Make sure the flow rate of the adsorption gas is

fully achieved before switching the crossover valve to reduce problems of gas

dispersion and delay. Always start recording data 60 sec earlier to introduction of

adsorption gas then subtract this value from the final time.

6. Wait at least 30 min to achieve the equilibrium weight uptake then stop recording

the data.

7. After achieving equilibrium, switch back the crossover valve to the helium purge

gas at 100 ml/min to desorb the gases for 1-2 h.

A4. Equilibrium Weight Uptake Measurement (Chapter 3)

On Silica fibers using a binary mixture of He and CO2 or C2H4 gases with partial

pressures between 0 to 1 atm at 27 ºC

1. Load 200-300 mg of the silica fibers to a light weight 200 mg Aluminum pan then

attach it to the arm of the microbalance as shown in Figure 3-1 using a light

platinum wire (Gauge 36, Fisher Scientific).

2. Replace the Pyrex balance tube and make sure the sample pan is not touching the

inner surface of the tube. Then mount the tubular furnace outside the Pyrex tube in

a way the sample pan is at its center.

200 3. Connect the Pyrex tube to the gas feed system. Heat the sample pan at 120 oC by

means of heating controller under 100 ml/min flow of helium purge for couple of

hours to desorb physically adsorbed water. Start recording the sample weight and

keep the process until the sample weight becomes constant.

4. When the weight becomes constant, stop recording the weight and cool down the

sample to room temperature by re-setting the temperature controller under the same

flow of the purge gas.

5. Introduce a 100 ml/min binary mixture of (adsorption gas+He) at zero atm partial

pressure of the adsorption gas (0 ml/min adsorption gas + 100 ml/min He) and start

recording the weight uptake data on the data acquisition system and wait at least 1 h

to achieve the equilibrium weight.

6. Increase the adsorption gas partial pressure to 0.2 atm by changing the feed flow

rates to (20 ml adsorption gas+80 ml He) and wait until the weight stabilizes.

7. Increase the adsorption gas partial pressure to 0.4 atm by changing the feed flow

rates to (40 ml adsorption gas+60 ml He) and wait until the weight stabilizes.

8. Increase the adsorption gas partial pressure to 0.6 atm by changing the feed flow

rates to (60 ml adsorption gas+40 ml He) and wait until the weight stabilizes.

9. Increase the adsorption gas partial pressure to 0.8 atm by changing the feed flow

rates to (80 ml adsorption gas+20 ml He) and wait until the weight stabilizes.

10. Increase the adsorption gas partial pressure to 1 atm by changing the feed flow rates

to (100 ml adsorption gas+0 ml He) and wait until the weight stabilizes.

11. Stop recording the data. From the weight data, calculate the equilibrium weight

uptake value corresponding to each gas partial pressure. The Weight uptake vs.

201 partial pressure plot can be used to obtain the adsorption equilibrium constant from

the slope.

A5. Preparation of α-Alumina Supports (Chapter 4)

1. Mix 10 g of the alumina powder (Alcoa A15 or A16 SG) thoroughly with 0.8 ml of

distilled water (as a binder) with a quartz mortar and pestle until the mixture

becomes homogeneous. This batch yields approximately 5 discs. If more supports

are required, repeat this step accordingly. Prepare the supports from the powder

mixture right away and do not leave it for the next day. Otherwise, the water will

evaporate and the pressed powder will fail to get intact.

2. Put 2.1 grams of the mixture in a stainless steel support mold and press at 1000 psi

(corresponding to 5000 Ib) for 1 minute using a pressing machine (Carver Inc.) then

flip the mold upside down and press at 3800 psi (20,000 Ib) for 3 minutes. Release

the pressure and remove the green (pressed) support. Repeat step 2 until all the

powder mixture is consumed.

3. Dry the supports in oven for 2 days at 40°C and 40 % relative humidity to remove

the water binder.

4. Put the supports carefully on a ceramic porous plate and insert into a high

temperature sintering furnace. Sinter the supports using the following program:

heating ramp to 600 °C at 60 °C/hr, then heating ramp to 1260 °C at 96°C/hr, then

cooling ramp to 75 °C at 96 °C/min, then heating ramped to 1150 °C at 60 °C/hr,

dwell at 1150 °C for 30 hrs, and finally cooling ramped to 200 at 60°C/hr. Always

keep the furnace at least at 200 ºC to extend the workability of the heating elements.

202 5. Polish the supports using a Metaserv 2000 grinder/polisher (Buehler). Polish the

side of the support to be used with 500 grit polishing paper by hand by gentle

holding on the polisher for 1 min minute. Rotate the support by 90° and polish for

another minute. Each support is rotated 4 times and is polished for 4 minutes.

Repeat the same polishing steps with 800 and 1200 grit polishing papers to obtain

smoother surface. Wash the polished supports with distilled water and dry them

for two days at 40°C.

A6. Preparation of 1M Boehmite Sol (Chapter 4)

1. In a 3-neck flask, heat 1 lit of distilled water to 70−90 °C while stirring using

magnetic bar.

2. Add 1 mol of aluminumtrisecbutoxide (ALTSB, Janssen, Mw=246.3 g/mol, 97%

purity) to the warm water in several sequential steps using a graduated cylinder or

plastic syringe while gradually increasing the stirring speed. Try to minimize

contacting the humidity-sensitive ALTSB with air.

3. Keep the solution at 90 °C for 1 hour.

4. Add 70 ml HNO3 (1 M) into the solution. Shutoff heating and slow down the

stirring speed.

5. Attach the 3-neck flask to a condenser and reflux at 90 °C overnight.

6. In some cases an oil-like layer may present on top of the Boehmite sol due to

impurities in ALTSB or from reaction alcohol byproducts. Remove the alcohol

layer using a Teflon stopcock separation funnel.

203 A7. Preparation of PVA Binder Solution (Chapter 4)

1. Pour 95 ml of distilled water in a 250 ml beaker.

2. Add 5 ml of HNO3 (1M).

3. Add 3 g of polyvinyl alcohol (PVA, Fluka Chemika, # 81384).

4. Stir for 15 min.

5. Cap the solution and heat to 80 °C (boiling point) while stirring. Avoid overheating

or, otherwise, the PVA will stick in the bottom of the beaker.

6. Heat with stirring until PVA dissolves.

A8. Preparation of γ-Alumina Supports by Dip Coating (Chapter 4)

1. Mix 13 ml PVA binder solution with 20 ml Boehmite sol.

2. Dip coat α-alumina support in the PVA/Boehmite mixture for 5 sec.

3. Dry extra sol from the support edges and dry in humid oven at 40 °C for 2 days.

4. Calcine membranes at 450 °C for 1 h, with heating and cooling rates of 0.5 °C/min.

5. Repeat dipping procedures as required to add additional layers or to cover-up

defects in the previous layer.

A9. Grafting of C18 silane agent on α-Alumina Supports (Chapter 4)

The supports are initially hydroxylated as described in the first 3 steps then the organic

groups are grafted as described in the following steps using C18 silane agent.

1. Boil the alumina supports in a mixture of 70 ml distilled water and 30 ml Hydrogen

Peroxide (H2O2) at 100 °C for 30 min.

2. Boil the supports in 100 ml distilled water at 100 °C for 30 min.

204 3. Dry supports in oven at 90 °C for 30 min.

4. Heat the hydroxylated supports at 150 ºC under vacuum for few hours.

5. Put 150 ml Toluene in a 3-neck flask and immerse it in an oil bath pre-heated to 120

ºC.

6. Put up to 4 hydroxylated alumina supports in the flask.

7. Add 12 ml Octadecyltrichlorosilane (C18, Gelest) quickly to the solution under N2

protection. Be careful with this material and take safety precautions.

8. Add 2 ml Pyridine. Be careful with this material and take safety precautions.

9. Attach the flask to a mechanical shaft mixer to the middle neck, a condenser to a

side neck to reflux back evaporated organics, and the last neck to a 20 ml/min

stream of N2 to reduce introduction of air to the air-sensitive material system.

10. Mix at 130 rpm under reflux for 20 hr.

11. Remove the supports from the flask and wash them in a vacuum filtration beaker or

sonicator according to the following sequential procedure:

ƒ Wash several times with toluene to remove excess organic silane.

ƒ Washed several times with ethanol to remove excess toluene.

ƒ Wash several times with distilled water to remove excess pyridine.

12. Dry the supports under vacuum at 90 ºC overnight to cure the covalent bonds of

grafting.

205 A10. Growth of Silica Membranes by the Counter Diffusion Self Assembly

Approach (Chapter 4)

Molar ratio: 1 TBOS: 0.025 CTAB: 2.29 HCl: 100 H2O at room temperature

1. Mix 23.12 ml distilled water with 0.156 g CTAB in a beaker to give a white

suspension.

2. Add 9.23 ml HCl (6M) slowly to the suspension and mix until it becomes clear.

3. Transfer the solution (water phase) to a small 50 ml beaker.

4. Fix the support, e.g., alumina disc, at the edge of a rubber tube.

5. Immerse the tube inside the beaker containing the water phase as shown in Fig. 4-4.

Make sure that no air bubbles are present on the support surface in contact with

water.

6. Wait for 5 minutes then add 620 µlit TBOS (Aldrich) inside the rubber tube on the

top of the support.

7. Age for 7−14 days.

8. Remove the support and wash gently with distilled water several times and dry in

air at room temperature for 1 day.

9. Perform any necessary characterization, e.g. gas permeation or XRD, on the as-

synthesized membrane.

10. Remove surfactant by calcination (with un-modified alumina supports) or Soxhlet

extraction (with silane modified supports) as follows:

ƒ Calcination: heat at 550 ºC for 6 h with heating and cooling rates of 1 ºC/min.

ƒ Soxhlet Extraction: in a mixture of 200 ml Ethanol and 2.8 g HCl at 120 ºC

for 1−2 days. Wrap the extractor with Aluminum foil to avoid heat loss.

206 A11. Steady State Single Gas Permeation (Chapters 4 and 5)

Differential pressure meter Pressure # 1 gauge ∆P High resolution P metering valve

Bubble flow meter Membrane cell Bubble flow He meter # 2

Bypass needle valve Gas Cylinder

1. Fix the membrane inside a stainless steel permeation cell with the active side of

membrane, e.g., γ-alumina layer or modified side, towards the upstream side. Use

O-ring (Viton) to seal the sides of the membrane as follows:

ƒ For 2 mm thick membranes, e.g., α-alumina, use 17.15 mm ID O-ring on each

side of the support or 2 concentric O-rings (12.6 mm ID inside 17.15 mm ID)

on each side of the membrane.

ƒ For thin membranes, e.g., 60 µm Anopores, use a stack of two 17.15 ID O-

rings on each side of the membrane.

2. Tight the permeation softly in a cross fashion to avoid breakage of the membrane.

3. Regulate the gas regulator pressure to 70 psig and calibrate the feed flow rate using

the mass flow controller to:

ƒ 5 to 10 ml/min for α- or γ-alumina supports

ƒ 230 to 400 ml/min for Anopore supports

ƒ 150 to 400 ml/min for large pore supports including borosilicate and quartz.

207 4. Open the metering valve (# 1) at the downstream side to full capacity.

5. Close the bypass valve (#2) and let the gas pass through the membrane cell.

6. Observe the change in the transmembrane pressure difference (∆P) using the

pressure readout. Make sure this value does not exceed 2.5 psig. If it does, open the

bypass valve (#2), decrease the feed flow rate and repeat step 6.

7. Wait until ∆P becomes constant. This equilibrium time varies depending on the

pore size. This time is approximately 1 h for alumina membranes, 5 min for

Anopore and 2 min for large pore supports.

8. When ∆P pressure becomes constant record the following three values:

ƒ ∆P in psi from pressure readout at location 1

ƒ Upstream pressure in psig from pressure readout at location 2

ƒ Gas flow rate using the bubble flow meter

9. Close the metering valve according to following sequence and repeat steps 5−8 for

each case:

ƒ 10, 1, 1, ½, and ½ cycles for alumina membranes

ƒ 2, 2, and 1 cycles for Anopore membranes

ƒ 5, 5, 1, and 1 cycles for large pore membranes

10. Calculate the average pressure across the membrane (Pav) and gas permeance (F/L)

corresponding at each valve opening using the equations:

ƒ Pav (Pa) = Pup − ∆P / 2

ƒ 2 Q (mol / s) F / L (mol / s.m .Pa) = 2 A(m )× (Pup − Pdown )

208 11. Draw (F/L) values versus Pav and regress the data with linear relation and obtain the

slope (β) and intercept (α) according to:

(F/L)=α+β Pav

12. Using the expressions and values of (α) and (β), obtain the average pore size of the

membrane from the ratio of (β/α) as follows:

ε rp α =1.06( ) τ L M w RT

2 ε rp β = 0.125( ) τ LηRT

RT  β  d p = 16.964η   M w  α 

where η and Mw are the gas viscosity and molecular weight respectively.

209 A12. Unsteady State Gas Permeation (Chapter 5)

Upstream To data acquisition pressure gauge Oven Downstream P 1 P pressure gauge 2

Gas tank 3 Membrane cell 4

Gas Cylinder 5 (CO2, N2, He) Vacuum pump

1. Fix the membrane inside a stainless steel permeation cell with the active side of

membrane towards the upstream side using O-ring (Viton) to seal the sides and tight

the cell softly to avoid breaking the membrane.

2. For high temperature permeation measurements, place the membrane cell inside a

furnace or oven and use thermal resistant O-rings that keep intact at the desired high

temperatures.

3. While valve 1 is closed, regulate the feed gas pressure to the desired pressure, e.g.,

10 psig.

4. Keep valve 1 closed, open valves 2, 3 and 4 and start the vacuum pump by opening

valve 5 to evacuate the upstream and downstream sides. Wait until the down stream

pressure achieves the minimum reading on the downstream pressure sensor.

5. Close valves 5 and observe the downstream pressure reading. If the value is

constant, proceed to next step. If it increases slowly, then there is a leak in the

system. Fix the leak and repeat from step 4.

210 6. Close valves 3 and 4 and keep valve 2 open.

7. Prepare a file on the data acquisition computer to save the data on (step size 10 sec).

8. Open valve 1 and start recording the data on the data acquisition file. The rate of

change of gas pressure in the 1 lit tank at the downstream side will be measured and

recorded in the file.

9. When an acceptable straight line of pressure change with time is obtained, normally

after 5 to 10 min, stop data recording and close valve 1.

10. For additional measurements at different pressure, regulate the feed pressure to the

new value, e.g., 20 or 30 psig, and repeat from step 4.

11. For additional measurements at different temperature, regulate the temperature of

the oven/furnace to the desired value and wait enough time until the temperature of

the permeation cell becomes uniform, then repeat from step 4.

12. Open the data acquisition file using Excel, draw the change in pressure with time,

regress the data with straight line and obtain the slope dPdown/dt (mmHg/s).

13. Calculate the permeance using following relation (assuming ideal gas):

F V dP = tan k down L RTAm (Pup − Pdown ) dt

where Vtank = 1 lit, Pdown ≈ 0 due to vacuum, Am= membranes area calculated from

the ID of the O-ring, dPdown/dt (unit: Pa/s) is the slope obtained from the straight

line part of the data acquisition file. T is the temperature of the gas collected in the

gas tank, i.e., room temperature, and not the temperature of the permeation cell.

211 A13. Removal of the surrounding polymer supporting ring from the Anopore

Membranes (Chapter 5)

1. Place the Anopore membranes on a porous support and put it inside a

programmable oven.

2. Calcine the membranes at 480 ºC for 4 h with heating and cooling rates of 1 ºC/min.

A14. CVD Modification of Anopore with Residual Gas (Chapter 5)

Pressure Sensor Tube Furnace Valve 5 1 4

2 3

Anopore

H2O TMA He Vacuum pump

1. Attach the Anopore membrane at the edge of the central ceramic tube by means of

aluminum foil around the perimeter of the membrane with the 20 nm pore side

facing the precursors upstream.

2. Place the Pyrex tube around the ceramic tube and attach it to the precursor feed inlet

and the vacuum pump as shown in the setup.

3. Place the Tubular furnace around the Pyrex tube and heat to 180 ºC using the

temperature controller.

4. Attach the vacuum pump discharge to the fume hood.

212 5. Open valves 1 and 2 and evacuate the system down to 1 mbar.

6. Close valves 1 and 2, then open valve 3 to introduce water vapor and for 5 min. The

pressure will increase to approximately 32 mbar.

7. Close valve 3, open valves 1 and 2 and evacuate the system down to 1 mbar

(residual pressure of water vapor) for 10 min.

8. Close valves 1 and 2, then open valve 4 to introduce trimethylaluminum (TMA

97%, Aldrich # 25 722-2) for 5 min.

9. Close valve 4, open valves 1 and 2 and evacuate the system down to 1 mbar

(residual pressure of TMA) for 10 min.

10. This corresponds to 1 cycle of CVD. For additional cycles, repeat steps 6 to 9 as

required.

11. At the end, Open valves 1 and 2 and evacuate the system down to 1 mbar for

enough time.

12. Purge the system with helium gas by opening valve 5 for short time then close the

valve.

13. Cool down the tubular furnace to room temperature by using the temperature

controller.

14. Disassemble the setup and take off the modified membrane.

213 A15. CVD Modification of Anopore with Purge (Chapter 5)

1. Attach the Anopore membrane at the edge of the central ceramic tube by means of

aluminum foil around the perimeter of the membrane with the 20 nm pore side

facing the precursors upstream.

2. Place the Pyrex tube around the ceramic tube and attach it to the precursor feed inlet

and the vacuum pump as shown in the setup.

3. Place the Tubular furnace around the Pyrex tube and heat to 180 ºC using the

temperature controller.

4. Attach the vacuum pump discharge to the fume hood.

5. Open valves 1 and 2 and evacuate the system down to 1 mbar.

6. Close valves 1 and 2, then open valve 3 to introduce water vapor and for 5 min. The

pressure will increase to approximately 32 mbar.

7. Close valve 3, open valves 1 and 2 and evacuate the system down to 1 mbar for 10

min.

8. Close valves 1 and 2, and open valve 5 to introduce a helium purge flow at

moderate pressure (~ 1500 mbar) for 2 min to clean the gas phase of the pore from

the residual water vapor.

9. Close valves 5, then open valve 4 to introduce trimethylaluminum (TMA 97%,

Aldrich # 25 722-2) for 5 min.

10. Close valve 4, then open valves 1 and 2 and evacuate the system down to 1 mbar

for 10 min.

11. Close valves 1 and 2, and open valve 5 to introduce a helium purge flow at 1500

mbar for 2 min to clean the gas phase of the pore from the residual TMA.

214 12. This corresponds to 1 cycle of CVD. For additional cycles, repeat steps 6 to 11 as

required.

13. At the end, Open valves 1 and 2 and evacuate the system down to 1 mbar for

enough time.

14. Purge the system with helium gas by opening valve 5 for short time then close the

valve.

15. Cool down the tubular furnace to room temperature by using the temperature

controller.

16. Disassemble the setup and take off the modified membrane.

215 A16. Separation of water vapor from air on Anopore Alumina Membranes using

Permporometry (Chapter 5)

Flow meter

To oxygen and moisture sensors Valve 1 2

3-way Oven Valve Membrane cell Water Sparger Air N2

1. Fix the membrane inside a stainless steel permeation cell with the active side of

membrane facing the upstream side using a stack of two Viton O-rings on each side.

Tight the cell softly to avoid breaking the membrane.

2. Place the membrane cell in an oven at 35-40 ºC to avoid condensation of water

vapor inside the tubes. For the same reason, all the outside stainless steel tubes are

wrapped with heating tapes and kept at 40 ºC.

3. Open the nitrogen sweep gas valve and pass the stream at a flow rate of 75 ml/min.

This flow rate is kept constant for the entire experiment.

4. Introduce a 100 ml/min total flow of air stream with 100% relative humidity (RH)

by mixing 100 ml/min saturated air with 0 ml/min dry air. The air stream flows in a

cross flow fashion relative to the sweep gas. The total feed flow is kept at 100

ml/min for the entire experiment.

5. Wait from 10 h to 1 day for equilibrium then measure the RH using humidity sensor

(Thermohygrometer, Cole-Parmer 37950-10) and oxygen content using oxygen

216 analyzer (6000 Oxygen Analyzer, Illinois Instruments) for both the upstream

(stream 1) and the down stream products (stream 2).

6. Conduct the experiment at 100 ml/min air feed of various RH in the range of

0−100%. The RH is changed by controlling the relative quantities of the dry to wet

air streams in the 100 ml/min feed flow.

7. To obtain X RH, the feed mixture flow should be [X ml/min wet air + (100−X)

ml/min dry air], e.g., 90%RH is obtained at 90 ml/min wet air + 10 ml/min dry air.

At each RH, repeat step 5.

8. Calculate the Oxygen and water permeances through the membrane according to

the relations:

Q (mol / s)×Y (F / L) = N2 O2 A× Patm × (X −Y )

where QN2 is the mole flow rate of the nitrogen sweep gas, Y is the mole fraction of

oxygen in the downstream product (stream 2), X is the mole fraction of oxygen in

the upstream product (stream 1), Patm is the pressure of both upstream and down

stream products= 1 atm.

Q (mol / s) × M × MR (F / L) = N 2 N 2 H 2O M × A× P (RH − RH ) /100 H 2O vo up down

where the molecular weights are in kg/mol, MR is the moisture ratio (kg H2O/kg

dry air) obtained from psychometric chart, Pvo is the water vapor pressure at the

temperature of product streams, and RH is the relative humidity.

217 APPENDIX B

SAMPLE OF CALCULATIONS

B1. Porosity of Silica Fibers (Chapter 3, p. 87)

Pore volume= 0.62 cm3/g and silica density= 2.15 g/cm3. Based on 1 g of the sample:

vol of voids Porosity = vol fraction of voids = vol of voids + vol of solid

0.62cm3 = = 0.571 0.62cm3 + (1g / 2.15 g / cm3 )

B2. Monolayer coverage of gases on silica fibers (Chapter 3, p. 89)

q× Ao × N A X monolayer = SBET −3 q = gas uptake capacity (mol / g) = 0.68×10 mol / g for CO2

o2 Ao = sorbate cross − sectional area estimated from the molecular diameter = 9 A 23 N A = Avogadro's number = 6.02×10 molecule/ mol 2 SBET = Surface area = 799 m / g 0.68×10−3 mol / g ×6.02×1023 molecule/ mol ×9Ao2 ×(10−10 m/ Ao )2 X = monolayer CO2 799 m2 / g = 4.6%

B3. Adsorption Equilibrium Constant (Chapter 3, p. 98)

K value in equation 3-3 is dimensionless. K value obtained from the slope of adsorption isotherm on Figure 3-5 has the unit (g/g.atm). This value is converted to dimensionless as follows:

218 R×T × ρ K (dimensioless) = slope (g / g.atm)× solid M

Slope= adsorption constant obtained from the slope of the equilibrium adsorption

-3 isotherm= 26.54×10 g/g.atm for CO2 gas.

R= ideal gas constant = 82.1 atm.ml/mol.K

T= temperature in K= 300 K

ρsolid= silica density= 2.15 ml/g

M= gas molecular weight= 44 g/mol for CO2 gas.

82.1×300× 2.15 K (dimensionless) = 26.54×10−3 × = 31.94 44

B4. Evaluation of Diffusion Model Constants (Chapter 3, p. 99)

D eff M = α + β K L2

0.027 = α + β ×31.94 for CO 2 0.030 = α + β × 41.66 for C2 H 4

Solution of two simultaneous equations with two unknowns gives:

α= 1.70×10-2 (g/mol)0.5/s and β= 3.17×10-4 (g/mol)0.5/s

B5. Evaluation of Diffusivities and Contributions (Chapter 3, p. 99)

This sample is for CO2 gas

2 2 L DK ,eff (cm / s) = α × M where α= 1.70×10-2 (g/mol)0.5/s

L= fiber length= 196 µm= 0.0196 cm and M= molecular weight= 44 g/mol for CO2

219 2 −2 0.0196 −7 2 DK ,eff CO2=1.70×10 × = 9.85×10 cm / s 44

DS, eff is evaluated from the difference between the total effective diffusivity and Knudsen diffusivity as follows:

1−ε D = D + K D eff K ,eff ε S,eff

-7 2 where Deff= 15.70×10 cm /s and K=31.94 for CO2 and ε=0.571

1− ε K D = 15.70 ×10−7 − 9.85×10−7 = 5.87 ×10−7 cm2 / s ε S,eff

5.87 ×10−7 0.571 D × = 2.48×10−8 cm2 / s S,eff = 31.94 1− 0.571

-7 DK ,eff 9.85×10 Contribution of Knudsen diffusion = = -7 = 63% Deff 15.70×10

(1−ε ) K D -7 ε S,eff 5.87×10 Contribution of Surface diffusion = = -7 = 37% Deff 15.70×10

B6. Effective and Real Diffusivities (Chapter 3, p. 101)

ε 1 DK,eff = DK ,real and DS,eff = DS,real κ g κ s

where κ is the tortuosity factor obtained from the parameter α. Here we assume that κg=κs

ε d 8R T α = p κ L2 9π M

α= 1.72×10-2 (g/mol)0.5/s

-9 dp= pore diameter of silica fibers= 2.74 nm= 2.74×10 m

220 L= fiber length= 196 µm= 196×10-6 m

T= temperature = 300 K

M= gas molecular weight= 44 g/mol for CO2

R= gas constant= 8.314 Pa.m3/mol.K= 8314 g.m2/s2.mol.K

0.571 2.74×10−9 8×8314×300 κ = × = 2012 1.7×10−2 (196×10−6 )2 9π

κ −7 2012 −3 2 DK ,real = DK ,eff × = 9.85×10 × = 3.47×10 cm / s ε 0.517 −8 −5 2 DS,real = DS,eff ×κ = 2.48×10 × 2012 = 5.00×10 cm / s

Small differences in calculated values could be due to rounding off errors.

B7. Internal Dimensions of Silica Fibers (Chapter 3, p. 105)

Tortuosity factor = tortuosity2 ⇒ κ = τ2

κ = 2012 =τ 2 ⇒ τ = 45

pore length l τ = = p = 45 ⇒ l = 8820 µm = 8.8 mm fiber length 196 µm p

To evaluate the pitch size (z) we use the following equation:

L f d f d f l p = π × d f × ⇒ z = π × = π × z l p τ L f where df= fiber diameter= 15 µm

Lf= fiber length= 196 µm

τ= 45

15 µm z = π × =1.05 µm 45

221 B8. Average pore size from steady state measurements (Chapter 4, p. 132)

For α-alumina support made from Alcoa-A15 powder, helium gas permeance data is shown in Figure 4-7. Linear regression of the data gave the following linear parameters:

-12 -6 (F/L)= 6.13×10 Pav + 4.55×10 where the slope (β) is in mol/s.m2.Pa2 and intercept (α) is in mol/s.m2.Pa

From Equations 4-2 and 4-3 the average pore diameter (dp) is calculated from the β/α ratio using the equation:

RT  β  d p =16.964η   M  α  where dp is the average pore size (m)

η is the He viscosity at 20 ºC (293 K)= 19.6µPa.s= 19.6×10-6 Pa.s

T= permeatin temperature= 20 ºC= 293 K

R= gas constant= 8.314 Pa.m3/mol.K or (kg.m2/s2.mol.K)

M= molecular weight of helium gas= 4 g/mol= 4×10-3 kg/mol

−12 −6 8.314× 293 6.13×10 −6 d p =16.964×19.6×10 × = 0.35×10 m = 0.35µm 4×10−3 4.55×10−6

B9. Permeance from Unsteady State Measurements (Chapter 4, p. 136)

This sample calculation is applied on the α-A15 alumina support with nitrogen gas.

(dP / dt)×V (F / L) = down tan k R×T × A×(Pup − Pdown )

(dPdn/dt) is the slope obtained from the rate of change of down stream pressure in the tank= 0.3205 mmHg/s= 0.3205 mmHg×101325 Pa/760 mmHg= 42.73 Pa/s.

222 -3 3 Vtank= 1 lit= 1×10 m

R= 8.314 Pa.m3/mol.K

T= 28 ºC= 28+273= 301 K

A= area exposed to gas based on ID of the O-ring= (π/4)×d2= (3.14/4)×(12.6×10-3)2

= 1.247×10-4 m2.

Pup= feed gas pressure= 10 psig= 10+14.7 psi= 24.7 psi×(101325 Pa/14.7 psi)

= 170254 Pa

Pdown= assumed to be ZERO due to vacuum

−3 3 42.73 Pa / s×1×10 m −7 2 F / L = = 8.04×10 mol / s.m .Pa 8.314×301 K ×1.247×10−4 m2 ×(170254 Pa − 0)

B10. Surface Coverage of Hydrophobic Groups (Chapter 4, p. 140)

This sample calculation is applied on C18-modified alumina membrane based on 100 g weight of support.

%wt loss of modified sample − %wt loss of hydroxylated sample Surface Coverage = Suppot wt.× S × M i where %wt loss of C18-modified sample= 0.34 from TGA curve (Fig. 4-10)

%wt loss of hydroxylated base sample= 0.227

Mi= Molecular weight of C18 group C18H37)= 253 g/mol

0.34 − 0.227 Surface Coverage = = 0.45×10−6 mol / m2 = 0.45 µmol / m2 100×10 m2 / g × 253

mol 6.02×1023 molecule/ mol 0.45×10−6 × = 0.27 molecule/ nm2 m2 (109 nm/ m)2

223 B11. Oxygen Permeance from Permporometry (Chapter 5, p. 177)

This sample of calculation is applied on the unmodified Anopore support at 30% relative humidity shown in Figure 5-5.

QN 2 ×Y (F / L)O2 = A×(Patm X − Patm ×Y)

-3 3 QN2= sweep gas molar flow rate= 75 ml/min= 75 ml/min×10 m /lit/(22.4 mol/lit×60 s/min)= 5.58×10-5 mol/s.

Y is the oxygen molar fraction in the downstream product= 5.41%

X is the oxygen fraction in the upstream product= 16.7%

A= membrane area based on ID of the O-ring= 1.54×10-4 m2

PatmY and PatmX are the oxygen partial pressure in the downstream and upstream products respectively. Patm in both stream is equivalent= 1 atm= 101325 Pa

−5 5.58×10 ×0.0541 −6 2 (F / L)O2 = =1.70×10 mol / s.m .Pa 1.54×10−4 m2 ×101325 Pa (0.167 − 0.0541)

B12. Water Vapor Permeance from Permporometry (Chapter 5, p. 179)

This sample of calculation is applied on the unmodified Anopore support at 30% relative humidity shown in Figure 5-6.

Q (mol / s) × M × MR (F / L) = N 2 N 2 H 2O M × A× P (RH − RH ) /100 H 2O vo up down

-3 3 QN2= sweep gas molar flow rate= 75 ml/min= 75 ml/min×10 m /lit/(22.4 mol/lit×60 s/min)= 5.58×10-5 mol/s.

MN2= molecular weight of the nitrogen sweep gas= 28 g/mol= 0.028 kg/mol

224 MR= moisture ratio of down stream product obtained using psychometric chart= 0.004 kg water vapor/kg dry air

MH2O= molecular weight of water= 18 g/mol= 0.018 kg/mol.

A= membrane area based on ID of the O-ring= 1.54×10-4 m2

Pvo= is the water vapor pressure of product streams at 35 ºC= 5626 Pa

RHup= relative humidity of the upstream product= 23.4%

RHdown= relative humidity of the downstream product= 11.8%

−5 5.58×10 mol / s×0.028 kg / mol ×0.004 kg H 2O / kg dry air (F / L) H O = 2 0.018kg / mol ×1.54×10−4 m2 ×5626Pa (23.4 −11.8)/100 = 3.45×10−6 mol / s.m2.Pa

225