Inorganic, Polymeric and Composite Membranes: Structure, Function and Other Correlations (Membrane Science and Technology, Volume 14)

Membrane Science and Technology Series

Inorganic, Polymeric and Composite Membranes Structure, Function and Other Correlations Volume 14

This page intentionally left blank

Membrane Science and Technology Series

Inorganic, Polymeric and Composite Membranes Structure, Function and Other Correlations Volume 14

Edited by

S. Ted Oyama Susan M. Stagg-Williams

Amsterdam • Boston • Heidelberg • London • New York • Oxford Paris • San Diego • San Francisco • Singapore • Sydney • Tokyo

Elsevier Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK First edition 2011 Copyright # 2011 Elsevier B.V. All rights reserved No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email: [email protected]. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material Notice No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress ISBN: 978-0-444-53728-7 ISSN: 0927-5193

For information on all Elsevier publications visit our web site at www.elsevierdirect.com

Printed and bound in Great Britain 11 12 13 14

10 9 8

7

6 5

4

3 2

1

To Hideko, for her sustaining love and support, and Monika and Leo for their inspiring energy and enthusiasm. To Michael, for his never-ending patience and love, and Mikey and Sammy for the laughter, hugs, and true blessings they are in my life.

This page intentionally left blank

Contents

Dedication Contributors Preface

1. Correlations

v xv xix

1

S. Ted Oyama Introduction Scientific Laws and Correlations Principles Theories Laws Properties Effects Equations Dimensionless Numbers Criteria Approximations, Factor, Curves Correlations Important Properties in Membrane Science Examples of Correlations in the Membrane Separation Field Summary Acknowledgments References

1 3 3 3 4 6 6 7 10 11 11 13 14 16 22 22 22

2. Review of Silica Membranes for Hydrogen Separation Prepared by Chemical Vapor Deposition

25

Sheima Jatib Khatib, S. Ted Oyama, Ka´tia R. de Souza and Fa´bio B. Noronha Introduction Silica Membranes for Hydrogen Separation Chemical Vapor Deposition: Principles Synthesis of Silica Membranes via Chemical Vapor Deposition Silica Membranes Supported on Vycor Glass Silica Membranes Supported on Alumina Conclusions Acknowledgments References

25 25 27 29 37 48 56 56 56

vii

viii

Contents

3. Amorphous Silica Membranes for H2 Separation Prepared by Chemical Vapor Deposition on Hollow Fiber Supports

61

Dmitri D. Iarikov, Pelin Hacarlioglu and S. Ted Oyama Introduction Experimental Results and Discussion Pure Hollow Fiber Support Properties Mesoporous Silica Layer Amorphous g-Alumina Layer Silica Precursor and Carrier Gas Flow Rate Effects on the Membrane Separation Performance Gas Separation Mechanism Conclusions Acknowledgments References

4. Ab Initio Studies of Silica-Based Membranes: Activation Energy of Permeation

61 65 66 67 69 72 73 74 75 76 76

79

Pelin Hacarlioglu, Luke Achenie and S. Ted Oyama Introduction Previous Theoretical Studies on Dense Silica-based Membranes Method of Calculation Results and Discussion Conclusions Acknowledgments References

5. Review of CO2/CH4 Separation Membranes

79 80 81 82 89 89 89

91

Dmitri D. Iarikov and S. Ted Oyama Introduction Discussion Zeolite Membranes and Carbon Molecular Sieves Silica Membranes Polymeric Membranes Mixed-matrix Membranes Supported ionic Liquid and Polyionic Membranes Overall Results Conclusions Acknowledgments References

91 93 93 97 98 102 103 107 109 110 110

Contents

6. Gas Permeation Properties of Helium, Hydrogen, and Polar Molecules Through Microporous Silica Membranes at High Temperatures: Correlation with Silica Network Structure

ix

117

Masakoto Kanezashi and Toshinori Tsuru Introduction Experimental Fabrication of Silica and Co-Doped Silica Membranes by Sol–Gel Method Gas Permeation/Separation Measurements for Silica Membranes Results and Discussion Improved Hydrothermal Stability of Amorphous Silica Membranes Helium and Hydrogen Permeation Properties Through Amorphous Silica Membranes Permeation Properties of Polar Molecules (NH3, H2O) Through Amorphous Silica Membranes Conclusions References

7. Correlation Between Pyrolysis Atmosphere and Carbon Molecular Sieve Membrane Performance Properties

117 119 119 120 120 120 125 127 133 134

137

Mayumi Kiyono, William J. Koros and Paul J. Williams Introduction Theory and Background Transport in CMS Membranes Structure of CMS Membranes Effect of Pyrolysis Atmosphere on Separation Performance of CMS Membranes Experimental Materials Characterization Methods Results and Discussion Correlation Between Oxygen Exposure and CMS Separation Performance Correlation Between Oxygen Concentration and CMS Separation Performance Possible Mechanism of Oxygen “Doping” Process During Pyrolysis Conclusion Acknowledgements References

137 138 138 139 140 142 142 145 146 146 161 169 170 171 171

x

Contents

8. Review on Prospects for Energy Saving in Distillation Process with Microporous Membranes

175

Masahiko Matsukata, Ken-ichi Sawamura, Yasushi Sekine and Eiichi Kikuchi Introduction Potential of Membrane Separation Technology for Large-Scale Reduction in Energy Consumption Why Zeolite Membranes are Promising Synthesis Technique of Zeolite Membranes Seeding Technique (Secondary Growth Method) Masking Technique Use of SDA for Microstructural Optimization De-Watering Technology Using Zeolite Membranes De-watering of Alcohol De-watering of Organic Acids De-watering for C1 Chemistry Concluding remarks References

9. Xylene Separation Performance of Composition-Gradient MFI Zeolite Membranes

175 176 179 181 181 182 183 184 184 186 187 188 189

195

Jessica O’Brien-Abraham, Mikel Duke and Y.S. Lin Introduction Experimental Bilayer Membrane Synthesis and Characterization Pervaporation Experiments Results and Discussion Membrane Characteristics Binary Pervaporation Through Single and Bilayer Membranes Reversal of Bilayer Structure Stability at Higher PX Feed Concentrations Conclusions Acknowledgments References

10. Membrane Extraction for Biofuel Production

195 198 198 200 201 201 202 207 209 210 211 211

213

David L. Grzenia, Xianghong Qian, Silvio Silverio da Silva, Xinying Wang and S. Ranil Wickramasinghe Introduction Removal of Acetic Acid from Biomass Hydrolysates Extraction of 5-Hydroxymethylfurfural Glycerol Extraction Material and Methods Removal of Acetic Acid from Biomass Hydrolysates

213 215 217 218 218 219

Contents

HMF Extraction Glycerol Extraction Results Conclusions Acknowledgments References

11. A Review of Mixed Ionic and Electronic Conducting Ceramic Membranes as Oxygen Sources for High-Temperature Reactors

xi 221 221 221 230 231 231

235

Qiying Jiang, Sedigheh Faraji, David A. Slade and Susan M. Stagg-Williams Introduction General Attributes of Oxygen-Conducting MIEC Ceramic Materials Oxygen Nonstoichiometry Self-Adjusting Phase Equilibria Chemical Expansivity Microstructure of Oxygen-MIEC Ceramics Common Oxygen-MIEC Membrane Materials Fluorites Perovskites SCF-Based Materials Dual-Phase Composite Materials Membrane Modifications to Improve Oxygen Flux Surface Modifications Membrane Thickness Reduction MIEC Membranes for Synthesis Gas Production Synthesis Gas Production Overview Benefits of MIEC Membranes for Synthesis Gas Production Overview of Work to Date Effect of Reaction Temperature on Membrane Performance Effect of Reaction Environment on Membrane Oxygen Flux Conclusions Acknowledgments References

12. Critical Factors Affecting Oxygen Permeation Through Dual-phase Membranes

235 236 236 237 238 239 240 240 241 243 248 249 252 253 255 255 257 258 260 261 263 264 264

275

Xuefeng Zhu and Weishen Yang Introduction Design of Dual-Phase Membranes with High Stability and Permeability Experimental Investigation of Dual-Phase Membranes Pure Electronic Conductor or Mixed Conductor?

275 278 280 280

xii

Contents

Surface Exchange Preparation Methods for Powders Sintering Temperature Ratio Between the Two Phases Other Factors Conclusions Acknowledgments References

13. High Temperature Gas Separations Using High Performance Polymers

281 282 286 288 289 290 290 291

295

John R. Klaehn, Christopher J. Orme, Eric S. Peterson, Frederick F. Stewart and Jagoda M. Urban-Klaehn Introduction Experimental Instrumentation Permeability Gas Testing Positron Annihilation Lifetime Spectroscopy Results and Discussion Conclusions Acknowledgments References

14. Using First-principles Models to Advance Development of Metal Membranes for High Temperature Hydrogen Purification

295 297 297 297 298 298 306 306 306

309

Sunggu Kang, Shiqiang Hao and David S. Sholl Introduction DFT-based Modeling of Crystalline Metal Membranes Cluster Expansion Methods Applications of DFT Calculations to Crystalline Membrane Materials Amorphous Metal Membranes Computational Approaches for Amorphous Metals Amorphous Structures Binding Energy of Interstitial H in Amorphous Alloys H–H Interactions H Solubility in Amorphous Alloys Hydrogen Diffusion in Amorphous Alloys Corrected Diffusivities The Thermodynamic Correction Factor Hydrogen Permeability Through Amorphous Alloys Conclusion Acknowledgments References

309 311 313 314 316 317 317 318 319 320 322 323 324 326 327 328 328

Contents

15. High Performance Ultrafiltration Membranes: Pore Geometry and Charge Effects

xiii

333

Andrew L. Zydney Introduction Pore Geometry Effects Fluid Flow Solute Transport Pore Size Distribution Effects Electrostatic Interactions Fluid Flow Solute Transport Concentration Polarization Effects Conclusions Acknowledgment References Subject Index

333 335 335 336 339 341 341 344 347 350 351 351 353

This page intentionally left blank

Contributors

Numbers in Parentheses indicate the pages on which the author’s contributions begin.

Luke Achenie (79), Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA Silvio Silverio da Silva (213), Department of Biotechnology, School of Engineering of Lorena, University of Sa˜o Paulo, Lorena/SP, Brazil Ka´tia R. de Souza (25), Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA, Instituto Nacional de Tecnologia—INT, Av. Venezuela 82, CEP 20081-312, Rio de Janeiro, Brazil Mikel Duke (195), Institute for Sustainability and Innovation, Victoria University, Werribee Campus, Melbourne, Victoria, Australia Sedigheh Faraji (235), Chemical Engineering Department, California State University, Long Beach, California, USA David L. Grzenia (213), Department of Chemical and Biological Engineering, Colorado State University, Fort Collins, Colorado, USA Pelin Hacarlioglu (61, 79), Environmental Catalysis and Nanomaterials Laboratory, Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA Shiqiang Hao (309), School of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA Dmitri D. Iarikov (61, 91), Environmental Catalysis and Nanomaterials Laboratory, Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA Qiying Jiang (235), Chemical and Petroleum Engineering Department, University of Kansas, Lawrence, Kansas, USA Masakoto Kanezashi (117), Department of Chemical Engineering, Hiroshima University, 1-4-1 Kagamiyama, Higashi-Hiroshima, 739–8527, Japan Sunggu Kang (309), School of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA Sheima Jatib Khatib (25), Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA Eiichi Kikuchi (175), Department of Applied Chemistry, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, Japan, Advanced Research Institute for Science and Engineering, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, Japan

xv

xvi

Contributors

Mayumi Kiyono (137), Department of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA John R. Klaehn (295), Idaho National Laboratory, Idaho Falls, Idaho, USA William J. Koros (137), Department of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA Y.S. Lin (195), Chemical Engineering, School for Engineering of Matter, Transport and Energy, Arizona State University, Tempe, Arizona, USA Masahiko Matsukata (175), Department of Applied Chemistry, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, Japan, Advanced Research Institute for Science and Engineering, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, Japan Fa´bio B. Noronha (25), Instituto Nacional de Tecnologia—INT, Av. Venezuela 82, CEP 20081-312, Rio de Janeiro, Brazil Jessica O’Brien-Abraham (195), Chemical Engineering, School for Engineering of Matter, Transport and Energy, Arizona State University, Tempe, Arizona, USA Christopher J. Orme (295), Idaho National Laboratory, Idaho Falls, Idaho, USA S. Ted Oyama (1, 25, 61, 79, 91), Environmental Catalysis and Nanomaterials Laboratory, Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA, Department of Chemical Systems Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan Eric S. Peterson (295), Idaho National Laboratory, Idaho Falls, Idaho, USA Xianghong Qian (213), Department of Mechanical Engineering, Colorado State University, Fort Collins, Colorado, USA Ken-ichi Sawamura (175), Department of Applied Chemistry, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, Japan Yasushi Sekine (175), Department of Applied Chemistry, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, Japan, Advanced Research Institute for Science and Engineering, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, Japan David S. Sholl (309), School of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA David A. Slade (235), Chemical and Petroleum Engineering Department, University of Kansas, Lawrence, Kansas, USA Susan M. Stagg-Williams (235), Chemical and Petroleum Engineering Department, University of Kansas, Lawrence, Kansas, USA Frederick F. Stewart (295), Idaho National Laboratory, Idaho Falls, Idaho, USA Toshinori Tsuru (117), Department of Chemical Engineering, Hiroshima University, 1-4-1 Kagamiyama, Higashi-Hiroshima, 739–8527, Japan Jagoda M. Urban-Klaehn (295), Idaho Accelerator Center, Pocatello, Idaho, USA Xinying Wang (213), Department of Chemical and Biological Engineering, Colorado State University, Fort Collins, Colorado, USA

Contributors

xvii

S. Ranil Wickramasinghe (213), Department of Chemical and Biological Engineering, Colorado State University, Fort Collins, Colorado, USA Paul J. Williams (137), Shell Projects and Technology, Houston, Texas, USA Weishen Yang (275), State Key Laboratory of Catalysis, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, China Xuefeng Zhu (275), State Key Laboratory of Catalysis, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, China Andrew L. Zydney (333), Department of Chemical Engineering, The Pennsylvania State University, University Park, Pennsylvania, USA

This page intentionally left blank

Preface

Membrane science is a vibrantly active experimental and theoretical field, with significant impact in chemical industry because of the development of improved separations and the advent of process intensification where membranes play an important role. This tome organizes and summarizes the results of many dispersed studies in inorganic, polymer, and composite membranes by presenting reviews of the literature and papers that present general correlations of results. Thus, the present treatise is a departure from conventional compilations of disparate papers from different areas because it presents a high level of analysis. To punctuate the effort, the book starts with a soliloquy on the nature and significance of correlations in intellectual fields of endeavor. Based on historical developments, it is suggested that, for “correlations” to rise to the level of “theories,” the correlations must be based on sound physical phenomena and must be capable of making predictions. The coverage then goes on to several frontier topics that range from molecular level descriptions of membranes to process improvement. Detailed coverage is provided of silica membranes, palladium alloy membranes, ionic liquids, mixed oxides, carbon sieves, polyimide materials as well as the enhancement of distillation performance by the addition of membrane units, and the production of bio-derived fuels. It is hoped that this book will contribute to a higher order of thinking in the field of membrane science and contribute to the search for general understanding. S. Ted Oyama Susan M. Stagg-Williams

xix

This page intentionally left blank

Chapter 1

Correlations S. Ted Oyama1,2,* Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA 2 Department of Chemical Systems Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan * Corresponding author: E-mail addresses: [email protected], [email protected] 1

INTRODUCTION Correlations are interesting because they relate different characteristics of a system and give understanding to complex phenomena. Correlations are also linked to causality, a subject that comes up in diverse fields. The more complex the phenomenon, the more difficult it is to unravel the true cause of an effect. The establishment of cause-and-effect relationships is a goal in many fields in the sciences and the humanities. It is particularly challenging in the humanities where quantitative measurements are not emphasized. Yet in disciplines like history, where it is clearly difficult to carry out controlled experiments, scientific methodologies are being developed [1,2]. In this respect, a field such as history bears a resemblance to astronomy, evolutionary biology, and geology [3], as well as psychology and economics, where effects are influenced by large numbers of variables. Some of the techniques to deal with this situation are statistical. A method used to deal with extremely complex phenomena is the “as if random” hypothesis. Although controlled experiments may not be possible, and, therefore, cases cannot be assigned to treatment and control groups, it can sometimes be claimed that the assignment of cases is “as if random” [2], and this leads to meaningful conclusions. Cause is a philosophical concept that goes back to antiquity. Aristotle considered that everything that existed had to have a cause, and described four categories, material cause, formal cause, efficient cause, and final cause. “Everything that moves is moved by another.” Aristotle (Metaphysics, c. 335 to 323 BCE)

The explanation of cause requires knowledge of its origin, and to choose between different possibilities, it was articulated by William of Ockham that Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

1

2

Inorganic, Polymeric and Composite Membranes

explanations should not have unnecessary postulates. This is often stated as “the simplest explanation is usually the correct one” and is an expression of the principle of parsimony. The concept has precedents with earlier thinkers, such as Maimonides and Thomas Aquinas, but became popular from its frequent use by Ockham [4] and was later referred to as Occam’s razor. The term razor comes from the need to shave away unnecessary elements of a theory or to cut apart two different explanations. “Entities must not be multiplied beyond necessity.” William of Ockham (c. 1285–1349)

Deduction and induction are the principal approaches of the scientific method as was expounded by Francis Bacon. Deduction deals with processes that involve going from generalities to specifics. Induction is the opposite and deals with deriving generalities from specifics [5]. Thus, causality is intimately related to these two approaches. Deduction is the explanation of the cause of a phenomenon using a previously identified law or principle of nature (vide infra). Induction is the derivation of a general relationship based on observation. Deduction is the process of inferring specifics from generalities. Induction is the process of deriving generalities from specifics. Francis Bacon (The Advancement of Learning, 1605)

Causality was also addressed by Isaac Newton in the first and most influential of his four rules for reasoning, in which he explained that there should be no more causes than required to explain a phenomenon [4]. This is the principle of minimality applied to cause. “We are to admit no more causes of natural things than such as are both true and sufficient to explain their appearances.” Isaac Newton (Principia Mathematica, 1687)

Causality is fundamental to the understanding of natural phenomena, and the elucidation of cause is thus a primary goal. Based on the foundations laid by Bacon and Newton, John Herschel describes an approach for the development of scientific ideas known as vera causa–true cause [4]. As described by Thomson [4], the establishment of the vera causa of a phenomenon has three components. The potential causal phenomenon has to be real and existing, it has to be demonstrated as capable of producing the effect under investigation, and then it has to be shown to be actually responsible. The third requirement is the hardest. Vera causa–True cause. The establishment of a causal relationship based on existence, capability, and verification. John Herschel (Preliminary Discourse on the Study of Natural Philosophy, 1830)

In the field of computation, the nature of information processes has been described in terms of representations [6]. Representations are patterns of symbols that stand for something. A representation that stands for a method

Chapter

1

Correlations

3

of evaluating a function is an algorithm, while a representation that stands for values is called data. A machine will manipulate an algorithm to transform an input data representation to an output data representation. Interestingly, it has been shown that there is no algorithm for finding the shortest possible representation of something [7]. If this has analogy to cause and effect, then there is no set or general means of deriving a cause. Correlations are a special case of cause-and effect-relationships, where the cause is not definitively established. They are generally presented as a relationship between two of more variables. In complex systems, an observed effect may arise from a multiplicity of causes, which might be further complicated by interactions. It is still of interest to deduce these causes. An ultimate goal is to determine if there is a way to prove that a relationship expressed in a correlation is due to a true cause rather than a simple coincidence. Analysis of the way cause-and-effect relationships are developed in science indicates that this must involve the development of a mathematical relation based on a physical model of the situation.

SCIENTIFIC LAWS AND CORRELATIONS It is useful to discuss first the simplest of cause-and-effect relationships, where the most clear manifestations of causality are found. These are well known as they are the familiar natural laws. There is actually a whole hierarchy of these relationships, which may be roughly classified as follows: Principles, Theories, Laws, Properties, Effects, Equations, Dimensionless Numbers, Criteria, Approximations, Factors, and Curves. In general, the lower down the hierarchy, the greater the number of assumptions needed for the obtention of the relationship. The exact order is easily off by one or two levels depending on the relationship (Figure 1.1). Correlations are also relationships, and although they may have a rationalization, they lack a deep understanding of all the factors that cause the phenomena.

Principles Principles may be considered to be fundamental axioms or tenets with foundations in logic. Coverage of this area is in the realm of philosophy and is beyond the scope of this chapter. Mention is only made of several principles: Principle of sufficient reason, principle of identity, principle of contradiction, principle of excluded middle. Sometimes, the term principle is used loosely to convey the idea of fundamentality in a given area.

Theories Theories are general frameworks for describing and understanding natural phenomena and although backed by considerable evidence and well accepted are generally difficult to prove. In fact, positive proof of veracity is not considered a

4

Inorganic, Polymeric and Composite Membranes

Principles

Theories Laws Properties

Generally, increasing assumptions

Effects Equations Criteria Approximations Factors Curves FIGURE 1.1 Hierarchy of relationships to explain natural phenomena.

possibility, but disproof by counterexample or deviation is possible. This reflects the nature of the scientific method, where a hypothesis is followed by experimental verification, which leads to confirmation or further refinement of the hypothesis. There are several levels of theories. Theories may be extremely general and all encompassing (number theory, classical electromagnetic theory, theory of relativity, theory of evolution, theory of plate tectonics), may be under development (string theory, gravitational theory), or may be accepted but relatively specific (Debye–Huckel theory of solutions, thermodynamic theory of corresponding states, Flory–Huggins polymer dissolution theory, Temkin nonuniform surface theory, Hume–Rothery alloy theory). In science, a theory is any broad hypothesis that provides an explanation for an observation. Thus, the term theory can have a high or a low level, with those at a lower level implicitly or explicitly depending on a higher-order theory.

Laws Laws are fundamental relationships between state variables, which are applicable over broad but generally not unlimited conditions (Table 1.1). State variables or properties are quantities that describe a system. Laws were often

Chapter

1

5

Correlations

TABLE 1.1 Laws. Fundamental Relationships Between State Variables Name

Equation

Variables

Gravitational law

F ¼ GM1M2/r

G ¼ gravitational constant M1, M2 ¼ masses r ¼ distance

Coulomb’s law

F ¼ keQ1Q2/r2,

ke ¼ Coulomb’s constant c ¼ speed of light mo ¼ magnetic constant Q1, Q2 ¼ charge of particles r ¼ distance between particles

2

ke ¼ c4pmo 2

Ideal gas law

pV ¼ nRT

p ¼ pressure V ¼ volume n ¼ number of moles R ¼ gas constant T ¼ temperature

First law of thermodynamics

E¼QþW

E ¼ energy of system Q ¼ heat input into the system W ¼ work done on the system

Raoult’s law

p ¼ pA*xA þ pB*xB

p ¼ total vapor pressure of ideal solution pi* ¼ vapor pressure of pure i xi ¼ mole fraction of i

Law of mass action

r ¼ kCACB

r ¼ rate of reaction k ¼ rate constant CA, CB ¼ concentrations of species

Hook’s law

F ¼ kx

F ¼ force, k ¼ spring constant k ¼ distance

Fick’s first law

J ¼ D d C/d x

J ¼ flux of species D ¼ diffusivity dC/dx ¼ gradient of concentration

Fick’s second law

@C @2C ¼D 2 @t @x

Darcy’s law

J¼

k @P m @x

@C/@t ¼ change in concentration with time D ¼ diffusivity @ 2C/@x2 ¼ second derivative of concentration J ¼ flux k ¼ permeability m ¼ viscosity @P/@x ¼ gradient in pressure

empirically discovered and, at the highest level, cannot be derived from more fundamental relations. It will be noted that many fundamental scientific equations represent a relationship between properties and are usually not thought as representing a causal situation. In statistics, this is known as an

6

Inorganic, Polymeric and Composite Membranes

identity relation. For example, the ideal gas law, pV ¼ nRT, which was derived by E´mile Clapeyron [8] from observations by Boyle and Charles, simply relates the pressure, volume, temperature, and number of moles of a dilute, noninteracting gas in a closed space. Some laws are not expressible as formulas, for example, the law of definite proportions, due to Joseph Proust [9], which describes the set stoichiometry by which certain reactions proceed.

Properties Properties are attributes of a system and may be classified as intensive or extensive. An intensive property is not dependent on the size or the amount of material in the system, that is, it is scale invariant. An extensive property does depend on the size or the amount of material in the system. Of the variables in the ideal gas law, pressure and temperature are intensive, while volume and number of moles are extensive, depending on the amount of material. Usually, quantities are expressed in terms of the intensive variables: V ¼ nRT/p. Thus, at constant number of moles, volume increases linearly with temperature and inversely with pressure. A class of properties, known as colligative properties, depend on the number of molecules in a given volume of solvent, and not on the properties (e.g., size or mass) of the molecules [10] (Table 1.2). The formulas can be derived from thermodynamic equations, such as the Clausius–Clapeyron equation and Raoult’s law.

Effects Effects are relations between an action or field and another action or field (Table 1.3). These are ubiquitous and about 5000 are known, with about a tenth used in practice by scientists and engineers (http://en.wikipedia.org/wiki/ List_of_effects). They often start as an observation of a natural phenomenon, TABLE 1.2 Colligative Properties Name

Equation

Variables

Boiling point elevation

DTb ¼ Kbmi, Kb ¼ RTb2M/DHv

Kb ¼ ebullioscopic constant Tb ¼ boiling point of solvent M ¼ molecular weight of solvent DHv ¼ heat of vaporization of solvent m ¼ molality i ¼ Van’t Hoff factor

Freezing point depression

DTf ¼ Kfmi

Kf ¼ cryoscopic constant m ¼ molality i ¼ Van’t Hoff factor

Osmotic pressure

P¼

nRTi V

n ¼ number of moles of solute i ¼ Van’t Hoff factor V ¼ volume

Chapter

1

7

Correlations

TABLE 1.3 Effects. Relations Between an Action or Field and Another Action or Field Name

Equation

Variables

Photoelectric effect

Emax ¼ hn  ’

Emax ¼ maximum energy of ejected electrons n ¼ frequency of light f ¼ work function

Compensation effect

lnA ¼ bE þ c

A ¼ pre-exponential factor E ¼ activation energy b, c ¼ empirical constants

Gibbs–Thomson effect

  p Rcritical , ¼ exp peq R Rcritical ¼

2gVatom kB T

p ¼ pressure peq ¼ equilibrium pressure R ¼ radius of droplet g ¼ surface tension Vatom ¼ atomic volume kB ¼ Boltzmann constant

followed by quantitation of the variables involved, and then a physical explanation. An example is the photoelectric effect, the ejection of electrons from matter by the application of light. It was first observed by Hertz in 1887, but quantitative measurements were made later by Alexandr Stoletov (1888–1891), J. J. Thomson (1899), and Philipp Lenard (1902), and a mathematical explanation was finally given by Albert Einstein (1905) [11] based on the concept of light quanta formulated by Max Planck (1901) to describe blackbody radiation. Another example is the Gibbs–Thomson effect [12]. In 1871, William Thomson (Lord Kelvin) explained that small samples of a material melt at a lower temperature than the bulk because of a combination of surface energy and surface curvature [13]. This was later placed in thermodynamic context using the methodology of Gibbs who showed the effect of curvature on equilibrium [14]. Some effects do not have accepted physical explanations, for example, the compensation effect in heterogeneous catalysis [15]. Some effects do not have associated equations. An example is the gauche effect in stereochemistry that describes the stability of conformational isomers where two vicinal groups are separated by a 60 torsional angle [16]. Another example is the Kirkendall effect that describes the diffusional motion of atomic markers when an alloy is placed in contact with a metal [17].

Equations Equations are explicit mathematical relations between variables that are either numerical fits to extensive empirical data, or derived expressions from certain physical models subject to various assumptions (Table 1.4). Examples are the

TABLE 1.4 Equations Explicit Mathematical Relations Between Variables Name

Equation

Langmuir equation

nads KP ¼ ntot 1 þ KP

BET equation

  1 c1 P 1 ¼ þ v½ðP0 =PÞ  1 vm P 0 vm c

Arrhenius equation

  E k ¼ A exp  RT

WLF master curve equation Antoine’s equation for vapor pressure

log10 aT ¼

Variables

17:44ðT  Tg Þ 51:6 þ T  Tg

A logðPv Þ ¼  þ B T

nads ¼ moles adsorbed ntot ¼ total number of sites K ¼ adsorption equilibrium constant P ¼ pressure P ¼ equilibrium pressure P0 ¼ saturation pressure v ¼ adsorption volume vm ¼ monolayer
adsorption volume EL c ¼ exp E1RT E1 ¼ heat of adsorption of first layer EL ¼ heat of adsorption of subsequent layers ¼ heat of liquefaction k ¼ rate constant, E ¼ activation energy R ¼ gas constant aT ¼ shift factor Tg ¼ glass transition temperature Pv ¼ vapor pressure A, B ¼ empirical constants

Clausius-Clapeyron equation

dT L ¼ dV TDV

re3

Hagen-Pouiseuille equation

P¼

Knudsen permeance equation

P¼

 1=2 edP 8 tL 9pMRT

Silica membrane permeation equation

P¼

  2 3=2  0:2 1 d2 h sh2 ðNS =NA Þ  eDEK =RT 2 6L h 2pmkT 8p IkT ðehn =2kT  ehn =2kT Þ2

2ð1  eÞ2 mta2v

L ¼ latent heat of phase transition T ¼ temperature of phase transition DV ¼ volume change of phase transition P ¼ permeance r ¼ density e ¼ void fraction m ¼ viscosity t ¼ tortuosity av ¼ pore area/membrane volume P ¼ permeance e ¼ void fraction dP ¼ pore diameter t ¼ tortuosity L ¼ membrane thickness M ¼ permeate molecular weight P ¼ permeance d ¼ jump distance h ¼ Planck’s constant m ¼ permeate mass k ¼ Boltzmann constant s ¼ symmetry factor n* ¼ vibrational frequency DEK ¼ activation barrier

10

Inorganic, Polymeric and Composite Membranes

TABLE 1.5 Dimensionless Numbers. Collections of Variables for Estimates of the Behavior of a System Name

Equation vLr m

Variables

Reynolds

Re ¼

Pe´clet

Pe ¼ Lv D

Ratio of rate of mass flow by advection (convection) to rate of mass diffusion L ¼ characteristic length, v ¼ velocity, D ¼ diffusion coefficient

Dahmko¨hler

Da ¼ kCon  1t

Ratio of reaction rate to mass transport rate k ¼ rate constant Co ¼ concentration n ¼ order of reaction t ¼ time

Prandtl

Pr ¼

Nusselt

NuL ¼ hL kf

n m=r ¼ a k=rcP

Ratio of fluid inertial and viscous forces v ¼ fluid velocity L ¼ characteristic distance r ¼ density m ¼ viscosity

Ratio of viscous diffusion rate to thermal diffusion rate n ¼ kinematic viscosity ¼ m/r [¼] m2 s 1 a ¼ thermal diffusivity ¼ k/rcP [¼] m2 s 1 m ¼ dynamic viscosity [¼] Pa s k ¼ thermal conductivity [¼] W m 1 K 1 cP ¼ specific heat [¼] J kg 1 K 1 r ¼ density [¼] kg m 3 Ratio of convective heat transfer coefficient to conductive heat transfer coefficient L ¼ characteristic length [¼] m kf ¼ thermal conductivity [¼] W m 1 K 1 h ¼ convective heat transfer coefficient [¼] W m2 K 1

Langmuir equation [18] for adsorption, the BET equation for surface area [19], the Arrhenius equation, the WLF master equation for stress relaxation time–temperature superposition in polymers [20], Antoine’s equation for vapor pressure [21], and the Clausius–Clapeyron equation for phase transitions. The permeation of gases in porous solids is described by the Knudsen equation [22,23] and the Hagen–Pouiseuille equation [24]. For vitreous glass and dense silica membranes, an equation derived from statistical mechanics fits the data over wide range of conditions and permeating species [25–27].

Dimensionless Numbers Dimensionless numbers in many fields of engineering are collections of variables that provide order-of-magnitude estimates about the behavior of a system (Table 1.5). They are often derived by combining coefficients from differential equations and are oftentimes a ratio between two physical quantities.

Chapter

1

11

Correlations

The dimensionless numbers can be related to other dimensionless variables or quantities through empirical relations. For example, the Dittus–Boelter equation is an explicit function for calculating the Nusselt number for turbulent flow from the Reynolds number and the Prandtl number [28]. NuD ¼ 0:023ReD Pr n 4=5

ð1:1Þ

where n ¼ 0.4 for heating of the fluid and n ¼ 0.3 for cooling of the fluid. The Sieder–Tate correlation is another means for estimating the Nusselt number for turbulent flow from the Reynolds number and Prandtl number, [29] where m is the fluid viscosity at the bulk fluid temperature and mS is the fluid viscosity at the heat-transfer boundary surface temperature  0:14 4=5 1=3 m ð1:2Þ NuD ¼ 0:027ReD Pr mS The Dittus–Boelter equation is an example of an explicit function, while the Sieder–Tate correlation is an example of an implicit function. Explicit functions give a prescription for determining the output value of a function y in terms of the input value x, that is, y ¼ f(x). For the Dittus–Boelter equation, knowledge of Re and Pr gives the Nusselt number. Implicit functions provide a value of y from solution of an equation of the form F(x,y) ¼ 0. The Sieder–Tate correlation analyzes the system as a nonlinear boundary value problem and needs to be solved iteratively, as the viscosity factor changes as the Nusselt number changes. These examples demonstrate that a quantity does not have to be expressed as an explicit function to be subject to a correlation. Cause and effect is neither necessarily linear nor explicit.

Criteria Criteria are relations that describe the range of parameter space where a phenomenon is expected to apply or to be valid in (Table 1.6). Some criteria are not expressible as a mathematical relation but give an experimental prescription. For example, the Koros–Nowak criterion for supported catalysts suggests that no mass transfer limitations are present if the activity of two samples with different concentrations of active phase at the same dispersion (particle size) are equal [30]. The Weisz Prater criterion prescribes when diffusion in catalyst pellets will be important [31]. The Order Hierarchy criterion makes a prediction on when a 1-dimensional or a 2-dimensional description of a membrane reactor (MR) is needed [32].

Approximations, Factor, Curves Approximations are usually assumptions used in solving equations (Table 1.7). An example is the steady-state approximation, which states that the rate of change of the concentration of an intermediate is zero at the steady state. Factors

12

Inorganic, Polymeric and Composite Membranes

TABLE 1.6 Criteria. Relations that Describe the Range of Validity in Parameter Space of a Phenomenon Name

Equation

Variables

Rr < 0:3 CS Deff 2

Weisz–Prater criterion

 Order hierarchy criterion

FDP DDC=r



R ¼ volumetric rate r ¼ catalyst radius CS ¼ surface concentration Deff ¼ effective diffusivity

  Pe uC0 < 0:01 dP L rb R

F ¼ permeation rate DP ¼ transmembrane partial pressure difference Deff ¼ diffusivity coefficient DC ¼ concentration difference r ¼ radius of the reactor Pe ¼ Peclet number, Pe ¼ udDP D ¼ dispersion coefficient dp ¼ catalyst particle diameter L ¼ length of the reactor u ¼ velocity of the gas C0 ¼ reactant initial concentration rb ¼ catalyst density R ¼ specific reaction rate reactor

TABLE 1.7 Approximations, Factor, Curves Stirling’s approximation

ln N ! ¼ N ln N  N

Steady-state approximation

dCI dt

Effectiveness factor

¼

¼0

CI ¼ concentration of intermediate

  3 1 1  , ’ tanh’ ’ 

’ ¼ RP Demand curve

N ¼ any positive number

k De

1=2

P ¼ a/b  1/bQ

RP ¼ radius of pellet k ¼ rate constant De ¼ effective diffusivity

P ¼ price Q ¼ quantity a, b ¼ empirical constants

are quantities that affect other variables (Table 1.7). The effectiveness factor in reactions in catalyst pellets measures the relative importance of diffusion rate and reaction rate. Curves are graphical representations of the relationship between

Chapter

1

13

Correlations

variables (Table 1.7). These are commonly employed in the field of economics. For example, the Beveridge curve is an inverse relation between job vacancies and jobless rate, the demand curve is the relationship between price and quantity, and the Phillips curve is the relationship between unemployment and inflation. These curves are empirical and shift with changing conditions.

Correlations In the field of statistics, correlations constitute any number of relations between two or more random variables or observed data values. Correlation refers to lack of independence between the variables and is measured in various ways. The most common is the Pearson correlation coefficient (r), which measures the linear linkage between two variables (even if the variables are not related linearly) [33]. It is obtained by dividing the covariance (cov) of two random variables (X, Y) with mean values (mX, mY) by the product of their standard deviations (sX, sY) and utilizes the expected value function (E). rXY ¼ corrðX; YÞ ¼

covðY; YÞ EðX  mX ÞðY  mY Þ ¼ sX sY sX sY

ð1:3Þ

The Pearson correlation ranges from þ 1 (for a perfectly positive linear relation) to  1 (for a decreasing relation) and is zero for uncorrelated independent variables. In statistics, it is well established that correlations cannot be used to infer a causal relation [34] as expressed in the maxim below. Correlation does not imply causation

This is because sometimes variables are highly correlated, but are not related by a causal relationship. For example, the ringing of a bell may be associated with the beginning of a recess, but there is no causality. This is not to say though that there is no potential cause. It is well understood that smoking and heart disease are correlated,but therelationshipbetween brainstructure andbehaviorisless manifest. The previous section reviewed a large number of types of equations on a hierarchical basis. Some represented fundamental relations between variables that were statements of identity and had no causal nature. Others were derived or observed relations that could be associated with a cause. Study of these relations leads to the following generalization. The only way to prove causality in a correlation is to uncover the fundamental physical phenomenon that causes it, to describe it mathematically, and to show that it leads to verifiable predictions.

It is fascinating that correlations may be part of a more general principle, and the search for that is a great motivation. An interesting question is how many data points are necessary to establish a correlation within reasonable probability, say above 90%. Take the values V1, V2, … Vn of a given property for a series of samples with some characteristic

14

Inorganic, Polymeric and Composite Membranes

C1, C2, … Cn, that say is increasing monotonically. It can be said that the property correlates positively with the characteristic if the series of V’s also increases monotonically (or negatively if it decreases). The following analysis will ignore the small chance of equal values, and the boundaries and extent of the values, and will assume small errors. If there are only two samples the probability of a positive correlation is ½, as the value of the second sample could be higher or lower than the first. For a third sample the value of the property can be above the highest value, below the lowest value, or in between, so the probability of a highest value is 1/3, and the overall probability of a positive correlation is 1/6. For a fourth sample a similar logic gives a probability of attaining a highest value of ¼ and an overall probability of a positive correlation of 1/24. For a fifth sample the corresponding values are 1/5 and 1/60. It should be noted that a perfect negative correlation is also acceptable, so the overall probabilities for obtaining a correlation should be multiplied by 2. It is obvious that a sample size of 3 is mostly insufficient to ascertain a correlation as there is 67% probability that this was by chance. What is surprising is that sample sizes of only 4 or 5 already give above 90% probability that the correlation is not by chance. This can be summarized as: The chances of randomly obtaining a perfectly positive (or negative) correlation decreases with sample size n as 2/n! A sample of 4 to 5 is enough to obtain a 92% or 97% certainty that a positive or negative correlation is not random. S. Ted Oyama (The Rule of 4 or 5)

IMPORTANT PROPERTIES IN MEMBRANE SCIENCE In chemistry, structure–function correlations are ubiquitous. In membrane science, function is most commonly associated with two properties, permeability and selectivity, so much work on correlations has revolved around understanding of these properties. The permeability, PMi [¼] mol m 1 s 1 Pa 1, refers to the intrinsic ability of a membrane to allow passage of a species i and relates the molar flux, Ni [¼] mol m 2 s 1, to the driving force, which is usually expressed as the pressure or the concentration difference across the membrane. The permeability divided by the thickness, L, of the membrane is the permeance [¼] mol m 2 s 1 Pa 1.   PMi ðdriving forceÞ ð1:4Þ Ni ¼ L ¼ PMi ðdriving forceÞ

ð1:5Þ

A fundamental expression for transport in membranes is derived from Fick’s First Law, which relates the flux of species i to the concentration in the inlet, cio and outlet, ciL, of a membrane of thickness L:

Chapter

1

15

Correlations

De;i ðcio  ciL Þ L

Ni ¼

ð1:6Þ

The diffusivity in Fick’s first law is the ordinary molecular diffusivity, Di [¼] m2 s 1, but in the case of membranes an effective diffusivity is used, where the porosity e and the tortuosity t of the membrane are taken into account, as well as the relative diameters of the diffusing molecules, dm, and pores, dP, through a restrictive factor Kr [35]. Dei ¼

eDi Kr t

   4 dm ðdm =dP Þ  1 Kr ¼ 1  dP

ð1:7Þ

ð1:8Þ

There are many expressions for permeance that depend on the mechanism by which the fluid species pass through the membrane. In the Hagen–Pouisselle mechanism, species transfer occurs by bulk fluid flow through large pores. Assuming that the fluid is Newtonian, and that the mean free path is small compared to pore diameter, the following expression for the average velocity n [¼] m s 1, may be derived, where dP is the diameter of the pore, m [¼] kg m s 1 is the viscosity, l is the length of the pore, po is the inlet pressure, and pL is the pressure at a distance L. v¼

dP2 ðpo  pL Þ 32ml

ð1:9Þ

With suitable manipulation, this gives rise to the following expression for the flux N¼

PM ðpo  pL Þ L

ð1:10Þ

Taking into account, the porosity e and the tortuosity t of the membrane, and the pore area per total volume, a, which is related to the pore area per membrane volume av. PM ¼

re3 2ð1  eÞ2 mta2v

av ¼

a ð1  eÞ

ð1:11Þ ð1:12Þ

In many of the laws and formulae presented so far, there have been explicit relations between variables, so the question is whether a closed-form equation is necessary to define a causal relation. The answer would appear to be “no.”

16

Inorganic, Polymeric and Composite Membranes

For example, a simple situation occurs for the case of equations that are nonlinear and, which, therefore cannot be solved for a quantity of interest. Another example is where multiple effects are coupled and is illustrated below. The flux of hydrogen in palladium membranes is commonly described by Sievert’s law [36,37], where p is a diffusion coefficient, PF is the feed H2 pressure, PP is the permeate H2 pressure, and the exponent n is 0.5. This accounts for permeance in thick membranes where the limiting process is diffusion of hydrogen atoms across the bulk.
ðnÞ JH2 ¼ pH2 PnF  PnP ð1:13Þ Oftentimes, exponents between 0.5 and 1 are observed, and this is because the overall flux is governed by a combination of external transport, surface processes, and bulk diffusion [38]. Drioli and coworkers explain that when these steps are taken into account, the following equation arises.    
1 b Diff 0:5 0:5 ðPF  PP Þ ð1:14Þ JH2 ¼ pH2 PF  PP þ aðTÞ þ 2 T   DEA ð1:15Þ aðTÞ ¼ ao exp 2RT Clearly, the foregoing expression for the flux accounts for the observed exponents even if an explicit expression for n cannot be obtained.

EXAMPLES OF CORRELATIONS IN THE MEMBRANE SEPARATION FIELD The systematic study of membrane properties has led to the observation of many correlations. A well-cited example is the work of Robeson showing plots of membrane permeability versus selectivity for gas separations with diverse polymer membranes [39]. The plots show much scatter, but the upper boundary traces a line of declining selectivity with increasing permeability, which has humorously been referred to as the line of death. Such a relationship between selectivity and permeability is commonly observed, and is readily explained by the presence of defects that allow indeterminate passage of gases, thus increasing permeance but reducing selectivity. In this section, several examples of correlations are presented that are not meant to be exhaustive. They illustrate a diversity of properties, including those of the membrane and the permeant. Overcoming the Robeson limit has been an objective in many studies and has stimulated the development of improved membranes. Park et al. have shown that control of free volume topologies in dense, vitreous polymeric membranes is possible by thermally driven segment rearrangement [40]. This results in membranes that have enhanced transport and separation performance compared

Chapter

1

17

Correlations

Permselectivity of He/CH4

10,000 Robeson’s line 1000 PPO-PVP-10 K PPO-PVP-29 K 100

PPO-PVP-40 K PPO PPO-PVP-55 K

10

PPO-PVP-90 K

1 10

100 1000 Permeability of He/Barrer

10,000

FIGURE 1.2 Correlation between permeability and permselectivity in the system of He and CH4 with Robeson’s line. Adapted from Ref. [26].

to conventional membranes. Another example is given from work by Lee et al. with carbon molecular sieve membranes [41] (Figure 1.2). By using a combination of the thermally stable polymer polyphenylene oxide (PPO) and the labile polymer polypyrrolidone (PVP) and controlling the pyrolysis temperature and molecular weight of PVP materials, both enhanced permeability and permselectivity were obtained. Higher molecular weights than 40 K gave enhanced diffusional pathways by removal of the PVP, while selectivity rose by improvement of the molecular sieving properties. The effect of permeant properties is illustrated from an example from the review by Wijmans and Baker on the solution–diffusion model [42]. A plot of permeability of various alkanes in a rubbery polymer versus the vapor pressure of the alkanes shows a maximum (Figure 1.3). This can be explained from an equation presented by the authors for the permeability coefficient of gases. Pi ¼

Di gi giðmÞ pisat

ð1:16Þ

where Di is the diffusion coefficient, gi is the affinity of the permeant for the gas phase, gi(m) is the affinity of the permeant for the membrane, and pisat is the saturation vapor pressure. The authors explain that both the saturation pressure of the permeant and the diffusion coefficient decrease with increasing molecular weight and so create competing effects on the permeability coefficient. In glassy polymers, the decrease in diffusion coefficient dominates other effects, but in rubbery polymers, the effects are more balanced. For molecular weights up to 100 permeabilities increase because Pisat dominates, but above molecular weights of 100 the diffusivity is more important. This is illustrated for simple alkanes in a silicone rubber membrane.

18

Inorganic, Polymeric and Composite Membranes

Permeability (Barrer)

100,000

C5H12 10,000 C8H18 C6H14

C4H10

C10H22

C3H8 C2H6

Increasing diffusion coefficient 1000

CH4

Increasing molecular weight

0.1 1 10 100 1000 1⫻10-3 0.01 Permeant saturation vapor pressure pisat (atm)

12

2.0

Separation factor

10 1.5 8 6

1.0

4 0.5 2 0

Permeance ( 10−8 mol m−2 s−1 Pa−1)

FIGURE 1.3 Permeability coefficient of n-alkanes in polydimethylsiloxane as a function of saturation pressure. Adapted from Ref. [42].

0.0 5

10 15 20 25 30 PDMS prepolymer composition (wt%)

FIGURE 1.4 Effect of PDMS prepolymer composition on performance of PDMS membranes. Adapted from Kim et al. using an estimated membrane thickness of 1 mm to convert GPU to mol m 2 s 1 Pa 1 using a conversion factor of 3.35  10 10 mol m 2 s 1 Pa 1/GPU.

Another work has addressed overcoming the inherent permeance–selectivity tradeoff. Kim et al. studied polydimethylsiloxane (PDMS)-coated membranes on a polysulfone support for propylene recovery from an off-gas stream [43]. These membranes showed the usual decline in selectivity with increase in permeance (Figure 1.4). However, the incorporation of fumed silica allowed breaking of the tradeoff with both permeance and selectivity increase (Figure 1.5). The authors

1

19

Correlations

9

1.6

Separation factor

Permeance (Sol-gel silica)

1.4

8

1.2

7

Selectivity (fumed silica) Permeance (fumed silica)

6

1.0 Selectivity (Sol-gel silica)

0.8

5 0

5

10

15

C3H6 permeance (10−8 mol m−2 s−1 Pa−1)

Chapter

20

Silica content (wt%) FIGURE 1.5 Effect of silica content on performance of composite membranes made of PDMS–silica nanoparticles coupled with mercaptosilane. Adapted from Kim et al. using an estimated membrane thickness of 1 mm to convert GPU to mol m 2 s 1 Pa 1 using a conversion factor of 3.35  10 10 mol m 2 s 1 Pa 1/GPU.

0.9

2000

0.8

1600 1400

0.7 1200 1000

0.6

Flux (kgm−2 h−1)

Separation factor

1800

800 600

0.5

400 0

2

4 6 PSS content (wt%)

8

10

FIGURE 1.6 Pervaporation performances of the hybrid membranes in the separation of 90 wt% THF solution at 50  C. Adapted from Ref. [44].

believe the fumed silica enhanced the sorption of propylene into the membrane matrix to increase the separation factor. At the same time, the silica disrupted the chain packing to increase the propylene permeance. An optimum silica nanoparticle content of 15 wt% was found, as excess silica caused problems in the coating process. Another example of overcoming the selectivity–permeance correlation is found in the work of Liu and coworkers [44] (Figure 1.6). They studied organic–inorganic hybrid membranes containing polyvinyl alcohol (PVA) and

20

Inorganic, Polymeric and Composite Membranes

polysilisesquioxane (PSS) for the dehydration of tetrahydrofuran by pervaporation. They found that both permeance and selectivity of the membranes increased up to PSS contents of 2 wt%, and attributed the increase in permeance to a decrease in the crystalline region in the hybrid membranes with increasing PSS content. The selectivity increase was thought to be due to increase in the hydrophilicity at PSS levels below 3 wt%. Other correlations relate membrane structure with performance. Permeability (P) is often correlated with the reciprocal fractional free volume (FFV) through the expression [45].   B P ¼ Aexp  ð1:17Þ FFV where A and B are adjustable parameters. An example is given below from the work by Senthilkumar and Reddy for a series of poly(dimethylsiloxane) (PDMS) films substituted with different bulky groups in the pendant chain having amino and hydroxyl functional groups [46] (Figure 1.7). The specific polymers used had aniline (DMA), phenylaniline (PPh), diphenylamine (DPP), and diisopropylphenylaniline (DIP) functionalities. Although the polymers adhered to the standard permeability–selectivity correlation, there was a dependence of permeability on structure. With the incorporation of increasingly bulkier substituents, the FFV, which characterizes the amount of free voids in the polymer matrix, decreased and the permeability coefficient also was found to decrease. In the membrane reactor (MR) field, the enhancement in yield or conversion due to simultaneous reaction and separation has been correlated with a parameter denoted as the operability level coefficient (OLC), defined as the ratio of product permeation and product formation rates, and related to the inverse combination of the Damko¨hler number and the Peclet number (1/DaPe) [47]. The OLCs for

18.0 DIP

In(P)

17.5

DMA DPP

17.0 PhP 16.5 PDMS 16.0 4.5

5.0 5.5 6.0 6.5 Reciprocal fractional free volume (1/FFV)

7.0

FIGURE 1.7 Plot showing correlation between O2 permeability against reciprocal fractional free volume for the functionalized films. Adapted from Ref. [46].

Chapter

1

21

Correlations

Conversion enhancement (%)

100

80

60

40

MSR, Tsuru et al. MSR, Tong and Matsumura MDR, Lee et al. MSR, Patil et al. EtOHSR, Lim and Oyama MSR, Tong and Matsumura MSR, Hacarlioglu et al. MDR, Irusta et al. MeOHSR, Basile et al. MeOHSR, Lee et al.

20

0 0.0

0.6 0.2 0.4 Operability level coefficient/OLC

0.8

FIGURE 1.8 Correlation between conversion enhancement and OLC. OLC ¼ Operability limit coefficient ¼ Permeance rate/Reaction rate. Adapted from Ref. [47].

product hydrogen formation in previously reported MRs for methane dryreforming (MDR), methane steam-reforming (MSR), methanol steam-reforming (MeOHSR), and ethanol steam-reforming (EtOHSR) were calculated. product permeation rate ðpermeanceÞðareaÞðDPÞ ¼ ðrateÞðvolumeÞ productformation rate 1 ¼ DaPe reaction conditions

OLC ¼

ð1:18Þ

For values of OLCs ranging from 0.03 to 0.78, a clear universal trend for increasing conversions and hydrogen yields with increasing OLC was observed for these different types of reforming reactions (Figure 1.8). The OLC curve calculated from a numerical simulation without adjustable parameters was found to closely approximate experimental data obtained from the MRs and was shown not to depend on the assumed kinetics. This study confirms that hydrogen selectivity has a substantial influence on conversion and hydrogen yield enhancements in an MR, and demonstrates that a hydrogen selectivity of 100 is sufficient to achieve high performance in a MR. The work on OLC was extended to the ethanol-reforming reaction by Lim et al. [48]. Both the OLC and the inverse product of the Pe and Da numbers correlated with the enhancement in yield or conversion (Figure 1.9). The aforementioned examples above about correlations in membranes and MRs illustrate the breadth of the variables that may be encountered. Although no explicit laws or formulas emerge, reasonable explanations with physical interpretations are still possible. Further analysis of these correlations could result in more general relationships.

22

Inorganic, Polymeric and Composite Membranes

Conversion or H2 yield enhancement (%)

1/DaPe = (Permeation rate/reaction rate)inlet 100

0

2

4

6

8

10

80 1/DaPe 60 OLC 40

20

0 0.0

0.6 0.8 0.2 0.4 OLC = (Permeation rate/reaction rate)actual

1.0

FIGURE 1.9 Correlation of H2 conversion or yield with ratios of permeation rate and conversion rate. Adapted from Ref. [48].

SUMMARY Correlations are used in many fields of endeavor to organize and present data, and it is important to understand the potential and the limitations of the method to use it effectively. Correlations are related to cause-and-effect relations, and the subject is explored in depth. A historical account of the development of the concept of causality is presented that ranges from the thinking of Aristotle to the contributions of William of Ockham, Bacon, Newton, and Herschel. Causality is tied to the explanation of phenomena, so the many categories of scientific concepts are reviewed and illustrated. It is concluded that for a correlation to be associated with a cause requires three conditions: a physical basis or explanation, a mathematical description, and the ability to make predictions.

ACKNOWLEDGMENTS For support of this work, the author acknowledges the Director, National Science Foundation, Division of Chemical, Bioengineering, Environmental, and Transport Systems (CBET) under grant CBET-084316, the National Energy Technology Laboratory under the NETL-RUA program grant number 5.681.884.001, the Mombukagakusho Kakenhi grant-in-aid Kiban kenkyu B 22-360,335.

REFERENCES [1] J. Diamond, & J.A. Robinson (Eds.), Natural Experiments of History, Harvard University Press, Cambridge, MA, 2010.

Chapter

1

Correlations

23

[2] J. Mahoney, History as laboratory, Science 327 (2010) 1578–1579. [3] J. Diamond, Guns, Germs, and Steel: The Fates of Human Societies, Norton, New York, 1977. [4] R. Ariew, Ockham’s Razor: A Historical and Philosophical Analysis of Ockham’s Principle of Parsimony, University of Illinois, Champaign-Urbana, 1976. [5] K. Thomson, Darwin’s literary models, Am. Sci. 98 (2010) 196–199. [6] P.J. Denning, The great principles of computing, Am. Sci. 98 (2010) 369–372. [7] G. Chaitin, Meta Math! The Quest for Omega, Vintage Press, New York, 2006. [8] E. Clapeyron, Me´moire sur la puissance motrice de la chaleur, Journal de l’E´cole Polytechnique XIV (1834) 153–190. [9] J.-L. Proust, Researches on copper, Ann. Chim. 32 (1799) 26–54. [10] P.W. Atkins, Physical Chemistry, fourth ed., Oxford University Press, Oxford, 1994, p. 222–225. ¨ ber einen die Erzeugung und Verwandlung des Lichtes [11] A. Einstein, U BetreffendenHeuristischen Gesichtspunkt, Ann. Phys. 17 (1905) 132–148. [12] M. Perez, Gibbs–Thomson effects in phase transformations, Scr. Mater. 52 (2005) 709–712. [13] W.T. Thomson, On the equilibrium of vapour at a curved surface of liquid, Philos. Mag. 42 (1871) 448–452. [14] J.W. Gibbs, The collected works of J. Willard Gibbs, Academic Press, New York, 1928. [15] M.C. Wilson, A.K. Galwey, Compensation effect in heterogeneous catalytic reactions including hydrocarbon formation on clays, Nature 343 (1973) 402–404. [16] S. Wolfe, Gauche effect. Stereochemical consequences of adjacent electron pairs and polar bonds, Acc. Chem. Res. 5 (1972) 102–111. [17] A.D. Smigelskas, E.O. Kirkendall, Zinc diffusion in alpha brass, Trans. AIME 171 (1947) 130–142. [18] I. Langmuir, The constitution and fundamental properties of solids and liquids. Part 1. solids, J. Am. Chem. Soc. 38 (1916) 2221–2295. [19] S. Brunauer, P.H. Emmett, E. Teller, Adsorption of gases in multimolecular layers, J. Am. Chem. Soc. 60 (1938) 309–319. [20] M.L. Williams, R.F. Landel, J.D. Ferry, The temperature dependence of relaxation mechanisms in amorphous polymers and other glass-forming liquids, J. Am. Chem. Soc. 77 (1955) 3701–3707. [21] C. Antoine, Tensions des vapeurs; nouvelle relation entre les tensions et les tempe´ratures, Comptes Rendus des Se´ances de l’Acade´mie des Sciences 107 (1888) 681–684, 778–780, 836–837. [22] M. Knudsen, The law of the molecular flow and viscosity of gases moving through tubes, Ann. Phys. 28 (1909) 75. [23] D. Lee, S.T. Oyama, Gas permeation characteristics of a hydrogen selective supported silica membrane, J. Membr. Sci. 5428 (2002) 1–16. [24] S.P. Sutera, R. Skalak, The history of Poiseuille’s law, Annu. Rev. Fluid Mech. 25 (1993) 1–19. [25] S.T. Oyama, D. Lee, P. Hacarlioglu, R.F. Saraf, High permeability in a nonporous system: theory of hydrogen permeability in dense silica membranes, J. Membr. Sci. 244 (2004) 45–53. [26] D. Lee, L. Zhang, S.T. Oyama, S. Niu, R.F. Saraf, Synthesis, characterization and gas permeation properties of a hydrogen permeable silica membrane supported on porous alumina, J. Membr. Sci. 231 (2004) 117. [27] Y. Gu, S.T. Oyama, High molecular permeance in a pore-less ceramic membrane, Adv. Mater. 19 (2007) 1636–1640.

24

Inorganic, Polymeric and Composite Membranes

[28] F. P. Incropera, D.P. DeWitt, Fundamentals of Heat and Mass Transfer, fourth ed., Wiley, New York, 1996, p. 493. [29] E.N. Seider, E. Tate, Heat transfer and pressure drop of liquids in tubes, Ind. Eng. Chem. 28 (1936) 1429. [30] R.M. Koros, E.J. Nowak, A diagnostic test of the kinetic regime in packed-bed reactors, Chem. Eng. Sci. 22 (1967) 470. [31] P.B. Weisz, C.D. Prater, Interpretation of measurements in heterogeneous catalysis, Adv. Catal. 6 (1959) 143. [32] S.T. Oyama, P. Hacarlioglu, The boundary between simple and complex descriptions of membrane reactors: the transition between 1-D and 2-D analysis, J. Membr. Sci. 337 (2009) 188–199. [33] J.L. Rodgers, W.A. Nicewander, Thirteen ways to look at the correlation coefficient, Am. Stat. 42 (1988) 59–66. [34] John Aldrich, Correlations genuine and spurious in Pearson and Yule, Stat. Sci. 10 (1995) 364–376. [35] R.E. Beck, J.S. Schultz, Hindered diffusion in microporous membranes with known pore geometry, Science 170 (1970) 1302–1305. [36] R.C. Hurlbert, J.O. Konecny, Diffusion of hydrogen through palladium, J. Chem. Phys. 34 (1961) 655. [37] A. Caravella, G. Barbieri, E. Drioli, Modelling and simulation of hydrogen permeation through supported Pd-based membranes with a multicomponent approach, Chem. Eng. Sci. 63 (2008) 2149–2160. [38] A. Caravella, F. Scura, G. Barbieri, E. Drioli, Sieverts law empirical exponent for Pd-based membranes: critical analysis in pure H2 permeation, J. Phys. Chem. B 114 (2010) 6033–6047. [39] L.M. Robeson, Correlation of separation factor versus permeability for polymeric membranes, J. Membr. Sci. 62 (1991) 165–185. [40] H.B. Park, C.H. Jung, Y.M. Lee, A.J. Hill, S.J. Pas, S.T. Mudie, et al., Polymers with cavities tuned for fast selective transport of small molecules and ions, Science 318 (2007) 254–258. [41] H.-J. Lee, H. Suda, K. Haraya, S.H. Moon, Gas permeation properties of carbon molecular sieving membranes derived from the polymer blend of polyphenylene oxide (PPO)/polyvinylpyrrolidone (PVP), J. Membr. Sci. 296 (2007) 139–146. [42] J.G. Wijmans, R.W. Baker, The solution-diffusion model: a review, J. Membr. Sci. 107 (1995) 1–21. [43] H. Kim, H.-G. Kim, S. Kim, S.S. Kim, PDMS–silica composite membranes with silane coupling for propylene separation, J. Membr. Sci. 344 (2009) 211–218. [44] Q.G. Zhang, Q.L. Liu, F.F. Shi, Y. Xiong, Structure and permeation of organic–inorganic hybrid membranes composed of poly(vinyl alcohol) and polysilisesquioxane, J. Mater. Chem. 18 (2008) 4646–4653. [45] L.G. Toy, K. Nagai, B.D. Freeman, I. Pinnau, Z. He, T. Masuda, et al., Pure-gas and vapor permeation and sorption properties of poly[1-phenyl-2-[p-(trimethylsilyl)phenyl]acetylene] (PTMSDPA), Macromolecules 33 (2000) 2516. [46] U. Senthilkumar, B.S.R. Reddy, Polysiloxanes with pendent bulky groups having aminohydroxy functionality: structure–permeability correlation, J. Membr. Sci. 292 (2007) 72–79. [47] S.T. Oyama, H. Lim, An operability level coefficient (OLC) as a useful tool for correlating the performance of membrane reactors, Chem. Eng. J. 151 (2009) 351–358. [48] H. Lim, Y. Gu, S.T. Oyama, Reaction of primary and secondary products in a membrane reactor: studies of ethanol steam reforming with a silica–alumina composite membrane, J. Membr. Sci. 351 (2010) 149–159.

Chapter 2

Review of Silica Membranes for Hydrogen Separation Prepared by Chemical Vapor Deposition Sheima Jatib Khatib1, S. Ted Oyama1,2,*, Ka´tia R. de Souza1,3 and Fa´bio B. Noronha3 Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA 2 Department of Chemical Systems Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan 3 Instituto Nacional de Tecnologia—INT, Av. Venezuela 82, CEP 20081-312, Rio de Janeiro, Brazil * Corresponding author: E-mail addresses: [email protected], [email protected] 1

INTRODUCTION Silica Membranes for Hydrogen Separation Hydrogen is an important industrial feedstock for the production of fuels and many chemicals [1,2] and in fuel cell applications [3], and the purification of hydrogen is an important unit operation. Hydrogen separation using membranes has emerged as an important technology, and although polymeric membranes have seen some application [4], they have limited permeance and selectivity, and inorganic materials have attracted attention. The most well-known material is palladium, but it suffers from high costs, and susceptibility to metallic failure by hardening, and poisoning by sulfur and other extraneous elements. This review focuses on silica-based membranes prepared by chemical vapor deposition (CVD), which are potentially low cost, are thermally stable, and are immune to poisons. Practical application of these membranes requires high permeation rates as well as good selectivities, which can be obtained with membranes of low thickness and an absence of cracks and pinholes. In addition, the membranes must be mechanically strong for use in practical equipment and have long life and resistance to poisons. To meet these requirements (mechanical strength and high permeability), ceramic membranes can be produced by applying a thin film Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

25

26

Inorganic, Polymeric and Composite Membranes

of the selective material on a thick porous support, usually in tubular form. For this purpose, film deposition technology has been exploited in membrane science. For potential industrial applications like high-temperature hydrogen separation and simultaneous reaction and separation, inorganic silica membranes offer unique advantages, such as high selectivities, and high stability at elevated temperatures and in chemically aggressive atmospheres. And, for this reason, considerable attention has been paid to them by researchers [5–18]. Moreover, the application of these inorganic membranes in membrane reactors, using catalytically active or passive membranes has proved to be promising, since yields above equilibrium have been obtained by the continuous separation of the hydrogen product from the reaction system [19–25]. Figure 2.1 shows a typical membrane reactor applied to the steam reforming of ethanol. The supported silica membranes are made by depositing the separation layer from suitable precursors suspended in a liquid or a gaseous medium. Deposition from the liquid phase involves techniques such as dip coating in polymeric or particulate suspensions (sol–gel technique), while deposition from the gas phase is usually carried out by CVD due to its versatility. Thus, silica membranes are usually in the form of silica layers placed on ceramic supports such as porous Vycor glass [5,10] and alumina [11,12,15–18], and are deposited usually by sol–gel [13–16] techniques or CVD [5–12] of silica precursors. Sol– gel modification provides relatively high gas permeation rates, mainly due to the very thin top layers, in the order of 50–100 nm and a good selectivity as opposed to CVD methods where there is an attendant loss of permeability, though the selectivity is enhanced. The sol–gel method, however, suffers from a lack of reproducibility. FIGURE 2.1 Schematic diagram of a membrane reactor applied to the steam reforming of ethanol reaction.

C2H5OH+H2O Sweep gas

Membrane H2

Catalyst

Sweep gas and permeate/H2 rich Retentate: C2H5OH, H2O, CO, CO2, CH4 ...

Chapter

2

Review of Silica Membranes for Hydrogen Separation

27

This review will pay attention to the work that different groups have carried out with supported silica membranes for hydrogen separation prepared by CVD. But before, some basic principles and notions of CVD will be described.

Chemical Vapor Deposition: Principles CVD is a technique that modifies the properties of membrane surfaces by depositing a layer of a solid product of the same or different compound through chemical reactions in a gaseous medium surrounding the component at an elevated temperature. Depending on the type of application, the product can be grown on flat substrates, fibers, or particles. Generally, CVD systems include a system for delivering a mixture of reactive and carrier gases, and a heated reaction chamber where film formation occurs. The gas mixture (which typically consists of hydrogen, nitrogen, or argon, and volatile reactive compounds such as metal halides, carbonyls, or alkoxides) is flowed over the substrate that is heated to the desired temperature. Over the past years, different types of CVD methods have been developed, including thermal CVD, plasma-assisted CVD, and laser CVD [26]. The deposition of coatings by CVD can be achieved in a number of ways such as decomposition, oxidation, hydrolysis, or compound formation. These reactions between various constituents may occur in the vapor phase over the heated substrate, with the products depositing over the surface and forming a film. Most commonly, they proceed by adsorption of the gaseous reactants on the solid substrate followed by surface reactions. Thus, the mechanisms of CVD reactions involve reactions taking place on the solid surface and sometimes in the gas phase. Figure 2.2 illustrates the seven mechanistic steps that have been hypothesized to occur during a vapor deposition process [27]. Although CVD procedures of membrane films have not been studied in such detail, empirical conditions for growing good-quality membranes have been identified and developed. All the same, some qualitative principles have been developed to help understand the CVD process for film formation [28]. A possible simplified reaction sequence for deposition is the following:

etc.

AðgÞ ! BðgÞ þ CðgÞ

ð2:1Þ

BðgÞ ! SðsÞ

ð2:2Þ

BðgÞ þ BðgÞ ! B2ðgÞ

ð2:3Þ

B2ðgÞ þ BðgÞ ! B3ðgÞ

ð2:4Þ

B2ðgÞ þ B2ðgÞ ! B4ðgÞ

ð2:5Þ

28

Inorganic, Polymeric and Composite Membranes

Bulk gas 1 7 2

3

6

4

Boundary layer

5

Coating substrate FIGURE 2.2 Schematic diagram of the mechanistic steps that occur during the CVD process: (1) transport of reactant gases into the reaction chamber, (2) intermediate reactants formed from reactant gases, (3) diffusion of reactant gases through the gaseous boundary layer to the substrate, (4) absorption of gases onto the substrate surface, (5) single or multistep reactions at the substrate surface, (6) desorption of product gases from the substrate surface, (7) forced exit of product gases from the system. (Figure adapted from Ref. [27]).

In the scheme, B is an active gas phase intermediate that can react on the solid or can form oligomers in the gas phase. The oligomerization reactions (2.3)–(2.5) can be purely physical, reversible events related to condensation, or can involve bond breaking and formation. The growth of particles in the gas phase follows a nucleation path where a small nucleus is formed and then grows. The surface film formation can follow any number of mechanisms such as island formation, or layer-by-layer growth. In both cases, lowering the reactant concentration will reduce the first-order steps (2.1) and (2.2) and more strongly, the oligomer and particle-forming steps (2.3)–(2.5). Another possible scheme involves two gaseous reactants, C and D, reacting as follows: CðgÞ þ DðgÞ ! EðgÞ þ FðgÞ

ð2:6Þ

EðgÞ þ EðgÞ ! E2ðgÞ

ð2:7Þ

EðgÞ ! EðsÞ

ð2:8Þ

etc. The formation of the membrane pore structure depends on several factors. Among them are the size and the shape of the reacting molecules, the mechanism by which the film grows, and the secondary transformations that the film

Chapter

2

Review of Silica Membranes for Hydrogen Separation

29

undergoes in subsequent drying and heat treatment steps [28]. The species responsible for growth can be either the unconverted feed molecules A or the species B formed by decomposition of A, small oligomers or clusters. The reactive species on the surface, which may include the oligomers and clusters, coalesce to form a network that may include pores. The interior of the pores may be filled by surface diffusion, but may also be blocked by occlusion of the pore mouths. After closure of the pore mouths, further deposition occurs on the outer surface, so the density of the film will not be uniform, but will be high in the vicinity of the surface and gradually diminish toward the interior. For higher concentrations of the reactant A, and longer residence times of the reactants prior to contacting the membrane support, substantial particle formation occurs, and the resulting structure of the deposit layer becomes more complex. Factors such as particle nucleation, growth, agglomeration, and mass transfer and adhesion to the surface become important, and film growth is much faster compared to the case where purely gaseous species impinge on the surface. Because of the complicated nature of the CVD process, it is important to control the reaction conditions to obtain high-quality membranes. Making a dense membrane requires suppression of particle formation so that growth proceeds strictly by surface reactions or similar reactions involving small clusters. For this purpose, the concentration of B must remain low and growth will be slow. These conditions are appropriate for growth on mesoporous supports with pore size 2–10 nm. For membrane growth on supports with larger pore size, carrying out the deposition in the absence of particles would take a long time so pore closure and pore penetration would be deep due to the larger diffusion coefficient. Thus, CVD in macroporous supports must generally use combined deposition of particles as well as gaseous species. The reaction conditions must be more delicately balanced to achieve a particle size appropriate for the particular pore size of the support. Although particle deposition is essential in this case, it must be supplemented by heterogeneous reaction in the deposit when the inter-particle voids are to be closed.

Synthesis of Silica Membranes via Chemical Vapor Deposition Silica membranes prepared by CVD have been studied extensively for applications in selective hydrogen separations [5–12], and less commonly for other separations like nitrogen/oxygen or air/hydrocarbon separations, where the obtained selectivities are not very high. This chapter will concentrate on hydrogen separation. The objective of most of the SiO2-CVD processes reported is to close the pores of the support over a narrow range enabling selective hydrogen separation. Over the past decades, several successful demonstrations of this idea have been reported; however, flux and stability improvements especially in the presence of water vapor and at high temperatures are needed for industrial applications. Different research groups have employed variations of the CVD

30

Inorganic, Polymeric and Composite Membranes

method under different conditions in order to prepare efficient and stable silica membranes on a porous substrate. Some common aspects that have been treated in the studies, which are presented in this chapter, are: (i) silica precursors that are employed for deposition, (ii) supports on which the silica film is deposited (iii) diffusion geometry of reactants, (iv) stability studies, (v) hydrogen permeation mechanism.

Silica Precursors The deposited silica film is a product of the reaction between the gaseous silica precursor and the reactive agent, which can be oxygen, air, water, ozone, etc. depending on the reaction chosen for the CVD. Some reactions that have been employed in the literature, which are presented in this chapter, are: SiH4 oxidation, SiCl4 hydrolysis, TEOS (tetraethylorthosilicate) oxidation, TEOS thermocracking, etc. All the processes employed in the literature for the preparation of silica membranes for hydrogen separation are summarized in Tables 2.1 and 2.2. Since the deposited silica film is a product of the reaction of the gaseous precursors, different conditions and reaction geometries have been studied for the different precursors in order to optimize the permeance/ selectivity and stability of the produced membranes. Supports Two kinds of supports have been used in the literature for silica membranes prepared by CVD, Vycor glass, and alumina. Initial work was reported on porous Vycor glass and was followed by other researchers, because in spite of its relative low permeability, that sets the upper limit for hydrogen fluxes through the composite membranes, it is known to be highly selective to hydrogen. This is due to its rather narrow mesopore size distribution. Vycor glass is compatible with SiO2 in terms of thermal expansion coefficients and is therefore expected to suffer less from cracking due to thermal cycling [28]. The low inherent permeance through Vycor glass led to the use of other supports with greater pore sizes and compatibility with silica such as a-alumina and g-alumina coated a-alumina tubes. Reactant Diffusion Geometry There are two major approaches for the deposition of oxide layers using CVD, the difference lying in the manner and the location in which the reactants are brought into contact with the support surface and pores. The first method employs a “onesided” geometry and constitutes what is known as the “standard low-pressure CVD,” where the streams of the reactants (one being the silica source and the other a reactive reagent), which are usually both diluted in an inert carrier gas, are introduced on the same side of a sample support surface. If the support is tubular, both ends are sealed to prevent reactant penetration to the inside of the porous tube other than by Knudsen diffusion through the tube wall.

Chapter

2

31

Review of Silica Membranes for Hydrogen Separation

TABLE 2.1 Summary of Work Conducted in the Area of Silica Membranes Supported on Porous Vycor Glass by CVD for H2 Separation Selectivity H2/N2

H2 Permeance (10 8 mol m 2 s 1 Pa 1)

Activation Energy (kJ mol 1)

Reference

TEOS þ O2 one-sided reactant deposition

11

0.5 at 473 K

5.3

[5,29]

SiH4 þ O2 counterdiffusion deposition

2000–3000

1.4 at 873 K

35

[6]

SiCl4 þ H2O one-sided reactant deposition

1000–7000

2.2 at 873 K

37

[30]

SiCl4 þ H2O alternating reactant deposition

1000

5.0 at 873 K

24

[31]

TEOS þ O2 one-sided reactant deposition

880

1.4 at 873 K

6.0

[7]

DES þ O2 DES þ N2O counterdiffusion deposition

800 (O2) 12 (N2O)

1.6 at 723 K

No report

[8]

TEOS þ O2 one-sided reactant deposition with evacuation

100

0.43 at 473 K

No report

[32]

TEOS hightemperature atmospheric decomposition

10,000

1.0 at 873 K

38.0

[10]

TEOS þ O3 þ O2

950

3.0 at 313 K

No report

[33]

Method

The second method uses a “countercurrent” geometry (sometimes called “opposing reactants”), and in this case, the two reactants are introduced from opposite sides of the support and diffuse toward each other and react within a narrow front inside the pores of the support in the tube wall, forming a thin SiO2 film. Depending on the ratio of the reactant vapor concentrations and the relative pressures, the deposition location can be varied from the surface to the inside of the substrate. When the deposition occurs inside the support,

TABLE 2.2 Summary of Work Conducted in the Area of Silica Membranes Supported on Porous Alumina by CVD for H2 Separation Method

Selectivity H2/N2

H2 Permeance (10 8 mol m 2 s 1 Pa 1)

Activation Energy (kJ mol 1)

Reference

TEOS thermocracking one-sided reactant deposition with evacuation

1000

1–10 at 873 K

28

[34]

TEOS thermocracking one-sided reactant deposition with evacuation

No N2 detected

1–10 at 873 K

6–25

[11]

TEOS thermocracking one-sided reactant deposition with evacuation

100

30 at 873 K

11–14

[35]

TEOS thermocracking one-sided reactant deposition with evacuation

–

100 at 873 K

No report

[36]

TEOS thermocracking one-sided reactant deposition with evacuation

100 7.7

30 at 873 K 40 at 873 K 70 at 873 K

No report

[12]

TEOS thermocracking one-sided reactant deposition with evacuation

160

0.6 at 873 K

14

[37]

TEOS thermocracking one-sided reactant deposition with evacuation

4

10 at 873 K

No report

[38]

TEOS high-temperature atmospheric decomposition

–

10 at 873 K

15

[17]

TEOS high-temperature atmospheric decomposition

–

50 at 873 K

No report

[39]

TEOS þ O2 TMOS þ O2 counterdiffusion deposition

600 840

3.6 at 423 K 35.0 at 423 K

4a 8a

[40]

TMOS þ O2 counterdiffusion deposition

1000

15.0 at 873 K

20

[22]

1st: sol–gel deposition 2nd: TMOS þ O2 counterdiffusion deposition

5000

32.7 at 873 K

16

[18]

TEOS þ O2 counterdiffusion deposition

57

70.0 at 673 K

2.2

[41]

TMOS þ O2 pressurized counterdiffusion deposition

–

5.0 at 573 K

13

[42]

TMOS þ O2 counterdiffusion deposition

12,200

2.98 at 798 K

17

[43]

0.1 10.0 100.0

19 11 Not activated

[44]

TMOS þ O2 counterdiffusion deposition PTMS þ O2 one-sided reactant deposition with evacuation DMDPS þ O2 one-sided reactant deposition with evacuation a

Values referred to He.

34

Inorganic, Polymeric and Composite Membranes

first a pore narrowing occurs and further deposition leads to pore plugging; hence, the reaction stops since the reactants cannot reach each other. Figure 2.3 shows how the molecules react for these geometries, respectively, and Figure 2.4 shows how these geometries are obtained in an experimental apparatus. Different rates of deposition would be obtained for the “one-sided” and “countercurrent” deposition, due to the different level of reactant concentrations at the location of maximum deposition rate. While in “one-sided” deposition, pore plugging will occur at or near the external surface where the reactant concentrations are near their free stream values, in the counterdiffusion “One - sided” deposition

“Counter diffusion” deposition

Reactants A + B

Reactant A

A

B A

Shell side

A

B

B

A

BA A

A

B

A A B

Support

B

A A

B

A B

A

B

A

A

A

B

Deposited film

B A

B

B

B

Deposited film Reactant B FIGURE 2.3 Schematic diagrams of the “one-sided” and “counterdiffusion” deposition geometries.

(a)

Reactants A and B

Support

(b)

Reactant A

Support Inert gas

Reactant B

Furnace Exhaust

Furnace Exhaust

Exhaust

Exhaust

FIGURE 2.4 Apparatus for film deposition by the: (a) “one-sided” deposition geometry and (b) “counterdiffusion” deposition geometry.

Chapter

2

Review of Silica Membranes for Hydrogen Separation

35

deposition, the reactant concentrations at the location of maximum rate are small fractions of their boundary values; these concentrations will be lower the faster the reaction, leading to a difference between the membranes obtained by the one-sided and the countercurrent geometries. In both methods (one-sided and countercurrent), the reactants are introduced simultaneously. A third approach is the so-called alternating reactant deposition, which consists in carrying out the deposition through vapor phase reaction by alternating flows of the reactants with the aim of minimizing homogeneous particle formation and deposition gradients and thus controlling strictly the stoichiometry. This method is slower since after surface saturation by the first reactant, the flow is switched to the second reactant until once more saturation occurs, completing one reaction cycle.

Stability Studies Another important aspect that needs to be considered when working with membranes is their stability under working conditions. Silica-modified membranes developed by many researchers have suffered loss of permeability (as much as 50% or more in the first 12 h) on exposure to moisture [36,45]. This has been attributed to the removal of Si–OH groups leading to the formation of Si–O–Si bonds that close the pore channels [46]. This phenomenon has been termed as “densification.” Moisture apparently catalyzes this reaction particularly at higher temperatures. Densification not only leads to lower permeability but also causes embrittlement of the silica film. Therefore, one of the aims in silica membrane studies has been the development of membranes that are stable at high temperatures in the presence of moisture. Great effort has been made to improve the stability of silica membranes; one approach is to make hydrophobic silica membranes by the incorporation of methyl groups in the silica microstructure [47]. Another approach involves keeping coated membranes in humid air for a few days and calcining them in steam [48]. Ni-doped silica membranes showed good steam tolerance with a H2 permeance of 2.0  10 7 mol m 2s 1Pa 1 and H2/N2 selectivity of 400 after 140 h on stream at 773 K [49]. Membranes prepared by sol–gel methods, composed of silica with other inorganic oxides such as Al2O3, TiO2, and ZrO2, have been reported to improve the stability of silica membranes under humid atmospheres at high temperatures [50–52]. This may be because Si–O–Al, Si–O–Ti, and Si–O–Zr bonds are energetically less susceptible to hydrolysis. Some of these concepts have also been applied to the CVDsilica membranes in order to obtain higher stability. Reaction Conditions Different groups have worked at different CVD conditions and the results have shown that different factors such as the reactive agents, reaction temperature, pressure/evacuation, deposition time have influence on the permselective performance of the formed membranes.

36

Inorganic, Polymeric and Composite Membranes

Hydrogen Permeation Mechanism Much investigation has been done on gas transport through membranes and extensive discussions are found in the literature [53–56]. The gas transport through the membranes has been described through various models that have different theoretical bases related with the details of the gas diffusion mechanism. There are two major factors that affect the overall gas transport in membranes: (i) sorption (solubility selectivity) and (ii) diffusion. While sorption describes the interactions between gas molecules and the membrane surface, diffusion deals with the rate of the gas passage through the membrane. The involvement of these steps has to be analyzed qualitatively and quantitatively in order to understand the overall gas transport mechanism, since each process can contribute to the total permeation rate depending on variables such as temperature, pressure, and composition. Sorption of the gas molecules from the bulk gaseous state to the surface of the membrane can occur physically (physisorption) or chemically (chemisorption) depending on the nature of the force between the gas molecules and the surface. Subsequently, the adsorbed molecules diffuse through the membrane in various ways, depending on the nature of the interaction between the diffusing gas molecules and the membrane, under the driving force of a concentration gradient. Figure 2.5 illustrates three different gas diffusion mechanisms that are often involved in gas transport through silica membranes. Each of these mechanisms will apply depending on the ratio of the molecular size of the diffusing gas and the pore diameter. If the pores are relatively large (0.1–10 mm) compared to the gas molecules, the gases permeate by convective flow and no separation takes place. If the pores are smaller than 0.1 mm, then the pore diameter is the same

(a)

Convective flow

(b) Porous membranes

Knudsen diffusion

(c)

Molecular sieving

(d) Dense membrane

Solution–diffusion

FIGURE 2.5 Mechanisms for permeation of gases through porous (a–c) and dense (d) gas separation membranes: (a) convective flow, (b) Knudsen diffusion, (c) molecular sieving (surface diffusion) and (d) solution–diffusion.

Chapter

2

Review of Silica Membranes for Hydrogen Separation

37

size as or smaller than the mean free path of the gas molecules and diffusion through such pores is governed by Knudsen diffusion, where the transport rate of any gas is inversely proportional to the square root of its molecular weight and temperature. If the membrane pores are extremely small (order of 0.2–5 nm), then the gases are separated by molecular sieving. Transport through this type of membrane is complex and includes both diffusion in the gas phase and diffusion of adsorbed species on the surface of the pores (surface diffusion). As the ratio of the pore size of the membrane versus the size of the diffusing gas molecule decreases, the separation through the dense film occurs by a solution–diffusion mechanism, in which a bound molecule in one sorption site will jump to other sorption sites with a specific vibrational frequency and jump length. The separation selectivity of microporous membranes is governed by Knudsen diffusion; however, this is not sufficient for membrane reactors or other demanding applications. Thus, all the research carried out with the silica membranes via CVD has the objective of modifying the pores of microporous membranes in order to move away from the Knudsen diffusion regime to activated diffusion as in the case of nonporous or ultramicroporous materials that can provide higher selectivities. The literature review, which is presented later, is split into two sections that include the works that have been published on silica membranes prepared on the two different supports that have been employed, Vycor and alumina. The most relevant information from these works is summarized in Tables 2.1 and 2.2.

SILICA MEMBRANES SUPPORTED ON VYCOR GLASS The first reported attempt on preparation of silica membranes by CVD was carried out by Okubo and Inoue [5]. These authors introduced tetraethoxysilane (TEOS) into the pores of a tubular porous glass support (with 2-nm pore size) at 473 K and 1 atm. The small changes in porosity indicated that TEOS decomposed in the pores near the surface of the porous glass tube, giving an asymmetrical membrane structure. This modification improved the selectivity in a gas separation experiment using a helium–oxygen mixture, where the separation factor changed from 3, in the original glass support, to 6 after the modification by CVD. The separation factor change is very small, so most of the pores were not filled, and this accounts for the lack of change in porosity. Gas separation experiments with helium–oxygen mixture were carried out at varying temperature and gas compositions, and the resulting helium permeation rate dependence on temperature proved to be the reverse of that obtained through the original porous glass. Also, the separation factor was improved (beyond Knudsen theoretical value) over a wide range of temperatures and compositions, and the dependence on these parameters also changed; these results led this group to suggest that a different permeation mechanism was operating in the pores (controlled finely by the modification) where deposition had taken place. In a subsequent publication [29], these workers studied the permeation properties of these membranes in more detail, obtaining a permeance

38

Inorganic, Polymeric and Composite Membranes

for H2 of 5.0  10 9 m 2 s 1 Pa 1 at 473 K with H2/N2 ratio of  11. This permeance is lower by 10-fold than the original glass substrate. Also, they observed that the permeability increased with temperature instead of decreasing as T 1/2 as predicted by Knudsen diffusion. The temperature dependence was of an Arrhenius type, indicating that permeation was an activated process, but the activation energy was smaller than that found in quartz or vitreous glass. Thus, it was concluded that the permeation was in a transition regime between gas and surface diffusion control. Shortly after the paper by Okubo and Inoue, the group of Gavalas published a series of systematic studies on the preparation of silica membranes by CVD [6,28]. This group used porous Vycor glass tubes as support, with a pore size of 4 nm, supplied by Corning glass. The tubes were sintered at the two ends to generate nonporous segments in order to connect with fittings to the gas lines. In their first work [6], they reported the deposition of H2-permselective films of SiO2 within the wall of the porous support tubes by the oxidation of SiH4 in what they named the “opposing reactants geometry” or “counterdiffusion geometry”1: SiH4 was passed inside the tube while O2 was passed outside. The two reactants diffused opposite to each other at a total pressure of 1 atm and reacted within a narrow front inside the tube wall, forming a thin SiO2 film. Once the pores were plugged, the reaction stopped since the reactants could not reach each other. At 723 K and 0.1 and 0.33 atm of SiH4 and O2, respectively, the reaction was complete within 15 min, whereas at lower temperatures the reaction was too slow, and at higher temperatures decomposition of silane to silicon occurred, forming an impermeable layer. The membranes prepared at 723 K proved to be selectively permeable to H2 with a permeability H2/N2 ratio as high as 3000 and an H2 permeance of 1.4  10 8 mol m 2 s 1 Pa 1. They identified the diffusion mechanism to be in the transition region between Knudsen and microporous. These membranes were stable upon further heat treatment at 723 K; however, when heated at higher temperature (873 K), they underwent densification that substantially reduced the permeability and the selectivity of the film. The permeation of H2 was believed to occur through an activated diffusion mechanism (solution–diffusion mechanism) and the activation energy was calculated to be 35 kJ mol 1 (this value suggests that the gases were separated by a molecular sieving effect). In this work, they also carried out deposition using the “one-sided geometry”; however, they abandoned this geometry because of the quick success they obtained using the counterdiffusion geometry. In order to overcome the densification produced with the silica films at high temperature and the corresponding loss in permeability and selectivity, this group carried out a series of experiments where they deposited SiO2, TiO2, 1

Some authors have used the terminology “opposing reactant geometry,” whereas others have used “counterdiffusion geometry” to describe this geometry of flows. Since the IUPAC terminology for membranes to describe this flow is “countercurrent flow”, from now in, in this text, this geometry will be referred to as “counterdiffusion geometry.”

Chapter

2

Review of Silica Membranes for Hydrogen Separation

39

Al2O3, and B2O3 films within the pores of Vycor tubes using the corresponding chlorides as precursors. In the case of the silica films, they tested another deposition reaction: the hydrolysis of SiCl4 for SiO2 deposition [30]. This reaction could be carried out at higher temperatures, as high as 1073 K, thus producing denser films with better thermal stability. Hydrogen permselective SiO2 membranes were formed by counterdiffusion deposition and by one-sided deposition, although the latter case presented membranes with thinner deposited layers and higher H2 permeation coefficients and showed a much higher deposition rate. In this work, the best membranes they obtained showed H2 permeances at 723 K of 2.2–0.7  10 8 mol m 2 s 1 Pa 1 for the membranes prepared in the one-sided geometry, and the H2/N2 permeation ratios were 1000–7000. The different results obtained with the different precursors (SiCl4, TiCl4, AlCl3, BCl3) and different geometries (“one-sided” versus “counterdiffusion”) led this group to tentatively explain the results considering that a direct heterogeneous deposition mechanism was responsible for the “counterdiffusion” deposition of all four oxides. This would explain the onesided deposition of SiO2. The order of permeance was: SiO2 ð2:1  108 mol m2 s1 Pa1 Þ > B2 O3 ð7:3  109 Þ > TiO2 ð4:4  109 Þ
> Al2 O3 2:2  109 The corresponding order of H2/N2 selectivity was: B2 O3 ð5000Þ > SiO2 ð4000Þ > TiO2 ð1600Þ > Al2 O3 ð480Þ Clearly, SiO2 has the best overall performance, followed by B2O3, TiO2, and Al2O3. The sharply different rates of the “one-sided” and “opposing reactant” deposition would be due to the different level of reactant concentrations at the location or the maximum deposition rate. In “one-sided” deposition, pore plugging occurs at or near the external surface where the reactant concentrations are near their free stream values. In the counterdiffusion deposition, the reactant concentrations at the location of maximum rate are small fractions of their boundary values; the faster the reaction, the lower these concentrations and, hence, the larger the difference between the one-sided and the opposing reactant geometries. Figure 2.6 illustrates concentration profiles in the “counterdiffusion” deposition. The deposit layer was found to have an asymmetric profile that was well explained by the heterogeneous nature of the reaction, which is initiated at surface silanols (–Si–OH groups). In the one-sided geometry, the situation is more complicated: although the heterogeneous reaction on the Vycor surface is still the dominant process, simultaneous generation of hydrous silicon oxide monomers, clusters and particles in the gas phase may also contribute to the deposit growth. Subsequent works of this group focused on the stability studies and microstructural characterization of silica membranes formed by “one-sided”

40

Inorganic, Polymeric and Composite Membranes

FIGURE 2.6 Schematic diagram of concentration profiles in the “counterdiffusion” deposition (diagram adapted from Ref. [30]).

Tube wall

Support tube

H2O

SiCl4

PH2O PSiCl4

Reaction region

deposition. While studying silica membranes prepared by one-sided deposition of silica through SiCl4 hydrolysis, where depositions were carried out at various temperatures and reactant concentrations, Tsapatsis and Gavalas [31] found that the membranes prepared were mechanically stable only when the deposit was confined inside the pores of the Vycor substrate. However, when deposition extended substantially outside of the porous matrix, the stresses induced by thermal cycling led to formation and propagation of cracks, decreasing the hydrogen permeance to a stable value 5–10 times below its initial value. In order to avoid crack formation, they deduced that low enough concentrations of reactants had to be used and the deposition did not have to be extended beyond the time that is required to achieve pore plugging. The best membrane they obtained at 873 K showed a hydrogen permeance of 3  10 8 mol m 2 s 1 Pa 1 and an H2/N2 permselectivity of 103. Membranes developed at 1023 K (H2 permeance of 2.9  10 8 mol m 2 s 1 Pa 1 and H2/N2 permselectivity ratio of 200) showed the best stability characteristics: after 9 days of hydrothermal treatment, these membranes were densified and exhibited a hydrogen permeance of 0.7  10 8 mol m 2 s 1 Pa 1 at 773 K and H2/N2 permselectivity above 500. Through electron microscopy, they found that the SiO2 deposit density was at its maximum in a region  0.5 mm thick, adjacent to the inner surface and gradually declines to zero within a depth of  10 mm from the surface. They stated that it was the thin region of maximum deposit that was responsible for permselectivity. The 0.5-mm sublayer contained a certain volume fraction (10%) of trapped porosity that became inaccessible at the end of deposition. This trapped porosity conferred to the sublayer a permeability that was much higher than that of the nonporous material, and estimates were made using percolation theory about the fraction of connected voids in the solid.

Chapter

2

Review of Silica Membranes for Hydrogen Separation

41

Kinetic studies had shown that the reaction of chlorosilanes with surface hydroxyls is very fast at high temperatures (above 873 K), and the fast depletion of chlorosilane reactant causes concentration gradients along the tube; so the longer the tube, the more severe is the mass transfer limitation, giving a nonuniform thickness along the length of the tube with the upstream part overdeposited to ensure adequate deposition in the downstream part. The overdeposition caused problems of reduced overall H2 permeance and stability (thermochemical stresses leading to crack formation appear). To solve this problem and to decrease the deposit layer thickness, Kim and Gavalas [57] continued with a study where they deposited SiO2 layers on the surface of porous Vycor glass by the method they named “alternating reactant deposition” that consisted in carrying out the deposition through vapor phase reaction by alternating flows of SiCl4 and H2O. This way, homogeneous particle formation and deposition gradients were minimized providing strict control of the stoichiometry. As commented earlier on, this method is slower since once the surface is saturated by the first reactant, the flow is switched to the second reactant until once more saturation occurs, completing one reaction cycle. Thus, several consecutive reaction cycles were necessary to build up a layer of the desired thickness. To limit the thickness of the deposited layer, one half of the reaction cycle was changed from continuous flow until saturation to a limited dosage of reactant. The membranes prepared by this technique had a H2 permeance of 2.2–2.9  10 8 mol m 2 s 1 Pa 1 and a H2/N2 selectivity of 500–1000 at 873 K. Over prolonged hydrothermal treatment, they observed that the membranes prepared by alternating deposition suffered a smaller decline in permeance (20–30% vs. 70–80%) and had a better H2/N2 selectivity than in the case of one-sided geometry in steady flow. However, numerous repetitions were required until a H2-permselective membrane was formed. These results suggested that the membranes prepared by alternating deposition were thinner and initially denser. Both types of membranes reached approximately equal densities after prolonged hydrothermal treatment. The smaller thickness of the membranes prepared by alternating deposition was mainly due to the elimination of axial reactant concentration gradients and gas-phase particle formation, both of which resulted in thicker deposit layers. Along with the “alternating reactant vapor deposition,” Jiang et al. [58] introduced an additional modification to the CVD (silica on porous Vycor support tubes) procedure in order to decrease the deposit layer thickness and improve the permeability characteristics: they temporarily decreased the diffusion coefficient of the support by introducing and carbonizing a polymer inside the pores of the support. The carbon barriers were introduced into the pores of the support by vapor deposition polymerization of furfuryl alcohol catalyzed by p-toluene sulfonic acid. The presence of these barriers limited the penetration of SiCl4 precursor inside the pore during CVD resulting in a thinner deposit layer. After CVD, the barriers were removed by oxidation. The hydrogen permeance of the deposit layer obtained using the carbon barriers, 5  10 8 mol m 2 s 1 Pa 1

42

Inorganic, Polymeric and Composite Membranes

and H2/N2 selectivity  1000, was two to four times higher than the permeance obtained in the absence of these barriers. From TGA experiments, these authors claimed that this method produced a layer of small carbon fragments ensconced inside the pores of the support, blocking about two-thirds of the surface hydroxyl groups and thus reduced the rate constant of silylation, although at the same time they reduced the SiCl4 diffusivity decreasing thus the depth of SiCl4 penetration. The volume of the carbon particles did not remain constant during CVD because of the continuous loss of carbon by carbonization and steam gasification. So the morphology of the silica deposit layer was the result of a complex coupling between carbonization, steam gasification, and silica deposition. The results in this study were not optimal and improvements could be derived from a better understanding of the carbon layer morphology. TGA experiments suggested that the carbon deposit reduced the reaction rate coefficient as well as the diffusion coefficient. These two changes have opposite effects on the precursor penetration, but the reduction of the diffusion coefficient seems to be the dominant effect in view of the higher permeance obtained in the presence of carbon. Additional studies on formation of silica membranes in Vycor tubes by CVD were carried out by Megiris and Glezer [59]. These workers employed oxidation of triisopropylsilane (TPS) at 1023 K in a counterdiffusion geometry that resulted in the deposition of nanoscale SiO2/C substructures within the mouth of the pores of porous Vycor glass. The reactant had an advantage of safety in contrast to the highly toxic and corrosive nature of SiCl4 and explosive nature of SiH4. They observed that the ratio of the partial pressure of the reactants affected the permselective properties of the membrane. Thus, membranes selective for hydrogen separation were obtained when 20 < r < 100, r ¼ O2/TPS. With increasing temperature, the flux of hydrogen increased indicating increasing contribution of activated diffusion through the film; thus, permeation of hydrogen through the SiO2/C nanoscale deposits was believed to follow an activated solution–diffusion mechanism. The films formed using this precursor showed an H2 permeance ranging 1–2  10 8 mol m 2 s 1 Pa 1 but contained a considerable amount of carbon deposit, and their selectivity for H2 over N2 was only in the range of 30–250, which was much lower than those of the films produced using SiH4 and SiCl4 as precursors. Figure 2.7 shows a schematic FIGURE 2.7 Schematic diagram of the deposition mechanism showing the modified CVD pore closure.

TEOS

TEOS

Chapter

2

Review of Silica Membranes for Hydrogen Separation

43

diagram of the CVD process; according to the suggested deposition mechanism, the selective properties of the membrane would depend on the solid deposits constricting the mouth of the substrate pores. Ha et al. [7] reported the deposition of H2-permselective films of SiO2 within the walls of porous glass tubes using TEOS as precursor at temperatures between 473 and 973 K, paying special attention to optimizing the deposition conditions in order to obtain higher H2 selectivity as compared to the work of Okubo and Inoue, who had used the same precursor. In this work, they employed both the “one-sided” and the “counterdiffusion” geometries. In the one-sided geometry, they observed that in the absence of oxygen, the SiO2 films produced at 973 K showed relatively low H2/N2 selectivities (hydrogen permeance obtained after 30 min of deposition was 1.4  10 8 mol m 2 s 1 Pa 1 and H2/N2 selectivity was 200), most likely due to the formation of microcracks in the films upon additional deposition of SiO2. The stability of these membranes depended on the deposition conditions (TEOS concentration, deposition time, etc.). In the presence of oxygen and in the “one-sided” deposition geometry, hydrogen-selective SiO2 films were successfully formed on the porous tubes at above 573 K and the film became denser with increasing temperature. At 873 K, the membrane exhibited permeances and H2/N2 selectivities of 1.7–3.1  10 8 mol m 2 s 1 Pa 1 and 880, respectively. The presence of oxygen was important in stabilizing the films during deposition. It was observed that unstable films were obtained at high concentration of TEOS by the “counterdiffusion” deposition since TEOS pyrolysis was dominant and oxygen concentration at the deposition region was low (at 973 K, hydrogen permeance was 2.6  10 8 mol m 2 s 1 Pa 1 and H2/N2 selectivity 3.8). When comparing SiO2 films produced by TEOS and O2 with the ones produced by SiCl4 hydrolysis, they observed that the former membranes had higher hydrogen permeation rates but lower selectivities. This could be explained by the analysis of the BET surface area results and gas permeation properties for the films deposited with the different precursors, which indicated that films formed by TEOS and O2 were located nearer to the pore entrance and less dense than those produced by SiCl4 and water. In the following years, Ha and coworkers tried to improve the chemical stability of the SiO2 membranes using porous Vycor glass as support, by studying a binary system of SiO2 with Al2O3 [60] and TiO2 [61]. In the first case, they reported the synthesis and properties of alumina–silica composite membranes formed by a metal-organic CVD method at atmospheric pressure, using aluminum tri-sec-butoxide (ATSB) and TEOS as the alumina and silica sources, respectively. They observed that the deposition rate of the alumina–silica layer was much faster than that of the pure silica layer. The gas permeation properties of the composite membranes were comparable to those of pure silica membranes (hydrogen permeance obtained was 4  10 9 mol m 2 s 1 Pa 1 and H2/N2 selectivity was 12). In the second case, films of TiO2/SiO2 were deposited on the inner surface of porous glass support tubes by decomposition

44

Inorganic, Polymeric and Composite Membranes

of tetraisopropyl titanate (TIPT) and TEOS at atmospheric pressure. Dense and hydrogen-permselective membranes were formed at 673–873 K. The permeation rates of H2 through the membrane at 873 K were 1.4–2.8  10 8 mol m 2 s 1 Pa 1 and H2/N2 permeation ratios were above 1000. It was observed that the addition of TIPT to the TEOS stream significantly accelerated the deposition rate and produced highly H2-selective films. The TiO2/SiO2 membranes formed at 873 K had permeation properties comparable to those of SiO2 membranes produced from TEOS. No stability data were reported. Levy et al. modified porous Vycor tubes with 4-nm pore diameter using lowpressure CVD of SiO2 films for gas separation [8] using diethylsilane (DES) as silicon precursor in conjunction with O2 or N2O as oxidizing agents. The latter was employed with the aim of preparing membranes to separate volatile organic compounds (VOC’s). Based on the idea that SiO2 formation within pores is a self-limited process, the use of the N2O precursor instead of O2 was considered in the counterflow geometry. At the point where the pore diameter approaches the size of the N2O molecule, no further reactions would be expected and film deposition would cease. The selection of N2O, with a diameter less than that of typical VOCs but greater than that of N2, would block the flow of the largersized molecules while still allowing the N2 flow through the membrane structure. The use of DES as precursor has the benefit that it is an environmentally benign substitute for silane, which is considered pyrophoric and highly explosive. Both one-sided and counterdiffusion geometries were investigated and it was observed that higher selectivities and better mechanical stability were obtained for membranes produced using the counterdiffusion geometry. For membranes prepared using both oxidants, H2 permeability coefficient in the order of 1.610 11 mol m 1 s 1 Pa 1 and selectivities in the order of 1000 were obtained for H2 and He over N2, Ar, and C7H8. The use of N2O caused improvements in the pore-narrowing rate and the N2:C7H8 selectivity compared to when O2 was used. Silica membranes deposited on porous glass tubes for methanol vapor separation in the gas phase were also studied by Kuraoka et al. In one of their works [62], they introduced a variation in the conventional CVD process as reported by Gavalas and coworkers; the inside of the membrane module was evacuated with a vacuum pump in order to deposit silica near the pores preferentially. TEOS vapor was fed to the reactor with nitrogen carrier as well as oxygen. While the nitrogen flow was kept constant, other conditions were changed to optimize the CVD conditions, such as O2 flow rate, TEOS concentrations, and reaction temperatures (673–773 K). The ratio of permeances, He/N2 of the membranes were studied for the different conditions, and it was found that this ratio was higher when a higher O2 flow rate was used (2000 cm3 m 1, compared to 1000 cm3 m 1). At this higher O2 flow, three different TEOS concentrations, 0.8%, 1.6%, and 2.3% were tested, and the highest He/N2 permeance ratio was obtained with the intermediate concentration (1.6%).

Chapter

2

Review of Silica Membranes for Hydrogen Separation

45

As for the reaction temperatures studied, in the interval of 673–773 K, the highest ratio was obtained at 723 K. The membranes thus prepared minimized permeation of methanol vapor and displayed molecular sieving performance. The separation factors of He/CH3OH were 7.1 at 373 K and 16.9 at 473 K, higher than Knudsen value (He/ CH3OH ¼ 2.8). In a subsequent work [32], this same group used the same type of membranes to measure gas separation using H2–10% CH3OH and CO–10% CH3OH gas mixtures. They used two different porous glass supports with different pore diameters, 4 and 2 nm, and their best results were obtained with the latter, in which, at 473 K, the H2/CH3OH separation factor was around 20, about five times larger than the theoretical Knudsen value. This was thought to be due to the small amounts of defects on this support. The smaller mean pore diameter of the 2-nm support resulted in a shorter CVD time than in the case of the 4-nm support; thus, the growth of silica became small and the silica deposition layer also became thin, lowering the possibility of generating defects. At 473 K, the H2 permeance for the membrane obtained with this support was 4.4  10 9 mol m 2 s 1 Pa 1 and showed an H2/N2 selectivity of 100. CVD in the counterdiffusion geometry was employed by Nakao et al. [9] to prepare silica membranes on porous Vycor glass using TEOS/O3 as reactants (O2 as carrier of O3); they studied the influence of reaction parameters such as the TEOS and O3 concentrations in the feed, the temperature and period of the CVD reaction, on the gas transport properties of the membranes prepared. To study the influence of reaction temperature, they carried out experiments at 150, 200, and 250  C and measured the permeance ratios of He/N2, He/O2, He/Ar, He/CO2. They observed that at 150  C, the temperature was not sufficient to modify the pore structure of Vycor glass since the permselectivity values hardly changed, even after 12 h of CVD. At 200 and 250  C, modification of pore size occurred within a few hours (3 and 1.5 h, respectively) and permselectivities increased, showing higher values for reaction at 250  C. They claimed that further experiments (not presented in the work) revealed that the maximum selectivity values that could be obtained under the present experimental conditions did not alter much with the variation in temperature from 200 to 250  C, and this led them to state that probably only kinetics is influenced by temperature change. Controlling the reaction parameters (temperature, flow rate) and other conditions (support treatments) enabled them to design defect-free microporous membranes with interesting gas transport properties. These membranes showed activated diffusion of N2, Ar, and O2 molecules and were capable of separating a variety of gas molecular combinations. This indicated the presence of a defectfree microporous membrane structure. Permselectivity values of 950, 26, and 14.3 were measured for He/N2, Ne/Ar, and O2/N2 gas molecular combinations and an He permeance of 3  10 8 mol m 2 s 1 Pa 1 at 313 K. Oyama and coworkers published a series of works related to the preparation of a modified Vycor glass membrane by CVD which they called “Nanosil.”

46

Inorganic, Polymeric and Composite Membranes

The difference in the new method they employed was that it involved the CVD of TEOS at 873 K in inert gas at atmospheric pressure, whereas previous work by other groups had employed O2 [5,6,8,32,60], O3 [9], or H2O [30,31,57,58] as reactive agents in the CVD process and generally used lower temperatures. In their first work [19], this group reported that the hydrogen selectivities obtained by the membranes produced by this new method were practically 100% with respect to other gases: CH4, CO, CO2, H2O, while there was practically no loss of permeability compared to the porous Vycor precursor. The membrane also showed high stability under hydrothermal conditions over prolonged times. They suggested that the absence of oxygen or water and the use of high temperatures during the CVD could probably give rise to silica of a different structure from previous work. The exact difference was uncertain but could be associated with a less crystalline structure with no pinhole defects and a reduction in the number of free hydroxyl groups. The apparent activation energy calculated from the permeability data was 2 kJ mol 1, unusually low for a surface diffusion mechanism. In later work [63], they further studied the permeation properties of these membranes toward the small gas molecules H2, He, Ne, CH4, CO, and CO2 and found that the permeation mechanism changed from a Knudsen-type mechanism for the Vycor glass to an activated diffusion mechanism for the Nanosil membrane. They also studied the surface topology of the Vycor and the Nanosil membranes using atomic force microscopy (AFM), applied for the first time to this kind of system, in order to obtain better insight into the nature of the Nanosil membrane. The images obtained showed that the Vycor substrate was made up of rectangular plate-like elements of size 90  30 nm and the edges between the plates likely constituted the pore entrances. This idea was supported by the porosimetry results, which indicated that the pores were slit- or rectangular-shaped and had 4-nm sizes. The deposited silica formed a thin layer on top of these plates since the surface structure after deposition consisted of globular, elongated particles that were more rounded and larger than the plates on the Vycor substrate (before CVD). This silica layer erased the fine structure and increased the average feature size to 110  50 nm. In other words, in this case, the silica layer was deposited outside the pores, instead of inside the pores as in other previous works by other researchers. Figure 2.8 shows a schematic diagram of the structures. In their next work, they carried out a deeper study of the gas permeation characteristics (a)

(b)

FIGURE 2.8 Schematic diagram of the (a) Vycor membrane and (b) Nanosil membrane.

Chapter

2

Review of Silica Membranes for Hydrogen Separation

47

of the gas molecules, He, Ne, H2, CH4, CO, and CO2, on the Vycor and the Nanosil membranes [10], and they applied model analysis to the experimental results for gas transport. They reported that after 12 h of silica deposition on the Vycor support, the membrane showed at least a 104 order of hydrogen selectivity over the larger gas molecules, CH4, CO, and CO2, while the permeance of hydrogen on the Nanosil membrane remained very high and comparable to the permeance on the fresh Vycor support (order of 10 8 mol m 2 s 1 Pa 1) at 873 K. The pore size distribution curves obtained from N2 desorption isotherms using the Barrett, Joyner, and Halenda (BJH) method for the fresh Vycor and Nanosil membranes showed little change in the pore size distribution by the deposition of the silica layer indicating that it was deposited on the outer surface of the Vycor glass without significant penetration into the pore structure. The gas permeance was found to be activated and increased with temperature; however, this permeance was limited at high temperatures because of the limited permeance on the Vycor support. The gas transport characteristics on the Vycor and Nanosil membranes were discussed with relevant diffusion models. The gas transport on the Vycor support was explained satisfactorily with a surface diffusion enhanced Knudsen diffusion model. Gas transport on the deposited silica layer was more accurately described by a statistical solid-state diffusion model that was able to account for the unusual permeation order He > H2 > Ne, which did not follow either mass or atomic size. The values for solubility site densities and activation energies calculated by these models matched experimental results. The theory of permeation that described the transport and physical properties of the silica-based membranes of the Nanosil type were further developed and extensively explained in another publication by Oyama and coworkers [64]. The theoretical model was based on the fact that in the Nanosil membranes the permeation of CO, CO2, and CH4 was inhibited relatively to that of smaller-sized species such as H2 and the lighter noble species, and that these membranes showed an unusual order of gas permeation, He > H2 > D2 > Ne, independent on size or mass of the diffusing species. This model was based on jumps of the permeating species between solubility sites in a solid matrix, assuming equilibrium sorption in the sites, random motion, and a transition state with two degrees of vibrational freedom and one degree of translational freedom. Agreement between the theoretical values and the experimentally obtained points indicated that the model equations were effective in describing the unusual behavior of hydrogen and the light noble gases. The density of the solubility sites and the jump distance were two important parameters in this analysis. The number of solubility sites was found to be smaller by about one order of magnitude than in vitreous glasses, while the vibrational frequencies remained unchanged. The activation energies for permeation through the silica membrane were also found to be smaller than the ones through glass and therefore indicated the presence of a different structure. These results led them to suggest that the Nanosil membrane was probably formed of larger Si–O rings that would result in more open and less dense structures.

48

Inorganic, Polymeric and Composite Membranes

Nitodas et al. carried out a study [33] where they employed the CVD method in a countercurrent configuration to prepare composite silica membranes and compare with two different kinds of support, porous Vycor tubes and alumina nanofiltration tubes. TEOS and O3 were used as deposition precursors. Most of the results shown focused on the silica layers deposited on the alumina substrate. They employed mercury porosimetry to measure the pore size of the substrates after CVD, and the results revealed that an important reduction of the initial size of the substrate pores, down to 30%, was achieved after 12.5 h of deposition in the case of the Vycor glass support. They claimed that their preliminary measurements of H2 permeance on Vycor/SiO2 membranes (not reported in the paper) led to the conclusion that the permeance of hydrogen in these membranes was smaller than that on the Al2O3/SiO2 membranes. For microporous supports like Vycor glass, the separation selectivity is limited by Knudsen diffusion (order of < 10), and this is not sufficient for membrane reactors or other demanding applications. For this reason, a substantial portion of the research carried out with silica membranes via CVD has the objective of modifying the pores of microporous membranes in order to move away from the Knudsen diffusion regime to activated diffusion. This is the reason much work shifted to alumina supports of higher porosity in order to achieve higher selectivities.

SILICA MEMBRANES SUPPORTED ON ALUMINA All the membranes presented in the previous section were limited by the small pore size (2–4 nm) of the Vycor glass support. Improvement in the permeability could be obtained by preparing membranes using alumina supports with larger pores that present much lower permeance resistance. Other advantages of these porous supports made by sintering a-alumina particles are that they are more economical than Vycor glass tubes, are much stronger, so have a higher resistance to high pressures. However, studies showed that their permselectivity was controlled by the Knudsen diffusion mechanism especially after modification by ultrafine g-alumina particles prepared by sol–gel technique [65–67]. Morooka et al. were the first in preparing alumina-supported membranes modified by a CVD process that gave high permeability and high selectivity. They published a series of studies where they prepared H2-permselective membranes by thermal cracking of TEOS under various conditions, using a-alumina tubes (NOK Corp. with average pore size  150 nm) as support [11,34]. Using the one-sided geometry (typical experimental set-up for this kind of CVD is shown in Figure 2.9), they modified the pores of the alumina supports and they compared the results obtained with evacuation and without evacuation of the reactants through the porous supports. They studied the effect of depositing a g-alumina layer (pore size: 7 nm) as an intermediate layer on the a-alumina support [34], before the silica layer was deposited. Their results showed that the permselectivity improved when evacuation was carried out

Chapter

2

49

Review of Silica Membranes for Hydrogen Separation

since the SiO2 infiltrated into the porous support and fewer pinholes were observed in the SiO2 film. When the CVD process was carried out at 873 K in the presence of the intermediate g-alumina layer, the membranes showed a permeability of 10 7–10 8 mol m 2 s 1 Pa 1 and a H2/N2 selectivity higher than 1000. When the silica deposition by CVD of TEOS was carried out on the same as-received a-alumina tubes, without the intermediate g-alumina coating at 873–923 K, the H2 permeance of the modified membrane at 873 K was of the order of 10 8 mol m 2 s 1 Pa 1, its value depending on the decomposition temperature and other CVD conditions such as the evacuation pressure, while the N2 permeance was below 10 11 mol m 2 s 1 Pa 1. Therefore, this method of membrane preparation proved to be effective to plug micropores of a sintered support that possessed a wide pore-size distribution. Figure 2.10 shows a sche-

Vent Furnace

Pirani gauge P

Temperature controller Vent Rotary vaccum pump Membrane

Mass flow controller Flow

MFC

Heating tape TEOS

Carrier gas FIGURE 2.9 Experimental apparatus for membrane modification by CVD with evacuation.

FIGURE 2.10 Schematic diagram showing the pore plugging model by CVD with forced evacuation.

After evacuation

50

Inorganic, Polymeric and Composite Membranes

matic diagram of the pore plugging with evacuation. The membranes prepared by this method had thermal stability and could endure cyclic temperature changes. Membranes prepared by the same method (CVD of TEOS with evacuation) at 873–973 K in a porous a-alumina tube (with pore size of 110–180 nm), with and without an intermediate g-alumina (with pore size of 6–8 nm) were prepared by Sea et al. [35,36] and were studied for hydrogen separation from a H2–H2O–HBr mixture produced by the thermochemical H2O decomposition process. They reported that the hydrogen permeance of the silica membranes formed in presence of the g-alumina intermediate film was higher (3  10 7 mol m 2 s 1 Pa 1) compared with the membrane formed in the pores of the a-alumina tube (10 8 mol m 2 s 1 Pa 1), and exhibited a H2/N2 selectivity > 100. They found that the final evacuation pressure played an important role in the quality of the membranes prepared and in the case of H2/H2O separation, found that when the pressure was 20 Pa and the temperature was 923 K, the water permeance was greatly reduced while the hydrogen permeance was tripled in the membrane with the g-alumina intermediate layer. In comparison with the membrane formed in the pores of the a-alumina tube, exhibited a H2/H2O selectivity of 7.6 that remained unchanged for 2 days in a mixture of H2–H2O–HBr at 623 K. The microporous silica membrane formed on the macroporous a-alumina tube and heat-treated at 1173 K in inert atmosphere showed a hydrogen permeance of 3–4  10 7 mol m 2 s 1 Pa 1 and a H2/H2O selectivity of 7–15 at permeation temperatures of 473–673 K. On the other hand, the microporous silica membrane that was formed on a mesoporous g-alumina intermediate layer was more water selective and showed a hydrogen permeance of the order of 10 6 mol m 2 s 1 Pa 1 and a H2O/H2 selectivity of 12–25 in the 473–673 K interval. This membrane rejected HBr at a H2O/HBr selectivity of 300–1000. Another aspect that Sea and coworkers investigated with these membranes was the influence of the CVD precursor molecule on the pore size of the amorphous silica layer. They prepared silica-based membranes with enlarged micropore size by using phenyl-substituted ethoxysilanes (phenyltriethoxysilane (PTES) and diphenyldiethoxysilane (DPDES)) as the Si source for the CVD process, the carbonaceous matters being removed by subsequent calcination [12]. The membrane formed with TEOS showed a H2 permeance of the order of 10 7 mol m 2 s 1 Pa 1 and H2/N2 selectivity of 100 at permeation temperatures of 773–873 K. The membrane prepared with DPDES showed an H2 permeance of 7  10 7 mol m 2 s 1 Pa 1 at this temperature range. The PTES membranes showed properties intermediate between the membranes prepared from TEOS and DPDES. So, they concluded that the use of PTES and DPDES was effective in controlling micropore size and separating large size molecules. Sea and coworkers continued studying more exhaustively the amorphous silica membranes deposited by thermal decomposition of TEOS at

Chapter

2

Review of Silica Membranes for Hydrogen Separation

51

873–923 K with evacuation on alumina supports [68–70] obtaining results that corroborated their previous findings, that is, using a forced cross-flow through the porous wall of the support (with evacuation) was effective in plugging the pores; the membranes with the intermediate g-alumina layer showed higher H2 permeance than the SiO2 membranes formed directly on the a-alumina tube. The permeation tests with CO2, N2, CH4, H2O, C3H8, and i-C4H10 showed that a small number of mesopores remained unplugged by the CVD; the permeation of hydrogen could be explained by activated diffusion and that of the other gases by Knudsen diffusion through the unplugged pores, so the total permeance comprised permeances due to the activated and Knudsen diffusion mechanisms. Long-term stability tests showed that the CVD-modified silica membrane was stable in an H2–N2 mixture of 723 K for 100 h. Apart from the extensive work published by these previous groups, other limited studies on modification of alumina supports by CVD of silica were presented by some other groups. Liu and coworkers [45,71] used the CVD of TEOS with O2 in the one-sided geometry to modify the g-alumina top layers (pore size of 4 nm) of asymmetric alumina tubes (supplied by US-Filter), obtaining porous membranes with pore size 0.6–1.5 nm or denser membranes (0.3–0.5 nm) that exhibited high permeance, while maintaining a moderate H2/N2 selectivity. For their best membranes, H2/N2 ratio ranged from 28–36 and H2 permeance was 1.5  10 7–7.3  10 8 mol m 2 s 1 Pa 1 at 573–873 K. Microporous silica membranes supported on mesoporous g-alumina membranes were prepared by Nijmeijer et al. [72] employing low-temperature chemical vapor infiltration (CVI). They used decomposition of silicon tetraacetate (SiAc4) at 548 K in the presence of O2 using the counterdiffusion geometry and compared it with the stateof-the-art microporous CVI membranes produced using silane (SiH4) precursor. They observed that the best membranes were obtained with the SiAc4 precursor showing a hydrogen permeance of 4  10 7 mol m 2 s 1 Pa 1 and H2/N2 selectivity of 43 at 523 K. The hydrogen separation characteristics of silica membranes prepared by CVD in a H2–H2O–HI gaseous mixture were evaluated by Hwang et al. with the aim of applying them to hydrogen iodide decomposition in the thermochemical iodine–sulfur (IS) process [37,38,73]. They employed TEOS decomposition with evacuation to modify g-alumina or a-alumina top layers of support tubes supplied by Noritake, obtaining higher hydrogen/nitrogen selectivities when the g-alumina intermediate layer was present. In their first work [37], they obtained a hydrogen permeance of 6  10 9 mol m 2 s 1 Pa 1 at 873 K; the H2/N2 selectivities at 873 K were 5.2 and 160 for the membranes with supports with a-alumina (with pore size of 100 nm)and g-alumina (with pore size of 10 nm) top layers. In their next publication [38], they prepared a series of similar membranes (support with a g-alumina top layer) controlling the CVD treatment using He–N2 selectivity as the indicator of the pore closure and using different tube lengths and deposition times. The membrane with the highest

52

Inorganic, Polymeric and Composite Membranes

H2 permeance was in the order of 10 7 mol m 2 s 1 Pa 1, and it also showed the highest separation factor of H2–HI (650 at 723 K). Next, this group focused on the study of the stability of these silica membrane for long exposure times in the HI–H2O gaseous mixture at 723 K [73] by carrying out the CVD treatment in two different ways: in one case, it was performed continuously until the membrane showed a certain He/N2 selectivity. In the other case, the CVD was carried out for a specific time, and then the membrane was cooled to room temperature and the treatment was repeated until the membrane showed the same He/N2 selectivity as that by the first method. The stability tests in HI–H2O gaseous mixture showed that the first, continuous method provided better membranes. Also, the membrane using a-alumina as support showed high stability in the HI–H2O gaseous mixture and had high H2/HI selectivity (240–2600 at 573–873 K) in the H2–H2O–HI gaseous mixture after the stability test. The group of Oyama extended the same CVD process they had employed in their previous work on “Nanosil” membranes on Vycor glass, to the preparation of silica membranes on porous alumina supports, that is, they obtained a highly hydrogen-permselective composite silica membrane by depositing a thin silica layer on a porous alumina support by CVD of TEOS at 873 K in inert gas at atmospheric pressure [17]. The alumina support had a graded multilayer structure with large macropores in the a-alumina layer to small mesopores on the top g-alumina layer. The resulting composite membrane exhibited a hydrogen permeance of the order of 10 7 mol m 2 s 1 Pa 1 at 873 K, and the H2 selectivities over CH4, CO, and CO2 were above 1000 at 873 K. While the experimental gas permeance data for the alumina support indicated Knudsen transport, in the case of the silica film the results showed that H2 and He permeance through the silica film was activated, showing good fits to the Arrhenius expression and giving activation energies of 15 and 9.8 kJ mol 1 for H2 and He, respectively. The mechanism of molecular differentiation by the silica layer was found to be through size selectivity. The mechanism of small gas permeation through the silica membrane was described by a permeation mechanism that involved the jumping of the diffusing molecules between adjacent solubility sites. The lower values in the number of solubility sites and the low activation energies for the silica layer compared to those for vitreous silica glasses indicated that the interstitial structure of the silica obtained from CVD was less dense, and this provided higher diffusivity. Later on, Oyama’s group prepared another series of membranes where special attention was paid to the intermediate g-alumina layer that was deposited on the a-alumina support before CVD [39]. A thin and defect-free g-alumina graded substrate with a three-layer structure was formed on a macroporous alumina support by sequential dipping and calcining with boehmite sols of decreasing particle sizes. A novel aspect in this work was that the size of the sol particles was tuned by controlling synthesis parameters such as acid type, concentration, and hydrolysis time, allowing control of the graded structure of the g-alumina intermediate multilayer. The silica layer was deposited on

Chapter

2

53

Review of Silica Membranes for Hydrogen Separation

the graded g-alumina structure by CVD of TEOS in inert gas. Cross-sectional images of the membranes observed in the SEM microphotographs showed that the intermediate alumina layers were formed from particles of progressively smaller size, forming a smooth interface between the rough support and the topmost amorphous silica layer. The SEM images showed that the use of dilute dipping solutions provided thin layers, four to six times thinner than conventional intermediate g-alumina layers, and the silica layer formed was ultrathin, with a thickness of 20–30 nm. The resulting membrane showed a H2 permeance of 5  10 7 mol m 2 s 1 Pa 1 and hydrogen selectivities over CH4, CO, and CO2 above 1500 at 873 K. Figure 2.11 shows the SEM images and a schematic diagram of the layered substrate. Nakao and coworkers carried out an extensive study of silica membranes supported on alumina prepared by counterdiffusion CVD with the aim to apply them in steam-reforming reactions, putting therefore special emphasis on studying the stability of the membranes in steam environments. In their first work of this series [40], they studied counterdiffusion CVD of TEOS/O3 and tetramethylorthosilicate (TMOS)/O3 on a-alumina porous tubes (NGK Co. Ltd.) coated with an intermediate g-alumina layer by the sol–gel method. The TMOS/O3 reacted membrane showed 10 times higher helium permeance (3.5  10 7 mol m 2 s 1 Pa 1 at 423 K) than the TEOS/O3 reacted membrane (3.6  10 8 mol m 2 s 1 Pa 1 at 423 K). This permeance was also high compared to previously published results. It also showed a higher He/N2 selectivity, 840 at 423 K, whereas in the case of the TEOS/O3 membrane it was 600. In another study, this group carried out another study where the deposition

SiO2 layer (L=20 nm)

High resolution

g-Al2O3 layer 100 nm

Silica layer Graded substrate

1000 nm

Support

FIGURE 2.11 Three-Layer Composite Membrane (a) cross-sectional SEM micrograph (b) schematic diagram, (c) cross-sectional SEM micrograph in higher resolution.

54

Inorganic, Polymeric and Composite Membranes

conditions of the counterdiffusion CVD method were investigated to prepare stable silica membranes at higher temperatures under steam conditions [22]. In this work, they achieved silica membranes on porous g-alumina substrates with a high H2/N2 permeance ratio of ca. 1000 by the counterdiffusion CVD method using TMOS/O2 at 873 K. The permeance at this temperature was 1.5  10 7 mol m 2 s 1 Pa 1. This membrane proved to be stable at 76 kPa of steam vapor at 773 K for 21 h without any reduction in H2/N2 permeance ratio. These conditions are one of the target conditions for the H2 permselective membrane reactors for H2 production from steam reforming of methane. The steam stability tests and steam selectivities of these silica membranes were further studied by this group [74], and showed that the silica membrane deposited at 873 K kept a H2/N2 permeance ratio for 82 h under 76 kPa of steam at 773 K. The H2/H2O permeance ratio through the silica membrane was ca. 300 at 773 K (much higher than through silica membrane prepared by sol–gel method). The effects of steam pretreatment of the g-alumina layer on CVD deposition were studied, and the results showed that this caused an increase in pore size and a decrease of the number of pores of the g-alumina layers, both of which caused the hydrogen permeance to fall after the CVD deposition; therefore, steam pretreatment was discarded. All in all, the results led to the conclusion that these silica membranes were suitable for application to a steam-reforming reaction both for high steam stability and for a high H2/H2O permeance ratio. In order to simultaneously improve membrane gas permeance and separation performance of these membranes, Gopalakrishnan et al. developed what they named the “new hybrid processing method,” which involved both sol–gel and CVD procedures [18] for the deposition of the silica layer. A thin g-alumina layer was applied to the capillary substrates (NOK, Japan) and silica layer was coated over the g-alumina layer also by the sol–gel processing. Then, these sol–gel coated substrates were subjected to CVD using TMOS and O2 as reactants at 873 K. The initial N2 permeance values through the membrane were brought down by four orders of magnitude to 2.7  10 10 mol m 2s 1Pa 1 within only 5 min of CVD, while maintaining the H2 permeance values as high as 6.4  10 7 mol m 2 s 1 Pa 1 (H2/N2 ¼ 2300). The H2/N2 separation values almost doubled to 5000 after 2 h of CVD reaction whereas the permeance dropped to 3.3  10 7 mol m 2 s 1 Pa 1. The permeance and the separation results coupled with XPS analysis led these workers to conclude that a reduction in the vapor phase deposition zone area (thickness) by the preexisting silica sol–gel layer effectively contributed to the quick membrane formation. Gopalakrishnan et al. also investigated the performance of high flux CVDsilica membranes for the separation of gas mixtures containing H2 and CO2 at various temperatures [41]. The membranes were prepared by counterdiffusion CVD with TEOS and O2 as reactants and g-alumina coated a-alumina tubes supplied by Noritake Co. Ltd., Japan, were used as substrates. Single gas performance through the membrane showed activated transport for the smaller

Chapter

2

Review of Silica Membranes for Hydrogen Separation

55

kinetic diameter gases (H2 and He) as permeance increased with temperature. On the other hand, the larger kinetic diameter gases, CO2 and N2, resulted in negative activation energy permeance, which they attributed to the membrane’s precise pore-size control and narrow pore-size distribution with an average pore size below the kinetic diameter of CO2 of 0.33 nm. The single gas permeation of H2 increased from 5.1  10 7 to 7.0  10 7 mol m 2 s 1Pa 1 in the temperature range 373–673 K, while H2/CO2 and H2/N2 values reached 36 and 57 at 673 K, respectively. A variation to the counterdiffusion method was introduced by Araki et al. [42] who prepared high-performance cylindrical silica membranes on a-alumina supports (Noritake, Japan) with a g-alumina interlayer deposited on the membrane side by the sol–gel method. They applied what they called the “extended counterdiffusion CVD method,” by which silica films were deposited under highly pressurized conditions applied to the membrane side where the precursor (in this case TMOS) was supplied. The oxygen used to promote the reaction was fed on the support side. The dense silica membrane prepared by this method showed a hydrogen permeance of 5.0  10 8 mol m 2 s 1 Pa 1 at 573 K, and a H2/CH4 selectivity above 24,000 and a H2/CO2 selectivity above 1200. They observed that pressurizing the membrane side of the reactor did not affect the hydrogen permeance, but CH4 and CO2 permeances were greatly reduced. Pressurizing reduced the number of defects and pinholes formed in the dense silica membranes and promoted the reaction between TMOS and O2, so the silica membrane formed was denser and more heat resistant when compared to the silica membranes prepared without pressurization. Nagano et al. prepared an amorphous silica membrane on g-alumina coated a-alumina support (NOK Co., Japan) by counterdiffusion CVD using TMOS and O2 at 873 K [43]. This membrane showed a high H2 permeance of 3.0  10 7 mol m 2 s 1 Pa 1 at temperatures equal to and above 798 K. Sometime after their studies with counterdiffusion CVD and TMOS precursor, Nakao and coworkers extended their studies to preparing silica membranes using tetramethoxysilane (TMOS), phenyltrimethoxysilane (PTMS), and dimethoxydiphenylsilane (DMDPS) as silicon sources [44], by two kinds of CVD methods with oxygen at 873 K: the counterdiffusion CVD method with TMOS precursor and the one-sided diffusion CVD method with PTMS and DMDPS precursors. These three organosilanes, with different numbers of phenyl groups, were used based on the concept that the pore size of the silica membrane was enlarged when the side-chain groups in the silicon source were large atomic groups. A porous a-alumina capillary with 100-nm pores was used as substrate (NOK Co., Japan) and a g-alumina intermediate layer was applied to reduce pore size. This group confirmed that changing the number of phenyl groups on the silicon source could allow controlling the pore size (the higher the number of phenyl groups, the larger the pore sizes). The DMDPS-derived membranes showed the highest hydrogen permeance at 573 K, of the order of 10 6 mol m 2 s 1 Pa 1, and hydrogen/sulfur hexafluoride permselectivity

56

Inorganic, Polymeric and Composite Membranes

over 6800 and nitrogen/sulfur hexafluoride selectivity of 45. The results from the permeation tests of single gases indicated that the PTMS-derived membrane had larger pores than the TMOS-derived membrane and that DMDPS-derived membrane had the largest pores of all. The order of permeances also increased in the same order, DMDPS > PTMS > TMOS, so the conclusion was reached that the DMDPS-derived membranes had much larger pores, looser structures, and wider pore distribution than the other two precursors. FE-SEM measurements and XPS results showed that the DMDPS-derived membrane had a thinner silica layer than the TMOS-derived membrane. The DMDPSderived membrane was stable under 3.4 kPa of steam vapor at 573 K for 266 h without any reduction in both the hydrogen permeance and the hydrogen/sulfur hexafluoride selectivity.

CONCLUSIONS Membranes for hydrogen separation based on silica have been studied extensively since their discovery in the late 1980s. Considerable progress has been made in their performance. The first membranes were supported on Vycor glass and had a permeance in the range of 10 9 mol m 2 s 1 Pa 1 and selectivity of H2/N2 of the order of 10. In the 1990s, a better support, porous alumina, was found and permeance increased 10-fold to the order of 10 8 mol m 2 s 1 Pa 1 and selectivity for H2/N2 in the 100s. In the new millennium, further improvements in the membranes were made with the use of graded intermediate layers and inert gas CVD and the permeance increased another 10-fold to the level of 10 7 mol m 2 s 1 Pa 1 and selectivity for H2 over other gases in the 1000s. This is now in the range of commercial interest. Current challenges include stability, cost, and ability to produce modules with high surface area to volume ratio.

ACKNOWLEDGMENTS For support of this work, the authors acknowledge the Director, National Science Foundation, Division of Chemical, Bioengineering, Environmental, and Transport Systems (CBET) under grant CBET-084316, the National Energy Technology Laboratory under the NETL-RUA program grant 5.681.884.001, the Kakenhi grant-in-aid Kiban kenkyu B 22-360,335 from the Mombukagakusho, and Petrobras under grant 0050.0055441.09.9 administered by Funcate.

REFERENCES [1] R.G. Lemus, J.M. Martı´nez Duart, Updated hydrogen production costs and parities for conventional and renewable technologies, Int. J. Hydrogen Energy 35 (2010) 3929–3936. [2] J.R. Bartels, M.B. Pate, N.K. Olson, An economic survey of hydrogen production from conventional and alternative energy sources, Int. J. Hydrogen Energy 35 (2010) 8371–8384. [3] P. Baptista, M. Toma´s, C. Silva, Plug-in hybrid fuel cell vehicles market penetration scenarios Int, J. Hydrogen Energy 35 (2010) 10024–10030.

Chapter

2

Review of Silica Membranes for Hydrogen Separation

57

[4] C.H. Lau, B.T. Low, L. Shao, T.-S. Chung, A vapor-phase surface modification method to enhance different types of hollow fiber membranes for industrial scale hydrogen separation, Int. J. Hydrogen Energy 35 (2010) 8970–8982. [5] T. Okubo, H. Inoue, Introduction of specific gas reactivity to porous glass membranes by treatment with tetraethoxysilane, J. Membr. Sci. 42 (1989) 109. [6] G.R. Gavalas, C.E. Megiris, S.W. Nam, Deposition of H2-permselective SiO2 films, Chem. Eng. Sci. 44 (1989) 1829. [7] H.Y. Ha, S.W. Nam, S.A. Hong, W.K. Lee, Chemical vapor deposition of hydrogen-permselective silica films on porous glass supports from tetraethylorthosilicate, J. Membr. Sci. 85 (1993) 279. [8] R.A. Levy, E.S. Ramos, L.N. Krasnoperov, A. Datta, J.M. Grow, Microporous SiO2/Vycor membranes for gas separation, J. Mater. Res. 11 (1996) 3164. [9] S.I. Nakao, T. Suzuki, T. Sugawara, T. Tsuru, S. Kimura, Preparation of microporous membranes by TEOS/O3 CVD in the opposing reactants geometry, Microp. Mesop. Mater. 37 (2000) 145. [10] D. Lee, S.T. Oyama, Gas permeation characteristics of a hydrogen selective supported silica membrane, J. Membr. Sci. 210 (2002) 291. [11] S. Morooka, S. Yan, K. Kusakabe, Y. Akiyama, Formation of hydrogen-permselective SiO2 membrane in macropores of a-alumina support tube by thermal decomposition of TEOS, J. Membr. Sci. 101 (1995) 89. [12] B.K. Sea, K. Kusakabe, S. Morooka, Pore size control and gas permeation kinetics of silica membranes by pyrolysis of phenyl-substituted ethoxysilanes with cross-flow through a porous support wall, J. Membr. Sci. 130 (1997) 41. [13] R.J.R. Ulhorn, K. Keizer, A.J. Burggraaf, Gas transport and separation with ceramic membranes. Part II: synthesis and separation properties of microporous membranes, J. Membr. Sci. 66 (2/3) (1992) 415. [14] R.S.A. de Lange, K.-N.P. Kumar, J.H.A. Hekkink, G.M.H. Van de Velde, K. Keizer, A.J. Burggraaf, et al., Microporous SiO2 and SiO2/MOx (M¼T, Zr, Al) for ceramic membrane applications: a microstructural study of the sol-stage and the consolidated state, J. Sol Gel Sci. Technol. 2 (1994) 489. [15] B.N. Nair, T. Yamaguchi, T. Okubo, H. Suematsu, K. Keizer, S.I. Nakao, Sol-gel synthesis of molecular sieving silica membranes, J. Membr. Sci. 135 (1997) 237. [16] R.M. de Vos, H. Verweij, High-selectivity, high-flux silica membranes for gas separation, Science 279 (1998) 1710. [17] D. Lee, L. Zhang, S.T. Oyama, S. Niu, R.F. Saraf, Synthesis, characterization, and gas permeation properties of a hydrogen permeable silica membrane supported on porous alumina, J. Membr. Sci. 231 (2004) 117. [18] S. Gopalakrishnan, Y. Yoshino, M. Nomura, B.N. Nair, S.I. Nakao, A hybrid processing method for high performance hydrogen-selective silica membranes, J. Membr. Sci. 297 (2007) 5. [19] A.K. Prabhu, S.T. Oyama, Highly hydrogen selective ceramic membranes: application to the transformation of greenhouse gases, J. Membr. Sci. 176 (2000) 233. [20] A.K. Prabhu, A. Liu, L.G. Lovell, S.T. Oyama, Modeling of the methane reforming reaction in hydrogen selective membrane reactors, J. Membr. Sci. 177 (2000) 83. [21] D. Lee, P. Hacarlioglu, S.T. Oyama, The effect of pressure in membrane reactors: trade-off in permeability and equilibrium conversion in the catalytic reforming of CH4 with CO2 at high pressure, Top. Catal. 29 (2004) 45.

58

Inorganic, Polymeric and Composite Membranes

[22] M. Nomura, K. Ono, S. Gopalakrishnan, T. Sugawara, S.I. Nakao, Preparation of a stable silica membrane by a counter diffusion chemical vapor deposition method, J. Membr. Sci. 251 (2005) 151. [23] M. Nomura, M. Seshimo, H. Aida, K. Nakatani, S. Gopalakrishnan, T. Sugawara, et al., Preparation of a catalyst composite silica membrane reactor for steam reforming reaction by using a counterdiffusion CVD method, Ind. Eng. Chem. Res. 45 (2006) 3950. [24] P. Hacarlioglu, Y. Gu, S.T. Oyama, Studies of the methane steam reforming reaction at high pressure in a ceramic membrane reactor, J. Nat. Gas Chem. 15 (2006) 73. [25] K. Akamatsu, M. Nakane, T. Sugawara, T. Hattori, S.I. Nakao, Development of a membrane reactor for decomposing hydrogen sulfide into hydrogen using a high-performance amorphous silica membrane, J. Membr. Sci. 325 (2008) 16. [26] K. F. Jensen, Chemical Vapor Deposition, in: D. W. Hess, K. F. Jensen (Eds). Microelectronics processing — chemical engineering aspects. Advances in Chemistry Series 221. American Chemical Society, Washington, DC, 1989, pp. 199–263. [27] A.K. Pattanaik, V.K. Sarin, Basic principles of CVD thermodynamics and kinetics, in: J.H. Park (Ed.), Chemical Vapor Deposition, Surf. Eng. Ser., Vol. 2, ASM International, Materials Park, 2001, pp. 23–43. [28] M. Tsapatsis, G.R. Gavalas, G. Xomeritakis, Chemical vapor deposition membranes, in: N.K. Kanellopoulos (Ed.), Recent Advances in Gas Separation by Microporous Ceramic Membranes, Membrane Science and Technology Series, vol. 6, Elsevier Science, Amsterdam, 2000, p. 397. [29] T. Okubo, H. Inoue, Single gas permeation through porous glass modified with tetraethoxysilane, AIChE J. 35 (1989) 845. [30] M. Tsapatsis, S. Kim, S.S. Nam, G. Gavalas, Synthesis of hydrogen permselective SiO2, TiO2, Al2O3, and B2O3 membranes from the chloride precursors, Ind. Eng. Chem. Res. 30 (1991) 2152. [31] M. Tsapatsis, G. Gavalas, Structure and aging characteristics of H2-permselective SiO2-Vycor membranes, J. Membr. Sci. 87 (1994) 281. [32] K. Kuraoka, T. Kakitani, T. Suetsugu, T. Yazawa, Methanol vapor separation through the silica membrane prepared by the CVD method with the aid of evacuation, Sep. Purif. Technol. 25 (2001) 161. [33] S.F. Nitodas, E.P. Favvas, G.E. Romanos, M.A. Papadopoulo, A.C. Mitropoulos, N.K. Kanellopoulos, Development and characterization of silica-based membranes for hydrogen separation, J. Porous Mater. 15 (2008) 551. [34] S. Yan, H. Maeda, K. Kusakabe, S. Morooka, Hydrogen-permselective SiO2 membrane formed in pores of alumina support tube by chemical vapor deposition with tetraethyl orthosilicate, Ind. Eng. Chem. Res. 33 (1994) 2096. [35] B.K. Sea, M. Watanabe, K. Kusakabe, S. Morooka, S.S. Kim, Formation of hydrogen permselective silica membrane for elevated temperature hydrogen recovery from a mixture containing steam, Gas Sep. Purif. 10 (1996) 187. [36] B.K. Sea, E. Soewito, M. Watanabe, K. Kusakabe, S. Morooka, S.S. Kim, Hydrogen recovery from a H2-H2O-HBr mixture utilizing silica-based membranes at elevated temperatures. 1. Preparation of H2O- and H2-selective membranes, Ind. Eng. Chem. Res. 37 (1998) 2502. [37] G.J. Hwang, K. Onuki, S. Shimizu, H. Ohya, Hydrogen separation in H2-H2O-HI gaseous mixture using the silica membrane prepared by chemical vapor deposition, J. Membr. Sci. 162 (1999) 83. [38] G.J. Hwang, K. Onuki, S. Shimizu, Separation of hydrogen from a H2-H2O-HI gaseous mixture using a silica membrane, AIChE J. 46 (2000) 92.

Chapter

2

Review of Silica Membranes for Hydrogen Separation

59

[39] Y. Gu, S.T. Oyama, Ultrathin, hydrogen-selective silica membranes deposited on aluminagraded structures prepared from size-controlled boehmite sols, J. Membr. Sci. 306 (2007) 216. [40] T. Yamaguchi, X. Ying, Y. Tokimasa, B.N. Nair, T. Sugawara, S.I. Nakao, Reaction control of tetraethylorthosilicate (TEOS)/O3 and tetramethylorthosilicate (TMOS)/O3 counter diffusion chemical vapour deposition for preparation of molecular-sieve membranes, Phys. Chem. Chem. Phys. 2 (2000) 4465. [41] S. Gopalakrishnan, J.D. Costa, Hydrogen gas mixture separation by CVD silica membrane, J. Membr. Sci. 323 (2008) 144. [42] S. Araki, N. Mohri, Y. Yoshimitsu, Y. Miyake, Synthesis, characterization and gas permeation properties of a silica membrane prepared by high-pressure chemical, vapor deposition, J. Membr. Sci. 290 (2007) 138. [43] T. Nagano, S. Fujisaki, K. Sato, K. Hataya, Y. Iwamoto, Relationship between the mesoporous intermediate layer structure and the gas permeation property of an amorphous silica membrane synthesized by counter diffusion chemical vapor deposition, J. Am. Ceram. Soc. 91 (2008) 71. [44] Y. Ohta, K. Akamatsu, T. Sugawara, A. Nakao, A. Miyoshi, S.I. Nakao, Development of pore size-controlled silica membranes for gas separation by chemical vapor deposition, J. Membr. Sci. 315 (2008) 93. [45] J.C.S. Wu, H. Sabol, G.W. Smith, D.L. Flowers, P.K.T. Liu, Characterization of hydrogenpermselective microporous ceramic membranes, J. Membr. Sci. 96 (1994) 275. [46] R.K. Iler, The Chemistry of Silica, Wiley, New York, 1979. [47] R.M. deVos, W.F. Maier, H. Verweij, Hydrophobic silica membranes for gas separation, J. Membr. Sci. 158 (1999) 277. [48] M. Asaeda, M. Kashimoto, Sol-gel derived silica membranes for separation of hydrogen at high temperature-separation performance and stability against steam, in: Proceeding of Fifth International Conference on Inorganic Membrane, Nagoya, Japan, 1998, p. 172. [49] M. Kanezashi, M. Asaeda, Hydrogen permeation characteristics and stability of Ni-doped silica membranes in steam at high temperature, J. Membr. Sci. 271 (2006) 86. [50] G.P. Fotou, Y.S. Lin, S.E. Pratsinis, Hydrothermal stability of pure and modified microporous silica membranes, J. Mater. Sci. 30 (1995) 2803. [51] J.H.A. Hekkink, R.S.A. de Lange, A.A. Ten Hoeve, P.J.A.M. Blankenvoorde, K. Keizer, A.J. Burggraaf, Characterization and permeation properties of binary SiO2-TiO2 and SiO2Al2O3 modified gamma-alumina membranes, Key Eng. Mater. 61 (1991) 375. [52] K. Yoshida, Y. Hirano, H. Fujii, T. Tsuru, M. Asaeda, Hydrothermal stability and performance of silica-zirconia membranes for hydrogen separation in hydrothermal conditions, J. Chem. Eng. Jpn. 34 (2001) 523. [53] S.T. Hwang, K. Kammermeyer, Surface diffusion in microporous media, Can. J. Chem. Eng. 44 (1966) 82. [54] A.B. Shelekhin, A.G. Dixon, Y.H. Ma, Theory of gas diffusion and permeation in inorganic molecular-sieve membranes, AIChE J. 41 (1995) 58. [55] M. Bhandarkhar, A.B. Shelekhin, A.G. Dixon, Y.H. Ma, Adsorption, permeation and diffusion of gases in microporous membranes. I. Adsorption of gases on microporous glass membranes, J. Membr. Sci. 75 (1992) 221. [56] R.S.A. de Lange, K. Keizer, A.J. Burggraaf, Analysis and theory of gas transport in microporous sol-gel derived ceramic membranes, J. Membr. Sci. 104 (1995) 81. [57] S. Kim, G.R. Gavalas, Preparation of H2 permselective silica membranes by alternating reactant vapor deposition, Ind. Eng. Chem. Res. 34 (1995) 168. [58] S. Jiang, Y. Yan, G.R. Gavalas, Temporary carbon barriers in the permeation of H2-permselective silica membranes, J. Membr. Sci. 103 (1995) 211.

60

Inorganic, Polymeric and Composite Membranes

[59] C.E. Megiris, J.H.E. Glezer, Synthesis of H2-permselective membranes by modified chemical vapor deposition. Microstructure and permselectivity of SiO2/C/Vycor membranes, Ind. Eng. Chem. Res. 31 (1992) 1293. [60] H.Y. Ha, J.S. Lee, S.W. Nam, I.W. Kim, S.A. Hong, Alumina composite membranes prepared by MOCVD, J. Mater. Sci. Lett. 16 (1997) 1023. [61] S.W. Nam, H.Y. Ha, S.P. Yoon, J. Han, T.H. Lim, I.H. Oh, et al., Hydrogen-permselective TiO2/SiO2 membranes formed by chemical vapor deposition, Korean Membr. J. 3 (2001) 69. [62] K. Kuroaka, Z. Shugen, K. Okita, T. Kakitani, T. Yazawa, Permeation of methanol vapor through silica membranes prepared by the CVD method with the aid of evacuation, J. Membr. Sci. 160 (1999) 31. [63] S.T. Oyama, D. Lee, S. Sugiyama, K. Fukui, Y. Iwasawa, Characterization of a highly selective hydrogen permeable silica membrane, J. Mater. Sci. 36 (2001) 5213. [64] S.T. Oyama, D. Lee, P. Hacarlioglu, R.F. Saraf, Theory of hydrogen permeability in nonporous silica membranes, J. Membr. Sci. 244 (2004) 45. [65] R.J.R. Ulhorn, K. Keizer, A.J. Burggraaf, Gas and surface diffusion in modified g-alumina systems, J. Membr. Sci. 46 (1989) 225. [66] T. Okubo, M. Watanabe, K. Kusakabe, S. Morooka, Preparation of g-alumina thin membrane by a sol-gel processing and its characterization by gas permeation, J. Mater. Sci. 25 (1990) 4822. [67] T. Okubo, K. Haruta, K. Kusakabe, S. Morooka, H. Anzai, S. Akiyama, Preparation of sol-gel derived thin membrane on a porous ceramic hollow filter by the filtration technique, J. Membr. Sci. 59 (1991) 73. [68] S.S. Kim, B.K. Sea, Gas permeation characteristics of silica/alumina composite membrane prepared by chemical vapor deposition, Korean J. Chem. Eng. 18 (2001) 322. [69] B. Sea, K.H. Lee, Modification of mesoporous g-alumina with silica and application for hydrogen separation at elevated temperature, J. Ind. Eng. Chem. 7 (2001) 417. [70] B. Sea, K.H. Lee, Molecular sieve silica membrane synthesized in mesoporous g-alumina layer, Bull. Korean Chem. Soc. 22 (2001) 1400. [71] C.L. Lin, D.L. Flowers, P.K.T. Liu, Characterization of ceramic membranes II. Modified com˚ , J. Membr. Sci. 92 (1994) 45. mercial membranes with pore size under 40 A [72] A. Nijmeijer, B.J. Bladergroen, H. Verweij, Low-temperature CVI modification of g-alumina membranes, Microp. Mesop. Mater. 25 (1998) 179. [73] G.J. Hwang, J.W. Kim, H.S. Choi, K. Onuki, Stability of a silica membrane prepared by CVD using g-and a-alumina tube as support tube in the HI-H2O gaseous mixture, J. Membr. Sci. 215 (2003) 293. [74] M. Nomura, H. Aida, S. Gopalakrishnan, T. Sugawara, S.I. Nakao, S. Yamazaki, et al., Steam stability of a silica membrane prepared by counter diffusion chemical vapor deposition, Desalination 193 (2006) 1.

Chapter 3

Amorphous Silica Membranes for H2 Separation Prepared by Chemical Vapor Deposition on Hollow Fiber Supports Dmitri D. Iarikov1, Pelin Hacarlioglu1 and S. Ted Oyama1,2,* Environmental Catalysis and Nanomaterials Laboratory, Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA 2 Department of Chemical Systems Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan * Corresponding author: E-mail addresses: [email protected], [email protected] 1

INTRODUCTION The separation of H2 is a critical industrial process because of its many uses in the hydrodesulfurization of petroleum, production of syngas, synthesis of ammonia, as well as in the Fischer Tropsch, hydrogenation, and hydrocracking reactions [1]. Current methods for H2 separation are pressure swing adsorption and cryogenic distillation, but membrane separation has potential because it is generally more energetically favorable and cost effective [2]. Inorganic membranes possess sufficient chemical and thermomechanical stability so that they can be used in separations at high temperatures and pressures, and therefore, they find applications in membrane reactors for reactive separation. This technology allows combination of separation and reaction functions in a single unit and increases the efficiency of the process. A current limitation is the low surface area of inorganic membranes, and to address this, work was carried out to develop hollow fiber (HF) membranes for H2 separation. Silica membranes for H2 separation prepared by chemical vapor deposition (CVD) have been studied extensively, and an overview of the work performed in the last several years is provided in Table 3.1. Almost all of the reports reviewed utilized a-alumina inorganic supports coated with one or more layers of g-alumina. In general, the H2 permeance was of the order of 10
7 mol m
2 s
1 Pa
1 with H2 selectivity ranging between 70 and 10,000. Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

61

TABLE 3.1 Overview of the H2 Silica Separation Membranes Prepared by CVD on a-Alumina Supports TCVD ( C)

CVD Precursor

Thermal decomposition at high temperature using the opposing reactants technique

600

20–30 nm permselective layer of silica deposited on an intermediate multilayer alumina substrate with a graded structure

H2 Permeance

H2/N2

Ref

TEOS TIP

3.0  10
7

50a

[6]

600

TEOS

5.0  10
7

5900a

[7]

A 30–40 nm layer of silica–alumina composition was deposited on a porous alumina support by CVD in an inert atmosphere at high temperature

600

TEOS ATSB

2.0–3.0  10
7

940a

[8]

A composite silica membrane obtained by depositing a thin silica layer on a porous alumina support by CVD in inert gas at atmospheric pressure

600

TEOS

1.2  10
7

2800a

[9]

Counterdiffusion CVD method

400

TEOS

O2

7.0  10
7

60

[20]

600

TMOS

O2

1.2  10
7

12000

[3]


7

CVD Details

Oxidant

Tubular a-alumina support

NOK hollow fiber a-alumina support Counterdiffusion CVD Rapid CVD modification of a sol–gel silica layer and g-alumina intermediate layer

600

TEOS

O2

6.4  10

2300

[10]

Counterdiffusion CVD, applying vacuum to tube side

600

TMOS

O2

7  10
9

70

[4]

1000

[18]

One-sided diffusion CVD assisted by vacuum

600

PTMS

O2


7

2.0  10

One-sided diffusion CVD assisted by vacuum Counterdiffusion CVD

600 300–600

DMDPS TMOS

O2 O2/O3

2.0  10
6
8

8.5  10


8

200

[18]

1200

[5]

The reactant was evacuated through the porous wall of the support with no g-alumina intermediate layer

600–650

TEOS

1  10

1000þ

[12]

Vacuum continuously applied to the tube side

400–600

TEOS

0.1–1  10
7

1000þ

[21]

Vacuum applied to the tube side

600–700

TEOS

1.2  10
7

200

[22]

Continuous evacuation with no intermediate g-alumina layer

600

TEOS

1  10
8

1000þ

[13]

a

H2/CH4.

H2O

64

Inorganic, Polymeric and Composite Membranes

Tetraethylorthosilicate (TEOS) was the most popular silica precursor used, although other precursors such as tetramethoxysilane (TMOS) and phenyltrimethoxysilane (PTMS) were also utilized successfully [3–5]. Most groups utilized O2 as an oxidizing agent to promote silica precursor decomposition, but one group reported silica precursor decomposition in an inert atmosphere [6–9]. In this work, asymmetric a-alumina HFs were used as the membrane support material. HFs have an advantage over traditional tubular ceramic membranes because of their high surface area to volume ratio. They can also be assembled into multiple-tube reactors that further enhance the volume utilization. Table 3.1 summarizes previous work on the development of H2 separation membranes on HF supports. In all of the reports cited, the silica separation layer was deposited on the outside of the fiber. This works well in separation but cannot be applied to catalytic membrane reactors because contact of the permselective layer with the catalyst pellets can lead to abrasion problems. In further examining the results reported in the literature, it appears that in all but two studies [5,10] reactants were evacuated through the pore walls during the CVD application. The application of vacuum is claimed to improve the coverage of defects during silica deposition [11]. In all but two studies [12,13], a g-alumina intermediate layer was applied to the a-alumina substrate prior to CVD, and it improved the H2 permeance by roughly an order of magnitude. The silica precursor molar flow rates varied by over two orders of magnitude, but the effects on H2 permeance or selectivity were not evaluated. Three of the studies [4,5,10] used O2 or O3 as an oxidant to promote silica precursor decomposition in the membrane pores. The use of the oxidant reagent had an apparent effect of increasing the final H2 permeance compared to the studies that relied on silica precursor thermal decomposition alone. The aim of this work is to prepare H2 separation amorphous silica membranes with the amorphous silica layer deposited inside the HF using previously established CVD methods [7]. This is a challenging endeavor because the interior of the HFs is considerably rougher than the outer surface, making the deposition of intermediate layers more difficult. In order to make the fibers applicable in reactive separation, a new type of intermediate layer comprised of g-alumina and mesoporous silica had to be developed to allow CVD deposition of amorphous silica deposition on the inside of the fiber. Different factors affecting the separation such as carrier flow rates, precursor concentration, and intermediate layer compositions were evaluated. In order to achieve good separation, it was necessary to apply two g-alumina layers with the total number of intermediate layers of at least four. It was found that using three layers of mesoporous silica was required to produce high quality H2 separation membranes. Membranes with high TEOS molar flow rate resulted in better permeance and selectivity compared with membranes with lower TEOS molar flow rate.

Chapter

3

Silica Membranes on Hollow Fiber Supports

65

EXPERIMENTAL The HF inorganic porous supports used in this study were obtained from the NOK Corporation. The fibers (2.9 mm O.D. by 2.2 mm I.D.) had an asymmetric structure with reported porosity of 52% and a nominal pore size of 150 nm [14]. The fibers were cut into 3–5 cm pieces and were connected to nonporous alumina tubes with glass glaze (Duncan IN 1001) by heating to 880  C (in air) at the rate of 3.6  C min
1 and soaking at 880  C for 40 min. The glass glaze was applied to the cross-section of the tubes. In order to fix the light alumina tubes in place before heat treatment, the fibers were secured with ceramic pieces using epoxy resin. At high temperature, the epoxy glue was burned off and the ceramic bits were removed. Boehmite (AlOOH) sols were prepared by hydrolysis of aluminum isopropoxide (Aldrich, 98þ%) followed by acid peptization of the boehmite precipitate [7]. A quantity of aluminum isopropoxide (0.2 mol) was slowly added to 300 ml of distilled water at 360 K. The mixture was hydrolyzed at this temperature overnight at 358 K to form a boehmite precipitate. The precipitate was then peptized at above 363 K with a quantity of acetic acid (Acros Organics, 99.8%) or nitric acid (Sigma Aldrich, 70%) to form a slightly translucent sol. Peptization as defined here is an acid treatment used to transform the macroscopic oxide precipitates into colloidal particles via hydrolysis and condensation reactions. The Hþ/Al ratio was varied between 0.04 and 0.2 (pH of 4.2–5.1) to control the final sol particle size as described in our previous work, which also provides the details of this synthesis [8]. The particle size distributions of the resultant sols were estimated using dynamic light scattering (DLS). The calibration of the DLS apparatus (Horiba Model LB-500) was verified with a standard polystyrene solution (Duke Scientific Corporation). The colloidal silica sols were prepared using base-catalyzed hydrolysis and condensation of TEOS. The TEOS–ethanol–water mixture with a catalytic amount of ammonium hydroxide was refluxed at 50  C for 3 h to form a stable colloidal silica sol. Silica sols with different particle size distributions were obtained by varying the reflux temperature and the amounts of ammonium hydroxide [15]. The mean particle diameter of the resultant silica sols decreased with increase in reflux temperature from 2  C (500 nm) to 95  C (40 nm). By using higher NH3/TEOS ratios, it is possible to obtain silica sols with higher particle size distributions. For example, at NH3/TEOS of 0.8, the mean particle size of the sol was approximately 60 nm, whereas for the sol prepared with NH3/ TEOS equal to 6.0, the mean particle size was around 500 nm. Reflux temperature of 20  C was used in both cases. A composite intermediate layer combining g-alumina and mesoporous silica materials was deposited on the a-alumina HFs using the following procedure. First, an amorphous intermediate layer consisting of one or more layers of galumina was placed on an HF by depositing boehmite sols of controlled sizes followed by a layer of colloidal silica sol. The sols were applied to the inside of

66

Inorganic, Polymeric and Composite Membranes

the a-alumina HF supports for approximately 20 s. This was done by connecting a small volume of the sol to the HF and letting it flow freely with gravity. After coating, the membranes were dried in ambient air in a vertical position and calcined at 973 K with 1 K min
1 heating and cooling rates. The coating and the calcining steps were repeated after each individual coating. After the intermediate layers were applied to the HF, an amorphous silica layer was formed by thermal decomposition of TEOS in an inert atmosphere. The HF membrane with the intermediate support layer deposited on the inside surface was placed in a quartz shell and tube reactor and sealed with Swagelok fittings with Teflon ferrules (Figure 3.1). The reactor was then placed inside a furnace and heated up to the desired CVD temperature at a rate of 1 K min
1. The carrier gas (argon, ultra high purity) was passed through a bubbler filled with the silica precursor. The bubbler temperature was varied to adjust the amount of TEOS in the argon stream. The silica precursor was carried to the membrane and decomposed at the surface on the tube side of the membrane. A pure argon stream was used as a balance gas and was passed through the shell side of the reactor. The deposition time varied depending on the carrier flow rate and the concentration of the precursor in the carrier stream. Higher TEOS molar flow rate resulted in lower deposition time because more of the silica precursor was fed into the CVD reactor per unit of time. The CVD system consisted of a high-temperature cylindrical furnace that housed the tubular CVD reactor (Figure 3.2). The furnace temperature was maintained with a temperature controller (Cole-Palmer Digi-Sense) to  0.2  C. The carrier and the balance gases were supplied through mass flow controllers (Brooks 5850E series). The carrier gas passed through a bubbler containing TEOS, and the saturated vapor was carried into the CVD reactor. Heating tape was used to heat the stainless steel lines to prevent condensation of TEOS.

RESULTS AND DISCUSSION Amorphous silica membranes were prepared in this study using thermal decomposition of TEOS in an inert atmosphere. The synthesis was performed in a CVD reactor, and several membranes with different intermediate support layers were evaluated. The overall details of membrane synthesis are provided in Table 3.2, and the effect of the synthesis conditions on the membrane separation performance will be discussed in the following sections.

FIGURE 3.1 HF membrane reactor assembly.

Chapter

3

67

Silica Membranes on Hollow Fiber Supports

Vent Vent Temperature controller

Furnace

Membrane

MFC Union cross

Heating tape MFC TEOS

Balance gas

Oil bath Carrier gas FIGURE 3.2 Schematic diagram of the CVD system.

Pure Hollow Fiber Support Properties The permeance properties of the pure a-alumina HF supports before application of the intermediate or CVD layers were obtained at various temperatures using H2, CO2, and CH4 (Figure 3.3). Figure 3.3A indicates that the permeance increased linearly with the inverse square root of the gas molecular weight, and Figure 3.3B shows that the gas permeance increased linearly with the inverse square root of temperature. The dependency on temperature and molecular weight is consistent with Knudsen diffusion. In the Knudsen diffusion regime, the number of molecular collisions with the membrane walls is much greater than the number of collisions between molecules that is expected for the pore size of the support (see below). The membrane permeance in the Knudsen regime can be estimated using the following equation [16]:  1=2 ed 8 P¼ 3tt pRT MW where P—membrane permeance (mol m
2 s
1 Pa
1), e—porosity, d—pore diameter (m), t—tortuosity (assumed to equal unity), R—universal gas constant (m3 Pa K
1 mol
1), T—temperature (K), and MW—gas molecular weight (kg mol
1). The pore diameter of the bare HF support was calculated with the Knudsen equation using the permeance data collected at different temperatures (Figure 3.4). The calculated pore diameter was approximately 160 nm, which agrees very well with the nominal reported pore size of these membranes of 150 nm [14]. The value of membrane tortuosity was assumed to be unity.

TABLE 3.2 Synthesis Details for the Amorphous Silica Membranes for H2 Separation Membrane

Intermediate Layer

Precursor Flowratea

H2/ CO2

H2 Permeanceb

M-1A-3S-1

One layer of g-alumina (140 nm), three layers of mesoporous silica

11.5

125

3.43  10
8

M-1A-3S-2

One layer of g-alumina (120 nm), three layers of mesoporous silica (450, 300, 300 nm)

2.22

45

3.65  10
8

M-2A-2S-1

Two layers of g-alumina (120 nm), two layers of mesoporous silica (65 nm)

2.68

18

1.05  10-7

M-2A-2S-2

Two layers of g-alumina (135 nm), two layers of mesoporous silica (70 nm) diluted fivefold in ethanol

2.43

22

1.26  10
7

M-2A-2S-3

Two layers of g-alumina (135 nm), two layers of mesoporous silica (70 nm) diluted fivefold in water

2.48

42

4.53  10
8

M-2A-3S-1

Two layers of g-alumina (140 nm), three layers of mesoporous silica (300 nm)

3.02

380

2.56  10
7

M-2A-3S-2

Two layers of g-alumina (140 nm), three layers of mesoporous silica (300, 200, 70 nm)

10.4

82

2.71  10
7

M-2A-3S-3

Two layers of g-alumina (135 nm), three layers of mesoporous silica (200, 200, 80 nm)

3.02

101

9.96  10
8

M-2A-3S-4

Two layers of g-alumina (135 nm), three layers of mesoporous silica (200 nm)

0.727

64

1.37  10
7

mmol s
1. mol m
2 s
1 Pa
1.

a

b

Chapter

3

(b) Permeance (mol m-2 s-1 Pa-1)

Permeance (mol m-2 s-1 Pa-1)

(a) 298 573 873

5.0 ´ 10-5 4.0 ´ 10-5 3.0 ´ 10-5 2.0 ´ 10

-5

1.0 ´ 10-5 0.1

69

Silica Membranes on Hollow Fiber Supports

0.2

0.3

0.4

0.6

0.5

0.7

0.8

6.0 ´ 10-5 H2 He CH4

5.0 ´ 10-5 4.0 ´ 10-5 3.0 ´ 10-5

298 400

-5

2.0 ´ 10

1.0 ´ 10-5

573 873

0.03

0.04

0.05

0.06

(Temperature)-1/2

(Molecular weight)-1/2

FIGURE 3.3 (a) Permeance of several gases at three different temperatures through the HF support plotted as a function of the inverse square root of gas molecular weight. (b) Permeance of H2, He, and CH4 plotted as a function of the inverse square root of temperature.

Pore diameter (nm)

300

200

100 298 400 573 873 0

2.0 ´ 10−5

0.0

Permeance (mol

4.0 ´ 10−5

6.0 ´ 10−5

m−2 s−1 Pa−1)

FIGURE 3.4 The pore diameter of the HF bare supports calculated using the Knudsen equation from the permeance data collected at different temperatures.

Mesoporous Silica Layer Several combinations of intermediate layers were evaluated to determine how their structure and morphology affected H2 separation. It was determined that applying less than four intermediate layers did not produce satisfactory results in H2 selectivity. Applying two layers of g-alumina produced better selectivity than one layer. It was also determined that using three layers of mesoporous silica on top of g-alumina produced better selectivity than using two layers of mesoporous silica. As seen in Figure 3.5, membranes with one g-alumina and three mesoporous silica layers resulted in the worst H2 permeance, with

70

Inorganic, Polymeric and Composite Membranes

the membranes produced with two g-alumina and three mesoporous silica layers resulting in the best permeance. The membranes with two g-alumina layers and three mesoporous silica layers had the best permeance and selectivity, and the membranes that had two mesoporous silica layers on top of g-alumina had the worst H2 selectivity. The SEM images of the cross-section and the inside surface of the HF show that there are large asymmetric pores transversing the inside surface of the fiber (Figure 3.6). The size of the large radial pores (Figure 3.6) is approximately 10–20 mm in diameter, which makes them too large to be the limiting pores obtained by fitting to the Knudsen equation. There are visible openings (Figure 3.6B) in the walls of these pores that are about one-tenth this size, and those are probably the dominant limiting pores involved in the transport of gases in the support. Other research groups that utilized these supports for membrane preparation traditionally deposited the selective layer on the outside of the fiber, which can be seen to be smoother. In the current study, a method 1000 1A-3S 2A-2S 2A-3S

1 × 10−7

100

1 × 10−8

10

H2/CO2 selectivity (−)

H2 permeance (mol m−2 s−1 Pa−1)

1 × 10−6

FIGURE 3.5 Permeance and selectivity for membranes with different numbers of intermediate layers of g-alumina (A) and mesoporous silica (S).

10 mm

100 mm

FIGURE 3.6 Cross-section scanning electron microscopy image of the uncoated HF.

Chapter

3

Silica Membranes on Hollow Fiber Supports

71

was developed to prepare an intermediate layer that would allow for successful preparation of the H2 separation membranes on the inside of the HF. A mesoporous silica layer was deposited on top of the g-alumina to correct the defects in the g-alumina layer. Membranes synthesized with one or two intermediate g-alumina layers with no mesoporous silica did not show H2 selectivity above the Knudsen value, indicating that there were large pore defects still remaining. Membranes with three g-alumina layers without mesoporous silica had slightly improved H2 selectivity that did not exceed 10. Furthermore, applying a mesoporous silica layer by itself did not produce H2 selectivity above Knudsen. However, a mesoporous silica layer deposited on top of the g-alumina layer corrected the defects and resulted in proper H2 separation. This occurs because the silica nanoparticles correct the defects in the underlying g-alumina layer and allow for a successful amorphous silica deposition by CVD on top of g-alumina. A schematic representation of the intermediate layer structure and an actual SEM photograph are shown in Figure 3.7 where one can clearly see the CVD (amorphous) silica layer on top of g-alumina. There were several membranes prepared with different combinations of g-alumina and mesoporous silica layers (Table 3.2). The nomenclature for the membrane samples indicates the number of alumina and silica intermediate layers that were used and the specimen number (e.g., M-3A-1-S-1). The membranes prepared with three mesoporous silica layers gave better H2 selectivity when compared with the membranes prepared with two mesoporous silica layers provided that the number of g-alumina layers was the same. This can be explained by examining the inside morphology of the HF. The ridges that run along the inside of the HF must be completely covered up with the mesoporous silica layer to ensure that the defects in the g-alumina layer are eliminated. This occurs only when a thick mesoporous silica layer is applied to the membrane (Figure 3.8B). Therefore, when less than three silica layers were applied, the silica nanoparticles did not cover the membrane surface entirely, particularly on top of the ridges of the HF (Figure 3.8B). This left the g-alumina layer exposed and resulted in lower H2 selectivity. For examples, for membranes g-Alumina

Defects

CVD silica

Mesoporous silica

1 mm

FIGURE 3.7 A schematic representation of how the mesoporous silica layer corrects the defects in the g-alumina. On the right is a picture of the CVD layer which is clearly visible on the SEM image.

72

Inorganic, Polymeric and Composite Membranes

(a)

(b) 10 mm

10 mm

1×10

1×10−6

20 15

1×10−7

10

−8

1×10

5

−9

1×10

0

1

2

3

4

CVD time (h)

1×10−4 1×10−5

40

−6

1×10

30 H2 CO2 CH4

1×10−7 1×10−8

20 10

1×10−9 −10

H2/CO2 selectivity (−)

H2 CO2 CH4

−5

Permeance (mol m−2 s−1 pa−1)

1×10−4

H2/CO2 selectivity (−)

Permeance (mol m−2 s−1 pa−1)

FIGURE 3.8 (a) A membrane prepared with three mesoporous silica layers and (b) a membrane prepared with just two silica layers.

0

1×10

0

1

2

3

4

CVD time (h)

FIGURE 3.9 Gas permeance plotted as a function of CVD time measured at 650  C for M-2A-2S-1 and M-2A-2S-2. The H2 selectivity is plotted on the same graph.

M-2A-2S-2 and M-2A-2S-3 that had two mesoporous silica layers prepared from diluted colloidal silica sols, the H2 selectivity did not exceed 40 (Figure 3.9).

Amorphous g-Alumina Layer It was determined that the membranes with only one intermediate g-alumina permeance of layer resulted in a relatively low H2 3.4–3.6  10
8 mol m
2 s
1 Pa
1 compared with the membranes prepared with two g-alumina layers, 1.0–2.7  10
7 mol m
2 s
1 Pa
1. This was because a defective surface required a thicker coating of amorphous silica to form a continuous layer causing the H2 permeance to decrease. Apart from having an effect on the membrane permeance, the intermediate layer composition also had an effect on the membrane selectivity. To illustrate this point, it is convenient to compare two membranes, M-2A-3S-3 and M-1A-3S-2, which were prepared in a similar fashion except that two layers of g-alumina were applied to M-2A-3S-3 and only one layer of g-alumina was applied to M-1A-3S-2 (Table 3.2). The membrane prepared with two g-alumina layers displayed a much higher H2 selectivity (101 vs. 45). This happened because the additional

Chapter

3

Silica Membranes on Hollow Fiber Supports

73

g-alumina layer reduced the number of defects and therefore significantly improved H2 selectivity.

Silica Precursor and Carrier Gas Flow Rate Effects on the Membrane Separation Performance The silica precursor molar flow rate is a function of the concentration of the precursor in the carrier gas stream and the carrier gas flow rate. The concentration of TEOS is adjusted by varying the oil bath temperature in which the TEOS bubbler is immersed and subsequently varying its vapor pressure. In order to evaluate the TEOS molar flow rate effects on the membrane synthesis, they were arbitrarily divided into three categories: low flow rates (up to 1 mmol s
1), medium flow rates (1–10 mmol s
1), and high flow rates (above 10 mmol s
1). Flow rates in mmol s
1 can be converted into cm3 (NTP)/min by multiplying by 1.5. The membrane M-1Al-3S-1 prepared with a high TEOS molar flow rate (11.5 mmol s
1) was compared with the membrane M-1A-3S-2 prepared with a medium flow rate (2.22 mmol s
1). The higher flow rate resulted in a higher H2 selectivity (125 vs. 45). The same trend was observed for the membranes with two g-alumina and two mesoporous silica intermediate layers. For example, the membranes with medium-to-high TEOS flow rates (3–10 mmol s
1) displayed higher H2 selectivity (80–380), whereas the membrane prepared with a low TEOS flow rate (0.7 mmol s
1) displayed an H2 selectivity of only 64. An alternative way to look at the membrane synthesis is to evaluate the carrier gas effect on the H2 selectivity. The membranes M-2A-3S-3 and M-2A-3S-4 were synthesized at the same conditions except for the carrier gas flow rates (22 and 76 standard cubic centimeters, sccm, respectively). Since the TEOS concentrations in the carrier stream for these two membranes were virtually the same, it is apparent that a higher carrier gas flow rate increases the H2 selectivity of the membrane. The effect of the silica precursor concentration in the carrier stream was also examined separately. The TEOS concentration in the carrier stream for the membrane M-1A-3S-1 was approximately an order of magnitude larger than the TEOS concentration used in the synthesis of M-1A-3S-2. The resultant H2 selectivity for M-1A-3S-1 was over 120 and for M-1A-3S-2 it was only 45. Since the H2 permeances for both these membranes were almost the same in the range of 3.4–3.6  10
8 mol m
2 s
1 Pa
1, the higher selectivity was attributed to a higher TEOS concentration in the carrier stream because it would give a thicker CVD layer to cover up defects in the support layer. The TEOS concentration that is too light may leave defects in the support that result in lower H2 selectivity [8]. The SEM images of the cross-sections and surfaces of the two membranes can be seen in Figure 3.10. Overall, the H2 permeance of the best sample was 2.56  10
7 mol m
2 s
1 Pa
1 with H2 selectivity of over 300. It is generally accepted that an H2 permeance of 10
7 mol m
2 s
1 Pa
1 is needed for commercialization [6] and that a selectivity of 100 is sufficient for high performance in a membrane reactor [17].

74

Inorganic, Polymeric and Composite Membranes

2 mm 2 mm

200 nm 1 mm

FIGURE 3.10 Membrane cross-section and mesoporous silica surface for M-1A-3S-1 (left) and M-1A-3S-2 (right).

Gas Separation Mechanism Just as the pure a-alumina support, the intermediate layers alone have Knudsen selectivity when separating gases and therefore cannot be used for effective separation without the CVD layer. The application of intermediate layers does not significantly decrease gas permeance with respect to the pure a-alumina support. For example, an application of two g-alumina layers reduced the gas permeance of the HF to 88% of the original value. After the application of CVD, the H2 permeance decreased dramatically from 10
5 to
7
2
1
1 10 mol m s Pa . The gas separation mechanism switched from Knudsen diffusion to activated diffusion and increased rather than decreased with temperature (Figure 3.11). The H2 activation energy calculated for a g-alumina membrane prepared on HF support was 18.0 kJ mol
1, which is comparable to other Nanosil membranes developed in our laboratory on either Vycor or a-alumina tubular supports (16.4 and 14.8 kJ mol
1, respectively) [18,19]. As the CVD continued, the H2 permeance decreased slower than other gases, which resulted in high H2 selectivity (Figure 3.12). However, after a certain point with further application of CVD, the H2 selectivity actually decreased. This is characteristic of membranes that contain defects in the intermediate layer, which are too large to be covered up by CVD. As the deposition continued, a thicker layer of CVD was deposited on the membrane surface that decreased the overall H2 permeance. The larger defects in the membrane that were responsible for the passage of larger gas molecules, such as CO2 and CH4, remained uncovered, which resulted in a decrease in H2 selectivity.

3

1´10-6

H2/CH4

H2 Permeance (m−2 s−1 Pa−1)

75

Silica Membranes on Hollow Fiber Supports

100

CO2 CH4

1´10-7

80 H2/CO2

60 40

1´10-8

Selectivity (−)

Chapter

20 1´10-9 100

0 200

300 400 500 Temperature (°C)

600

FIGURE 3.11 Gas permeance and selectivity for H2, CO2, and CH4 measured at 600 and 150  C for M-2A-3S-2.

1´10-4 60

CO2

1´10-5 1´10-6

40

1´10-7 20

Selectivity (−)

Permeance (mol m−2 s−1 Pa−1)

H2

1´10-8 0 1´10-9 0

2

4 6 CVD time (h)

8

10

12

FIGURE 3.12 H2 and CO2 permeance and selectivity for M-2A-3S-4 plotted as a function of CVD time.

CONCLUSIONS Amorphous silica membranes prepared by CVD with novel intermediate layer morphology were developed. In the preparation of these membranes, a-alumina asymmetric HFs were first coated with boehmite sols and then with colloidal silica sols to establish a composite g-alumina mesoporous silica intermediate layer. The mesoporous silica nanoparticles reduced the defects in the g-alumina

76

Inorganic, Polymeric and Composite Membranes

intermediate layer, which resulted in an increase in H2 selectivity. Following the intermediate layer deposition, a permselective silica layer was deposited by CVD at atmospheric pressure using TEOS as the silica precursor. The resultant membranes displayed H2 permeance on the order of 1  10
7 mol m
2 s
1 Pa
1 and H2 selectivity of over 100 and were therefore suitable for H2 separation. The synthesis conditions that allowed for optimal H2 selectivity and permeance were examined to improve the membrane quality. It was determined that high TEOS molar flow rate improved the H2 permeance and selectivity. It was also determined that the membranes prepared with two g-alumina layers exhibited higher H2 selectivity and permeance when compared with the membranes prepared with one g-alumina layer due to the decrease in surface defects. The gas permeation mechanism for the membranes before CVD application followed Knudsen diffusion and switched to activated diffusion following the application of CVD.

ACKNOWLEDGMENTS For support of this work, the author acknowledges the Director, National Science Foundation, Division of Chemical, Bioengineering, Environmental, and Transport Systems (CBET) under grant CBET-084316, the National Energy Technology Laboratory under the NETL-RUA program grant number 5.681.884.001, the Mombukagakusho Kakenhi grant-in-aid Kiban kenkyu B 22-360,335.

REFERENCES [1] T.V. Choudhary, V.R. Choudhary, Energy-efficient syngas production through catalytic oxymethane reforming reactions, Angew. Chem. Int. Ed. 47 (2008) 1828–1847. [2] T.M. Nenoff, N.W. Ockwig, Membranes for hydrogen separation, Chem. Rev. 107 (2007) 4078–4110. [3] T. Nagano, S. Fujisaki, K. Sato, K. Hataya, Y. Iwamoto, Relationship between the mesoporous intermediate layer structure and the gas permeation property of an amorphous silica membrane synthesized by counter diffusion chemical vapor deposition, J. Am. Ceram. Soc. 91 (2008) 71–76. [4] Y. Ohta, K. Akamatsu, T. Sugawara, A. Nakao, A. Miyoshi, S. Nakao, Development of pore size-controlled silica membranes for gas separation by chemical vapor deposition, J. Membr. Sci. 315 (2008) 93–99. [5] M. Nomura, K. Ono, S. Gopalakrishnan, T. Sugawara, S. Nakao, Preparation of a stable silica membrane by a counter diffusion chemical vapor deposition method, J. Membr. Sci. 251 (2005) 151–158. [6] Y. Gu, S.T. Oyama, Permeation properties and hydrothermal stability of silica–titania membranes supported on porous alumina substrates, J. Membr. Sci. 345 (2009) 267–275. [7] Y. Gu, S.T. Oyama, Ultrathin, hydrogen-selective silica membranes deposited on aluminagraded structures prepared from size-controlled boehmite sols, J. Membr. Sci. 306 (2007) 216–227.

Chapter

3

Silica Membranes on Hollow Fiber Supports

77

[8] Y. Gu, P. Hacarlioglu, S.T. Oyama, Hydrothermally stable silica-alumina composite membranes for hydrogen separation, J. Membr. Sci. 310 (2008) 28–37. [9] D. Lee, L. Zhang, S.T. Oyama, S. Niu, R.F. Saraf, Synthesis, characterization, and gas permeation properties of a hydrogen permeable silica membrane supported on porous alumina, J. Membr. Sci. 231 (2004) 117–126. [10] S. Gopalakrishnan, Y. Yoshino, M. Nomura, B.N. Nair, S.I. Nakao, A hybrid processing method for high performance hydrogen-selective silica membranes, J. Membr. Sci. 297 (2007) 5–9. [11] B.-K. Sea, K. Kusakabe, S. Morooka, Pore size control and gas permeation of silica membranes by pyrolysis of phenyl-substituted ethoxysilanes with cross-flow through a porous support wall, J. Membr. Sci. 130 (1997) 41–52. [12] S. Morooka, S. Yan, K. Kusakabe, Y. Akiyama, Formation of hydrogen-permselective SiO2 membrane in macropores of a-alumina support tube by thermal decomposition of TEOS, J. Membr. Sci. 101 (1995) 89–98. [13] S. Morooka, S.S. Kim, S. Yan, K. Kusakabe, M. Watanabe, Separation of hydrogen from an H2-H2O-HBr system with an SiO2 membrane formed in macropores of an a-alumina support tube, Int. J. Hydrogen Energy 21 (1996) 183–188. [14] J. Tong, L. Su, K. Haraya, H. Suda, Thin and defect-free Pd-based composite membrane without any interlayer and substrate penetration by a combined organic and inorganic process, Chem. Commun. (2006) 1142–1144. [15] S. Park, D. Lee, C. Yu, K. Lee, K. Lee, Effect of silica particle size on performance of porous stainless-steel-supported silica membranes prepared by the DRFF and SRFF method, Ind. Eng. Chem. Res. 47 (2008) 6211–6215. [16] A.J. Burggraaf, Transport and separation properties of membranes with gases and vapours, in: A.J. Burggraaf, L. Cot (Eds.), Fundamentals of Inorganic Membrane Science and Technology, Elsevier, Amsterdam, 1996, pp. 331–433. [17] S.T. Oyama, H. Lim, An operability level coefficient (OLC) as a useful tool for correlating the performance of membrane reactors, Chem. Eng. Sci. 151 (2009) 351–358. [18] D. Lee, S.T. Oyama, Gas permeation characteristics of a hydrogen selective supported silica membrane, J. Membr. Sci. 210 (2002) 291–306. [19] S.T. Oyama, D. Lee, P. Hacarlioglu, R.F. Saraf, Theory of hydrogen permeability in nonporous silica membranes, J. Membr. Sci. 244 (2004) 45–53. [20] S. Gopalakrishnan, J.C. Diniz da Costa, Hydrogen gas mixture separation by CVD silica membrane, J. Membr. Sci. 323 (2008) 144–147. [21] S. Yan, H. Maeda, K. Kusakabe, S. Morooka, Y. Akiyama, Hydrogen-permselective SiO2 membrane formed in pores of alumina support tube by chemical vapor deposition with tetraethylorthosilicate, Ind. Eng. Chem. Res. 33 (1994) 2096–2101. [22] B. Sea, M. Watanabe, K. Kusakabe, S. Morooka, S. Kim, Formation of hydrogen permselective silica membrane for elevated temperature hydrogen recovery from a mixture containing steam, Gas Sep. Purif. 10 (1996) 187–195.

This page intentionally left blank

Chapter 4

Ab Initio Studies of Silica-Based Membranes: Activation Energy of Permeation Pelin Hacarlioglu1, Luke Achenie1 and S. Ted Oyama1,2,* Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA 2 Department of Chemical Systems Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan * Corresponding author: E-mail addresses: [email protected], [email protected] 1

INTRODUCTION Highly hydrogen-permeable silica-based membranes supported on a porous alumina substrate have been obtained by high-temperature chemical vapor deposition (CVD) of alkoxide precursors [1–3]. As opposed to previous studies that used O2 or H2O as co-reactants at low temperatures [4,5], the use of an inert carrier gas and high temperatures results in the deposition of a thin amorphous silica layer on the surface of the porous alumina support rather than in the bulk. This gives rise to high permeance and selectivity. A concern with these silica membranes is their loss of stability in the presence of water vapor at high temperatures. Water vapor catalyzes the formation of Si–O–Si bonds from silanol groups (Si–OH) that leads to the densification of the silica membranes and the shrinkage of pores [6]. It has been reported that the preparation of composite membranes composed of silica and other inorganic oxides such as alumina (Al2O3), titania (TiO2), and zirconia (ZrO2) by sol–gel methods improves the hydrothermal stability of the silica membranes. Fotou et al. [7] observed an increase in hydrothermal stability after doping the starting silica sol with 3% alumina. They also reported that silica membranes doped with 6% alumina or magnesia did not show improved permeance due to a substantial decrease in the surface area compared to the pure silica membranes. Yoshida et al. [8] examined the hydrothermal stability of composite silica membranes prepared by sol–gel deposition of the silica with varying amounts of zirconia (10–50 mol%). A 70% drop in permeance was observed with the Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

79

80

Inorganic, Polymeric and Composite Membranes

10 mol% zirconia–silica composite membrane with an increased H2/N2 selectivity of 190, while the H2 permeance of a 50 mol% zirconia–silica membrane stayed constant after a 20-h exposure to 13–33 mol% steam at a temperature of 773 K. Hekkink et al. [9] reported that the permeances of sol–gel-derived 10 mol% alumina–silica and 10 mol% titania–silica composite membranes were 6
10 8 and 2.2
10 7 mol m 2 s 1 Pa 1, respectively, which were high. However, these composite membranes had a very limited selectivity of H2 over CO, which is commonly observed with sol–gel-derived membranes. In recent studies, Gu and Oyama [6,10] showed that composite membranes prepared by the multicomponent CVD of silica–alumina and silica–titania showed excellent stability with an H2 permeance of 10 7 mol m 2 s 1 Pa 1 after a 500-h exposure to 16 mol% steam at 873 K. Much attention has been given to the gas permeation mechanism in microporous inorganic membranes, including zeolite membranes. These consist of classical Knudsen [11], surface diffusion [12], and activated-type Knudsen diffusion mechanisms [13–15]. However, for dense silica membranes, the diffusion of gas species was described as involving random jumps between adjacent solubility sites connected by passageways. This mechanism accounted for the unusual permeance order of He, H2, and Ne, which does not follow the mass or molecular size of the diffusing species [16]. This chapter deals with the modeling of the structure of composite silica membranes using an ab initio method performed in a previous study [17]. The permeating species are considered to pass through a single critical ring opening in the process of hopping from one solubility site to another. The ring structures and the diffusing species were optimized using a hybrid functional in density functional theory (DFT) with a highly accurate basis set. In this study, the same simulation technique was used to calculate the activation energies of various gaseous species (He, H2, Ne, CO, CO2, CH4) through silica composite membranes containing aluminum, boron, silicon, titanium, yttrium, and zirconium.

PREVIOUS THEORETICAL STUDIES ON DENSE SILICA-BASED MEMBRANES The structure of the silica membranes obtained by high-temperature CVD of SiO2 on porous supports can be considered to be similar to that of vitreous silica glass, which can be visualized as a disordered form of b-cristobalite [18]. The structure of b-cristobalite is formed from six-membered rings, but bonds break and reform in the process of glass formation to create a random network containing five-, six-, seven-, and eight-membered silicate rings. The structure has solubility sites of approximately 0.3 nm in diameter (Figure 4.1). Basically, the restricted size of the interstitial solubility sites in the dense silica membranes allows accommodation of only small species (He ¼ 0.26 nm, Ne ¼ 0.275 nm, H2 ¼ 0.289 nm) [19] that are smaller than the size of the solubility sites, while prohibiting the entrance of large gas molecules (CO2 ¼ 0.33 nm, CO ¼ 0.376 nm,

Chapter

4

Ab Initio Studies of Silica-Based Membranes

81

FIGURE 4.1 Solubility site in b-cristobalite. (a) Ball-and-stick model of the solubility site. Large spheres, oxygen; small spheres, silicon. (b) Polyhedral representation of the same site. Each vertex is an oxygen atom. There are four 6-membered rings (doorways) that are tetrahedrally oriented.

CH4 ¼ 0.38 nm) [19]. The unusual permeance order was explained by a trade-off between molecular size and mobility. The smaller species (e.g., He and H2) have a larger number of solubility sites able to accommodate them and a smaller activation barrier for hopping, leading to higher permeance. The heavier species (e.g., Ne and He) have a lower jump frequency between sites. It was also found that the activation energies for permeation through the membranes were much lower (He ¼ 8.0 kJ mol 1, H2 ¼ 14.8 kJ mol 1, Ne ¼ 16.6 kJ mol 1) than those for vitreous glass reported in the literature (H2 ¼ 37.2–38.8 kJ mol 1 [20,21], He ¼ 17.8–21.3 kJ mol 1 [22–24], Ne ¼ 33.8–39.5 kJ mol 1 [25,26]). The activation energy of permeation of different species through the silica membranes was obtained empirically by fitting permeation data at different temperatures to an expression derived for jumps between solubility sites [16].   2 3=2  a 1 d2 h sh2 ðNS =NA Þ eDEK =RT ð4:1Þ P¼ 2pmkT 8p2 IkT ðehv =2kT  ehv =2kT Þ2 6L h In this equation, P is the permeance of the gas species (mol m 2 s 1 Pa 1), L is the thickness of the membrane (m), d is the jump distance (m), m is the mass of the species (kg), h is Planck’s constant (J s), k is Boltzmann’s constant (J K 1), v* is the vibrational frequency of the species in the passageways between the sorption sites (s 1), T is temperature (K), NS is the number of solubility sites available per m3 of silica volume (m 3), NA is Avogadro’s number, R is the gas constant (kJ mol 1 K 1), DEK is the activation energy for hopping between sorption sites (kJ mol 1), I is the moment of inertia (kg m2), and s is the symmetry number of the permeating species, with s ¼ 2 in the case of hydrogen. The exponent, a, accounts for incomplete loss of rotation, with a ¼ 0 for a nonrotating transition state.

METHOD OF CALCULATION It had been found earlier that the siloxane rings present in the membranes were predominantly six-membered [16], so this ring size was used to model the structure of the multicomponent silica rings (H12Si5O6X, X ¼ Si, Al, B, Ti, Y, Zr).

82

Inorganic, Polymeric and Composite Membranes

The geometries of the ring clusters were obtained by energy minimization of the structures, using a DFT method in Gaussian 03. The optimizations were conducted using the Becke3LYP hybrid functional in which the exchange functional is Becke’s three-parameter formulation (a linear combination of HartreeFock, local and gradient-corrected exchange terms) [27], and the nonlocal term is provided by the gradient-corrected correlation functional, LYP, of Lee, Yang, and Parr [28]. A 6-311G (2d, p) polarized basis set, which puts 2d functions on heavy atoms and p functions on hydrogens, was applied to optimize both the siloxane rings with Al, B, Ti and the diffusing species. In the case of Zr and Y, which are post-third-row elements, a LANL2DZ basis set was used to optimize the siloxane rings and the diffusing species. For these very large nuclei, electrons near the nucleus are treated in an approximate way, via effective core potentials (ECPs). This treatment includes some relativistic effects, which are important in these atoms. The diffusion of gas molecules through the siloxane rings occurs by passage through a diffusion saddle point in the ring structures. In the optimization calculations, the positions of the atoms that build up the siloxane rings were allowed to move. The activation energy was calculated as the interaction energy for the permeating species passing through the diffusion saddle point of each siloxane ring. The interaction energy is defined as the difference in energy between the combined siloxane ring and the diffusing molecule system and the independent siloxane rings and the permeating molecules.

RESULTS AND DISCUSSION A six-membered siloxane ring that is composed of at least five Si atoms and one guest atom (X ¼ Al, B, Ti, Y, Zr) was used as a simplified model of the composite silica-based membranes. This model was based on an earlier study in which n-membered siloxane rings were used to describe the pure silica membranes [16]. It was proposed that the planar siloxane rings form a network of solubility sites through which the diffusing species pass. The structures of the geometrically optimized buckled siloxane-based rings are shown in Figure 4.2. The rings are terminated with hydrogen, as this had been demonstrated to give accurate results in modeling silicate structures [29–31]. The siloxane rings with the added elements had a minimum energy in a buckled conformation as a result of the sp3 hybridization of the composing Si orbitals. A striking observation of the geometry optimization is that the pure siloxane rings have two distinct low-energy forms: one, a local minimum, which has a planar starlike shape with an opening of 0.52 nm; another local minimum (possibly a global minimum), which has a buckled structure with a reduced opening of 0.41 nm. The only similarity between the planar and the buckled rings is their symmetry around the axis through opposite Si atoms. The siloxane-based rings with guest atoms B, Al, Si, and Ti have a boat conformation in their structures, while the siloxane-based rings with Y has a

Chapter

4

83

Ab Initio Studies of Silica-Based Membranes

0.35 nm

0.49 nm

Alumina

Boria

0.47 nm 0.53 nm

Titania

Silica

0.59 nm

0.53 nm

Zirconia

Yttria

FIGURE 4.2 Ball-and-stick models of the geometry-optimized six-membered buckled siloxanebased rings (one-coordinate atoms, H; two-coordinate atoms, O; three-coordinate atoms, B, Al; four-coordinate atoms, Si, Ti, Zr, Y).

twist-boat conformation. The siloxane-based ring with Zr is the only one which has a boat conformation with bilateral symmetry. The energies and bond lengths of the siloxane-based rings are summarized in Table 4.1. The first column identifies the composition. The second column reports the lengths of the guest atom–oxygen bond, while the third column reports the lengths of the silicon–oxygen bond. The Si–O bond lengths of the siloxane rings are in the range of 0.163–0.164 nm. The B–O bond length is much shorter at 0.135 nm than a typical Si–O bond length in a siloxane ring. The bond length of other inorganic oxides follow the order Al < Ti < Zr < Y. An increase in the length of the remaining Si–O bonds (0.164–0.169 nm) that form the siloxane-based ring is also observed with the presence of the guest inorganic element. The fourth column reports the formation enthalpy of the siloxane-based rings. The formation enthalpy gradually increased across the periodic table: B < Al < Si < Ti with a basis set of 6-311G. Gaussian requires the usage of the LANL2DZ basis set for the optimization of Y or Zr containing siloxane-based rings that resulted in much lower values for the formation enthalpy. LANL2DZ is a double-zeta basis set containing an ECP representation of electrons near the nuclei for post-third-row atoms. The formation enthalpies were in the order Y < Zr.

84

Inorganic, Polymeric and Composite Membranes

TABLE 4.1 Bond Length and Total Energy of Siloxane-Based Rings Silica Ring Clusters (Inorganic Oxide, X)

X–O Bond Length (nm)

Si–O Bond Length (nm)

Formation Enthalpy (Hartree)

H10Si5O6X, X ¼ B

0.135

0.163–0.165

1931.1557

H12Si5O6X, X ¼ Si

0.164

0.163–0.164

2196.4179

H10Si5O6X, X ¼ Al

0.170

0.162–0.165

2148.7796

H10Si5O6X, X ¼ Ti

0.178

0.163–0.164

2756.3438

H10Si5O6X, X ¼ Zr

0.194

0.165–0.169

524.99188

H10Si5O6X, X ¼ Y

0.203

0.165–0.168

516.29212

TABLE 4.2 Bond Angles of Siloxane-Based Rings Silica Ring Clusters (Inorganic Oxide, X)

O–X–O Bond Angle (deg)

X–O–Si Bond Angle (deg)

O–Si–O Si–O–Si Bond Angle Bond Angle (deg) (deg)

H9Si5O6X, X ¼ B

120.0

131.7–132.8

109.7–111.1 142.3–145.5

H12Si5O6X, X ¼ Si

110.01

142.8–145.3

110.0–111.0 142.0–145.3

H9Si5O6X, X ¼ Al

119.16

147.4–153.3

110.1–114.3 139.6–148.0

H10Si5O6X, X ¼ Ti

117.5

168.3–175.13

108.8–111.1 140.4–143.1

H10Si5O6X, X ¼ Zr

117.6

177.5–177.8

110.7–111.2 172.5–174.6

H10Si5O6X, X ¼ Y

114.7

171.8–173.1

111.4–112.0 170.7–176.7

The bond angles are summarized in Table 4.2. The first column indicates the composition. The second and third columns report the O–X–O and X–O–Si bond angles, while the last two columns report the O–Si–O and Si–O–Si bond angles in the siloxane-based rings. The difference in the bond angles of O–X–O is in the range of 10 , while the corresponding Si–O bond angles of the siloxane rings are quite similar, which shows that they are not affected significantly by the introduction of the inorganic oxides. The difference in the bond lengths of X–O–Si is much larger. The X–O–Si bond angles of guest atoms B and Al show similar values to the corresponding Si–O–Si bond angles. The X–O–Si bond angle of Ti is quite large compared to the Si–O–Si angle. However, the X–O–Si bond angles of the guest atoms Y and Zr are the largest, which indicates that the addition of these atoms changes the structure of the siloxane rings completely.

Chapter

4

(a)

(b) 1200

Activation energy (kJ mol−1)

Buckled B six-membered ring 1000

CH4 CO2 CO Ne H2 He

800 600 400 200 0 −1

0

1

2

3

4

5

6

7

Distance (10 nm)

(c) 1200 1000

CH4 CO2 CO Ne H2 He

600 400 200 0 −1

0

1

2

3

4

5

Distance (10 nm)

6

7

8

CH4 CO2 CO Ne H2 He

800 600 400 200 0 −1

(d)

Buckled Si six-membered rings

800

Buckled AI six-membered ring 1000

8

Activation energy (kJ mol−1)

Activation energy (kJ mol−1)

1200

Activation energy (kJ mol−1)

85

Ab Initio Studies of Silica-Based Membranes

0

1

2

3

4

5

6

7

8

Distance (10 nm) 1200 Buckled Ti six-membered rings 1000

CH4 CO2 CO Ne H2 He

800 600 400 200 0 −1

0

1

2

3

4

5

6

7

8

Distance (10 nm)

FIGURE 4.3 Evolution of activation energy for diffusion of various gases as a function of distance between diffusing species and the center of the siloxane-based rings.

Figure 4.3 presents the activation energies of various gases through the siloxane-based rings. In Figure 4.3a–d, the interaction energy for the diffusion of He, H2, Ne, CH4, CO, and CO2 through the siloxane-based rings with guest atoms, B, Al, Si, and Ti, are presented as a function of the distance from the diffusing atom (C is chosen in the case of diffusion of CO, CO2, and CH4) orthogonally to the center of the siloxane-based rings. The activation energies of permeation through the siloxane rings with boria, alumina, silica, and titania increased from 60 to 930 kJ mol 1 with the size of the diffusing species. Much higher energies are obtained for the siloxane rings with B, Al, and Ti than the pure silica ring. The kinetic diameters of the diffusing species are presented in Figure 4.4. The results show that the interaction energy increases as the diffusing molecules approach the center of the siloxane-based rings. The center of the rings is defined as the center of mass of the overall structure. For all diffusing species, but especially the larger ones, the interaction energy starts to increase at a distance of 0.2 nm and reaches a peak value between 0.1 and 0.2 nm, before the geometric center of the ring. This is because of the interaction between the diffusing species and the terminal hydrogen atoms. As mentioned earlier, these siloxane rings were terminated with hydrogen to save calculation time. However, the introduction of guest atoms into the silica structure distorted

86

Inorganic, Polymeric and Composite Membranes

He

Ne

H2

0.260 0.275 0.289

CO2

CO

CH4

0.330

0.376

0.380

FIGURE 4.4 Kinetic Diameters of Diffusing Species (nm).

1200

1200

Buckled Y six-membered ring

1000

Activation energy (kJ mol−1)

−1

Activation energy (kJ mol )

Buckled Zr six-membered rings CH4 CO2 CO Ne H2

800 600 400 200 0

1000 CH4 CO2 CO Ne H2

800 600 400 200 0

−1

0

1

2

3

4

5

Distance (10 nm)

6

7

8

−1

0

1

2

3

4

5

6

7

8

Distance (10 nm)

FIGURE 4.5 Evolution of activation energy for diffusion of various gases as a function of distance between diffusing species and the center of the siloxane-based rings.

the planar structure of the pure silica rings and formed a buckled structure. The structure of the silica composite membranes may be viewed as a network of the siloxane-based rings of different sizes with no termination hydrogen molecules. Although the evolution curves are thus only estimates of the activation energies, they are still useful tools to compare the differences in activation energies of permeation between various gases. Figure 4.5 presents the evolution of activation energies of various gases through siloxane-based rings with zirconia and yttria. The activation energies of permeation through the siloxane-based rings are estimated from 20 to 400 kJ mol 1 with the increasing size of the diffusing species. These values are much lower than the activation energies of permeation through the siloxane rings with B, Al, Si, and Ti. For all gases, the interaction energy also starts to increase at a distance of 0.2 nm and reaches a peak value between 0.1 and 0.2 nm, but the peaks are not as definitive as before because of the much larger opening and the more symmetrical structure. However, the evolution curves follow a similar trend for all diffusing species with much closer values, which indicates that the siloxane rings with zirconia and yttria might have high permeances with lower selectivities over larger diffusing species. Table 4.3 reports the distances between Si and the guest atoms on the opposite side and between O atoms in the opposite sides of the rings to determine the

Chapter

4

87

Ab Initio Studies of Silica-Based Membranes

TABLE 4.3 Distances in Buckled Siloxane Rings and Calculated Activation Energies of Permeation X–Si (Opposite Sides) Length (nm)

O–O (Opposite Sides) Length (nm) He

H10Si5O6X, X¼B

0.490

0.530

H12Si5O6X, X ¼ Si

0.530

H10Si5O6X, X ¼ Al

Silica Ring Clusters (Inorganic Oxide, X)

CH4

CO2

CO

61.5 102.2 128.1

899.9

527.3

405.1

0.414

104.1 118.6 151.3

700.9

562.2

393.2

0.350

0.406

282.5 392.3 403.7

929.2

690.4

569.7

H10Si5O6X, X ¼ Ti

0.474

0.599

144.8 182.0 360.2

886.6

795.8

573.1

H10Si5O6X, X ¼ Zr

0.538

0.580

36.5

43.5

381.1

240.5

163.0

H10Si5O6X, X¼Y

0.593

0.571

34.1

36.4

333.0

143.5

119.4

H2

Ne

critical opening of each siloxane-based ring. The siloxane ring with the guest atom Al has the smallest critical opening, and it is followed by the guest atoms B, Si, and Ti. The openings of the siloxane-based rings with guest atoms Zr and Y are significantly larger when compared to the other siloxane rings. The activation energies of permeation for He, H2, Ne, CO, CO2, and CH4 are also tabulated in the following columns. The smaller-sized gases, He, H2, Ne, which had shown good permeance through the silica-based membranes [2,6] have substantially smaller activation energies than the other gases. A striking point is that an increase of 0.1 nm in the cross section of the siloxane rings results in a one order of magnitude decrease in the activation energies. The activation energies of permeation for CO, CO2, and CH4 through all types of siloxane rings are significantly higher because these diffusing species have larger kinetic diameters (> 0.3 nm) compared to He, H2, and Ne. The siloxane ring with B shows the lowest activation energy for permeation and is followed by the activation energy of permeation through the rings with Si, Ti, and then Al. As mentioned earlier, the LANL2DZ basis set was used for the post-third-row elements but cannot be used with He. For this reason, the activation energies of He through the siloxane rings for Zr and Y were not obtained. The activation energies of permeation for H2 and Ne have slightly higher values when compared to the values for He for all types of siloxane rings.

88

Inorganic, Polymeric and Composite Membranes

The highest activation energies for CH4 permeation are obtained with the guest atoms Al, B, and Ti, and they are followed closely by Si. In comparison, the activation energies of siloxane rings with guest atoms Zr and Y were only about half of these. Similar trends are observed for the activation energies of permeation for CO and CO2. These observations indicate that the general trend of activation energies through various siloxane-based rings is strongly dependent on the differences between the cross-sectional size of the opening and the size of the various diffusing species. Table 4.4 shows the experimental permeation performances at 873 K with a DP of 1 bar of the various silica-based membranes prepared by high-temperature CVD. The second column reports the H2 permeance values in mol m 2 s 1 Pa 1, while the third and fourth columns show the selectivity of H2/CH4 and H2/CO2, which is obtained by the ratio of the permeances of each gas. The H2 permeance of the pure silica membrane has the highest value with significantly higher H2 selectivities for both CH4 and CO2. The silica–alumina membrane has a slightly lower H2 permeance with appreciable H2/CH4 and H2/CO2 selectivities of 900 and 600. The H2 permeance of the silica–boron composite membrane is reduced to half, but similar selectivities for both CH4 and CO2 are obtained. According to the optimization calculations, the rings with guest atoms, Si, Al, and B, have higher values of activation energies for CH4, CO2, and CO, which suggests that they might have very low permeances toward these gas species. A significant difference in activation energies between H2 and CH4 or CO2 indicates that these rings might have high H2/CH4 and H2/CO2 selectivities. These experimental findings support the predictions of the activation energies of permeation through the siloxane-based rings with guest atoms Si, Al, and B. Even though similar trends are observed for the siloxane rings with Ti, the silica–titania membranes had very low selectivities compared to the silica-based membranes. This suggests that further studies should be performed on these membranes to optimize the conditions of the membrane preparation. The silica–zirconia membranes show a moderate H2 permeance with lower H2/CH4 and H2/CO2 selectivities, which also fits the predictions of the activation energy for permeation. The siloxane ring with Zr has much lower activation energies of permeation for both small and larger gas species; thus it does not offer a high propensity for the separation of these molecules. TABLE 4.4 Permeation Performances of Various Silica-Based Membranes Membranes

H2 Permeance (mol m 2 s 1 Pa 1)

Selectivity H2/CH4

Selectivity H2/CO2

Silica–boria

5.8
10 8

Silica Silica–alumina Silica–titania Silica–zirconia

700

2700

7

5900

1500

7

900

600

7

16

24

7

50

50

5.0
10

1.6
10 1.4
10

2.1
10

Chapter

4

Ab Initio Studies of Silica-Based Membranes

89

CONCLUSIONS The equilibrium geometries of buckled siloxane-based rings with guest atoms, Si, B, Al, Ti, Ge, Zr, and Y, were optimized by using a DFT with high-level basis sets. In addition, the activation energies of permeation for He, H2, Ne, CH4, CO2, and CO were obtained by using the difference in the formation energies of the optimized siloxane-based rings and the rings with the diffusing species. The activation energies of the buckled silica–yttria and silica–zirconia rings (20–40 kJ mol 1) were found to be much lower than the activation energies for the buckled silica–alumina, silica–boron, and silica–titania rings (60–100 kJ mol 1). However, the difference between the activation energies of smaller and larger gases was considerably higher for the siloxane rings with silica, alumina, boron, and titania when compared to the rings with zirconia and yttria. Experimental studies performed on the gas permeation performance of the pure silica, silica–alumina, silica–boria, and silica–zirconia membranes were in agreement with the predictions of activation energies of permeation for the various gases.

ACKNOWLEDGMENTS For supporting this work, the author acknowledges the Director, Division of Chemical, Bioengineering, Environmental, and Transport Systems (CBET), National Science Foundation (grant CBET-084316); the National Energy Technology Laboratory (NETL-RUA program grant number 5.681.884.001); the Mombukagakusho Kakenhi (grant-in-aid Kiban kenkyu B 22-360,335); and the International Collaborative Program of the Japan Science and Technology Agency.

REFERENCES [1] Y. Gu, S.T. Oyama, High molecular permeance in a pore-less ceramic membrane, Adv. Mater. 19 (2007) 1636. [2] D. Lee, S.T. Oyama, Gas permeation characteristics of a hydrogen selective supported silica membrane, J. Membr. Sci. 210 (2002) 291. [3] D. Lee, L. Zhang, S.T. Oyama, S. Niu, R.F. Saraf, Synthesis, characterization and gas permeation properties of a hydrogen permeable silica membrane supported on porous alumina, J. Membr. Sci. 231 (2004) 117. [4] T. Okubo, H. Inoue, Introduction of specific gas selectivity to porous glass membranes by treatment with tetraethoxysilane, J. Membr. Sci. 42 (1989) 109–117. [5] G.R. Gavalaas, C.E. Megiris, S.W. Nam, Deposition of H2-permselective SiO2 films, Chem. Eng. Sci. 44 (1989) 1829–1835. [6] Y. Gu, P. Hacarlioglu, S.T. Oyama, Hydrothermally-stable silica-alumina composite membranes for hydrogen separation, J. Membr. Sci. 310 (2008) 28. [7] G.P. Fotou, Y.S. Lin, S.E. Pratsinis, Hydrothermal stability of pure and modified microporous silica membranes, J. Mater. Sci. 30 (1995) 2803–2808. [8] K. Yoshida, Y. Hirano, H. Fujii, T. Tsuru, M. Asaeda, Hydrothermal stability and performance of silica-zirconia membranes for hydrogen separation in hydrothermal conditions, J. Chem. Eng. Jpn. 34 (2001) 523–530. [9] J.H.A. Hekkink, R.S.A. De Lange, A.A. Ten Hoeve, P.J.A.M. Blankenvoorde, K. Keizer, A.J. Burggraaf, Characterization and permeation properties of binary SiO2-TiO2 and SiO2Al2O3 modified gamma-alumina membranes, Key Eng. Mater. 61&62 (1991) 375–378.

90

Inorganic, Polymeric and Composite Membranes

[10] Y. Gu, S.T. Oyama, Permeation properties and hydrothermal stability of silica–titania membranes supported on porous alumina substrates, J. Membr. Sci. 345 (2009) 267–275. [11] M. Knudsen, The law of the molecular flow and viscosity of gases moving through tubes, Ann. Phys. 28 (1909) 75. [12] S.-T. Hwang, K. Kammermeyer, Surface diffusion in microporous media, Can. J. Chem. Eng. 44 (1966) 82. [13] J. Xiao, J. Wei, Diffusion mechanism of hydrocarbons in zeolites-I. Theory, Chem. Eng. Sci. 47 (5) (1992) 1123. [14] A.B. Shelekhin, A.G. Dixon, Y.H. Ma, Theory of gas diffusion and permeation in inorganic molecular-sieve membranes, AlChE J. 41 (1995) 58. [15] A.J. Burggraaf, Single gas permeation of thin zeolite (MFI) membranes: theory and analysis of experimental observations, J. Membr. Sci. 155 (1999) 45. [16] S.T. Oyama, D. Lee, P. Hacarlioglu, R.F. Saraf, Theory of hydrogen permeability in nonporous silica membranes, J. Membr. Sci. 244 (2004) 45–53. [17] P. Hacarlioglu, D. Lee, J.G. Gibbs, S.T. Oyama, Activation energies for permeation of hydrogen and helium through silica membranes: an ab initio calculation study, J. Membr. Sci. 313 (2008) 278–283. [18] R.M. Barrer, D.E.W. Vaughan, Solution and diffusion of helium and neon in tridymite and cristobalite, Trans. Faraday Soc. 63 (1967) 2275. [19] D.W. Breck, Zeolite Molecular Sieves: Structure, Chemistry and Use, John Wiley & Sons, New York, 1974, p. 636. [20] R.W. Lee, R.C. Frank, D.E. Swets, Diffusion of hydrogen and deuterium in fused quartz, J. Chem. Phys. 36 (1962) 1062. [21] R.W. Lee, Diffusion of hydrogen in natural and synthetic fused quartz, J. Chem. Phys. 38 (1963) 448. [22] J.E. Shelby, Helium migration in natural and synthetic vitreous silica, J. Am. Ceram. Soc. 55 (1972) 61. [23] J.E. Shelby, Temperature dependence of He diffusion in Vitreous SiO2, J. Am. Ceram. Soc. 54 (1972) 125. [24] R.W. Lee, R.C. Frank, D.E. Swets, Diffusion coefficients of helium in fused quartz, J. Chem. Phys. 34 (1961) 17. [25] J.E. Shelby, Neon migration in TiO2-SiO2 glasses, J. Am. Ceram. Soc. 56 (6) (1973) 340. [26] W.G. Perkins, D.R. Begeal, Diffusion and permeation of He, Ne, Ar, Kr, and D2 through silicon oxide thin films, J. Chem. Phys. 54 (1971) 1683. [27] A.D. Becke, Density functional thermochemistry III The role of exact exchange, Chem. Phys. Lett. 157 (1989) 200. [28] C. Lee, W. Yang, R.G. Parr, Development of the Colle-Salvetti correlation energy formula into a functional of the electron density, Phys. Rev. B 37 (1988) 785. [29] G.V. Gibbs, M.B. Boisen Jr., L.L. Beverly, K.M. Rosso, A computational quantum chemical study of the bonded interactions in earth materials and structurally and chemically related molecules, in: R.T. Cygan, J.D. Kubicki (Eds.), Molecular Modeling Theory: Applications in the Geosciences, Rev. Mineral. Geochem. vol. 42, 2001, p. 345. [30] G.V. Gibbs, M.B. Boisen, A molecular modeling of bonded interactions in crystalline silica, in: Z. Rappoport, Y. Apeloig (Eds.), The Chemistry of Organic Silicon Compounds, Wiley, Chichester, 1998, p. 103. [31] K.M. Rosso, G.V. Gibbs, M.B. Boisen, Si-O bonded interactions in coesite: a comparison of crystalline, molecular, and experimental electron density distributions, Phys. Chem. Miner. 26 (1999) 264.

Chapter 5

Review of CO2/CH4 Separation Membranes Dmitri D. Iarikov1 and S. Ted Oyama1,2,*

Environmental Catalysis and Nanomaterials Laboratory, Department of Chemical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA 2 Department of Chemical Systems Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan * Corresponding author: E-mail addresses: [email protected], [email protected] 1

INTRODUCTION Membrane technology has a promising future in the refining, petrochemical, and natural gas industries [1]. One area of importance is the separation of CH4 and CO2. There are over 20 trillion standard cubic feet of natural gas produced in the US annually, and approximately 20% of that gas requires significant treatment [2]. Separation of CO2 is an important industrial process because in some natural gas wells CO2 can comprise up to 70% of the total gas volume [3]. Current US pipeline specifications dictate that CO2 content be under 2% [2]. At higher concentrations, it contributes to pipeline corrosion and decreases the calorific value of natural gas. Traditional methods for CO2 capture are based on reversible absorption, such as amine scrubbing, but these processes are energy intensive and pose environmental concerns. The advantages of membrane gas separation over traditional techniques include low-energy consumption, ease of operation, and low environmental impact [4]. However, the flux and the selectivity of current commercial polymeric membranes are too low to process large volumes of gas [5], and they are only able to compete with small-scale amine plants operating at less than 30 million of standard cubic feet per day [2]. Other limitations associated with current membrane systems include insufficient thermal and chemical stability and susceptibility to plasticization. Any improvement in membrane system efficiency could lead to considerable financial savings [6] and further enhance the role of membrane systems in gas separation. Despite decades of research on the subject, today there are less than 10 different types of polymers used for commercial gas separation, and none of those polymers were designed specifically for gas Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

91

92

Inorganic, Polymeric and Composite Membranes

separation. In order to be commercially competitive, new membrane materials must offer significant improvements in CO2 permeance and CO2/CH4 selectivity compared to existing systems. Further, membranes must have excellent thermal and chemical stability, resistance to plasticization (for polymeric membranes), resistance to aging, low cost, and ease of scale-up [7]. Membrane systems reviewed in this chapter can be separated into three categories: inorganic, polymeric, and composite. Inorganic membranes are made up of inorganic materials and include zeolite membranes, carbon molecular sieves, and amorphous silica membranes. Polymeric membranes include membranes composed of polymers and polymer blends. Composite membranes include organic–inorganic or mixed-matrix membranes that consist of an inorganic phase integrated into a continuous polymer matrix and also supported ionic liquid membranes (SILMs) that consist of room temperature ionic liquids (RTILs) impregnated into porous polymeric or inorganic supports. Different membrane types will be described later with the purpose of highlighting structural properties, membrane morphologies, gas transport mechanisms, and performance differences for CO2/CH4 separation. Membrane separation performance is evaluated based on gas permeance and selectivity. Membrane permeance is expressed as the amount of gas (in moles) that goes through a membrane of known area per unit time per cross-membrane pressure. Selectivity is defined here as the ratio of single-gas permeances of any two species permeating through the membrane and is used as a measure of separation efficiency. Unfortunately, multiple-unit conventions are used to describe membrane permeance in the literature. In this chapter, permeance results are reported in both SI units and in Barrers. Barrers were converted into SI units if the membrane thickness was provided by the authors. For systems where a range of thicknesses was reported, an average thickness was used. Another popular gas permeance unit is the GPU (gas permeation unit) that can be converted directly into SI units. Permeance values quoted in this chapter were reported at or around room temperature although some studies reported permeances at other temperatures. 1 Barrer ¼ 1  1010

cm3 (STPÞcm mol ¼ 3:35  1010 2 ; 2 cm s cm Hg m s Pa

for a 1 mm thickness. 1 GPU ¼ 1  106

cm3 (STPÞ mol ¼ 3:35  1010 2 : 2 cm s cm Hg m s Pa

There has been much work done in the field of CO2/CH4 gas separation, but extensive reviews on this subject are rare [4]. This chapter provides a comprehensive scientific assessment of the latest literature results with different types of gas separation membranes.

Chapter

5

Review of CO2/CH4 Separation Membranes

93

DISCUSSION Zeolite Membranes and Carbon Molecular Sieves Zeolites are crystalline aluminosilicates with a well-defined repeating pore structure. Zeolite membranes generally consist of a thin zeolite layer several micrometers thick deposited on a support such as porous a-alumina or stainless steel. These membranes offer an advantage over traditional polymeric gas separation membranes due to their excellent thermal and mechanical stabilities and chemical resistance [8]. Gas separation in zeolite membranes is achieved by molecular sieving accompanied by surface diffusion. The channels and the cages that make up the zeolite pore structure have the ability to separate gases based on their kinetic diameters. In the case of CO2, separation is enhanced at lower temperatures due to preferential adsorption [9]. The zeolite membranes surveyed in this review displayed excellent results for both permeance and selectivity. Carbon membranes are porous random networks that are typically produced by pyrolysis of thermosetting polymers under controlled conditions and possess good thermal and chemical stability [9]. Pyrolysis results in membranes with narrow pore size distributions with pores of molecular dimensions (molecular sieves). The pore sizes in carbon fibers used for gas separation vary from 0.35 to 1 nm depending on preparation conditions [5]. The pore structure can be described as consisting of relatively large openings with narrow “necks” that are similar in size to gas molecules. The principal gas transport mechanism is molecular sieving where molecules of smaller dimensions permeate preferentially in comparison to molecules of larger dimensions [10]. Permeation of larger molecules is severely restricted sterically, while permeation of molecules smaller than a certain size occurs much faster. The interactions of gas molecules with the walls of carbon molecular sieve consist of both repulsive and attractive forces. Gas molecules are required to overcome an activation energy barrier originating from the repulsive forces in order to pass through the pore openings. Large changes in permeance caused by small changes in molecular size can thus be explained by activated diffusion through the membrane pores. The second transport mechanism in carbon membranes is attributed to surface diffusion with selective adsorption. Carbon membranes that rely on the selective surface diffusion mechanism generally have larger pore dimensions and are able to separate strongly adsorbable gases (NH3, CO2, H2S) from weakly adsorbable species (CH4, O2, N2) [9]. Carbon membranes can either be supported (on tubular or flat supports) or unsupported. Unsupported (flat or tubular) carbon membranes are brittle and difficult to handle and, moreover, are thicker than supported membranes and thus display lower gas permeance. Inorganic membranes showed the best potential for CO2/CH4 separation based on permeance and selectivity (Table 5.1). Zeolite membranes supported on either a-alumina or stainless steel displayed the highest values for both CO2

94

Inorganic, Polymeric and Composite Membranes

TABLE 5.1 Zeolite and Carbon Membranes for CO2/CH4 Separation CO2 Permeance (Barrer)

CO2 Permeance (mol m 2 s 1 Pa 1)

a(CO2/ CH4)

Reference

Highly 2200 hydrophobic DDRtype zeolite membrane on a-alumina support

2.9  10 7

420

[8]

SAPO-34 zeolite membrane on aalumina porous support

20,000–40,000

1.0  10 6–2.0  10 6

86–171

[11]

DD3R zeolite membrane on aalumina disk support

1000

7.0  10 8

600

[12]

SAPO membranes synthesized by in situ crystallization onto porous, tubular stainless steel support

930–3000

6.2  10 8–2.0  10 7

32–95

[13]

3.9  10 8–5.9  10 8

36–120

[14]

6.5  10 8

400

[15]

4.2  10 9

150

[16]

Membrane Type

Zeolite molecular sieve membranes

SAPO-34 ionexchanged membranes synthesized by in situ crystallization on the inside of a stainless steel tubular support DDR membrane synthesized on porous alumina support (disk)

970

Carbon molecular sieve membranes Dual-layer hollow fiber membrane with closely packed beta zeolite nanoparticles

Continued

Chapter

5

95

Review of CO2/CH4 Separation Membranes

TABLE 5.1 Zeolite and Carbon Membranes for CO2/CH4 Separation— Cont’d

Membrane Type Carbon molecular sieve membrane derived from poly (phenylene oxide) substituted with trimethylsilyl

CO2 Permeance (Barrer) 218–529

CO2 Permeance (mol m 2 s 1 Pa 1) 9

2.8  10

9

–7.5  10

a(CO2/ CH4)

Reference

100–130 [17]

NaA zeolite 1000 incorporated into carbon nanocomposite thin film on porous a-alumina support

3.4  10 7

28

[18]

Zeolite L/carbon 600 nanocomposite membranes obtained on porous alumina tubes by incorporating nanosized zeolite L into PFA

5.7  10 8

36

[19]

Silver-doped ionexchanged SPAEK with different counterions used as polymeric precursors to fabricate carbon membranes

6.4  10 10

67

[20]

50–200

[21]

96

Carbon molecular 40–300 sieve membranes derived from blends of PBI with various polyimides

permeance and CO2/CH4 selectivity. The groups of Falconer and Noble were able to synthesize a SAPO-34 zeolite membrane on a-alumina porous supports with CO2 permeances of 1.0–2.0  10 6 mol m 2 s 1 Pa 1 (20,000–40,000 Barrer) and CO2/CH4 selectivity of 86–171 [11]. In two other reports from the same researchers, SAPO membranes with CO2 permeances of

96

Inorganic, Polymeric and Composite Membranes

3.9  10 8–2.0  10 7 mol m 2 s 1 Pa 1 (930–3000 Barrer) and CO2/CH4 selectivity of 32–118 were described [13,14]. These membranes were prepared by in situ crystallization on porous, tubular stainless steel supports. It is apparent that using a-alumina supported zeolites results in a much higher permeance than with the stainless steel supports, which can be explained by large pores and an asymmetric structure of a-alumina supports. In another work, Himeno et al. described a highly hydrophobic DDR-type zeolite membrane on a-alumina support with CO2 permeance of 2.9  10 7 mol m 2 s 1 Pa 1 (2200 Barrer) with CO2/CH4 selectivity of over 400 at room temperature [8]. In two separate articles by Van der Bergh et al. [12,15], a DDR zeolite membrane on a-alumina disk support was described with CO2 permeance on the order of 7  10 8 (1000 Barrer) and CO2/CH4 selectivity of over 400. It appears that DDR membranes give a better CO2/CH4 selectivity than SAPO membranes but show a significantly lower gas permeance. Overall, zeolite membranes displayed drastic improvements in gas separation properties compared to current commercial polymeric membranes. This can be attributed to the inherent zeolite gas transport properties. Since zeolites have a well-defined porous network, high gas diffusivity can be achieved if the membrane separation layer is thin. In zeolite membranes, gases are separated based on size exclusion and selective adsorption that give superior selectivity compared to polymeric membranes that follow the solution–diffusion mechanism. Zeolites do not exhibit plasticization and, therefore, excellent CO2/CH4 selectivity can be achieved even at high pressures. Unlike polymeric membranes, zeolite membranes are able to withstand relatively high temperatures in the presence of oxygen [2]. It is important to note that CO2/CH4 selectivity does decrease with temperature because the selective adsorption of CO2 decreases. Additionally, zeolite membranes are very expensive, difficult to process, and difficult to handle, which offsets some of their attractive separation characteristics. Due to their high cost, zeolite membranes for gas separation are used only on the laboratory scale. Commercial implementation is unlikely until the price of zeolite membranes is reduced by approximately an order of magnitude [2]. Carbon molecular sieve membranes also displayed excellent CO2/CH4 separation properties (Table 5.1). Unsupported carbon membranes synthesized by Yoshimune et al. gave very high CO2/CH4 selectivity of 100–130 but low CO2 permeance (2.8  10 9–7.5  10 9 mol m 2 s 1 Pa 1 or 220–530 Barrer) due to a high membrane thickness (23–25 mm) [17]. In addition to low permeance, unsupported membranes tend to be extremely brittle and difficult to handle as explained earlier. Supported carbon membranes exhibit much better permeance due to a decrease in the active layer thickness that reduces resistance to gas flow. Jiang et al. synthesized a dual-layer hollow fiber membrane on a Matrimid support [16]. The CO2 permeance of this membrane was rather low (4.2  10 9 mol m 2 s 1 Pa 1) due to the collapse of micropores in the support layer, but the CO2/CH4 selectivity was high (150). Zhou et al. manufactured a carbon membrane supported on a-alumina with incorporated

Chapter

5

Review of CO2/CH4 Separation Membranes

97

NaA zeolite nanoparticles [18]. The membrane showed excellent permeance of 3.4  10 7 mol m 2 s 1 Pa 1 (1000 Barrer) with CO2/CH4 selectivity of 28. The membrane selectivity suffered from defects in the thin (1 mm) carbon separation layer supported on a-alumina, but the results were nevertheless promising. In the work of Yin et al., carbon nanocomposite membranes were synthesized on a-alumina tubes by incorporating nanosized zeolite L into a polyfurfuryl alcohol (PFA) matrix [19] using one-step coating and pyrolysis. It was found that the addition of the zeolite into the carbon matrix improved the separation properties of the membranes as demonstrated with CO2 adsorption isotherms. The membranes showed the CO2 permeance of 5.7  10 8 mol m 2 s 1 Pa 1 (600 Barrer) with CO2/CH4 selectivity of 36. In the article by Xiao et al., an ion-exchanged sulfonated poly(aryl ether ketone) (SPAEK) carbon membrane was synthesized using different counterions including silver. The silver doping was found to increase gas permeability dramatically, possibly due to strong interactions between the silver ions and the polar groups within the polymeric matrix as demonstrated by d-spacing analysis with XRD. The resultant membranes displayed CO2 permeance of 6.4  10 10 (96 Barrer) with CO2/CH4 selectivity of 67. In a report by Hosseini and Chung, carbon membranes were derived from blends of poly(benzimidazole) (PBI) with several different polyimides. It was determined that the PBI/Matrimid combination was the best choice for CO2/CH4 separation and displayed the selectivity of 50–200 with CO2 permeance of 40–300 Barrer. The thin active carbon separation layer must be defect-free; otherwise, membrane selectivity decreases significantly. Carbon membranes suffer from many of the same difficulties as zeolite membranes. They are expensive, hard to process, and difficult to handle. Other drawbacks include the vulnerability of carbon membranes to oxidizing agents and water vapor that results in performance deterioration with time [5]. In order to synthesize good carbon membranes, a technology must be in place to first manufacture good polymeric membranes. The polymeric membranes must then be pyrolyzed under very rigorous conditions, which may be difficult to scale up. The cost of materials and the difficulty of processing prevent wide commercialization of these membranes.

Silica Membranes Two types of silica membranes will be described in this review which can be categorized according to their morphology as amorphous or mesoporous membranes. These membranes are generally prepared from silica precursors by either CVD or sol–gel synthesis in the case of amorphous silica or by sol–gel or hydrothermal synthesis in the case of mesoporous silica. The membranes prepared by CVD using standard silica precursors such as tetraethyl orthosilicate (TEOS) or tetramethyl orthosilicate (TMOS) generally result in a microporous or a nonporous silica thin film suitable for H2 or He separation [22].

98

Inorganic, Polymeric and Composite Membranes

These membranes are highly permeable to small gas molecules, but the permeance of larger gas molecules such as CO2 is very low. In order to achieve higher CO2 permeation rates by the CVD route, it is necessary to use silica precursors that contain large organic groups, that is, phenyl groups. Two of the studies that are surveyed here utilized phenyl-containing precursors that created a pore structure that was loosely packed compared to the regular CVD silica [23,24]. The mesoporous membranes that are prepared by the sol–gel techniques have well-defined pore structures and act as molecular sieves. The sol–gel route is a convenient way to generate silica layers; however, the final morphology of the membrane depends greatly on the synthesis parameters. This allows for very precise control of the pore morphology and structure but also makes the final membranes sensitive to synthesis conditions and therefore difficult to reproduce. In general, the gas permeance is higher in membranes prepared by the sol–gel methods compared with CVD membranes, but the film stability and durability is significantly better in the CVD membranes [25]. The past work in the area of silica membranes for CO2 separation has been surveyed and is presented in Table 5.2. There are certain issues associated with the use of amorphous silica membrane synthesis, such as defect formation due to particles in the preparation atmosphere or due to thermal cracking during heat treatment. These problems could potentially affect membrane reproducibility [31]. It is also well known that the silica membranes suffer from hydrothermal instability at elevated temperatures. The gas flux and the selectivity often decrease dramatically in the first few hours of operation when the membranes are exposed to water vapor. The instability is usually explained by the deterioration of pore structure due to the vulnerability of the surface silanol groups to water molecules [32].

Polymeric Membranes Organic polymers are the most widely used materials in membrane gas separation [33]. Polymers can be separated into two general categories—those operating above their glass transition temperature (rubbery polymers) and those operating below their glass transition temperature (glassy polymers). Glassy polymers are able to effectively separate molecules based on small differences in molecular dimensions [34]. They are innately more size and shape selective than rubbery polymers and therefore better suited for CO2 separation [2]. The sorption of gases in rubbery polymers follows Henry’s law; the sorption in glassy polymers can be described by complex sorption isotherms linked to the unrelaxed volume of the matrix when the materials are quenched below their glass transition temperature. Detailed descriptions of the solubility in glassy polymers can be found elsewhere [34]. A summary of the performance of polymer-based membranes is given in Table 5.3. Commercial polymeric membranes generally have an asymmetric structure with a very thin selective layer supported on a thicker porous layer. A thin

Chapter

5

99

Review of CO2/CH4 Separation Membranes

TABLE 5.2 Silica Membranes for CO2/CH4 Separation Membrane type

CO2 Permeance a(CO2/ a(CO2/ Reference (mol m 2 s 1 Pa 1) CH4) N2)

Hybrid silica membranes incorporating phenyl functional groups, prepared via sol–gel and dip coating on a-alumina substrates

6  10 8

Sol–gel organic–inorganic membranes on PAN support prepared by cohydrolysis of organoalkoxide with TMOS

2.34  10 6

Amorphous silica deposited by CVD in the g-alumina film coated on an a-alumina tube, by evacuating TEOS, PTES, or DPDES through the porous wall

8.1  10 8

11

Dual-layer microporous silica membrane prepared by a novel sol–gel dip-coating process

6.7  10 8

240

Mesoporous silica membranes prepared on alumina supports by hydrothermal and sol–gel spin-coating methods.

4.5

6.0

[26]

14

[27]

[23]

60

[28]

1  10 9

800

[29]

Pure, amine-derivatized and nickel-doped sol–gel silica membranes deposited on Membralox-type ceramic supports

1.5  10 7

80

[30]

Hybrid organic–inorganic thin layers of silica incorporating aromatic groups deposited on porous alumina support by CVD of PTES or DPDES in an inert atmosphere at high temperature

5.8  10 8

380

[24]

selective layer allows higher gas fluxes through the membrane, while the thick support layer ensures structural integrity. Recently, there have been more publications on systems that combine (blend) different polymers to produce

TABLE 5.3 Polymeric and Copolymeric Membranes for CO2/CH4 Separation Membrane Type

CO2 Permeance (Barrer)

Film composite of poly(ether-block-amide) on PVDF ultraporous 0.12–1.7 substrate Sulfonated polycarbonate

CO2 Permeance (mol m 2 s 1 Pa 1)  11

4.0  10

 10

–5.7  10

1.5–6.3  11

 10

–4.2  10

a(CO2/ CH4)

Reference

27–52

[35]

27–75

[36]

18–85

[3]

Three different types of polyimide membranes (Matrimid, Kapton, 8.5–63 P84)

5.7  10

Pure Matrimid, blends of Matrimid, cellulose acetate, polyphenyleneoxide (asymmetric hollow fibers)

3.4  10 9–4.0  10 9

28–31

[6]

8.4  10 10–1.0  10 9

93–170

[37]

Matrimid 5218 chemically cross-linked at room temperature by 5.4–120 poly(propylene glycol) block poly(ethylene glycol) block poly (propylene glycol) diamine

2.4  10 11–5.1  10 10

17–36

[38]

Poly(urethane) and poly(vinyl acetate) blends

8.9  10 11

48

[39]

4.1  10 9–1.1  10 8

15–36

[40]

Polycarbonate

Poly(ether-block-amide) with silver tetrafluoroborate on polyvinylidene fluoride ultrafiltration substrate

230–560

26

Fluorinated copolyimides synthesized with various diamine compositions by chemical imidization

40–550

23–57

[41]

Fluorinated polyimide nanocomposite membranes synthesized from 6FDA and ODA with nanoparticulate additives

14–16

60–67

[42]

29–38

[43]

Polysulfone copolymer membranes based on equimolar mixtures 5.2–7.0 of rigid/compact naphthalene moiety with bulky connectors Membranes were prepared with a partially cross-linked Matrimid 8.3 5218 selective top layer deposited on a PTMSP intermediate layer on top of PAN porous support

8.4  10 9

37

[44]

Fluorinated PEI membranes prepared with incorporation of heterocyclic moieties

37–52

1.2  10 10–3.0  10 10

24–28

[45]

Polyimide membrane scaffold with homogeneous pseudointerpenetrating polymer networks

7–315

20–50

[46]

Dense fluoroaniline copolymer membranes

0.6–1.4

40–55

[47]

36–75

[48]

Hyperbranched copolyimides based on ODPA, TAP, and ODA

1–6

2.0  10 12–4.7  10 12  12

8.7  10

 11

–5.4  10

102

Inorganic, Polymeric and Composite Membranes

membranes with improved CO2/CH4 permeances and selectivities. In the case of blended membranes, a glassy polymer is incorporated into a rubbery polymer matrix to combine the high selectivity of the former with the high diffusivity of the latter [39]. Additionally, the glassy polymer provides structural and mechanical support for the rubbery matrix [49]. Polymeric membranes are generally nonporous, and gas permeation is governed by the solution–diffusion mechanism, which, as the name implies, is determined by solubility and diffusivity of gases within the polymer matrix [49]. Gas molecules diffuse from the feed side to the gas-membrane interface where they dissolve and permeate across the membrane by random molecular diffusion, followed by desorption and diffusion into the permeate bulk stream. Diffusion occurs in free-volume elements (0.2 and 0.5 nm in size) between the polymer chains that appear and disappear continuously due to thermal motion. Gas diffusion through the nonporous polymeric structures is inherently slow; thus, polymeric membranes exhibit low CO2 permeance with moderate CO2/ CH4 selectivity. Polymeric membranes currently dominate the gas separation market due to the low cost of materials and ease of processing. Compared to other membrane systems described earlier, polymeric membranes are easier to scale up. The studies surveyed in this chapter demonstrate that polymeric membranes possess poor CO2 permeance with moderate CO2/CH4 selectivity (Table 5.3). Polymeric membrane systems surveyed in this chapter generally exhibited permeance below 1  10 8 mol m 2 s 1 Pa 1 (300 Barrer). Poor gas permeance in polymeric membranes can be understood from the gas diffusion mechanism. Permeance is inherently low because gases must dissolve in and diffuse through nonporous solids. Additionally, there exists a well-known tradeoff between the selectivity and the permeance for polymeric membranes. Despite vigorous research in membrane gas separations in recent years, polymeric membranes have not been able to significantly advance beyond the so-called line of death of the Robeson plot that underlines the permeability–selectivity tradeoff in polymeric membranes [2,50]. Figure 5.2 illustrates the CO2/CH4 separation performance of the polymeric membranes and shows the original 1991 Robeson curve and how it has shifted over the past two decades of membrane research.

Mixed-matrix Membranes Composite organic–inorganic membranes or mixed-matrix membranes consist of inorganic particles incorporated into a continuous polymer matrix. Mixedmatrix membranes attempt to combine the ease of processing of polymeric materials with the excellent gas separation properties of molecular sieves [9]. Gases are transported through both polymeric and inorganic phases. Inorganic particles can act as molecular sieves, disrupt the polymer structure thus increasing the permeance, or serve as a barrier reducing gas permeance [49]. At low-tointermediate loading of the inorganic phase, molecular transport occurs through

Chapter

5

Review of CO2/CH4 Separation Membranes

103

diffusion across the polymer matrix and through the inorganic particles. At very high inorganic loading, transport occurs primarily through the inorganic phase. There are also mixed-matrix membranes that rely upon selective sorption of different gases by the inorganic fillers [2]. A summary of the performance of mixed-matrix membranes is given in Table 5.4. At low loadings of the inorganic phase in mixed-matrix membranes, gas transport primarily occurs through the polymeric phase. This accounts for the low CO2 permeation rates and CO2/CH4 selectivity that is comparable with polymeric membranes. The inorganic phases incorporated into the polymer matrix included zeolites [51–56], silica particles [57,58], metal organic frameworks [60,61], and carbon nanotubes [59]. The CO2 permeances for mixed-matrix membrane systems surveyed in this chapter generally fell below 1  10 8 mol m 2 s 1 Pa 1 (200 Barrer). Overall, the organic–inorganic composite membranes did not show a significant improvement in gas permeance or CO2/CH4 selectivity over polymeric membranes.

Supported ionic Liquid and Polyionic Membranes SILMs are composed of ionic liquids impregnated into polymeric or inorganic supports. RTILs are special compounds that can be characterized as organic salts that are liquid at room temperature. RTILs are nonflammable, thermally stable, and possess no measurable vapor pressure [62]. These properties make RTILs ideal candidates as the separation medium for supported liquid membranes. Traditional supported liquid membranes that use volatile solvents suffer from stability problems under pressurized conditions or under vacuum [63] and are subject to drying out, which disrupts performance. Gas transport in the RTIL membranes can be described by the solution–diffusion mechanism similar to polymeric membranes. Gas molecules on the feed side are absorbed into the liquid phase, diffuse across the membrane, and desorb at the permeate side [64]. In some cases, CO2 can participate in a reversible chemical reaction with a mobile or a fixed carrier component inside the membrane [49]. Separation is then governed by facilitated transport, and CO2 diffuses through the membrane both on its own and as a reacted species. Facilitated membranes thus exhibit higher CO2/CH4 selectivities compared with nonreactive membranes. The SILMs surveyed in this chapter (Table 5.5) displayed permeance comparable to those of the polymeric membranes, generally below 1  10 8 mol m 2 s 1 Pa 1. The CO2/CH4 selectivities of SILMs reported in the literature were in the range of 4–20 [64] and compared poorly with the other membrane types. Hanioka et al. reported high CO2/CH4 selectivity of over 100 using a SILM with ionic liquids containing amine groups that allowed for facilitated transport of CO2 [70]. Their best results were obtained at very low partial pressures of CO2 because, at higher pressures, the carrier became saturated and selectivity decreased significantly. In addition to SILMs, there was a significant amount of work done with polymerizable ionic liquid

TABLE 5.4 Mixed-Matrix Membranes for CO2/CH4 Separation a(CO2/

Membrane Type

CO2 Permeance (Barrer)

CO2 Permeance (mol m 2 s 1 Pa 1)

CH4)

Reference

Zeolites NaA and AgA dispersed in polyethersulfone

1.0–2.7

5.2  10 12–1.4  10 11

32–60

[51]

Organic–inorganic asymmetric hollow fiber, with HSSZ-13 zeolites dispersed in UltemÒ 1000 polyetherimide matrix

3.8  10 9–4.4  10 9

36–40

[52]

Pure polycarbonate (PC), PC/p-nitroaniline (pNA), PC/zeolite 4A, and PC/ 4.0–8.8 pNA/zeolite 4A membranes

2.5  10 11–5.6  10 11

24–53

[53]

Polyethersulfone (PES)–zeolite 4A hollow fiber membrane

1.9–5.6

2.2  10 9–6.8  10 9

15–29

[54]

SSZ-13 zeolite dispersed in 3:2 6FDA-DAM-DABA chemically modified with 1,3-propane diol polyethersulfone–zeolite NaA and polyethersulfone–zeolite AgA

56–87

37–50

[55]

Polyethersulfone–zeolite NaA and polyethersulfone–zeolite AgA

1.7–3.4

30–39

[56]

54–95

[57]

18–29

[58]

End group-modified 6FDA-TAPOB hyperbranched polyimide–silica hybrid with tetramethoxysilane as precursor

7.4–26

Polysulfone-containing embedded nonporous fumed silica nanoparticles 6.3–20

8.8  10 12–1.7  10 11  10

1.2  10

 10

–4.4  10

3.0  10 11–9.4  10 11

Polysulfone and functionalized single walled carbon nanotube mixed matrix

3.9–5.2

16–24

[59]

Microporous MOF (Cu–BPY–HFS) combined with Matrimid polymer to form free-standing films

7.8–15

32–26

[60]

Matrimid continuous phase embedded with MOF-5 nanocrystals

11–20

40–50

[61]

1.1  10 10–1.9  10 10

TABLE 5.5 Supported Ionic Liquid and Polyionic Membranes for CO2/CH4 Separation Membrane Type

CO2 Permeance (Barrer)

CO2 Permeance (mol m 2 s 1 Pa 1)  11

Reference

14–38

[65]

Polymerizable room temperature ionic liquids on porous polymeric support

9.2–32

2.1  10

Polyethylene glycol grafted on ionic polymers

15–40

3.6  10 11–9.6  10 11

22–31

[66]

Gemini room temperature ionic liquids photocross-linked into thin films

4

9.2  10 12

32

[67]

Polymer membranes fabricated from RTIL monomers containing oligo(ethylene glycol) or nitrile-terminated alkyl substituents tethered to imidazolium cations

4–22

9.5  10 12–5.1  10 11

29–37

[68]

Membrane was prepared by solution casting an ionic liquid with a poly(RTIL) on Supor 200 porous polymer support

44

7.4  10 11

39

[69]

Task-specific ionic liquids on porous hydrophilic polytetrafluoroethylene support

800

7.7  10 9

55

[70]

Room temperature ionic liquids photopolymerized with divinylbenzene cross-linker on porous supports

22

5.1  10 11

29

[68]

Imidazolium-based ionic liquids in polysulfone support

330–740

7.3  10 10–1.6  10 9

4.7–6.0

[71]

6.6–10

[62]

3.4–5

[72]

 10

–7.4  10

a(CO2/CH4)

 11

–2.2  10

9

Imidazolium-based ionic liquids in polyethersulfone microfiltration support

360–980

8.0  10

Phosphonium-based ionic liquids supported on glass fiber disk filters

230–690

1.2  10 10–3.5  10 10

Continued

TABLE 5.5 Supported Ionic Liquid and Polyionic Membranes for CO2/CH4 Separation—Cont’d Membrane Type

CO2 Permeance (Barrer)

CO2 Permeance (mol m 2 s 1 Pa 1)  10

Supported ionic liquid membranes on polyvinylidene fluorolide support

94–750

2.1  10

Supported ionic liquid membranes on PES porous disk support

310–740

7.9  10 10–1.9  10 9  10

Fluoroalkyl-functionalized supported room-temperature ionic liquid membranes

210–320

4.9  10

Supported ionic liquid membranes prepared on polyvinylidene fluorolide support

6

5.0  10 11

–1.7  10

9

–7.4  10

 10

a(CO2/CH4)

Reference

5.7–23

[73]

10–27

[74]

13–19

[75]

30

[76]

Chapter

5

Review of CO2/CH4 Separation Membranes

107

membranes where the ionic liquid monomers were cross-linked to form a polymer matrix. Although these membranes could be classified as polymeric, they were listed in the same category as the SILMs (Table 5.5). SILMs are most effective at low pressures because the CO2/CH4 selectivity decreases dramatically with increasing total feed pressure for a binary gas mixture. Although the solubility of pure CO2 in 1-hexyl-3-methylimidazolium bis (trifluoromethylsulfonyl)imide is much higher than pure CH4, the presence of CO2 actually increases the CH4 solubility in the ionic liquid in binary mixtures [77]. Since the separation mechanism at work for SILMs is solution–diffusion, the separation becomes ineffective at relatively mild pressures. Fundamental properties of different ionic liquids reported in the literature were surveyed to evaluate the behavior of CO2/CH4 solubility as a function of temperature. The permeances of both CO2 and CH4 generally increase with temperature; however, the permeance of CH4 increases more rapidly, which results in lower selectivity at elevated temperatures. For example, in 1-butyl-3methylimidazolium tetrafluoroborate [78], the CO2 solubility is much higher than that of other gases. However, it decreases dramatically with increasing temperature compared to CH4. The same trend was observed for 1-butyl-3methylimidazolium hexafluorophosphate in another study [79]. It was also shown that the solubility selectivity of different gases in imidazolium-based ionic liquids decreases with temperature [80]. It was also determined in a separate study for 1-hexyl-3-methylpyridinium bis(trifluoromethylsulfonyl)imide [81] where Henry’s constant for CO2 increased with temperature, but Henry’s constant for CH4 did not show an increase in the temperature range. In the work done on 1-n-butyl-3-methylimidazolium hexafluorophosphate [82], it was shown that the CO2 Henry’s constant increased more than twofold, while Henry’s constant for CH4 actually decreased in that temperature range. Even though the onset of thermal decomposition for some ionic liquids starts well above 400  C [83–86], their separation efficiency decreases at these high temperatures due to the gas solubility behavior.

OVERALL RESULTS Overall results for inorganic, organic, composite, and SILMs are presented in Figure 5.1 using SI units and in Figure 5.2 using the traditional permeability unit of Barrer. The membrane systems reviewed in this chapter were compared based on their CO2 permeance and ideal CO2/CH4 selectivity on a single plot. CO2 permeance spanned six orders of magnitude, and the CO2/CH4 selectivity varied by three orders of magnitude depending on the membrane type. The shaded area indicates the membrane performance region that is the most desirable, with permeance above 1  10 9 mol m 2 s 1 Pa 1 (above 100 Barrer) and the ideal CO2/CH4 selectivity of over 40. This region represents membranes with a superior performance compared to commercially available systems. The plot reveals that zeolite molecular sieves are more promising compared to other

108

Inorganic, Polymeric and Composite Membranes

1000

CO2/CH4(–)

100

10

Carbon Mixed matrix Polymeric RTIL Zeolite Silica

1 10−12

10−11

10−10

10−9

10−8

10−7

10−6

10−5

CO2 permeance (mol m−2 Pa−1 s−1) FIGURE 5.1 CO2/CH4 ideal selectivity plotted versus pure CO2 gas permeance (in mol m 2 Pa 1 s 1) for the membrane systems surveyed in this chapter.

1000

CO2/CH4(–)

100

10

1 10−1

Carbon Mixed matrix Polymeric RTIL Zeolite 100

101 102 103 CO2 permeance (Barrer)

2010

1991

104

105

FIGURE 5.2 CO2/CH4 ideal selectivity plotted against CO2 permeance (in Barrers) for different membrane systems. Note the original 1991 Robeson selectivity–permeance tradeoff line, and how it changed over almost 20 years of membrane research. Inorganic membranes were plotted in the same figure for comparison.

Chapter

5

Review of CO2/CH4 Separation Membranes

109

membrane types for CO2/CH4 separation in terms of their separation performance. Some of these systems exhibited CO2 permeance above 1  10 6 mol m 2 s 1 Pa 1 (above 10,000 Barrer) with CO2/CH4 ideal selectivity exceeding 100. The results achieved with hybrid organic–inorganic silica membranes prepared by thermal decomposition of PTES made it clear that these membranes performed very well compared with the other membrane types. The carbon molecular sieves exhibited CO2 permeance comparable to the best of their polymeric counterparts with very high CO2/CH4 selectivity that placed them well above the Robeson curve (Figure 5.2). In comparison, organic (polymeric) membranes exhibited low CO2 permeances in the range of 1  10 11–1  10 8 mol m 2 s 1 Pa 1 (1–300 Barrer) and CO2/CH4 selectivity of between 15 and 80. Mixed-matrix membranes did not show a significant improvement in CO2 separation when compared with polymeric membranes. The CO2 permeance was generally under 1  10 9 mol m 2 s 1 Pa 1 (200 Barrer) and the CO2/CH4 ideal selectivity was below 100. SILMs and polyionic membranes demonstrated inferior CO2/CH4 separation performance when compared with other membrane separation systems. The CO2 permeance of these membranes was generally better than those of polymeric or mixed-matrix membranes, but the CO2/CH4 selectivity was subpar. The membrane systems surveyed here encompassed advances in CO2/CH4 separation over the past several years (2007–2010). Figure 5.2 shows the original Robeson line and how it has advanced over the past two decades of membrane research.

CONCLUSIONS State-of-the-art membrane systems for CO2/CH4 separation were surveyed in this chapter. Membrane systems fell into three general categories: inorganic (zeolite and carbon molecular sieves and amorphous silica membranes), polymeric, and composite (mixed-matrix and SILMs). Of all the membrane types, zeolites offered the best combination of CO2 separation properties. Zeolite membranes prepared on a-alumina supports resulted in much higher permeance when compared to the zeolite membranes synthesized on stainless steel. This could be attributed to the large pores and asymmetric structure of a-alumina support tubes. It was also evident that the DDR membranes resulted in better CO2/CH4 selectivity than the SAPO membranes but showed a significantly lower gas permeance. Widespread commercialization of zeolite membranes is currently curtailed by their high cost. Polymeric membranes displayed poor CO2 permeance and moderate CO2/CH4 selectivity and suffered from a permeance and selectivity tradeoff where high permeance and high selectivity could not be achieved simultaneously. However, due to the ease of processing and cost considerations, they remain the prevailing choice for industrial gas separation. Mixed-matrix membranes surveyed in this chapter did not offer drastic improvements in separation efficiency when compared to polymeric membranes. Carbon molecular sieve membranes provided superior separation

110

Inorganic, Polymeric and Composite Membranes

qualities when compared with polymer and mixed-matrix membranes but were very brittle and difficult to process. Supported carbon membranes exhibited better permeance due to a decrease in the separation layer thickness when compared to the unsupported carbon membranes and were more structurally robust. Amorphous silica membranes showed excellent CO2/CH4 separation properties with the selectivity of over 200 and the CO2 permeance of almost two orders of magnitude higher than that of the polymeric membranes. Issues with stability in steam-rich environments especially at elevated temperatures need to be resolved before these membranes find widespread application. Supported liquid membranes displayed separation properties comparable to those of the polymeric systems in terms of CO2 permeance, but the CO2/CH4 selectivity was generally lower. These membranes were deemed to be a poor choice for CO2/CH4 separation from natural gas streams because they failed to adequately separate gases at high pressure.

ACKNOWLEDGMENTS For support of this work, the author acknowledges the Director, National Science Foundation, Division of Chemical, Bioengineering, Environmental, and Transport Systems (CBET) under grant CBET-084316, the National Energy Technology Laboratory under the NETL-RUA program grant number 5.681.884.001, and the Mombukagakusho Kakenhi grant-in-aid Kiban kenkyu B 22-360,335.

REFERENCES [1] M. Arruebo, R. Mallada, M.P. Pina, Zeolite membranes: synthesis, characterization, important applications, and recent advances, in: A.K. Pabby, S.S.H. Rizvi, A.M. Sastre (Eds.), Handbook of Membrane Separations: Chemical, Pharmaceutical, Food, and Biotechnological Applications, CRC Press, Boca Raton, FL, 2009, pp. 269–323. [2] R.W. Baker, Membrane Technology and Applications, John Wiley & Sons, Ltd., West Sussex, England, 2004. [3] S. Sridhar, R.S. Veerapur, M.B. Patil, K.B. Gudasi, T.M. Aminabhavi, Matrimid polyimide membranes for the separation of carbon dioxide from methane, J. Appl. Polym. Sci. 106 (2007) 1585–1594. [4] S. Sridhar, B. Smitha, T.M. Aminabhavi, Separation of carbon dioxide from natural gas mixtures through polymeric membranes—a review, Sep. Purif. Rev. 36 (2007) 113–174. [5] M. Hagg, Membranes in gas separation, in: A.K. Pabby, S.S.H. Rizvi, A.M. Sastre (Eds.), Handbook of Membrane Separations: Chemical, Pharmaceutical, Food, and Biotechnological Applications, CRC Press, Boca Raton, FL, 2009, pp. 65–105. [6] T. Visser, N. Masetto, M. Wessling, Materials dependence of mixed gas plasticization behavior in asymmetric membranes, J. Membr. Sci. 306 (2007) 16–28. [7] C.E. Powell, G.G. Qiao, Polymeric CO2/N2 gas separation membranes for the capture of carbon dioxide from power plant flue gases, J. Membr. Sci. 279 (2006) 1–49. [8] S. Himeno, T. Tomita, K. Suzuki, K. Nakayama, K. Yajima, S. Yoshida, Synthesis and permeation properties of a DDR-Type zeolite membrane for separation of CO2/CH4 gaseous mixtures, Ind. Eng. Chem. Res. 45 (2007) 6989–6997.

Chapter

5

Review of CO2/CH4 Separation Membranes

111

[9] D. Shekhawat, D.R. Luebke, H.W. Pennline, A review of carbon dioxide selective membranes, (2003) DOE/NETL-2003/1200. [10] A.F. Ismail, L.I.B. David, A review on the latest development of carbon membranes for gas separation, J. Membr. Sci. 193 (2001) 1–18. [11] M.A. Carreon, S. Li, J.L. Falconer, R.D. Noble, Alumina-supported SAPO-34 membranes for CO2/CH4 separation, J. Am. Chem. Soc. 130 (2008) 5412–5413. [12] J. van den Bergh, W. Zhu, J. Gascon, J.A. Moulijn, F. Kapteijn, Separation and permeation characteristics of a DD3R zeolite membrane, J. Membr. Sci. 316 (2008) 35–45. [13] S. Li, J.L. Falconer, R.D. Noble, SAPO-34 membranes for CO2/CH4 separations: effect of Si/ Al ratio, Microporous Mesoporous Mater. 110 (2008) 310–317. [14] M. Hong, S. Li, H.F. Funke, J.L. Falconer, R.D. Noble, Ion-exchanged SAPO-34 membranes for light gas separations, Microporous Mesoporous Mater. 106 (2007) 140–146. [15] J. Van Den Bergh, W. Zhu, F. Kapteijn, J.A. Moulijn, K. Yajima, K. Nakayama, et al., Separation of CO2 and CH4 by a DDR membrane, Res. Chem. Intermed. 34 (2008) 467–474. [16] L.Y. Jiang, T.S. Chung, R. Rajagopalan, Dual-layer hollow carbon fiber membranes for gas separation consisting of carbon and mixed matrix layers, Carbon 45 (2007) 166–172. [17] M. Yoshimune, I. Fujiwara, K. Haraya, Carbon molecular sieve membranes derived from trimethylsilyl substituted poly(phenylene oxide) for gas separation, Carbon 45 (2007) 553–560. [18] Z.H. Zhou, J.H. Yang, L.F. Chang, Y. Zhang, W.G. Sun, J.Q. Wang, Novel preparation of NaA/carbon nanocomposite thin films with high permeance for CO2/CH4 separation, Chin. Chem. Lett. 18 (2007) 455–457. [19] X. Yin, J. Wang, N. Chu, J. Yang, J. Lu, Y. Zhang, et al., Zeolite L/carbon nanocomposite membranes on the porous alumina tubes and their gas separation properties, J. Membr. Sci. 348 (2010) 181–189. [20] Y. Xiao, M.L. Chung, T. Chung, M. Toriida, S. Tamai, H. Chen, et al., Asymmetric structure and enhanced gas separation performance induced by in situ growth of silver nanoparticles in carbon membranes, Carbon 48 (2010) 408–416. [21] S.S. Hosseini, T.S. Chung, Carbon membranes from blends of PBI and polyimides for N2/CH4 and CO2/CH4 separation and hydrogen purification, J. Membr. Sci. 328 (2009) 174–185. [22] A. Ayral, A. Julbe, V. Rouessac, S. Roualdes, J. Durand, Microporous silica membrane: basic principles and recent advances, in: R. Mallada, M. Menendez (Eds.), Inorganic Membranes: Synthesis, Characterization, and Applications, Elsevier, Amsterdam, 2008, pp. 33–73. [23] B.-K. Sea, K. Kusakabe, S. Morooka, Pore size control and gas permeation of silica membranes by pyrolysis of phenyl-substituted ethoxysilanes with cross-flow through a porous support wall, J. Membr. Sci. 130 (1997) 41–52. [24] S. T. Oyama, Y. Gu, J. D. Allison, G. C. Gunter, S. A. Scholten, Hybrid organic-inorganic gas separation membranes, US Pat. Appl. 205,500, August 20, 2009. [25] S. Gopalakrishnan, Y. Yoshino, M. Nomura, B.N. Nair, S. Nakao, A hybrid processing method for high performance hydrogen-selective silica membranes, J. Membr. Sci. 297 (2007) 5–9. [26] T. Okui, Y. Saito, T. Okubo, M. Sadakta, Gas permeation of porous organic/inorganic hybrid membranes, J. Sol Gel Sci. Technol. 5 (1995) 127–134. [27] M. Smaı¨hi, T. Jermoumi, J. Marignan, R.D. Noble, Organic-inorganic gas separation membranes: preparation and characterization, J. Membr. Sci. 116 (1996) 211–220. [28] C. Tsai, S. Tam, Y. Lu, C.J. Brinker, Dual-layer asymmetric microporous silica membranes, J. Membr. Sci. 169 (2000) 255–268. [29] Y. Sakamoto, K. Nagata, K. Yogo, K. Yamada, Preparation and CO2 separation properties of amine-modified mesoporous silica membranes, Microporous Mesoporous Mater. 101 (2007) 303–311.

112

Inorganic, Polymeric and Composite Membranes

[30] G. Xomeritakis, C.Y. Tsai, Y.B. Jiang, C.J. Brinker, Tubular ceramic-supported sol-gel silicabased membranes for flue gas carbon dioxide capture and sequestration, J. Membr. Sci. 341 (2009) 30–36. [31] R.M. de Vos, H. Verweij, Improved performance of silica membranes for gas separation, J. Membr. Sci. 143 (1998) 37–51. [32] Q. Wei, F. Wang, Z. Nie, C. Song, Y. Wang, Q. Li, Highly hydrothermally stable microporous silica membranes for hydrogen separation, J. Phys. Chem. B 112 (2008) 9354–9359. [33] V. Abetz, T. Brinkmann, M. Dijkstra, K. Ebert, D. Fritsch, K. Ohlrogge, Developments in membrane research: From material via process design to industrial application, Adv. Eng. Mater. 8 (2006) 328–358. [34] R.R. Zolandz, G.K. Fleming, Theory, in: C.D. Ho, K.K. Sirkar (Eds.), Membrane Handbook, Van Nostrand Reinhold, New York, 1992, pp. 25–53. [35] S. Sridhar, R. Suryamurali, B. Smitha, T.M. Aminabhavi, Development of crosslinked poly (ether-block-amide) membrane for CO2/CH4 separation, Colloids Surf., A 297 (2007) 267–274. [36] S. Sridhar, T.M. Aminabhavi, M. Ramakrishna, Separation of binary mixtures of carbon dioxide and methane through sulfonated polycarbonate membranes, J. Appl. Polym. Sci. 105 (2007) 1749–1756. [37] M. Iqbal, Z. Man, H. Mukhtar, B.K. Dutta, Solvent effect on morphology and CO2/CH4 separation performance of asymmetric polycarbonate membranes, J. Membr. Sci. 318 (2008) 167–175. [38] H. Zhao, Y. Cao, X. Dinga, M. Zhou, J. Liu, Q. Yuan, Poly(ethylene oxide) induced crosslinking modification of Matrimid membranes for selective separation of CO2, J. Membr. Sci. 320 (2008) 179–184. [39] B. Ghalei, M.A. Semsarzadeh, A novel nano structured blend membrane for gas separation, Macromol. Symp. 249–250 (2007) 330–335. [40] S. Sridhar, T.M. Aminabhavi, S.J. Mayor, M. Ramakrishna, Permeation of carbon dioxide and methane gases through novel silver-incorporated thin film composite pebax membranes, Ind. Eng. Chem. Res. 46 (2007) 8144–8151. [41] L. Wang, Y. Cao, M. Zhou, S.J. Zhou, Q. Yuan, Novel copolyimide membranes for gas separation, J. Membr. Sci. 305 (2007) 338–346. [42] M. Rubal, C.W. Wilkins, P.E. Cassidy, C. Lansford, Y. Yamada, Fluorinated polyimide nanocomposites for CO2/CH4 separation, Polym. Adv. Technol. 19 (2008) 1033–1039. [43] C. Camacho-Zuniga, F.A. Ruiz-Trevino, S. Herna´ndez-Lo´pez, M.G. Zolotukhin, F.H.J. Maurer, A. Gonza´lez-Montiel, Aromatic polysulfone copolymers for gas separation membrane applications, J. Membr. Sci. 340 (2009) 221–226. [44] J. Peter, K.-V. Peinemann, Multilayer composite membranes for gas separation based on crosslinked PTMSP gutter layer and partially crosslinked Matrimid 5218 selective layer, J. Membr. Sci. 340 (2009) 62–72. [45] S.K. Sen, B. Dasgupta, S. Banerjee, Effect of introduction of heterocyclic moieties into polymer backbone on gas transport properties of fluorinated poly(ether imide) membranes, J. Membr. Sci. 343 (2009) 97–103. [46] B.T. Low, T.S. Chung, H. Chen, Y. Jean, K.P. Pramoda, Tuning the free volume cavities of polyimide membranes via the construction of pseudo-interpenetrating networks for enhanced gas separation performance, Macromolecules 42 (2009) 7042–7054. [47] K. Huang, C. Shiu, Y. Su, C. Yang, J. Yeh, Y. Wei, et al., Preparation and gas transport properties of dense fluoroaniline copolymer membranes, J. Membr. Sci. 339 (2009) 171–176.

Chapter

5

Review of CO2/CH4 Separation Membranes

113

[48] J. Peter, A. Khalyavina, J. Krı´z, M. Bleha, Synthesis and gas transport properties of ODPA–TAP–ODA hyperbranched polyimides with various comonomer ratios, Eur. Polym. J. 45 (2009) 1716–1727. [49] C.A. Scholes, S.E. Kentish, G.W. Stevens, Carbon dioxide separation through polymeric membrane systems for flue gas applications, Rec. Pat. Chem. Eng. 1 (2008) 52–66. [50] L.M. Robeson, Correlation of separation factor versus permeability for polymeric membranes, J. Membr. Sci. 62 (1991) 165–185. [51] Y. Li, T.S. Chung, Novel Agþ-zeolite/polymer mixed matrix membranes with a high CO2/ CH4 selectivity, AIChE J. 53 (2007) 610–616. [52] S. Husain, W.J. Koros, Mixed matrix hollow fiber membranes made with modified HSSZ-13 zeolite in polyetherimide polymer matrix for gas separation, J. Membr. Sci. 288 (2007) 195–207. [53] D. Sen, H. Kalıpcılar, L. Yilmaz, Development of polycarbonate based zeolite 4A filled mixed matrix gas separation membranes, J. Membr. Sci. 303 (2007) 194–203. [54] A.F. Ismail, T.D. Kusworo, A. Mustafa, Enhanced gas permeation performance of polyethersulfone mixed matrix hollow fiber membranes using novel Dynasylan Ameo silane agent, J. Membr. Sci. 319 (2008) 306–312. [55] A.M.W. Hillock, S.J. Miller, W.J. Koros, Crosslinked mixed matrix membranes for the purification of natural gas: Effects of sieve surface modification, J. Membr. Sci. 314 (2008) 193–199. [56] Y. Li, T.S. Chung, Highly selective sulfonated polyethersulfone (SPES)-based membranes with transition metal counterions for hydrogen recovery and natural gas separation, J. Membr. Sci. 308 (2008) 128–135. [57] T. Suzuki, Y. Yamada, Effect of end group modification on gas transport properties of 6FDATAPOB hyperbranched polyimide–silica hybrid membranes, High Perform. Polym. 19 (2007) 553–564. [58] J. Ahn, W.J. Chung, I. Pinnau, M.D. Guiver, Polysulfone/silica nanoparticle mixed-matrix membranes for gas separation, J. Membr. Sci. 314 (2008) 123–133. [59] S. Kim, L. Chen, J.K. Johnson, E. Marand, Polysulfone and functionalized carbon nanotube mixed matrix membranes for gas separation: theory and experiment, J. Membr. Sci. 294 (2007) 147–158. [60] Y. Zhang, I.H. Musselman, J.P. Ferraris, K.J. Balkus Jr., Gas permeability properties of Matrimid membranes containing the metal-organic framework Cu–BPY–HFS, J. Membr. Sci. 313 (2008) 170–181. [61] E.V. Perez, K.J. Balkus Jr., J.P. Ferraris, I.H. Musselman, Mixed-matrix membranes containing MOF-5 for gas separations, J. Membr. Sci. 328 (2009) 165–173. [62] Y.Y. Jiang, Z. Zhou, Z. Jiao, L. Li, Y.T. Wu, Z.B. Zhang, SO2 gas separation using supported ionic liquid membranes, J. Phys. Chem. B 111 (2007) 5058–5061. [63] A. Ito, S. Duan, Y. Ikenori, A. Ohkawa, Permeation of wet CO2/CH4 mixed gas through a liquid membrane supported on surface of a hydrophobic microporous membrane, Sep. Purif. Technol. 24 (2001) 235–242. [64] P. Scovazzo, J. Kieft, D.A. Finan, C. Koval, D. DuBois, R.D. Noble, Gas separations using non-hexafluorophosphate [PF6]- anion supported ionic liquid membranes, J. Membr. Sci. 238 (2004) 57–63. [65] J.E. Bara, S. Lessmann, C.J. Gabriel, E.S. Hatakeyama, R.D. Noble, D.L. Gin, Synthesis and performance of polymerizable room-temperature ionic liquids as gas separation membranes, Ind. Eng. Chem. Res. 46 (2007) 5397–5404.

114

Inorganic, Polymeric and Composite Membranes

[66] X. Hu, J. Tang, A. Blasig, Y. Shen, M. Radosz, CO2 permeability, diffusivity and solubility in polyethylene glycol-grafted polyionic membranes and their CO2 selectivity relative to methane and nitrogen, J. Membr. Sci. 281 (2006) 130–138. [67] J.E. Bara, E.S. Hatakeyama, C.J. Gabriel, X. Zeng, S. Lessmann, D.L. Gin, et al., Synthesis and light gas separations in cross-linked gemini room temperature ionic liquid polymer membranes, J. Membr. Sci. 316 (2008) 186–191. [68] J.E. Bara, C.J. Gabriel, E.S. Hatakeyama, T.K. Carlisle, S. Lessmann, R.D. Noble, et al., Improving CO2 selectivity in polymerized room-temperature ionic liquid gas separation membranes through incorporation of polar substituents, J. Membr. Sci. 321 (2008) 3–7. [69] J.E. Bara, E.S. Hatakeyama, D.L. Gin, R.D. Noble, Improving CO2 permeability in polymerized room-temperature ionic liquid gas separation membranes through the formation of a solid composite with a room-temperature ionic liquid, Polym. Adv. Technol. 19 (2008) 1415–1420. [70] S. Hanioka, T. Maruyama, T. Sotani, M. Teramoto, H. Matsuyama, K. Nakashima, et al., CO2 separation facilitated by task-specific ionic liquids using a supported liquid membrane, J. Membr. Sci. 314 (2008) 1–4. [71] J. Ilconich, C. Myers, H. Pennline, D. Luebke, Experimental investigation of the permeability and selectivity of supported ionic liquid membranes for CO2/He separation at temperatures up to 125  C, J. Membr. Sci. 298 (2007) 41–47. [72] L. Ferguson, P. Scovazzo, Solubility, diffusivity, and permeability of gases in phosphoniumbased room temperature ionic liquids: data and correlations, Ind. Eng. Chem. Res. 46 (2007) 1369–1374. [73] P. Cserje´si, N. Nemesto´thy, K. Be´lafi-Bako´, Gas separation properties of supported liquid membranes prepared with unconventional ionic liquids, J. Membr. Sci. 349 (2010) 6–11. [74] P. Scovazzo, D. Havard, M. McShea, S. Mixon, D. Morgan, Long-term, continuous mixed-gas dry fed CO2/CH4 and CO2/N2 separation performance and selectivities for room temperature ionic liquid membranes, J. Membr. Sci. 327 (2009) 41–48. [75] J.E. Bara, C.J. Gabriel, T.K. Carlisle, D.E. Camper, A. Finotello, D.L. Gin, et al., Gas separations in fluoroalkyl-functionalized room-temperature ionic liquids using supported liquid membranes, Chem. Eng. J. 147 (2009) 43–50. [76] Y. Park, B. Kim, Y. Byun, S. Lee, E. Lee, J. Lee, Preparation of supported ionic liquid membranes (SILMs) for the removal of acidic gases from crude natural gas, Desalination 236 (2009) 342–348. [77] D.G. Hert, J.L. Anderson, S.N.V. Aki, J.F. Brennecke, Enhancement of oxygen and methane solubility in 1-hexyl-3-methylimidazolium bis(trifluoromethylsulfonyl) imide using carbon dioxide, Chem. Commun. (2005) 2603–2605. [78] J. Jacquemin, M.F. Costa Gomes, P. Husson, V. Majer, Solubility of carbon dioxide, ethane, methane, oxygen, nitrogen, hydrogen, argon, and carbon monoxide in 1-butyl-3methylimidazolium tetrafluoroborate between temperatures 283 K and 343 K and at pressures close to atmospheric, J. Chem. Thermodyn. 38 (2006) 490–502. [79] J. Jacquemin, P. Husson, V. Majer, M.F. Costa Gomes, Low-pressure solubilities and thermodynamics of solvation of eight gases in 1-butyl-3-methylimidazolium hexafluorophosphate, Fluid Phase Equilib. 240 (2006) 87–95. [80] A. Finotello, J.E. Bara, D. Camper, R.D. Noble, Room-temperature ionic liquids: temperature dependence of gas solubility selectivity, Ind. Eng. Chem. Res. 47 (2008) 3453–3459. [81] J.L. Anderson, J.K. Dixon, J.F. Brennecke, Solubility of CO2, CH4, C2H6, C2H4, O2, and N2 in 1-hexyl-3-methylpyridinium bis(trifluoromethylsulfonyl)imide: comparison to other ionic liquids, Acc. Chem. Res. 40 (2007) 1208–1216.

Chapter

5

Review of CO2/CH4 Separation Membranes

115

[82] J. Anthony, E.J. Maginn, J.F. Brennecke, Thermodynamic properties of gases in the Ionic liquid 1-n-butyl-3-methylimidazolium hexafluorophosphate, J. Phys. Chem. B 106 (2002) 7315–7320. [83] T.J. Wooster, K.M. Johanson, K.J. Fraser, D.R. MacFarlane, J.L. Scott, Thermal degradation of cyano containing ionic liquids, Green Chem. 8 (2006) 691–696. [84] J.G. Huddleston, A.E. Visser, W.M. Reichert, H.D. Willauer, G.A. Broker, R.D. Rogers, Characterization and comparison of hydrophilic and hydrophobic room temperature ionic liquids incorporating the imidazolium cation, Green Chem. 3 (2001) 156–164. [85] C.P. Fredlake, J.M. Crosthwaite, D.G. Hert, S.N.V.K. Aki, J.F. Brennecke, Thermophysical properties of imidazolium-based ionic liquids, J. Chem. Eng. Data 49 (2004) 954–964. [86] H. Tokuda, K. Hayamizu, K. Ishii, Md.A.B.H. Susan, M. Watanabe, Physicochemical properties and structures of room temperature ionic liquids. 2. Variation of alkyl chain length in imidazolium, J. Phys. Chem. B. 109 (2005) 6103–6110.

This page intentionally left blank

Chapter 6

Gas Permeation Properties of Helium, Hydrogen, and Polar Molecules Through Microporous Silica Membranes at High Temperatures: Correlation with Silica Network Structure Masakoto Kanezashi and Toshinori Tsuru Department of Chemical Engineering, Hiroshima University, 1-4-1 Kagamiyama, Higashi-Hiroshima, 739-8527, Japan E-mail addresses: [email protected], [email protected]

INTRODUCTION Amorphous silica consists of an amorphous silica network structure that is stable up to 1000
C and that allows permeation of the small molecules such as helium (kinetic diameter: 0.26 nm) and hydrogen (0.289 nm) but not larger molecules such as nitrogen (0.364 nm) [1–3]. Silica membranes for hydrogen separation were first reported in 1989 by chemical vapor deposition (CVD) [4] and in 1990 by sol–gel processing [5]. Both techniques allow formation of amorphous silica networks connected by Si–O–Si bonding and containing –SiOH groups. Although silica membranes have been extensively developed in the past two decades, there still remain two technical challenges: pore size control and hydrothermal stability [3,6,7]. Tetraethoxysilane (TEOS) has been most commonly used for preparation of silica membranes in CVD and sol–gel process [4–16]. Sol–gel processing has the great advantage of pore size control based on the size of silica sols and the starting precursors and can allow for the low-temperature synthesis of mixed oxides by mixing metal ions into silica [5–8,10,12,15,16]. A new concept to control pore sizes has been proposed using organic–inorganic hybrid alkoxides, which contain organic groups between two silicon atoms, such as bis(triethoxysilyl)ethane [17,18]. Using organic–inorganic hybrid alkoxides, Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

117

118

Inorganic, Polymeric and Composite Membranes

very highly permeable hydrogen separation membranes were successfully prepared by tuning the size of the silica network [17,18]. Although very highly permeable hydrogen separation membranes were successfully prepared by tuning the size of the silica network, organic–inorganic hybrid silica membranes cannot be applied at high temperatures due to the presence of organic groups. Concerning the hydrothermal stability of silica membranes, it has been reported that they are not stable in humidified air, even at room temperature; H2 permeance decreases with time under hydrothermal conditions [12,19–23]. This is thought to be caused by densification of the silica network; siloxane bonds are broken to form silanol functional groups, which reduce the effective pore size and can lead to the rearrangement of silanol groups, which results in a denser structure. Extensive research on CVD [24,25] and hydrophobic membranes [26–28] has been carried out to develop hydrogen separation membranes that can be used at high temperatures in a steam atmosphere for possible use in the production of hydrogen for fuel cell systems. We previously proposed composite oxides with silica and various types of metal ions such as Zr, Al, Mg, and Ni, for improved stability in hydrothermal conditions, prepared by sol–gel processing [16,22,29–31] Among the metals, Ni was the first reported to be effective in improving hydrothermal stability [16,29]. We also reported improved hydrothermal stability by doping Co into silica with different Si/Co compositions by sol–gel processing and applied this method to hydrogen separation [30,31]. The Co-doped silica with improved hydrothermal stability was also investigated by other research groups based on permeation experiments [32]. However, the permeation properties through microporous silica membranes that consist of silica networks, including hydrogen and helium at high temperatures, are not well understood. Moreover, the permeation of water vapor has been reported in a limited number of papers, probably due to the difficulty of measuring the permeation rate of water vapor and the low stability of microporous silica membranes in a steam atmosphere [16,29,33,34]. For this chapter, the permeation characteristics of various types of gaseous molecules including NH3 [35] were investigated, and they are discussed here based on the activation energy of permeation and selectivity of gaseous molecules, which are intrinsic properties that reflect the pore size and pore chemistry of microporous silica. From the viewpoint of the transport mechanism through porous membranes, according to the size of the membrane pores and the mean free path, gaseous molecules permeate porous membranes via four types of permeation mechanisms: viscous flow, Knudsen flow, surface diffusion, and activated diffusion [36–39]. In these transport mechanisms, activated diffusion, where high separation can be expected, is the permeation mechanism in which permeance (mol m 2 s 1 Pa 1), defined as the permeation rate (mol s 1) per unit membrane area (m2) and unit pressure difference (Pa), increases with temperature. Activated diffusion can be caused by two mechanisms: vibration of the membrane matrix, mostly in the case of dense polymeric membranes, and a molecular sieving mechanism, in the case of

Chapter

6

Gas Permeation Through Silica Membranes

119

porous inorganic membranes. Obviously, the molecular size of permeating molecules is very important in determining the transport mechanism through porous membranes [36–39]. Several ways of defining molecular size are available, including kinetic diameter and Lennard-Jones collision diameter [40–42]. Kinetic diameter has been used most commonly to discuss adsorption in adsorbents and permeation through microporous materials. This diameter is determined as the molecular distance at the minimum potential using the Lennard-Jones potential for nonpolar spherical molecules (He, H2, N2). The kinetic diameters of polar gases (NH3, H2O) are determined based on the Stockmayer potential, which considers both the Lennard-Jones potential and the electrostatic interaction. However, very few papers have discussed which molecular sizes are appropriate for permeation and diffusion through amorphous silica networks where only the smallest molecules are allowed to permeate. For example, it is reported that the order of kinetic diameter for NH3, H2, and N2 is as follows; NH3 (0.26 nm), H2 (0.289 nm), and N2 (0.364 nm) which are obtained by Stockmayer potential [40]. However, Leeuwen [42] reported the molecular size of NH3 is 0.326 nm, which is also obtained by the Stockmayer potential. If the NH3 molecular size of 0.26 nm [40] is appropriate in the case of permeation through porous inorganic membranes, NH3 molecules could be selectively separated from gaseous mixtures of NH3 and H2 by the molecular sieving mechanism. On the contrary, if the NH3 molecular size of 0.326 nm [42] is appropriate, NH3 molecules could be selectively separated just by preferential adsorption of NH3 and its blocking effect on H2 molecules. In the present study, the transport mechanisms of He and H2 through amorphous silica networks are discussed extensively and compared with those through CVD silica membranes. He and H2 were used as probe molecules, since the permeation properties of the smallest molecules will offer structural information about the amorphous silica networks, and we propose a novel correlation between the activation energy of permeation and selectivity of gaseous molecules, both of which are intrinsic properties of microporous silica membranes. Moreover, the molecular sizes of polar molecules such as H2O and NH3 are also discussed.

EXPERIMENTAL Fabrication of Silica and Co-Doped Silica Membranes by Sol–Gel Method Silica colloidal sols were prepared by the following procedures [12,35]. TEOS was added to water with nitric acid as a catalyst at a molar ratio of TEOS:H2O: HNO3 ¼ 0.1:11.5–23:0.016–0.033, and then the solution was hydrolyzed at room temperature for 12 h. To convert the polymeric sol to colloidal sol, additional nitric acid and water were added and the solutions were boiled for another 8 h, keeping the total amount of solution constant (2.0, 1.5, 1.0 wt% of

120

Inorganic, Polymeric and Composite Membranes

equivalent TEOS in final sol). Cobalt-doped silica sols were prepared by mixing a specific amount of metal nitrate (Co(NO3)26H2O) and TEOS with Si/Co ¼ 1/ 1–3/1 as the starting chemicals [30,31]. Porous a-alumina cylindrical microfiltration membranes (length: 90 mm; outer diameter: 10 mm; inner diameter: 8 mm; nominal pore size: 1 mm; porosity: 50%) were used as substrates of hydrogen separation membranes. First, a-alumina particles of 0.2 mm in diameter were deposited on the microfiltration membrane and fired at 550
C to reduce the pore size enough for further sol coating. Next, colloidal sol solutions of SiO2–ZrO2 (Si/Zr¼1/1) were coated on the substrates to fabricate intermediate layers having approximate pore sizes of 1 nm. Finally, silica or Co-doped silica sols were further coated and fired for 30–60 min at 500–600
C under dry air or a steam partial pressure of 90 kPa. The detailed procedure may be found in a previous paper [12,30,31,35].

Gas Permeation/Separation Measurements for Silica Membranes Single gases (He, H2, CO2, N2, SF6) were fed on the outside (upstream) of a cylindrical membrane module at 200–300 kPa, keeping the down stream pressure constant at the atmospheric pressure. The permeation rate was measured by a bubble film meter. For hydrothermal stability tests, a mixture of steam and nitrogen was fed for a specific time, followed by pure nitrogen to dry the membrane completely, and then the permeation rates of pure He, H2, N2 were measured. By repeating this procedure, the time course of permeation rates was recorded [16,30,31]. Steam partial pressure was controlled by the total pressure in the range of 90–400 kPa. After a steady state was reached, the temperature dependencies of permeation rates were measured from 300 to 500
C to determine the activation energy of permeating gases. The water permeation rate was measured in the mixture of H2/H2O and N2/H2O (mole ratio of 1/1) under a total pressure of 200–500 kPa with the permeate stream at atmospheric pressure. The permeate flux of water was determined in two ways: a weighing method and a gas chromatographic method [31]. In the case of NH3 permeation, sweep gas (Ar) was used for measurement of NH3 permeation [35].

RESULTS AND DISCUSSION Improved Hydrothermal Stability of Amorphous Silica Membranes Pore size distribution of porous silica membranes was estimated by measuring several gas permeances at 200
C, where the effect of surface flow can be negligible. Figure 6.1 shows gas permeances for pure silica membranes (Si-1, Si-2, Si-3) at 200
C as a function of kinetic diameter [35]. The Si-1 membrane shows a high hydrogen permeance of 2.16  10 6 mol m 2 s 1 Pa 1 with a high H2/SF6 permeance ratio of 1600 and a low H2/N2 permeance ratio ( 30) at 200
C. However, the Si-3 membrane shows a one order of magnitude

Chapter

6

121

Gas Permeation Through Silica Membranes

10−5

He H2 CO2 N2

SF6

Permeance (mol m−2 s−1 Pa−1)

10−6

Si-1 10−7

10−8

Si-2

Si-3

10−9

10−10 0.2

0.3

0.4 0.5 Kinetic diameter (nm)

0.6

FIGURE 6.1 Gas permeances for silica membranes (Si-1, Si-2, Si-3) at 200
C as a function of kinetic diameter [35].

lower H2 permeance than Si-1 membrane with a high permeance ratio of H2 to N2 of 1000. For the Si-2 membrane, the gas permeance of He was comparable to that of the Si-1 membrane with an H2/N2 permeance ratio of  100 and an H2/SF6 permeance ratio of 2500, which are much higher than the selectivity for the Si-1 membrane but lower than those for the Si-3 membrane. The average pore size of the Si-1 membrane can be estimated to be approximately in the range of 0.5–0.6 nm, while that of the Si-3 membrane can be estimated to be approximately 0.3 nm and that of the Si-2 membrane 0.4–0.5 nm, suggesting that a high H2/N2 permeance ratio can be obtained by membranes with average pore size of 0.3 nm, while a high H2/SF6 permeance ratio can be produced by membranes with much larger average pore size. The average pore size of silica membranes could be successfully controlled by coating colloidal sols of different colloidal diameter. Silica membranes have the great advantage of controllability of pore size in the sub-nanometer range; however, a serious problem yet to be resolved is that of hydrothermal stability. Under hydrothermal conditions, gaseous permeance decreases by densification of the silica network through which helium and hydrogen can permeate. The decreased flux under hydrothermal conditions is one of the most important problems to be solved for the practical application of microporous silica membranes, since one of the possible applications would

122

Inorganic, Polymeric and Composite Membranes

include the separation of mixed gases that contain steam at high pressures, such as steam reforming of methane [33,34]. Metal doping such as Ni and Co was proposed to increase the hydrothermal stability by several groups [16,29–32]. Figure 6.2a shows the TEM cross-sectional image of the top layer with 10 mm depth of a Co-doped silica membrane, which was observed using a TOPCON (Model EM002b, Tokyo, Japan) [31]. The sample was prepared using a focused ion beam (FIB), and tungsten was deposited on the surface of the top layer. The Co-doped silica membrane showed a composite membrane structure, consisting of an a-alumina substrate that was prepared using a-alumina particles several micrometers in size (pore size: 1 mm), an intermediate layer prepared using a-alumina fine particles (200 nm in diameter), an SiO2–ZrO2 layer, and a Co-doped silica top layer for hydrogen separation according to the preparation procedure. A thin Co-doped SiO2 layer was observed as a layer having a finer morphology than the SiO2–ZrO2 layer (Figure 6.2b). The estimated thickness of the Co-doped SiO2 layer was approximately 50 nm, and that of the SiO2–ZrO2 layer was several hundred nanometers. Figure 6.3 shows the time course of permeance through a Co-doped silica (Si/Co ¼ 2/1) membrane (which was prepared by firing under steam partial pressure of 90 kPa) at a steam partial pressure of 300 kPa [31]. The Co-doped silica membrane showed slightly decreased He and H2 permeance during the first several hours under hydrothermal conditions and reached a steady permeance. Nitrogen showed a very low permeance and did not display an appreciable change during hydrothermal treatment. H2 showed a permeance of 1.8  10 7 mol m 2 s 1 Pa 1 with an H2/N2 separation factor of 730 after hydrothermal treatment (steam: 300 kPa, 500
C), while helium showed 5.5  10 7 mol m 2 s 1 Pa 1 with an He/N2 separation factor of 2230.

(a)

Co–SiO2/SiO2ZrO2

Tungsten

(b)

Co–SiO2

a-Al2O3 particle deposits (particle size: 0.2 mm)

SiO2ZrO2

a-Al2O3 particle deposits (particle size: 0.2 mm)

50 nm a-Al2O3 substrate (pore size: 1 mm)

1 mm

FIGURE 6.2 Transmission electron microscopy (TEM) image of Co-doped silica membranes (a) cross-section with 10 mm depth and (b) top layer cross-section at high magnification [31].

Chapter

6

123

Gas Permeation Through Silica Membranes

Permeance (mol m−2 s−1 Pa−1)

10−5 10−6

He

10−7

H2

10−8 10−9 10−10

N2 0

10

20

30 40 Time (h)

50

60

70

FIGURE 6.3 Time course of gaseous permeance through a Co-doped silica membrane (MemCoSi-1, steam partial pressure of 300 kPa, 500
C) [31].

The activation energy of H2 permeation (Ep) was obtained by regressing Equation (6.1) [38] with the experimental single permeation data between 300 and 500
C.   DEp k0 P ¼ pffiffiffiffiffiffiffiffiffiffi exp  ð6:1Þ RT MRT where DEp is the activation energy of H2 permeation, k0 is the structural parameter, M is the molecular weight of the permeating molecules, and R is the gas constant. Equation (6.1) covers activated diffusion (DEp > 0), surface diffusion (DEp < 0), and Knudsen diffusion (DEp ¼ 0). The activation energy can be determined by interactions between permeating molecules and the pore wall, based on the Lennard-Jones potential using the size (membrane pore size, molecular size of permeating molecules) and the interaction parameters [37,38]. Recently, Oyama et al. [14,43] applied a different type of permeation equation, which was originally proposed by Barrer [44], to the analysis of gaseous permeance through microporous silica membranes prepared by the CVD technique. Although the physical meaning of this permeation equation is clear, at least three parameters are required. In the present study, Equation (6.1), which requires two parameters, was applied for simplicity. Figure 6.4 shows H2 permeance at 500
C for microporous silica membranes (CVD [4,21], sol–gel method [16]) and metal-doped silica membranes (Nidoped [16,29], Co-doped [30,31], SiO2–Al2O3 [25]) after exposure to steam (closed symbols: steam: 300–400 kPa, open symbols: steam: < 100 kPa) as a function of activation energy of H2 permeation. Since helium and hydrogen are small molecules that show no or a quite small interaction with silica, the activation energy of permeation is mostly determined by that of diffusion.

124

Inorganic, Polymeric and Composite Membranes

H2 permeance (mol m−2 s−1 Pa−1)

10−6

10−7

SiO2 (Sol–gel [16], CVD [4,21]) Ni-doped silica [16,29] Co-doped silica [30,31] SiO2–Al2O3 [25] 10−8 10

20 30 40 Activation energy of H2 permeation (kJ mol−1)

FIGURE 6.4 H2 permeance at 500
C for microporous silica (CVD [4,21], sol–gel method [16]) and metal-doped silica membranes (Ni-doped [16,29], Co-doped [30,31], SiO2–Al2O3 [25]) after exposure to steam (closed symbols: steam: 300–400 kPa, open symbols: steam: <100 kPa) as a function of activation energy of H2 permeation.

Even though some reported data is scattered probably because of different membrane thickness and number of pores, the permeance of H2 decreases with an increase of activation energy of H2 permeation. Oyama et al. [45] reported ab initio calculation of the activation energy of H2 permeance through H2nSinOn (n ¼ 4–8) cyclosiloxane n-membered rings, which correspond to the pore sizes of silica networks. They clearly showed that activation energy increased with a decrease in silica member rings. Because the repelling force increases as pore size decreases, the activation energy of H2 permeation corresponds with the pore size that is effective for H2 permeation. The activation energy of H2 permeation for Ni-doped and Co-doped silica membranes slightly increases with an increase of partial pressure of steam (steam: 90–400 kPa). However, Nidoped and Co-doped silica membranes after exposure to steam (steam: 300–400 kPa) clearly show high H2 permeances with small activation energy in comparison with those for pure silica membranes after exposure to low steam (steam: < 100 kPa). Since the densification of amorphous silica networks under steam atmosphere accelerates with the higher partial pressure of steam, that is, activation energy increases with the higher partial pressure of steam [16], this suggests improved hydrothermal stability of silica by Ni and Co doping. Based on XRD analysis of Co-doped SiO2 powder, the doped Co was impregnated in the silica matrix, while only a certain amount of Co formed crystalline Co3O4, approximately 20 nm in size and existed outside the silica

Chapter

6

Gas Permeation Through Silica Membranes

125

network [30]. Co-doped SiO2 showed excellent stability under hydrothermal conditions and in a reductive/oxidative atmosphere, which was attributed to the doped Co, which could be impregnated in the silica matrix as metal ions [46] covalently bound compounds [47] such as Si–O–Co–, or as tiny crystals that cannot be detected by XRD. One possible mechanism that could prevent the densification of a silica network [30] is that doped cobalt, which may exist as metal ions, covalently bound compounds, or as tiny crystals, could increase stability under hydrothermal conditions. Doped cobalt might reduce the attack of water vapor and prevent thermally induced movement of silanol groups, such as rotation, followed by reduced recombination, as has been suggested for carbonized-template silica membranes by Duke et al. [27,28].

Helium and Hydrogen Permeation Properties Through Amorphous Silica Membranes Permeation of gases through microporous silica membranes can be explained based on the assumption of a bimodal pore size distribution consisting of intraparticle pores and interparticle pores [3]. Small molecules such as helium and hydrogen can permeate through the intraparticle pores, that is, an amorphous silica network, by activated diffusion. However, large molecules such as nitrogen can permeate through relatively large pores or pinholes in the Knudsen diffusion mechanism. Therefore, small molecules such as helium and hydrogen can permeate through both intraparticle and interparticle pores. Roughly speaking, the approximate sizes of silica networks available for He and H2 permeation could be from 0.25 to 0.3 nm, based on the kinetic diameters of permeating molecules. Since the permeation through interparticle pores obeys the Knudsen mechanism, it is possible to obtain permeance through silica network pores. Therefore, to discuss the permeation mechanism through silica networks, the following correction (Equation (6.2)) was made for Knudsen diffusion via relatively large pores through which nitrogen can permeate. Although correction of Knudsen flow was made in the present study, the activation energy without correlation was approximately the same as that without correction when the permeance ratio was greater than 100, due to the small contribution of permeation through large pores.   pffiffiffiffiffi k0 E2 ð6:2Þ P  14PN2 ¼ pffiffiffiffiffiffiffiffiffiffi exp  RT MRT Figure 6.5 shows He/H2 permeance ratio at 500
C as a function of activation energy of H2 permeation for microporous silica membranes prepared by sol–gel [16,22] CVD [13,48–50] and vitreous silica [51] using various materials including silica and silica composite oxides with Zr, Ni, and Co [31]. It should be noted that each point corresponds to one membrane. Since the molecular size of He is smaller than that of H2, the pore area effective for permeation is larger

126

Inorganic, Polymeric and Composite Membranes

7 Sol–gel methods : This work, [16, 22] CVD methods : [13] [48] [49] [50] Vitrification:

He/H2 permeance ratio (–)

6

[51]

5

4

3

2

1

0

10 20 30 40 Activation energy of H2 permeation (kJ mol−1)

FIGURE 6.5 He/H2 permeance ratio at 500
C as a function of the activation energy of H2 permeation for microporous silica membranes prepared by sol–gel [16,22] CVD [13,48–50] and vitreous silica [51] using different types of materials [31].

for He than it is for H2. Therefore, the He/H2 permeance ratio can be considered as a measure of the pore size distribution of a silica network. However, the hydrogen activation energy corresponds to the membrane pore sizes through which H2 permeates, as discussed in the previous section. The correlation between the activation energy of permeation and the selectivity of gaseous molecules, both of which are intrinsic properties of microporous silica, is newly proposed in the present study, and the applicability can be examined using various types of silica membranes, including sol–gel derived membranes, CVD, and vitreous glass silica. Surprisingly, irrespective of preparation methods and types of membrane materials, all types of membranes can be expressed with one correlation curve; the permeance ratio of He/H2 increased with an increase in the activation energy of hydrogen for any type of silica membrane. The reason that no effect of membrane materials was observed can be explained as follows. Compared with some metals, including Si, oxygen is electronegative and attracts electrons showing a larger size than other metals [40]. In other words, the pore walls of silica and silica composite oxides are assumed to be covered by oxygen. Another important point that should be emphasized is that no dependency on the preparation method was observed in the correlation curve. We can conclude that amorphous silica networks have similar pore size distributions, irrespective of membrane materials and preparation technique. Figure 6.6 shows the activation energy of He permeation through silica membranes prepared by sol–gel, CVD, and vitreous glass, as a function of that of H2, showing an approximately linear correlation between them [31]. The larger the activation energy of helium, the larger that of hydrogen. The

6

127

Gas Permeation Through Silica Membranes

Activation energy of He permeation (kJ mol−1)

Chapter

30

20

10

0

0

10

20

30

40

Activation energy of H2 permeation (kJ mol−1) FIGURE 6.6 Correlation of activation energy of He and H2 permeation through microporous silica membranes [31].

difference in activation energy between helium and hydrogen is 5 and 10 kJ mol 1 for silica membranes having H2 activation energy of 20 and 30 kJ mol 1, respectively, suggesting that with a decrease in pore size, the activation energy of hydrogen increases more rapidly than that of He due to larger molecular size. This trend is consistent with the ab intio calculation given by Oyama et al. [45].

Permeation Properties of Polar Molecules (NH3, H2O) Through Amorphous Silica Membranes Figure 6.7 shows the temperature dependency of gas permeances for the Si-1 membrane (a) and the Si-3 membrane (b) in the temperature range 50–400
C [35]. The permeances of He and H2 for the Si-1 membrane are independent of temperature, while those for the Si-3 membrane drastically increase with increasing temperature, which is characteristic of an activated permeation mechanism. For the NH3 permeation behavior of the Si-3 membrane, there is a maximum point at around 200
C. The permeances of N2 for the Si-1 and Si-3 membranes show Knudsen type permeation behavior with permeance that slightly increases with decreasing temperature. The permeance of SF6 for the Si-1 membrane also shows Knudsen permeation behavior. The reason for the occurrence of a maximum for NH3 permeance can be explained as follows; the number of adsorbed molecules decreases with increasing temperature (above the maximum point), while at low temperature, especially below the critical temperature of NH3 molecules (< 130
C), the mobility of adsorbed

128

Inorganic, Polymeric and Composite Membranes

(a)

Temperature [°C] 100

(b) 50

Si-1 H2 10−6

Temperature [°C] 10−5

400 300 200

100

50

Si-3 Permeance (mol m−2 s−1 Pa−1)

Permeance (mol m−2 s−1 Pa−1)

10−5

400 300 200

He NH3

10−7 N2 10−8 SF6 10−9

10−6 He 10−7

H2

10−8 NH3 10−9 N2

10−10

1.0

1.5

2.0

2.5

1000/T (K−1)

3.0

3.5

10−10

1.0

1.5

2.0

2.5

3.0

3.5

1000/T (K−1)

FIGURE 6.7 Temperature dependency of gas permeances for the Si-1 membrane (a) and the Si-3 membrane (b) in the temperature range of 50–400
C [35].

molecules decreases because of liquid-like diffusion of pore filling molecules. It is reported that the order of kinetic diameter for He, H2, NH3, and N2 is as follows; He (0.26 nm), NH3 (0.26 nm), H2 (0.289 nm), and N2 (0.364 nm) [40]. However, the order of gas permeances at high temperatures for silica membranes was He > H2 > NH3 > N2 and independent of average pore size of membranes. Both silica membranes show higher H2 permselectivity over NH3; the H2/NH3 permeance ratio for the Si-1 membrane at 200
C was 6, while that for the Si-3 membrane at 200 and 400
C were 40 and 178 due to the activated permeation of H2 molecules. It should be noted that the H2/NH3 permeance ratio obtained by binary component gas separation of NH3 and H2 at 400
C agrees well with that obtained by single gas permeation, suggesting that no adsorption of NH3 molecules occurs on silica at high temperatures. Figure 6.8 shows the time courses for the permeances of He, H2, and N2 through a Co-doped silica membrane (Si/Co ¼ 1/1), which were measured using pure gases after drying the membrane with nitrogen [31]. The permeance of steam was measured using a mixture of steam and nitrogen with an H2O/N2 ratio of 50/50. The Co-doped silica membrane again showed stable permeance under hydrothermal conditions (500
C and a steam partial pressure of 300 kPa). The permeance decreased in the order of He > H2 > H2O > N2. The permeance ratio of hydrogen to water vapor, which is an important factor for actual application in membrane reactors for steam reforming of methane, was 30. As mentioned previously, the kinetic diameters, which have mostly been accepted in gas separation using porous materials such as adsorption and membrane separation, were decreased in the order of He (0.26 nm) <

Chapter

6

129

Gas Permeation Through Silica Membranes

Permeance (mol m−2 s−1 Pa−1)

10−6 He 10−7

H2

19.6 kJ mol−1 28.6 kJ mol−1

10−8 H2O

H2/H2O = 30

10−9 N2 10−10

0

10

20 30 Time (h)

40

50

FIGURE 6.8 Time course of permeance through a Co-doped silica membrane (Si/Co ¼ 1/1) in steam atmosphere (Mem-CoSi-2, steam partial pressure of 300 kPa) [31].

H2O (0.265 nm) < H2 (0.289 nm) < N2 (0.364 nm) [40]. Although water vapor is reported to be smaller than hydrogen, the permeance of water vapor was found to be lower than that of hydrogen. That is, the order of permeance does not follow the kinetic diameters. Therefore, another possible permeation mechanism, the surface diffusion mechanism, should be examined. Water vapor is a polar molecule that is expected to have a large interaction with silica. The effect of partial pressure on the permeance of water vapor was examined in a mixture of N2/H2O and H2/H2O systems, to determine whether the surface diffusion mechanism could be the transport mechanism. Wakabayashi et al. [52] measured diffusivity in the range of 0–20 kPa, which is much lower than the partial pressure that is required for actual applications such as steam reforming of methane, and pointed out the possible contribution of surface diffusion to the transport at high temperatures, based on the adsorption measurement. If the surface diffusion is responsible for membrane permeation, permeance should decrease with an increase in the partial pressure of water vapor. Experimental results of Co-doped silica reveal that, as shown in Figure 6.9, the permeance of water vapor was approximately constant, irrespective of H2O partial pressure in the range of 10–300 kPa at 500
C for both N2/H2O and H2/H2O systems [31]. No contribution of surface diffusion to permeation through the silica membranes was confirmed in the range of H2O partial pressure of 10–300 kPa at temperatures from 300 to 500
C. It should also be noted that hydrogen and nitrogen mixed with water vapor showed approximately the same permeance of hydrogen and nitrogen as that measured in a single gas. Decreased permeance in mixtures, compared with single components, is reported to be due to the blocking effect of adsorptive gases on the permeation of less adsorptive gases, such as hydrogen and nitrogen, especially at low temperatures. The lack of effect of mixing the two molecules

130

Inorganic, Polymeric and Composite Membranes

Permeance (mol m−2 s−1 Pa−1)

10−6 closed symbols: H2–H2O H2 (single) open symbols: N2–H2O 10−7

H2O

10−8

10−9

10−10

N2 (single)

0

100 200 Feed partial pressure of H2O (kPa)

300

FIGURE 6.9 Effect of the partial pressure of water on permeance at 500
C through Co-doped silica (Mem-CoSi-3) membrane in mixed H2/H2O and N2/H2O systems [31].

suggests a very small interaction between inorganic gases and water vapor, as well as between membrane and water vapor. Figure 6.10 shows the temperature dependency of permeance, including water vapor using a membrane with Si/Co ¼ 2/1 (Mem-CoSi-3) [31]. As discussed before, the activation energy of the smaller molecule, that is, He, was lower than that of H2, while nitrogen showed a decreased permeance with an increase in temperature. The explanation may be that He and H2 permeate through the small pores in the silica network, while nitrogen was considered to permeate through relatively large pores, such as pinholes, in the Knudsen diffusion mechanism. Water vapor appeared to follow the activated diffusion mechanism. The permeance ratio of hydrogen to water vapor was approximately 20, again showing that hydrogen was more permeable than water vapor. The permeance slope of water vapor is positive but lower than that of hydrogen and higher than that of nitrogen. This intermediate tendency can be explained by the water vapor being able to permeate both the silica network (small pores) in activated diffusion and relatively large pores in Knudsen diffusion. To date, only a limited number of papers have reported the permeance of water vapor, and no work has been reported previously on the NH3 permeation behavior for porous inorganic membranes at high temperatures probably because of the difficulty of measurement and the low stability of microporous silica membranes [16,29,31,33–35]. Nomura et al. [34] reported permeance ratios of hydrogen to water vapor from 30 to 100 for silica membranes prepared by the CVD method. We reported that silica and silica–zirconia microporous membranes showed a higher permeance for hydrogen than for water vapor [16,29,31,33]. There are several possibilities: (1) a large interaction with the

Chapter

6

131

Gas Permeation Through Silica Membranes

Temperature (°C) 10−6

500

400

300

He

Permeance (mol m−2 s−1 Pa−1)

10−7

H2

10−8 H2O

10−9 N2

Knudsen 10−10 1.2

1.4

1.6

1.8

1000/T (K−1) FIGURE 6.10 Arrhenius plot of permeance for a Co-doped silica membrane (Si/Co ¼ 1/1) in a steam atmosphere (Mem-CoSi-3) [31].

silica surface, (2) differences in molecular shapes of polar molecules (NH3 and H2O) and hydrogen, (3) solid vibration of the silica network, and (4) differences in a kinetic diameter of permeating molecules through silica membranes. Water and ammonia are polar molecules that possibly have a larger interaction with silica than do hydrogen, and therefore, the molecular motion of water vapor and ammonia may be restricted on the silica surface. However, the interaction with silica cannot account for the order of permeances because, at high temperatures, there was negligible adsorption of NH3 molecules on silica, and water did not show the permeation characteristics of a surface diffusion. Using CVD silica membranes, Oyama et al. [14] reported that the order of permeance for noble gases was unusual, He > H2 > Ne, since it followed neither the order of kinetic diameter (He: 0.26 nm, H2: 0.289 nm, Ne: 0.275 nm) nor the molecular weight (He: 4.0, H2: 2, Ne: 20.1). They explained the order using a theory involving jumps between solubility sites. The vibration frequency of permeating gas between the sorption sites was smallest for Ne due to its having the largest molecular weight, which also accounts for its having the lowest

132

Inorganic, Polymeric and Composite Membranes

permeance of the three molecules. However, the applicability of the theory to polar nonspherical molecules remains unclear. Regarding the kinetic diameters, there are several methods for determining the molecular size that are available for discussion of adsorption and membrane permeation in/through microporous materials, such as collision diameter and kinetic diameter. Based on the extended simple point charge (SPC/E) model proposed by Brendesen et al. [41], a molecular size of 0.317 nm was obtained for water using its heat of evaporation. van Leeuwen [42] obtained a molecular size of 0.2995 nm for water, based on the Stockmayer potential using phasecoexistence data (vapor–liquid phase). Breck [40] and van Leeuwen [42] used the Stockmayer potential using gaseous viscosity data and vapor–liquid equilibrium data and obtained sizes of 0.265 and 0.2995 nm, respectively. Brendesen et al. [41] also used thermodynamic data in the liquid phase. For the kinetic diameters of ammonia and hydrogen of 0.26 and 0.289 nm, respectively, hydrogen showed a higher permeance than ammonia. Based on permeation properties, which is also obtained from vapor–liquid equilibrium data by Leeuwen [42], a value of 0.326 nm seems to be suitable for the molecular size of NH3 permeating through amorphous SiO2 membranes at high temperature. The reason is not clear at this moment, but when NH3 and H2O molecules permeate through amorphous silica structure with pore diameter of 0.3–0.5 nm at high temperatures, the interactions between permeating molecules (NH3 and H2O) and amorphous silica can be dominant, that is, the interactions between permeating molecules (NH3–NH3, H2O–H2O) (mean free path: 220 nm, 500
C, 101.3 kPa) are less than those between permeating molecules (NH3 and H2O) and amorphous silica. Therefore, the concentration of NH3 and H2O molecules in a membrane is considered to be quite low because of the mean free path and positive activation energy of NH3 and H2O permeation. This suggests that the molecular size determined by viscosity (the interaction between permeating molecules) may not be applicable to silica microporous membranes. The following is another possible reason. Determination of molecular size using experimental data in liquid phase appears to be appropriate for the evaluation of the size of molecules permeating in silica networks, probably because of the possibility of more precise calculation by considering the interaction of coulombic interaction in dense fluid data. Figure 6.11 shows gas permeances for Co-doped silica (Mem-CoSi-1, MemCoSi-2, Mem-CoSi-3) (a) and silica membranes (Si-1, Si-2, Si-3) (b) at 200
C as a function of molecular size. H2O with molecular size of 0.2995 and/or 0.317 nm and NH3 with 0.326 nm are in good agreement with experimentally obtained permeances using microporous silica and Co-doped silica membranes and are clearly independent of membrane properties, suggesting that the order of molecular sizes through amorphous silica membranes is He (0.26 nm) < H2 (0.289 nm) < H2O (0.2995, 0.317 nm) < NH3 (0.326nm) < N2 (0.364 nm).

Chapter

(a)

6

He

H2 H2O

(b)

N2

10−5 Permeance (mol m−2 s−1 Pa−1)

Permeance (mol m−2 s−1 Pa−1)

10−6

10−7

10−8

10−9

10−10 0.2

133

Gas Permeation Through Silica Membranes

0.25

0.3

0.35

Molecular size (nm)

0.4

He H2 NH3 N2

SF6

10−6

10−7

10−8

10−9

10−10 0.2

0.3

0.4

0.5

0.6

Molecular size (nm)

FIGURE 6.11 Gas permeances for Co-doped silica (Mem-CoSi-1, Mem-CoSi-2, Mem-CoSi-3) (a) and silica membranes (Si-1, Si-2, Si-3) (b) at 200
C as a function of molecular size.

CONCLUSIONS Silica and cobalt-doped silica membranes were successfully prepared to study the permeation mechanism of gas molecules, focusing particularly on helium, hydrogen, water vapor, and ammonia in the 300–500
C temperature range and are discussed based on the activation energy of permeation and the selectivity of gaseous molecules. 1. The sol–gel method was applied for the preparation of silica membranes with different average pore size. The average pore size of silica membranes was successfully controlled by coating with colloidal sols of different colloidal diameter. Cobalt-doped silica membranes that showed high permeance, in the range of 1.8  10 7 mol m 2 s 1 Pa 1 and an H2/N2 permeance ratio of approximately 730, with excellent hydrothermal stability under steam pressure of 300 kPa, were successfully prepared. 2. The activation energy of H2 permeance through cobalt-doped silica membranes with varying Co contents, which ranged from 10 to 30 kJ mol 1, was found to correlate well with the permeance ratio of He/ H2 for porous silica membranes prepared by sol–gel processing as well as CVD and vitreous glasses, suggesting similar amorphous silica network structures, irrespective of preparation techniques. 3. It was found that the order of gas permeances for silica membranes was He > H2 > NH3 > N2, which did not follow the order of kinetic diameter (He: 0.26 nm, NH3: 0.26 nm, H2: 0.289 nm, N2: 0.364 nm), and independent of average pore size of membranes. This suggests that molecular size of NH3 is larger than that of H2 in the case of permeation through amorphous silica structure. A value of 0.326 nm, which is obtained from vapor–liquid

134

Inorganic, Polymeric and Composite Membranes

equilibrium data, seems to be suitable for the molecular size of NH3 permeating through amorphous SiO2 membranes at high temperature. 4. The permeance ratios of H2/H2O were found to range from 5 to 40, that is, hydrogen (kinetic diameter: 0.289 nm) was always more permeable than water vapor (0.265 nm). This suggests that the H2O molecular size of 0.265 nm may not be appropriate for permeation through amorphous silica networks.

REFERENCES [1] R.W. Baker, Membrane Technology and Applications, second ed., John Wiley & Sons, Ltd, West Sussex, 2004. [2] C.J. Brinker, G.W. Scherer, Sol-Gel Science: The Physics and Chemistry of Sol-Gel Processing, Academic Press, San Diego, 1990. [3] T. Tsuru, Nano/subnano-tuning of porous ceramic membranes for molecular separation, J. Sol Gel Sci. Technol. 46 (2008) 349. [4] G.R. Gavalas, C.E. Megiris, S.W. Nam, Deposition of H2 permselective SiO2 films, Chem. Eng. Sci. 44 (1989) 1829. [5] S. Kitao, H. Kameda, M. Asaeda, Gas separation by thin porous silica membrane of ultra fine pores at high temperature, MAKU (Membrane) 15 (1990) 222. [6] Y.S. Lin, I. Kumakiri, B.N. Nair, H. Alsyouri, Microporous inorganic membranes, Sep. Purif. Methods 31 (2002) 229. [7] N.W. Ockwig, T.M. Nenoff, Membranes for hydrogen separation, Chem. Rev. 107 (2007) 4078. [8] R.J.R. Uhlhorn, K. Keizer, A.J. Burggraaf, Gas transport and separation with ceramic membranes. Part II. Synthesis and separation properties of microporous membranes, J. Membr. Sci. 66 (1992) 271. [9] S. Yan, H. Maeda, K. Kusakabe, S. Morooka, Hydrogen-permselective SiO2 membrane formed in pores of alumina support tube by chemical vapor deposition with tetraethylorthosilicate, Ind. Eng. Chem. Res. 33 (1994) 2096. [10] R.S.A. de Lange, K. Keizer, A.J. Burggraaf, Analysis and theory of gas transport in microporous sol-gel derived ceramic membranes, J. Membr. Sci. 104 (1995) 81. [11] S. Nakao, T. Suzuki, T. Sugawara, T. Tsuru, S. Kimura, Preparation of microporous membranes by TEOS/O3 CVD in the opposing reactants geometry, Microporous Mesoporous Mater. 37 (2000) 145. [12] M. Asaeda, S. Yamasaki, Separation of inorganic/organic gas mixtures by porous silica membranes, Sep. Purif. Technol. 25 (2001) 151. [13] D. Lee, L. Zhang, S.T. Oyama, S. Niu, R.F. Saraf, Synthesis, characterization, and gas permeation properties of a hydrogen permeable silica membrane supported on porous alumina, J. Membr. Sci. 231 (2004) 117. [14] Y. Gu, S.T. Oyama, High molecular permeance in a poreless ceramic membrane, Adv. Mater. 19 (2007) 1636. [15] R.M. de Vos, H. Verweij, High-selectivity, high-flux silica membranes for gas separation, Science 279 (1998) 1710. [16] M. Kanezashi, M. Asaeda, Stability of H2-permselective Ni-doped silica membranes in steam at high temperature, J. Chem. Eng. Jpn. 38 (2005) 908.

Chapter

6

Gas Permeation Through Silica Membranes

135

[17] M. Kanezashi, K. Yada, T. Yoshioka, T. Tsuru, Design of silica networks for development of highly permeable hydrogen separation membranes with hydrothermal stability, J. Am. Chem. Soc. 131 (2009) 414. [18] M. Kanezashi, K. Yada, T. Yoshioka, T. Tsuru, Organic-inorganic hybrid silica membranes with controlled silica network size: preparation and gas permeation characteristics, J. Membr. Sci. 348 (2010) 310. [19] J.C.S. Wu, H. Sabol, G.W. Smith, D.L. Flowers, P.K.T. Liu, Characterization of hydrogenpermselective microporous ceramic membrane, J. Membr. Sci. 96 (1994) 275. [20] S. Kim, G.R. Gavalas, Preparation of H2 permselective silica membranes by alternating reactant vapor deposition, Ind. Eng. Chem. Res. 34 (1995) 168. [21] B.K. Sea, M. Watanabe, K. Kusakabe, S. Morooka, S.S. Kim, Formation of hydrogen permselective silica membrane for elevated temperature hydrogen recovery from a mixture containing steam, Gas Sep. Purif. 10 (1996) 187. [22] K. Yoshida, Y. Hirano, H. Fujii, T. Tsuru, M. Asaeda, Hydrothermal stability and performance of silica–zirconia membranes for hydrogen separation in hydrothermal conditions, J. Chem. Eng. Jpn. 34 (2001) 523. [23] T. Tsuru, T. Tsuge, S. Kubota, K. Yoshida, T. Yoshioka, M. Asaeda, Catalytic membrane reaction for methane steam reforming using porous silica membranes, Sep. Sci. Technol. 36 (2001) 3721. [24] M. Nomura, K. Ono, S. Gopalakrishnan, T. Sugawara, S.-I. Nakao, Preparation of a stable silica membrane by a counter diffusion chemical vapor deposition method, J. Membr. Sci. 251 (2005) 151. [25] Y. Gu, P. Hacarlioglu, S.T. Oyama, Hydrothermally stable silica-alumina composite membranes for hydrogen separation, J. Membr. Sci. 310 (2008) 28. [26] R.M. de Vos, W.F. Maier, H. Verweij, Hydrophobic silica membranes for gas separation, J. Membr. Sci. 158 (1999) 277. [27] M.C. Duke, J.C.D. da Costa, G.Q. Lu, M. Petch, P. Gray, Carbonised template molecular sieve silica membranes in fuel processing systems: permeation, hydrostability and regeneration, J. Membr. Sci. 241 (2004) 325. [28] M.C. Duke, J.C.D. da Costa, D.D. Do, P. Gray, G.Q. Lu, Hydrothermally robust molecular sieving silica for wet gas separation, Adv. Funct. Mater. 16 (2006) 1215. [29] M. Kanezashi, M. Asaeda, Hydrogen permeation characteristics and stability of Ni-doped silica membranes in steam at high temperature, J. Membr. Sci. 271 (2006) 86. [30] R. Igi, T. Yoshioka, Y.H. Ikuhara, Y. Iwamoto, T. Tsuru, Characterization of Co-doped silica for improved hydrothermal stability and application to hydrogen separation membranes at high temperatures, J. Am. Ceram. Soc. 91 (2008) 2975. [31] T. Tsuru, R. Igi, M. Kanezashi, T. Yoshioka, S. Fujisaki, Y. Iwamoto, Permeation properties of hydrogen and water vapor through porous silica membranes at high temperatures, AIChE J. DOI: 10.1002/aic.12298. [32] D. Uhlmann, S. Liu, B.P. Ladewigb, J.C.D. da Costa, Cobalt-doped silica membranes for gas separation, J. Membr. Sci. 326 (2009) 316. [33] T. Tsuru, K. Yamaguchi, T. Yoshioka, M. Asaeda, Methane steam reforming by microporous catalytic membrane reactors, AIChE J. 50 (2005) 2794. [34] M. Nomura, M. Seshimo, H. Aida, K. Nakatani, S. Gopalakrishnan, T. Sugawara, et al., Preparation of a catalyst composite silica membrane reactor for steam reforming reaction by using a counter-diffusion CVD method, Ind. Eng. Chem. Res. 45 (2006) 3950. [35] M. Kanezashi, A. Yamamoto, T. Yoshioka, T. Tsuru, Characteristics of ammonia permeation through porous silica membranes, AIChE J. 56 (2010) 1204.

136

Inorganic, Polymeric and Composite Membranes

[36] J. Xiao, J. Wei, Diffusion mechanism of hydrocarbons in zeolites-I. Theory, Chem. Eng. Sci. 47 (1992) 1123. [37] A.B. Shelekhin, A.G. Dixon, Y.H. Ma, Theory of gas diffusion and permeation in inorganic molecular-sieve membranes, AIChE J. 41 (1995) 58. [38] T. Yoshioka, E. Nakanishi, T. Tsuru, M. Asaeda, Experimental study of gas permeation through microporous silica membranes, AIChE. J. 47 (2001) 2052. [39] H. Verweij, Y.S. Lin, J. Dong, Microporous silica and zeolite membranes for hydrogen purification, MRS Bull. 31 (2006) 1. [40] D.W. Breck, Zeolite Molecular Sieves, Structure, Chemistry and Use, John Wiley, New York, 1974. [41] H.J.C. Brendesen, J.R. Grigera, T.P. Strasstsma, The missing term in effective pair potentials, J. Phys. Chem. 91 (1987) 6269. [42] M.E. van Leeuwen, Derivation of Stockmayer potential parameters for polar fluids, Fluid Phase Equilib. 99 (1994) 1. [43] S.T. Oyama, D. Lee, P. Hacarlioglu, R.F. Saraf, Theory of hydrogen permeability in nonporous silica membranes, J. Membr. Sci. 244 (2004) 45. [44] R.M. Barrer, D.E.W. Vaughan, Solution and diffusion of helium and neon in tridymite and cristobalite, Trans. Faraday Soc. 63 (1967) 2275. [45] P. Hacarlioglu, D. Lee, G.V. Gibbs, S.T. Oyama, Activation energy for permeation of He and H2 through silica membranes: an ab initio calculation study, J. Membr. Sci. 313 (2008) 277. [46] S. Esposito, M. Turco, G. Ramis, B. Giovanni, P. Pernice, C. Pagliuca, et al., Cobalt-silicon mixed oxide nanocomposites by modified sol-gel method, J. Solid State Chem. 180 (2007) 3341. [47] G. Ortega-Zarzosa, C. Araujo-Andrade, M.E. Compea´n-Jasso, J.R. Martı´nez, F. Ruiz, Cobalt oxide/silica xerogels powders: X-ray diffraction, infrared and visible absorption studies, J. Sol Gel Sci. Technol. 24 (2002) 23. [48] G.-J. Hwang, K. Onuki, S. Shimizu, H. Ohya, Hydrogen separation in H2-H2O-HI gaseous mixture using the silica membrane prepared by chemical vapor deposition, J. Membr. Sci. 162 (1999) 83. [49] S. Araki, N. Mohri, Y. Yoshimitsu, Y. Miyake, Synthesis, characterization and gas permeation properties of a silica membrane prepared by high-pressure chemical vapor deposition, J. Membr. Sci. 290 (2007) 138. [50] T. Nagano, S. Fujisaki, K. Sato, K. Hataya, Y. Iwamoto, M. Nomura, et al., Relationship between the mesoporous intermediate layer structure and the gas permeation property of an amorphous silica membrane synthesized by counter diffusion chemical vapor deposition, J. Am. Ceram. Soc. 91 (2008) 71. [51] S.T. Hwang, K. Kammermeyer, Membrane in Separation, Wiley, New York, 1975. [52] H. Wakabayashi, M. Tomozawa, Diffusion of water into silica Glass at low temperature, J. Am. Ceram. Soc. 72 (1989) 1850.

Chapter 7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular Sieve Membrane Performance Properties Mayumi Kiyono1, William J. Koros1,* and Paul J. Williams2 Department of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA 2 Shell Projects and Technology, Houston, Texas, USA * Corresponding author: E-mail address: [email protected] 1

INTRODUCTION As an alternative to thermally driven gas separations, membrane-based gas separations have been introduced in pursuit of lower environmental impact, relatively low cost and lower energy requirements. Membrane use in natural gas separation allows removal of contaminants, the most prevalent of which comprises CO2 [1,2]. Currently, polymers are the dominant materials used for membranes, due to easy processability and adequate permselectivity for many gas pairs. Transport properties of glassy polymer membranes can be tailored by introducing packing–inhibiting bulky groups and intrinsically rigid linkages in the polymer backbones [3,4]. The separation performance of processable polymeric materials is limited by an “upper bound” trade-off curve relating permeability and selectivity [5], while performance of carbon molecular sieve (CMS) membranes can exceed this trade-off curve for challenging gas pairs, such as O2/N2, CO2/CH4, C3H6/C3H8 [6,7]. CMS membranes are formed by thermal decomposition of polymer precursors that result in amorphous carbon membranes [8]. This chapter describes development of a pyrolysis method to form CMS membranes with performance enhancement of more than 50–100 times in productivity and double in efficiency, compared with polymeric precursors. The authors describe the development and assessment of the pyrolysis method with oxygen exposure, tested on two different polymer precursors: in-house synthesized polymer, Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

137

138

Inorganic, Polymeric and Composite Membranes

6FDA/BPDA-DAM, and a commercially available polymer, MatrimidÒ. In addition, importance of oxygen concentration level prior to the pyrolysis process is described, which is a key factor for scale-up process. Lastly, a hypothetical mechanism of the pyrolysis process with respect to oxygen exposure during the formation of CMS membranes is proposed.

THEORY AND BACKGROUND Transport in CMS Membranes Gas transport through CMS membranes is modeled by the sorption–diffusion mechanism. Specifically, gas molecules sorb into the membrane at the upstream, then diffuse under the influence of a chemical potential gradient, and finally desorb from the membrane at the downstream. Two intrinsic properties, “permeability” and “selectivity,” are used to evaluate the performance of membrane materials. Permeability is a measure of the membrane material’s intrinsic productivity, and the selectivity is a measure of the membrane’s separation efficiency. The permeability equals the pressure and thickness normalized flux, described as: Pi ¼

ni l : Dpi

ð7:1Þ

where ni represents the flux of gas molecule component “i” through the membrane of thickness, l, and Dpi is the transmembrane partial pressure difference that acts as the driving force for component i across the membrane. The permeability can be described in terms of the governing kinetic and thermodynamic parameters, namely the diffusion coefficient, Di, and the sorption coefficient, Si, by: Pi ¼ Di Si :

ð7:2Þ

The diffusion coefficient, Di, in CMS membranes is a strong function of penetrate size and ideally enables molecular sieving discrimination between similarly sized penetrants. The sorption coefficient equals the concentration of gas sorbed divided by the penetrant partial pressure at equilibrium and depends on the condensability of the gas penetrant and its interactions with the membrane material. For CMSs with rigid saturable capacities, a Langmuir isotherm is commonly used, so the sorption coefficient can be expressed as: Si ¼

Ci CHi 0 bi ¼ : pi 1 þ bi pi

ð7:3Þ

where Ci is the equilibrium uptake of penetrant i by the sorbent, pi is the partial pressure, CHi0 is the Langmuir hole filling capacity, and bi is the Langmuir affinity constant for component i.

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

139

Structure of CMS Membranes As previously mentioned, CMS membranes are formed by thermal decomposition of polymer precursors that results in an almost pure carbon material, in many cases above 95% carbon [9–13]. When polymers are pyrolyzed, either coke or char is formed [8,9]. While coke forms graphite at a temperature above  2473 K, char remains in an amorphous structure [8,9]. Such amorphous materials are believed to have a highly aromatic structure comprised of disordered sp2 hybridized carbon sheets with angles of disorientation that can attain values of several degrees as illustrated in Figure 7.1a and b [8]. The structure can be envisioned to comprise roughly parallel layers of condensed hexagonal rings with no three-dimensional crystalline order. Pores are formed from packing imperfections between microcrystalline regions in the material [14,15], and the pore structure in CMS membranes is described as “slit-like” with an “idealized” pore structure illustrated in Figure 7.1c. This pore structure can be further represented as shown in Figure 7.1d in terms of an idealized bimodal pore distri˚ connected by smaller bution. This distribution consists of large pores of 6–20 A pores known as “ultramicropores” [16]. Such a combination of ultramicropores and micropores is believed to provide the combined molecular sieving function and high permeability characteristic of these unusual materials. The disordered structure of the carbon material is different from zeolites, which have a uniform, well defined set of pores. Despite the distribution of the ultramicropores, CMS materials offer the important advantage of facile formation of defect free (a)

(b)

(c)

(d) Micropore

Critical ultramicropore

m of pores

Critical ultramicropore

Micropore

Pore size FIGURE 7.1 Structures of carbon material and an idealized pore size distribution of carbon materials, adapted from Jenkins and Kawamura [8] and Steel and Koros [16]. In (b), the notations correspond to dc: ultramicropore dimension, dtv: size of adsorptive pore dimension, and dl: jump length dimension.

140

Inorganic, Polymeric and Composite Membranes

membranes for use in gas separation applications. Structures of CMS membranes have been investigated by many researchers using traditional techniques, such as X-ray diffraction (XRD), transmission electron microscopy (TEM), and adsorption. Unfortunately, due to the amorphous nature of CMS, it has been difficult to determine the structure, especially the ultramicropore region that governs the molecular sieving process. When XRD was performed, as shown in Figure 7.2, broad peaks were observed due to its amorphous structure. Therefore, the precision with which one can deconvolute information to estimate actual pore structure is poor [11,17–19]. The results of high resolution TEM were also inconclusive due to the amorphous structure, and the image hardly showed any pores of the material [10,16]. Chen, Loo, Wang, and Do conducted an argon adsorption isotherm to obtain a pore size distribution of CMS membranes, but the argon molecule was too large to analyze the selective ultramicropore region [20]. Similarly, Steel and Koros and Campo and Mendes concluded that the pore size distribution derived from CO2 adsorption equilibrium may not be enough to explain CMS pore structures that are responsible for molecular sieving [17,21]. A part of this chapter describes an investigation of the CMS structures, namely the applicable ultramicropore distribution, by using various gas molecules as probes. Details of the investigation are described in Section Molecular Ruler.

Effect of Pyrolysis Atmosphere on Separation Performance of CMS Membranes Review Gas separation performance of CMS membranes depends on the critical pore size and the pore size distribution. These properties are known to be affected by many factors, such as polymer precursor [18,22–25] and pyrolysis 800 3.8 Å

700

2.08 Å

600 500

800 °C, 2 h

400 300

5.7Å

200

550 °C, 2 h

100 6FDA/BPDA-DAM polymer

0 0

10

20

30

40

50

60

70

2q FIGURE 7.2 WAXRD of carbon material pyrolyzed with different temperatures using 6FDA/ BPDA-DAM polymer precursor [11].

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

141

temperature [18,25,26]. The effect of pyrolysis atmosphere has also been investigated and identified as one of the key factors. When vacuum and inert pyrolysis are compared, some researchers observed that an inert pyrolysis provides much higher permeability and lower selectivity [14,16,25]. To the contrary, Suda and Haraya reported that CMS films prepared from polyimide KaptonÒ at  1273 K in either argon or in high vacuum of  1.3  10 3 Pa showed negligible differences [10]. Geiszler and Koros conducted work on asymmetric hollow fiber CMS membranes using 6FDA/BPDA-DAM and reported negligible differences between fibers produced under argon, helium, and carbon dioxide inerts for both O2/N2 and H2/N2 separations [14]. When high versus low inert flows were compared, Geizler observed that the inert purge flow rate per se does not have a significant impact on a permeability of CMS membranes within the standard deviations of  20% typically observed for CMS membranes [14] formed under the same conditions. These various discrepancies reveal the need to identify at least one additional factor to unite and explain the rather disparate observations in this CMS field, and this chapter addresses this need.

Effect of Oxygen Exposure on Separation Performance To explain the discrepancies between reported studies regarding effects of the pyrolysis atmosphere, we propose a hypothesis regarding CMS formation. The hypothesis relates the formation mechanism to the amount of oxygen exposure during pyrolysis as the key missing factor controlling CMS properties. As mentioned in the section on carbon structure, CMS consists of irregularly packed sheets of sp2 hybridized carbon. The CMS structures illustrated in Figure 7.1 have micropores providing adsorption sites and ultramicropores acting as molecular sieve sites. These ultramicropores are speculated to be created at “kinks” in the carbon sheet or at the edge of a carbon sheet. These sites have more reactive unpaired sigma electrons prone to oxidation as compared to other sites in the membrane [27–29]. In fact, studies on the electrochemical oxidation of carbon give evidence that the rate of gas-phase oxidation of carbon on the edge plane is about 17 times greater than on the basal plane [27]. In addition, various characterizations have been performed by prior researchers to indicate the existence of chemisorbed oxygen. Ishiguro et al. showed FTIR spectrum that indicates existence of the C–O bonding group on carbon samples heated at  773 K [30]. When pyrolyzed in the presence of oxygen, it is hypothesized that oxygen reacts and binds to the reactive sites on CMS membranes at locations comprising the ultramicropores, thereby narrowing the average ultramicropore size. This process, in turn, is hypothesized to lower the permeability of the membrane and to increase the selectivity. A schematic drawing of this hypothetical process is depicted in Figure 7.3. Tests of the above hypothesis are discussed in the Results and Discussion section.

142

Inorganic, Polymeric and Composite Membranes

“Intrinsic” CMS

“Doped and stabilized” CMS following exposure at pyrolysis

Chemisorbed oxygen or oxidation at active ultramicropore sites

FIGURE 7.3 Schematic of oxygen “doping” process during pyrolysis.

The exact mechanism of the oxidation reaction that takes place during formation of pyrolytic carbon is not fully known, but as mentioned above, studies have shown that three stable oxides, CO, CO2, and surface oxides, such as carbonyls, are produced by oxygen molecules during the electrochemical oxidation on carbon [28,29,31]. de Soete showed measured adsorption and desorption rates of oxides at temperatures between 623 and 973 K using high surface area of active carbons, and concluded that the adsorption is the limiting step [32]. During CMS formation, the main source of oxygen in the pyrolysis system comes from the pyrolysis atmosphere. As the highly reactive pores on the surface are oxidized, oxygen diffuses into the membrane and reacts with the next available site. The limiting step of this complex carbon–oxygen reaction is investigated and hypothetical reaction mechanisms are proposed in a later section.

EXPERIMENTAL Materials Polymer Precursor Films The polymer precursor MatrimidÒ 5218 was provided by Vantico, Inc. and 6FDA/BPDA-DAM was synthesized in-house. The chemical structure and the characteristics of the polymer are shown in Table 7.1. The synthesis of 6FDA/BPDA-DAM was conducted via a polycondensation reaction by addition of the dianhydride, 5,50 -[2,2,2-trifluoro-1-(trifluoromethyl)ethylidene]bis-1,3isobenzofurandione (6FDA), and diamines, 2,4,6-trimethyl-1,3-phenylene diamine (DAM) and 3,30 ,4,40 -biphenyl tetracarboxylic dianhydride (BPDA), in solution using the solvent, n-methylpyrrolidone (NMP). All monomers were purified by sublimation prior to use. This sublimation was essential for the end product to have high molecular weight and lower polydispersity index (PDI). In this study, the reaction stoichiometry was adjusted to have the ratio of BPDA to DAM of 1:1. The polycondensation is sensitive to water, and minimal

TABLE 7.1 Chemical Structures and Characteristics of Polyimide Discussed in this Study Polymer

Chemical Structure

6FDA/BPDA-DAM O

CF3

CF3

N

O

O

CH3 N

N

O CH3

O

O

Tdecomp

627 K

723 K

575 K

698 K

CH3 N

O CH3

O

H3C

Tg

X

H3C Y

2,4,6-trimethyl-1,3-phenylene diamine (DAM), 3,30 ,4,40 -biphenyl tetracarboxylic dianhydride (BPDA), and 5,50 -[2,2,2-trifluoro-1-(trifluoromethyl)ethylidene]bis-1,3-isobenzofurandione (6FDA). The ration of X to Y is 1:1. BTDA-DAPI (MatrimidÒ)

H3C

CH3 O

H3C

O

O

N

N O

O

n

144

Inorganic, Polymeric and Composite Membranes

exposure to moisture was ensured. The reaction produced polyamic acid, which is the precursor to polyimide. Thermal imidization was chosen to dehydrate the polyamic acid to form a polyimide. The reaction solution was heated to a high temperature of  453 K in order for the imidization to occur [33]. Gel permeation chromatography (GPC) showed the polyimide product with molecular weight of 103,000 and PDI of 2.2 as a result of the polymer synthesis. Differential scanning calorimetry (DSC) was performed to determine the glass transition temperature (Tg) on both MatrimidÒ and 6FDA/BPDA-DAM polymers. Thermal gravimetric analysis combined with Fourier transform infrared spectroscopy (TGA-FTIR, provided from Netzsch, STA 409 PC Luxx TGA/DSC) was performed to characterize the decomposition process of polymers. On TGA analysis, the temperature, where a significant weight loss occurs, is termed as “decomposition temperature, Tdecomp.”

Formation of Polymer Precursor Films Homogeneous polymeric dense films were prepared by first drying the polymer powder in a vacuum oven at  383 K for at least 12 h to remove moisture. Immediately after removal from the oven, a polymer solution (3–5 wt%) was prepared by dissolving in dichloromethane (Aldrich) and by placing it on rollers for at least 12 h. After mixing, polymer dense films were prepared by a solution casting method in a glove bag at room temperature to achieve a slow solvent evaporation rate. After solvent was evaporated, films were removed from the casting setting and placed in the vacuum oven at 383 K for at least 12 h to remove residual solvent. Once the films were removed from the oven, they were cut into small disks with a diameter of 2.54 cm. Formation of Carbon Molecular Sieve Films The polymer films were placed on a corrugated quartz plate, which was ridged to allow for the diffusion of volatile by-products from the top and bottom of the films into the effluent stream and loaded into the pyrolysis setup. In order to maintain consistency, a pyrolysis temperature of 823 K and a 2-h soak time was used, which was the same temperature protocol used in the work by Geiszler and Koros [14] and Vu and Koros [13]. A pyrolysis system shown in Figure 7.4 was used to monitor oxygen concentration in the well-controlled system. A temperature controller (Omega Engineering, Inc., model CN8241 series, Stamford, CT) was used to heat a furnace (Thermocraft, Inc., model 23-24-1ZH, Winston-Salem, NC) and a quartz tube (National Scientific Co., 55 mm ID and 4 ft long, Quakertown, PA). An assembly of a metal flange with silicon O-rings (MTI Corporation, model EQ-FI-60, Richmond, CA) was used on both ends of the quartz tube. This sealing method was used to further reduce leaks when performing pyrolysis under vacuum. An oxygen analyzer (Cambridge Sensotec Ltd., Rapidox 2100 series, Cambridge, England,  1%

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

145

Micro needle valve Pressure transducer

Inert gas Mass flow

controller Quartz tube Pressure transducer Digital reader

O2 sensor Vacuum or vent

Thermocouple

Temperature controller

Data acquisition

FIGURE 7.4 Schematic diagram of a new pyrolysis system used to generate CMS membranes.

accuracy between 10 20 ppm and 100% [34]) was integrated to monitor oxygen concentration during the pyrolysis process. Pyrolysis atmospheres consisted of either continuous vacuum or continuous inert gas. For vacuum pyrolysis, a pump (Edwards, model RV3, Tewksbury, MA) was used to create a low pressure below 0.66 Pa, and a liquid nitrogen trap was used to prevent any back diffusion of oil vapor. The pressure inside the tube was monitored by a pressure transducer (MKS Instruments, 628B capacitance manometer, Andover, MA with 0.5% accuracy below 1 torr) attached to a readout (MKS Instruments, PDR2000). For experiments using purged gas during pyrolysis, the flow rate of the gas was controlled with a mass flow controller (MKS Instruments, typ. 247) and confirmed with a bubble flow meter (Fisher Scientific, model 520, Pittsburgh, PA) before and after each experiment. Between experiments, the quartz tube and plate were rinsed with acetone (Aldrich) and baked in air at 1073 K to remove any deposited materials which could affect consecutive runs.

Characterization Methods Permeation Once CMS films were prepared, they were immediately loaded into permeation cells. The carbon films were first masked using impermeable aluminum tape, and only a specific area was exposed for permeation. Five-minute epoxy (Devcon, Danvers, MA) was applied at the interface of the tape and the film to further minimize any gas leak. This assembly was placed in a double O-ring flange permeation cell. The cell was placed in a permeation system in which a constant-volume variable-pressure method [35,36] was applied. Both upstream and downstream of the permeation system was evacuated for at least 12 h and a leak rate was measured, which was always less than 1% of the permeate rate of the slowest gas. Once the whole system was evacuated, the upstream was pressurized with a testing gas while the downstream was maintained at vacuum, but isolated from the vacuum pump. The pressure rise in a standard volume on

146

Inorganic, Polymeric and Composite Membranes

the downstream was monitored with time by LabView (National Instruments, Austin, TX), and permeability was calculated using Equation (7.1). The system was again evacuated each time before experiments with different gases for at least 12 h.

Sorption Gas sorption measurements were made using a pressure decay method [37,38], where the equilibrium sorbed concentration at a given pressure can be used to calculate the solubility coefficient. Once the CMS films were loaded, the system was evacuated for  24 h. A feed reservoir was pressurized with a certain amount of gas allowing the system to equilibrate thermally. The entire system was kept in a heated water bath with a circulator to maintain constant temperature. Once the feed reservoir came to equilibrium, the pressure valve between the feed and the sample cell was opened and then quickly closed to introduce a dose of the feed gas into the cell. The pressure in both chambers was monitored with pressure transducers. The amount of gas sorbed was then calculated using a mole balance.

RESULTS AND DISCUSSION Correlation Between Oxygen Exposure and CMS Separation Performance 6FDA/BPDA-DAM CMS Membranes Correlation Between Oxygen Exposure and Consumption The previous section hypothesizes that the controlled oxidation of carbon at ultramicropores occurs during pyrolysis. We hypothesized to control the amount of “dopants” on carbon edges by varying total amount of oxygen exposure during pyrolysis. The total amount of oxygen is defined to be “total oxygen exposure coefficient, qO2, tot,” that expresses the amount of oxygen present to react during the pyrolysis process. Using the synthesized polymer 6FDA/ BPDA-DAM, a mixed gas of argon and a specific amount of oxygen in ppm level, supplied by Airgas, were used at a flow rate of 200 cm3(STP)/min to perform a control study on the amount of oxygen available during pyrolysis. All experiments were conducted at least three times to demonstrate repeatability. During each experiment, 0.12  0.02 mg of 6FDA/BPDA-DAM polymer films were pyrolyzed under specified atmospheres. Once the oxygen profiles were obtained, such as shown in Figure 7.5, the total amount of oxygen available, which is equivalent to the aforementioned parameter qO2, tot, and the amount of oxygen consumed were calculated using Equations (7.4) and (7.5).   Total amount of O2 (q (O2 ;totalÞ Þ ¼ Q_ purge  (ppm O2 Þfeed  t ð7:4Þ

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

147

Oxygen concentration (ppm) in pyrolysis system

70

60

50 ppm O2 /Ar inert

4 ppm O2 /Ar inert

30 ppm O2 /Ar inert

Vacuum

8 ppm O2 /Ar inert

50

40

30

20

10

0

0

100

200

300 400 Time (min)

500

600

700

FIGURE 7.5 Oxygen time profile during pyrolysis of 6FDA/BPDA-DAM. A specified amount of oxygen levels in argon gas was used as inert pyrolysis, and vacuum pyrolysis was conducted at 0.67 Pa.

Amount of O2 (q (O ;consumedÞ Þ 2

¼ Q_ purge 
 (ppm O2 Þfeed  (ppm O2 measured   t ð7:5Þ

The parameter Q_ purge is the volumetric flow rate of the “inert” gas containing the controlled amount of O2 impurity. For vacuum pyrolysis, Q_ purge was determined by a curve supplied by BOC Edwards and by direct measurement by bubble flow meter for inert pyrolysis. The values (ppm O2)feed and (ppm O2)measured are obtained by the oxygen analyzer at the entrance and exit of the pyrolysis chamber. The value (ppm O2)feed was the measured oxygen entering the pyrolysis chamber for both vacuum and inert pyrolysis prior to the initiation of the pyrolysis. A duration of 720 min, time from onset of heating to cooling, was used to make an initial attempt to observe oxygen consumption during pyrolysis of polymer membranes. The oxygen consumption profile in Figure 7.5 shows that the oxygen level prior to the pyrolysis process, between 4 and 30 ppm level, dropped to an order of 10 16 ppm once the temperature reached the polymer decomposition temperature either in inert or vacuum pyrolysis. The oxygen consumption slowly decreased when the temperature started to cool down and eventually stopped. Considering all pyrolysis atmospheres, the polymer precursor tends to consume

148

Inorganic, Polymeric and Composite Membranes

0.002 Vacuum pyrolysis

20

0.001

50 ppm O2/Ar 0

15

0.02

0

0.04

30 ppm O2/Ar

10

5

8 ppm O2/Ar

2

qO ,consumed (cm3(STP)/(g of polymer precursor))

25

4 ppm O2/Ar

0

0

10

20

30

40

50

60

qO2,tot (cm3(STP)/(g of polymer precursor)) FIGURE 7.6 A correlation between the total amount of oxygen and the amount of oxygen consumption normalized by weight of polymer precursors. A linear relationship was observed with R2 ¼ 99.7%.

more oxygen as the amount of oxygen fed increased. This is illustrated as the linearity in Figure 7.6 between the amount of consumed, qO2 ;consumed , and the total amount of oxygen available for consumption, qO2 ;tot We speculated that a majority of the oxygen supplied is consumed by the by-products evolved during pyrolysis process since the polymer precursor loses 40–45% of the initial weight as a result of pyrolysis [16]. This indicates that the relatively large amount of “total amount of oxygen available” was consumed by the byproducts, rather than as “dopants.” In addition, TGA analysis by Steel showed that most of the weight loss occurred just before the temperature protocol reaches the 2 h soak period [16]: it is speculated that majority of the by-products evolve before the 2-h soaking period and consume a majority of the oxygen supplied in the bulk flow. Once the by-products flow out of the pyrolysis system, oxygen is speculated to be used mainly to treat the active carbon edges. Correlation Between Total Amount of Oxygen Exposure and CMS Separation Performance Figure 7.7 shows separation properties of homogeneous dense CMS films pyrolyzed under each pyrolysis condition. Each data point on the figure represents an average of at least three CMS dense films, and standard deviations of less than  10% in both permeability and selectivity were achieved. It was

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

149

1000

CO2 /CH4 selectivity

Polymer precursor CMS (vacuum pyrolysis) CMS (inert pyrolysis) Robeson curve 08

100

30 ppm O2/Ar

8 ppm O2/Ar 4 ppm O2/Ar

10

1 10-8

50 ppm O2/Ar

10-7 10-6 CO2 permeance (mol m-2 s-1 Pa-1)

10-5

FIGURE 7.7 Separation performance of 6FDA/BPDA-DAM CMS dense films. All were tested at 308 K. A circular data point represents polymeric properties, and rectangular points represent properties of CMS membranes. Pyrolysis atmospheres are indicated with data points for inert pyrolysis. A dash line represents a trend between separation performance and oxygen concentration prior to pyrolysis. The Robeson curve shows the trade-off from currently available polymer [5]. The unit permeability was converted into an unit of mol m 2 s 1 Pa 1 permeance with an assumed thickness of 1 mm.

demonstrated that CMS films pyrolyzed under inert gases with an oxygen concentration in a range of 4–30 ppm perform as attractively as CMS films pyrolyzed under vacuum, well exceeding the upper bound curve. This finding leads us to believe that a selectivity as high as was obtained under vacuum pyrolysis can be accomplished by tuning the oxygen exposure. Figure 7.7 clearly shows that a strong relationship exists between the total amount of oxygen and the transport properties for inert pyrolysis. Performance differences between the various “inert” atmospheres are especially revealing. Selectivity increases and permeability decreases as the amount of oxygen in the inert gas increases. However, there is a cut-off point on the benefit to the amount of oxygen exposure such that once it exceeds a critical level, low values of both selectivity and permeability are observed. This phenomenon is shown clearly by data with 50 ppm O2/Ar inert pyrolysis, possibly suggesting that the amount of oxygen was so high in this case that most of the oxygen used for the purpose of “doping” may have filled most of the “active” sites of ultramicropores, reducing both selectivity and permeability. This hypothetical explanation can also be understood in terms of a shift in the ultramicropore size distribution as shown in Figure 7.8. Specifically, the few remaining pathways in such a case are hypothesized to be larger and less susceptible to selectivity

Number of pores

150

Inorganic, Polymeric and Composite Membranes

Intrinsic structure

Optimal qO ,tot 2

qO ,tot slightly too high 2

Ultramicropores accessible to molecules CO2 CH4

CO2 CH4

CO2 CH4

O2 “Doping”

Overdoping

Pore size

Undoped

FIGURE 7.8 Hypothetical ultramicropore distribution along with CMS slit like structures: (a) undoped “intrinsic” structure when pyrolyzed under no oxygen presents, (b) optimal selective structure with adequate amount of oxygen doped, (c) overdoped structure when slightly higher oxygen was introduced during pyrolysis.

enhancement by the oxygen chemisorption mechanism described earlier. In such a case, the permeability decreases as the total amount of oxygen exposed increases, but eventually selectivity actually drops due to only less selective paths remaining open for transport. The previous section discussed a correlation between total amount of oxygen available and oxygen consumed. One may note a discrepancy between the total amount of oxygen available and the CMS separation performance when inert and vacuum pyrolysis atmospheres are compared. According to Figure 7.6, vacuum pyrolysis supplied the lowest total amount of oxygen available. Theoretically, this would result in the most permeable and least selective CMS membranes when compared with other pyrolysis experiments conducted in this work; however, the experimental results show that the separation performance of CMS membranes via vacuum pyrolysis lies between CMS membranes pyrolyzed under 8–30 ppm oxygen level. The oxygen profile in Figure 7.5 shows the oxygen level prior to the vacuum pyrolysis to be about 10 ppm. The oxygen level measured in this work under the vacuum is equivalent to an oxygen concentration measured at 1 atm pressure; that is, a total pressure of 0.66 Pa with 21% O2 is equivalent to an oxygen concentration of 10 ppm at 105 Pa. This suggests the possibility of the oxygen concentration being a governing factor of “doping” process, rather than total amount of oxygen exposed: further investigations were conducted and discussed in a later section.

MatrimidÒ CMS Membranes Similar to the experiments conducted on 6FDA/BPDA-DAM polymer, the effect of oxygen exposure was investigated experimentally on commercially available polymer, MatrimidÒ. The polymer precursor was pyrolyzed in the same manner under (i) vacuum and (ii) an inert gas of argon with

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

151

3–100 ppm level oxygen. A correlation between the total amount of oxygen available and the amount of consumption existed as illustrated in Figure 7.6 for 6FDA/BPDA-DAM. This section further seeks to verify the oxygen “doping” process by conducting FTIR analysis and evaluating separation performance on CMS membranes with various oxygen exposures. FTIR Figure 7.9 shows the FTIR result from CMS samples prepared with inert pyrolysis. While typical CMS membranes show no significant peaks [15], the spectra in Figure 7.9 shows C¼O group appearing in the vicinity of 1700 cm 1, indicating that the oxygen “doping” method was successfully applied to CMS membranes.

5.00

CMS (100 ppm O2/Ar pyrolysis) 0.00 5.00

Absorbance (a.u.)

CMS (50 ppm O2/Ar pyrolysis) 0.00 5.00

CMS (30 ppm O2/Ar pyrolysis) 0.00 5.00

CMS (10 ppm O2/Ar pyrolysis) 0.00 5.00

CMS (3 ppm O2/Ar pyrolysis) 0.00 2400

2200 2000 1800 Wavenumber (cm–1)

FIGURE 7.9 FTIR spectra of MatrimidÒ CMS membranes [39].

1600

152

Inorganic, Polymeric and Composite Membranes

Correlation Between Oxygen Exposure and CMS Separation Performance The result of CO2/CH4 separation performance is shown in Figure 7.10. It demonstrates that the attractive separation performance, which is above the upper bound curve, can be achieved with pyrolysis under (i) vacuum and (ii) inert pyrolysis when the pyrolysis atmosphere is properly controlled. These results are consistent with the observation with 6FDA/BPDA-DAM that oxygen exposure has a correlation with separation performance. The results on Figure 7.10 show that the oxygen doping method caused decreases in both permeability and selectivity. This trend is different from the trend the authors observed with the 6FDA/BPDA-DAM CMS membranes (shown on Figure 7.7): we speculate that the discrepancy is caused by the difference in intrinsic CMS structures for the two samples. We hypothesize that MatrimidÒ CMS membranes, created from an intrinsically lower free volume precursor, have a less sorptive and less selective intrinsic structure than 6FDA/BPDA-DAM CMS membranes. Therefore, oxygen doping not only reduces pore sizes of ultramicropore and micropores but also further closes selective pores, leading to a decrease in both permeability and selectivity. This hypothesis was tested by several characterizations as discussed in the next section. 1000 Polymer precursor CMS (vacuum pyrolysis)

CO2/CH4 selectivity

CMS (inert pyrolysis O2 concentration in Ar purge listed) Robeson curve (2008)

100

1 ppm 3 ppm 10 ppm

10

1 10−9

30 ppm 50 ppm 100 ppm

10−6 10−8 10−7 CO2 permeance (mol m−2 s−1 Pa−1)

10−5

FIGURE 7.10 Separation performance of MatrimidÒ CMS dense films. Tests were conducted at 308 K with an upstream pressure of 3.5  105 Pa. A circular data point represents polymeric properties, and square points represent properties of CMS membranes. The arrow line represents a trend between separation performance and oxygen concentration during pyrolysis [39]. The unit permeability was converted to the unit permeance (mol m 2 s 1 Pa 1) with an assumed thickness of 1 mm.

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

153

Structural Differences Between 6FDA/BPDA-DAM and MatrimidÒ CMS Membranes TGA-FTIR The decomposition process of each polymer was investigated using TGA-FTIR [39]. Polymeric samples were heated under argon purge, and the evolved gases were sent to an FTIR chamber to analyze the chemical composition. Figures 7.11 and 7.12 show decomposition results of 6FDA/BPDA-DAM and MatrimidÒ. Polymer 6FDA/BPDA-DAM produced HF (4250–4500 cm 1), CH4 (3017 cm 1), CHF3 (1150, 1178 cm 1), CO2 (2110 cm 1), and CO (2190 cm 1). However, MatrimidÒ produced mainly CH4 (3017 cm 1), CO2 (2110 cm 1), and CO (2190 cm 1). There are three stages of pyrolysis when polymers are heated up to 1473 K: (i) the precarbonation, (ii) the carbonization, and (iii) dehydrogenation [8]. A temperature range between 373 K and up to the polymer decomposition temperatures, that is, 723 K for 6FDA/BPDA-DAM and 698 K for MatrimidÒ, would be considered as the precarbonation stage (see Table 7.1). During this first stage, molecules, such as excess monomer and solvent, are removed. Polymeric films turn black and linear conjugated C–C systems start to form near the decomposition temperature [8]. In the carbonization stage, rapid weight loss is observed due to the removal of entities, such as oxygen, nitrogen, and CF3 [8]. As Figures 7.11 and 7.12 show, major by-products evolve in this regime. An exact temperature range of this stage is difficult to define. However, it is speculated to be between the decomposition temperature and the temperature where the rate of weight loss is significantly reduced (an elbow of the decomposition weight loss curve). Evolution of CO, CO2, and CH4 were also observed for FIGURE 7.11 TGA-FTIR result of 6FDA/BPDA-DAM [15]. Y-axis in arbitrary intensity units.

CHF3 0.3 HF

CO/CO2

0.2 0.1 0

8000 6000 4000

–0.1 1000

s nd

co

Se

2000

2000

3000 4000 5000

W

-1 )

cm

ber (

um aven

154

Inorganic, Polymeric and Composite Membranes

0.10 0.05 0.00 –0.05

8000

ds

on

c Se

6000

–0.10

4000 1000

2000

2000 3000 4000

1 m- )

ber (c

num Wave

FIGURE 7.12 TGA-FTIR result of MatrimidÒ [41]. Y-axis in arbitrary intensity units.

MatrimidÒ by Barsema et al [40]. At the end of this stage, a loose network of linear conjugated systems is formed [8]. In the dehydrogenation stage, hydrogen is gradually eliminated, typically between 773 and 1473 K. Jenkins and Kawamura note that the rate of the hydrogen removal is a characteristic of a given heat-treatment temperature [8]. Elemental analysis on carbon membranes pyrolyzed in similar manners shows 95–99% of aromatic carbon content [10–13]; moreover, the percentage of carbon element is dependent on pyrolysis temperature [8]. As the large fluorinated compounds are produced and diffuse out of membrane films, more open ultramicropore structures are believed to be formed with 6FDA/BPDA-DAM than with MatrimidÒ, which lacks these fluorinated moieties. Sorption Isotherms Gas sorption was examined to characterize sorption coefficients, hole filling capacity, and affinity constant [39]. Each polymeric film was pyrolyzed under a flow of 200 cc/min 1 ppm O2/Ar to produce CMS membranes. The lowest concentration of oxygen in argon mixture available was chosen to prepare CMS membranes with structures closest to their “intrinsic structures.” The sorption coefficients depend on the micropore sizes of carbon material, the critical temperature of the molecule, which reflects condensability and affinity of penetrants for the carbon. If the ultramicropore size is small enough to effectively exclude one gas molecule while transmitting the smaller molecule in a pair, true molecular sieving occurs. In this case, no uptake of the sieved

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

155

component would be shown on the sorption isotherm. For carbon material, however, a distribution of ultramicropore exists. Therefore, only regions connected with perfect molecular sieving pore windows would be inaccessible to a larger gas molecule. Figures 7.13 and 7.14 show sorption isotherms measured with six gases, He, CO2, O2, N2, CH4, and SF6. The sorption isotherms were fitted to the Langmuir equation model in Equation (7.3). The hole filling capacities, CH0 , and affinity constants, b, are shown in Tables 7.2 and 7.3, along with the kinetic diameters of the test gases. In all gas measurements, 6FDA/BPDA-DAM CMS membranes showed higher sorption coefficients and hole filling capacities than MatrimidÒ CMS membranes. This indicates a larger available micropore volume in 6FDA/ BPDA-DAM CMS membranes than for MatrimidÒ CMS membranes. The ratio of (CH 0 )CO2 values for 6FDA/BPDA-DAM versus MatrimidÒ CMS membranes is about 1.3. This ratio is similar to the reported value of 1.4 for the ratio of micropore volume for the 6FDA/BPDA-DAM versus Matrimid CMS samples by DFT analysis in Steel and Koros [21], which is consistent with our data. The trend of isotherms for both sets of CMS membranes is similar. The similarity can also be seen with the order of magnitudes on Langmuir affinity constants in both CMS membranes. This is likely caused by the fact that CMS samples were treated in the same manner and resulted in similar overall gross structure. As previously mentioned, pyrolysis treatment above 823 K results in more than 95% carbon in the final structure [10,11], and the percentage

100

C (cm3(STP)/cm3 CMS)

80

60

40

20

0 0

500

1 ´ 103 1.5 ´ 103 Pressure (kPa)

2 ´ 103

FIGURE 7.13 Sorption isotherms for 6FDA/BPDA-DAM CMS membranes prepared with 1 ppm O2/Ar inert pyrolysis. The isotherms were obtained with six different gases: ○ CO2, d CH4, ■ O2, □ N2, D He, m SF6. The experiments were repeated and fitted with Langmuir isotherm model described in dot lines [39].

156

Inorganic, Polymeric and Composite Membranes

100

C (cm3(STP)/cm3 CMS)

80

60

40

20

0 0

500

1.5 ´ 103 1 ´ 103 Pressure (kPa)

2 ´ 103

FIGURE 7.14 Sorption isotherms for MatrimidÒ CMS membranes prepared with 1 ppm O2/Ar inert pyrolysis. The isotherms were repeated and obtained with six different gases: ○ CO2, d CH4, ■ O2, □ N2, D He, m SF6. The experiments were repeated and fitted with Langmuir isotherm model described in dot lines [39].

TABLE 7.2 Langmuir Hole Filling Capacity CH0 and Langmuir Affinity Constant b Calculated Based on 6FDA/BPDA-DAM CMS Sorption Isotherms [39] Gas

Kinetic diameter (A˚) [55]

CH0 (cc(STP)/cc CMS)

b (MPa 1)

He

2.6

13.0

0.73

CO2

3.3

98.0

8.70

O2

3.46

72.2

0.87

N2

3.64

69.4

1.45

CH4

3.8

78.0

3.63

SF6

5.5

12.7

0.29

presumably depends upon the intensity of heat treatment [8]. Despite its small molecular size, the sorption coefficients for He were significantly lower than other gases. This is attributed to its noncondensable nature. However, the lowest sorption coefficients observed for both CMS materials were found for the highly condensable gas, SF6. It may be caused by the small population of ultra˚ . In fact, bimodal pore size micropores in the range between 5 and 6 A

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

157

TABLE 7.3 Langmuir Hole Filling Capacity CH0 and Langmuir Affinity Constant b Calculated Based on MatrimidÒ CMS Sorption Isotherms [39] Gas

Kinetic Diameter (A˚) [55]

CH0 (cc(STP)/cc CMS)

b (MPa 1)

He

2.6

11.7

0.44

CO2

3.3

74.3

2.90

O2

3.46

68.5

0.87

N2

3.64

27.1

1.45

CH4

3.8

57.9

0.73

SF6

5.5

11.6

0.07

distributions seen by various researchers suggest that the minimum for typical CMS materials lies around the size of an SF6 molecule [17,41,42]. This also explains the order of Langmuir affinity constants in both CMS membranes being CO2 > CH4 > N2 > O2 > He > SF6, since sorption of a large spherical molecule SF6 is simply too low to enable accurate measurement of the affinity constants for typical micropores. A comparison of R2 values for the affinity constants on SF6 versus other gases shows poor fitting of the model:  0.6 and 0.95, respectively, providing further evidence of this observation. “Molecular-ruler” Pore size distributions in the ultramicropore region were investigated in this study using various gas molecules as probes [39]. Diffusion coefficients were obtained from permeation and sorption experiments using Equation (7.2). Similar to sorption isotherm experiments, CMS membranes were prepared under inert pyrolysis of 200 cc/min 1 ppm O2/Ar to produce CMS membranes with close to “intrinsic” structures. Once samples were prepared, the experiments were conducted at 308 K with a pressure of 3.5  105 Pa using ˚ ), CO2 (3.3 A ˚ ), O2 the test gases listed in Table 7.2, namely He (2.6 A ˚ ˚ ˚ ˚ (3.46 A), N2 (3.64 A), CH4 (3.80 A), and SF6 (5.5 A). Tables 7.4, 7.5 and 7.6 list permeability, sorption and diffusion coefficients, respectively for 6FDA/BPDA-DAM and MatrimidÒ CMS membranes. Permeabilities of 6FDA/BPDA-DAM CMS membranes are higher than for MatrimidÒ CMS membranes; possibly due to higher free volume of the polymer precursor and resultant larger CH0 as a result of evolution of CF3 group during the heating process as seen on TGA-FTIR. Table 7.6 confirms that transport in CMS membranes is not sorption dominant with regard to selectivity. For CO2/ CH4 separation, sorption selectivity is less than half of diffusion selectivity. In the case of O2/N2 separation on both 6FDA/BPDA-DAM and MatrimidÒ CMS

158

Inorganic, Polymeric and Composite Membranes

TABLE 7.4 Permeability of 6FDA/BPDA-DAM and MatrimidÒ CMS Membranes Pyrolyzed under 1 ppm O2/Ar Inert Gas [39] Polymer Precursor

He

CO2

O2

N2

CH4

SF6

6FDA/BPDA-DAM

2530

7170

1530

204

247

0.6

605

1049

301

63

17

0.13

Ò

Matrimid

Tests were conducted at  308 K with a pressure of  7.3 kPa. Units are in barrer with percentage deviation of less than 10%.

TABLE 7.5 Sorption Coefficients of 6FDA/BPDA-DAM and MatrimidÒ CMS Membranes in ccSTP/(ccCMS-psia) [39] Polymer Precursor

He

CO2

O2

N2

CH4

SF6

6FDA/BPDA-DAM

0.049

1.41

0.43

0.46

0.82

0.018

0.031

1.28

0.37

0.18

0.23

0.006

Ò

Matrimid

Experiments were repeated and had less than 10% of deviation.

TABLE 7.6 Diffusion Coefficients of 6FDA/BPDA-DAM and MatrimidÒ CMS Membranes in 10 8 cm2/s [39] Polymer Precursor

He

CO2

O2

N2

CH4

SF6

6FDA/BPDA-DAM

2680

262

237

22.8

15.5

1.72

1020

42.4

42.1

18.0

3.85

1.14

Ò

Matrimid

Experiments were repeated and had less than 10% of deviation.

membranes, the sorption selectivity is in the range between 0.9 and 1, which is similar to results observed by Singh and Koros [43]. The results of diffusion coefficients suggest that 6FDA/BPDA-DAM CMS membranes have a larger average number of accessible ultramicropore windows than MatrimidÒ CMS membranes. For all tested gases, diffusion coefficients are much higher for 6FDA/BPDA-DAM CMS membranes than that of MatrimidÒ CMS membranes. For instance, DHe is more than double, and DCO2 is more than six times higher than MatrimidÒ CMS membranes. In addition, diffusion selectivity indicates that 6FDA/BPDA-DAM CMS membranes have more selective pore structures. For CO2/CH4 separation, the diffusion selectivity is  17 on 6FDA/BPDA-DAM CMS membranes while it is  11 for MatrimidÒ CMS

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

159

membranes. By interpreting results of diffusion coefficients, effective semiquantitative pore size distributions for the ultramicropore region were constructed and are shown in Figure 7.15. The distributions were drawn to match the ratio of diffusion coefficients relative to the area of accessible ultramicropores for each respective molecule for the gas separations among CO2, O2, N2, CH4 gases. In addition, the total area under the curve was adjusted to be about 2.6 times larger for 6FDA/BPDA-DAM CMS membranes than that of MatrimidÒ CMS membranes to reflect the relative diffusion coefficients of He in the two polymers. This is based on an assumption that He samples all pores accessible to any gas molecule in both CMS membranes from the two precursors. Specifically, the overall shape of the curve was built with an assumption that the number of ultramicropores that are accessible to the SF6 molecule provides a useful metric of the minimal number of large size ultramicropores to which a value of unity was assigned. Then the area which represents number of additional ultramicropores accessible for the rest of the gas molecules was scaled to be proportional to the diffusion coefficients. Finally, distribution curves were drawn to satisfy the ratios of diffusion coefficients among challenging separation gas pairs: He/N2, CO2/CH4, and ˚ ) and CO2 (3.3 A ˚) O2/N2. The shape of the distribution between He (2.6 A was drawn for convenience with uncertainty and is not important for this discussion. Clarification of the shape in this range requires additional work with ˚ , such as a neon molecule. This diffusion coeffiprobes between 2.6 and 3.3 A cient-based ultramicropore size distribution suggest that 6FDA/BPDA-DAM

Relative number of ultramicropores

6FDA/BPDA-DAM CMS Matrimid® CMS

2.6 He

3.0

3.3 3.46 3.64 3.8 4.0 CO2 O2

N2 CH4

5.0

5.5 SF6

Pore size (Å) FIGURE 7.15 Diffusion coefficient based ultramicropore size distribution. The dot line represents the distribution of 6FDA/BPDA-DAM CMS membrane, and the solid line represents the distribution of MatrimidÒ CMS membranes. The x-axis is linearly scaled with an indication of molecule’s kinetic diameters [39].

160

Inorganic, Polymeric and Composite Membranes

CMS membranes have a larger number of large pores and a more selective pore structure than MatrimidÒ CMS membranes.

Effect of Pyrolysis Temperature Previous sections have shown that MatrimidÒ CMS membranes have closed and less selective intrinsic pore structures that result in a decrease in both permeability and selectivity with an increase in oxygen exposure. In order to utilize the oxygen doping method on the MatrimidÒ precursor, experiments were designed to create CMS membranes with more open intrinsic pore structures. Researchers have shown that higher temperatures tend to produce more selective yet less permeable CMS membranes [16,25,26] presumably due to systematic relaxation of the CMS matrix. This suggests that higher pyrolysis temperatures result in more selective and less permeable CMS structure. In principle, lowering the pyrolysis temperature should lead to more open intrinsic CMS structure so that one can take advantage of the oxygen doping method. In order to test this hypothesis, a slightly lower pyrolysis temperature of 773 K with 1 ppm oxygen in argon gas was chosen to demonstrate the effect of temperature and the doping process. Results are shown in Figure 7.16. As predicted, the lower pyrolysis temperature produces more permeable but less selective CMS membranes with “near-intrinsic” structures when CMS 1000 CMS (500 °C pyrolysis)

CO2/CH4 selectivity

CMS (550 °C pyrolysis)

30 ppm

100 50 ppm

1 ppm 3 ppm

10

1 10−9

100 ppm

10−8 10−7 10−6 CO2 permeance (mol m–2 s–1 Pa–1)

10−5

FIGURE 7.16 Separation performance of MatrimidÒ CMS membranes pyrolyzed by different temperatures with oxygen “doping” process. A dot line represents a trend observed for CMS films pyrolyzed at 723 K. All data were repeated and had deviation of less than 10%. The unit permeability was converted into an unit mol m 2 s 1 Pa 1 permeance with an assumed thickness of 1 mm.

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

161

membranes are exposed to low amounts of oxygen. In addition, exposure to higher oxygen concentration enables use of the “doping” method, since the selectivity enhancement was more than double from 1 ppm O2/Ar to 30 ppm O2/Ar.

Correlation Between Oxygen Concentration and CMS Separation Performance The method above allows one to enhance membrane performance by > 100 times in permeability with doubled selectivity when CMS pyrolysis atmosphere is optimized, compared with separation performance of polymeric precursors [39,44]. Although the total amount of oxygen exposure was hypothesized to govern effectiveness and amount of oxygen chemisorbed at the selective pore windows, we have refined this version to suggest that the oxygen–carbon reaction at the edges may be equilibrium limited, rather than kinetically controlled by observations from Figures 7.7 and 7.10. Oxidation of carbon is complex due to the fact that the oxygen–carbon reaction involves chemical kinetics with consequent heat transfer and mass transfer processes at a number of levels. Gaseous oxygen molecules move from the surrounding bulk atmosphere to the carbon surface and are adsorbed to form surface intermediates, which may rearrange, desorb, and return to the gas phase [28,29,45]. Moreover, emitted decomposition by-products can be reacted in the external bulk gas phase. This section seeks to identify the limiting factor of the doping process during formation of CMS membranes.

Review Three regimes are considered during the oxidation process: (i) oxygen fully penetrates the solid and all active sites are available for the reaction, (ii) oxidant penetration is partial and oxygen diffusion into the solid is insufficient to supply all reaction needs, and (iii) reaction only takes place at the outer surface [46,47] for the available contact time. Investigation of the influence of diffusional resistance relative to reaction resistance can be carried out using a fundamental analysis of the Thiele modulus, y, which is defined as:   d sP kb 1=2 ð7:6Þ y¼ 2 De sOX where d is the particle diameter in meters, sP is its density in kg/m3, k is the carbon reaction rate in kg/kg-s, b is the stoichiometric mass ratio of oxygen to carbon, De is the effective diffusion coefficient in m2/s, and sOX is the bulk gas concentration of oxygen in kg/m3 [48]. The Thiele modulus is applied by means of an effectiveness factor, which varies between 0 and 1 [48,49]. The effectiveness factor represents the fraction of internal surface which can react when exposed to the surface concentration of reactant gas [49]. When the Thiele

162

Inorganic, Polymeric and Composite Membranes

modulus is larger than 1, the reaction rate hinders the ability of diffusion to supply oxygen to the reactive surface, and the effectiveness factor is much smaller than 1. When the Thiele modulus is smaller than unity, there is no resistance to pore diffusion. In terms of formation of CMS membranes, factors like pyrolysis temperature, film thickness, and pore structure strongly influence the process and determine the three regimes described earlier. A preliminary calculation can be made to identify one of three oxidation process regimes from the Thiele modules in Equation (7.6). The bulk oxygen concentration, sOX, of interest in our work is in the range from 1–100 ppm, which corresponds from 1.41  10 6 to 1.4110 4 kg m 3. The density, sP, of CMS membranes is reported to be in the order of 103 kg m 3, depending on polymeric precursors [16]. At temperatures of 773–1073 K, where typical CMS membranes are produced, carbon oxidation produces CO as the dominant product [50]. According to Stanmore et al., this results in a b value of 4/3 [45]. Effective diffusion coefficients of oxygen in CMS membranes at high temperatures are in the order of 10 8 m2 s 1, according to Singh [11]. According to de Soete, the reaction is in the first order when oxygen adsorption dominates, and overall reaction rate can be in the range of 0.5–0.9 s 1 [32,51]. When either homogeneous dense films or asymmetric hollow fibers have a selective skin thickness of d ¼ 0.1–30 mm, calculation of the Thiele modulus based on these literature values results in a value of 0.6–2500. Since this range clearly spans values less than and greater than one, it suggests the possibility of pore diffusion resistance control of the oxidation reaction. Nevertheless, there is considerable uncertainty in the above parameters, so this study probes this issue experimentally in detail. While the above theoretical calculation suggests possible pore diffusion resistance of oxygen during high temperature pyrolysis in carbon membranes, some researchers have seen inconsistent separation performance of CMS membranes pyrolyzed under atmosphere in which oxygen exists for possible oxidation reaction. Variations have been reported in CMS separation performance with respect to polymer precursor thickness [15], and researchers have seen separation performance independent of inert flow rates as mentioned in the Review section [14,15,52] covering the effect of pyrolysis atmosphere on separation performance of CMS membranes, so observations and theoretical calculation need to be rationalized. Therefore, we seek to draw a more definite conclusion by identifying the controlling factor in the oxygen doping process by a series of well-controlled experiments, specifically, by means of oxygen consumption during pyrolysis and separation performance of the resulting CMS membranes.

Effect of Thermal Soak Time Effect of thermal soak time on the oxygen doping process was investigated by comparing oxygen consumption during pyrolysis and the separation performance between CMS membranes prepared at 823 K with two different thermal soak times of 2–8 h. First, 6FDA/BPDA-DAM polymer precursor was

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

163

pyrolyzed with an 8-h soak time under 200 cc(STP)/min inert flow of 7 ppm oxygen in argon gas. The result was compared with data of a 2-h thermal soak time previously reported [44]. Table 7.7 shows amounts of oxygen available and consumed during the pyrolysis process. While the total amount of oxygen available during the 8-h soak time is higher than during the 2-h soak, the consumption amount stays almost the same. Moreover, Figure 7.17 shows that the data falls within the range of correlation observed in the previous study between

TABLE 7.7 Normalized Values of the Total Amount of Oxygen at Different Soak Time Obtained with 6FDA/BPDA-DAM Thermal Soak Time (h)

Total O2 Available (ccSTP/g)

Total O2 Consumed (ccSTP/g)

2

11.8

5.24

8

12.5

6.75

Inert pyrolysis under 200 cc/min of 7 ppm O2/Ar was conducted for 8 h soak time while the other was pyrolyzed for 2 h 8 ppm O2/Ar. Experiments were repeated and have a standard deviation of less than 10%.

qO2,consumed (cm3(STP)/(g of polymer precursor))

25

20 50 ppm O2/Ar 2h

15

30 ppm O2/Ar 2h 7 ppm O2/Ar 8h

10

8 ppm O2/Ar 2h 4 ppm O2/Ar 2h

5

0

0

10 20 30 40 50 qO2,tot (cm3(STP)/(g of polymer precursor))

60

FIGURE 7.17 A correlation between the total amount of oxygen and the amount of oxygen consumption normalized by weight of polymer precursors. Circles (d) represent data of 2-h thermal soak time from previous study [44], and the rectangular (□) represents data of 8-h thermal soak time. Inert compositions are listed along with data points. The data with longer thermal soak time falls within the trend seen with 2-h soak time [56].

164

Inorganic, Polymeric and Composite Membranes

the total amount of oxygen and the amount of oxygen consumption. This result indicates that a longer thermal soak time led to only a slight increase in the oxygen consumption. In addition, the separation performance of 6FDA/BPDA-DAM CMS membranes from two thermal soak periods was evaluated. As shown in Figure 7.18, CMS membranes prepared with an 8-h soak time have a slightly higher CO2 permeability and lower CO2/CH4 selectivity compared with that of a 2-h soak time. Based on these facts, it was speculated that the oxygen concentration may play a more major role during the doping process with oxygen than the period of soaking. The effect of thermal soak time on MatrimidÒ as a polymer precursor was also investigated. A pyrolysis atmosphere of 200 cc(STP)/min inert flow with  30 ppm oxygen in argon was used. Oxygen consumption and separation performance results are shown in Table 7.8 and Figure 7.19, respectively. The results show that samples with a shorter soak time consumed a smaller amount of oxygen compared with a longer one; however, Figure 7.19 indicates that their separation performances are similar. During the MatrimidÒ pyrolysis process, we observed large amounts of tan colored by-products adsorbed on the pyrolysis tube wall that we did not see during the pyrolysis of 6FDA/BPDA-DAM. Based 1000

CO2 /CH4 selectivity

4 ppm O2 /Ar 8 ppm O2 /Ar 30 ppm O2 /Ar 50 ppm O2 /Ar 7 ppm O2/Ar 8 h Robeson curve (2008) 30 ppm

O2/Ar

100

7 ppm O2/Ar 8 h

4 ppm O2/Ar

50 ppm O2/Ar

10

1 10−8

8 ppm O2/Ar

10−7

10−6

10−5

CO2 permeance (mol m−2 s−1 Pa−1) FIGURE 7.18 Separation performance of 6FDA/BPDA-DAM CMS films. A filled circle (d) represents CMS pyrolyzed with 8-h thermal soak time, and open circles (○) represent CMS pyrolyzed with 2-h soak time. Each was repeated at 308 K [56], and CO2 permeability was converted into an unit mol m 2 s 1 Pa 1 permeance with an assumed thickness of 1 mm.

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

165

TABLE 7.8 Normalized Values of the Total Amount of Oxygen at Different Soak Time Obtained with MatrimidÒ Thermal Soak Time (h)

Total O2 Available (ccSTP/g)

Total O2 Consumed (ccSTP/g)

2

154.2

50.0

8

216.1

90.2

Inert pyrolysis under 200 cc/min of  30 ppm O2/Ar was conducted in both cases. Experiments were repeated and have a standard deviation of less than 10%.

40

200 cc/min 2h

CO2 /CH4 selectivity

30

20

200 cc/min 8h

10 10−8

10−7 CO2 permeance (mol m−2 s−1 Pa−1)

10−6

FIGURE 7.19 Separation performance of MatrimidÒ CMS films pyrolyzed with two different thermal soak times of 2 h (d) and 8 h (○). The thickness of the films was 80 mm. Inert carrier of 30 ppm O2/Ar was used with a flow rate of 200 cc(STP)/min. All was repeated at 308 K [56].

on this observation, we speculate that the difference in oxygen consumption could be caused by the oxidation of the by-products, which has little effect on the membrane properties. A combination of the oxygen consumption and the separation performance properties indicate that the effect of duration of oxygen exposure to the separation performance is relatively small. This led us to hypothesize that the oxygen doping process during the pyrolysis to produce attractive CMS membranes is governed by oxygen concentration rather than total amount of oxygen exposed. A series of well-controlled experiments was conducted to identify the limiting

166

Inorganic, Polymeric and Composite Membranes

factor on the oxygen “doping” process to support this hypothesis. Three limiting factors are considered regarding the oxygen doping effect in dense films: external transport, internal transport, and chemical reaction. They were studied with respect to oxygen consumption and CO2/CH4 separation performance.

Effect of Inert Flow Rate: External Transfer Limitation Researchers have shown that inert flow rate during pyrolysis does not affect separation performances [14,15]. These studies showed that the external transport is not the limiting factor in determining the resulting CMS membrane performance. In order to demonstrate this phenomenon in terms of the oxygen exposure, two sets of experiments were conducted as shown in Figure 7.20. A gas mixture of 30 ppm of oxygen in argon was used as an “inert” with two different flow rates: 50 and 200 cc(STP)/min. Two polymeric films (total of  0.02 g) were pyrolyzed with the thermal protocol of 823 K and a 2-h thermal soak time for consistency. It was hypothesized that if external mass transfer dominates the mechanism, oxygen consumption and separation performance would be dependent on the inert flow rates. Table 7.9 shows the result of oxygen consumption on the two different flow rates. This further confirms that the flow rates do not affect the amount of oxygen consumption as noted above. In addition, Figure 7.21 shows that the separation performances are almost the same for different flow rates, which also Experiment #1

Experiment #2

~30 ppm O2/Ar 50 cc(STP)/min

~30 ppm O2/Ar 200 cc(STP)/min

m1

m2

l

m3

m4

l

FIGURE 7.20 A schematic of experiments to investigate external transfer limitation by applying two different inert flow rates. Total of 0.02 g polymeric films, m1 þ m2 ¼ m3 þ m4 were pyrolyzed at 823 K with 2-h soak time. Experiments were repeated twice.

TABLE 7.9 Normalized Values of the Total Amount of Oxygen at Different Inert Flow Rates of 50 and 200 cc(STP)/min with MatrimidÒ Inert Flowrate (cc(STP)/min)

Total O2 Available (ccSTP/g)

Total O2 Consumed (ccSTP/g)

50

67.3

45.4

200

154.2

50.0

Inert gas of  30 ppm O2/Ar was used in both cases. Experiments were repeated and have a standard deviation of less than 10%.

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

167

40

50 cc/min 2h

CO2 /CH4 selectivity

30

20 200 cc/min 2h

10 10−8

10−7 CO2 permeance (mol m−2 s−1 Pa−1)

10−6

FIGURE 7.21 Separation performance of MatrimidÒ CMS films pyrolyzed with two different inert flow rates of 200 (d) and 50 (○) cc(STP)/min. Inert carrier of 30 ppm O2/Ar was used. All was repeated at 308 K [56], and CO2 permeability was converted into an unit mol m 2 s 1 Pa 1 permeance with an assumed thickness of 1 mm.

support our hypothesis that rate effects due to external transport resistance are negligible factors in fixing membrane performance. Clearly, when CMS membranes experience longer thermal soak times, both the total amount of oxygen available and consumed should increase; however, consumption stayed almost consistent despite a higher inert (and oxygen) flow rate. Since the separation performances for the two different thermal soak times are almost the same, this discrepancy can be explained by suggesting that CMS oxidation may be equilibrium limited while by-products, or “molecular debris,” externally follow a kinetically limited reaction mechanism.

Effect of Precursor Film Thickness: Internal Transfer Limitation The above investigation shows that the transport mechanism is not likely dominated by the external transport. Next, we conducted another investigation to determine any role of internal mass transfer limitations. This investigation consisted of two experiments as shown in Figure 7.22. The first experiment consisted of pyrolysis of two polymer films (m1 and m2) each of which film had a thickness of 60 mm. The second experiment consisted of a film (m3) whose mass was essentially a sum of m1 and m2 with a thickness of 120 mm. It was hypothesized that if the internal mass transfer dominated the mechanism, the oxygen consumption and separation performance would depend strongly on the film thickness. Recall that about four times longer exposure to a given amount of

168

Inorganic, Polymeric and Composite Membranes

oxygen showed almost no effect on CMS oxygen uptake, so these experiments sought to further probe if the process of doping is an equilibrium limited, internal reaction process. The results are shown in Table 7.10 and Figure 7.23. Results of oxygen consumption in Table 7.10 indicate that oxygen consumption is almost the same regardless of the film thickness as long as the sample masses are the same. In addition, the separation performance of CO2/CH4 in Figure 7.23 is similar as well. This indicates that the oxygen doping process is unlikely to be internal mass transfer limited.

Oxygen–Carbon Reaction Mechanism: Chemical Reaction Limitation Consideration The above experiments show that the CMS membrane separation performance and oxygen consumption during the high temperature pyrolysis are not likely influenced by the thickness of the polymer precursors, inert, or oxygen, flow rates, nor thermal soak time during pyrolysis. The summation of all these facts implies that a carbon–oxygen equilibrium reaction governs the oxygen doping process via a chemisorption process. The oxygen chemisorption most likely takes place at the same time polymer precursors decompose, and evolved products diffuse out of the membranes. This indicates that the actual full mechanism can be very complex. Experiment #1

Experiment #2 ~30 ppm O2/Ar 200 cc(STP)/min

~30 O2/Ar 200 cc(STP)/min m1

m2

l

m3

2l

FIGURE 7.22 Schematic design of experiments testing the internal mass transfer limitation. About 0.02 g of polymer precursor, m1 þ m2 ¼ m3, was prepared. Gas mixture of argon and a slightly higher than 30 ppm of oxygen in argon gas were used as an inert with the flow of 200 cc(STP)/min. Experiments were repeated twice.

TABLE 7.10 Normalized Values of the Total Amount of Oxygen with Different Polymer Precursor Thickness Thickness of Polymer Precursor

Total O2 Available (ccSTP/g)

Total O2 Consumed (ccSTP/g)

60 mm

180

47

120 mm

182

52

Inert gas of  35 ppm O2/Ar was used in both cases. Experiments were repeated and have a standard deviation of less than 10%.

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

169

40

CO2/CH4 selectivity

30

200 cc/min 120 micron

20

200 cc/min 60 micron

10 5 ´ 10−8

10−7 CO2 permeance (mol

m−2 s−1 Pa−1)

FIGURE 7.23 Separation performance of MatrimidÒ CMS films pyrolyzed with two different precursor thickness, 60 (d) and 120 (○) mm. Inert carrier of 35 ppm O2/Ar was used. All was repeated at 308 K [56], and CO2 permeability was converted into an unit mol m 2 s 1 Pa 1 permeance with an assumed thickness of 1 mm.

Possible Mechanism of Oxygen “Doping” Process During Pyrolysis As previous sections described, we successfully developed an oxygen doping process to tune separation properties of CMS membranes by controlling the CMS structure. The method was built based on well-known scientific facts that (i) oxygen reacts with active carbon edges at high temperature during pyrolysis [28,29,31,45] and that (ii) the adsorption step which involves surface oxides dominates the reaction process and is endothermic and reversible in the temperature range between 623 and 923 K [32,53]. As described above, the oxygen–carbon mechanism can be very complex. Here, we seek to understand the pyrolysis process of polymer membranes with oxygen exposure by normalization of literature and our findings. The three stages that occur during the polymer decomposition process are : precarbonation, carbonation, and dehydrogenation [8]. In this study, oxygen was continuously supplied during the pyrolysis process. Precarbonation mostly involves removal of excess solvent and monomer [8], and as illustrated in Figure 7.5, consumption of oxygen does not start until temperature is close to the decomposition temperature. During the carbonation stages, it is believed that a majority of the oxygen was consumed by by-products. Meanwhile, a transformation from polymeric to CMS membranes takes place [8], and the “intrinsic” CMS structure results a product of the high temperature pyrolysis process.

170

Inorganic, Polymeric and Composite Membranes

Once most of the by-products are evolved, the dehydrogenation process begins, but consumption of oxygen by the by-products continues at a significantly decreased rate, and a larger amount of oxygen became available to the CMS membranes compared with previous two stages. The oxidation of the by-products is likely controlled kinetically while the oxidation of the active carbon edges for “doping” is likely equilibrium controlled. Therefore, the likelihood of oxygen molecules adsorbing on the surface and penetrating through CMS membranes to react with active carbon edges depends on oxygen concentration in the bulk flow. The oxidation “doping” process also depends on the temperature which is related to the energetics of the chemisorption [32,54]. Effects of the temperature on the “doping” process are complex since one also needs to consider the effect of pyrolysis temperature on the formation of the “intrinsic” CMS structure in the absence of oxygen chemisorption. The section dealing with effects of pyrolysis temperature consider an experiment involving the effects of a pyrolysis temperature of 773K. Comparing separation performance of CMS membranes with 823 K, it seems that the effect of temperature to the “doping” process is less significant compared to its importance on the formation of the intrinsic CMS structures. One should note, however, that a distribution of reactivity of carbon edges is speculated to exist to form carbonyl groups among ultramicropores, and the overall “doping” process is complex.

CONCLUSION In this chapter, the authors presented correlations observed during the development of a pyrolysis method that enables the control of separation performance of CMS membranes. This method utilized oxygen chemisorptions or “O2 doping” on selective pore windows at high temperature. Initially, the method was developed using a 6FDA-based polymer and extended to a commercially available polymer MatrimidÒ. Observed correlations were shown (i) between the total oxygen exposure qO2, tot and oxygen consumption, qO2 consumed and (ii) between qO2, tot and CMS CO2/CH4 separation performance. Moreover, a series of controlled experiments allowed us to show that the oxygen–carbon reaction appears to be equilibrium limited and that a correlation exists between oxygen concentration in the inert gas and CMS CO2/CH4 separation performance. These findings are significant for two major reasons: (i) they enable one to predict the trends in separation performance with the doping method after knowing the intrinsic structures created in low oxygen exposure pyrolysis conditions at a given temperature and (ii) they require only monitoring a primary variable of the oxygen concentration during the pyrolysis process for the doping method in inert pyrolysis.

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

171

ACKNOWLEDGEMENTS The authors gratefully acknowledge support for this work by Shell Global Solutions (US) Inc., and by award no. KUS-I1-011-21 made by King Abdullah University of Science and Technology (KAUST).

REFERENCES [1] R.W. Baker, Future directions of membrane gas separation technology, Ind. Eng. Chem. Res. 41 (2002) 1393–1411. [2] R.W. Baker, K. Lokhandwala, Natural gas processing with membranes: an overview, Ind. Eng. Chem. Res. 47 (2008) 2109–2121. [3] M.R. Coleman, W.J. Koros, The transport properties of polyimide isomers containing hexafluoroisopropylene in the diamine residue, J. Polym. Sci., Part B 32 (1994) 1915. [4] M.R. Pixton, D.R. Paul, Gas transport properties of polyarylates: substituent size and symmetry effects, Macromolecules 28 (1995) 8277. [5] L.M. Robeson, The upper bound revisited, J. Membr. Sci. 320 (1–2) (2008) 390–400. [6] R.W. Baker, Membrane Technology and Applications, McGraw-Hill, New York, 2000. [7] A.F. Ismail, L.I.B. David, A review on the latest development of carbon membranes for gas separation, J. Membr. Sci. 193 (1) (2001) 1–18. [8] G.M. Jenkins, K. Kawamura, Polymeric Carbons—Carbon Fiber, Glass and Char, Cambridge University Press, London, 1976. [9] H.O. Pierson, Handbook of Carbon, Graphite, Diamond, and Fullerenes, Noyes Publication, New York, 1993. [10] H. Suda, K. Haraya, Gas permeation through micropores of carbon molecular sieve membranes derived from Kapton polyimide, J. Phys. Chem. B 101 (20) (1997) 3988–3994. [11] A. Singh, Membrane Materials with Enhanced Selectivity: An Entropic Interpretation, University of Texas at Austin, Austin, TX, 1997. [12] C.W. Jones, W.J. Koros, Carbon molecular sieve gas separation membranes-I. Preparation and characterizaton based on polyimide precursors, Carbon 32 (8) (1994) 1419–1425. [13] D.Q. Vu, W.J. Koros, S.J. Miller, High pressure CO2/CH4 separation using carbon molecular sieve hollow fiber membranes, Ind. Eng. Chem. Res. 41 (3) (2002) 367–380. [14] V.C. Geiszler, W.J. Koros, Effects of polyimide pyrolysis conditions on carbon molecular sieve membrane properties, Ind. Eng. Chem. Res. 35 (9) (1996) 2999–3003. [15] P.J. Williams, Analysis of Factors Influencing the Performance of CMS Membrane for Gas Separation, Georgia Institute of Technology, Atlanta, GA, USA, 2006, PhD thesis. [16] K.M. Steel, Carbon Membranes for Challenging Gas Separations, The University of Texas at Austin, Austin, TX, USA, 2000, PhD thesis. [17] M.C. Campo, F.D. Magalhaes, A. Mendes, Comparative study between a CMS membrane and a CMS adsorbent: Part I—morphology, adsorption equilibrium and kinetics, J. Membr. Sci. 346 (1) (2010) 15–25. [18] K.M. Steel, W.J. Koros, Investigation of porosity of carbon materials and related effects on gas separation properties, Carbon 41 (2) (2003) 253–266. [19] H.B. Park, Y.K. Kim, J.M. Lee, S.Y. Lee, Y.M. Lee, Relationship between chemical structure of aromatic polyimides and gas permeation properties of their carbon molecular sieve membranes, J. Membr. Sci. 229 (1–2) (2004) 117–127. [20] J. Chen, L.S. Loo, K. Wang, D.D. Do, The structural characterization of a CMS membrane using Ar sorption and permeation, J. Membr. Sci. 335 (1–2) (2009) 1–4.

172

Inorganic, Polymeric and Composite Membranes

[21] K.M. Steel, W.J. Koros, An investigation of the effects of pyrolysis parameters on gas separation properties of carbon materials, Carbon 43 (2005) 1843–1856. [22] T.A. Centeno, A.B. Fuertes, Supported carbon molecular sieve membranes based on a phenolic resin, J. Membr. Sci. 160 (1999) 201–211. [23] J.E. Koresh, A. Soffer, Molecular sieve permselective membrane. Part I. Presentation of a new device for gas mixture separation, Sep. Sci. Technol. 18 (8) (1983) 723–734. [24] M.G. Sedigh, M. Jahangiri, P.K. Liu, M. Sahimi, T.T. Tsotsis, Structural characterization of polyetherimide-based carbon molecular sieve membranes, AICHE J. 46 (11) (2000) 2245–2255. [25] D.Q. Vu, Formation and Characterization of Asymmetric Carbon Molecular Sieve and Mixed Matrix Membranes for Natural Gas Purification, University of Texas at Austin, Austin, TX, USA, 2001, PhD thesis. [26] J.N. Barsema, N.F.A. Vegt, G.H. Koops, M. Wessling, Carbon molecular sieve membranes prepared from porous fiber precursor, J. Membr. Sci. 205 (1–2) (2002) 239–246. [27] R.O. Grisdale, The properties of carbon contacts, J. Appl. Phys. 24 (10) (1953) 1288–1296. [28] K. Kinoshita, Carbon: Electrochemical and Physicochemical Properties, Wiley-Interscience, Berkeley, CA, USA, 1988. [29] H. Marsh, Introduction to Carbon Science, Butterworth & Co. Ltd., Borough Green, England, 1989. [30] T. Ishiguro, N. Suzuki, Y. Fujitani, H. Morimoto, Microstructural changes of diesel soot during oxidation, Combust. Flame 85 (1–2) (1991) 1–6. [31] N.R. Laine, F.J. Vastola, P.L.J. Walker, The importance of active surface area in the carbonoxygen reaction, J. Phys. Chem. 67 (10) (1963) 2030–2034. [32] G. de Soete, Catalysis of soot combustion by metal oxides, in: Western States Section Meeting of the Combustion Institute, 1988 March 21–22, 1988; Salt Lake City, UT, USA. [33] I.C. Omole, Crosslinked Polyimide Hollow Fiber Membranes for Aggressive Natural Gas Feed Streams, Georgia Institute of Technology, Atlanta, GA, USA, 2008, PhD thesis. [34] Rapidox 2100 Oxygen Analyser Instruction Manual, Cambridge Sensotec Limited, Cambridge, England, 2004. [35] D.G. Pye, H.H. Hoehn, M. Panar, Measurement of gas permeability of polymers. I. Permeabilities in constant volume/variable pressure apparatus, J. Appl. Polym. Sci. 20 (7) (1976) 1921–1931. [36] D.G. Pye, H.H. Hoehn, M. Panar, Measurement of gas permeability of polymers. II. Apparatus for the determination of permeabilities of mixed gases and vapors, J. Appl. Polym. Sci. 20 (2) (1976) 287–301. [37] L.M. Costello, W.J. Koros, Temperature dependence of gas sorption and transport properties in polymers: measurement and applications, Ind. Eng. Chem. Res. 31 (12) (1992) 2708–2714. [38] W.J. Koros, D.R. Paul, Design considerations for measurements of gas sorption in polymers by pressure decay, J. Polym. Sci., Polym. Phys. Ed. 14 (10) (1976) 1903–1907. [39] M. Kiyono, P.J. Williams, W.J. Koros, Effect of polymer precursor on carbon molecular sieve structure and separation performance properties, accepted, Carbon 48 (15) (2010) 4432–4441. [40] J.N. Barsema, S.D. Klijnstra, J.H. Balster, N.F.A. Van der Vegt, G.H. Koops, M. Wessling, Intermediate polymer to carbon gas separation membranes based on Matrimid PI, J. Membr. Sci. 328 (1–2) (2004) 93–102. [41] H. Lee, H. Suda, K. Haraya, Characterization of the post-oxidized carbon membranes derived from poly(2, 4-dimethyl-1, 4-phenylene oxide) and their gas permeation properties, Sep. Purif. Technol. 59 (2) (2008) 190–196.

Chapter

7

Correlation Between Pyrolysis Atmosphere and Carbon Molecular

173

[42] F. Stoeckli, T.A. Centeno, On the characterization of microporous carbons by immersion calorimetry alone, Carbon 35 (8) (1997) 1097–1100. [43] A. Singh, W.J. Koros, Significance of entropic selectivity for advanced gas separation membranes, Ind. Eng. Chem. Res. 35 (4) (1996) 1231–1234. [44] M. Kiyono, P.J. Williams, W.J. Koros, Effect of pyrolysis atmosphere on separation performance of carbon molecular sieve membranes, J. Membr. Sci. (2009) (Membranes for CO2). [45] B.R. Stanmore, J.F. Brilhac, P. Gilot, The oxidation of soot: a review of experiments, mechanisms and models, Carbon 38 (15) (2001) 2247–2268. [46] B.C. Smith, Infrared Spectral Interpretation: A Systematic Approach, first ed., CRC Press, New York, 1999. [47] J.R. Welty, C.E. Wicks, R.E. Wilson, G.L. Rorrer, Fundamentals of Momentum, Heat, and Mass Transfer, John Wiley & Sons, Inc., New York, 2001. [48] P.J. Walker, F.J. Rusinko, L.G. Austin, Gas reactions of carbon, in: D.D. Eley, W.G. Frankenburg (Eds.), Advances in Catalysis and Related Subjects, Academic Press Inc., New York, 1959, pp. 133–176. [49] O. Levenspiel, Chemical Reaction Engineering, third ed., John Wiley & Sons, Inc., Hoboken, NJ, 1999, p. 376–417. [50] A.E. Lear, T.C. Brown, B.S. Haynes, Formation of metastable oxide complex during the oxidation of carbons at low temperatures, in: Twenty-third International Symposium on Combustion, The combustion Institute, Pittsburge, 1990, pp. 1191–1197. [51] J. Neeft, T. Nijhuis, E. Smakman, M. Makkee, J. Moulijn, Kinetics of the oxidation of diesel soot, Fuel 76 (12) (1997) 1129–1136. [52] H.B. Park, S.Y. Nam, J.G. Jang, Y.M. Lee, Preparation, characterization, and gas permeation properties of carbon molecular sieve membranes derived from dense P84-polyimide film, Korean Membr. J. 4 (1) (2002) 25–35. [53] J.N. Ong, On the kinetics of oxidation of graphite, Carbon 2 (1964) 281–297. [54] G. Charles, J. Hill, An Introduction to Chemical Engineering Kinetics & Reactor Design, John Wiley & Sons, Inc., New York, 1977. [55] M. Mulder, Basic Principles of Membrane Technology, Kluwer Academic Publishers, Netherlands, 2000. [56] M. Kiyono, P.J. Williams, W.J. Koros, Generalization of effect of oxygen exposure on formation and performance of carbon molecular sieve membranes, accepted, Carbon 48 (15) (2010) 4442–4449.

This page intentionally left blank

Chapter 8

Review on Prospects for Energy Saving in Distillation Process with Microporous Membranes Masahiko Matsukata1,2,*, Ken-ichi Sawamura1, Yasushi Sekine1,2 and Eiichi Kikuchi1,2 Department of Applied Chemistry, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, Japan Advanced Research Institute for Science and Engineering, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, Japan * Corresponding author: E-mail address: [email protected] 1 2

INTRODUCTION The past decades have seen a worldwide expansion in industrial output in energyintensive industries such as steel, chemicals, and petroleum. The case of Japan is special, since despite expansion in the industrial sector, energy consumption has risen little since the 1970s. This has been due to substantial efforts in energy savings. In part, this was motivated by increasing pressure from the government and the general public for reducing energy consumption by industry to help mitigate CO2 emissions. Overall, however, it seems that most measures for energy savings have already been taken, and there is little room left for large-scale reductions in energy consumption with existing technologies. It is, thus, essential to develop novel alternatives which can enable large-scale reductions in energy use. As will be discussed, membrane technology offers one such possibility. In the chemical and petroleum industries, it is estimated that about 40% of the total energy consumption is expended in separation processes. Since more than 90% of this energy is used in distillation, a reduction in energy demand for distillation will have a great impact. While membrane separation technology has not been broadly commercialized in the chemical industry, the analysis in this chapter will demonstrate that membrane separations offer great opportunities to significantly reduce energy consumption in existing and future chemical and petroleum processes. In particular, microporous aluminosilicate materials such as zeolites are attractive candidates as membrane materials.

Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

175

176

Inorganic, Polymeric and Composite Membranes

POTENTIAL OF MEMBRANE SEPARATION TECHNOLOGY FOR LARGE-SCALE REDUCTION IN ENERGY CONSUMPTION A conventional distillation separation system inevitably consumes a large amount of energy because of the need to vaporize a substantial portion of the feed. The energy requirement is roughly given by the following equation: Energy demand ¼ (Reflux ratio þ 1Þ  (Heat of vaporizationÞ

ð8:1Þ

As will be shown, the installation of a membrane separation unit in combination with a distillation tower can decrease the reflux ratio, resulting in a substantial decrease in energy consumption when the relative volatilities of the substances to be separated are small. Here, the separation of water and acetic acid (AcOH) will be considered as an example. Separation of water/AcOH mixtures is an important operation in the production of such chemicals as terephthalic acid, acetate esters, and AcOH itself, and is one of the systems where membrane separation is highly desired [1–6] due to the low relative volatility of water and AcOH, 1.3–2.7. Figure 8.1a shows the estimated energy necessary for separating and concentrating a 100 kg h 1 stream of a 50/50 wt% water/AcOH mixture to 99 wt% of water and AcOH. Since the relative volatility of water and AcOH is low, concentrating the components to 99 wt% of water and AcOH requires a large part of the top product of the distillation column to be refluxed. For a reflux ratio ¼ 5, obtaining 49.5 kg h 1 of 99 wt% AcOH requires 250 kg h 1 of the top product of the distillation column to be refluxed, which is about five times as large as the amount of concentrated AcOH obtained. Since heavier AcOH travels to the bottom of the distillation column while lighter water travels to the top of the distillation column, a large amount of water is required to be repeatedly vaporized at the top of the distillation column. The repeated vaporization requires a significantly large amount of energy, 162,000 kcal h 1, for supplying the thermal energy for vaporization. Figure 8.1b and c shows the effect of introducing a membrane separation on the estimated energy necessary for separating and concentrating the water/ AcOH mixture. If distillation is completely replaced with a pervaporation (PV)-based membrane separation with a high performance de-watering membrane (separation factor, a ¼ 400), a drastic reduction of energy demand can be possible (Figure 8.1b). Since PV-based membrane separation requires only the heat of vaporization of the permeate molecules, the energy necessary for the separation is estimated to be only 27,000 kcal h 1, cutting the energy demand by about 85%. However, the complete replacement of distillation with PVbased membrane separation requires significantly high performance of the membrane and sizable investment cost, which may hinder its practical application. However, this preliminary study suggests that just installing a membrane separation unit in the latter part of the distillation tower is effective enough to largely reduce the energy demand for the separation (Figure 8.1c). In this

Chapter

8

177

Review on Prospects for Energy Saving

(a) Conventional distillation 300 kg h-1

100 kg h-1

50 kg h-1

Water: 49.5 AcOH: 0.5 -1 250 kg h Water: 50 AcOH: 50 162,000 [kcal h-1] Water: 0.5 AcOH: 49.5 (Reflux ratio: 5) (b) Pervaporation (PV) Membrane a(water/AcOH) = 400

27,000 [kcal h-1] 100 kg h-1

Water: 49.5 kg h-1 (99%) AcOH: 0.5 kg h-1 (1%)

Water: 50 AcOH: 50

(c)

Water: 0.5 kg h-1 (1%) AcOH: 49.5 kg h-1 (99%)

Distillation–membrane (vapor permeation, VP) hybrid system

Water: 57.2 kg h-1 (60%) AcOH: 38.1 kg h-1

100 kg h-1 Water: 50 AcOH: 50

Membrane a(water/AcOH) = 150

Water: 7.65 kg h-1 (50%) AcOH: 7.65 kg h-1 (50%) 35,000 [kcal h-1] Water: 0.44 kg h-1 (2.2%) AcOH: 19.5 kg h-1 (97.8%)

Water: 49.5 kg h-1 (99%) AcOH: 0.5 kg h-1 (1%) Water: 7.71 kg h-1 (17%) AcOH: 37.6 kg h-1 (83%) Membrane a(water/AcOH) = 20 Water: 0.06 kg h-1 (0.2%) AcOH: 30.0 kg h-1 (99.8%)

FIGURE 8.1 Estimated energy demands for separation of a 100 kg h 1 of 50/50 wt% water/ AcOH mixture to 99 wt% of water and AcOH by (a) a conventional distillation system, (b) a membrane (PV-mode) separation system, and (c) a distillation-membrane (VP-mode) hybrid separation system.

calculation, de-watering membranes with separation factors of only 20 and 150 (not very high) are installed in the exit section of the distillation tower. The separation factors of 20 or 150 are practical target values for researchers and technicians to achieve. In this case, a part of the top product of the distillation column needs to be refluxed because it cannot concentrate to 99 wt% in a onepass operation. However, the water concentration of the top product of the distillation column necessary for obtaining 99 wt% of AcOH is about 60 wt%, reducing the energy demand to 35,000 kcal h 1. In this system, a vacuum pump

178

Inorganic, Polymeric and Composite Membranes

may be necessary for maintaining a low pressure at the permeate side to increase the permeation driving force. Even considering the energy consumption for operating the vacuum pump, the introduction of the membrane separation unit in the distillation tower enables us to cut about 70% of the energy demand for the separation. The efficiency of this distillation–membrane hybrid separation system highly depends on the value of a; therefore, development of higher performance membranes enables further energy savings. In short, this distillation–membrane hybrid separation system enables a large-scale reduction in energy utilization in an existing distillation unit with minimum risk by the addition of a membrane separation unit. This is due to the following important features of membrane separation technology: (1) Requirements for the membrane performance are not so high. The process can be flexibly modified depending on the membrane performance. (2) Large-scale reduction in energy consumption is possible without scrapping an already working distillation unit by the addition of a membrane separation unit. (3) The performance of the distillation–membrane hybrid separation system can be incrementally enhanced by replacing membranes as the membrane properties are improved. Our preliminary estimation described above shows that proper combination of distillation and membrane separation units enables large-scale reduction of energy consumption without scrapping distillation towers by just installing a membrane separation unit. De-watering is essential in the petrochemical industry because water is a reactant or by-product in the production of chemicals such as ethanol, ethylene glycol, acetic acid, and ethyl acetate via partial oxidation and hydration. Therefore, improvement of de-watering efficiency by a distillation–membrane hybrid separation system should contribute to sustainability in the petrochemical industry. We described prospects of membrane separation as retrofit technology in distillation units. Membrane separation is also promising for separation of mixtures which are difficult to separate by distillation as in the following cases: (1) Azeotropic mixtures (e.g., ethanol/water system, azeotrope composition 96 wt% of ethanol) (2) Mixtures with close boiling points (e.g., xylene isomers, boiling points 144.41  C for o-xylene, 139.10  C for m-xylene, and 138.35  C for p-xylene) (3) Gas mixtures (e.g., CH4/CO2, propylene/propane, H2). In the first case, de-watering technology for ethanol using membranes has recently been commercialized, as described later in detail. For the second case, microporous membranes have been studied intensively, to utilize their molecular-sieving properties. Gas mixtures need to be liquefied by consuming low-

Chapter

8

Review on Prospects for Energy Saving

179

temperature energy for distillation, and such low-temperature energy is difficult to recover and reuse. Besides, some components in gas mixtures have close boiling points. In these cases, the impact on energy savings by installing membrane separation units would be tremendous if gas separation becomes possible. In particular, important applications are separations of natural gas (CH4, CO2, N2, C2–C3 hydrocarbons, and steam), light hydrocarbon mixtures (C1–C4 hydrocarbons) produced from naphtha and ethane crackers for the production of petrochemical raw materials and hydrogen. In addition to the above applications, membrane separations are also promising for applications in C1 chemistry, as will be mentioned later.

WHY ZEOLITE MEMBRANES ARE PROMISING The history of membrane separation technology dates to the 1940s. Porous alumina membranes found a practical application, in the separation of UF6 isotopes on the basis of the difference in their molecular weights. In the 1960s, research and development in organic polymer membranes began, and most efforts have been directed to these materials, as inorganic membranes were found to be difficult for practical applications for several decades. Since then, organic polymer membranes have been industrially used in large-scale for seawater desalination and purification of hydrogen. In the 1980s, organic polymer membranes were tried for separations in the chemical and petroleum industries, but the results revealed that these membranes could not satisfy requirements for separation due to their instability in organic solvents and at high temperature. However, there has been growing interest in inorganic membranes for applications to separation processes, owing to their superior thermal, chemical, and structural stability. To develop new generation membranes with suitable durability, researchers have returned to R&D of inorganic membranes. Inorganic membranes can be classified into two types, nonporous (dense) and porous membranes. Among nonporous membranes, palladium-based membranes [7,8], which are permeable only to hydrogen, have been well studied due to their high gas permselectivity. Following palladium-based membranes, other types of dense inorganic membranes permselective to oxygen, for example, silver and stabilized zirconia, have been developed [9]. However, the types of permselective nonporous membranes developed so far have been limited. Porous materials may be classified into three groups depending on the size of their pores: macroporous, mesoporous, and microporous. Based on the definition by International Union of Pure and Applied Chemistry (IUPAC), this classification corresponds to the following pore sizes: (1) Pores with widths exceeding about 50 nm are called macropores; (2) Pores with widths between 2 and 50 nm are called mesopores; and (3) Pores with width not exceeding about 2 nm are called micropores.

180

Inorganic, Polymeric and Composite Membranes

Among materials having these types of pores, microporous materials should be more appropriate as membrane materials because molecular-sieving properties appear when the pore size is comparable to molecular dimensions and, in addition, preferential adsorption in micropores can lead to high permselectivity. Since size and physical properties are different from molecule to molecule, membranes with different pore sizes and adsorption properties should be developed to match the properties of targeted mixtures. Increasing attention has been paid to development of zeolite membranes owing to the controllability of their physicochemical properties [10–12]. Zeolites are a class of microporous aluminosilicate crystals that offer great potential as membrane materials because their physicochemical properties can be widely controllable by choice of structure, Si/Al ratio, and type of exchanged cation. In the structure of zeolites, the pore diameters are determined by the aperture of the cages, with 8, 10 or 12-membered rings of oxygen anions being the most common. The maximum values of the pore openings have been calculated to be 0.42, 0.63, and 0.74 nm, respectively. Figure 8.2 shows the pore dimensions in zeolites together with the sizes of several molecules [13,14]. Zeolites can be classified into three main groups: small-pore zeolites such as A (LTA); medium-pore zeolites such as ZSM-5 and silicalite (MFI), and large-pore zeolites such as X, Y (FAU) and mordenite (MOR). A suitable structure should be selected depending on the sizes of the molecules to separate.

nm 1.0 0.9

1,3,5-Trimethyllbenzene

o/m-Xylene

0.8

X

FAU Y

0.7

BEA

MOR

Double branched alkanes p-Xylene, benzene SF6

0.6

Single branched alkanes IPA EtOH, AcOH, n-alkanes MeOH, CH4 N2 O2 CO

0.5

MFI (ZSM-5) LTA (A)

DDR

FER (ferrierite)

0.4

CHA

2

H2O

H2 He

0.3 ANA (analcime) 0.2 1

5

10 100 Si/Al ratio

200

FIGURE 8.2 Pore dimensions of zeolites in relation to the size of several molecules.

∞

Chapter

8

Review on Prospects for Energy Saving

181

Second, the hydrophilic/hydrophobic nature and stability of zeolites can be tuned by changing the Si/Al ratio in the framework of the zeolites. The cation that balances the negative charge associated with framework aluminum ions creates an electrostatic field in the zeolite. Increase of Si/Al ratio of zeolite membranes can enhance its thermal and acid stability, though the hydrophilicity properties decrease. Third, both the actual pore size and the affinity between adsorbed molecules and the pore-wall depend on the type of cation and level of ion-exchange.

SYNTHESIS TECHNIQUE OF ZEOLITE MEMBRANES Zeolite membranes are often prepared at conditions similar to those of zeolite powder synthesis. For a polycrystalline zeolite membrane, however, the intercrystalline voids must be embedded by the intergrowth of crystals, which is not required for zeolite powder synthesis. Control of crystal growth of zeolites is essential for preparation of a compact zeolite membrane. Since the mechanical strength of self-supporting, thin zeolite membranes is insufficient for practical use, zeolite membranes are generally prepared on porous supports. The physicochemical properties of the porous support are, thus, important for zeolite membrane synthesis. Generally, a-alumina is used as the support material. To control the zeolite membrane thickness, compactness, and orientation, several synthesis methods have been developed as follows.

Seeding Technique (Secondary Growth Method) The seeding technique [15–39] enables better control of nucleation and crystal growth steps. The seeds can grow in low concentration solutions where homogeneous nucleation is suppressed, leading to better control of zeolite crystal growth for membranes. Seeding is also effective to obtain some zeolites membranes free from [17–22] or with lesser amounts of [39] structure directing agents (SDAs). It should be noted that organic SDAs are corrosive and expensive; therefore, organic SDA-free synthesis of a zeolite membrane is important for both the environment and the cost. In some cases, the kind of zeolite formed was critically influenced by the type of seed employed [22,23]. For example, Li et al. showed that mordenite or ZSM-5 membranes could be prepared from identical solution under the same hydrothermal conditions by using either mordenite or ZSM-5 seeds. Seeding is generally performed with zeolite colloidal particles by dip-coating, while some groups employed other methods, including rubbing the support surface with zeolite crystals [23–27] or pulsed laser ablation of zeolite powder [28–30]. Among the methods, seeding with colloidal particles by dip-coating seems to be the most effective method for industrial applications. The size of seed crystals is one of the most critical factors to prepare a compact zeolite membrane. Zhang et al. [31] investigated the influence of seed size

182

Inorganic, Polymeric and Composite Membranes

on the microstructure of zeolite membranes by seeded growth, with five silicalite-1 seeds of around 100, 600 nm, 1.5, 3.0, and 7.5 mm. They claimed that the seed layer and secondary grown membrane from the smallest seed (100 nm) were the most uniform and densest, though no permeation and separation tests were performed for supporting their claim. Li et al. [32] developed a method to obtain zeolite nanoseed crystals for seeding. The positive points of this method are that it is simple and applicable to any kind of zeolite. The procedure of this method is as follows: (1) First, crushing the zeolite powder in an agate mortar. (2) Second, mixing the crushed powder with an appropriate amount of water in a beaker to form a slurry. This slurry is then treated in an ultrasonic bath. (3) Finally, this slurry is kept at room temperature for a few days. Due to gravitational effect, bigger particles are settled down to the bottom of the beaker, while smaller particles are dispersed in the upper part of the slurry to form a colloidal suspension. Using this method, Li et al. obtained about 100 nm of mordenite and ZSM-5 seed crystals. By controlling the growth of zeolite seed crystals, compact zeolite membranes can be synthesized on a porous support. Recently, the group of Matsukata reported synthesis of compact zeolite membranes with medium Si/Al ratio, a mordenite membrane (Si/Al ¼ 5–6) [40] and a ZSM-5 membrane (Si/Al ¼ 12) [41], on the surface of porous a-alumina tubes by controlling the growth of zeolite seed crystals.

Masking Technique Hedlund et al. [42,43] discussed a method to prepare an ultra-thin (less than 1 mm) membrane without defects on an open support and summarized important factors as follows: (1) The support should have a high mechanical strength, but at the same time should also have a high porosity and large pores to ensure a high flux. The support surface should be smooth to enable complete coating by a thin membrane. (2) The support should be masked to prevent support invasion and leaching. (3) The membrane should be continuous and significantly thinner than 1 mm to ensure high flux. (4) Interference from dust or crystals nucleating in the liquid should be avoided. (5) The membrane should have a sufficiently low defect density to be isomer selective, preferably without reparation.

Chapter

8

Review on Prospects for Energy Saving

183

To prepare thin zeolite membranes with few defects, considering all factors listed above, Hedlund et al. developed a masking technique, called two-step masking. This procedure relies on a masking approach that fills all support pores with wax while leaving the top surface free for deposition of the zeolite membrane, thus, protecting the support from the synthesis mixture. After hydrothermal synthesis of the zeolite membrane, the wax is removed by calcination. By this method, Hedlund et al. successfully obtained a silicalite-1 membrane with a thickness of 500 nm and few defects and achieved a high permeance of n-hexane (5.6  10 7 mol s 1 m 2 Pa 1) with a high separation factor of n-hexane/2,2-dimethyl-butane of 227 at 400  C. Another effective technique was reported by Lai et al. [44]. They coated the top surface of an a-alumina support (pore size 200 nm) with a mesoporous silica layer (pore size 2 nm) by using the sol–gel technique developed by Brinker and coworkers [45]. The mesoporous silica layer provided a smooth surface that can be functionalized for the deposition of the seeds. In addition, it acts as a barrier for avoiding zeolite deposit formation in the interior of the support and also reduces leaching of the aluminum from the support during secondary growth synthesis. They also found that the presence of the silica layer eliminated stress-induced crack formation during calcination at the support/zeolite interface.

Use of SDA for Microstructural Optimization Siliceous zeolites are generally synthesized by using organic SDAs. For example, MFI type zeolites are usually synthesized by using tetrapropylammonium hydroxide (TPAOH) as SDA. In zeolite membrane synthesis, other kinds of SDAs are also employed to optimize the microstructure of zeolite membranes such as pore orientation and membrane compactness. In terms of pore orientation, membrane orientation can be critical when the membrane is composed of zeolites with anisotropic pore structures. ZSM-5, for example, has straight 0.54  0.56 nm channels along the b-axis and sinusoidal 0.51  0.55 nm channels along the a-axis. Both experimental data and simulation studies [46–48] suggest that the fastest diffusion pathway in ZSM-5 crystals is the straight channel along the b-axis. Therefore, b-oriented ZSM-5 membranes seem favorable in the separation of xylene isomers by obtaining a high flux of p-xylene, at the same time blocking o-xylene permeation at the window of the ZSM-5 pores. Lai et al. [44] prepared siliceous ZSM-5 membranes oriented with straight 0.54  0.56 nm pores perpendicular to the support using trimer-TPAOH as crystal shape modifiers by the secondary growth of b-oriented seed layers and also prepared ZSM-5 membranes using monomer-TPA with mixed grain orientations, namely b&a- or b&a&h0h/c orientation depending on the temperature of the secondary growth. The b-oriented ZSM-5 membrane exhibited higher separation factors of p-/o-xylene (up to 450 at 220  C) while maintaining high permeance of p-xylene, in comparison with ZSM-5 membranes with different pore orientations.

184

Inorganic, Polymeric and Composite Membranes

In terms of compactness, Noack et al. claimed a zeolite with a higher aluminum content was difficult to grow into a compact membrane because its highly negatively charged surface prevents the transport of negatively charged precursors to the crystal boundary [49,50]. They proposed the concept of crystal intergrowth supporting substances (ISS) for enhancing crystal intergrowth in zeolite membranes. In this concept, the negatively charged crystal surfaces of a growing zeolite membrane are neutralized by the positively charged ISS for better crystal intergrowth. They used N,N,N,N’,N’,N’-hexamethylethylene diammoniumdiiodide as ISS and enhanced crystal intergrowth in ZSM-5 membranes with an Si/Al ratio from 96 to 200. However, a compact zeolite membrane with an Si/Al ratio less than 96 is still difficult to obtain even with ISS. However, ISS may not be essential for the preparation of compact zeolite membranes with high aluminum content because compact zeolite membranes with relatively high aluminum content were obtained by the group of Matsukata, namely a mordenite membrane (Si/Al ¼ 5–6) [40] and a ZSM-5 membrane (Si/Al ¼ 12) [41] without the use of ISS by controlling the growth of seed crystals, as mentioned above.

DE-WATERING TECHNOLOGY USING ZEOLITE MEMBRANES Zeolite membranes are expected to be applied in various separations, and in particular, in de-watering, because it is a requirement in many separation systems. In this section, state-of-the-art of de-watering technology using zeolite membranes is described.

De-watering of Alcohol Zeolite membranes were first proposed for separations in the 1980s [51–54]. However, research on zeolite membranes is relatively new and just started to appear frequently in academic journals in the early 1990s. In 1997, an A-type zeolite membrane was commercialized as a de-watering membrane for isopropanol (IPA) and several organic solvents in the semiconductor industry by Mitsui Engineering and Shipbuilding Co. Ltd. in Japan. This commercialization was due to the strong hydrophilic nature of A-type zeolites which enabled the easy obtention of a high-purity organic solvent. So far, three Japanese companies, Mitsui Engineering and Shipbuilding Co. Ltd., Mitsubishi Chemical Co.(previous Bussan Nanotech Research Institute Inc.), and Hitz Hitachi Zosen Corporation; a German company, Inoceramic Inc. (previous The Smart Chemical Company of UK); and a Singapore company, Hyflux Ltd., have been involved in the development of A-type zeolite membranes. One of their aims is production of fuel grade ethanol by dewatering ethanol. Figure 8.3 shows tubular zeolite membranes. The diameter of the tubular membranes varies from 12 to 17 mm, depending on the company. A tubular

Chapter

8

Review on Prospects for Energy Saving

185

FIGURE 8.3 Commercial tubular zeolite membranes on porous a-alumina support.

(a)

(b)

FIGURE 8.4 SEM images of (a) surface and (c) cross-section of an A-type zeolite membrane on porous a-alumina support.

membrane with larger diameter has larger surface area but has less uniformity in diameter and straightness. Therefore, each company is choosing their own optimum size. For enlarging membrane surface area, monolith supports are also being developed. Figure 8.4 shows SEM images of the surface and cross-sections of an A-type zeolite membrane. The surface of the membrane seems to be covered with a large number of cubic crystals, which are typical A-type crystals. The SEM image of the cross-section of the membrane shows tightly ordered A-type crystals forming a thin membrane of a few micrometers thickness. Since the zeolite layer is polycrystalline, there may be voids or other structures along the grain boundaries, which might act either as conduits or barriers for the permeation of molecules. Such fine structures are usually too small to be detected by SEM; therefore, TEM is the only method for studying the fine structure. TEM observation reveals that these A-type crystals form column crystals and that an unknown phase with a width of ca. 5 nm exists at the grain boundary of the column crystals [55]. The effect of the boundary layer on permeation and separation property is yet to be clarified.

186

Inorganic, Polymeric and Composite Membranes

In de-watering of ethanol, the separation factor of an A-type zeolite membrane is extremely high and can exceed 100,000 in the laboratory. In addition, water flux can exceed 10 kg m 2 h 1 at some conditions, which is two orders of magnitude larger than that of organic polymer membranes for de-watering. Recently, Hitachi Zosen Corporation reported the development of an A-type zeolite membrane with extremely high water flux (> 50 kg m 2 h 1) which maintained high water permselectivity (> 1000) at 130  C with an ethanol/water mixture of 90/10 wt%. As discussed here, A-type zeolite membranes show excellent performance as de-watering membranes and, therefore, have been commercialized at an early stage in the development of zeolite membranes. Figure 8.5 shows a schematic diagram of a pilot plant for production of fuel grade bioethanol operated in the E3 demonstration project in Miyako Island, Okinawa prefecture in Japan. This plant uses triple distillation systems for maximizing the efficiency of distillation. In addition, A-type zeolite membrane units for de-watering are also installed in the latter part of the distillation towers. In this process, the distillation towers concentrate ethanol to 88 wt%, and the zeolite membrane units concentrate ethanol to 99.6 wt%, overcoming the azeotropic limitation. It should be noted that the third distillation tower is operated under pressurized condition (500 kPa) to provide a driving-force for membrane permeation. In this process, more than 80% of energy saving has been achieved, in comparison with a simple azeotropic distillation. As shown here, properly allocated distillation–membrane hybrid separation systems can have great impact in reducing energy consumption.

De-watering of Organic Acids As mentioned earlier, the development of membranes for de-watering organic acids, such as acetic acid, are highly desired for their significant impact on energy saving. Though hydrophilic NaA (Si/Al ¼ ) type zeolite membranes Dilution tank

Fermentor

Molasses Water

Triple effect distillation

Zeolite

Stripper Stripper Rectifier membrane (Vacuum) (Atms.) (4barG) Ethanol 88 wt% (Vap.) Ethanol 45 wt% (Liq.) Feed (Sugar 15%)

Beer (Ethanol 6 wt%)

Product ethanol (99.5 wt%)

Distillation Residue FIGURE 8.5 Schematic diagram of a process flow of a fuel grade bioethanol production plant installed with a de-watering membrane separation unit in Miyako Island, Okinawa, Japan.

Chapter

8

187

Review on Prospects for Energy Saving

have been applied industrially for the dehydration of alcohols [56,57], the membranes are not stable at high temperature or with organic acid systems. As mentioned above, increase of the Si/Al ratio of a zeolite can enhance its thermal and acid stability, though there is a trade-off with hydrophilicity. Therefore, zeolite membranes with medium Si/Al ratio, such as T [58], CHA [59], mordenite [5,60], and ZSM-5 membranes [6,61], seem appropriate for de-watering of organic acids. For this reason, increasing effort has been paid to the development of these zeolite membranes for potential applications where A-type zeolite membranes cannot be used.

De-watering for C1 Chemistry De-watering technology is also important for C1 chemistry (Figure 8.6), such as Fischer–Tropsch (FT) syntheses [62–65] and methanol [66–70] synthesis. These syngas reactions (C1 chemistry) require further technical development considering future growth in alternate fuel sectors, driven by both strategic and environmental considerations. In FT synthesis (Equation (8.2)), water, a by-product, oxidizes iron-based catalysts, thus causing their deactivation. nCO þ 2nH2 ! (  CH2 Þn þ nH2 O

ð8:2Þ

In addition, water adsorbed on iron-based catalysts inhibits the FT reaction. Therefore, keeping the water partial pressure at a minimum level in the reaction system, by simultaneously removing water with the membrane, would prevent catalyst deactivation and increase the reaction rate. In methanol synthesis

CH4/LPG

Hydrogen removal 450 °C, 0.1 MPa

De-watering 280 °C, 5 MPa De-watering, 250 °C, 5 MPa

CO+H

2 Air separation RT, 0.1-1 MPa Alcohol removal, de-watering 250 °C, 5 MPa MeOH

Air/O2

220 °C, 1 MPa Hydrogen removal 180 °C, 0.1 MPa

DMC Alcohol removal, 170 °C, 1 MPa

PMC EtOH

DME

De-watering

De-watering, 180 °C, 1 MPa

BioEtO

FT

Hydrogen removal 200 °C, 0.5 MPa

H2

Hydrogen removal 450 °C, 0.1 MPa

De-watering, 250 °C, 1 MPa

DEC

De-watering, 150 °C, 0.1 MPa

Energy source

Value-added products

FIGURE 8.6 Applications of membrane separation in C1 chemistry.

188

Inorganic, Polymeric and Composite Membranes

(Equations (8.3)–(8.5)), one-pass methanol yield is limited by the thermodynamic equilibrium relations of the following three reactions. CO þ 2H2 ! CH3 OH D H 0 ¼ 90:7 kJ mol1

ð8:3Þ

CO2 þ 3H2 ! CH3 OH þ H2 O D H 0 ¼ 49:5 kJ mol1

ð8:4Þ

CO þ H2 O ! CO2 þ H2 D H 0 ¼ 41:2 kJ mol1

ð8:5Þ

Removal of products (water and methanol) from the reaction system can overcome the equilibrium limitation and achieve an energy-efficient and costeffective process by improving the conversion level, downsizing the synthesis reactor, and minimizing energy consumption to recycle unconverted reaction gases. Considering these advantages, attempts to develop membrane reactors for FT syntheses [63,64] and methanol [67,69,70] were made using various membranes; however, the performance of these membrane reactors were very limited because of the lack of thermal stability and poor separation performance of the membranes used. Since these reactions are generally operated above 200  C, membranes should work above 200  C. Also, since small amounts of acid can also be produced in actual reaction systems [65], membranes should have acid resistance. An A-type zeolite membrane reported by Aoki et al. [71] showed high separation factors (awater/hydrogen > 160 at 30–200  C) though its stability above 200  C is as of yet unknown. Zhu et al. reported that heating a zeolite-4A membrane at 180  C in the absence of water vapor caused irreversible formation of defects [64]. Considering the stability required, therefore, A-type zeolite membrane may not be suitable in those applications. Rezai et al. [72] prepared compact siliceous ZSM-5 (Si/Al ¼ 62 and 157) membranes and investigated water/hydrogen separation properties in the temperature range of 20–350  C. The water/hydrogen separation factors were 14.3–19.7 at 20  C but approached 1 above 180  C. The observed water selectivity was attributed to weak adsorption of water on polar sites. Recently, the group of Matsukata has developed membranes with promising separation performance at reaction temperatures above 200  C by utilizing Na cations occluded in zeolitic ordered frameworks [41,73]. The membranes developed enables selective permeation and separation of water and/or methanol, from hydrogen above 200  C (Figure 8.7; [41]). This type of membrane is attractive because it can maintain high partial pressure of the reactant hydrogen, which is required for these reactions.

CONCLUDING REMARKS In this review, we have shown the substantial positive impact that membrane separations can have on the reduction of energy demand in distillation processes. Our preliminary study suggest that even though complete replacement

Chapter

8

189

Review on Prospects for Energy Saving

Zeolite framework

H2

Na

Na

H2O

CH3OH

FIGURE 8.7 Schematic illustration of selective permeation and separation of water and methanol from hydrogen. Water and/or methanol adsorbed on Na cation blocks hydrogen permeation through zeolite pores, while water and methanol permeate through.

of distillation with membrane separation is difficult due to the high investment risk, the energy saving obtained by just installing a membrane separation unit at the latter part of a distillation tower is large enough to reduce energy demand, by more than 70% in comparison with conventional distillation. To realize this innovative distillation–membrane separation system, research and development of zeolite-based separation membranes is recommended due to their high potential as membrane materials. There are many tasks to be solved for the large-scale applications of zeolite membranes in future chemical and petroleum industries, including improvement of membrane performance, development of membrane synthesis technology from the laboratory level to the mass-production level, and design of optimum process flow schemes. Therefore, the authors hope that continuous, collaborative, and increasing efforts will be paid to the development of zeolite-based membrane separation system.

REFERENCES [1] G.H. Koops, J.A.M. Nolten, M.H.V. Mulder, C.A. Smolders, Poly (vinyl chloride) poly acrylonitrile composite membranes for the dehydration of acetic acid, J. Membr. Sci. 81 (1993) 57–70. [2] S.K. Ray, S.B. Sawant, J.B. Joshi, V.G. Pangarkar, Dehydration of acetic acid by pervaporation, J. Membr. Sci. 138 (1998) 1–17. [3] S.P. Kusumocahyo, M. Sudoh, Dehydration of acetic acid by pervaporation with charged membranes, J. Membr. Sci. 161 (1999) 77–83. [4] T. Masuda, S. Otani, T. Tsuji, M. Kitamura, S.R. Mukai, Preparation of hydrophilic and acidproof silicalite-1 zeolite membrane and its application to selective separation of water from water solutions of concentrated acetic acid by pervaporation, Sep. Purif. Technol. 32 (2003) 181–189. [5] G. Li, E. Kikuchi, M. Matsukata, Separation of water-acetic acid mixtures by pervaporation using a thin mordenite membrane, Sep. Purif. Technol. 32 (2003) 199–206. [6] G. Li, E. Kikuchi, M. Matsukata, A study on the pervaporation of water-acetic acid mixtures through ZSM-5 zeolite membranes, J. Membr. Sci. 218 (2003) 185–194.

190

Inorganic, Polymeric and Composite Membranes

[7] B.J. Grashoff, C.E. Pilkington, C.W. Corti, The purification of hydrogen: a review of the technology emphasising the current status of palladium membrane diffusion, Platinum Met. Rev. 27 (1983) 157–169. [8] S. Uemiya, T. Matsuda, E. Kikuchi, Hydrogen permeable palladium-silver alloy membrane supported on porous ceramics, J. Membr. Sci. 56 (1991) 315–325. [9] Y.S. Lin, A.J. Burggraaf, CVD of solid oxides in porous substrates for ceramic membrane modification, AIChE J. 38 (1992) 445–454. [10] M. Matsukata, E. Kikuchi, Zeolitic membranes: synthesis, properties, and prospects, Bull. Chem. Soc. Jpn. 70 (1997) 2341–2356. [11] E.E. McLeary, J.C. Jansen, F. Kapteijn, Zeolite based films, membranes, and membrane reactors: progress and prospects, Microporous Mesoporous Mater. 90 (2006) 198–220. [12] J. Caro, M. Noack, Zeolite membranes—recent developments and progress, Microporous Mesoporpous Mater. 115 (2008) 215–233. [13] M. Matsukata, State-of-the art of hydrocarbon separation through inorganic membranes, Membrane(Maku) 23 (1998) 55–61. [14] M.E. van Leeuwen, Derivation of Stockmayer potential parameters for polar fluids, Fluid Phase Equlib. 99 (1994) 1–18. [15] M.C. Lovallo, M. Tsapatsis, T. Okubo, Preparation of an asymmetric zeolite-L film, Chem. Mater. 8 (1996) 1579–1583. [16] Y. Takata, T. Tsuru, T. Yoshioka, M. Asaeda, Gas permeation properties of MFI zeolite membranes prepared by the secondary growth of colloidal silicate and application to the methylation of toluene, Microporous Mesoporous Mater. 54 (2002) 257–268. [17] J. Hedlund, M. Noack, P. Kolsch, D. Creaser, J. Caro, J. Sterte, ZSM-5 membranes synthesized without organic templates using a seeding technique, J. Membr. Sci. 159 (1999) 263–273. [18] R. Lai, G.R. Gavalas, ZSM-5 membrane synthesis with organic-free mixtures, Microporous Mesoporous Mater. 38 (2000) 239–245. [19] M. Lassinantti, F. Jareman, J. Hedlund, D. Creaser, J. Sterte, Preparation and evaluation of thin ZSM-5 membranes synthesized in the absence of organic template molecules, Catal. Today 67 (2001) 109–119. [20] G. Li, E. Kikuchi, M. Matsukata, ZSM-5 zeolite membranes prepared from a clear templatefree solution, Microporous Mesoporous Mater. 60 (2003) 225–235. [21] M. Kanezashi, J.O. Brien, Y.S. Lin, Template-free synthesis of MFI-type zeolite membranes: permeation characteristics and thermal stability improvement of membrane structure, J. Membr. Sci. 286 (2006) 213–222. [22] G. Li, E. Kikuchi, M. Matsukata, The control of phase and orientation in zeolite membranes by the secondary growth method, Microporous Mesoporous Mater. 62 (2003) 211–220. [23] I. Kumakiri, T. Yamaguchi, S. Nakao, Preparation of zeolite A and faujasite membrane from a clear solution, Ind. Eng. Chem. Res. 38 (1999) 4682–4688. [24] H. Kita, K. Horii, Y. Ohtoshi, K. Tanaka, K. Okamoto, Synthesis of a zeolite NaA membrane for pervaporation of water-organic liquid mixtures, J. Membr. Sci. Lett. 14 (1995) 206–208. [25] K. Aoki, K. Kusakabe, S. Morooka, Gas permeation properties of A-type zeolite membrane formed on porous substrate by hydrothermal synthesis, J. Membr. Sci. 141 (1998) 197–205. [26] K. Aoki, K. Kusakabe, S. Morooka, Preparation of oriented A-type zeolite membranes, AIChE J. 46 (2000) 221–224. [27] K. Kusakabe, T. Kuroda, A. Murata, S. Morooka, Formation of a Y-type zeolite membrane on a porous a-alumina tube for gas separation, Ind. Eng. Chem. Res. 36 (1997) 649–655. [28] L. Washmonkriel, K.J. Balkus, Preparation and characterization of oriented SAPO-39 membranes, Microporous Mesoporous Mater. 38 (2000) 107–121.

Chapter

8

Review on Prospects for Energy Saving

191

[29] K.J. Balkus, G. Gbery, Z.S. Deng, Preparation of partially oriented zeolite MCM-22 membranes via pulsed laser deposition, Microporous Mesoporous Mater. 52 (2002) 141–150. [30] T. Munoz Jr., K.J. Balkus, Preparation of oriented zeolite UTD-1 membranes via pulsed laser ablation, J. Am. Chem. Soc. 121 (1999) 139–146. [31] X. Zhang, H. Liu, K.L. Yeung, Influence of seed size on the formation and microstructure of zeolite silicalite-1 membranes by seeded growth, Mater. Chem. Phys. 96 (2006) 42–50. [32] G. Li, Thesis, Waseda University, 2003. [33] M.C. Lovallo, A. Gouzinis, M. Tsapatsis, Synthesis and characterization of oriented MFI membranes prepared by secondary growth, AIChE J. 44 (1998) 1903–1913. [34] R. Lai, G.R. Gavalas, Surface seeding in ZSM-5 membrane preparation, Ind. Eng. Chem. Res. 37 (1998) 4275–7283. [35] G. Xomeritakis, A. Gouzinis, S. Nair, T. Okubo, M. He, R.M. Overney, et al., Growth, microstructure, and permeation properties of supported zeolite (MFI) films and membranes prepared by secondary growth, Chem. Eng. Sci. 54 (1999) 3521–3531. [36] G. Bonilla, D.G. Vlachos, M. Tsapatsis, Simulation and experiments on the growth and microstructure of zeolite MFI films and membranes made by secondary growth, Microporous Mesoporous Mater. 42 (2001) 191–203. [37] C. Algieri, G. Golemme, S. Kallus, J.D.F. Ramsay, Preparation of thin supported MFI membranes by in situ nucleation and secondary growth, Microporous Mesoporous Mater. 47 (2001) 127–134. [38] S. Nair, Z. Lai, V. Nikolakis, G. Xomeritakis, G. Bonilla, M. Tsapatsis, Separation of closeboiling hydrocarbon mixtures by MFI and FAU membranes made by secondary growth, Microporous Mesoporous Mater. 48 (2001) 219–228. [39] G. Shao, J. Yang, X. Zhang, G. Zhu, J. Wang, C. Liu, Seeded growth of beta zeolite membranes using zeolite structure-directing agent, Mater. Lett. 61 (2007) 1443–1445. [40] M. Matsukata, K. Sawamura, T. Shirai, M. Takada, Y. Sekine, E. Kikuchi, Controlled growth for synthesizing a compact mordenite membrane, J. Membr. Sci. 316 (2008) 18–27. [41] K. Sawamura, T. Izumi, K. Kawasaki, S. Daikohara, T. Ohsuna, M. Takada, et al., Chem. Asian J. 4 (2009) 1070–1077. [42] J. Hedlund, J. Sterte, M. Anthonis, A.-J. Bons, B. Carstensen, N. Corcoran, et al., High flux MFI membranes, Microporous Mesoporous Mater. 52 (2002) 179–189. [43] J. Hedlund, F. Jareman, A.-J. Bons, M. Anthonis, A masking technique for high quality MFI membranes, J. Membr. Sci. 222 (2003) 163–179. [44] Z. Lai, G. Bonilla, I. Diaz, J.G. Nery, K. Sujaoti, M.A. Amat, et al., Microstructual optimization of a zeolite membrane for organic vapor separation, Science 300 (2003) 456–460. [45] Y. Lu, R. Ganguli, C.A. Drewien, M.T. Anderson, C.J. Brinker, W. Gong, et al., Continuous formation of supported cubic and hexagonal mesoporous films by sol–gel dip-coating, Nature 389 (1997) 364–368. [46] J. Caro, M. Noack, J. Richter-Mendau, F. Marlow, D. Petersohn, M. Griepentrog, et al., Selective sorption uptake kinetics of n-hexane on ZSM-5—a new method for measuring anisotropic diffusivities, J. Phys. Chem. 97 (1993) 13685–13690. [47] J. Karger, Random walk through two-channel networks: a simple means to correlate the coefficients of anisotropic diffusion in ZSM-5 type zeolites, J. Phys. Chem. 95 (1991) 5558–5560. [48] L.J. Song, Z.L. Sun, L.V.C. Rees, Experimental and molecular simulation studies of adsorption and diffusion of cyclic hydrocarbons in silicalite-1, Microporous Mesoporous Mater. 55 (2002) 31–49.

192

Inorganic, Polymeric and Composite Membranes

[49] M. Noack, P. Kolsch, A. Dittmar, M. Stohr, G. Georgi, R. Eckelt, et al., Effect of crystal intergrowth supporting substances (ISS) on the permeation properties of MFI membranes with enhanced Al-content, Microporous Mesoporous Mater. 97 (2006) 88–96. [50] M. Noack, P. Kolsch, A. Dittmar, M. Stohr, G. Georgi, M. Schneider, et al., Proof of the ISSconcept for LTA and FAU membranes and their characterization by extended gas permeation studies, Microporous Mesoporous Mater. 102 (2007) 1–20. [51] H. Suzuki, Composite membrane having a surface layer of an ultrathin film of cage-shaped zeolite and processes for production thereof, (1987) U.S. Patent 4699892. [52] I.M. Lachman, M.D. Patil, Method of crystallizing a zeolite on the surface of a monolithic ceramic substrate, (1989) U.S. Patent 4800187. [53] S. Sakurada, N. Tagaya, Maejima, T. Isoda, Membranous synthetic zeolite and its manufacture, (1983) Japanese Patent, JP 59-213615. [54] H. Abe, Y. Fujita, Separation membrane, (1984) Japanese Patent, JP 63-287504. [55] Z. Liu, T. Ohsuna, K. Sato, T. Mizuno, T. Kyotani, T. Nakane, et al., Transmission electron microscopy observation on fine structure of zeolite NaA membrane, Chem. Mater. 18 (2006) 922–927. [56] Y. Morigami, M. Kondo, J. Abe, H. Kita, K. Okamoto, The first large-scale pervaporation plant using tubular-type module with zeolite NaA membrane, Sep. Purif. Technol. 25 (2001) 251–260. [57] K. Sato, K. Aoki, K. Sugimoto, K. Izumi, S. Inoue, J. Saito, et al., Dehydrating performance of commercial LTA zeolite membranes and application to fuel grade bio-ethanol production by hybrid distillation/vapor permeation process, Microporous Mesoporpous Mater. 115 (2008) 184–188. [58] Y. Cui, H. Kita, K. Okamoto, Zeolite T membrane: preparation, characterization, pervaporation of water/organic liquid mixtures and acid stability, J. Membr. Sci. 236 (2004) 17–27. [59] T. Inoue, T. Nagase, Y. Hasegawa, Y. Kiyozumi, K. Sato, M. Nishioka, et al., Stoichiometric ester condensation reaction processes by pervaporative water removal via acid-tolerant zeolite membranes, Ind. Eng. Chem. Res. 46 (2007) 3743–3750. [60] K. Sawamura, T. Shirai, T. Ohsuna, T. Hagino, M. Takada, Y. Sekine, et al., Separation behavior of steam from hydrogen and methanol through mordenite membrane, J. Chem. Eng. Jpn. 41 (2008) 870–877. [61] X. Li, H. Kita, H. Zhu, Z. Zhang, K. Tanaka, Synthesis of long-term acid-stable zeolite membranes and their potential application to esterification reactions, J. Membr. Sci. 339 (2009) 224–232. [62] R.L. Espinoza, E. du Toit, J. Santamaria, M. Menendez, J. Coronas, S. Irusta, Use of membranes in Fischer-Tropsch reactors, Stud. Surf. Sci. Catal. 130 (2000) 389–394. [63] M.P. Rohde, D. Unruh, G. Schaub, Membrane application in Fischer-Tropsch synthesis to enhance CO2 hydrogenation, Catal. Today 106 (2005) 143–148. [64] W. Zhu, L. Gora, A.W.C. van den Berg, F. Kaptein, J.C. Jansen, J.A. Moulijn, Water vapor separation from permanent gases by a zeolite-4A membrane, J. Membr. Sci. 253 (2005) 57–66. [65] Y. Li, H. Chen, J. Liu, H. Li, W. Yang, Pervaporation and vapor permeation dehydration of Fischer-Tropsch mixed-alcohols by LTA zeolite membranes, Sep. Purif. Technol. 57 (2007) 140–146. [66] R.P.W.J. Struis, S. Stucki, A membrane reactor for methanol synthesis, J. Membr. Sci. 113 (1996) 93–100. [67] R.P.W.J. Struis, S. Stucki, M. Wiedorn, Verification of the membrane reactor concept for the methanol synthesis, Appl. Catal. A Gen. 216 (2001) 117–129.

Chapter

8

Review on Prospects for Energy Saving

193

[68] G. Barbieri, G. Marigliano, G. Golemme, E. Drioli, Simulation of CO2 hydrogenation with CH3OH removal in a zeolite membrane reactor, Chem. Eng. J. 85 (2002) 53–59. [69] G. Chen, Q. Yuan, Methanol synthesis from CO2 using a silicone rubber/ceramic composite membrane reactor, Sep. Purif. Technol. 34 (2004) 227–237. [70] F. Gaullucci, L. Paturzo, A. Basile, An experimental study of CO2 hydrogenation into methanol involving a zeolite membrane reactor, Chem. Eng. Proc. 43 (2004) 1029–1036. [71] K. Aoki, K. Kusakabe, S. Morooka, Separation of gases with a A-type zeolite membrane, Ind. Eng. Chem. Res. 39 (2000) 2245–2251. [72] S.A.S. Rezai, J. Lindmark, C. Andersson, F. Jareman, K. Moller, J. Hedlund, Water/hydrogen/ hexane multicomponent selectivity of thin MFI membranes with different Si/Al ratios, Microporous Mesoporous Mater. 108 (2008) 136–142. [73] K. Sawamura, T. Shirai, M. Takada, Y. Sekine, E. Kikuchi, M. Matsukata, Selective permeation and separation of steam from water/methanol/hydrogen gas mixtures through mordenite membrane, Catal. Today 132 (2008) 182–187.

This page intentionally left blank

Chapter 9

Xylene Separation Performance of Composition-Gradient MFI Zeolite Membranes Jessica O’Brien-Abraham1, Mikel Duke2 and Y.S. Lin1,* Chemical Engineering, School for Engineering of Matter, Transport and Energy, Arizona State University, Tempe, Arizona, USA 2 Institute for Sustainability and Innovation, Victoria University, Werribee Campus, Melbourne, Victoria, Australia * Corresponding author: E-mail address: [email protected] 1

INTRODUCTION Supported MFI-type zeolite membranes are considered ideal candidates for separation of xylene isomers due to their molecular sieving separation capability. Given the average pore size, a high-quality MFI-type zeolite membrane is expected to be able to selectively pass p-xylene (PX, kinetic diameter dk ¼ 0.59 nm) while excluding the bulkier o- and m-xylene (OX and MX) isomers (dk ¼ 0.68 nm). The pore structure MFI-type zeolite consists of two channels: straight channels with circular openings of 0.54  0.56 nm along the b-axis and sinusoidal channels with elliptical openings of 0.51  0.55 nm along the a-axis. Transport is also possible through the channel intersections along the c-axis [1]. Previous work with MFI-type zeolite membranes demonstrates that the membranes are capable of performing the separation of xylene isomers under a variety of conditions, including vapor permeation and pervaporation. The former was more commonly investigated, and selectivities as high as 500 have been observed [2]. The high-vapor permeation selectivities are found only when operating at lower xylene partial pressures (0.27–2.5 kPa,  Pactual/Psat’d < 1%) [3]. Xomeritakis et al. [4] reported that the PX selectivity of MFI-type membranes dropped significantly as the vapor pressure of the feed was increased. Shape/size selectivity of the MFI-type membrane is therefore highly dependent on the PX loading. At high loadings of PX, the MFI-type zeolite framework experiences distortion which changes the pore shape and ultimately Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

195

196

Inorganic, Polymeric and Composite Membranes

allowing greater mobility of the larger isomers in the zeolitic pores [4–7]. At low partial pressures (i.e., low loadings), the deformation can be avoided, and the effect of the PX–framework interaction is negligible leading to molecular sieving of the isomers. Unfortunately, operating at such conditions can lead to impractically low fluxes of xylene through the membrane [4–6]. There are fewer results in current literature for pervaporation separation of xylene isomers using MFI-type zeolite membranes, and in general, lower selectivities are observed for this mode of operation. Lin and co-workers [8] found silicalite membranes synthesized with template-free secondary growth procedures to possess a maximum ideal selectivity of 69 and a separation factor of 40 with a 50%PX/50%OX feed, which is the highest reported so far. However, these data were accumulated within the first 3–5 h of experimentation on fresh membranes and no long term study of performance was conducted [8]. More commonly, moderate to low PX selectivities have been observed for pervaporation separation regardless of growth method [7,9,10]. Matsufuji et al. [7] conducted pervaporation studies on MFI-type zeolite membranes prepared by the vapor transport method. Initially, their membranes were found to be slightly PX selective in both equimolar binary (PX, MX) and ternary mixtures (PX, MX, OX) at 30  C but over time the PX permeance dropped below that of the other isomers, reversing the selectivity of the membrane. This was explained by enhanced adsorption of MX and OX which essentially blocked PX from transporting through the membrane [7]. The conventional approach for circumventing the low selectivity and lack of stability during pervaporation studies is to modify the synthesis conditions to improve the membrane quality, given that it has long been believed that poor membrane performance is due to the presence of nonselective microporous defects. An important observation made in studies of silicalite crystals exposed to xylene isomers is that the adsorption of PX results in significant framework distortion that is associated with structural phase changes [11–14]. It is known that the flexibility of the MFI-type framework arises from the flexibility of the Si–O–Si joints linking rigid SiO4 tetrahedra. This feature allows for the effective pore size of zeolite to vary continuously under the influence of the adsorbing species close in size to the pore diameter [15]. The loading of silicalite is marked by two distinct phase changes: the MONO to ORTHO transition at 4 PX molecules per unit cell (m.u.c.) and the ORTHO to PARA transition which occurs between 4 and 7.6 m.u.c. [16]. For the xylene–zeolite system, the two most favorable adsorption sites are in the channel intersections and the sinusoidal channels; at significantly high loadings, the straight channels will also be occupied [17]. In the ORTHO phase, single-component PX resides solely in the channel intersections [11,12] which increases the elliptical nature of the sinusoidal channels and renders them less accessible to further PX loading. Nair et al. studied the adsorption of single-component OX and MX in silicalite and determined that the molecules are able to pack in

Chapter

9

Zeolite Membrane for Xylene Separation

197

the channel intersections which cause severe distortion of the straight channels whereas it was previously thought that the molecules were excluded from the zeolite pores [15]. The transition from ORTHO to PARA is associated with larger changes to the zeolite microstructure. At this point, the channel intersections are filled to capacity, and the PX molecules must fit themselves into either the straight or sinusoidal (preferred) channels which causes displacement of framework ions [11]. Under these conditions, both the straight and sinusoidal channels have an increase in effective diameters, and these dimensional changes are large enough to significantly change the accessibility of OX and MX into the pore network [14,16,17]. Another factor that accounts for xylene isomer behavior is single file diffusion where molecules are unable to pass one another in the narrow zeolitic passageways. When low PX selectivity (or selectivity reversal) is observed, both molecules reside in the channel intersections, but OX will occupy a majority of the sinusoidal channels due to steric hindrance in the straight channels [14]. This forces PX to travel in the straight channels where movement from intersection to intersection is limited [11]. The overall effect is a faster OX transport rate than PX. At low concentrations of PX in the feed, the effect is reversed, and the membrane is once again PX selective [10]. A schematic representation of the transport difference which occurs as the loading of the membrane increases can be seen in Figure 9.1. Here, we see that as OX occupies favorable adsorption sites, the movement of the PX molecule becomes limited. Overall, the unit cell changes experienced for MFI-type zeolite saturated with PX conditions cause crystal expansion of approximately 0.39%, with the a- and caxes expanding 0.09% and 0.52%, respectively, and b-axis contracting 0.23% [10]. A secondary effect of these unit cell changes is that they are significant enough to effectively “seal” 60–90% of intercrystalline defects less than 0.86 nm in size for randomly oriented MFI-type zeolite membranes as they expand the effective zeolite pore diameter [10,18]. This distortion fundamentally OX in sinusoidal channels

PX in channel intersections and straight channels

OX in channel intersections FIGURE 9.1 Schematic representation of the position of OX and PX in the MFI framework for silicalite at high loading conditions. In this configuration, the framework is distorted to allow OX to enter the MFI pores, and it will occupy the channel intersections and sinusoidal channels preferentially (molecules and framework not to scale; for pictorial purpose only).

198

Inorganic, Polymeric and Composite Membranes

changes the access of the xylene isomers to transport pathways, allowing OX to travel through the zeolite pores and forcing competitive adsorption between the isomers for favorable sites [10]. The difficulty in achieving a highly selective, stable membrane is a result of these microstructure changes, which are inherent to the membrane material and independent of defect concentration [18]. The first attempt by our group to address the performance and stability issues of xylene pervaporation involved synthesizing MFI-type zeolite membranes with Al and B isomorphously substituted into the crystalline framework [18]. It was discovered that the unit cell changes associated with substitution were able to enhance the ability of the MFI-type zeolite framework to resist deformation at high loadings of PX, but there was little improvement to stability [18]. When Al replaces smaller Si in the framework, an overall unit cell expansion of  1.3% occurs as a result of the large increase in dSi–Al and dAl–O bond lengths. This is accompanied by a reduction in the Si–O–Al angle, which induces localized framework rigidity [19]. These minor structural changes are only beneficial at lower loadings because, under this condition, the PX molecules will preferentially adsorb on intersections that contain SiOHAl groups, taking advantage of the localized rigidity and preventing OX transport. These differences are shown schematically in Figure 9.2. At higher loadings, molecules are forced into intersections where the bridging groups do not exist [20]. As a result, there is no resistance to distortion, and steady-state behavior similar to unsubstituted silicalite is observed [19]. It is important to develop a membrane and identify operating conditions that prevent/mitigate the effects of framework distortion and favor the transport of PX over OX at high xylene loadings of the MFI-type zeolite framework. In this work, we report a new approach to improve the MFI zeolite membrane for xylene separation. The approach taken here is to prepare bilayer MFI-type zeolite membranes, with top layer having high Si/Al ratio (silicalite) and the sublayer with a low Si/Al ratio (ZSM-5). In this new bilayer silicalite/ZSM-5 membrane, the top layer operates at high xylene loading where adsorption affinity dominates separation; the sublayer operates at low xylene loading, exploiting the molecular sieving function of MFI-type zeolite. The present paper reports results of multicomponent pervaporation of PX and OX for the silicalite/ZSM5 bilayer membranes and the effect of feed composition and experiment duration on membrane quality and separation results. The results show enhanced selectivity and stability with flux comparable to single layer MFI-type zeolite membranes.

EXPERIMENTAL Bilayer Membrane Synthesis and Characterization MFI-type zeolite layers were synthesized using a template-free secondary growth method as developed by Lin and co-workers [21] on disk-shaped a-Al2O3 supports. The disks, 20 mm in diameter and 2 mm thick (average pore

Chapter

9

199

Zeolite Membrane for Xylene Separation

(a) Low Loading PX loaded at type 2 intersections

Type 1 intersection OX excluded from framework at both type 1 and 2 intersections

Type 2 intersection (b) High Loading OX access to sinusoidal channels through type 1 intersections

PX loaded at type 1 and 2 intersections

OX restricted in type 2 intersections but not type 1 FIGURE 9.2 Schematic representations of the position of OX and PX in the MFI framework for silicalite at low and high loading conditions. In the low loading condition, PX prefers to adsorb at the type 2 intersections which distort less than type 1 causing restriction of OX transport and a PX selective membrane. At high loading, PX will adsorb in both type 1 and 2 intersections, and while type 2 intersections offer more resistance, OX still gains access to the MFI pores via nonsubstituted intersections.

diameter; 0.2 mm, porosity 45%) [3], were made with calcined alumina powder (Alcoa, A-16). Details of support preparation can be found in previous publications [9,22]. The bilayer membranes were constructed in a layer by layer fashion with an initial deposition of silicalite seeds followed by two consecutive growths of either ZSM-5 or silicalite depending on the desired membrane structure. Seed layer deposition onto polished supports was conducted using a silicalite seed suspension with the composition 10SiO2:2.4TPAOH:1NaOH:110H2O synthesized hydrothermally at 125  C for 8 h and diluted to 1–2 wt% in

200

Inorganic, Polymeric and Composite Membranes

deionized water. To ensure significant coverage, the supports were dip coated at least three times as demonstrated in a previous publication by our group [3]. Combinations of silicalite and ZSM-5 (Si/Al ¼ 20) layers were grown on the seeded supports through subsequent growths after a 24-h drying time at 40  C (40% relative humidity). The silicalite layers were grown by hydrothermal synthesis (8 h, 175  C) with a clear sol, previously aged for 24 h, composed of 1SiO2:xNaOH:35H2O, where x ranged from 0.05 to 0.6. Similarly, the ZSM5 layers were synthesized (8 h, 175  C) using a solution with the molar composition 6SiO2:1.4NaOH:200H2O: 0.3NaAlO2 yielding an Si/Al ratio of 20. All membrane growth conditions (i.e., temperature, duration) were kept the same to maintain relatively similar thicknesses. Membrane microstructure was characterized by X-ray diffraction (XRD) (Bruker AXS-D8, Cu Ka radiation). Thickness and morphology were examined by scanning electron microscopy (SEM; Philips, FEI XL 30).

Pervaporation Experiments Pervaporation experiments of xylene isomers were conducted on single layer and bilayer membranes. Experiments were carried out in two modes: (1) transient operation where the feed was changed after 2 h sampling at each composition (no regeneration between samples) and (2) steady-state operation holding a single feed composition for a maximum of 96 h. The reliability and reproducibility of these techniques are published elsewhere [10]. Binary feeds composed of pure component PX (99%, Aldrich) and OX (99%, Aldrich) were used for pervaporation studies at room temperature. The pervaporation apparatus used in this work is presented in detail in a previous publication [22]. The membrane was sealed in the vertical stainless steel cell with top layer facing upwards. The liquid feed was contained in a reservoir above, while vacuum was applied to the downstream side. Permeate vapors were captured in liquid nitrogen cold traps, and measurements were taken by weighing the cold trap before and after each run. Membrane regeneration was performed between experiments by heat treating at 150  C for 24 h in a vacuum oven (Precision model 6500, Thermo Electron Corporation) [10]. Sample processing involved removing a 1-mL permeate sample from the cold trap and dilution in 10 mL of toluene (99.5%, Mallinckrodt). Compositional analysis was conducted using a gas chromatograph (6890N, Agilent) with mass spectrometer (5973N, Agilent) fitted with a DB-Waxetr capillary column (J&W Scientific-Agilent Technologies) using ultra-high purity He as a carrier gas. Feed composition was evaluated in the same manner both pre- and postrun to ensure that the component concentrations remained constant [10]. The membrane selectivity for component i over component j in a mixture, represented by selectivity factor, aij, is defined as

Chapter

9

201

Zeolite Membrane for Xylene Separation

aij ¼

yi =yj xi =xj

ð9:1Þ

where yi and yj are the mole fractions of components i and j in the permeate, and xi and xj are the mole fractions of i and j in the feed. For this process, the error associated with the flux and selectivity measurements is  10%.

RESULTS AND DISCUSSION Membrane Characteristics Table 9.1 summarizes the structures of the membranes prepared in this. For clarity, membrane notation will consist of top layer_sublayer; where silicalite and ZSM-5 are denoted by Sil(Na/Si ratio) and Z, respectively. For example, membrane Sil(0.35)_Z possesses a top layer of silicalite with an Na/Si ratio of 0.35 and a sublayer of ZSM-5 (Si/Al ¼ 20). The membranes prepared in this work include pure silicalite and ZSM-5 membranes, and two bilayer silicalite/ZSM-5 membranes with silicalite layer either in top or in bottom, as summarized in Table 9.1. Figure 9.3 shows SEM image of the cross-sectional view of a silicalite/ ZSM-5 bilayer membrane on alumina support. The SEM image shows a thickness of about 4 mm for each layer, with a total thickness of about 7–9 mm for both MFI layers. The XRD patterns show randomly oriented MFI microstructure. The XRD peaks for ZSM-5 sublayer, in comparison with the peaks for alumina, are much smaller than those for the silicalite/ZSM-5 bilayer, because of the smaller membrane thickness for the former. A typical EDS spectrum for silicalite/ZSM-5 bilayer membrane is given in Figure 9.4. The data confirms the differences in composition observed throughout the membrane. The top layer of silicalite has an approximate Si/Al ratio of 180 which decreases with depth into the membrane as a result of increasing Al content. It must be noted that due to the moderate spot size and high voltage used for optimum imaging,

TABLE 9.1 MFI-Type Membranes (Single and Bilayer) Synthesized Toplayer, 8 h (175  C)

Sublayer, 8 h (175  C)

Membrane

Si/Al ()

Na/Si ()

Si/Al ()

Na/Si ()

Silicalite

–

–

1

0.25

ZSM-5

–

–

20

0.23

Sil(0.35)_Z

1

0.35

20

0.23

Z_Sil(0.25)

20

0.23

1

0.25

Molar ratios denote composition in synthesis solution.

202

Inorganic, Polymeric and Composite Membranes

(a)

Silicalite layer

ZSM-5 sublayer

(b)

*

Relative intensity (a.u.)

*

ZSM-5 sublayer

*

* *

* Silicalite/ZSM-5 bilayer

5

10

15

20

*

25 2q (°)

*

30

35

40

45

FIGURE 9.3 SEM image of the cross-section of silicalite/ZSM-5 bilayer membrane on alumina support (a), and XRD spectra for ZSM-5 sublayer and silicalite/ZSM-5 bilayer on alumia support (asterisks denotes a-Al2O3 support peaks) (b).

there is a large volume of interaction on the cross-section surface which affects the elemental analysis at a particular location in the membrane. However, composition differences through the thickness are verified.

Binary Pervaporation Through Single and Bilayer Membranes Figure 9.5 demonstrates the performance of single layer ZSM-5, silicalite, and bilayer Sil(0.35)_Z as a binary (PX, OX) feed composition was varied. While neither the silicalite nor the ZSM-5 layer alone provides the selectivity or stability desired for the pervaporation separation of xylene isomers, the performance of Sil(0.35)_Z indicates a synergistic effect when transport is in series through the layers. Enhanced performance is observed for PX feed concentrations between  5% and 20% where the selectivity reaches a maximum of 150 for Sil(0.35)_Z. It can be seen that the performance of the bilayer

9

203

Zeolite Membrane for Xylene Separation

250

50

200

40

150

30

100

20

50

10

0

0

2

4 6 Membrane thickness (mm)

Al content (At%)

Si/Al (molar ratio)

Chapter

0 10

8

FIGURE 9.4 Elemental analysis for silicalite/ZSM-5/support bilayer membrane including the molar ratio (Si/Al) and Al determined as a function of membrane depth from the zeolite layer.

150

Selectivity (aPX/OX)

125

100

Selectivity (aPX/OX)

150 Sil(0.35)_Z

125 100 75 50 25

75 0 0 50

24

48 Time (h)

72

96

ZSM-5 25 Silicalite 0 0

0.2

0.6 0.4 Mass fraction PX in feed (–)

0.8

1

FIGURE 9.5 Observed PX selectivity in a binary PX–OX system for single layer silicalite, single layer ZSM-5, and bilayer Sil(0.35)_Z as a function of mass fraction of PX in the feed, inset shows the selectivity of each membrane with time at a feed composition of 5%PX:95%OX over 96 h of operation.

membrane does not vary significantly from its single layer counterparts at higher concentrations of PX in the feed. The profiles exhibited by all MFITYPE membranes are a result of the interplay between PX–OX and PX–framework interactions and are observed regardless of quality, microstructure, and composition [18].

204

Inorganic, Polymeric and Composite Membranes

At sufficiently low PX concentrations ( 5 wt% PX), the membranes are highly selective for PX. As the PX concentration in the feed is increased so is the loading of the framework which induces greater structural deformation and mobility of OX. At the highest concentrations of PX in the feed, selectivity reversal is observed (a < 1), as the larger isomer has greater transport capability than PX. While it may seem counter intuitive that the MFI-type zeolite membrane would exhibit selectivity for a larger molecule, this selectivity reversal can be explained by adsorbate–adsorbate interactions that determine which molecule will permeate fastest [17]. In a study on the sorbate interactions with zeolite, Monson et al. [23] found that the ordering of a fluid within a pore system is dependent on both molecular size and adsorbate–pore interactions. In general, a smaller molecule will be expected to have promoted adsorption. In a mixture where the various components have similar adsorbate–pore interactions, the intermolecular interactions between components will play a significant role [23]. In the case of xylene isomers traveling through MFI-type zeolite membrane, the separation mechanism exhibited by the membrane changes with adsorbate concentration within the framework. At low occupancy, PX has the favorable adsorbate–pore interaction because OX is excluded from the zeolite pores, given its size and the result is a molecular sieving separation mechanism. As the distortion increases with membrane loading and OX gains access to the zeolite pores, the separation mechanism changes from molecular sieving to size/shape selectivity. When both isomers can enter the framework, they must compete with one another for favorable adsorption sites and diffusion pathways. Under these conditions, the molecule with the better fit into the pore space (regardless of size) will offer the fastest transport. As has been seen in the past, this fit is highly dependent on component concentration within the framework [4,10]. Figure 9.5 (inset) shows the performance of the single layer membranes and bilayer Sil(0.35)_Z subjected to a feed composition of 5%PX:95%OX for a period of 96 h. In general, it is known that the high selectivities observed (in the low PX regions) are not stable due to a gradual building of PX in the MFI-type zeolite framework which allows for eventual mobility of OX and subsequent degradation of performance [10]. This is the case for both the silicalite and ZSM-5 single layer membranes where after 96 h, little to no PX selectivity is observed. In contrast, Sil(0.35)_Z experiences an initial drop in selectivity but exhibits a final aPX/OX  60, demonstrating significantly enhanced stability. The component flux profiles during the steady-state operation (5%PX:95% OX, 96 h) for single layer silicalite, single layer ZSM-5, and bilayer Sil(0.35)_Z are shown in Figure 9.6a–c, along with illustrations of the various membrane structures. Figure 9.6a and b shows that within the first 48 h, OX dominates the overall transport through the membrane as the flux of PX is significantly reduced. A very different component flux profile emerges for the bilayer

Chapter

9

205

Zeolite Membrane for Xylene Separation

(a)

0.1 0.08 0.06 0.04 0.02 0 0

24

48

72

96

24

48

72

96

24

48

72

96

24

48 Time (h)

72

96

(b)

0.2 0.15

Component flux (kg m-2 h-1)

0.1 0.05 0 0 (c)

0.2 0.16 0.12 0.08 0.04 0 0

(d)

0.2 0.16 0.12 0.08 0.04 0 0

FIGURE 9.6 PX (dotted) and OX (solid) flux as a function of time through MFI-type zeolite membranes of varying composition (a-silicalite on alumina, b-ZSM-5 on alumina, c-silicalite/ ZSM-5 on alumina, and d-ZSM-5/silicalite on alumina).

206

Inorganic, Polymeric and Composite Membranes

membrane; it can be seen that the PX and OX fluxes stabilize after 48 h, and the membrane remains highly PX selective (Figure 9.5c). The behavior of the xylene isomers through the bilayer membrane is shown schematically in Figure 9.7. It is believed that in this case, the top layer silicalite is acting as a sacrificial layer where it will saturate fully in the presence of the liquid binary feed and will experience the associated distortion and crystallite expansion similar to case Figure 9.1. While the silicalite top layer is essentially nonselective and serves to reduce the loading on the ZSM-5 sublayer, the ZSM5 layer is able to retain its low loading characteristics with preferential adsorption occurring only in the intersections containing SiOHAl groups where the local framework is stiffer. This has the effect of reducing the distortion and blocking OX from transporting through MFI-type zeolite pathways allowing the sublayer to act in a permselective manner similar to low loading condition in Figure 9.2. In this case, the high loading restriction on selectivity is circumvented, and the ZSM-5 layer performance resembles that of vapor permeation where the highest PX selectivities have been reported [2,4]. The data presented in Figure 9.8 validate this theory. The component flux data for the sublayer ZSM-5 (acting alone) and after silicalite deposition (Sil (0.35)_Z) are shown as a function of mass fraction of PX in the feed. It can be seen that for both single and bilayer, the PX dominates the overall transport but the ZSM-5 layer experiences significantly higher OX flux as a single layer than after silicalite growth. The bilayer demonstrates a higher PX flux due to the fact that the smaller isomer no longer has to engage in competitive adsorption. The ZSM-5 sublayer is able to limit the transport of OX in the zeolite pores which allows PX to travel freely.

OX in intersections and sinusoidal channels

PX in channel intersections and straight channels

Silicalite ZSM5

OX access to some intersections

PX loaded at type 2 intersections and sinusoidal channels

FIGURE 9.7 Schematic representations of the position of OX and PX in the MFI framework for silicalite/ZSM-5 bilayer at high loading. The top silicalite layer behaves as it does in single layer form at high concentrations (Figure 9.1). The bottom ZSM-5 layer will behave as it does in a single layer form at low concentrations (Figure 9.2). This synergy allows for significantly enhanced membrane selectivity.

Chapter

9

207

Zeolite Membrane for Xylene Separation

Component flux (kg m-2 h-1)

0.2

0.16 Sil(0.35)_Z 0.12

0.08

0.04 Single layer ZSM-5 0 0

0.2

0.4 0.6 Mass fraction of PX in feed (-)

0.8

1

FIGURE 9.8 PX (open symbols) and OX (closed symbols) flux as a function of feed composition for bilayer Sil(0.35)_Z and single layer ZSM-5.

Reversal of Bilayer Structure Reversal of the membrane structure (silicalite on ZSM-5 to ZSM-5 on silicalite) also results in improved membrane performance over single layer membranes. In this case, the ZSM-5 layer is acting as the sacrificial layer for the underlying silicalite. Figure 9.9 shows the selectivity behavior of Z_Sil(0.25) as a function of mass fraction of PX in the feed. It was found that at almost every composition, the bilayer membrane performed better than the silicalite sublayer (acting alone). However, at a feed composition of 5%PX, reduced performance of the bilayer was observed when compared to its single layer counterpart. Similar behavior was found for ZSM-5 membranes in previous work [18] and was determined to be the result of a reduction in the effectiveness of defect sealing. Given that ZSM-5 has some resistance to deformation at low loadings of PX, the crystallite expansion experienced by the membrane is reduced, allowing flow through microporous defects that are present. In the bilayer assembly, defects in the top layer essentially short circuit the bilayer membrane allowing significantly greater transport of both isomers from one layer to the other and increasing the loading on the underlying silicalite layer which has no resistance to deformation. It is known that the presence of defects (that cannot be sealed through crystallite expansion) is detrimental to the selectivity. It has been suggested that despite the fact that PX is smaller and has a greater degree of symmetry, its bulky aromatic group prevents it from being able to pack efficiently in irregular microporous defects, which limits its ability to be able to block OX from transport. Gump et al. [24] proposed the affinity of the silicalite for PX over OX leads to adsorption of PX through

208

Inorganic, Polymeric and Composite Membranes

40

Selectivity (a PX/OX)

40 Silicalite 30 Z_Sil(0.25)

Selectivity (a PX/OX)

50

30 20 10 0

20

0

24

48 Time (h)

72

96

10

0 0

0.2

0.4 0.6 Mass fraction of PX in feed (-)

0.8

1

FIGURE 9.9 Observed PX selectivity for single layer silicalite shown as reference and bilayer Z_Sil(0.25) as a function of mass fraction of PX in the feed; inset shows the selectivity of both membranes with time at a feed composition of 5%PX:95%OX for 96 h.

the walls of the MFI-type zeolite crystals which acts as a sink for the continuous removal of PX from the intercrystalline gaps and enhances the OX permeance through defect pores [24]. Membrane Z_Sil(0.25) was also subjected to steady-state testing with a binary feed composition of 5%PX:95%OX for 96 h (Figure 9.8, inset). It was found that Z_Sil(0.25) did not exhibit a large decrease in membrane performance over 96 h and demonstrated a final selectivity  17. The component flux as a function of time at 5%PX in the feed for Z_Sil(0.25) is shown in Figure 9.6d. OX is dominating the overall transport which demonstrates that the ZSM-5 layer is not as effective at protecting the underlying silicalite layer (i.e., preventing significant loading and competitive adsorption). These data suggest that the ZSM-5/silicalite bilayer has enhanced performance over single layer silicalite due to a reduction in loading at the interface but without an increase in framework rigidity, the sublayer is still susceptible to deformation. As a result, the larger OX has the ability to travel at a faster ratio than PX through Z_Sil(0.25). From these results, we find the sacrificial top layer possesses two roles in the bilayer structure: the first is the sealing of intercrystalline defects from transport, and the second is the reduction of the loading on the underlying layer to prevent the transport of OX at the expense of PX. To accomplish these functions, the top layer must be continuous and well intergrown with as few defects as possible. Table 9.2 summarizes steady-state pervaporation results for both single and bilayer membranes investigated by our group. Unlike during vapor permeation operation, selectivity improvement in this work is not at the expense of the PX

Chapter

9

209

Zeolite Membrane for Xylene Separation

TABLE 9.2 Summary of Single and Bilayer Membrane Results when Subjected to Pervaporation for 96 h with a Feed Composition of 5%PX:95% OX at 25  C Membrane

PX Flux ( 10 2 kg m 2 h 1)

OX Flux (10 2 kg m 2 h 1)

aPX/OX

Silicalite

3

28

2

ZSM-5

7

143

1

Sil(0.6)_Z

5

4

20

Sil(0.35)_Z

13

4

61

Sil(0.05)_Z

1

15

1

Z_Sil(0.25)

8

11

17

flux through the membrane. This is due to only slight increase in thickness for the multilayer structures and a reduction in competitive adsorption between xylene isomers. Sil(0.35)_Z is the only membrane to show an appreciable dominance of PX transport over OX after 96 h of experimentation and is determined to be the best composition for xylene separation within the context of this work. The pervaporation performance of the bilayer membranes is among the best to be reported so far in the literature and the first to demonstrate high selectivity over a period of time.

Stability at Higher PX Feed Concentrations Table 9.3 lists performance values for Sil(0.35)_Z and Z_Sil(0.25) subjected to pervaporation for a period of 96 h at various feed compositions; membrane regeneration was conducted between experimental runs. The values of the steady-state flux of OX and PX demonstrate the effect of reduced framework distortion experienced by both bilayer structures. For Sil(0.35)_Z, there is only a slight difference in isomer transport as the PX concentration is increased, indicating that the microstructure is not deforming further with increased PX exposure. The component flux behavior for Z_Sil(0.25) confirms the presence of unsealed defects. As the driving force (concentration) of PX increases, so does the flux of PX through the membrane (similar behavior is observed for OX). When this phenomenon is observed, it is an indication of free flow of both isomers through membrane defects where the restrictions (i.e., adsorption, sieving, etc.) posed by traveling through the micropores are not determining the rate of transport [18]. Regardless, both bilayer membrane types exhibit a decrease in PX selectivity as the amount of PX in the feed is increased. It is clear that while the bilayer can provide enhanced stability at sufficiently low PX loadings, under more

210

Inorganic, Polymeric and Composite Membranes

TABLE 9.3 Summary of Performance Results for Bilayer Membranes at 25  C After 96 h of Pervaporation at the Listed Feed Composition; Balance OX PX Flux ( 10 2 kg m 2 h 1)

OX Flux ( 10 2 kg m 2 h 1)

aPX/

Membrane

Feed Composition (%PX)

Sil(0.35)_Z

5

13

4

61

23

12

6

7

60

11

7

1

5

8

11

17

23

10

8

4

60

17

2

4

90

16

3

1

Z_Sil(0.25)

OX

severe conditions, the membrane performance still suffers. To operate at higher loadings during pervaporation, further optimization of the bilayer structure is needed. For example, a finer parametric study varying zeolite Na and Al composition would be valuable in interpreting further their effects in this application.

CONCLUSIONS Enhanced xylene separation performance in terms of membrane selectivity and stability was observed for membranes with the two-layer (bilayer) zeolite structure. Silicalite on ZSM-5 proved to be the optimum layer combination exhibiting steady-state aPX/OX of 20–60 with moderate PX fluxes  0.05–0.13 kg m 2h 1 (5%PX feed concentration, 96 h). The stability improvement was a result of the top layer silicalite acting as a sacrificial layer where it will fully saturate in the presence of the liquid binary feed and will experience full distortion and crystallite expansion. The silicalite behaves as it does when it is present as a single layer possessing low PX selectivity but it serves to reduce the loading on the ZSM-5 sublayer. The ZSM-5 layer was therefore acting in the selective role whereby under lower loading conditions, preferential adsorption occurs only in the intersections containing SiOHAl groups where the local framework is stiffer, therefore reducing the distortion and effectively blocking OX from transporting through MFI-type zeolite pathways. The result is a membrane that exhibits high PX selectivity which appears to be stable over the time investigated. While the bilayer can provide enhanced stability at sufficiently low PX loadings, the membrane performance suffers under more severe conditions.

Chapter

9

Zeolite Membrane for Xylene Separation

211

ACKNOWLEDGMENTS The authors would like to thank the Department of Energy for their support (DE-PS3603GO93007). MCD acknowledges the support from the Australian Research Council Linkage International Fellowship (LX0775930).

REFERENCES [1] Y.S. Lin, I. Kumakiri, B.N. Nair, H. Alsyouri, Microporous inorganic membranes, Sep. Purif. Technol. 31 (2002) 229. [2] Z. Lai, G. Bonilla, I. Diaz, J.G. Nery, K. Sujaoti, M.A. Amat, et al., Microstructural optimization of a zeolite membrane for organic vapor separation, Science 300 (2003) 456. [3] M. Kanezashi, J. O’Brien, Y.S. Lin, Template-free synthesis of MFI-zeolite membranes: permeation characteristics and thermal stability improvement of membrane structure, J. Membr. Sci. 286 (2006) 213. [4] G. Xomeritakis, M. Tsapatsis, Permeation of aromatic isomer vapors through oriented MFItype membranes made by secondary growth, Chem. Mater. 11 (1999) 875. [5] X. Gu, J. Dong, T.M. Nenoff, D.E. Ozokwelu, Separation of p-xylene from multicomponent vapor mixtures using tubular MFI zeolite membranes, J. Membr. Sci. 280 (2006) 624. [6] J. Hedlund, J. Sterte, M. Anthonis, A.J. Bons, et al., High-flux MFI membranes, Microporous Mesoporous Mater. 52 (2002) 179. [7] T. Matsufuji, N. Nishiyama, M. Matsukata, K. Ueyama, Separation of butane and xylene isomers with MFI-type zeolitic membrane synthesized by a vapor-phase transport method, J. Membr. Sci. 178 (2000) 25. [8] W. Yuan, Y.S. Lin, W. Yang, Molecular sieving MFI-type membranes for pervaporation separation of xylene isomers, J. Am. Chem. Soc. 126 (2004) 4776. [9] K. Wegner, J. Dong, Y.S. Lin, Polycrystalline MFI zeolite membranes: xylene pervaporation and its implication on membrane microstructure, J. Membr. Sci. 158 (1999) 17. [10] J. O’Brien-Abraham, M. Kanezashi, Y.S. Lin, Effects of adsorption-induced microstructural changes on separation of xylenes isomers through MFI-type zeolite membranes, J. Membr. Sci. 320 (2008) 505. [11] H. van Koningsveld, F. Tuinstra, The location of p-xylene in a single crystal of zeolite H-ZSM-5 with a new, sorbate-induced, orthorhombic framework symmetry, Acta Crystallogr. B45 (1989) 423. [12] S. Nair, Z. Lai, V. Nikolakis, G. Xomeritakis, G. Bonilla, M. Tsapatsis, Separation of closeboiling hydrocarbon mixtures by MFI and FAU membranes made by secondary growth, Microporous Mesoporous Mater. 48 (2001) 219. [13] S. Mohanty, H.T. Davis, A.V. McCormick, Shape selective adsorption in atomistic nanopores—a study of xylene isomers in silicalite, Chem. Eng. Sci. 55 (2000) 2779. [14] S. Chempath, R.Q. Snurr, J.J. Low, Molecular modeling of binary liquid-phase adsorption of aromatics in silicalite, AIChE J. 50 (2004) 463. [15] S. Nair, M. Tsapatsis, The location of o- and m-xylene in silicalite by powder x-ray diffraction, J. Phys. Chem. B 104 (2000) 8982. [16] B.F. Mentzen, P. Gelin, The silicalite/p-xylene system: part I—flexibility of the MFI framework and sorption mechanism observed during p-xylene pore-filling by x-ray powder diffraction at room temperature, Mater. Res. Bull. 30 (1995) 373. [17] A.V. Mohanty, McCormick, Prospects for principles of size and shape selective separations, Chem. Eng. J. 74 (1999) 1.

212

Inorganic, Polymeric and Composite Membranes

[18] J. O’Brien, A study of the microstructure-property relationship for MFI-type zeolite membranes for xylene separation, Ph.D. Dissertation, Arizona State University, 2009. [19] R.R. Mukti, A. Jentys, J.A. Lercher, Orientation of alkyl-substituted aromatic molecules during sorption in the pores of H/ZSM-5 zeolites, J. Phys. Chem. C 111 (2007) 3973. [20] C.J. Gump, V.A. Tuan, R.D. Noble, J.L. Falconer, Aromatic permeation through crystalline molecular sieve membranes, Ind. Eng. Chem. Res. 40 (2001) 565. [21] M. Pan, Y.S. Lin, Template-free secondary growth synthesis of MFI type zeolite membranes, Microporous Mesoporous Mater. 43 (2001) 319. [22] J. O’Brien-Abraham, M. Kanezashi, Y.S. Lin, A comparative study on permeation and mechanical properties of random and oriented MFI-type zeolite membranes, Microporous Mesoporous Mater. 105 (2007) 140. [23] P.A. Monson, The properties of inhomogeneous square-well mixtures in one dimension, Mol. Phys. 70 (1990) 401. [24] G. Valerio, J. Plevert, A. Goursot, F. di Renzo, Modeling of boron substitution in zeolites and implications on lattice parameters, Phys. Chem. Chem. Phys. 2 (2000) 1091.

Chapter 10

Membrane Extraction for Biofuel Production David L. Grzenia1, Xianghong Qian2, Silvio Silverio da Silva3, Xinying Wang1 and S. Ranil Wickramasinghe1,* Department of Chemical and Biological Engineering, Colorado State University, Fort Collins, Colorado, USA 2 Department of Mechanical Engineering, Colorado State University, Fort Collins, Colorado, USA 3 Department of Biotechnology, School of Engineering of Lorena, University of Sa˜o Paulo, Lorena/SP, Brazil * Corresponding author: E-mail address: [email protected] 1

INTRODUCTION Development of energy-efficient process for sustainable production of fuels and chemicals is critical, given declining world petroleum reserves, increasing energy needs by emerging economies, and political and environmental concerns associated with using fossil fuels [1]. Figure 10.1 is a summary of the breakdown of the US energy supply for 2003 [2]. Fossil fuels (petroleum, coal, and natural gas) accounted for about 86% of the US energy supply in 2003, while renewable sources accounted for combined 6%, 47% of which came from biomass. In addition, energy demand is projected to grow by more than 50% by 2025 [3]. Clearly, finite petroleum resources cannot meet the increasing energy demand. Solar, 1% Wind, 2% Geothermal, 5%

Natural gas, 24%

Coal, 23% Hydroelectric 45%

Biomass, 47%

Nuclear, 8% Petroleum, 39%

Renewable energy, 6%

FIGURE 10.1 Summary of biomass resource consumption modified from Ref. [2]. Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

213

214

Inorganic, Polymeric and Composite Membranes

The form of renewable energy that can contribute substantially to energy needs at costs competitive to fossil fuels is the solar energy captured by photosynthesis and stored in biomass [4]. Currently, the United States produces about 4.5
109 gallons of ethanol annually from about 90 corn grain-to-ethanol refineries (noncellulosic). It is estimated that 130
109 gallons of fuel ethanol could be produced from cellulosic biomass in the United States [4]. Improving biomass to ethanol conversion processes to minimize processing costs will have a significant impact on the economy. A major obstacle to the large-scale conversion of lignocellulosic biomass to fuels and chemicals is the lack of cost-effective separations for product isolation and purification, as separations can account for up to 60–80% of the processing costs. While distillation dominates in petroleum refinery separations, given the nonvolatile nature of most biomass components, it is likely that solvent extraction will dominate in future biorefineries [5]. Solvent (liquid–liquid) extraction involves the transfer of the solute species from the feed to the extracting solvent [6]. Two immiscible phases are brought together to promote good mass transfer. This is achieved by dispersing one phase in the other. Consequently, the dispersed phase must be coalesced and the phases separated after extraction of the solute. The amount of solute transferred to the solvent will be limited by its partition coefficient between the feed and solvent phases. Solvent extraction is generally conducted using a number of mixer settlers in series or a continuous countercurrent extraction column. However, irrespective of the equipment used, conventional extraction operations suffer from a number of disadvantages. These include dispersion of one phase in the other, which requires subsequent coalescence and phase separation, emulsification problems, flooding and loading concerns, and scale-up related issues [7]. Membrane-based solvent extraction, using microporous hollow fibers, overcomes all of these disadvantages. The membrane physically separates the two immiscible phases. Thus, independent variation of the two-phase flow rates over a large range of flow rates is possible without regard to flooding or loading limitations. The hollow fiber membrane immobilizes the liquid–liquid interface. In the experiments described here, hydrophobic polypropylene hollow fibers are used. Thus the membrane pores are filled with the organic phase. Loss of organic phase is prevented by ensuring that the aqueous side pressure is higher than the organic side pressure. In addition, extraction is achieved without dispersion of one phase in the other, thus, no coalescence step is required. Schlosser et al. [8] have reviewed the use of membrane-based solvent extraction for recovery and separation of organic acids. The theory of membrane-based solvent extraction is well developed [9]. Transfer of the solute species from the feed to the solvent phase through the membrane is described by an overall mass transfer coefficient. The overall mass transfer coefficient depends on three individual mass transfer coefficients which describe mass transfer across the concentration boundary layer in each of the phases and through the membrane pores. Numerous correlations are available in the literature for

Chapter

10

215

Membrane Extraction for Biofuel Production

predicting the mass transfer coefficient inside the fibers, outside the fibers, and in the membrane pores [10]. Here, we present three applications for membrane-based solvent extraction in the manufacture of biofuels. Table 10.1 summarizes these applications. The first application focuses on the production of bioethanol by fermentation of lignocellulosic biomass. The second application explores the feasibility of extraction of 5-hydroxymethylfurfural (HMF) from an aqueous solution into methyl isobutyl ketone (MIBK). While the first two examples focus on extraction of a target compound into an organic phase, the third application involves the extraction of glycerol from 2-butanol into an aqueous phase. Efficient removal of glycerol during the manufacture of biodiesel is essential for the development of cost-effective technologies for the production of biodiesel. The potential role for membrane extraction in each of these examples is described below in more detail.

Removal of Acetic Acid from Biomass Hydrolysates Lignocellulosic biomass consists of three main polymers: cellulose, hemicellulose, and lignin. There are three main routes for conversion of lignocellulosic biomass into liquid fuels: gasification, pyrolysis or liquefaction, and hydrolysis [1]. TABLE 10.1 Summary of Membrane-Based Solvent Extraction Experiments Conducted Comment Pretreatment Conditions Feed

Target Compound

Pressure (atm)

Holding Time (min)

Initial pH

Sugarcane bagasse

Acetic acid

1

10

1.15

Sorghum

1

10

1.3

Oat

1

10

1.37

Coffee husk

1

30

1.32

Corn fiber

0.5

10

0.84

Corn leaves

1

10

0.6

Organic phase 85:15 octanol:alamine 336 (weight ratio) HMF in water or water/ DMSO mixture

HMF

Organic phase MIBK

Glycerol in 2-butanol

Glycerol

Glycerol extracted from 2-butanol into water

Pretreatment was conducted at 120  C.

216

Inorganic, Polymeric and Composite Membranes

Here, we focus on hydrolysis. Hydrolysis involves the combination of thermochemical and biochemical processing steps to convert the cellulose into glucose and hemicellulose into five carbon sugars (mainly xylose) [11]. Dilute sulfuric acid is used to hydrolyze the lignocellulosic biomass, as it has been shown to effectively hydrolyze the hemicellulose and increase the enzymatic digestibility of the cellulose [12]. All of the different lignocellulosic biomass samples were provided by the School of Engineering of Lorena (EEL), University of Sa˜o Paulo, Lorena, SP, Brazil. Dilute sulfuric acid was used to pretreat the lignocellulosic biomass. During pretreatment, compounds are produced that are toxic to the microorganism used to ferment the sugars to the desired products (ethanol). Here, we focus on removal of acetic acid which is produced as a result of hydrolysis of acetyl groups present in the hemicellulose. In its protonated form, acetic acid can diffuse through the cytoplasmic membrane of cells affecting the cell metabolism [13,14]. It can further diffuse through the cell cytoplasm where it lowers the intracellular pH, resulting in impaired transport of various ions and increased energy requirements and reduced ethanol yields [15]. This is particularly important as the viability of a cellulosic ethanol plant will depend on maximizing ethanol yields. In our earlier work, we have explored the feasibility of using membranebased solvent extraction for removal of acetic acid from corn stover hydrolysates [16,17]. Here, we present results for extraction of acetic acid from a number of different hydrolysates readily available in Brazil. As described in our earlier work, the organic phase consists of Alamine 336, a long chain water insoluble tri-octyl/decyl amine dissolved in octanol. Long chain aliphatic amines may be used for the extraction of carboxylic acids from dilute aqueous solutions [18–27]. Table 10.1 summarizes the various lignocellulosic biomass hydrolysates we have investigated. In addition, pretreatment conditions are listed. These pretreatment conditions were developed by EEL and were found to produce a xylose-rich hydrolysate. There is a considerable amount of literature available on the extraction of acetic acid in the presence of sulfate and bisulfate anions using aliphatic tertiary amines [16]. Briefly, for effective extraction of acetic acid, the aliphatic amine (Alamine 336 used here) must be insoluble in the aqueous phase. Further, for the hydrolysates listed in Table 10.1, after pretreatment, the pH will be between 0.6 and 1.4. Thus the sulfuric acid will be present almost entirely as HSO4 and SO42. Eyal and Canri [27] describe four main mechanisms for acid extraction by amine-based extractants. In the case of sulfuric acid, a strong mineral acid, ion pair formation occurs where the amine binds a proton to form an ammonium cation. As shown in our earlier work, this mechanism leads to significant extraction of sulfuric acid according to the following mechanism: R3 N (organicÞ þ Hþ(aqÞ ! R3 NHþ(organicÞ

Chapter

10

Membrane Extraction for Biofuel Production

217

þ  R3 NHþ ðorganicÞ þ HSO 4ðaqÞ ! R3 NH HSO4ðorganicÞ

Alamine 336 will extract the protonated form of acetic acid. Barrow and Yerger [28] indicate that the acetic acid molecule reacts with the amine to form an ion pair. As we have shown in our earlier work [16], sulfuric acid will be extracted preferentially. This leads to a rapid increase in pH until the pH reaches a value of 2, the pKa value for the second dissociation constant for sulfuric acid. After the hydrolysate pH goes above 2, the rate of extraction of acetic acid increases. As the pH approaches 4.75, the pKa of acetic acid, the rate of extraction of acetic acid rapidly decreases. Here, we extend our previous results obtained for corn stover-based hydrolysates to the hydrolysates listed in Table 10.1. We show that membrane extraction is a highly flexible unit operation that may be used to remove acetic acid from a variety of lignocellulosic biomass hydrolysates. Further, the rate of extraction may be correlated using mass transfer coefficients.

Extraction of 5-Hydroxymethylfurfural The previous example focused on fermentation of lignocellulosic biomass hydrolysates to bioethanol. Detoxification (conditioning) of the hydrolysate was required to remove compounds that are toxic to the microorganism. Alternatively, aqueous phase processing could be used to chemically convert the aqueous sugar-rich hydrolysate into potential transportation fuels such as dimethylfuran (DMF). Aqueous phase processing, pioneered by Dumesic and co-workers, involves the catalytic conversion of sugars, sugar alcohols, and polyols into hydrogen or alkanes [29–37]. Huber et al. [1] describe a self-sustaining biomass biorefinery for conversion of biomass into liquid alkanes using aqueous phase processing. Acid dehydration of six carbon sugars can lead to the production of HMF. HMF is a critical and versatile intermediate in the conversion of biomass to liquid biofuels such as DMF, liquid alkanes, and many other value-added products. Removal of HMF from the aqueous reaction media will greatly enhance HMF yield by driving the forward reaction and preventing HMF degradation. We have explored the feasibility of HMF removal by membrane-based solvent extraction. Extraction of HMF is complicated by the fact that HMF is highly hydrophilic. An ideal extractant should have low water solubility. In addition, the extractant should be polar to efficiently extract the highly polar HMF. We have used MIBK as the organic extractant. Previous investigators have used MIBK as the organic phase in a biphasic reactor designed to produce HMF from fructose [38–40]. Aqueous phases consisting of water and a water DMSO (dimethyl sulfoxide) mixture have been used. Use of a phase modifier such as DMSO has been shown to improve reaction yields [38]. The results again highlight the versatility of membrane-based solvent extraction. Our results may be correlated using mass transfer coefficients.

218

Inorganic, Polymeric and Composite Membranes

Glycerol Extraction Biodiesel is a “green fuel” that has several advantages over conventional petroleum-based diesel. Engines using biodiesel will have a significant decrease in carbon dioxide and many other green house gas emissions. Current commercial biodiesel production uses base catalyzed transesterification to convert triglycerides to biodiesel. Each triglyceride molecule reacts with three molecules of alcohol (commonly methanol) to produce three esters and one glycerol molecule. NaOH and KOH catalysts are commonly used. Removal of glycerol from the reaction mixture is essential. Base-catalyzed transesterification reactions have little tolerance for free fatty acids (FFA) and water. FFAs will cause soap formation and generate water. In turn, water will cause ester hydrolysis to produce more FFAs. Consequently, product recovery can be difficult and costly. While acid catalysis is slower than base catalysis, the fact that acid catalysts can catalyze esterification of FFAs and do not lead to saponification reactions is a major advantage [41]. Further use of higher alcohols such as butanol rather than methanol is an advantage as butanol may be produced via biomass fermentation leading to a greener process. Additionally, it is hydrophobic and has a higher boiling point allowing the transesterification reaction to be conduced at higher temperature. Development of alternative transesterification schemes will require separation of glycerol from a reaction medium containing different components. Here, we have investigated glycerol extraction from an organic phase consisting of butanol as would be the case if butanol were the alcohol used for the transesterification reaction. Unlike the previous examples, membrane-based solvent extraction is used to extract glycerol from an organic phase into an aqueous phase. However, the same mass transfer correlations apply.

MATERIAL AND METHODS The membrane extraction setup used for all the experiments is given in Figure 10.2. Two Liqui-Cel membrane contactors MiniModuleÒ 1
5.5 and MiniModuleÒ 1.7
5.5 (Membrana, Charlotte, USA) were used for extraction of acetic acid and glycerol, respectively. The 1
5.5 and 1.7
5.5 modules contain 0.18 and 0.58 m2 of active membrane surface area. HMF extraction was conducted using a Liqui-CelÒ Extra-Flow 2.5
8 module. This module contains 1.4 m2 of membrane surface as well as a central baffle to promote mixing of the shell side fluid. All three modules contained polypropylene fibers; 300 mm OD, 220 mm ID, porosity 40%, pore size 0.04 mm. The larger ExtraFlow 2.5
8 module was compatible with MIBK which was used as the organic phase for extraction of HMF. Flexible chemical resistant Masterflex, precision silicone 6410-18 and tygon 2075 tubing (Cole-Parmer, Vernon, IL), was used to connect a Watson

Chapter

10

219

Membrane Extraction for Biofuel Production

P pH probe Hollow fiber module Aqueous phase

Organic phase

P P

P

Pump 1

Pump 2

FIGURE 10.2 Schematic representation of experimental setup.

Marlow 505 U and 505 S peristaltic pump (Cole-Parmer, Vernon, IL) to the MiniModule 1
1.5. For the Minimodule 1.7
5.5 and the Extra-flow 2.5
8 module, a Masterflex 77601-10 pump head equipped with a 20-650 RPM drive (Nr 7591-50, Cole-Parmer) and a Watson Marlow 503 U peristaltic pump were used with Masterflex 96410-73 precision silicone tubing (Cole-Parmer). Pressure gauges, 0–100 kPa (McMaster-Carr, Atlanta, GA), were used to monitor the aqueous and organic phase pressures. All experiments were conducted at 25  C. The system was started by first turning on the aqueous and then the organic phase pumps. The aqueous phase was pumped through the fiber lumen, while the organic phase was pumped on the shell side. Table 10.2 gives the various aqueous and organic phase flow rates investigated. In all experiments, the aqueous phase pressure was above the organic phase pressure at any given point in the module to prevent passage of the organic phase into the aqueous phase. For extraction of acetic acid, pH was monitored using a Thermo Orion 520 pH meter (Thermo Fisher Scientific, Waltham, MA) equipped with a Metler Toledo pH probe (Cole-Parmer). Samples were taken from the aqueous phase at frequent intervals for high pressure liquid chromatography (HPLC) analysis (extraction of acetic acid and HMF) or refractive index measurement (extraction of glycerol) as described below.

Removal of Acetic Acid from Biomass Hydrolysates The various biomass samples listed in Table 10.1 were air dried and milled to a mesh size of 10. Water was added such that the solid:liquid ratio was 1:10. Hydrolysis of the biomass was conducted using dilute sulfuric acid at a ratio of 100 mg of acid to 1 g of biomass in a 200 mL stainless steel container. Holding times are given in Table 10.1. After reaction, the remaining solids were removed by filtration using 0.45 mm pore size filter paper (Millipore, Bedford, MA).

220

Inorganic, Polymeric and Composite Membranes

TABLE 10.2 Summary of Experimental Conditions Used for Extraction of Acetic Acid, HMF, and Glycerol

Feed

Aqueous Flow Rate (L h 1)

Organic Flow Rate (L h 1)

Initial Target Compound Concentration (g L 1)

Sugar bagasse

48

23

2.93

Sorghum

3.61

Oat

5.94

Coffee husk

0.45

Corn fiber

2.74

Corn leaves

1.21

HMF in water and water/DMSO mixture

Glycerol in 2-butanol

Comments Hydrolysate filtered using 0.45 mm filter

Water:DMSO (mole ratio) 32

32

4.6

1:0

64

16

4.4

1:0

48

24

4.5

1:0

32

16

4.2

15:1

10.8

24.5

2

27.7

24.5

2

2-butanol was saturated with water

12.7

49

2

27.7

49

2

Extraction of acetic acid from different biomass hydrolysates was conducted using 333 g of organic phase consisting of 85% octanol (Sigma-Aldrich Corporation, St. Louis, MO) and 15% Alamine 336 (w/w) (Cognis, Cincinnati, OH). The aqueous phase consisted of 500 g of hydrolysate. The aqueous and organic phase flow rates were 48 and 23 L h 1, respectively. The aqueous side inlet and outlet pressures were maintained at 21 and 3.5 kPa, while the organic side inlet and outlet pressures were maintained at 14 and 2 kPa. During extraction, 2 mL samples of the aqueous phase were removed at frequent intervals for analysis (2424 HPLC system equipped with a refractive index detector, Waters Corporation, Milford, MA). An Aminex HPX-87 H column (Biorad Hercules, CA) was used, with a mobile phase consisting of 0.01 N

Chapter

10

Membrane Extraction for Biofuel Production

221

sulfuric acid at a flow rate of 0.6 mL min 1. The column temperature was set to 45  C with an injection volume of 20 mL.

HMF Extraction HMF was obtained from SAFCÒ (St. Louis, MO). The initial feed concentration was  5 g/L. The aqueous phase (500 mL) comprised either DI water or 15:1 DI water:DMSO (molar ratio) mixture. Dimethyl sulfoxide (DMSO) was purchased from Thermo Fisher Scientific. The organic phase consisted of 500 mL of MIBK (Sigma-Aldrich). Aqueous and organic phase flow rates from 16 to 64 L h 1 were investigated (see Table 10.2). The HMF concentration in the aqueous phase was measured using HP 1050 HPLC equipped with a refractive index detector HP 1047A (Agilent Technologies, Santa Clara, CA). The column temperature was set to 55  C with an injection volume of 6 mL.

Glycerol Extraction Glycerol (Mallinckrodt Baker, NJ) was extracted from 2-butanol (SigmaAldrich). The organic phase consisted of 86.2 g glycerol dissolved in 340.8 g 2-butanol. Organic phase flow rates of 24.5 and 49 L h 1 were investigated. The inlet and outlet organic phase pressures were controlled at 14 and 3.5 kPa. The aqueous phase consisted of 275.0 g of 2-butanol saturated DI water. Aqueous flow rates between 10 and 28 L h 1 were investigated. The aqueous phase inlet and outlet pressures were set at 21 and 9.5 kPa, respectively. Samples (1 mL) were taken at frequent intervals from the aqueous phase for glycerol analysis using a refractive index meter (Bausch & Lomb, Rochester, NY) at 20  C.

RESULTS Acetic acid extraction from various biomass hydrolysates is shown in Figures 10.3 and 10.4. Filled symbols linked by a line represent the change in acetic acid concentration in the hydrolysate and are read using the left hand side y-axis. To compare results for the three different extractions considered here (ethanol, HMF, and glycerol), the results are normalized by dividing the measured acetic acid concentration in the aqueous phase (g L 1) by the molecular weight of acetic acid (60.05 g mol 1) and the membrane surface area (0.18 m2) and multiplying by the organic phase flow rate (23 L h 1). Thus the change in acetic acid concentration is given in terms of an acetic acid flux. Open unconnected symbols give the variation of hydrolysate pH with time and are read using the right hand side y-axis. Figures 10.3 and 10.4 indicate that the initial acetic acid concentration varies considerably between the different hydrolysates. The initial acetic acid concentration depends on the lignocellulosic biomass as well as the severity (temperature, sulfuric acid concentration,

222

Inorganic, Polymeric and Composite Membranes

Sorghum Sugarcane bagasse pH Sorghum pH Sugarcane bagasse

0.004

5

4 0.003 0.0025

3 pH

Acetic acid (mol m-2 s-1)

0.0035

Oat Coffee husk pH Oat pH Coffee husk

0.002 2

0.0015 0.001

1 0.0005 0

0 0

50

100

150

200 250 300 Time (min)

350

400

450

FIGURE 10.3 Acetic acid extraction from sorghum, oat, coffee husk. and sugar bagasse hydrolysates using 15% Alamine 336 in octanol as the organic phase. Aqueous and organic phase flow rates were 48 and 23 L h 1.

5

0.004

0.003 0.0025

Corn leaf

pH Corn fiber

pH Corn leaf

4

3 pH

Acetic acid (mol m-2 s-1)

0.0035

Corn fiber

0.002 2

0.0015 0.001

1 0.0005 0 0

50

0 100 150 200 250 300 350 400 450 500 550 600 Time (min)

FIGURE 10.4 Acetic acid extraction from corn fiber and corn leaf hydrolysates 15% Alamine 336 in octanol as the organic phase. Aqueous and organic phase flow rates were 48 and 23 L h 1.

and time) of the pretreatment [11,42]. Consequently, the development of a hydrolysate detoxification process depends on the lignocellulosic biomass and the hydrolysis conditions. All six hydrolysates indicate an increase in pH during extraction. As described in our earlier work [16,17] for corn stover

Chapter

10

223

Membrane Extraction for Biofuel Production

hydrolysates, sulfuric acid will be preferentially extracted over acetic acid which results in an increase in hydrolysate pH. A mass balance around the aqueous feed reservoir for acetic acid gives: V

dC ¼ KA (C  C Þ dt

ð10:1Þ

where V is the volume of hydrolysate (500 mL), K is the overall mass transfer coefficient based on the aqueous phase, A is the membrane surface area (0.18 m2), C is the acetic acid concentration in the hydrolysate and C* is the acetic acid concentration in the hydrolysate that would be in equilibrium with the concentration in the organic phase. Since fresh organic phase was used for each experiment, initially C*is zero. We also assume that the aqueous and organic phase reservoirs are fully mixed, and the rate of change of acetic acid concentration in the hydrolysate per pass through the module is small. Integration of Equation (10.1) leads to:   C KAt ð10:2Þ ¼ Ln C0 V Plotting the left-hand side of Equation (10.2) against time should lead to a straight line, the slope of which is proportional to the overall mass transfer coefficient as given in Figure 10.5. For short extraction times, the results for all six hydrolysates fall on the same straight line. However, for longer run times, significant deviations from this straight line are observed. For hydrolysates containing higher initial concentrations of acetic acid such as oat and sorghum, assuming the acetic acid concentration in the organic phase is zero will not be valid at longer run times.

Ln (Co/C)

1

Sorghum

Oat

Sugarcane bagasse

Coffee husk

Corn fiber

Corn leaf

0 0

50

100 150 200 250 300 350 400 450 500 550 600 Time (min)

FIGURE 10.5 Determination of overall mass transfer coefficient for removal of acetic acid from biomass hydrolysates.

224

Inorganic, Polymeric and Composite Membranes

The decrease in the overall mass transfer coefficient at longer run times may also be due to fouling of the membrane by particulate matter in the hydrolysate. In this work, the hydrolysate was prefiltered using a 0.45 mm filter prior to membrane extraction. Our results indicate that inclusion of a membrane detoxification step will lead to modifications in the unit operations before and after the membrane step and the effect of these changes must be accounted for when determining the economic viability of membrane detoxification. The overall mass transfer coefficient based on the aqueous phase is made up of three individual mass transfer coefficients: 1 1 m m ¼ þ þ K ka km ko

ð10:3Þ

where ka, km, and ko are the aqueous, membrane, and organic phase mass transfer coefficients, respectively, and m is the distribution coefficient of acetic acid between the phases defined as the acetic acid concentration in the aqueous phase divided by the concentration in the organic phase. The aqueous and organic phase mass transfer coefficients depend on the aqueous and organic phase flow rates [43], while the membrane mass transfer coefficient does not depend on the phase flow rates. The membrane mass transfer coefficient is given by: km ¼

De tl

ð10:4Þ

where D is the diffusion coefficient of acetic acid in the membrane pores, e is the membrane porosity (40%), t is the membrane tortuosity, and l is the wall thickness of the hollow fibers (40 mm). Membrane tortuosity factors ranging from 3 to 12 have been reported [44]. Here, we use a value of 3 as is commonly used for polypropylene membranes [45]. In our earlier work [17], we have shown that, for extraction of acetic acid, the membrane mass transfer coefficient controls the rate of acetic acid extraction. Thus the overall mass transfer coefficient is independent of the aqueous and organic phase flow rates over a large range of flow rates. Consequently, the overall mass transfer coefficient may be approximated by km/m (Equation (10.3)). The experimentally determined value of the overall mass transfer coefficient from Figure 10.5 is given in Table 10.3. The membrane mass transfer coefficient may be calculated using Equation (10.4). The diffusion coefficient of acetic acid is estimated using the Wilke–Chang equation [46,47] D¼

7:4
108 (fM2 Þ0:5 T mV 0:6

ð10:5Þ

where f is the association parameter assumed to be 1.5 for alcohols, M2 is the molecular weight of octanol (130 g mol 1), T is the temperature (295 K), m is

Chapter

10

225

Membrane Extraction for Biofuel Production

TABLE 10.3 Calculated and Experimentally Determined Mass Transfer Coefficients

Compound

KA/V (s 1)

Experimental Mass Transfer Coefficient (m s 1)

Calculated Mass Transfer Coefficient (m s 1)

Acetic acid

0.003

1.39E-07

1.49E-06

HMF

0.1415

8.42E-07

7.08E-06

Glycerol

0.0526

4.16E-07

5.88E-07

the viscosity of the solvent, and V is the molar volume of the solute. The presence of Alamine 336 is ignored and the viscosity of octanol at 25  C is used (6.09
10 3 Pa s). The molar volume of acetic acid is calculated from its density and molecular weight (see Table 10.4) [48]. The calculated value of the overall mass transfer coefficient is also given in Table 10.3. The experimentally determined mass transfer coefficient is lower than the calculated mass transfer coefficient. There are, however, a number of simplifications that have been made in estimating the mass transfer coefficient. Errors in diffusion coefficients predicted by the Wilke–Chang Equation are much higher for nonaqueous solutions [47]. Further, the viscosity of octanol was assumed though the organic phase contains 15% Alamine 336 which has a much higher viscosity. The experimentally determined mass transfer coefficient is based on the aqueous phase concentrations. However, the membrane pores are filled with the organic phase. Thus, a distribution coefficient for acetic acid between the two phases should be included in the calculated overall mass transfer coefficient [51]. Estimation of the distribution coefficient for acetic acid is complicated, as the concentration of the dissociated species in the aqueous phase as well as acid bound to amine in the organic phase must be included, resulting in the distribution coefficient being a function of pH. Consequently, the effect of the distribution coefficient on the membrane mass transfer coefficient has been ignored, and the membrane mass transfer coefficient is used to approximate the overall mass transfer coefficient. In addition, fouling of the membrane by particulate matter that passed through the 0.45 mm filter used to prefilter the hydrolysate could explain the lower experimentally determined mass transfer coefficient compared to the calculated mass transfer coefficient. In earlier work [17] for corn stover-based hydrolysates, the experimental and calculated mass transfer coefficients were within 60% of each other. In this work, the hydrolysate was prefiltered using a 0.22-mm filter. It is also worth noting that mass transfer coefficients are typically accurate to within 40% [47]. Thus, Table 10.3 indicates that the mass transfer coefficient for acetic acid may be estimated using Equations (10.3) and (10.4).

TABLE 10.4 Values of Parameters Used to Calculate Diffusion Coefficients

Compound

Molecular Weight of Solute (g mol 1)

Molecular Weight of Solvent

Density of Solute (g cm 3)

Molar Volume of Solute (cm3 mol 1)

Association Parameter

Viscosity of Solvent (mPa s)

D (cm2 s 1)

Acetic acid

60.1

130.2

1.0

57.2

1.5

6.1

4.50E-06

HMF

126.1

74.1

1.2

104.6

1.0

0.5

2.50E-05

Glycerol

92.1

100.2

1.3

73.0

1.5

3.0

5.90E-06

Physical data from Refs. [48–50].

Chapter

10

227

Membrane Extraction for Biofuel Production

Figure 10.6 gives results for extraction of HMF from water into MIBK. Analogous to Figures 10.3 and 10.4, to compare results for HMF extraction with extraction of acetic acid and glycerol, the results are normalized by dividing the measured HMF concentration in the aqueous phase (g L 1) by the molecular weight of HMF (126.11 g mol 1) and the membrane area (1.4 m2) and multiplying by the organic phase flow rate. If the rate of transfer of HMF across the membrane is independent of the aqueous and organic phase flow rates, since the measured HMF concentration in the aqueous phase is being multiplied by the organic phase flow rate to normalize the results, the HMF flux should be directly proportional to the organic phase flow rate for the same extraction time. Figure 10.6 indicates that this is in fact the case. In addition, the rate of HMF extraction from either water or a 15:1 mol ratio water:DMSO solution is the same. Thus the presence of DMSO in the feed has no effect on the rate of HMF extraction. Consequently, as was the case for acetic acid extraction from biomass hydrolysates, the membrane mass transfer coefficient dominates and we approximate the overall mass transfer coefficient by km/m (Equation (10.3)). Thus, Equations (10.1)–(10.4) apply for the extraction of HMF from water and water DMSO solutions. Figure 10.7 is a plot of the left hand side of Equation (10.2) against time. The results fall on approximately the same curve. As was the case for extraction of acetic acid, the overall mass transfer coefficient may be determined by fitting a straight line to the initial data. Table 10.3 gives the experimentally determined overall mass transfer coefficient for extraction of HMF. Equation (10.4) may be used to calculate the overall mass transfer coefficient. Table 10.4 gives values for the molecular weight and viscosity of MIBK and the density, molecular weight, and molar volume of HMF [48,49].

HMF (mol m-2 s-1)

0.0003 Water:DMSO 1:0 (32/32)

Water:DMSO 1:0 (64/16)

Water:DMSO 1:0 (48/24)

Water:DMSO 15:1 (32/16)

0.0002

0.0001

0 0

5

10 Time (min)

15

20

FIGURE 10.6 HMF extraction results from water and water/DMSO mixture (mol fraction) at various (aqueous/organic) phase flow rates in liters per hour.

228

Inorganic, Polymeric and Composite Membranes

Ln (Co/C)

1

Water:DMSO 1:0 (32/32)

Water:DMSO 1:0 (64/16)

Water:DMSO 1:0 (48/24)

Water:DMSO 15:1 (32/16)

0 0

5

10 Time (min)

15

20

FIGURE 10.7 Determination of overall mass transfer coefficient for removal of HMF from water and water/DMSO mixtures (mol fraction) at various (aqueous/organic) flow rates in liters per hour.

The association parameter f was taken to be 1.0 as is usually assumed for organic compounds [47]. The distribution coefficient, m, for HMF was experimentally determined to be 1.17 (concentration of HMF in the water and water/ DMSO mixture divided by concentration in MIBK). Reasonable agreement is obtained between the predicted and experimentally determined mass transfer coefficients given the numerous assumptions that have been made, as discussed above. Results for glycerol extraction from butanol into water are given in Figure 10.8. The measured glycerol concentration in water is divided by the molecular weight of glycerol (92.1 g mol 1) and the membrane surface area (0.58 m2). However, unlike extraction of acetic acid and HMF, the aqueous extractant phase flow rate is used to normalize the measured glycerol concentration. Analogous to extraction of HMF, if the transfer of glycerol across the membrane is independent of the aqueous and organic phase flow rates, since the measured glycerol concentration in the aqueous phase is being multiplied by the aqueous phase flow rate to normalize the results, the glycerol flux should be directly proportional to the aqueous phase flow rate for the same extraction time. Figure 10.8 indicates that this is in fact the case. As can be seen for a range of aqueous (10–28 L h 1) and organic (24–49 L h 1) phase flow rates tested, no effect is observed on the rate of mass transfer. Thus, again the overall mass transfer coefficient may be approximated by km/m (Equation (10.3)). Since glycerol is being extracted from the organic phase into the aqueous phase, a mass balance around the aqueous feed reservoir for glycerol leads to a slightly modified form of Equation (10.1) V

dC ¼ KA (C  CÞ dt

ð10:6Þ

Chapter

10

229

Membrane Extraction for Biofuel Production

Glycerol (mol m−2 s−1)

0.04

0.03

0.02

0.01 Water:Butanol (10.8/24.5)

Water:Butanol (27.7/24.5)

Water:Butanol(12.7/49.0)

Water:Butanol (27.7/49.0)

0 0

20

40

60 Time (min)

80

100

120

FIGURE 10.8 Glycerol extraction results from 2-butanol into water at various (aqueous/organic) flow rates in liters per hour.

where V is the volume of water, (275 mL), K is the overall mass transfer coefficient based on the aqueous phase, A is the membrane surface area (0.58 m2), C is the glycerol concentration in aqueous phase, and C* is the glycerol concentration in the aqueous phase that would be in equilibrium with the concentration in the organic phase. Integration of Equation (10.6) leads to   C KAt ¼ ð10:7Þ Ln 1  C V Plotting Ln((C  C*)/C*) against time should lead to a straight line, the slope of which is proportional to the overall mass transfer coefficient as given in Figure 10.9. The experimentally determined overall mass transfer coefficient is given in Table 10.3. Equation (10.4) is again used to calculate the membrane mass transfer coefficient. Table 10.4 gives values for the molecular weight and viscosity of butanol and the density, molecular weight, and molar volume of glycerol [50,48]. The association parameter f was taken to be 1.5 as is usually assumed for alcohols [46]. The distribution coefficient, m, for glycerol was experimentally determined to be 3.34 (concentration of glycerol in the water divided by concentration in 2-butonol). As can be seen, good agreement is obtained between the predicted and experimentally mass transfer coefficients. Taken together, our results highlight the versatility of membrane extraction as a unit operation in future biorefineries. The overall mass transfer coefficient may be predicted using Equations (10.3) and (10.4). These mass transfer coefficients will be important when estimating the rate of removal of the target compound, in scaling up the process and when correlating the experimental data. This in turn will enable easy design and scale up of larger extraction systems.

230

Inorganic, Polymeric and Composite Membranes

10

Ln (C - C*/C*)

8

10.8/24.5 L/h (O/W)

27.7/24.5 L/h (O/W)

12.7/49.0 L/h (O/W)

27.7/49.0 L/h (O/W)

6 4 2 0 0

20

40

60 Time (min)

80

100

120

FIGURE 10.9 Determination of overall mass transfer coefficient for removal of glycerol from 2-butanol at various (aqueous/organic) flow rates in liters per hour.

When using membrane-based solvent extraction to detoxify biomass hydrolysates, it is essential to minimize losses of the organic phase in the hydrolysate, as the organic phase is toxic to the microorganism used for subsequent fermentation. The fact that membrane-based solvent extraction is dispersion-free is therefore a significant benefit [17]. The actual economic viability of an extraction process, however, depends on the actual extraction being conducted and integration of membrane-based solvent extraction into the overall manufacturing process.

CONCLUSIONS Extraction of acetic acid from aqueous biomass hydrolysates into an organic phase consisting of a mixture of octanol and Alamine 336 has been investigated. Removal of acetic acid from biomass hydrolysates is important to maximize ethanol yields, as it is toxic to the microorganisms used to convert the sugars present in the hydrolysate to ethanol. Extraction of HMF from an aqueous phase into MIBK has been investigated. HMF is a very valuable intermediate during the thermochemical conversion of sugars present in biomass hydrolysates into transportation fuels. Extraction of glycerol from an organic phase consisting of 2-butanol into water has also been investigated. This extraction could be of significance in the production of biodiesel. For each of the extractions studied in this work, the overall and calculated mass transfer coefficients are in good agreement. These mass transfer coefficients are important when estimating the rate of extraction which will be essential in designing a larger-scale process. They will also be important when correlating the experimental data. Our results highlight the versatility and tremendous possibilities for dispersion-free membrane-based solvent extraction as a unit operation in future biorefineries.

Chapter

10

Membrane Extraction for Biofuel Production

231

ACKNOWLEDGMENTS Funding for this work was provided by the National Science Foundation (CBET 045683) and Colorado State University through the Sustainable Bioenergy Development Center (SG-0903) and the Clean Energy Supercluster. Funding in Brazil was obtained from FAPESP (2008/57926-4) and CNPq (490351/2007-7). In addition, the authors would like to thank Mr. Michael Becker and Mr. Flavio Ferraz for their assistance.

REFERENCES [1] G.W. Huber, S. Iborra, A. Corma, Synthesis of transportation fuels from biomass: chemistry, catalysis and engineering, Chem. Rev. 106 (9) (2006) 4044–4098. [2] United State Department of Agriculture, Biomass as feedstock for a biorefinery and bioproducts industry: the technical feasibility of a billion-ton annual supply, April 2005. [3] A.J. Ragauskas, C.K. Williams, B.H. Davison, G. Britovsek, J. Cairney, C.A. Eckert, et al., The path forward for biofuels and biomaterials, Science 311 (5760) (2006) 484–489. [4] C. Somerville, The billion-ton biofuels vision, Science 312 (5778) (2006) 1277. [5] J.D. Seader, E.J. Henley, Separation Process Principles, John Wiley & Sons, Hoboken, NJ, 2006. [6] W.L. McCabe, J.C. Smith, P. Harriott, Unit Operations of Chemical Engineering, fifth ed., McGraw Hill, Inc., New York, NY, 1993. [7] K.K. Sirkar, Membrane contactors, in: N.N. Li, A.G. Fane, W.S.W. Ho, T. Matsuura (Eds.), Advanced Membrane Technology, John Wiley & Sons, Hoboken, NJ, 2008. [8] Sˇ. Schlosser, R. Kerte´sz, J. Marta´k, Review: recovery and separation of organic acids by membrane based solvent extraction and pertraction—an overview with a case study on recovery of MPCA, Sep. Purif. Technol. 41 (2005) 237–266. [9] R. Prasad, K.K. Sirkar, Membrane-based solvent extraction, in: W.S.W. Ho, K.K. Sirkar (Eds.), Membrane Handbook, van Nostrand Reinhold, New York, NY, 1992. [10] S.R. Wickramasinghe, M.J. Semmens, E.L. Cussler, Mass transfer in various hollow fiber geometries, J. Membr. Sci. 69 (1992) 235–250. [11] S.E. Bower, S.R. Wickramasinghe, N.J. Nagel, D.J. Schell, Modeling sucrose hydrolysis in dilute sulfuric acid solutions at pretreatment conditions for lignocellulosic biomass, Bioresour. Technol. 99 (2008) 7354–7362. [12] J.C. Parajo, H. Dominguez, J.M. Dominguez, Biotechnological production of xylitol. Part 3. Operation in culture media made from lignocellulosic hydrolysates, Bioresour. Technol. 66 (1) (1998) 25–40. [13] D. Nabarlatz, X. Farriol, D. Montane´, Kinetic modeling of the autohydrolysis of lignocellulosic biomass for the production of hemicellulose-derived oligosaccharides, Ind. Eng. Chem. 43 (2004) 4124–4131. [14] E.M. Lohmeirer-Vogel, C.R. Sopher, H. Lee, Intracellular acidification as a mechanism for the inhibition by acid hydrolysis-derived inhibitors of xylose fermentation by yeasts, J Ind. Microbiol. Biotechnol. 20 (1998) 75–81. [15] E. Palmqvist, B. Hahn-Hagerdal, Fermentation of lignocellulosic hydrolysates. I: inhibition and detoxification, Bioresour. Technol. 74 (1) (2000) 17–24. [16] D.L. Grzenia, D.J. Schell, S.R. Wickramasinghe, Membrane extraction for removal of acetic acid from biomass hydrolysates, J. Memb. Sci. 322 (2008) 189–195. [17] D.L. Grzenia, D.J. Schell, S.R. Wickramasinghe, Detoxification of biomass hydrolysates by reactive membrane extraction, J. Memb. Sci. 348 (2010) 6–12.

232

Inorganic, Polymeric and Composite Membranes

[18] A.S. Kertes, J.C. King, Extraction chemistry of fermentation product carboxylic acids, Biotechnol. Bioeng. 28 (1986) 269–282. [19] J.A. Tamada, A.S. Kertes, C.J. King, Extraction of carboxylic acids with amine extractants. 1. Equilibria and law of mass action modeling, Ind. Eng. Chem. Res. 29 (1990) 1319–1326. [20] J.A. Tamada, J.C. King, Extraction of carboxylic acids with amine extractants. 2. Chemical interactions and interpretation of data, Ind. Eng. Chem. Res. 29 (1990) 1327–1333. [21] J.A. Tamada, J.C. King, Extraction of carboxylic acids with amine extractants. 3. Effect of temperature, water coextraction and process conditions, Ind. Eng. Chem. Res. 29 (1990) 1333–1338. [22] R.-S. Juang, W.-T. Huang, Kinetic studies on the extraction of citric acid from aqueous solutions with tri-n-octylamine, J. Chem. Eng. Jpn 28 (3) (1995) 274–281. [23] R. Canari, A.M. Eyal, Extraction of carboxylic acids by amine-based extractants: apparent extractant basicity according to the pH of half-neutralization, Ind. Eng. Chem. Res. 42 (2003) 1285–1292. [24] A. Senol, Effect of diluent on amine extraction of acetic acid: modeling considerations, Ind. Eng. Chem. Res. 43 (2004) 6496–6506. [25] X. Shan, W. Qin, Y. Dai, Dependence of extraction equilibrium of monocarboxylic acid from aqueous solutions on the relative basicity of extractant, Chem. Eng. Sci. 61 (2006) 2574–2581. [26] H. Reisinger, C.J. King, Extraction and sorption of acetic acid at pH above pKa to form calcium magnesium acetate, Ind. Eng. Chem. 34 (1995) 845–852. [27] A.M. Eyal, R. Canari, pH dependence of carboxylic and mineral acid extraction by amine based extractants: effects of pKa, amine basicity and diluent properties, Ind. Eng. Chem. Res. 34 (1995) 1789–1798. [28] G.M. Barrow, E.A. Yerger, Acid-base reactions in non-dissociating solvents. Acetic acid and triethylamine in carbon tetrachloride and chloroform, Am. Chem. Soc. 76 (20) (1954) 5211–5216. [29] G.W. Huber, J.N. Chheda, J.C. Barrett, J.A. Dumesic, Production of liquid alkanes by aqueousphase processing of biomass-derived carbohydrates, Science 308 (2005) 1446–1450. [30] G.W. Huber, J.A. Dumesic, An overview of aqueous-phase catalytic processes for production of hydrogen and alkanes in a biorefinery, Catal. Today 111 (2006) 119–132. [31] R.D. Cortright, R.R. Davda, J.A. Dumesic, Hydrogen from catalytic reforming of biomassderived hydrocarbons in liquid water, Nature 418 (2002) 964–967. [32] G.W. Huber, R.D. Cortright, J.A. Dimesic, Renewable alkenes by aqueous-phase reforming of biomass-derived oxygenates, Angew. Chem., Int. Ed. 43 (2004) 1549–1551. [33] J.W. Shabaker, R.R. Davda, G.W. Huber, R.D. Cortright, J.A. Dimesic, Aqueous-phase reforming of methanol and ethylene glycol over alumina-supported platinum catalysts, J. Catal. 215 (2003) 344–352. [34] R.R. Davda, J.W. Shabaker, G.W. Huber, R.D. Cortright, J.A. Dumesic, A review of catalytic issues and process conditions for renewable hydrogen and alkanes by aqueous-phase reforming of oxygenated hydrocarbons over supported metal catalysts, Appl. Catl., B 56 (2005) 171–186. [35] G.W. Huber, J.W. Shabaker, J.A. Dumesic, Raney Ni-Sn catalyst for H2 production from biomass-derived hydrocarbons, Science 300 (2003) 2075–2077. [36] R.R. Davda, J.W. Shabaker, G.W. Huber, R.D. Cortright, J.A. Dumesic, Aqueous-phase reforming of ethylene glycol on silica-supported metal catalysts, Appl. Catal., B 43 (1) (2003) 13–26. [37] J.W. Shabaker, J.A. Dumesic, Kinetics of aqueous-phase reforming of oxygenated hydrocarbons: Pt/Al(2)O(3)and Sn-modified Ni catalysts, Ind. Eng. Chem. Res. 43 (12) (2004) 3105–3112.

Chapter

10

Membrane Extraction for Biofuel Production

233

[38] Y. Roman-Leshkov, J.N. Chheda, J.A. Dumesic, Phase modifiers promote efficient production of hydroxymethylfurfural from fructose, Science 312 (2006) 1933–1937. [39] J.N. Chheda, J.A. Dumesic, An overview of dehydration, aldol-condensation and hydrogenation process for production of liquid alkanes from biomass-derived carbohydrates, Catal. Today 123 (2007) 59–70. [40] M. Kroger, U. Prusse, K.D. Vorlop, A new approach for the production of 2,5-furandicarboxylic acid by in situ oxidation of 5-hydroxymethylfurfural starting from fructose, Top. Catal. 13 (2000) 237–242. [41] A. Srivastava, R. Prasad, Triglycerides-based diesel fuels, Renewable Sustainable Energy Rev. 4 (2000) 111–133. [42] Q.A. Nguyen, M.P. Tucker, F.A. Keller, F.P. Eddy, Two-stage dilute acid pretreatment of softwoods, in: 21st Symposium on Biotechnology for Fuels and Chemicals, 1999. [43] C.F. Kenfield, R. Qin, M.J. Semmens, E.L. Cussler, Cyanide recovery across hollow fiber gas membranes, Environ. Sci. Technol. 22 (10) (1988) 1151–1155. [44] Q. Zhang, E.L. Cussler, Microporous hollow fibers for gas absorption II. Mass transfer across the membrane, J. Membr. Sci. 23 (3) (1985) 333–345. [45] Z. Shen, B. Han, S.R. Wickramasinghe, Cyanide removal from wastewater using gas membranes: pilot-scale study, Water Environ. Res. 76 (1) (2004) 15–22. [46] C.R. Wilke, P. Chang, Correlation of diffusion coefficients in dilute solutions, AIChE J. 1 (1955) 264–270. [47] R.H. Perry, C.H. Chilton, Chemical Engineering Handbook, fifth ed., McGraw Hill International Book Company, Tokyo, Japan, 1983. [48] S. Budavari (Ed.), The Merck Index, twelfth ed., Merck and Company, Whitehouse Station, NJ, 1996. [49] R. Riggio, H.E. Martinez, Excess volumes, viscosities, enthalpies, and Gibbs free energies for mixtures of methyl isobutyl ketone þ n-pentanol and methyl isobutyl ketone þ isoamyl alcohol at 298.15 L, Can. J. Chem. 64 (1986) 1595–1598. [50] A. Mariano, A. Camacho, M. Postigo, A. Valen, H. Artigas, F.M. Royo, et al., Viscosities and excess energy of activation for viscous flow for binary mixtures of tetrahydrofuran with 1-butanol, 2-butonal, and 1-chlorobutane at 283.15, 298.15 and 313.15 K, Braz. J. Chem. Eng. 17 (4–7) (2000), Sa˜o Paulo. [51] B. Han, Z. Shen, S.R. Wickramasinghe, Cyanide removal from industrial wastewaters using gas membranes, J. Membr. Sci. 257 (1–2) (2005) 171–181.

This page intentionally left blank

Chapter 11

A Review of Mixed Ionic and Electronic Conducting Ceramic Membranes as Oxygen Sources for High-Temperature Reactors Qiying Jiang1, Sedigheh Faraji2, David A. Slade1 and Susan M. Stagg-Williams1,* Chemical and Petroleum Engineering Department, University of Kansas, Lawrence, Kansas, USA Chemical Engineering Department, California State University, Long Beach, California, USA * Corresponding author: E-mail address: [email protected] 1 2

INTRODUCTION Mixed ionic–electronic conducting (MIEC) ceramic membranes have received substantial interest in recent decades for various applications requiring gas separations [1–3]. Dense (i.e., nonporous) oxygen-conducting MIEC ceramic membranes allow oxygen ion diffusion through the solid ceramic lattice, resulting in theoretically infinite selectivity for oxygen. Unlike the oxygenconducting ceramic membranes used in solid-oxide fuel cells (SOFCs) such as yttria-stabilized zirconia (YSZ), the ability of MIECs to conduct electron and oxygen ions simultaneously enables these membranes to operate without an external electrical circuit. An oxygen-MIEC ceramic membrane requires only different gas-phase environments on its two surfaces and a sufficiently high temperature to transport oxygen from the high oxygen content surface to the low oxygen content surface [1–4]. Under an oxygen potential gradient and above some threshold activity temperature, a fully densified oxygen-conducting membrane would exhibit perfectly selective oxygen production into the low oxygen environment [5]. At temperatures above their threshold activity temperature (typically greater than 600
C [6]), dense oxygen-conducting ceramics can be characterized in general as exhibiting (1) highly mobile lattice oxygen ions; (2) continuously variable oxygen content (i.e., oxygen nonstoichiometry) via the ability to support lattice oxygen defects and/or undergo gradual, dispersed phase changes; and (3) the surface abilities to dissociate molecular oxygen while incorporating oxygen ions and Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

235

236

Inorganic, Polymeric and Composite Membranes

to reassociate oxygen ions while evolving molecular oxygen [7]. Between their respective activity threshold temperatures and degradation temperatures, oxygen-MIEC ceramic materials exhibit oxygen fluxes that generally increase with both temperature and the magnitude of the imposed oxygen gradient. Oxygen-MIEC membranes have been explored for a variety of applications, including oxygen supply via air separation [2,8–11], energy production [4,12,13], SOFC anodes [14–20], and oxygen sources for high-temperature reactions. The high-temperature reactions of interest include partial oxidation of methane (POM) [21–26], other hydrogen production reactions [13,27–29], oxidative dehydrogenation of light alkanes [30–32], and oxidative coupling of methane [33–38]. Current commercial barriers for utilizing membranes in high-temperature hydrocarbon conversion reactors include low membrane oxygen flux, poor mechanical stability, and high membrane fabrication costs. The area of greatest interest in the research community to date, and of particular focus in this chapter, is the use of oxygen-MIEC membrane reactors for synthesis gas (syngas) production, particularly via POM. POM requires pure oxygen as a feedstock, and the high energy and safety-related costs associated with industrial-scale oxygen production provide a substantial incentive to develop oxygen-MIEC membranes as alternative oxygen sources for syngas production reactors [5,6,10,21,24,25,39–43]. The major challenge in developing membranes for this application involves the historically incompatible requirements of high oxygen transport and high material stability. These requirements have proven particularly difficult to reconcile in the strongly reducing environments of methane conversion reactors. After providing an overview of oxygen-conducting MIEC ceramic membrane materials, this chapter discusses recent work to overcome the unique challenges involved in incorporating these membranes into syngas production reactors. The scope includes the relationships between crystal structure, oxygen permeation, stability, and reactor performance.

GENERAL ATTRIBUTES OF OXYGEN-CONDUCTING MIEC CERAMIC MATERIALS Oxygen Nonstoichiometry The oxygen conduction phenomenon exhibited by oxygen-MIEC membranes is attributed to their ability to support oxygen vacancies and lattice disorder, which allows the relatively rapid and sustainable transport of oxygen ions and holes under the appropriate conditions [2–4,19,31]. The number of oxygen vacancies is a function of material composition, initial lattice structure, temperature, and ambient gas composition. Membrane material composition, temperature, and oxygen gradient are the dominant factors in determining oxygen flux [14,15,44–46]. The oxygen transport mechanism of oxygen-MIEC membranes is shown in Figure 11.1. First, an oxygen molecule from the gas phase must adsorb on the high oxygen potential surface of the ceramic, at which time it dissociates into

Chapter

11

237

Review of MIEC Ceramic Membranes

Reaction on high [O2-] surface O2 + 4e- ® 2O2Air

Reaction on low [O2-] surface 4e- + 2O2- ® O2 e-

Sweep gas

O2Oyxgen-depleted air

Sweep gas enriched with

FIGURE 11.1 O2 transport mechanism in a dense MIEC membrane.

two oxygen ions by receiving electrons from the membrane surface. Oxygen ions then migrate from the high to low oxygen potential surface via the oxygen vacancies in the lattice. In an inert environment, the oxygen ions lose electrons at the low oxygen potential surface, recombine to form molecular oxygen, and desorb into the gas phase. In a reactive environment, the low oxygen surface reactions will depend on the other chemical species present in the gas phase and adsorbed on the low oxygen surface. It is well established that equilibrium oxygen ion content in these materials varies with both temperature and ambient oxygen partial pressure, and oxygen content losses beyond a material’s phase stability limit can cause chemical decomposition. Under a constant oxygen partial pressure, equilibrium oxygen ion content decreases with increasing temperature. In some cases, certain constituents can even be reduced to their metallic state by temperature alone [47]. A reducing atmosphere (e.g., one containing hydrogen and/or methane or, according to Ma and Balachandran [48], one with an oxygen partial pressure < 10 6 atm) can produce the same effect at even lower temperatures [35,49].

Self-Adjusting Phase Equilibria For metal oxide materials with labile oxygen ions, changes in oxygen content following changes in temperature and/or oxygen environment can be manifested as phase adjustments. Studies using in situ X-ray diffraction (XRD) have confirmed that responsive phase adjustments in oxygen-MIEC ceramics are common. Depending on both the material and the new environment, individual phase changes can occur throughout the entire membrane or they can be highly localized (e.g., only at a surface or as distributed pockets of a newly formed phase in an equilibrium mixture of phases) [35,42,49]. For a material with a large variety of phase options that is exposed to a continuous oxygen partial pressure gradient, it is reasonable to expect the formation of a phase composition gradient that reflects the electrochemical oxygen potential distribution in the membrane.

238

Inorganic, Polymeric and Composite Membranes

Numerous researchers have confirmed that certain materials, most notably SrFeCo0.5O3d (SFC), can consist entirely of intimately intermixed phases. Such mixtures are referred to as “phase assemblages” or “solid solutions,” and they will adjust their phase distribution according to their environment [50–53]. Phase distribution changes in oxygen-MIEC ceramics have been observed to be readily reversible [25,48,54], with the consequence that equilibrium phase distributions can be highly sensitive to both temperature and gaseous environment [51]. However, phase transition kinetics for solid-phase equilibrium transitions are slow enough at lower temperatures that phase change reversibility can be masked or suspended [55]. This allows nonequilibrium compositions to be maintained at temperatures below the activity threshold. Another consequence of the phase lability of MIEC ceramics is their potential to exhibit known phase changes at lower temperatures than expected in response to gas-phase oxygen content manipulation [53]. This phenomenon confirms that equilibrium phase composition and phase distribution are determined by membrane oxygen content, which depends on both temperature and gaseous environment. This feature increases both the opportunities for, and the challenges associated with, oxygen-MIEC membrane applications. Membrane fracture, which is the common shortcoming of ceramic oxygen-MIEC materials, particularly in partial oxidation applications [5], can result from chemical decomposition in the form of phase conversion and segregation or from lattice expansion mismatch within either a single phase or a set of phases [39,54].

Chemical Expansivity Oxygen-MIEC ceramics generally expand as their oxygen content decreases. A decrease in oxygen content increases lattice oxygen vacancies and reduces the oxidation states of a portion of the membrane’s metal ions, both of which diminish the overall binding forces within the lattice and thus allow it to expand even under isothermal conditions [56–59]. To illustrate the potential extent of lattice expansion from compositional changes, the perovskite-to-brownmillerite phase transition has been observed to produce a unit-cell volume increase of up to 6% for SrCo0.8Fe0.2O3d (SCF) [57]. Adler has proposed that this “chemical expansivity” should be considered a new physical property for these materials. He concludes that chemical expansion from oxygen content decrease can be significantly greater than thermal expansion alone and argues that labeling as thermal expansion— the volume increases of oxygen-conducting MIECs during temperature increases—oversimplifies the phenomenon [56]. Oxygen-MIEC nonuniform chemical expansion is a much more likely explanation for a fracture-inducing density difference than nonuniform thermal expansion. At high temperatures, electronic conductivity is a characteristic feature of these mixed-conducting materials and is necessary to their oxygen

Chapter

11

Review of MIEC Ceramic Membranes

239

transport function. The Wiedemann–Franz law states that the ratio of thermal conductivity to electronic conductivity increases with temperature for solid materials. At temperatures high enough to conduct electrons (i.e., approaching 1000 K), these materials should therefore exhibit sufficient thermal conductivity to support the assumption of nearly isothermal behavior at steady state. The membrane’s lattice oxygen content, on the other hand, is clearly not isocratic, so a chemical expansion gradient is inevitable. The likelihood that an imposed chemical gradient will be more extreme than a possible thermal gradient supports the argument that chemical expansion under operating conditions is both greater and more important than thermal expansion.

Microstructure of Oxygen-MIEC Ceramics Researchers have been trying to relate oxygen transport through MIEC membranes to membrane microstructure and then use this understanding of the relationship to design membranes with higher oxygen flux [60–66]. The microstructure properties of interest include grain size, the distribution of the grain boundary, and inhomogeneous grain boundary composition (i.e., differences between grain boundary and bulk compositions). Recent studies have suggested that the grain boundary provides the pathway for oxygen transport in the membrane material and that sintering conditions can significantly impact grain boundaries. One study on La0.5Sr0.5FeO3d showed that as sintering temperature and time increased, grain size increased while grain boundary decreased [61]. The resulting oxygen permeation experiments showed that membranes with larger grain sizes exhibited lower oxygen fluxes. A similar phenomenon was observed on LaCoO3d membranes [62]. It was suggested in these studies that the grain boundary acts as a pathway for oxygen transport, so that a membrane with smaller grain boundaries produces lower oxygen flux than a membrane of the same bulk material with more extensive grain boundaries. Similar to the observations with the lanthanum-based membranes, studies on CaTi0.8Fe0.2O3d, Ba1xSrxCo0.8Fe0.2O3d, and Ba0.5Sr0.5Fe0.8 Zn0.2O3d membranes have shown that increasing sintering temperature and time produces larger grain size [60,65,66]. However, in contrast to the lanthanum-based membrane studies, increased grain size was correlated with significantly increased oxygen fluxes with the Ca- and Ba-based membrane materials. In these cases, the grain boundary seemed to act as a barrier to oxygen transport. The seeming contradiction in the effect of sintering conditions on the oxygen flux can be explained by differences in the role of the grain boundary in oxygen transport. For materials in which the grain boundary acts as a barrier to diffusion, larger grains and thus, smaller grain boundaries, will result in increased oxygen flux [64]. In contrast, for materials in which the oxygen transport is facilitated by the grain boundaries, smaller grains with larger grain boundaries will increase oxygen flux. While this explanation suggests that the grain

240

Inorganic, Polymeric and Composite Membranes

boundary has a profound effect on the oxygen flux, it does not reveal the true mechanism of oxygen transport through the grain boundary or provide a simple way to the effect of grain boundary on flux. More research in this area is necessary to answer these questions.

COMMON OXYGEN-MIEC MEMBRANE MATERIALS Oxygen-MIEC membrane materials can be classified into two main groups based on crystal structure: fluorite and perovskite. In recent years, dual-phase composite MIEC membranes have also been developed in an attempt to create membranes with both high oxygen permeability and good mechanical stability. These dual-phase membranes are the focus of Chapter 12 and will only be introduced briefly here.

Fluorites The fluorite oxide structure is represented by AO2, where A is the large fourvalent cation such as Zr4þ and Ce4þ [67]. Fluorite contains oxygen anions in simple cubic packing, with half of the interstices occupied by metal cations. The metal cations occupy the cube centers and the oxygen anions are tetrahedrally coordinated to the metal cations [68]. Two fluorite oxides known to have particularly high oxygen ion mobility, ZrO2 and Bi2O3, have been extensively investigated for use as oxygen-MIEC membrane materials. At room temperature, ZrO2 has a monoclinic crystal structure. As the temperature increases, the crystal structure of ZrO2 transforms to the tetragonal (> 1000
C) and cubic structures (> 2300
C) [69]. It is generally believed that cubic ZrO2 has a high oxygen ionic conductivity [70]. To maintain the cubic phase at room temperature and thus eliminate the mechanical stresses created by volume expansion from phase transition, dopant materials such as CaO, MgO, and Y2O3 can be added to pure ZrO2 [69,71,72]. In this scenario, the solid solution is referred to as “stabilized ZrO2.” Bi2O3 has four polymorphs, designated a-, d-, b-, and g-Bi2O3. Among the four phases, the d-phase Bi2O3 exhibits high oxygen ionic conductivity [73]. As with ZrO2, compounds such as Y2O3 and Er2O3 are added to pure Bi2O3 to maintain the d-phase at low temperatures [74,75]. Compared to ZrO2, Bi2O3 shows higher oxygen ionic conductivity at intermediate temperatures (< 800
C). The doped Bi2O3 oxides, especially BIMEVOX (where “ME” represents a metal ion, such as copper or cobalt, and “V” represents the vanadium), are suggested to be some of the highest ionic conductors at intermediate temperatures [8,76,77]. Even though ZrO2 and Bi2O3 have high oxygen ionic conductivity, their electronic conductivity is low (the electronic transference number is close to zero). These materials are thus commonly used as the electrolyte or anode in SOFC systems. To form oxygen-MIEC ceramic membranes, a dopant

Chapter

11

Review of MIEC Ceramic Membranes

241

compound with a high electronic conductivity has to be introduced to ZrO2 or Bi2O3. If the added compound has similar thermal properties and can form a solid solution with ZrO2 or Bi2O3, the doped ZrO2 or Bi2O3 becomes a single-phase oxygen-MIEC material. CeO2 is a primary dopant because cerium has multivalent states (þ 4, þ 3) and the transition from Ce4þ and Ce3þ increases the electronic conductivity of the material [12,74,78,79]. However, even with the addition of cerium, doped fluorite compounds still show much lower electronic conductivity than ionic conductivity. For a relatively thick fluorite-based oxygen-MIEC membrane, the oxygen permeation is therefore determined by electron conduction [12,74,80–82]. Table 11.1 provides an overview of the more common fluorite-type oxygen-MIEC membrane materials reported in the literature to date.

Perovskites Any metal oxide with the general formula ABO3, where A is the larger cation with a 12-fold oxygen ion coordination and B is the smaller cation with a 6-fold coordination, is considered to be a perovskite oxide. Twelve oxygen anions are cuboctahedrally coordinated to the A-site cations and six oxygen ions are octahedrally coordinated to the B-site cations. After doping with other metal cations, the perovksite can be symbolized by the formula AxA10  xByB10  yO3  d [85]. Generally speaking, A-site ions are rare-earth metals, A0 -site ions are alkaline-earth metals such as Ca2þ, Sr2þ, and Ba2þ, and B- and B0 -site ions are transition metals such as Co3þ and Fe3þ. A single-phase perovskite compound with five or more constituent metal species is rare, while compounds with three and four metal species are quite common. Goldschmidt defined a tolerance factor t to evaluate the structural stability of perovskite materials: RA þ RO2 t ¼ pffiffiffi ; 2 RB þ RO2

ð11:1Þ

where RO2 is the radius of the oxygen ions, RA is the radius of the A-site ion, and RB is the radius of the B-site ion. In general, structural stability of a material increases as t increases, with t-values between 0.75 and 1.0 representing a stable material [86]. Oxygen transport in perovskites is largely determined by the activation energy for oxygen ion conduction. Reducing the activation energy will effectively increase oxygen ion diffusion and enhance net oxygen flux. Researchers believe that this activation energy depends on the average metal-oxygen bond energy, the lattice free volume, and the radius of opening between the two A-site cations and one B-site cation [14,15]. Low metal-oxygen bond energy, large lattice free volume, and large openings are associated with lower activation energy. Lattice free volume

TABLE 11.1 Comparison of Oxygen Fluxes for Various Fluorite Membranes Reported in the Literature Membrane Material Bi1.5Y0.3Sm0.2O3 [34] (Bi2O3)0.73(CaO)0.27 [75] (Bi2O3)0.75(Y2O3)0.25 [83]

Temperature (
C) 750–950 600–680 800–950

Oxygen Flux (mol s 1 m 2) 3.5  10

5

4

–4.3  10

Disk

1.27

1.230  10

5

Disk

1.2

7.840  10

5

4

Disk

1.4

Disk

2

Tube

1.5

Tube

1.5

Tube

2

Disk

2

Disk

3  10 3–16  10 3

–3.300  10

1.0  10

0.8ZrO2–0.1TiO2–0.1Y2O3 [71]

1305–1481

0.018–0.12 3

1305–1481

9  10

[(ZrO2)0.7(CeO2)0.3]0.9(MgO)0.1 [72]

1027–1477

0.12–3.8

900–1000

5

–7.6  10

0.825ZrO2–0.075TiO2–0.1Y2O3 [71]

(ZrO2)0.8(Y2O3)0.20 [84]

–3.351  10

5

650–850

900

Thickness (mm)

5

BiY0.5Cu0.5O3 [81]

(ZrO2)0.7(Tb2O3.5)0.3 [80]

Membrane Configuration

2

–4.2  10

2.6  10

7

2.1  10

5

5

–5.5  10

To convert to molar fluxes, all volumetric fluxes reported in the literature were assumed to be at T ¼ 25
C and P ¼ 1 atm unless otherwise stated in the reference.

Chapter

11

Review of MIEC Ceramic Membranes

243

increases as the radii of the A- and B-site cations increase. Furthermore, the combination of large A-site cations and small B-site cations will increase the size of the opening between the two A-site cations and one B-site cation. Two principles should be noted when doping A- and B-site cations (1) the radii of dopants must match the radii of the cations that the dopants replace and (2) the concentration of heterovalent dopant within the perovskite lattice must less than the level at which the introduced lattice vacancies become ordered [15,44,45]. One example of the effect of lattice order on oxygen conductivity can be found in the orthorhombic brownmillerite structure (ABO2.5). A brownmillerite is essentially a highly ordered perovskite derivative that can occur when a variable oxygen content perovskite phase approaches an oxygen deficiency of 16.7% (i.e., one-sixth) [5]. It is generally believed that the highly ordered oxygen ion vacancies in brownmillerites lead to low oxygen flux because the brownmillerite phase is more stable than the related perovskite and does not support oxygen hole transport. However, at high enough temperature brownmillerite phases may disorder sufficiently to return the material to a high-vacancy perovskite structure with higher oxygen transport capability [11,39,49,87,88]. The difference in behavior observed between perovskites and their associated brownmillerite phases confirms that oxygen vacancy disorder can be a determining factor for oxygen conductivity. A- and B-site perovskite doping studies with heterovalent metal ions have produced many oxygen-MIEC compounds. Table 11.2 provides the compositions and properties of some of these perovskite-based membrane materials. Because a detailed review covering the entire progress of these perovskite oxygen-MIEC materials is beyond the scope of this book, the discussion will be limited to the most commonly studied perovskite or perovskite-like materials.

SCF-Based Materials Perovskites made from strontium, iron, and cobalt have attracted substantial attention since Teraoka [46] published his pioneering work with this family in 1985. Compared to other early oxygen-MIEC materials, Sr–Co–Fe perovskite materials, particularly SrCo0.8Fe0.2O3d, show a notably high oxygen flux. Teraoka et al. [46] reported an oxygen flux of about 0.02 mol m 2 s 1 in an air: helium gradient at 900
C with a 1-mm-thick disk-shaped SrCo0.8Fe0.2O3d membrane. It should be noted that SCF is not a perovskite-type compound at room temperature. However, it undergoes a phase transition from the brownmillerite phase to the defect perovskite phase at elevated temperatures [101]. This phase transition represents an increase in lattice oxygen vacancy disorder, which is responsible for the observed oxygen flux increase. Although SCF has exhibited high oxygen flux, additional work demonstrated that the SCF material suffers from poor chemical and structural stability,

TABLE 11.2 Comparison of Oxygen Fluxes for Various Perovskite Membranes Reported in the Literature Membrane Material

Temperature (
C)

Oxygen Flux (mol s 1 m 2)

Membrane Configuration

Thickness (mm)

(BiO1.5)0.86BaO [90]

650–950

3.7  10 5–9  10 3

Disk

0.55

Disk

1.1

Disk

1.65

650–950 650–950 Bi0.85Sr0.15FeO3 [90] Bi0.7Sr0.3FeO3 [90] Bi0.4Sr0.6FeO3 [90] Bi0.2Sr0.8FeO3 [90] BaCe0.4Fe0.6O3d [91] BaCo0.4Fe0.5Zr0.1O3d [9] Ba0.5Sr0.5Co0.8Fe0.2O3d [92] Ba0.5Sr0.5Zn0.2Fe0.8O3d [93] CaTi0.8Fe0.2O3d [17] La0.7Ca0.3CrO3d [20] LaCo0.8Fe0.2O3d [94] La0.6Sr0.4Co0.8Cr0.2O3d [94]

800 800 800 800 800–950 700–950 850–900 800–975 800–1000 800 860 860

5

3.7  10

5

3.7  10

3

–7.3  10

3

–6  10

4

4.2  10

Disk

1

4

Disk

1

4

Disk

1

7.7  10 2.2  10

3

1.1  10 7.4  10

4 3

2.0  10 9.0  10

4 2

1.1  10

5

8.0  10

1

Disk

1

3

Disk

1

2

Disk

1.8

2

Disk

1.45

4

Disk

1

Disk

1.07

Disk

1

Disk

1.5

–1.8  10 –6.8  10 –1.6  10 –2.6  10 –2.2  10

5

6  10

4

1.8  10

4.21  10

Disk 3

3

La0.6Sr0.4Co0.2Fe0.8O3d [95] La0.8Sr0.2Ga0.7Co0.3O3d [96] (2%wt)Y0.2Ce0.8O1.9–(98%wt)La0.7Ca0.3CrO3d [20] (4%wt)Y0.2Ce0.8O1.9–(96%wt)La0.7Ca0.3CrO3d [20] (6%wt)Y0.2Ce0.8O1.9–(94%wt)La0.7Ca0.3CrO3d [20] (8%wt)Y0.2Ce0.8O1.9–(92%wt)La0.7Ca0.3CrO3d [20] (10%wt)Y0.2Ce0.8O1.9–(90%wt)La0.7Ca0.3CrO3d [20] (La0.75Sr0.25)0.95Cr0.5Mn0.5O3d [16] SrSc0.5Co0.95O3d [97] La0.6Sr0.4Fe0.9Ga0.1O3d [98] (98%vol)La0.6Sr0.4Fe0.9Ga0.1O3d–(2%vol)MgO [98] (95%vol)La0.6Sr0.4Fe0.9Ga0.1O3sd–(5%vol)MgO [98] La0.4Sr0.6Co3d [46] SrCo0.9Nb0.1O3d [99]

850–900 700–1000 800 800 800 800 800 950–1000 675–900 875–975 825–925 825–925 870 700–900 700–900 700–900

SrFe0.6Cu0.3O3d [100]

750–950

4  10 4–1.1  10 3 3

2.3  10

2

–1.1  10

5

1.4  10

5

3.5  10

5

6  10

5

6.4  10

5

7  10

7

1  10

5

–1.3  10

3

1.8  10

5

2.2  10

5

4.4  10

4

1.2  10

2

–2.2  10

4

–1.2  10

5  10

0.5

Disk

1.07

Disk

1.07

Disk

1.07

Disk

1.07

Disk

1.07

Disk

1

Disk

1 1

Disk

1

4

Disk

–2.7  10 –4.2  10

2

–1.8  10

3

Disk

Disk

3.8  10

2.5  10

0.219

4

3 3

Tube

Disk

1

Disk

1.5

2

–2.7  10

3

5.5  10

3

2.6  10

1

2

–3.2  10

3

–7.1  10


0.7 Disk

1.5

To convert to molar fluxes, all volumetric fluxes reported in the literature were assumed to be at T ¼ 25 C and P ¼ 1 atm unless otherwise stated in the reference.

246

Inorganic, Polymeric and Composite Membranes

especially in a reducing atmosphere [54,88]. For example, severe cracks occurred in an SCF tubular membrane after only a few minutes of exposure to a POM environment [54]. Extensive work has been performed to improve the stability of SCF with the addition of promoters.

Promoted SCF Materials The addition of small amounts of silver to an SCF material led to an increase in oxygen flux, with the highest flux obtained at a silver content of 5 mol%. Upon further investigation of the Ag–SCF membrane, the added silver was determined to have increased the surface oxygen exchange rate [102]. However, the addition of silver did not impact the phase transition between the brownmillerite and perovskite phases, and the stability of the Ag–SCF membrane was not significantly improved. Wu et al. [103] attempted to utilize the thermal and chemical stability of Al2O3 to improve the stability of SCF. They were able to significantly improve the stability of the SFC in low oxygen partial pressure, high-temperature environments by introducing Al2O3. However, because Al2O3 reacts with cobalt to form a spinel phase (CoAl2O4) at elevated temperatures (1200
C), the doping of Al2O3 into SrCo0.8Fe0.2O3d resulted in a loss of cobalt from the perovskite phase and a dramatic decrease in membrane oxygen flux. Fan et al. [104] investigated the use of tin (Sn) to improve the stability of SCF. The introduction of tin as SrSnO3 not only decreased the thermal expansion coefficient of the material, but also reduced the onset temperature of the oxygen permeation to 560
C. Decreased thermal expansion is associated with greater mechanical stability, and a decrease in activity threshold temperature increases the potential utility of a membrane material. However, the oxygen permeability of the SrSnO3-doped SCF was lower than that of the parent SCF at temperatures higher than 850
C. Substituted SCF Materials Other researchers have modified the SrCo0.8Fe0.2O3d material by partially substituting for strontium in the A site of the perovskite structure. One example of this is La0.6Sr0.4Co0.2Fe0.8O3d [105,106]. Oxygen flux increased with temperature in a series of tests with a La0.6Sr0.4Co0.2Fe0.8O3d tubular membrane with 1.5-mm wall thickness. At 850
C, the oxygen flux was 9.67  10 4 mol m 2 s 1 under a helium:air gradient. However, the performance of the LSCF membrane deteriorated over time under this gradient, and the loss of flux was ascribed to the perovskite phase on the surface of the membrane decomposing to SrSO4, CoSO4, SrO, Co2O3, and La2O3 over the 110 h of testing [106]. Some researchers have substituted cobalt-doped materials with less reducible ions such as Ga4þ as a means of improving the chemical stability of the material. It was believed that materials with lower thermal expansion

Chapter

11

Review of MIEC Ceramic Membranes

247

coefficients would have less lattice expansion and might be more stable in reducing environments [99,107]. Oxygen flux testing on a disk-shaped La0.7Sr0.3Ga0.6Fe0.4O3d membrane with 0.5-mm thickness showed that at 1000
C, oxygen fluxes of 0.012 mol m 2 s 1 under an air:helium gradient were achieved [107]. Although the study indicated that the high-density La0.7Sr0.3Ga0.6Fe0.4O3d ceramic had a low thermal expansion coefficient, it also showed that this material is susceptible to decomposition to LaSrGaO4 at temperatures higher than 700
C in a reducing atmosphere [108]. The most successful modification of SrCo0.8Fe0.2O3d (SCF) has been Ba0.5Sr0.5Co0.8Fe0.2O3d (BSCF). A 1.8-mm-thick disk-shaped BSCF membrane has exhibited an oxygen flux of 9.5  10 3 mol m 2 s 1 in an air:argon gradient at 850
C [92]. This partial substitution of the A-site strontium cation with barium has been shown to improve phase stability by preventing oxidation of the B-site cation, thus maintaining perovskite content in the membrane material [109]. However, some recent studies [60,110,111] have shown that the BSCF membranes are sensitive to CO2 because of the alkaline-earth elements in the structure. For example, Jiang et al. [111] investigated the performance of the BSCF membrane in the CO2 reforming reaction. The results showed that the presence of a large amount of CO2 leads to surface restructuring of the BSCF membrane, which decreases the oxygen flux. This restructuring is ascribed to the formation of carbonate on the reaction surface of the membrane.

Multiphase SrFeCo0.5Ox Balachandran et al. [112,113] adjusted the composition of Sr–Co–Fe perovskite to develop a new material, SrFeCo0.5Ox (SFC). Compared to SrCo0.8Fe0.2O3d, SFC showed good mechanical integrity under reaction conditions with a strongly reducing environment, and was reported to be quite stable even under the extreme oxygen potential gradients in a POM membrane reactor [115]. Initial reports of SFC’s high oxygen flux by Ma et al. and Maiya et al. led to numerous investigations of SFC, but the subsequent work demonstrated a large variation in oxygen flux with SFC membranes, with most results much lower those initially reported [39,40,50,51,55,115–117]. Stoichiometrically, SFC is not a perovskite material. However, SFC typically occurs as an adaptable three-phase solid solution (or “phase assemblage”) with small, highly intermixed phase moieties, including perovskite phases [48,50–53,55,114,116]. The two-phase categories observed in SFC samples in addition to the perovskite phases (SrFe1xCoxO3d) are spinel phases (Co3xFexO4) and the so-called “intergrowth” phases (SrFe1.5xCoxO3.25d). The name “intergrowth” refers to the structural appearance of this phase as interspersed perovskite and brownmillerite units [50]. All three phase types have adjustable stoichiometry, leading to a highly interdependent distribution of cations and phase shifts within phase categories as well as between phase categories. It has been shown that SFC’s initial phase

248

Inorganic, Polymeric and Composite Membranes

composition can vary substantially with both material preparation and membrane fabrication methods [116]. The ratios and compositions of the three common phase types depend on the synthesis method, the preparation procedure, and the ultimate environment of the prepared material. One additional phase type has also been reported to appear on highly reduced membrane surfaces: the Co–Fe spinel phase can decompose to a cubic “rocksalt” phase (Co1xFexO) at temperatures greater than 900
C in air or at lower temperatures in reducing environments [52,53]. The overall body of work has determined conclusively that oxygen flux through SFC membranes correlates directly with its perovskite phase content [50,51]. Xia et al. [51] observed the creation of greater amounts of perovskite phase with annealing temperatures above 1150
C and also noted that the oxygen flux for SrFeCo0.5Ox slowly decreases over time due to a slow transition of perovskite phase to the intergrowth phase. Other researchers have noted increased perovskite content with exposure to high temperature/low pO2 environments (i.e., conditions that lead to low equilibrium oxygen content in the SFC material) [48,50,53]. Ikeguchi et al. [116] reported a 17% decrease in oxygen flux with a decrease in membrane sintering temperature from 1200 to 1150
C, with a decrease in perovskite content as the presumptive cause of the flux decrease. Armstrong et al. studied carefully prepared samples of known phase composition for each of the three phase types, and the oxygen flux results correlated emphatically with perovskite phase content. Oxygen flux through the intergrowth phase samples was two orders of magnitude lower than that with the representative perovskite material (SrFe0.75Co0.25O3d) and one order of magnitude lower than that of a typical three-phase SFC “phase assemblage” material [50]. Xia et al. [51] also reported slow conversion back to an intergrowth phase at temperatures below 1000
C. The relatively slow transition to the intergrowth phases at lower temperatures allows “meta-stable” perovskite phases to persist at lower temperatures than expected.

Dual-Phase Composite Materials It is clear from the literature that several key challenges still remain with singlephase oxygen-MIEC membranes. The oxygen flux of the MIEC membranes is highly correlated with the ionic and electronic conductivity of the material, with both high ionic and electronic conductivity necessary to achieve high oxygen flux. However, limited success has been found in identifying materials with both high ionic and electronic conductivity. Furthermore, materials that exhibit high oxygen flux often suffer from lower thermal or chemical stability when exposed to the large oxygen partial pressure gradients present for most membrane reactor applications. This is especially true when the membrane is used in a hydrocarbon conversion reactor, where one surface is exposed to a highly reducing reaction environment and the other surface is exposed to an oxygenrich environment.

Chapter

11

Review of MIEC Ceramic Membranes

249

As work to overcome the challenges in single-phase oxygen-MIEC membrane development continues, researchers have been evaluating dual-phase composite materials as an alternative approach to fabricating membranes with high ionic and electronic conductivity while maintaining chemical and thermal stability [118–123]. A dual-phase composite membrane combines a material with high oxygen ionic conductivity with a second material with high electronic conductivity. In these materials, oxygen ion and electron transport occur in different phases. The ion- and electron-conducting phases must form a continuous and intimately intermixed network to allow simultaneous and balanced transport of oxygen ions and electrons under the electroneutrality constraint of zero net charge. The conductivity of the composite material therefore depends strongly on the volume ratio of two phases. Theoretically, it is easy to design high oxygen flux membranes by selecting appropriate materials and adjusting the ratio of two phases. As mentioned earlier, fluorite materials such as YSZ have high oxygen ionic conductivity. Mazanec et al. [120] were able to achieve high oxygen fluxes by incorporating noble metals such as platinum into YSZ. However, further study showed that at least 30 vol% of the metal has to be added to the ceramic material to form an effective electron-conducting phase network [121]. From a practical perspective, the high price of noble metals restricts the development and application of composite membranes that include a noble metal. Replacing expensive noble metals with perovskite materials that exhibit high electronic conductivity is a promising alternative. For example, Sirman et al. [122] produced a dual-phase composite with a very low thermal expansion coefficient by mixing Ce0.8Gd0.2O1.9 (a fluorite oxygen ion conductor) and La0.8Sr0.2Fe0.8Co0.2O3d (a perovskite electron conductor) in equal parts by volume and sintering in the temperature range of 1200–1400
C. This dualphase composite membrane exhibited an oxygen flux of 0.042 mol m 2 s 1 at 1000
C. Wang et al. [123] successfully prepared a La0.15Sr0.85Ga0.3 Fe0.7O3d–Ba0.5Sr0.5Co0.8Fe0.2O3d (LSGF–BSCF) composite that exhibited an oxygen flux of 3.1  10 3 mol m 2 s 1 at 915
C, which was about nine times greater than the oxygen flux through a single-phase LSGF membrane. Table 11.3 provides an overview of the dual-phase composite membrane work available in the literature and a more detailed discussion of recent work in dual-phase membranes is provided in Chapter 12.

Membrane Modifications to Improve Oxygen Flux Oxygen transport through dense oxygen-MIEC membranes requires multiple steps, as shown previously in Figure 11.1. For a given membrane, if the ratelimiting step of the oxygen permeation is determined, appropriate modification techniques can be employed to enhance the performance of the membrane. For example, if oxygen transport is limited by the surface exchange reactions, a surface modification such as a catalyst coating or an increase in surface area can increase the surface exchange of oxygen and thus increase membrane oxygen

TABLE 11.3 Comparison of Oxygen Fluxes for Various Dual-Phase Composite Membranes Reported in the Literature Membrane Material

Temperature (
C)

Oxygen Flux (mol s 1 m 2)

Membrane Configuration

Thickness (mm)

((ZrO2)0.94(Y2O3)0.06)0.5–Pd0.5 [120]

1100

1.5  10 2–1.6  10 2

Disk

0.8

((ZrO2)0.94(Y2O3)0.06)0.5–Pt0.5 [120]

1100

1.4  10 2

Disk

0.8

(ESB)0.6–(Ag)0.4 [75] (ESB¼(Bi2O3)0.75(Er2O3)0.25)

750–850

8.5  10 4–3.1  10 3

Disk

0.23–1.6

(ESB)0.6–(Au)0.4 [124] (ESB¼(Bi2O3)0.75(Er2O3)0.25)

650–850

2.8  10 4–5  10 4

Disk

1

Zr0.84Y0.16O1.92 [125]

850–1050

4  10 4–2  10 3

Tube

0.16

Bi1.6Y0.4O3–Ag [126]

650–950

2  10 3–6  10 3

Disk

1.44

(Y2O3)0.08(ZrO2)0.92–Ni [127]

950–1000

3  10 3–4.2  10 3

Disk

0.7

950–1000

2.5  10 3–3  10 3

Disk

1.52

Disk

2.82

Tube

1.25

Disk

0.3

950–1000 (SrFeO3s)0.7(SrAl2O4)0.3 [128] Ce0.8Sm0.2O2s–La0.8Sr0.2CrO3d [129]

800–900 850–950 850–1000 850–1000 850–1000

SrSc0.2Co0.8O3s–Sm0.5Sr0.5CoO3d [130]

700–900

1.1  10

3

3

1.6  10

4

3

–1.5  10 –2.5  10

4

5  10

4

4  10

3

–1.6  10

3

–2  10

Disk

0.6

3.2  10

4

3

Disk

1

1.3  10

4

4

Disk

1.7

Disk

0.85

–1.3  10

3

2  10

–6.3  10

3

–7.5  10

Ce0.8Gd0.2O0.19–Gd0.2Sr0.8FeO3d [131]

800–1000

1.1  10 3–9  10 3 4

4

Disk

0.5

Disk

1.0

(50%wt)Ce0.8Gd0.2O2d–(50%wt)La0.8Sr0.2Fe0.8Co0.2O3d [121]

750–950

2.0  10

(50%wt)Ce0.8Gd0.2O2s–(50%wt)La0.7Sr0.3FMnCo0.2O3d [121]

750–950

1  10 4–6  10 4

Disk

1.0

((ZrO2)0.94(Y2O3)0.06)0.5–Pd0.5 [120]

1100

1.5  10 2–1.6  10 2

Disk

0.8

–6  10

To convert to molar fluxes, all volumetric fluxes reported in the literature were assumed to be at T ¼ 25
C and P ¼ 1 atm unless otherwise stated in the reference.

252

Inorganic, Polymeric and Composite Membranes

flux. For membranes limited by bulk diffusion rates, reducing the thickness of the membrane can increase the oxygen flux significantly. If both surface exchange and bulk diffusion limit oxygen transport, surface modification and membrane thickness reduction can both be applied to increase oxygen flux. Recent studies have shown that surface modification and/or thinner membranes can both increase oxygen transport through MIEC ceramic membranes.

Surface Modifications When oxygen transport is limited by the rate of oxygen exchange on the surface of the membrane, increasing surface area can be one strategy to increase the oxygen flux. Polishing the membrane surface with different media can lead to the roughened surface which increases the surface-to-volume ratio. Experiments on a La0.1Sr0.9Co0.9Fe0.1O3d membrane have shown that an increase in the oxygen flux can be accomplished by using this simple method [132]. Another effective means to increase the surface area is coating a porous layer on the dense oxygen-MIEC membranes. The porous coating material can be the same as the membrane material or it can be another oxygen-MIEC material [133–135]. Compared to the simple roughening method, coating a porous layer on the dense membrane has been shown to be more effective and can lead to greater oxygen flux enhancement. Coating an oxygen dissociation catalyst on the membrane surface can facilitate the adsorption and dissociation of oxygen and increase the oxygen flux. Researchers showed that La0.6Sr0.4Co0.2Fe0.8O3d membranes [136] coated with a silver catalyst on the air side surface of the membrane improved the oxygen permeability of the membrane significantly. Compared to unmodified membranes, oxygen flux through silver-coated membranes at 1000
C in an air:helium gradient increased by a factor of 2.5. Similarly, studies of the effect of both a platinum pattern [115] and a very thin platinum film (< 1 nm) [137] showed that the deposited catalyst substantially enhanced membrane oxygen flux. As shown in Figure 11.2, an SFC membrane with a pattern of 20-nm-thick platinum disks exhibited an oxygen flux two times higher at 800
C than an identical membrane without the platinum pattern [115]. Studies with Sr0.97Ti0.6Fe0.4O3d membranes also demonstrated an increase in oxygen flux following the application of a platinum-based oxygen dissociation catalyst to the membrane surface [138]. In contrast to the studies on the SFC and Sr0.97Ti0.6Fe0.4O3d membranes, platinum and silver coatings on the air side surfaces of 1.8-mm-thick BSCF membranes did not result in an improvement in oxygen flux. Additional studies showed that oxygen transport through BSCF membranes greater than 1.8 mm in thickness is not limited by surface exchange on the air side of the membrane but is limited by bulk diffusion of oxygen ions through the membrane.

11

Membrane oxygen flux (mol m-2 s-1)

Chapter

253

Review of MIEC Ceramic Membranes

1 ⫻ 10-3 8 ⫻ 10-4

Plain membrane Pt-patterned membrane

6 ⫻ 10-4 4 ⫻ 10-4 2 ⫻ 10-4 0 450

550

650 Temperature (°C)

750

850

FIGURE 11.2 Oxygen flux results for plain and Pt-patterned membranes (Pt pattern installed on oxygen supply side).

Membrane Thickness Reduction Under bulk diffusion control, oxygen flux across an oxygen-MIEC membrane can be correlated with the inverse of membrane thickness as described by the Wagner equation (Equation 11.2) [139]: ð lnP0 O2 RT si se d lnP; ð11:2Þ JO2 ¼ 2 00 16F L lnPO si þ se 2

where R is the gas constant (J mol 1 K 1), T is the temperature (K), F is the Faraday constant (C mol 1), L is the thickness of membrane (m), si is the oxygen ionic conductivity (S m 1), se is the electronic conductivity (S m 1), and P0O2 and 00 PO2 are the oxygen supply side and oxygen permeate side partial pressures, respectively (Pa). In theory, then, reducing membrane thickness can increase oxygen flux when oxygen transport is limited by bulk diffusion. However, preparing a thin or ultra-thin membrane is not always feasible, and the thickness of the membrane can have a significant role in the mechanical stability of the membrane under operating conditions. For this reason, a substantial amount of research has focused on asymmetric membrane fabrication. Asymmetric membranes contain a thin film supported on a porous substrate. The initial studies in the literature concentrated on coating thin layers of oxygen-MIEC materials on conventional porous substrates such as Al2O3 because of the low cost and availability of the porous substrate [140]. However, to avoid the physical and chemical incompatibility between the thin layer and substrate, fabricating a thin layer and a porous substrate from the same materials is desirable. To date, many fabrication methods such as tape-casting [141], screen printing [142], slurry-coating [143], acid etching [144], and electrochemical vapor deposition (EVD) [84] have been used to prepare asymmetric oxygen-MIEC

254

Inorganic, Polymeric and Composite Membranes

membranes. For example, recent work on a BSCF membrane demonstrated that the oxygen flux of a BSCF asymmetric membrane was effectively enhanced as the thickness of the dense BSCF layer in the asymmetric membrane decreased (Figure 11.3) [111]. This increase in oxygen flux was attributed to the reduction in the thickness of the dense BSCF layer in the asymmetric membrane, leading to a reduction in diffusion resistance for oxygen transport through the membrane. In these studies, the thickness of the porous substrates for each membrane was kept the same to isolate the effect of the dense layer thickness. The observed oxygen fluxes in Figure 11.3 are much lower than theoretical values predicted by the Wagner equation. Additionally, the deviation of the measured oxygen fluxes from theory increases as the thickness of the thin layer decreases. Two factors could be responsible for the lower-thanpredicted oxygen fluxes. The first is that the porous support could still exert some resistance on oxygen transport. The second is that, as the thickness of the thin layer decreases, the influence of surface exchange could increase, causing a transition to mixed control scenario in which both bulk diffusion and surface exchange limit oxygen transport. In the mixed control scenario, oxygen flux is not inversely proportional to thickness and, thus, deviations from the theoretical oxygen flux values predicted by Wagner’s equation would be expected. New techniques for fabricating ultra-thin dense layers on thin yet robust porous supports continue to be explored. Novel hollow fiber oxygen-permeable ceramic membranes have been reported which have several advantages over membranes prepared using traditional fabrication techniques [144–148]. Compared to conventional tubular or disk-shaped membranes, hollow fiber membranes have a higher surface-to-volume ratio. The thinner membrane wall

Oxygen flux (mol m-2 s-1)

8 ⫻ 10-3

6 ⫻ 10-3

4 ⫻ 10-3

Asymmetric membrane Dense membrane

2 ⫻ 10-3 Theorectical oxygen fluxes 0 0

2 4 Inverse of thin layer thickness (mm−1)

6

FIGURE 11.3 The effect of the thicknesses of the dense layers of BSCF asymmetric membranes to the oxygen flux of BSCF membranes (T ¼ 800
C).

Chapter

11

Review of MIEC Ceramic Membranes

255

and asymmetric structure of hollow fiber membranes are favorable to oxygen permeation, especially for materials limited by bulk diffusion. Finally, the hollow fiber membrane configuration has been suggested to be more suitable for large-scale fabrication and for utilization in membrane reactors. Hollow fiber membranes can be prepared using a phase inversion spinning and sintering method [136,149]. The typical morphology of a hollow fiber membrane has finger-like structures near both the inner and outer membrane walls and a sponge-like thin layer at the center of the membrane wall. This asymmetric structure can be attributed to rapid precipitation at both the inner and outer walls close to the coagulant, which results in short finger-like pores, while slow precipitation at the center of the wall leads to the sponge-like structure. Compared to conventional tubular or disk-shaped membranes, a hollow fiber membrane with the same composition shows higher oxygen fluxes. For example, Tan et al. [145] reported that at 900
C, the oxygen flux of a La0.6Sr0.4Co0.2Fe0.8O3d hollow fiber membrane fiber is about 5.5  10 3 mol m 2 s 1 in an air:argon gradient, which is about 3.5 times greater than a conventional tubular La0.6Sr0.4Co0.2Fe0.8O3d membrane. Oxygen flux increases with hollow fiber membranes have been attributed to the significant reductions in wall thickness with the hollow fiber membranes and the attendant reduction in bulk diffusion limitations [136,146].

MIEC MEMBRANES FOR SYNTHESIS GAS PRODUCTION Reforming and partial oxidation of hydrocarbons to produce mixtures of hydrogen and carbon monoxide known as synthesis gas (or “syngas”) have been widely studied in MIEC membrane reactors [29,35,43,110,150–160] because of the potential for an oxygen-MIEC membrane to provide cheap, safe, and distributed oxygen. From a membrane performance perspective, motivations for incorporating MIEC membranes into hydrocarbon conversion reactors include (1) the high reaction temperature of the production of the synthesis gas allows effective membrane oxygen transport without a requirement for additional energy and (2) the strongly reducing atmosphere created by the reaction products provides a large oxygen potential gradient to facilitate oxygen transport through a membrane. Figure 11.4 illustrates a simple scheme for syngas production from methane in an oxygen-permeable membrane reactor. This figure shows the membrane in a disk configuration, but it should be noted that tubular configurations are also widely studied in the literature and are much more likely to be utilized in an industrial process.

Synthesis Gas Production Overview Syngas is an important industrial feedstock for hydroformylation and Fischer–Tropsch processes to make chemicals and fuels. Generally speaking,

256

Inorganic, Polymeric and Composite Membranes

syngas is produced by either partial oxidation or reforming of hydrocarbons. Among the hydrocarbons typically studied, methane (CH4) has received the most attention because it is both the simplest alkane and the main component of natural gas [43,150,155,159,161]. The three primary reactions used to produce synthesis gas from methane are shown in Table 11.4 [27,152]. POM yields a H2:CO product ratio of 2:1 that is ideal for Fischer–Tropsch and direct methanol syntheses of liquid fuels, which are also referred to as gasto-liquids (GTL) processes. One disadvantage of partial oxidation in general is its large oxygen requirement. Pure gas-phase oxygen is an expensive feedstock and the nitrogen in air precludes its direct use. About one-third of a POM synthesis gas facility’s operating costs and up to 45% of its capital costs can come from the air separation unit alone [39,162]. Any reduction in the cost of supplying oxygen to such a system would directly lower its production costs and thus increase its competitiveness. For example, Dyer et al. [10] estimated in 2000 that a 25% reduction in the cost of current GTL technology including oxygen separation would make GTL products competitive with oil at $20 per barrel, which could substantially increase GTL production. FIGURE 11.4 Membrane reactor scheme for the methane conversion to syngas using an MIEC ceramic membrane.

CH4

H2, CO, CO2, H2O

Catalyst

Ceramic membrane

e−

O2−

Air

TABLE 11.4 Comparison of Oxygen Fluxes for Various Dual-Phase Composite Membranes Reported in the Literature Reaction

Chemical Equation

DH298 K∘(kJ mol 1)

Product H2: CO Ratio

Partial oxidation

CH4 þ 12 O2 ! CO þ 2H2

36

2

Steam reforming

CH4 þ H2O ! CO þ 3H2

þ206

3

CO2 reforming

CH4 þ CO2 ! 2CO þ 2H2

þ247

1

To convert to molar fluxes, all volumetric fluxes reported in the literature were assumed to be at T ¼ 25
C and P ¼ 1 atm unless otherwise stated in the reference.

Chapter

11

Review of MIEC Ceramic Membranes

257

The most common method for producing synthesis gas in industry is steam reforming [163]. The high H2:CO ratio of the products is ideal for ammonia and urea production and also makes this the most desirable reaction for hydrogen production from hydrocarbons. However, steam reforming is highly endothermic, which is an expensive attribute with target reaction temperatures of at least 750
C for steam reforming of methane. In addition to the energy required to maintain the reaction temperature, a high steam to hydrocarbon feed ratio is typically required to minimize catalyst deactivation [27]. The CO2 reforming reaction utilizes methane and carbon dioxide, both of which are key greenhouse gases, to produce synthesis gas, and has been suggested as an environmentally benign strategy for reusing CO2 produced by other processes. The lower H2:CO ratio (1:1 or less) of the CO2 reforming reaction products also provides an opportunity to adjust the composition of synthesis gas to meet the needs of downstream processes such as hydroformylation that require lower H2:CO ratios than those produced by steam reforming and partial oxidation. Moreover, CO2 reforming can utilize the CO2 that naturally exists in natural gas. This removes the need to separate the CO2 from natural gas, which is particularly beneficial for oilfield gas sources that can contain large amounts of CO2. The major disadvantages for the practical application of CO2 reforming include the potential for rapid catalyst deactivation from carbon deposition, high reaction temperatures (beyond 1073 K), and unfavorable energetics. Like steam reforming, CO2 reforming is a highly endothermic reaction. A more recent approach to producing H2:CO ratios between one and two is a combination of partial oxidation and CO2 reforming known as combined reforming. Combined reforming has been explored with the intention of decreasing the endothermicity and poor catalyst life associated with straight CO2 reforming [164–176]. It also offers the potential for “tuning” the H2:CO ratio via manipulation of the ratio of O2 and CO2 in the reactor feed. Adding oxygen to promote catalyst performance and decrease the energy requirements of CO2 reforming could bring this potentially valuable syngas production reaction closer to mainstream industrial application. However, combined reforming requires pure oxygen as a feedstock, and as previously stated, an energy-intensive air separation unit represents a substantial capital and operating cost. Oxygen-MIEC membranes offer the potential to replace separated oxygen gas in future partial oxidation and combined reforming processes, and thereby reduce synthesis gas process operating costs.

Benefits of MIEC Membranes for Synthesis Gas Production In addition to the potential economic benefits, membrane-supplied oxygen incorporates several significant environmental and safety benefits over gasphase oxygen supplies for synthesis gas operations (1) a substantial reduction

258

Inorganic, Polymeric and Composite Membranes

in process energy consumption; (2) safer oxygen (i.e., no hotspot or flammability issues); (3) inherently distributed oxygen introduction, which can increase the selectivity of oxidation reactions, produce a more uniform and predictable reactor temperature profile, and possibly even reduce overall oxygen consumption; and (4) a reduction in homogeneous thermochemical reactions involving gas-phase oxygen which can produce soot-forming precursors. The potential advantages of membrane oxygen have led to numerous investigations of oxygen-MIEC membrane reactors for synthesis gas production. Membrane material composition and the effect of composition and structure on oxygen flux have a significant impact on the activity and stability of oxygenMIEC membranes in synthesis gas production environments. Several requirements must be met by industrially feasible membranes, including (1) high and steady oxygen flux during reaction, (2) considerable long-term mechanical and thermal stability under a reducing environment, and (3) inexpensive starting materials and fabrication methods [31].

Overview of Work to Date The majority of the work published to date on the use of oxygen-MIEC membranes for the production of synthesis gas has focused on the POM reaction. While a comprehensive review of all POM studies on oxygen-MIEC membranes is beyond the scope of this chapter, a brief review of selected membranes with high fluxes and/or high stability is included in this section. Table 11.5 provides a summary of the oxygen fluxes observed for the membrane materials discussed under air:inert and air:POM gradients. Early work utilizing MIEC membranes in POM reactions sought to take advantage of the high oxygen fluxes exhibited by SCF membranes [39,89]. However, when these membranes were exposed to reaction environments at high temperature, they fractured quickly. Changing the composition to the mixed-phase SFC resulted in membranes that were more stable in reducing environments. However, the oxygen flux of the SFC material is significantly less than that of the perovskite SCF, and SFC is thus not a viable candidate for POM applications [39,50,51,55,116]. Other studies with La0.6Sr0.4Co0.8Fe0.2O3d [35], La0.2Ba0.8Fe0.8Co0.2O3d [42], and Ba0.5Sr0.5Co0.8Fe0.2O3d [43,177] have found that doping higher valence metal ions such as La2þ and Ba2þ into the A site of SCF perovskites can achieve oxygen fluxes in air:inert gradients that are similar to the SCF membranes, but with significantly improved stability under reaction conditions. For example, Ba0.5Sr0.5Co0.8Fe0.2O3d maintained an oxygen flux of 7.8  10 2 mol m 2 s 1 at 850
C for over 1000 h in an air:helium environment and achieved > 97% methane conversion at 875
C for over 500 h [43]. The suggested reaction mechanism for this work was combustion of a portion of the CH4 to CO2 and H2O by membrane oxygen followed by CO2 and steam reforming of the remaining CH4 to form synthesis gas.

Chapter

11

259

Review of MIEC Ceramic Membranes

TABLE 11.5 Oxygen Flux Results Under Reaction Conditions for Various Membrane Materials

Membrane Material

Oxygen Flux Air: Inert Gradient (mol s 1 m 2)

Oxygen Flux Air:POM Gradient (mol s 1 m 2)

Thickness (mm)

SrCo0.8Fe0.2O3d [25,39,88]

7.4  10 4–6.7  10 3 Fractured quickly [25,39]

1

SrFeCo0.5O3d [39,50,51,55,116]

7.4  10 6–9.7  10 4 2.2  10 3 [39]

1

La0.6Sr0.4Co0.8Fe0.2O3d [35]

7.4  10 4 (900
C)

7.4  10 4

1

La0.2Ba0.8Fe0.8Co0.2O3d [42]

5.9  10 3

3.0  10 2

1

Ba0.5Sr0.5Co0.8Fe0.2O3d [43,177]

7.8  10 3–1.0  10 2 7.1  10 2 [43]

2.2  10 2 [39]

1.8

5.8  10 2 [177]

La0.8Sr0.2Co0.1Fe0.8Cr0.1O3d [5] Not reported

9.9  10 2

Not reported

BaCo0.4Fe0.4Zr0.2O3d [25]

4.1  10 3

3.8  10 2

1

BaCo0:7Fe0:2Nb0:1O3d [22]

Not reported

>0.14

1

BaCo0.7Fe0.2Ta0.1O3d [23] BaCe0.1Co0.4Fe0.5O3d [156]

2

1.4  10

2

< 1.0  10

0.11–0.12 2

6.5  10

0.6–0.7 1

To convert to molar fluxes, all volumetric fluxes reported in the literature were assumed to be at T ¼ 25
C and P ¼ 1 atm unless otherwise stated in the reference.

More recent investigations have focused on doping higher valance metal ions into both the A and B sites. The majority of the studies have used La2þ and Ba2þ in the A site with Cr7þ, Nb5þ, Ta5þ, and Zr4þ in the B site. La0.8Sr0.2Co0.1Fe0.8Cr0.1O3d membranes exhibited oxygen fluxes of 9.9  10 2 mol m 2 s 1 in an air:syngas gradient and were stable for the POM reaction at 900
C for 340 h [5]. The addition of Zr4þ was observed to increase stability further, with a BaCo0.4Fe0.4Zr0.2O3d membrane operating for over 2200 h with 96–98% methane conversion and 98–99% selectivity of carbon to CO. The flux during reaction was reported to be 3.7  10 2–4.0  10 2 mol m 2 s 1 at 850
C [25]. Substituting Nb5þ [22] and Ta5þ [23] in the B site resulted in membranes that were stable under POM reaction conditions for 300 h and 400 h, respectively, with unusually high oxygen fluxes. At 900
C, the BaCo0.7Fe0.2Ta0.1O3d membrane achieved > 99% CH4 conversion with 94% selectivity to CO and a flux of 0.11 mol m 2 s 1 during the reaction, while fluxes > 0.14 mol m 2 s 1 were

260

Inorganic, Polymeric and Composite Membranes

observed under reaction conditions for the Nb-doped membrane. These fluxes are some of the highest reported in the literature to date.

Effect of Reaction Temperature on Membrane Performance Reactor temperature is a key factor for syngas production with oxygen-MIEC membranes because both catalyst activity and membrane oxygen flux are strongly dependent on temperature. Differences in reaction conditions among studies prevent a direct comparison of the effect of reaction temperature on oxygen flux for specific membrane compositions. However, general correlations between oxygen flux, methane conversion, and temperature are evident for the majority of the studies. Figure 11.5 depicts the effect of temperature on methane conversion and oxygen flux for the POM reaction using a nickel-based catalyst and a BaCo0.7Fe0.2Ta0.1O3d membrane [23]. Increasing the reaction temperature resulted in a significant increase in both oxygen flux and methane conversion, with conversion ultimately leveling off near 90% at 950
C. These same trends of increasing oxygen flux and methane conversion with increasing temperature were observed with other catalysts and membranes. Examples include the use of a LiLaNiOx/g-Al2O3 catalyst in conjunction with BaCo0.4Fe0.4Zr0.2O3d [25] and BaCe0.1Co0.4Fe0.5O3d [156] membranes. Tong et al. ascribed the increased oxygen flux to an enhancement of both the diffusion rate of lattice oxygen vacancies in the bulk membrane material and the surface exchange rate. One interesting point in these studies is the trend in CO selectivity. As the temperature, flux, and conversion increase, product selectivity to CO decreases.

O2 flux (ml min-1 cm-2)

19

Oxygen flux Methane conversion

90

17 15

80

13

70

11 60

9 7 750

800

850 900 Temperature (°C)

950

CH4 conversion (%)

100

21

50 1000

FIGURE 11.5 O2 flux and CH4 conversion as a function of temperature for BaCo0.7Fe0.2Ta0.1O3d during partial oxidation of methane. Experimental conditions: air: 200 ml min 1; 48.0% helium diluted methane: 50 ml min 1; and membrane thickness: 0.7 mm [23]. (To convert to molar fluxes, all volumetric fluxes were assumed to be at T ¼ 25
C and P ¼ 1 atm.)

Chapter

11

Review of MIEC Ceramic Membranes

261

This drop in CO selectivity is due to the ratio of methane and oxygen in the reactor. When CH4 is in excess (i.e., at relatively low temperatures), any increase in oxygen flux results in a direct increase in methane conversion and CO selectivity remains high. As the temperature is increased and additional membrane oxygen becomes available, the methane excess diminishes, which can lead to deeper oxidation of the reaction products and a reduction in CO selectivity. The exact temperature at which CO selectivity begins to decline and the rate at which CO selectivity decreases depend on the membrane material and also on other factors such as the catalyst used and the composition of the reactor feed.

Effect of Reaction Environment on Membrane Oxygen Flux In general, regardless of membrane composition, oxygen fluxes for MIEC membranes under POM reaction conditions are higher than fluxes observed under an air:inert gradient. The increase in oxygen flux during the POM reaction is attributed to the rapid consumption of oxygen on the reaction side of the membrane, which leads to a vanishingly low oxygen partial pressure in the gasphase environment and a low oxygen potential in the membrane surface exposed to the reaction environment. The lower oxygen potential in the reaction surface than in the low oxygen surface under an inert atmosphere produces a larger oxygen potential gradient across the membrane, which ultimately leads to higher flux. CH4, CO, and H2 are the most prominent reducing agents in the synthesis gas production process. Studies have been performed in different gas environments to investigate the effect of a reducing gas on membrane oxygen flux. The order of increase in oxygen flux in the presence of different reducing gases was reported to be H2 > CO > CH4 [154] and this order has been suggested to correlate with the tendency of the gases to react with surface oxygen ions. Previous studies have shown that carbon monoxide and hydrogen readily react with lattice oxygen from the membrane to produce CO2 and water while CH4, even at high temperatures, exhibits very little activity with the membrane surface in the absence of a reforming catalyst [153]. These studies have also shown that the Ba0.5Sr0.5Co0.8Fe0.2O3d membrane material reacts more readily with hydrogen than with CO, so the surface reaction results correlate with the oxygen flux increase results. It is interesting to note that some researchers have seen a shift in the limiting factor for oxygen transport by changing the permeate side environment. By varying membrane thickness, Luo et al. [23] found that oxygen flux through BaCo0.7Fe0.2Ta0.1O3d membranes with thicknesses between 0.6 and 1.2 mm was limited by bulk diffusion in an air:helium gradient at 900
C. However, when membrane thickness was varied with an air:POM gradient, no significant change in oxygen flux was observed. This observation suggests that the BaCo0.7Fe0.2Ta0.1O3d membrane switched from a bulk diffusion-limited mode

262

Inorganic, Polymeric and Composite Membranes

to a surface exchange-limited mode when the permeate side environment was switched from inert gas to POM/synthesis gas. The effect of changing the permeate side environment during the POM reaction by varying the flow rate of a pure methane feed has been investigated using a BaCe0.1Co0.4Fe0.5O3d membrane. Figure 11.6 shows the effect of feed flow rate on oxygen flux, methane conversion, and CO selectivity, where a higher methane feed rate is assumed to correspond to a higher methane concentration in the reactor. Similar to the temperature effect results discussed in the earlier section, oxygen flux increases significantly and then plateaus at approximately 8.2  10 2 mol m 2 s 1 as methane flow rate continues to increase. However, unlike in the temperature results, as the methane flow rate is increased, methane conversion decreases while CO selectivity increases. The decrease in methane conversion and corresponding increase in selectivity have been ascribed to the change in the methane:oxygen ratio in the reactor [25,156]. When the methane flow rate is low, the amount of available oxygen is sufficient to react with almost all of the methane. However, a decrease in the methane:oxygen ratio favors deeper oxidation of methane and thus produces lower CO selectivity. As the feed flow rate of CH4 is increased, so does the methane: oxygen ratio in the reactor. Eventually, the reaction becomes limited by the oxygen supplied through the membrane. While oxygen flux and CO selectivity remain constant as the flow rate is increased further, methane conversion continues to decrease. These trends have also been observed with other membrane materials and with diluted methane feeds, suggesting that a general relationship between reactant feed and oxygen flux, hydrocarbon conversion, and reaction selectivity may exist regardless of the membrane material. 120

20

100

16 14

80

12 60

10 8

40

6

Oxygen flux Conversion CO selectivity

4 2 0 5

25 10 15 20 Methane flow rate (mI min-1)

20

Conversion/selectivity (%)

O2 flux (mI min-1 cm-2)

18

0 30

FIGURE 11.6 O2 flux, CH4 conversion, and CO selectivity as a function of CH4 flow rate for BaCe0.1Co0.4Fe0.5O3d during partial oxidation of methane. Experimental conditions: air flow rate: 250 ml min 1; thickness of the membrane: 1.0 mm; temperature: 875
C [156]. (To convert to molar fluxes, all volumetric fluxes were assumed to be at T ¼ 25
C and P ¼ 1 atm.)

Chapter

11

Review of MIEC Ceramic Membranes

263

The effect of varying the air flow rate on the oxygen source side has also been investigated. Studies showed that at low air flow rates, oxygen flux increases with increasing air flow rate. These results indicate that oxygen transport from the gas phase to the membrane surface is a rate-controlling factor when air flow rate is low. However, as the air flow rate continues to increase, oxygen flux again reaches a plateau. The flow rate at which flux levels off depends on the membrane material, but in general was found to be in the range of 200–250 ml min 1 [23,156]. Methane conversion and CO selectivity were also found to be affected by air flow rate. However, the change in these values was on the order of 1–2% whereas a much larger change was observed when methane flow rate was varied, as discussed earlier. Thus, air flow rate has a much smaller impact on oxygen flux, methane conversion, and selectivity than the composition of the reactor contents. An isothermal membrane exposed simultaneously to two different oxygen partial pressures experiences nonuniform chemical expansion (or contraction) to a degree that depends on the material phases present and the ratio of the oxygen partial pressures [54,59]. The oxygen potential gradient across a partial oxidation reactor membrane is much more extreme than the gradient during a typical oxygen flux test. In a POM membrane reactor, the membrane’s permeate side can be exposed to an oxygen partial pressure of between 10 17 and 10 30 atm at 850
C [21,24,40], compared to a permeate side oxygen partial pressure of 10 3–10 5 atm in an oxygen flux test under an air:inert gradient [24,55,88]. Considering the much larger chemical expansion gradient within a membrane in a POM reactor, among other differences, it is to be expected that fewer viable membranes have been proposed for membrane reactor than for oxygen separation applications. The very properties that give some materials high oxygen diffusivity can make them unfit for use in methane conversion applications [6].

CONCLUSIONS Oxygen-MIEC membranes have been extensively studied for over three decades and significant advances have been, and continue to be, made. Reported oxygen fluxes are approaching levels that have commercial application, and the new oxygen-MIEC materials being tested have also been reported to exhibit greatly improved mechanical stability compared with early membrane material candidates. However, substantial work is still needed to develop and thoroughly evaluate these more robust, higher flux membrane materials. The largest practical challenge facing designers of oxygen-conducting ceramic membranes is clearly the need to simultaneously maximize oxygen conduction—which is expected to correlate positively with bulk oxygen absorption/desorption capacity and the resulting lattice contractions/expansions—and membrane mechanical stability, which is expected to correlate negatively with these characteristics. The commercial potential of new membrane materials,

264

Inorganic, Polymeric and Composite Membranes

based on both flux and mechanical stability, can only be evaluated properly under reaction conditions. Membrane stability must also be assessed under relevant industrial constraints, both economic and practical. Catalyst and membrane systems must be inexpensive to fabricate and must demonstrate long-term, low maintenance operation. Also, test reactor feed composition must represent actual commercial supplies. For example, a majority of the reaction studies reported in the literature use diluted methane as a reactor feed which is not consistent with industrial operations. Industrial hydrocarbon feedstocks will also contain impurities such as CO2 that can have a significant impact on oxygen flux, which means that the true commercial viability of new MIEC membrane reactor candidate materials must be evaluated in the presence of these impurities. Finally, reactor system design and engineering will become even more important as membranes become thinner in the quest for increased oxygen flux. Not only does a reduction in membrane thickness have an impact on the mechanical strength and possible configuration of a membrane reactor, but it may also amplify the effect of any catalyst–membrane interactions, as thinner membranes bring a greater proportion of the total membrane material in contact with the catalyst. Understanding and predicting long-term catalyst–membrane interactions is a critical research area with any membrane material or configuration, but thinner membranes make it even more urgent. Future work will benefit from a combined knowledge of material science, reaction engineering, and catalyst design.

ACKNOWLEDGMENTS Financial support for this work was provided by the Office of Naval Research (N00014-03-10601) and the US Department of Transportation Research Innovative Technology Administration (DTOS59-06-G-0047).

REFERENCES [1] M. Freemantle, Membranes for gas separations, Chem. Eng. News 83 (2005) 9. [2] S. Gopalan, Using ceramic mixed ionic and electronic conductors for gas separation, J. Miner. Met. Mater. Soc. 54 (2002) 26–29. [3] A.J. Jacobson, S. Kim, A. Medina, Y.L. Yang, B. Abeles, Dense oxide membranes for oxygen separation and methane conversion, Mater. Res. Soc. Symp. Proc. 497 (1998) 29–34. [4] A. Leo, S.M. Liu, J.C.D. da Costa, Development of mixed conducting membranes for clean coal energy delivery, Int. J. Greenhouse Gas Control 3 (2009) 357–367. [5] H.J. Bouwmeester, Dense ceramic membranes for methane conversion, Catal. Today 82 (2003) 141–150. [6] E.T. Robinson, S. Aasland, J. Chen, The application of oxygen transport membranes for syngas or hydrogen production, in: 2004 AIChE Spring Meeting, 2004, p. 68d. [7] D.A. Slade, Mixed Ionic/Electronic Conducting Ceramic Membranes for Oxygen-Assisted CO2 Reforming, Department of Chemical and Petroleum Engineering, University of Kansas, Lawrence, KS, 2010, p. 332.

Chapter

11

Review of MIEC Ceramic Membranes

265

[8] J.C. Boivin, C. Pirovano, G. Nowogerocki, G. Mairesse, P. Labrune, G. Lagerange, Electrode–electrolyte BIMEVOX system for moderate temperature oxygen separation, Solid State Ionics 113 (1998) 639–651. [9] J. Tong, W. Yang, B. Zhu, R. Cai, Investigation of ideal zirconium-doped perovskite-type ceramic membrane materials for oxygen separation, J. Membr. Sci. 203 (2002) 175–189. [10] P.N. Dyer, R.E. Richards, S.L. Russek, D.M. Taylor, Ion transport membrane technology for oxygen separation and syngas production, Solid State Ionics 134 (2000) 21–33. [11] B.C.H. Steele, Oxygen ion conductors and their technological applications, Mater. Sci. Eng. B13 (1992) 79–87. [12] E. Politova, G. Suchaneck, Physicochemical problems encountered in the development of oxygen-conducting oxide materials for new devices for chemical energy conversion, Russ. Chem. Bull. 57 (2008) 1126–1137. [13] W. Zhu, W. Han, G. Xiong, W. Yang, Mixed reforming of heptane to syngas in the Ba0.5Sr0.5Co0.8Fe0.2O3 membrane reactor, Catal. Today 104 (2005) 5. [14] R.L. Cook, R.C. MacDuff, A.F. Sammells, Perovskite solid electrolytes for intermediate temperature solid oxide fuel cells, J. Electrochem. Soc. 137 (1990) 3309–3310. [15] R.L. Cook, A.F. Sammells, On the systematic selection of perovskite solid electrolytes for intermediate temperature fuel cells, Solid State Ionics 45 (1991) 311–321. [16] V.V. Kharton, E.V. Tsipis, I.P. Marozau, A.P. Viskup, J.R. Frade, J.T.S. Irvine, Mixed conductivity and electrochemical behavior of (La0.75Sr0.25)0.95Cr0.5Mn0.5O3d, Solid State Ionics 178 (2007) 101–113. [17] S. Xie, W. Liu, K. Wu, P.H. Yang, G.Y. Meng, C.S. Chen, Mixed oxygen ionic and electronic conduction in CaFe0.2Ti0.8O3d: a combined oxygen permeation and electrical conductivity study, Solid State Ionics 118 (1999) 23–28. [18] Z. Yang, A.S. Harveya, L.J. Gauckler, Influence of CO2 on Ba0.2Sr0.8Co0.8Fe0.2O3d at elevated temperatures, Scr. Mater. 61 (2009) 4. [19] W. Zhou, R. Rana, Z. Shao, Progress in understanding and development of Ba0.5Sr0.5Co0.8Fe0.2O3d-based cathodes for intermediate-temperature solid-oxide fuel cells: a review, J. Power Sour. 192 (2009) 231–246. [20] X. Zhou, M. Zhu, F. Deng, G. Meng, X. Liu, Electrical properties, sintering and thermal expansion behavior of composite ceramic interconnecting materials, La0.7Ca0.3CrO3d/ Y0.2Ce0.8O1.9 for SOFCs, Acta Mater. 55 (2007) 2113–2118. [21] S.J. Feng, S. Ran, D.C. Zhu, W. Liu, C.S. Chen, Synthesis gas production from methane with SrFeCo0.5Oy membrane reactor, Energy Fuels 18 (2004) 385–389. [22] M. Harada, K. Domen, M. Hara, T. Takashi, Ba1.0Co0.7Fe0.2Nb0.1O3d dense ceramic as an oxygen permeable membrane for partial oxidation of methane to synthesis gas, Chem. Lett. 35 (2006) 1326–1327. [23] H. Luo, Y. Wei, H. Jiang, W. Yuan, Y. Lv, J. Caro, H. Wang, Performance of a ceramic membrane reactor with high oxygen flux Ta-containing perovskite for the partial oxidation of methane to syngas, J. Membr. Sci. 350 (2010) 154–160. [24] J.T. Ritchie, J.T. Richardson, D. Luss, Ceramic membrane reactor for synthesis gas production, AIChE J. 47 (2001) 2092–2101. [25] J. Tong, W. Yang, R. Cai, B. Zhu, L. Lin, Novel and ideal zirconium-based dense membrane reactors for partial oxidation of methane to syngas, Catal. Lett. 78 (2002) 129–137. [26] J. Tong, W. Yang, H. Suda, K. Haraya, Initiation of oxygen permeation and POM reaction in different mixed conducting ceramic membrane reactors, Catal. Today 118 (2006) 144–150. [27] J.D. Holladay, J. Hu, D.L. King, Y. Wang, An overview of hydrogen production technologies, Catal. Today 139 (2009) 244–260.

266

Inorganic, Polymeric and Composite Membranes

[28] W. Zhu, W. Han, G. Xiong, W. Yang, Mixed reforming of simulated gasoline to hydrogen in a BSCFO membrane reactor, Catal. Today 118 (2006) 39–43. [29] H. Wang, Y. Cong, W. Yang, Partial oxidation of ethane to syngas in an oxygen-permeable membrane reactor, J. Membr. Sci. 209 (2002) 143–152. [30] H.H. Wang, Y. Cong, W.S. Yang, Continuous oxygen ion transfer medium as a catalyst for high selective oxidative dehydrogenation of ethane, Catal. Lett. 84 (2002) 101–106. [31] W. Yang, H. Wang, X. Zhu, L. Lin, Development and application of oxygen permeable membrane in selective oxidation of light alkanes, Top. Catal. 35 (2005) 155–167. [32] M. Rebeilleau-Dassonneville, S. Rosini, A.C. van Veen, D. Farrusseng, C. Mirodatos, Oxidative activation of ethane on catalytic modified dense ionic oxygen conducting membranes, Catal. Today 104 (2005) 131–137. [33] L. Oliviera, S. Haaga, C. Mirodatosa, A.C. van Veen, Oxidative coupling of methane using catalyst modified dense perovskite membrane reactors, Catal. Today 142 (2009) 8. [34] Y. Zeng, Y.S. Lin, Oxidative coupling of methane on improved bismuth oxide membrane reactors, AIChE J. 47 (2001) 436–444. [35] J.E. ten Elshof, H.J. Bouwmeester, H. Verweij, Oxidative coupling of methane in a mixed-conducting perovskite membrane reactor, Appl. Catal. A 130 (1995) 195–212. [36] F.T. Akin, Y.S. Lin, Y. Zeng, Comparative study on oxygen permeation and oxidative coupling of methane on disk-shaped and tubular dense ceramic membrane reactors, Ind. Eng. Chem. Res. 40 (2001) 5908–5916. [37] Y.S. Lin, Y. Zeng, Catalytic properties of oxygen semipermeable perovskite-type ceramic membrane materials for oxidative coupling of methane, J. Catal. 164 (1996) 220–231. [38] K. Omata, O. Yamazaki, K. Tomita, K. Fujimoto, Oxidative coupling of methane on an ABO3 type oxide with mixed conductivity, J. Chem. Soc. Chem. Commun. 14 (1994) 1647–1648. [39] U. Balachandran, J.T. Dusek, P.S. Maiya, B. Ma, R.L. Mieville, M.S. Kleefisch, et al., Ceramic membrane reactor for converting methane to syngas, Catal. Today 36 (1997) 265–272. [40] P.S. Maiya, U. Balachandran, J.T. Dusek, R.L. Mieville, M.S. Kleefisch, C.A. Udovich, Oxygen transport by oxygen potential gradient in dense ceramic oxide membranes, Solid State Ionics 99 (1997) 1–7. [41] A.F. Sammells, M. Schwartz, R.A. Mackay, T.F. Barton, D.R. Peterson, Catalytic membrane reactors for spontaneous synthesis gas production, Catal. Today 56 (2000) 325–328. [42] C.-Y. Tsai, A.G. Dixon, Y.H. Ma, W.R. Moser, M.R. Pascucci, Dense perovskite La1  xAx0 Fe1  yCoyO3  d (A0 ¼ Ba, Sr, Ca) membrane synthesis, applications, and characterization, J. Am. Ceram. Soc. 81 (1998) 1437–1444. [43] H. Dong, Z. Shao, G. Xiong, J. Tong, S. Sheng, W. Yang, Investigation on POM reaction in a new perovskite membrane reactor, Catal. Today 67 (2001) 3–13. [44] J. Dong, W. Yang, Methods for selection perovskite-type oxides used as mixed-conducting membranes for oxygen separation, Membr. Sci. Technol. 23 (2003) 33–42. [45] J. Sunarso, S. Baumann, J.M. Serra, W.A. Meulenberg, S. Liu, Y.S. Lin, J.C. Diniz da Costa, Mixed ionic-electronic conducting (MIEC) ceramic-based membranes for oxygen separation, J. Membr. Sci. 320 (2008) 13–41. [46] Y. Teraoka, H.-M. Zhang, S. Furukawa, N. Yamazoe, Oxygen permeation through perovskitetype oxides, Chem. Lett. (1985) 1743–1746. [47] F.J. Berry, J.F. Marco, X. Ren, Reduction properties of phases in the system La0.5Sr0.5MO (M ¼ Fe, Co), Solid State Ionics 178 (2005) 961–969. [48] B. Ma, U. Balachandran, Phase stability of SrFeCo0.5Ox in reducing environments, Mater. Res. Bull. 33 (1998) 223–236.

Chapter

11

Review of MIEC Ceramic Membranes

267

[49] S.J. Xu, W.J. Thomson, Stability of La0.6Sr0.4Co0.2Fe0.8O3d perovskite membranes in reducing and nonreducing environments, Ind. Eng. Chem. Res. 37 (1998) 1290–1299. [50] T. Armstrong, F. Prado, Y. Xia, A. Manthiram, Role of perovskite phase on the oxygen permeation properties of the Sr4Fe6xCoxO13þd system, J. Electrochem. Soc. 147 (2000) 435–438. [51] Y. Xia, T. Armstrong, F. Prado, A. Manthiram, Sol-gel synthesis, phase relationships, and oxygen permeation properties of Sr4Fe6xCoxO13þd (0  x  3), Solid State Ionics 130 (2000) 81–90. [52] A. Fossdal, L.T. Sagdahl, M.-A. Einarsrud, K. Wiik, T. Grande, P.H. Larsen, et al., Phase equilibria and microstructure in Sr4Fe6xCoxO13 (0 x  4) mixed conductors, Solid State Ionics 143 (2001) 367–377. [53] B. Ma, N.I. Victory, U. Balachandran, B.J. Mitchell, J.W. Richardson, Study of the mixedconducting SrFeCo0.5Oy system, J. Am. Ceram. Soc. 85 (2002) 2641–2645. [54] S. Pei, M.S. Kleefisch, T.P. Kobylinski, J. Faber, C.A. Udovich, V. Zhang-McCoy, et al., Failure mechanisms of ceramic membrane reactors in partial oxidation of methane to synthesis gas, Catal. Lett. 30 (1995) 201. [55] S. Kim, Y.L. Yang, R. Christoffersen, A.J. Jacobson, Determination of oxygen permeation kinetics in a ceramic membrane based on the composition SrFeCo0.5O3.25d, Solid State Ionics 109 (1998) 187–196. [56] S.B. Adler, Chemical expansivity of electrochemical ceramics, J. Am. Ceram. Soc. 84 (2001) 2117–2119. [57] J.P. Hodges, J.D. Jorgensen, D.J. Miller, B. Ma, U. Balachandran, J.W. Richardson, Crystal structures of mixed-conducting oxides present in the Sr–Fe–Co–O system, Mater. Res. Soc. Symp. Proc. 496 (1998) 173–178. [58] H. Ullmann, N. Trofimenko, F. Tietz, D. Stoever, A. Ahmad-Khanlou, Correlation between thermal expansion and oxide ion transport in mixed conducting perovskite-type oxides for SOFC cathodes, Solid State Ionics 138 (2000) 79–90. [59] A. Atkinson, T.M.G.M. Ramos, Chemically-induced stresses in ceramic oxygen ion-conducting membranes, Solid State Ionics 129 (2000) 259–269. [60] M. Arnold, J. Martynczuk, K. Efimov, H. Wang, A. Feldhoff, Grain boundaries as barrier for oxygen transport in perovskite-type membranes, J. Membr. Sci. 316 (2008) 137–144. [61] S. Diethelm, J. Van Herle, J. Sfeir, P. Buffat, Correlation between oxygen transport properties and microstructure in La0.5Sr0.5FeO3d, J. Eur. Ceram. Soc. 25 (2005) 2191–2196. [62] V.V. Kharton, F.M.B. Marques, Mixed ionic–electronic conductors: effects of ceramic microstructure on transport properties, Curr. Opin. Solid State Mater. Sci. 6 (2002) 261–269. [63] J. Maier, Transport in electroceramics: micro- and nano-structural aspects, J. Eur. Ceram. Soc. 24 (2004) 1251–1257. [64] J. Martynczuk, M. Arnold, A. Feldhoff, Influence of grain size on the oxygen permeation performance of perovskite-type (Ba0.5Sr0.5)(Fe0.8Zn0.2)O3d membranes, J. Membr. Sci. 322 (2008) 375–382. [65] A.L. Shaula, R.O. Fuentes, F.M. Figueiredo, V.V. Kharton, F.M.B. Marques, J.R. Frade, Grain size effects on oxygen permeation in submicrometric CaTi0.8Fe0.2O3d ceramics obtained by mechanical activation, J. Eur. Ceram. Soc. 25 (2005) 2613–2616. [66] H. Wang, C. Tablet, A. Feldhoff, J. Caro, Investigation of phase structure, sintering, and permeability of perovskite-type Ba0.5Sr0.5Co0.8Fe0.2O3d membranes, J. Membr. Sci. 262 (2005) 20–26. [67] M. Mogensen, D. Lybye, N. Bonanos, P.V. Hendriksen, F.W. Poulsen, Factors controlling the oxide ion conductivity of fluorite and perovskite structured oxides, Solid State Ionics 174 (2004) 279–286.

268

Inorganic, Polymeric and Composite Membranes

[68] W.D. Kingery, J. Pappis, M.E. Doty, D.C. Hill, Oxygen ion mobility in cubic Zr0.85Ca0.15O1.85, J. Am. Ceram. Soc. 42 (1959) 393–398. [69] D. Wang, Y. Guo, K. Liang, K. Tao, Crystal structure of zirconia by Rietveld refinement, Sci. China Ser. A Math. 42 (1999) 80–86. [70] S. Dou, C.R. Masson, P.D. Pacey, Mechanism of oxygen permeation through lime-stabilized zirconia, J. Electrochem. Soc. 132 (1985) 1843–1849. [71] H. Arashi, H. Naito, Oxygen permeability in ZrO2–TiO2–Y2O3 system, Solid State Ionics 53–56 (1992) 431–435. [72] Y. Nigara, Y. Kosaka, K. Kawamura, J. Mizusaki, M. Ishigame, Oxygen permeability in ZrO2–CeO2–MgO at high temperatures, Solid State Ionics 86 (1996) 739–744. [73] H. Cerva, W. Russwurm, Microstructure and crystal structure of bismuth oxide phases in zinc oxide varistor ceramics, J. Am. Ceram. Soc. 71 (1988) 522–530. [74] H.J.M. Bouwmeester, H. Kruidhof, A.J. Burggraaf, P.J. Gellings, Oxygen semipermeability of erbia-stabilized bismuth oxide, Solid State Ionics 53–56 (1992) 460–468. [75] E. Capoen, M.C. Steil, G. Nowogrocki, M. Malys, C. Pirovano, A. Lo¨fberg, et al., Oxygen permeation in bismuth-based materials. Part I: Sintering and oxygen permeation fluxes, Solid State Ionics 177 (2006) 483–488. [76] V. Zhukovskii, E. Buyanova, Y. Emel’yanova, M. Morozova, R. Shafigina, R. Zakharov, et al., Synthesis, structure, and conductivity of BIMEVOX oxide ceramics, Russ. J. Electrochem. 45 (2009) 512–519. [77] C. Pirovano, A. Lo¨fberg, H. Bodet, E. Bordes-Richard, M.C. Steil, R.N. Vannier, BIMEVOX as dense membrane in catalytic reactor (ME ¼ Co, Cu, Ta), Solid State Ionics 177 (2006) 2241–2244. [78] Y. Nigara, J. Mizusaki, M. Ishigame, Measurement of oxygen permeability in CEO2 doped CSZ, Solid State Ionics 79 (1995) 208–211. [79] B. Cales, J.E. Baumard, Mixed conduction and defect structure of ZrO2–Y2O3–CeO2 solid solution, J. Electrochem. Soc. 131 (1984) 2407–2413. [80] G.Z. Cao, Electrical conductivity and oxygen semipermeability of terbia and yttria stabilized zirconia, J. Appl. Electrochem. 24 (1994) 1222–1227. [81] J. Han, Y. Zeng, Y.S. Lin, Oxygen permeation through fluorite-type bismuth–yttrium–copper oxide membranes, J. Membr. Sci. 132 (1997) 235–243. [82] D.P. Fagg, A.L. Shaula, V.V. Kharton, J.R. Frade, High oxygen permeability in fluorite-type Ce0.8Pr0.2O2d via the use of sintering aids, J. Membr. Sci. 299 (2007) 1–7. [83] Y. Zeng, Y.S. Lin, Oxygen permeation and oxidative coupling of methane in yttria doped bismuth oxide membrane reactor, J. Catal. 193 (2000) 58–64. [84] J.H. Han, G. Xomeritakis, Y.S. Lin, Oxygen permeation through thin zirconia/yttria membranes prepared by EVD, Solid State Ionics 93 (1997) 263–272. [85] O. Muller, R. Ahmadi, The Major Ternary Structural Families, Springer, Berlin, 1974. [86] V. Goldschmidt, Geochemische Verteilungsgesetze der Elemente, Norske Videnskap, Oslo, 1927. [87] Y.H. Ma, Dense palladium and perovskite membranes and membrane reactors, MRS Bull. 24 (1999) 46–49. [88] L. Qiu, T.H. Lee, L.-M. Liu, Y.L. Yang, A.J. Jacobson, Oxygen permeation studies of SrCo0.8Fe0.2O3d, Solid State Ionics 76 (1995) 321–329. [89] J. Sunarso, S. Liu, Y.S. Lin, J.C.D. da Costa, Oxygen permeation performance of BaBiO3d ceramic membranes, J. Membr. Sci. 344 (2009) 281–287. [90] K. Brinkman, T. Iijima, H. Takamura, The oxygen permeation characteristics of Bi1xSrxFeO3 mixed ionic and electronic conducting ceramics, Solid State Ionics 181 (2010) 53–58.

Chapter

11

Review of MIEC Ceramic Membranes

269

[91] X. Zhu, H. Wang, W. Yang, Novel cobalt-free oxygen permeable membrane, Chem. Commun. (2004) 1130–1131. [92] Z. Shao, W. Yang, Y. Cong, H. Dong, J. Tong, G. Xiong, Investigation of the permeation behavior and stability of a Ba0.5Sr0.5Co0.8Fe0.2O3d oxygen membrane, J. Membr. Sci. 172 (2000) 177–188. [93] H. Wang, C. Tablet, A. Feldhoff, J. Caro, A cobalt-free oxygen-permeable membrane based on the perovskite-type oxide Ba0.5Sr0.5Zn0.2Fe0.8O3d, Adv. Mater. 17 (2005) 1785–1788. [94] Y. Teraoka, T. Nobunaga, N. Yamazoe, Effect of cation substitution on the oxygen semipermeability of perovskite-type oxides, Chem. Lett. (1988) 503–506. [95] K. Li, X. Tan, Y. Liu, Single-step fabrication of ceramic hollow fibers for oxygen permeation, J. Membr. Sci. 272 (2006) 1–5. [96] T. Ishihara, T. Yamada, H. Arikawa, H. Nishiguchi, Y. Takita, Mixed electronic-oxide ionic conductivity and oxygen permeating property of Fe-, Co- or Ni-doped LaGaO3 perovskite oxide, Solid State Ionics 135 (2000) 631–636. [97] P. Zeng, R. Ran, Z. Chen, H. Gu, Z. Shao, Novel mixed conducting SrSc0.05Co0.95O3d ceramic membrane for oxygen separation, AIChE J. 53 (2007) 3116–3124. [98] G. Etchegoyen, T. Chartier, A. Julian, P. Del-Gallo, Microstructure and oxygen permeability of a La0.6Sr0.4Fe0.9Ga0.1O3d membrane containing magnesia as dispersed second phase particles, J. Membr. Sci. 268 (2006) 86–95. [99] K. Zhang, R. Ran, L. Ge, Z. Shao, W. Jin, N. Xu, Systematic investigation on new SrCo1yNbyO3d ceramic membranes with high oxygen semi-permeability, J. Membr. Sci. 323 (2008) 436–443. [100] X. Dong, H. Zhang, W. Lin, Preparation and characterization of a perovskite-type mixed conducting SrFe0.6Cu0.3Ti0.1O3d membrane for partial oxidation of methane to syngas, Chin. J. Chem. Eng. 16 (2008) 411–415. [101] H. Kruidhof, H.J.M. Bouwmeester, R.H.E. Vondoorn, A.J. Burggraaf, Influence of orderdisorder transitions on oxygen permeability through selected nonstoichiometric perovskitetype oxides, Sol. St. Ionics 63 (1993) 816–822. [102] Z.T. Wu, W.Q. Jin, N.P. Xu, Oxygen permeability and stability of Al2O3-doped SrCo0.8Fe0.2O3d mixed conducting oxides, J. Membr. Sci. 279 (2006) 320–327. [103] L. Tan, L. Yang, X.H. Gu, W.Q. Jin, L.X. Zhang, N.P. Xu, Structure and oxygen permeability of Ag-doped SrCo0.8Fe0.2O3d oxides, AIChE J. 50 (2004) 701–707. [104] C.G. Fan, Z.Q. Deng, Y.B. Zuo, W. Liu, C.S. Chen, Preparation and characterization of SrCo0.8Fe0.2O3d–SrSnO3 oxygen-permeable composite membrane, Solid State Ionics 166 (2004) 339–342. [105] S. Diethelm, J. Van Herle, Oxygen transport through dense La0.6Sr0.4Fe0.8Co0.2O3d perovskite-type permeation membranes, J. Eur. Ceram. Soc. 24 (2004) 1319–1323. [106] S.G. Li, H. Qi, N.P. Xu, J. Shi, Tubular dense perovskite type membranes. Preparation, sealing, and oxygen permeation properties, Ind. Eng. Chem. Res. 38 (1999) 5028–5033. [107] T. Ishihara, Y. Tsuruta, T. Todaka, H. Nishiguchi, Y. Takita, Fe doped LaGaO3 perovskite oxide as an oxygen separating membrane for CH4 partial oxidation, Solid State Ionics 152 (2002) 709–714. [108] S. Koutcheiko, P. Whitfield, I. Davidson, Electrical and thermal properties of La0.7Sr0.3Ga0.6Fe0.4O3 ceramics, Ceram. Int. 32 (2006) 339–344. [109] J.F. Vente, S. McIntosh, W.G. Haije, H.J.M. Bouwmeester, Properties and performance of BaxSr1xCo0.8Fe0.2O3d materials for oxygen transport membranes, J. Solid State Electrochem. 10 (2006) 581–588.

270

Inorganic, Polymeric and Composite Membranes

[110] Y. Zhang, Q. Li, P. Shen, Y. Liu, Z. Yang, W. Ding, et al., Hydrogen amplification of coke oven gas by reforming of methane in a ceramic membrane reactor, Int. J. Hydrogen Energy 33 (2008) 3311–3319. [111] Q. Jiang, S.M. Williams, K.J. Nordheden, Oxygen Permeation Studies of Asymmetric Ba0.5Sr0.5Co0.8Fe0.2Ox Membranes, AIChE, Philadelphia, PA, 2008. [112] Q. Jiang, S.M. Williams, K.J. Nordheden, Oxygen Permeation Study and Improvement of Ba0.5Sr0.5Co0.8Fe0.2Ox Perovskite Membranes, J. Membrane Science, 369 (2011) 174–181. [113] U. Balachandran, J.T. Dusek, S.M. Sweeney, R.B. Poeppel, R.L. Mieville, P.S. Maiya, et al., Methane to syngas via ceramic membranes, Am. Ceram. Soc. Bull. 74 (1995) 71–75. [114] B. Ma, U. Balachandran, Oxygen nonstoichiometry in mixed-conducting SrFeCo0.5Ox, Solid State Ionics 100 (1997) 53–62. [115] S.M. Murphy, D.A. Slade, K.J. Nordheden, S.M. Stagg-Williams, Increasing oxygen flux through a dense oxygen permeable membrane by photolithographic patterning of platinum, J. Membr. Sci. 277 (2006) 94–98. [116] M. Ikeguchi, K. Yoshino, K. Kanie, M. Nomura, E. Kikuchi, M. Matsukata, Effects of preparation method on oxygen permeation properties of SrFeCo0.5Ox membrane, Sep. Purif. Technol. 32 (2003) 313–318. [117] B. Ma, U. Balachandran, J.-H. Park, C.U. Segre, Determination of chemical diffusion coefficient of SrFeCo0.5Ox by the conductivity relaxation method, Solid State Ionics 83 (1996) 65–71. [118] V.V. Kharton, A.V. Kovalevsky, A.P. Viskup, A.L. Shaula, F.M. Figueiredo, et al., Oxygen transport in Ce0.8Gd0.2O2d-based composite membranes, Solid State Ionics 160 (2003) 247–258. [119] V.V. Kharton, A.V. Kovalevsky, A.P. Viskup, A.A. Yaremchenko, E.N. Naumovich, F.M. B. Marques, Oxygen permeability of Ce0.8Gd0.2O2d–La0.7Sr0.3MnO3d composite membranes, J. Electrochem. Soc. 147 (2000) 2814–2821. [120] T.J. Mazanec, T.L. Cable, J.G. Frye, Electrocatalytic cells for chemical reaction, Solid State Ionics 53–56 (1992) 111–118. [121] A. Shaula, V. Kharton, F. Marques, A. Kovalevsky, A. Viskup, E. Naumovich, Oxygen permeability of mixed-conducting composite membranes: effects of phase interaction, J. Solid State Electrochem. 10 (2006) 28–40. [122] J. Sirman, J. Chen, Ceramic membrane structure and oxygen separation method, (2003) US Patent 6,514,314. [123] H. Wang, W.S. Yang, Y. Cong, X. Zhu, Y.S. Lin, Structure and oxygen permeability of a dual-phase membrane, J. Membr. Sci. 224 (2003) 107–115. [124] C. Chen, H. Kruidhof, H.J.M. Bouwmeester, H. Verweij, A.J. Burggraaf, Oxygen permeation through oxygen ion oxide-noble metal dual phase composites, Solid State Ionics 86–88 (1996) 569–572. [125] W. Li, J.-J. Liu, C.-S. Chen, Hollow fiber membrane of yttrium-stabilized zirconia and strontium-doped lanthanum manganite dual-phase composite for oxygen separation, J. Membr. Sci. 340 (2009) 266–271. [126] K. Kobayashi, T. Tsunoda, Oxygen permeation and electrical transport properties of 60 vol.% Bi1.6Y0.4O3 and 40 vol.% Ag composite prepared by the sol-gel method, Solid State Ionics 175 (2004) 405–408. [127] Y.-b. Zuo, J.-h. Li, Q. Zeng, W. Liu, C.-s. Chen, Estimation of oxygen ionic conductivity in nickel–zirconia composite by oxygen permeation method, J. Solid State Electrochem. 14 (2010) 945–949.

Chapter

11

Review of MIEC Ceramic Membranes

271

[128] A.A. Yaremchenko, V.V. Kharton, A.A. Valente, F.M.M. Snijkers, J.F.C. Cooymans, J.J. Luyten, et al., Performance of tubular SrFe(Al)O3d–SrAl2O4 composite membranes in CO2- and CH4-containing atmospheres, J. Membr. Sci. 319 (2008) 141–148. [129] J. Yi, Y. Zuo, W. Liu, L. Winnubst, C. Chen, Oxygen permeation through a Ce0.8Sm0.2O2[delta]-La0.8Sr0.2CrO3-[delta] dual-phase composite membrane, J. Membr. Sci. 280 (2006) 849–855. [130] Y. Guo, H. Shi, R. Ran, Z. Shao, Performance of SrSc0.2Co0.8O3-[delta] þ Sm0.5Sr0.5CoO3[delta] mixed-conducting composite electrodes for oxygen reduction at intermediate temperatures, International Journal of Hydrogen Energy 34 (2009) 9496–9504. [131] X. Zhu, W. Yang, Composite membrane based on ionic conductor and mixed conductor for oxygen permeation, AIChE J. 54 (2008) 665–672. [132] H. Kusaba, Y. Shibata, K. Sasaki, Y. Teraoka, Surface effect on oxygen permeation through dense membrane of mixed-conductive LSCF perovskite-type oxide, Solid State Ionics 177 (2006) 2249–2253. [133] W. Ito, T. Nagai, T. Sakon, Oxygen separation from compressed air using a mixed conducting perovskite-type oxide membrane, Solid State Ionics 178 (2007) 809–816. [134] N. Miura, Y. Okamoto, J. Tamaki, K. Morinaga, N. Yamazoe, Oxygen semipermeability of mixed-conductive oxide thick-film prepared by slip casting, Solid State Ionics 79 (1995) 195–200. [135] Y. Teraoka, Y. Honbe, J. Ishii, H. Furukawa, I. Moriguchi, Catalytic effects in oxygen permeation through mixed-conductive LSCF perovskite membranes, Solid State Ionics 152–153 (2002) 681–687. [136] X.Y. Tan, Z.G. Wang, H. Liu, S.M. Liu, Enhancement of oxygen permeation through La0.6Sr0.4Co0.2Fe0.8O3d hollow fibre membranes by surface modifications, J. Membr. Sci. 324 (2008) 128–135. [137] K. Gerdes, D. Luss, Oxygen flux increases through MIEC membranes by enhanced surface exchange, AIChE J. 53 (2007) 1389–1391. [138] V.V. Kharton, A.V. Kovalevsky, A.P. Viskup, F.M. Figueiredo, J.R. Frade, A.A. Yaremchenko, et al., Faradaic efficiency and oxygen permeability of Sr0.97Ti0.60Fe0.40O3d perovskite, Solid State Ionics 128 (2000) 117–130. [139] C. Wagner, Equations for transport in solid oxides and sulfides of transition metals, Prog. Solid State Chem. 10 (1975) 3–16. [140] C.H. Chen, H.J.M. Bouwmeester, H. Kruidhof, J.E. tenElshof, A.J. Burggraaf, Fabrication of La1xSrxCoO3d thin layers on porous supports by a polymeric sol-gel process, J. Mater. Chem. 6 (1996) 815–819. [141] A. Julian, E. Juste, P.M. Geffroy, V. Coudert, S. Degot, P. Del Gallo, et al., Elaboration of La0.8Sr0.2Fe0.7Ga0.3O3-d/La0.8M0.2FeO3-d (M ¼ Ca, Sr and Ba) asymmetric membranes by tape-casting and co-firing, J. Membr. Sci. 333 (2009) 132–140. [142] O. Buchler, J.M. Serra, W.A. Meulenberg, D. Sebold, H.P. Buchkremer, Preparation and properties of thin La1xSrxCo1yFeyO3d perovskitic membranes supported on tailored ceramic substrates, Solid State Ionics 178 (2007) 91–99. [143] W.Q. Jin, S.G. Li, P. Huang, N.P. Xu, J. Shi, Preparation of an asymmetric perovskite-type membrane and its oxygen permeability, J. Membr. Sci. 185 (2001) 237–243. [144] Z.G. Wang, H. Liu, X.Y. Tan, Y.G. Jin, S.M. Liu, Improvement of the oxygen permeation through perovskite hollow fibre membranes by surface acid-modification, J. Membr. Sci. 345 (2009) 65–73. [145] X.Y. Tan, Y.T. Liu, K. Li, Preparation of LSCF ceramic hollow-fiber membranes for oxygen production by a phase-inversion/sintering technique, Ind. Eng. Chem. Res. 44 (2005) 61–66.

272

Inorganic, Polymeric and Composite Membranes

[146] S.M. Liu, G.R. Gavalas, Preparation of oxygen ion conducting ceramic hollow-fiber membranes, Ind. Eng. Chem. Res. 44 (2005) 7633–7637. [147] A. Leo, S.M. Liu, J.C.D. da Costa, The enhancement of oxygen flux on Ba0.5Sr0.5Co0.8Fe0.2O3d (BSCF) hollow fibers using silver surface modification, J. Membr. Sci. 340 (2009) 148–153. [148] S.M. Liu, X.Y. Tan, K. Li, R. Hughes, Preparation and characterisation of SrCe0.95Yb0.05O2.975 hollow fibre membranes, J. Membr. Sci. 193 (2001) 249–260. [149] S.P. Deshmukh, K. Li, Effect of ethanol composition in water coagulation bath on morphology of PVDF hollow fiber membranes, J. Membr. Sci. 150 (1998) 75–85. [150] H. Wang, Y. Cong, W. Yang, Investigation on the partial oxidation of methane to syngas in a tubular Ba0.5Sr0.5Fe0.2O3d membrane reactor, Catal. Today 82 (2003) 157–166. [151] D.A. Slade, A.M. Duncan, K.J. Nordheden, S.M. Stagg-Williams, Mixed-conducting oxygen permeable ceramic membranes for carbon dioxide reforming of methane, Green Chem. 9 (2007) 5. [152] M.-S. Fan, A.Z. Abdullah, S. Bhatia, Catalytic technology for carbon dioxide reforming of methane to synthesis gas, ChemCatChem 1 (2009) 192–208. [153] S. Faraji, K.J. Nordheden, S.M. Stagg-Williams, The interaction between SrFeCo0.5Ox ceramic membranes and Pt/CeZrO2 during syngas production from methane, Catal. Lett. 131 (2009) 114–121. [154] M. Ikeguchi, T. Mimura, Y. Sekine, E. Kikuchi, M. Matsukata, Reaction and oxygen permeation studies in Sm0.4Ba0.6Fe0.8Co0.2O3d membrane reactor for partial oxidation of methane to syngas, Appl. Catal. A 290 (2005) 212–220. [155] J.-M. Kim, G.-J. Hwang, S.-H. Lee, C.-S. Park, J.-W. Kim, Y.-H. Kim, Properties of oxygen permeation and partial oxidation of methane in La0.6Sr0.4CoO3d (LSC)–La0.7Sr0.3Ga0.6Fe0.4O3d (LSGF) membrane, J. Membr. Sci. 250 (2005) 11–16. [156] Q. Li, X. Zhu, Y. He, W. Yang, Partial oxidation of methane in BaCe0.1Co0.4Fe0.5O3d membrane reactor, Catal. Today 149 (2010) 185–190. [157] Z. Shao, H. Dong, G. Xiong, Y. Cong, W. Yang, Performance of a mixed-conducting ceramic membrane reactor with high oxygen permeability for methane conversion, J. Membr. Sci. 183 (2001) 181–192. [158] C.-Y. Tsai, A.G. Dixon, W.R. Moser, Y.H. Ma, Dense perovskite membrane reactors for partial oxidation of methane to syngas, AIChE J. 43 (1997) 2741–2750. [159] X. Yin, L. Hong, Z.-L. Liu, Asymmetric tubular oxygen-permeable ceramic membrane reactor for partial oxidation of methane, J. Phys. Chem. C 111 (2007) 9194–9202. [160] P. Zeng, Z. Chen, W. Zhou, H. Gu, Z. Shao, S. Liu, Re-evaluation of Ba0.5Sr0.5Co0.8Fe0.2O3d perovskite as oxygen semi-permeable membrane, J. Membr. Sci. 291 (2007) 148–156. [161] M. Ikeguchi, K. Ishii, Y. Sekine, E. Kikuchi, M. Matsukata, Improving oxygen permeation in SrFeCo0.5Ox asymmetric membranes by modifying support-layer porous structure, Mater. Lett. 59 (2005) 1356–1360. [162] D.J. Wilhelm, D.R. Simbeck, A.D. Karp, R.L. Dickenson, Syngas production for gas-toliquids applications: technologies, issues, and outlook, Fuel Process. Technol. 71 (2001) 139–148. [163] M. Balat, Possible methods for hydrogen production, Energy Sources A 31 (2009) 39–50. [164] A.M. O’Connor, J.R.H. Ross, The effect of O2 addition on the carbon dioxide reforming of methane over Pt/ZrO2 catalysts, Catal. Today 46 (1998) 203–210. [165] M.M.V.M. Souza, M. Schmal, Combination of carbon dioxide reforming and partial oxidation of methane over supported platinum catalysts, Appl. Catal. A 255 (2003) 83–92.

Chapter

11

Review of MIEC Ceramic Membranes

273

[166] W. Wang, S.M. Stagg-Williams, F.B. Noronha, L.V. Mattos, F.B. Passos, Partial oxidation and combined reforming of methane of Ce-promoted catalysts, Catal. Today 98 (2004) 553–563. [167] V.R. Choudhary, K.C. Mondal, CO2 reforming of methane combined with steam reforming or partial oxidation of methane to syngas over NdCoO3 perovskite-type mixed metal-oxide catalyst, Appl. Energy 83 (2006) 1024–1032. [168] V.R. Choudhary, K.C. Mondal, T.V. Choudhary, Oxy-CO2 reforming of methane to syngas over CoOx/CeO2/SA-5205 catalyst, Energy Fuels 20 (2006) 1753–1756. [169] V.R. Choudhary, K.C. Mondal, T.V. Choudhary, Partial oxidation of methane to syngas with or without simultaneous steam or CO2 reforming over a high-temperature stable NiCoMgCeOx supported on zirconia–hafnia catalyst, Appl. Catal. A 306 (2006) 45–50. [170] Q. Jing, H. Lou, L. Mo, J. Fei, X. Zheng, Combination of CO2 reforming and partial oxidation of methane over Ni/BaO-SiO2 catalysts to produce low H2/CO ratio syngas using a fluidized bed reactor, J. Mol. Catal. A Chem. 212 (2004) 211–217. [171] Q.S. Jing, X.M. Zheng, Combined catalytic partial oxidation and CO2 reforming of methane over ZrO2-modified Ni/SiO2 catalysts using fluidized-bed reactor, Energy 31 (2006) 2184–2192. [172] S. He, Q. Jing, W. Yu, L. Mo, H. Lou, X. Zheng, Combination of CO2 reforming and partial oxidation of methane to produce syngas over Ni/SiO2 prepared with nickel citrate precursor, Catal. Today 148 (2009) 130–133. [173] S. He, H. Wu, W. Yu, L. Mo, H. Lou, X. Zheng, Combination of CO2 reforming and partial oxidation of methane to produce syngas over Ni/SiO2 and Ni-Al2O3/SiO2 catalysts with different precursors, Int. J. Hydrogen Energy 34 (2009) 839–843. [174] P.D.F. Vernon, M.L.H. Green, A.K. Cheetham, A.T. Ashcroft, Partial oxidation of methane to synthesis gas, and carbon dioxide as an oxidising agent for methane conversion, Catal. Today 13 (1992) 417–426. [175] J. Guo, Z. Hou, J. Gao, X. Zheng, Syngas production via combined oxy-CO2 reforming of methane over Gd2O3-modified Ni/SiO2 catalysts in a fluidized-bed reactor, Fuel 87 (2008) 1348–1354. [176] B.C. Michael, A. Donazzi, L.D. Schmidt, Effects of H2O and CO2 addition in catalytic partial oxidation of methane on Rh, J. Catal. 265 (2009) 117–129. [177] Z.P. Shao, H. Dong, G.X. Xiong, W.S. Yang, Syngas production using an oxygen-permeating membrane reactor with cofeed of methane and carbon dioxide, Chin. Chem. Lett. 11 (2000) 631–634.

This page intentionally left blank

Chapter 12

Critical Factors Affecting Oxygen Permeation Through Dual-phase Membranes Xuefeng Zhu and Weishen Yang* State Key Laboratory of Catalysis, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, China * Corresponding author: E-mail address: [email protected]

INTRODUCTION Dense ceramic oxygen permeable membranes made from mixed oxygen ionic and electronic conductors can separate oxygen from air at elevated temperature with infinite permeation selectivity. The dense membrane separator is exposed to an asymmetrical oxygen partial pressure environment. The oxygen chemical potential gradient across the membrane causes the oxygen ions to be directionally transported from the high-oxygen partial pressure side to the low-oxygen partial pressure side, accompanied by the reverse transfer of electrons, as shown in Figure 12.1. Perovskite-type and fluorite oxides are promising materials with high-oxygen ionic conductivity for oxygen permeation. Figure 12.2 shows their crystal structures. Perovskite oxides are usually denoted as ABO3
d, where A-site ions usually are alkaline earth metal ions (Ba2þ, Sr2þ, Ca2þ) and/or lanthanide metal ions (La3þ, Sm3þ, Ga3þ, etc.). Fluorite oxides are often denoted as MO2
d, and the typical fluorite oxides are ZrO2, CeO2, Bi2O3 and their alkaline earth or rare earth doped oxides. Mixed ionic/ electronic conduction is required for oxygen transport through membranes. However, fluorite oxides with low electronic conductivity exhibit poor permeability unless a potential is applied across the membrane. During recent years, considerable effort has been focused on the investigation of single-phase mixed-conducting membranes with perovskite-type structures, especially materials with the composition Ln1
x(Ba, Sr, Ca)xCo1
yFeyO3
dd (Ln: lanthanide) [1–10]. For example, the membrane materials Ba1
xSrxCo0.8Fe0.2O3
d and La1
xSrxCo1
yFeyO3
d have been intensively investigated by many research Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

275

276

Inorganic, Polymeric and Composite Membranes

High oxygen partial pressure side N2 N2 N 2

Pressurized air

O2 N2

O2 O2

N2

N2

O2

O2 + 4e− Dense ceramic membrane

O2 N 2 N2

O2

2O2− O2−

e− O2 + 4e−

Oxygen depleted air

Mixed conductors

2O2− O2

O2

O2

Low oxygen partial pressure side FIGURE 12.1 Schematic illustration of oxygen permeation through mixed-conducting ceramic membranes.

A M B O O

Fluorite: MO2

Perovskite: ABO 3

FIGURE 12.2 Structure of fluorite oxides and perovskite oxides.

groups and have been shown to have high-oxygen permeability. However, these Co-based perovskite-type membranes are not stable when they are employed as membrane reactors for natural gas conversion to syngas (containing CO, H2, and minor amounts of CO2, CH4, and H2O) at elevated temperatures [11–13]. Clearly, the most promising application of oxygen permeable membrane technology is for the generation of syngas at high temperature, not for pure oxygen production since the mature technologies of cryogenic separation and pressure swing adsorption for air separation are more cost effective. Stability is more significant than permeability for the commercialization of membrane technology. For perovskite-type mixed conductors, the improvement of stability is at the cost of permeability. Less reducible metal ions, such as Zn2þ, Al3þ, Ga3þ, Ti4þ, Zr4þ, Ce4þ, etc., have been proposed for the partial substitution of the reducible ions (Co3þ/4þ, Fe3þ/4þ) to enhance the membrane stability [14–24]. This strategy works but not dramatically. For example, the Zr-doped perovskite oxide Ba(Co,Fe)O3 membrane changes its composition and morphology to a depth of 100 mm after long-term use for syngas generation [20]. The cobalt-free

Chapter

12

Critical Factors Affecting Oxygen Permeation

277

materials, SrFe0.7Al0.3O3 and BaCe0.15Fe0.85O3, can retain their perovskite structure under an H2-containing atmosphere at high temperature, but their surface morphologies and microstructures are still destroyed by syngas to a depth of tens of microns [15,25]. The overall stability of perovskite oxides is made up of a combination of thermodynamic and kinetic stability. The thermodynamic stability depends mainly on the average bond energy (ABE) of the B-site ions. High ABE values lead to high stability, but the motion of oxygen ions in the lattice becomes difficult. The kinetic stability of perovskite-type membranes relies on the self-diffusion coefficients of the metallic cations. When a chemical potential gradient for oxygen is applied to the membranes, not only oxygen ions but also metallic cations are driven from one side to the other side. The kinetic demixing rate largely depends on the differences in the self-diffusion coefficients between the cations, the temperature, the oxygen chemical potential gradient across the membrane, the microstructure, and oxygen exchange rates on the surfaces. However, adjustment of the microstructure can only marginally improve the stability of perovskite-type membranes. Dual-phase membranes, with a separate transport path for electrons and oxygen ions, have been suggested as another choice for oxygen permeation. Initial researches were focused on the cermet dual-phase membranes made of noble metals (such as Au, Pd, Pt, Ag, etc.) and oxygen ionic conductors (such as Bi2O3, ZrO2, and CeO2-based solid electrolytes) [26–28]. However, because of their high cost and low permeability and, in contrast, the high permeability of perovskite-type membranes, not much attention was paid at the time. Ceramic dual-phase membranes, consisting of an oxide for electronic transport together with an oxygen ionic conductor, were first reported by Kharton et al. [29], who reported a study of the Ce0.8Gd0.2O1.9–La0.7Sr0.3MnO3 composite membrane [29]. Soon after then, the dual-phase membranes Ce0.8Sm0.2O1.9–La0.8 Sr0.2Co0.2Fe0.8O3, Ce0.8Gd0.2O1.9–Gd0.7Ca0.3CoO3, La0.8Sr0.2Ga0.8Mg0.2O3– La0.8Sr0.2Co0.2Fe0.8O3, Ce0.8Sm0.2O1.9–La0.8Sr0.2CrO3, Zr0.6Y0.4O1.8–La0.7 Sr0.3MnO3, and others were described by several groups [30–35]. All these membranes have high stability but poor permeability. In addition, the oxygen permeation process is unstable over the prolonged periods. For Ce0.8Gd0.2O1.9–La0.7Sr0.3MnO3, the oxygen flux decreases with time and does not reach a steady state [29]. However, the oxygen fluxes of Ce0.8Sm0.2O1.9–La0.8Sr0.2Co0.2Fe0.8O3 and Ce0.8Sm0.2O1.9–La0.8Sr0.2CrO3 initially increase and then achieve steady states [30,34]. There has been no in-depth investigation of these dual-phase membranes because their permeation fluxes are far less than that required for practical uses. Recently, a new group of dual-phase composite membranes, comprising an oxide ionic conductor and a mixed conductor, was reported to have higher permeability and stability [36–41]. The oxide ionic conductor transports oxide ions, and the mixed conductor transports both oxide oxygen ions and electrons, rather than purely electrons as a noble metal does. This is the key to the improvement.

278

Inorganic, Polymeric and Composite Membranes

Poor permeability of former dual-phase membranes can be attributed to the blockage of the transport of oxide ions through the bulk membrane by the electron-conducting phase (noble metal or perovskite oxide). When a mixed conductor is used instead of a pure electronic conductor, this blockage disappears [36]. Therefore, it is expected that the new dual-phase composite membranes, consisting of an oxide ionic conductor and a mixed conductor, will have a high-oxygen permeation flux. Since there has been no previous review of dual-phase membranes, we are taking this opportunity to summarize the achievements of our group and other researchers and discuss the critical factors affecting oxygen permeation through dual-phase ceramic membranes and also throw light on the development of new dual-phase membranes.

DESIGN OF DUAL-PHASE MEMBRANES WITH HIGH STABILITY AND PERMEABILITY For dual-phase membranes, the transport of electrons and oxide ions takes place separately in the two phases. This facilitates tailoring the stability and permeability of the membranes by choosing an ionic conducting phase and electronic conducting phase, respectively. However, the design of dual-phase membranes must follow the basic principles of solid state chemistry. Figure 12.3 shows the ideal of designing new dual-phase membranes with high stability and permeability. First of all, an oxide ionic conductor must be selected for dual-phase membranes since high-oxygen ionic conductivity is a prerequisite to high permeability. Among known ionic conductors, Y or Sc-stabilized zirconia, doped ceria, doped bismuth oxides, La2Mo2O9, and La1
xSrxGa1
yMgyO3
d (0  x  0.2, 0  y  0.2) possess high ionic conductivity and are suitable as (a)

Oxide ionic conductors

Electronic conductors

(b)

Zr1-x(Ca,Y,Sc)xO2-d , Ce1-xLnxO2-d , Bi2-x(M, Ln)xO3-d, Bi4V1.8Cu0.2O11-d, La1-xCaxAlO3, La1-xSrxGa1-yMgyO3-d , La2Mo2O9, La10Si6O26+d ... ...

Mixed conductors

Pure electronic conductors

Ionic conductivity Zr1-x(Y,Sc)O2-d , Ce1-xLnxO2-d , Bi2-x(M, Ln)xO3-d , Bi4V1.8Cu0.2O11-d , La1-xSrxGa1-yMgyO3-d , La2Mo2O9

Ln1-w(Ba,Sr,Ca)wFe1-vCovO3 Stability under reducing environments

Zr1-x(Y,Sc)O2-d , Ce1-xLnxO2-d , La1-xSrxGa1-yMgyO3-d

Ln1-w(Ba,Sr,Ca)wFeO3 Compatibility with another phase

Ce1-xLnxO2-d Ce1-xGdxO2-d

Ln1-wSrwFeO3

Ce1-xSmxO2-d

Gd1-wSrwFeO3

Sm1-wSrwFeO3

Ce1-xGdxO2-d -Gd1-wSrwFeO3 Ce1-xSmxO2-d -Sm1-wSrwFeO3

FIGURE 12.3 Selection of oxide ionic conductors and perovskite-type electronic conductors for dual-phase composite membranes. Ln: rare earth element.

Chapter

12

Critical Factors Affecting Oxygen Permeation

279

the ionic conducting phase for a dual-phase membrane. However, when the stability under reducing environments is considered, only Y or Sc-stabilized zirconia, doped ceria, and La1
xSrxGa1
yMgyO3
d are suitable. Although bismuthbased fluorite or Aurivillius oxides have the highest ionic conductivities among the known oxygen ionic conductors, they are all easily reduced to metals even with a low partial pressure of oxygen. Their stability is nowhere near that required for membrane reactors for syngas generation. La2Mo2O9 and its substitutes are also unstable in reducing environments at high temperatures [42]. In addition, the volatilization of MoO3 during sintering and high temperature operation limits their applications. Perovskite-type or spinel oxides (such as MnFe2O4, MnCo2O4, CuCo2O4, etc.) are good electronic conductors at high temperatures. Therefore, when the chemical compatibility between the ionic and electronic conducting phases is considered, only doped ceria remains. Zirconia easily reacts with perovskite-type oxides to produce zirconates and pyrochlores when sintered at elevated temperatures, and the resulting new compounds have poor ionic and electronic conductivity, so block the transport of both ions and electrons. The reactions between Y/Sc-stabilized zirconia and spinel oxides are not as severe as those with perovskite-type oxides, but the cubic/tetragonal structures with high ionic conductivity are degraded to monoclinic structures with poor ionic conductivity. The elements in electronic conducting perovskites easily diffuse into the La1
xSrxGa1
yMgyO3
d lattice and produce a new perovskite oxide, and the new materials usually have normal stability and permeability. In the investigation of La0.8Sr0.2Ga0.8Mg0.2O3–La0.8Sr0.2Co0.2Fe0.8O3, Shaula et al. found that the composite material almost becomes a pure perovskite phase after sintering at high temperature [33]. The similar crystal structures of the above two perovskite oxides make for ready diffusion of metal ions from one oxide to the other. Similarly, almost all the elements that make up spinel oxides can be substituted by Ga3þ and Mg2þ to produce new composite systems such as La0.8Sr0.2(Ga, Mg, Mn, Fe, Co)O3 þ (Ga, Mg, Fe, Co)O4, or even more complex phases. Therefore, these two kinds of ionic conductors (zirconia-based or LaGaO3-based perovskite oxides) are improperly used as oxygen ionic conducting phases in dual-phase membranes except when compatible electronic conducting oxides are developed. Electronic conducting oxides can be classified into two groups, pure electronic conductors (or those where the ionic conductivity is extremely low, such as the perovskite-type La1
xSrxMn(Cr)O3 and spinels (M,M0 )3O4) and mixed conductors (such as Lnx(Ba, Sr, Ca)1
xCoyFe1
yO3
d). If a pure electronic conductor or mixed conductor with poor ionic conductivity is used as the second phase in the dual-phase membrane, it blocks the transport of oxygen ions in the membrane bulk between ionic conductor grains. For example, when La0.8Sr0.2Co0.2Fe0.8O3, a mixed conductor with low ionic conductivity, is used in a dual-phase system, the permeation flux of the membrane is as low as 3.7  10
8 mol cm
2 s
1 at 950  C for a thickness of 1 mm, which is even lower than that of the Ce0.8Gd0.2O1.9–La0.7Sr0.3MnO3 dual-phase membrane under the same conditions [29,30]. However, when a mixed conductor with

280

Inorganic, Polymeric and Composite Membranes

Electronic conductor

Island

O2− e−

Mixed conductor

O2−

e− O2−

Ionic conductor

Ionic conductor

FIGURE 12.4 Schematic illustration of oxygen transport in ionic–electronic conductor dual-phase membrane and ionic-mixed conductor dual-phase membrane [36].

significant ionic conductivity is used as the second phase, it will contribute to, rather than block, the transport of oxide ions, as illustrated in Figure 12.4. It is expected that dual-phase membranes made up of an ionic conductor and a mixed conductor have higher oxygen permeation fluxes than those made of a pure electronic conductor or a mixed conductor with poor ionic conductivity [36]. If the stability under a reducing environment is considered, perovskites doped with cobalt or other reducible metal ions are excluded, and only Fe-based perovskite oxides are left. If the compatibility between doped ceria and Lnx(Ba, Sr,Ca)1
xFeO3
d perovskite is considered, barium and calcium should be excluded because barium-doped perovskite oxides easily react with ceria to produce BaCeO3, and calcium can easily diffuse into the ceria lattice and decrease the ionic conductivity. Of course, strontium-doped perovskite oxides can also react with ceria to produce SrCeO3. However, the low tolerance factor of 0.87 gives Sr2þ a higher chemical potential in SrCeO3 than in Ln1
xSrxFeO3, so SrCeO3 does not appear in the list of dual-phase composite membranes. This has been verified in our previously published papers [37–41]. Therefore, we can obtain dual-phase membranes with the composition Ce1
xLnxO2
d–Ln1
ySryFeO3
d. Among doped ceria, Gd3þ and Sm3þ doping produces the highest conductivities, so the matched perovskite oxides need to be Gd1
ySryFeO3
d and Sm1
ySryFeO3
d. The stability of the perovskite can be improved by doping Al3þ or Ga3þ into the B-site.

EXPERIMENTAL INVESTIGATION OF DUAL-PHASE MEMBRANES Pure Electronic Conductor or Mixed Conductor? Experimental investigations of the designed dual-phase membranes were performed in our laboratory, and some interesting results were obtained. The key to

Chapter

12

Critical Factors Affecting Oxygen Permeation

281

the design of dual-phase membranes is the selection of pure electronic conductors or mixed conductors for electronic transport through the membranes. The above analysis of the design of dual-phase membranes shows that mixed conductors are better than pure electronic conductors, because mixed conductors with significant ionic conductivity would contribute to, rather than block, the transport of oxide ions. The following two dual-phase membranes were selected to illustrate the difference experimentally. The material 75 wt% Ce0.85Sm0.15O1.925–25 wt% Sm0.6Sr0.4FeO3
d is designated as SDC75–SSF25 and 75 wt% Ce0.85Sm0.15O1.925–25 wt% Sm0.6Sr0.4CrO3
d as SDC75–SSFCr25. Only one element is different between the two dual-phase membranes. It makes the former a mixed conductor–ionic conductor dual-phase membrane and the latter a pure electronic conductor–ionic conductor dualphase membrane. The two membranes were prepared by the same one-pot method used for the synthesis of the composite powders, followed by pressing and sintering at the same pressure and temperature. All the metallic ions required are mixed in one beaker, so that every element has the same chemical potential in each phase after synthesis. Therefore, the diffusion of metallic elements between the two phases during synthesis is complete. However, the composition of each phase differed slightly from the target values, as revealed by EDX analysis [38]. XRD data from the as-synthesized powders reveal that there is no new phase emerging, such as SrCeO3. This demonstrates that the perovskite phases have good structural compatibility with the fluorite phase even though the powders were prepared using the one-pot method. The good compatibility arises since the two phases contain the same rare earth element samarium, as well as strontium having low dissolution energy in the perovskite lattice [43,44]. The oxygen permeation flux for SDC75–SSF25 is more than four times that of SDC75–SSCr25 under the same conditions, as shown in Figure 12.5. The oxygen permeation activation energy of SDC75–SSF25 (66  1 kJ mol
1) is smaller than that of SDC75–SSCr25 (94  2 kJ mol
1), and close to the reported oxygen ionic conduction activation energy of SDC [45].This reveals that the occurrence of a mixed-conducting phase, SSF, in the dual-phase membrane system has little influence on the ionic transport through the bulk of membrane, but the appearance of a pure electronic conductor in the dual-phase system will block the ionic transport.

Surface Exchange Different groups have reported variable oxygen permeation through a number of dual-phase membranes in the initial stages. Usually, it takes tens or hundreds of hours to reach oxygen permeation steady state [30,34,36,38]. The oxygen permeation flux at steady state is often twice the initial value. The variable oxygen permeation state for single-phase perovskite-type membranes is related to the readjustment of the lattice structure from the as-prepared state to an

282

Inorganic, Polymeric and Composite Membranes

SDC75–SSF25 SDC75–SSCr25

JO2 (ml cm−2 min−1)

0.8

0.6

0.4

0.2

0.0 750

800

850 Temperature (°C)

900

950

FIGURE 12.5 Temperature dependence of oxygen permeation fluxes of SDC75–SSF25 and SDC75–SSCr25 dual-phase membranes with both sides coated with La0.6Sr0.4CoO3 porous layers. Thickness: 0.5 mm; air flow rate: 100 ml min
1; He flow rate: 30 ml min
1.

invariant state under permeation conditions or to the reduction of surface oxygen. However, this process is still not clear for dual-phase membranes. A porous layer of La0.6Sr0.4CoO3 (LSC)  20 mm thick was coated on one side or both sides of the dual-phase membranes to speed up the oxygen exchange on the membrane surfaces [46]. The results revealed that the membrane with both side coated with LSC reaches the permeation steady state immediately and shows the highest oxygen flux, greater than those with only one side coated or naked membranes. It takes more time for the naked membranes than membranes with LSC layers on the feed side or sweep side to reach the steady state. For example, the naked membrane needs about 42 h to reach steady state, but the membrane with its sweep side coated with an LSC porous layer needs only 8 h [46]. This unsteady period for the SDC75–SSF25 dual-phase membranes is irreversible. When the membranes reached their permeation steady state, the helium was replaced by air on the permeation side, and after about 50 h, the helium flow was restored. However, the oxygen flux maintained its steady value. From the above results, we deduce that the unsteady permeation of dual-phase membranes is related to the surface oxygen exchange on the membrane surface.

Preparation Methods for Powders Even for the same oxygen permeable materials, there is little reproducibility of values of the oxygen permeability in the literature. The method of powder synthesis is also an important factor. There are many methods available to prepare perovskite powders, such as the solid state reaction (SSR) method, the EDTA-

Chapter

12

Critical Factors Affecting Oxygen Permeation

283

citric acid (EC) process, the glycine–nitrate combustion process (GNP), and the chemical coprecipitation (CP) method, and others. Powders synthesized by different processes have different particle sizes, sintering activity, and chemical components. These factors will lead to differences in microstructure of the ceramic membranes. Kharton et al. reported that SrCo0.6Fe0.25Cu0.15O3
d, synthesized by the SSR method (giving a large grain size), had higher oxygen permeability than that synthesized via a cellulose precursor [47]. Tan et al. reported that the oxygen permeability of Ba0.5Sr0.5Co0.8Fe0.2O3
d synthesized by the SSR method was higher than that synthesized by the EC process [48]. They attributed this to high ionic transport in the bulk rather than along the grain boundaries. Further, it was reported that mixed-conducting oxides synthesized by different methods had different phase compositions. For example, the phase composition of SrFeCo0.5Oy is largely determined by the synthetic method. Materials synthesized by the SSR method possess the perovskite phase SrFe1
xCoxO3
d, the intergrowth phase Sr4Fe6
xCoxO13þd, and spinel Co3
xFexO4; materials synthesized by the sol–gel process usually produced Sr4Fe6
xCoxO13þd as the main phase [49], while materials synthesized by the GNP process usually generated SrFe1
xCoxO3
d as the main phase [50]. Dual-phase membranes are usually prepared by simply mixing the two oxide powders followed by pressing and sintering. However, the two oxide powders cannot be uniformly mixed using mechanical mixing, so the local thermal expansion and mechanical strength of the composite membranes are variable. The effects of powder preparation methods on the microstructures, chemical composition, defects, conductivity, and oxygen permeability of perovskite materials have been extensively investigated. The influence of the synthetic method on the oxygen permeability and homogeneity of dual-phase membranes is still unclear, so in this section, powder preparation methods for dual-phase membranes are discussed, since they are clearly important factors affecting oxygen transport through dual-phase membranes.

Solution-Based Synthesis Methods Four methods were used to prepare the composite powders of 75 wt% GDC–25 wt % GSF (corresponding to a volume ratio of 71:29) [38]. In method 1, the required fluorite and perovskite oxides were prepared via the EC procedure and then mixed with a weight ratio of 75:25. In method 2, the GDC powder obtained in method 1 is mixed with a complex solution containing Gd3þ, Sr2þ, and Fe3þ. Similarly, in method 3, a complex solution containing Gd3þ and Ce3þ is mixed with the GSF powder obtained in method 1. In method 4, all the required metal nitrates are mixed in one beaker and then follows the EC procedure. After evaporation, all the resultant gels were calcined on an electrical furnace at 600–800  C to remove most of organics and then calcined in muffle at 900  C for 5 h. The derived dual-phase membranes are labeled as GDC–GSF-1–4, respectively. The morphologies of the dual-phase membranes were characterized with BSEM, and the results are shown in Figure 12.6. The white areas are fluorite

284

Inorganic, Polymeric and Composite Membranes

GDC–GSF-1

GDC–GSF-2

GDC–GSF-3

GDC–GSF-4

FIGURE 12.6 Backscattered electron images of the dual-phase membranes prepared via mixing two powders (GDC–GSF-1); GDC powder and GSF precursor solution (GDC–GSF-2); GDC precursor solution and GSF powder (GDC–GSF-3); one-pot method (GDC–GSF-4). Scale bar: 5 mm, adapted from Ref. [38].

grains, and the black areas are perovskite grains. It is easy to see that the uniformity and grain sizes of the two phases are different for the four membranes derived from the different methods. The uniformity of GDC–GSF-4 is better than the others. The grain sizes of these membranes are mainly in the range 0.53 mm, and the grain size of the perovskite phase is usually smaller than that of the fluorite phase for all membranes. GDC–GSF-4 has a relatively narrow grain size distribution, around 1 mm for the perovskite phase and 12 mm for the fluorite phase. Oxygen transport through the membranes is controlled by the dipolar conductivity of the oxide ions and electrons. In inhomogeneous membranes, like GDC–GSF-2, some perovskite grains were enclosed in fluorite grains, forming islands, as shown in Figure 12.4. It is certain that the islands make little contribution to oxygen permeation since the transport of electrons between the perovskite grains is blocked by fluorite. However, the influence of this blockage is less for homogeneous membranes than inhomogeneous ones. Therefore, the dipolar conductivity of the homogeneous sample is higher than that of the inhomogeneous sample, which ultimately affects the oxygen permeation flux. The permeation fluxes of all membranes showed an increase during the first  50 h, and then reached steady states. The preparation methods have little influence on the permeation stability but a significant effect on the oxygen permeability. Figure 12.7 shows the temperature dependence of the oxygen permeation fluxes of

Chapter

12

285

Critical Factors Affecting Oxygen Permeation

−6.2

logJO2 (mol cm−2 s−1)

−6.4 −6.6 −6.8 −7.0 −7.2 −7.4 0.80

GDC-GSF-1, Ea = 95.0 kJ mol-1 GDC-GSF-2, Ea = 124.5 kJ mol-1 GDC-GSF-3, Ea = 93.1 kJ mol-1 GDC-GSF-4, Ea = 92.2 kJ mol-1

0.84

0.88 0.92 1000/T (K−1)

0.96

1.00

FIGURE 12.7 Arrhenius plots of the oxygen permeation fluxes through GDC75–GSF25 dualphase membranes prepared via different methods. Oxygen partial pressure gradient: 21k Pa/ 0.5 kPa; thickness: 0.5 mm [38].

the four membranes. Greater homogeneous mixing of the two phases leads to higher oxygen permeability and lower permeation activation energy.

Solid State Reaction Method The SSR method is a standard process for preparing ceramic powders. It is very commonly used in preparing mixed-conducting ceramic membranes. Usually, membranes derived using the SSR method have higher oxygen permeability than those derived from liquid phase synthesis. The main reason is that the membrane usually has more defects at the crystal boundaries than in the lattice, and these cationic/anionic defects can enhance oxygen ion conductivity and improve oxygen exchange. For example, LaCoO3
d ceramic membranes can be prepared by several methods, and the membrane derived using the SSR method shows the highest oxygen permeability. This is believed to be caused by a low-grain-boundary resistance to the transport of oxide ions [47]. Figure 12.8 shows the temperature dependence of oxygen permeation flux of 75 wt% Ce0.85Sm0.15O1.925–25 wt% Sm0.6Sr0.4Al0.3Fe0.7O3
d (SDC75–SSAF25) dual-phase membranes prepared using different methods, namely, the EC method, the SSR, and the CP method. In all these methods, the required amounts of precursors are mixed together, followed by the relevant steps. Changes in the processing route may affect the surface exchange, including the surface concentration of active adsorption centers. Here, both sides of all membranes were coated with LSC porous layers to eliminate oxygen exchange limitations, providing a direct comparison of oxygen ionic conductivity of these dual-phase membranes. Overall, the oxygen permeability within the examined temperature

286

Inorganic, Polymeric and Composite Membranes

0.8 Solid state reaction method Co–precipitation method EDTA–citric acid method

JO2 (ml cm−2 min−1)

0.7 0.6 0.5 0.4 0.3 0.2 0.1 800

825

850 875 900 Temperature (°C)

925

950

FIGURE 12.8 Oxygen permeation fluxes of the SDC75–SSAF25 membranes prepared by different methods with both sides coated with La0.6Sr0.4CoO3 porous layers. Thickness: 0.5 mm; air flow rate: 100 ml min
1; He flow rate: 30 ml min
1.

range increases in the sequence M-EC < M-CP < M-SSR. When comparing the membranes prepared by the standard ceramic and liquid phase techniques, this behavior may be attributed to increasing average grain size, in the order M-EC < M-CP < M-SSR, and lower grain-boundary resistance to ionic transport. The different grain-boundary resistances arise from the various synthetic techniques for the powders. Grain boundaries are, in fact, a complex entity, and simple correlation between grain size and grain-boundary performance is only possible when the remaining processing steps are similar. For the SDC75–SSAF25 dualphase membrane, grain boundaries provide a positive influence on the total oxygen ionic conductivity. It should be emphasized that when the grain-boundary resistance to ionic transport is low enough, ionic diffusion may take place along grain boundaries and provide an additional contribution to the oxygen transport. This behavior is often observed in single-phase perovskite mixed conductors. The grain-boundary oxygen diffusion coefficient in A-site-deficient LaMnO3 and La0.6Sr0.4Co0.2Fe0.8O3
d is 2–4 orders higher than that for the bulk materials [47]. A detailed investigation of the microstructure of dual-phase membranes is needed to understand the influence of the grain boundary, grain size, and defects on oxide ionic and electronic transport.

Sintering Temperature Another fact that should be noted is that the microstructure and texture of ceramic membranes developed are greatly influenced by the sintering temperature with effects on their oxygen permeability and structural stability. This is

Chapter

12

287

Critical Factors Affecting Oxygen Permeation

despite the fact that the oxygen permeability of a given material can vary by several orders of magnitude due to differences in microstructure. It has been reported that reducing the ceramic grain size will decrease the electrical conductivity and oxygen permeability through perovskite-type membranes. As an example, Wang et al. investigated the effects of sintering temperature on Ba0.5Sr0.5Co0.8Fe0.2O3
d membranes and found that the oxygen permeability increases with grain size [51]. Usually, for perovskite-type mixed-conducting ceramics, the grain-boundary resistance plays an important role in ionic and electronic transport and decreases with an increase of grain size (or sintering temperature). However, the above investigations on the sintering temperature were all focused on single-phase membrane materials. The components of dual-phase membranes are more complex, so the influence of sintering temperature on the microstructure and permeability of dual-phase membranes is more complex. Figure 12.9 shows the oxygen permeation flux of SDC75–SSF25 dual-phase membranes sintered at different temperatures [52]. All the membranes were prepared using the one-pot method. It can be seen that the oxygen permeation flux of the membrane disks at a given operating temperature varies with sintering temperature. The membrane sintered at 1425  C exhibits the highest oxygen permeation flux (950  C, 3.1  10
7 mol cm
2 s
1), and the membrane sintered at 1525  C has the lowest oxygen permeation flux (950  C, 1.6  10
7 mol cm
2 s
1). It was found that the membrane surface is composed of fluorite and perovskite grains after sintering at 1400 and 1425  C, but only the fluorite phase is detected when the sintering temperature rises to 1450  C. After polishing, which removes about 200 mm thickness, all the

950 °C 850 °C

0.3

0.2

2

JO (ml cm−2 min−1)

0.4

0.1

0.0 1400

1425

1475 1500 1450 Sintering temperature (°C)

1525

FIGURE 12.9 Oxygen permeation fluxes of dual-phase membranes versus the various sintering temperature at 950 and 850  C. Thickness: 0.7 mm; air flow rate: 100 ml min
1; He flow rate: 30 ml min
1 [52].

288

Inorganic, Polymeric and Composite Membranes

membranes, irrespective of sintering temperature, show both the fluorite and perovskite phases. The EDX analysis shows that the increase of sintering temperature leads to a discrepancy in elemental composition between the membrane surface and membrane bulk, which impairs the homogenous distribution of SSF and SDC grains in the membranes. In the dual-phase membranes, the oxygen permeation process depends on the cooperation between SDC and SSF. It should be noted that a homogenous distribution of the dual-phase grains improves the oxygen permeability of the dual-phase membranes through ionic transport and surface exchange. However, enhancement of the sintering temperature leads to a growth in grain size. This degrades the homogeneous distribution of the two kinds of grains and destroys the percolation of the mixed-conducting phase through the membranes. This phenomenon is significantly different from that in the single-phase perovskite membranes because crystal grains of single-phase membranes can simultaneously transport oxygen ions and electrons.

Ratio Between the Two Phases Ceria-based fluorite ionic conductors have high conductivity and stability at elevated temperatures, so it was hoped that dual-phase membranes containing more fluorite oxides would give good performance because the fraction of the mixed-conducting phase could exceed the percolation limit. For comparison purposes, both sides of dual-phase membranes were coated with LSC porous layers to eliminate the disturbance of surface exchange. The oxygen permeability within the investigated temperature range decreases in the sequence SDC65–SSF35  SDC75–SSF25 > SDC85–SSF15. The ionic conductivity of SDC at 950  C is about 0.1 S cm
1, and the ionic conductivity of SSF is also about 0.1 S cm
1 at this temperature, based on published research [40], so the ionic conductivity of the SDC–SSF dual-phase membranes is close to 0.1 S cm
1 no matter what the volume ratio of the two phases, if the slight elemental diffusion between the two phases is neglected. From the total conductivity data and keeping the Wagner equation in mind, it was found that the electronic conductivities are high enough for SDC65–SSF35 ( 1.6 S cm
1) and SDC75–SSF25 ( 0.5 S cm
1) but insufficient for SDC85–SSF15 ( 0.17 S cm
1), so the former two materials with similar ionic conductivity have comparable oxygen permeation flux and even activation energy. A value of oxygen permeation flux as high as 0.6  10
7 mol cm
2 s
1 was achieved for the SDC65–SSF35 and SDC75–SSF25 dual-phase membrane if the surface reactions were activated. The optimal ratio was around 75:25 for the SDC–SSF dual-phase membrane system but may be different for others. The total conductivity of SSF at 950  C is about 21 S cm
1, which is low compared with most mixed-conducting perovskite oxides. If a high conductivity mixed conductor is used in the dual-phase membranes, the proportion of ionic conductor can be increased to maximize the oxygen permeability of the dual-phase membranes.

Chapter

12

Critical Factors Affecting Oxygen Permeation

289

Other Factors As well as the factors discussed above, there are many others that may influence the oxygen transport through dual-phase membranes, such as the elemental composition of both phases, grain sizes, the grain size ratio between the two phases, the conductivity of the mixed-conducting phase, the microstructure, thickness, etc. Elemental diffusion between the two phases cannot be avoided during the synthesis and high temperature operation. However, the correct choice of elemental composition of dual-phase membranes can prevent the variable permeation rate induced by elemental diffusion under operating conditions. Al3þ was doped in the SSF to improve its stability for syngas generation. However, the permeation flux of the resultant new dual-phase membrane is only 60% of the undoped one even though it was prepared using the same procedure. The doping of Al3þ decreases the total conductivity of the dual-phase membrane and reveals insufficient electronic conductivity for oxygen permeation. The grain size of the dual-phase membranes is dependent on the sintering temperature, preparation method, and the volume ratio between the two phases, as is the grain size ratio between the two phases. An appropriate grain size ratio between the two phases is needed to provide a percolation path for the perovskite phase when using the least amount. For the SDC75–SSF25 dual-phase membranes, the increase of sintering temperature leads to a growth of the grains and a simultaneous decrease of the grain size ratio. Usually, 30 vol.% of the electronic conducting phase is needed to exceed the percolation threshold for a dual-phase system where the grain sizes of the two phases are comparable. A low percolation threshold will be obtained for a high grain size ratio between the two phases. New methods need to be developed for construction of such a microstructure. The defect structures, including lattice defects and boundary defects, provide another undetermined factor affecting oxygen permeation through dual-phase membranes; this depends mainly on the preparation procedure for a given composition. Defects sometimes favor the ionic and electronic transport in the membrane bulk and oxygen exchange on the membrane surfaces, but sometimes it results in variable permeation due to their instability at elevated temperatures. For example, the permeation flux of the GDC–LSM dual-phase membrane degrades with time due to the enrichment in Mn-containing oxides at the SDC boundaries. If the limitation of oxygen exchange on both sides can be totally removed by coating with catalysts, the oxygen permeation flux of single-phase membranes increases linearly with the reciprocal of thickness. However, this is not suitable for dual-phase membranes. The percolation threshold will decrease quickly with a reduction of thickness when the membrane thickness is comparable to the grain size. In addition, this leads to a rapid increase of both electronic and ionic conductivity, especially for the dual-phase membranes made of pure electronic and ionic conductors. Under these circumstances, a small difference in oxygen permeation permeability will be found between the two kinds of dual-phase

290

Inorganic, Polymeric and Composite Membranes

membranes, such as SDC–SSF and SDC–SSCr. There may be other factors not discussed here which affect oxygen transport through dual-phase membranes, since investigation of dual-phase membranes is just beginning.

CONCLUSIONS Though dual-phase membranes have high stability and provide potential applications for air separation and membrane reactors for light hydrocarbon conversion, their permeability remains far lower than most perovskite-type membranes and the requirements for commercialization of the oxygen membrane technology. Therefore, in this work, we discuss the critical factors affecting oxygen permeation through dual-phase membranes to discover a route for the improvement of the permeability of dual-phase membranes. Through design and analysis based on current knowledge in solid state electrochemistry, two series of dual-phase membranes made of ceria-based ionic conductors and Fe-based perovskite mixed conductors are suggested as promising dual-phase materials. Pure electronic conductors are inappropriate for dual-phase membranes unless the thickness of the membrane is comparable to the grain size. Oxygen exchange on the membrane surface, preparation methods, sintering temperature, and phase composition are all critical factors, verified by experiments. Other factors, such as the elemental composition of the phases, grain size, grain size ratio between the two phases, the conductivity of the mixed-conducting phase, the microstructure, and thickness, all affect the oxygen transport process and deserve thorough investigation. Some factors are closely correlated to each other, so it is difficult to separately investigate the influence of one factor. To improve the stability of the perovskite mixed-conducting phase, one can dope some Al3þ/Ga3þ or other less reducible ions into the B-sites. The change of composition usually leads to a decrease in total conductivity and an increase in optimal sintering temperature for the dual-phase membranes. The effects of boundary defects on oxygen transport in bulk dual-phase membranes have been unclear until now, especially how oxygen ions transport between the perovskite and fluorite phase. An investigation of these problems using high resolution electron microscopy may disclose some important information and help the development and optimization of mixedconducting ceramic materials for high-temperature electrochemical applications. New materials and techniques need to be developed to improve the permeation flux, and new ideas for the selection of dual-phase materials are also needed to make further progress.

ACKNOWLEDGMENTS The authors gratefully acknowledge the financial support of National Natural Science Fund (20801053), China National Natural Science Fund for Distinguished Young Scientists (20725313), and the Ministry of Science and Technology of China (Grant No. 2005CB221404).

Chapter

12

Critical Factors Affecting Oxygen Permeation

291

REFERENCES [1] H.J.M. Bouwmeester, A.J. Burggraaf, Dense ceramic membranes for oxygen separation, in: The CRC handbook of Solid State Electrochemistry, CRC Press, Inc., New York, 1997, p. 481. [2] Y. Teraoka, H.M. Zhang, S. Furuka, N. Yamazoe, Oxygen permeation through perovskite-type oxides, Chem. Lett. (1985) 1743. [3] Y. Teraoka, T. Nobunaga, N. Yamazoe, Effect of cation substitution on the oxygen semipermeability of perovskite-type oxides, Chem. Lett. (1988) 503. [4] C.Y. Tsai, A.G. Dixon, Y.H. Ma, W.R. Moser, M.R. Pascucci, Dense perovskite, La1
xAxFe1
yCoyO3
d (A¼ Ba, Sr, Ca), membrane synthesis, applications, and characterization, J. Am. Ceram. Soc. 81 (1998) 1437. [5] V.V. Kharton, A.A. Yaremchenko, A.V. Kovalevsky, A.P. Viskup, E.N. Naumovich, P.F. Kerko, Perovskite-type oxides for high-temperature oxygen separation membranes, J. Membr. Sci. 163 (1999) 307. [6] Z.P. Shao, W.S. Yang, Y. Cong, H. Dong, J.H. Tong, G.X. Xiong, Investigation of the permeation behavior and stability of a Ba0.5Sr0.5Co0.8Fe0.2O3
d oxygen membrane, J. Membr. Sci. 172 (2000) 177. [7] Y.S. Chou, J.W. Stevenson, T.R. Armstrong, L.R. Pederson, Mechanical properties of La1
xSrxCo0.2Fe0.8O3 mixed-conducting perovskites made by the combustion synthesis technique, J. Am. Ceram. Soc. 83 (2000) 1457. [8] S.J. Xu, W.J. Thomson, Stability of La0.6Sr0.4Co0.2Fe0.8O3-d perovskite membranes in reducing and nonreducing environments, Ind. Eng. Chem. Res. 37 (1998) 1290. [9] K. Zhang, Y.L. Yang, D. Ponnusamy, A.J. Jacobson, K. Salama, Effect of microstructure on oxygen permeation in SrCo0.8Fe0.2O3–d, J. Mater. Sci. 34 (1999) 1367. [10] J.W. Stevenson, T.R. Armstrong, R.D. Carnein, L.R. Pederson, W.J. Weber, Electrochemical properties of mixed conducting perovskites La1-xMxCo1-yFeyO3-d (M¼Ba, Sr, Ca), J. Electrochem. Soc. 143 (1996) 2722. [11] S. Pei, M.S. Kleefisch, T.P. Kobylinski, J. Faber, C.A. Udovich, V. Zhang-McCoy, et al., Failure mechanisms of ceramic membrane reactors in partial oxidation of methane to synthesis gas, Catal. Lett. 30 (1995) 201. [12] W. Jin, S. Li, P. Huang, N. Xu, J. Shi, Y.S. Lin, Tubular lanthanum cobaltite perovskite-type membrane reactors for partial oxidation of methane to syngas, J. Membr. Sci. 166 (2000) 13. [13] H.J.M. Bouwmeester, Dense ceramic membranes for methane conversion, Catal. Today 82 (2003) 141. [14] H.H. Wang, C. Tablet, A. Feldhoff, J. Caro, A cobalt-free oxygen-permeable membrane based on perovskite-type oxide Ba0.5Sr0.5Zn0.2Fe0.8O3-d, Adv. Mater. 17 (2005) 1785. [15] V.V. Kharton, A.L. Shaula, F.M.M. Snijkers, J.F.C. Cooymans, J.J. Luyten, A.A. Yaremchenko, et al., Processing, stability and oxygen permeability of Sr(Fe, Al)O3-d based ceramic membranes, J. Membr. Sci. 252 (2005) 215. [16] A.L. Shaula, V.V. Kharton, N.P. Vyshatko, E.V. Tsipis, M.V. Patrakeev, F.M.B. Marques, et al., Oxygen ionic transport in SrFe1
yAlyO3
d and Sr1
xCaxFe0.5Al0.5O3
d ceramics, J. Eur. Ceram. Soc. 25 (2005) 489. [17] A.V. Kovalevsky, V.V. Kharton, F. Maxim, A.L. Shaula, J.R. Frade, Processing and characterization of La0.5Sr0.5FeO3-supported Sr1
xFe(Al)O3–SrAl2O4 composite membranes, J. Membr. Sci. 278 (2006) 162. [18] T. Ishihara, Y. Tsuruta, T. Todaka, H. Nishiguchi, Y. Takita, Fe doped LaGaO3 perovskite oxides as an oxygen separating membrane for CH4 partial oxidation, Solid State Ionics 152 (2002) 709.

292

Inorganic, Polymeric and Composite Membranes

[19] T. Ishihara, T. Yamada, H. Arikawa, H. Nishiguchi, Y. Takita, Mixed electronic-oxide conductivity and oxygen permeating property of Fe-, Co- or Ni-doped LaGaO3 perovskite oxide, Solid State Ionics 135 (2000) 631. [20] J.H. Tong, W.S. Yang, B.C. Zhu, R. Cai, Investigation of ideal zirconium-doped perovskitetype ceramic membrane materials for oxygen separation, J. Membr. Sci. 203 (2002) 175. [21] X. Gu, W. Jin, C. Chen, N. Xu, J. Shi, YSZ-SrCo0.4Fe0.6O3-d membranes for the partial oxidation of methane to syngas, AIChE J. 48 (2002) 2051. [22] E. Chinarro, J.R. Jurado, F.M. Figueiredo, J.R. Frade, Bulk and grain boundary conductivity of Ca0.97Ti1-xFexO3-d materials, Solid State Ionics 160 (2003) 161. [23] X.F. Zhu, Y. Cong, W.S. Yang, Oxygen permeability and structural stability of BaCe0.15Fe0.85O3-d membranes, J. Membr. Sci. 283 (2006) 38. [24] X.F. Zhu, H.H. Wang, W.S. Yang, Novel cobalt-free oxygen permeable membrane, Chem. Commun. 9 (2004) 1130. [25] X.F. Zhu, H.H. Wang, Y. Cong, W.S. Yang, Partial oxidation of methane to syngas in BaCe0.15Fe0.85O3-d membrane reactors, Catal. Lett. 111 (2006) 179. [26] T.H. Lee, Y.L. Yang, A.J. Jacobson, Electrical conductivity and oxygen permeation of Ag/ BaBi8O13 composites, Solid State Ionics 134 (2000) 331. [27] C.S. Chen, H. Kruidhof, H.J.M. Bouwmeester, Thickness dependence of oxygen permeation through stabilized bismuth oxide-silver composite, Solid State Ionics 99 (1997) 215. [28] C.S. Chen, B.A. Boukamp, H.J.M. Bouwmeester, G.Z. Cao, H. Kruidhof, A.J.A. Winnubst, et al., Microstructural development, electrical properties and oxygen permeation of zirconia-palladium composites, Solid State Ionics 76 (1995) 23. [29] V.V. Kharton, A.V. Kovalevsky, A.P. Viskup, A.A. Yaremchenko, E.N. Naumovich, F.M.B. Marques, Oxygen permeability of Ce0.8Gd0.2O2-d-La0.7Sr0.3MnO3-d composite membranes, J. Electrochem. Soc. 147 (2000) 2814. [30] V.V. Kharton, A.V. Kovalevsky, A.P. Viskup, A.L. Shaula, F.M. Figueiredo, E.N. Naumovich, et al., Oxygen transport in Ce0.8Gd0.2O2-d based composite membranes, Solid State Ionics 160 (2003) 247. [31] J.D. Sirman, J.C. Chen, Ceramic membrane structure and oxygen separation method, (2003) US Patent, 6,514,314, Assigned to Eltron Research & Development Inc. [32] U. Nigge, H.D. Wiemhofer, E.W.J. Romer, H.J.M. Bouwmeester, T.R. Schulte, Composites of Ce0.8Gd0.2O1.9 and Gd0.7Ca0.3CoO3-delta as oxygen permeable membranes for exhaust gas sensors, Solid State Ionics 146 (2002) 163. [33] A.L. Shaula, V.V. Kharton, F.M.B. Marques, Phase interaction and oxygen transport in La0.8Sr0.2Fe0.8Co0.2O3–(La0.9Sr0.1)0.98Ga0.8Mg0.2O3 composites, J. Eur. Ceram. Soc. 24 (2004) 2631. [34] J. Yi, Y. Zuo, W. Liu, L. Winnubst, C. Chen, Oxygen permeation through a Ce0.8Sm0.2O2-dLa0.8Sr0.2CrO3-d dual-phase composite membrane, J. Membr. Sci. 280 (2006) 849. [35] B. Wang, M.C. Zhan, D.C. Zhu, W. Liu, C.S. Chen, Oxygen permeation and stability of Zr0.8Y0.2O1.9-La0.8Sr0.2CrO3-delta dual-phase composite, J. Solid State Electrochem. 10 (2006) 625. [36] X.F. Zhu, W.S. Yang, Composite membrane based on ionic conductor and mixed conductor for oxygen permeation, AIChE J. 54 (2008) 665. [37] X.F. Zhu, Q.M. Li, Y. Cong, W.S. Yang, Syngas generation in a membrane reactor with a highly stable ceramic composite membrane, Catal. Comm. 10 (2008) 309. [38] X.F. Zhu, H.H. Wang, W.S. Yang, Relationship between homogeneity and oxygen permeability of composite membranes, J. Membr. Sci. 309 (2008) 120.

Chapter

12

Critical Factors Affecting Oxygen Permeation

293

[39] Q.M. Li, X.F. Zhu, W.S. Yang, Single-step fabrication of asymmetric dual-phase composite membranes for oxygen separation, J. Membr. Sci. 325 (2008) 11. [40] X.F. Zhu, Q.M. Li, Y.F. He, Y. Cong, W.S. Yang, Oxygen permeation and partial oxidation of methane in dual-phase membrane reactors, J. Membr. Sci. 360 (2010) 454. [41] Q.M. Li, X.F. Zhu, W.S. Yang, Effects of sintering temperature on properties of dual-phase oxygen permeable membranes, J. Membr. Sci. 367 (2011) 134. [42] D. Li, X.P. Wang, Z. Zhuang, J.X. Wang, C. Li, Q.F. Fang, Reducibility study of oxide-ion conductors La(2-x)Ba(x)Mo(2-y)A(y)O(9-delta) (A ¼ W, Al, Ga) assessed by impedance spectroscopy, Mater. Res. Bull. 44 (2009) 446. [43] M.S.D. Read, M.S. Islam, G.W. Watson, F. King, F.E. Hancock, Defect chemistry and surface properties of LaCoO3, J. Mater. Chem. 10 (2000) 2298. [44] M.S. Islam, R.A. Davies, Atomistic study of dopant site-selectivity and defect association in the lanthanum gallate perovskite, J. Mater. Chem. 14 (2004) 86. [45] R. Peng, C. Xia, Q. Fu, G. Meng, D. Peng, Sintering and electrical properties of (CeO2)0.8(Sm2O3)0.1powders prepared by glycine–nitrate process, Mater. Lett. 56 (2002) 1043. [46] X.F. Zhu, Development and applications of novel high stability oxygen permeable ceramic membranes, (2006). PhD thesis, Dalian Institute of Chemical Physics. [47] V.V. Kharton, F.M.B. Marques, Mixed ionic-electronic conductors: effects of ceramic microstrucutre on transport properties, Curr. Opin. Solid State Mater. Sci. 6 (2002) 261. [48] L. Tan, X. Gu, L. Yang, W. Jin, L. Xiong, N. Xu, Influence of powder synthesis methods on microstructure and oxygen permeation performance of Ba0.5Sr0.5Co0.8Fe0.2O3-d perovskitetype membranes, J. Membr. Sci. 212 (2003) 157. [49] Y. Xia, T. Armstrong, F. Prado, A. Manthiram, Sol-gel synthesis, phase relationship, and oxygen permeation properties of Sr4Fe6-xCoxO13þd (0x3), Solid State Ionics 130 (2000) 81. [50] Z.Q. Deng, W. Liu, D.K. Peng, C.S. Chen, W.S. Yang, Combustion synthesis, annealing, and oxygen permeation properties of SrFeCo0.5Oy membranes, Mater. Res. Bull. 39 (2004) 963. [51] H.H. Wang, C. Tablet, A. Feldhoff, J. Caro, Investigation of phase structure, sintering, and permeability of perovskite-type Ba0.5Sr0.5Co0.8Fe0.2O3
d membranes, J. Membr. Sci. 262 (2005) 20. [52] Q.M. Li, The novel mixed conducting oxygen pereameble membrane and prepration of dualphase asymmetric membrane, (2009). PhD thesis, Dalian Institute of Chemical Physics.

This page intentionally left blank

Chapter 13

High Temperature Gas Separations Using High Performance Polymers John R. Klaehn1,*, Christopher J. Orme1, Eric S. Peterson1, Frederick F. Stewart1 and Jagoda M. Urban-Klaehn2 1

Idaho National Laboratory, Idaho Falls, Idaho, USA Idaho Accelerator Center, Pocatello, Idaho, USA * Corresponding author: E-mail address: [email protected] 2

INTRODUCTION Polymer membranes are effective, economic, and flexible for many gas separations. For example, the commercial use of polymer membranes for air separation, the recovery of hydrogen (H2) from nitrogen (N2) and methane (CH4) mixtures, and the removal of carbon dioxide (CO2) from natural gas has been widely reported [1–13]. In each of these applications, high fluxes and excellent selectivities have relied on glassy polymer membranes. This high selectivity is due to the combination of size-based and solubility-based molecular separation, a “dual mode” mechanism not shared by size-driven inorganic membrane separations. This “dual mode” mechanism can be derived from the basic relationship of gas transport properties: permeability (P), solubility (S), and diffusivity (D) (P ¼ DS) [1–20]. The solubility factor (S) is controlled by the intimate interactions between the gases and the polymer matrix (sorption). Diffusivity (D) is controlled by the pressure-induced transport of gases through the polymer matrix (molecular sieving). Permeability (P) is the resulting factor for overall gas transport. Most of the high performance (HP) polymers have inherently high glass transitions (Tg) that should limit the interactions of the polymer matrix and the gases; however, solubility can still have an effect. To date, gas separation research using glassy HP polymer membranes has focused on optimizing materials at near ambient conditions. Therefore, the development of high temperature polymeric membranes for gas separations is a vital contribution for economic and environmental reasons. Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

295

296

HN

Inorganic, Polymeric and Composite Membranes

O

O

N C H

C n

Nomex®

O

O

N

N

O

O

N

H N

N H

N

O n

n

Kapton®

Polyamides (Nylon, Nomex®)

PBI

Polyimides (Kapton®, Matrimid®, VTEC)

Polybenzazoles (Polybenzimidazole [PBI], PBO)

N P

P

N

N P

R

R

N P

N

N

R

P

P

N

P

P

O

CH3

N

P

O

O CH3

N

S O

n

Poly (ether sulfone) Cyclomatrix Phosphazenes (Cyclomatrix derivatives)

Polysulfones (Polyphenylsulfone)

FIGURE 13.1 High-performance (HP) polymer structures.

High temperature regimes (> 150  C) are reserved for a few polymers that can operate and perform effectively at these temperature. Some of the best polymers that operate effectively at these high temperatures are glassy polymers that contain aromatic groups in their backbones. These glassy aromatic-based polymers are commonly called HP polymers. HP polymers (see Figure 13.1) are typically glassy polymers that include five very general polymer families: polyazoles (polybenzimidazole [PBI]), phosphazenes (cyclomatrix derivatives), sulfur containing polymers (sulfones, phenylsulfones), polyamides (Nylon, Nomex), and polyimides (Kapton, VTEC PI series). Most of these polymers are known to have high thermal stabilities (some having operating temperatures up to 400  C). Their high temperature stabilities allow their use in applications unsuitable for nearly all other organic polymers. Another important attribute, which is needed for some gas separations, is polymer resistance to many organic solvents, including some acids and bases. In addition, many HP polymers are mechanically robust with high compressive strength. Even with such stability and strength, these polymers can be processed into thin membranes, and some of them have properties that allow blending with copolymers and modifiers, as well as maintain solubility in certain solvent systems. Because of the temperature stability and robust nature of these HP polymers, these vital attributes make them ideal candidates for membrane separations. Many glassy polymers are known to be effective for gas separations; however, many gas permeability experiments show low gas fluxes at ambient temperatures. Generally, glassy polymer membranes have higher permeability at increased temperatures; however, as a rule, when gas permeabilities (fluxes) increase, the gas pair separation factors (selectivities) decrease for most polymer membranes. This permeability/selectivity tradeoff is demonstrated using Robeson’s selectivity/permeability correlation plot [14–16], but most of these

Chapter

13

High Temperature Gas Separations

297

data sets were acquired near ambient temperatures. Temperature is not typically shown with this tradeoff; nevertheless, it should be noted when comparing gas permeability/selectivity data with Robeson’s plot. Early reports of high temperature gas separations using HP polymers are found in the literature during the 1970s using binary gas mixtures at about 100  C [21]. From this early information, the data show that glassy polymers can provide good gas separations, even at temperatures beyond ambient. During the past decade, interest in high temperature gas separations (such as CO2) has provided new opportunities for application of HP polymers. Overall, there are a limited number of gas separation studies using HP polymers at high temperatures, but temperatures beyond 150  C were not previously investigated. The investigations described in this chapter concentrate on the HP polymer classes containing the building blocks needed for a successful membrane material. This includes the potential materials to form complex-shaped films utilizing common membrane casting methods. Initial research focused on the modification (N-substitution) of the base PBI, which substantially increases polymer solubility in common organic solvents [22]. From these studies, the development of several other HP polymers, such as the VTEC PI series, was explored. Preliminary studies have shown that these promising HP polymers gave exciting results for high temperature gas separations (this work). The VTEC polymers can be formed into films and tested for their gas transport properties at elevated temperatures (> 250  C). They are attractive polymers for challenging applications, which have shown relatively high resistance to acids and bases as well as elevated temperatures (up to 450  C). It is interesting to note that the N-substituted PBIs could be blended with the VTEC polyimides to give new polymer membranes for our testing procedures. Several experiments were performed using membranes that were processed under different conditions in order to enhance their gas transport properties.

EXPERIMENTAL Instrumentation Thermal analyses were obtained using a TA Instruments Model 2910 differential scanning calorimeter (DSC), TA Instruments Model 2940 thermomechanical analysis (TMA), and a Model 2950 thermogravimetric analyzer (TGA). Dynamic mechanical analysis (DMA) was acquired using a TA Instruments Model Q800.

Permeability Gas Testing Pure gas permeability results were obtained using the pressure-rise time-lag method [17–20]. Membranes were exposed to individual gases, like H2, CH4, and CO2. In a typical experiment, both sides of the membrane are evacuated to an equal vacuum. The test cell is then isolated, and the pressure at zero time

298

Inorganic, Polymeric and Composite Membranes

is used as the baseline. The feed side is exposed to the test gas, and the pressure buildup on the permeate side of the membrane is recorded as a function of time. The two gas transport properties that are determined directly from the pure gas test system are time lag and permeability. The permeability is the rate at which the gas permeates through the membrane after the gas has come to equilibrium in the polymer. The time lag is the time it takes the gas to permeate from the feed side of the membrane to the permeate side and can be used to calculate the diffusivity. In the mixed gas analysis, permeation is determined analytically by gas chromatography rather than barometrically. The mixed gas analysis uses a predefined mixture of gases (percent by volume), like H2, CH4, and CO2 along with helium as the balance gas. The pure gas system can reach about 70  C, whereas mixed gas system can achieve 400  C. The high temperature gas permeation studies (greater than 70  C) were conducted on the mixed gas system.

Positron Annihilation Lifetime Spectroscopy Standard Positron Lifetime Spectroscopy consisted of Na-22 positron source and the samples in so-called sandwich configuration—source was in the middle between two samples of polymers. Two nickel backings 1 mm thick were placed outside the samples to make sure that all positrons are absorbed in this arrangement. Na-22 positron source emits a positron and a high energy gamma ray with the energy 1.27 MeV in coincidence. The high-energy gamma served as the “start” signal for lifetime electronics, and one of the annihilation quanta of 0.511 MeV from the annihilation event (positron–electron annihilation) was the “stop”—after applying an appropriate delay. CFDD (Constant Fraction Differential Discriminators) selected the proper energy, either 1.27 or 0.511 MeV into the appropriate channel, and then the TAC (Time-Amplitude Converter) converted the length of time between the two pulses (“START” and “STOP”) into the amplitude of the pulse. The time between the creation of positron and its subsequent annihilation depends on the microstructure of the sample. PAL spectra were measured for nonheated and heated samples. The polymers were placed in the oven for about 1 h at about 150  C.

RESULTS AND DISCUSSION From previous work, development of polymeric and/or metallic composite membrane structures was needed to achieve the critical combination of high selectivity, high permeability, chemical, and mechanical stabilities, at elevated temperatures. PBI was studied since it met the described criteria. A problem with PBI (resin) is that it has minimal solubility properties in most solvents and requires aggressive refluxing in polar aprotic solvents, like dimethylacetamide (DMAc) to achieve dissolution. This suggests that PBI resin is more crystalline or densely packed, which causes it to be difficult to dissolve. Therefore, it is necessary to break up the packing of the parent PBI for full dissolution. This directed

Chapter

13

299

High Temperature Gas Separations

the investigation to the synthesis of N-substituted PBI polymers. A number of PBI-based polymers were successfully synthesized, characterized, and tested with the following functional groups: RMe2SiCH2–, where R ¼ alkyl, phenyl (Scheme 13.1) [22]. The INL developed synthetic pathway provides moderate yields with a high degree of substitution with organosilane functional groups (namely, trimethylsilylmethyl groups). The yields among the substitutions are 50%, but substitution reactions with larger functional groups (phenyl) provided low yields, possibly due to steric hindrance. The N-substituted PBI polymers are soluble in common organic solvents, such as tetrahydrofuran (THF) and chloroform, which is exciting since there are a limited number of literature citations describing soluble PBI materials [23,24]. The N-substituted PBI polymers can be readily used for film casting. However, when N-substituted PBI gas permeation analyses were completed, the data showed increased fluxes coupled with decreased gas pair separation factors compared to the parent PBI (Klaehn, J. R, Orme, C. J, Peterson, E. S. unpublished results). Thus, N-substituted PBI polymers showed no advantage for high temperature gas separations; however, these substituted polymers are shown to be useful for blending with their parent PBI or other HP polymers.

HN

N

NaH (2 eq.) DMAc/room temp.

N

N H

n

(R)Me2SiCH2CI (xs) -N

N-

DMAc/room temp. N

N

R = Methyl, Phenyl, Vinyl, Allyl, Hexyl, Decyl n

(1) R = Methyl (2) R = Phenyl (3) R = Vinyl (4) R = Allyl (5) R = Hexyl (6) R = Decyl

(R)Me2SiH2C N N

N N

(R)Me2SiH2C n

SCHEME 13.1 Synthetic route for N-substituent modification of polybenzimidazole with organosilane groups.

300

Inorganic, Polymeric and Composite Membranes

Due to the limitations of N-substituted PBI, the focus of these investigations changed to polyimides because they offer better film forming properties coupled with their intrinsic HP nature. We selected a commercially available material with the trade name of “VTEC.” The VTEC PI series of polymers are sold as solutions of polyamic acids, which can be easily cast into films. The films have to be heat treated, as described in Scheme 13.2, which completes the condensation reaction forming the polyimide while eliminating water as a by-product. The polyimide film that is formed is highly heat resistant and virtually insoluble in common organic solvents. The VTEC polyimide film possesses thermal properties similar to those of PBI (up to 550  C). These films are robust and flexible even after three thermal cycles (up to 400  C). VTEC polyimides are manufactured for various high temperature applications, like enamel coatings for wires and metal, but they have not been investigated for gas permeation membrane applications. The major problem for polyimides is their low solubility in common solvents, thus making membrane development a serious problem because film casting can be difficult. VTEC polyimides have very good thermal resistance, chemical stability, and mechanical stability at high temperature. Interestingly, VTEC polymer films do not discolor after being heat processed in a convection oven at 250  C for 24 h; most organic polymers degrade at these high temperatures while they are exposed to the atmosphere for extended periods. They are attractive polymers for challenging applications, which have shown relatively high

O

OH O N H

H N

Heating

O

−2H2O

O HO

n

O

Polyamic acid (Prepolymer) to Kapton®

O

O

N

N

O

O

O

n

Kapton® SCHEME 13.2 Polyamic acid (prepolymer) goes through a condensation reaction yielding the polyimide polymer.

Chapter

13

High Temperature Gas Separations

301

resistance to acids and bases as well as elevated temperatures (up to 450  C). The current work shows that the VTEC PI series of polymers can be blended with many different types of polymers and additives, like N-substituted PBI and functionalized polyhedral oligosilsesquioxanes (POSS), without difficulty. In some cases, the parent VTEC was blended with a secondary polymer, polyethylene glycol (PEG), without physical property loss of the polymer. This is significant because it demonstrates the stability of the parent VTEC polymers even when blended with polymers that are known to be unstable at high temperature (> 250  C). The resulting gas transport data from VTEC PI series polymers are quite interesting and competitive with other HP gas separation polymers; however, it became apparent that polymer processing methodologies for membrane preparation (free-standing films) are critical for consistent gas permeation data. Thus, a heat treatment regimen was developed that gives consistent gas permeation data. The regimen consists of polymer film formation on glass plates, evaporating the solvent from the film over night, heating to 100  C to remove any water from the system. Then, the films are heated to 150  C for 24 h to remove any additional solvent from the system, followed by continued heating to 220–250  C (for 6–24 h) to cure the polyimide. The films were lifted from the glass plates using water to give free-standing films, the films were finally dried at 150  C. Gas permeability analyses were conducted immediately following the final drying step to avoid water sorption from the atmosphere. The thermal studies with these polymers are shown in Tables 13.1 and 13.2. The onset of decomposition (under nitrogen) for the VTEC PI series of polymers shows similar decomposition temperatures to KaptonÒ (Table 13.1), and the decomposition temperature does not change dramatically when combining various additives or polymers (as long as they are stable at higher temperatures). Thermogravimetric analysis (TGA) data was obtained for the VTEC polyamic acid (unprocessed polyimide) and TMA data for the thermal processed VTEC polyimide films (Table 13.2). All of the VTEC polymers exhibit high thermal decomposition temperatures (> 500  C), and the TGA shows that the evolution of water (polyimide condensation reaction) occurs above 240  C. Differential scanning calorimetry (DSC) shows subtle thermal features with the processed films; therefore, it is not useful for finding thermal transitions with these polymers. Instead, the TMA data for the VTEC polymers show small dimensional changes when heated to 400  C with thermal transitions at about 250  C. The observed thermal feature can be identified as a glass transition, but similar thermal features are not obvious in the DSC data. DMA, shown in Figure 13.2, provides better evidence for thermal transitions on VTEC polyimides. When thin strips of the processed polyimide films are used, DMA provides an apparent thermal transition (changes in the storage modulus and loss modulus). The thermal transition is identified as the glass transition for these polymers. The thermal transition differences between the TMA and DMA data could be due to polymer processing issues, but it is found

302

Inorganic, Polymeric and Composite Membranes

TABLE 13.1 Thermal Decomposition Temperatures of Various High Performance Polymers Determined by Thermogravimetric Analysis (TGA)

Polymer Films

TGA Onset of Decomposition— Under Nitrogen ( C)

KaptonÒ (DuPont)

574

VTEC PI 80-051

524

VTEC PI 851

512

VTEC PI 1388

529

PBI (Celanese; PBI Performance Products, Inc.)

512

VTEC PI 80-051 w/10 wt% Bromine

551

VTEC PI 80-051 w/20 wt% PBI–TMS (N-substituted PBI)

465

PBI–TMS (Me3SiCH2–PBI) (N-substituted PBI)

448

TABLE 13.2 Thermogravimetric Analysis (TGA), Thermomechanical Analysis (TMA), and Dynamic Mechanical Analysis (DMA) of VTEC Polyimide

Polymer Films

TGA Weight Loss under N2 Before Heat Treatment; Onset to Decomposition

TMA Thermal Transition; Heat DMA Thermal Processed Films Transition; Heat (Dimensional Change) Processed Films

VTEC PI 851

243  C (–H2O); 512  C

243  C (2.5 mm; to 400  C)

265  C

VTEC PI 80-051 237  C (–H2O); 524  C

255  C (5.0 mm; to 400  C)

290  C

241  C (–H2O); 529  C

285  C (1.0 mm; to 400  C)

286  C

VTEC PI 1388

that DMA gives more distinct thermal transitions. Therefore, DMA is preferred for observing thermal transitions on these polymers. Table 13.3 contains pure gas permeation results and explains the fundamental gas transport properties of the polymers under ideal conditions. These HP polymers at ambient temperatures (Table 13.3) exhibit low permeabilities; however, the focus of these studies is high temperature gas separations. The permeation rate differences among these polymers are somewhat subtle, regardless if it is in the native state or as blends or with additives. Most of the blended

Chapter

13

303

High Temperature Gas Separations

200

2500 297.30 °C 285.18 °C

2000

0.8

1500 Tan detta

0.6 275.06 °C

1000

100 0.4

284.17 °C(I)

Loss modulus (MPa)

Storage modulus (MPa)

150

50 0.2

500 292.49 °C

0 400

0 0

50

100

150

200

250

300

350

Universal V3.8B TA Instruments

Temperature (°C)

FIGURE 13.2 Dynamic mechanical analysis of VTEC 80-051 parent polymer film.

TABLE 13.3 Pure Gas Data Collected at 30  C Permeabilitya

Selectivity a

Ar

N2

O2

Kapton 30 mm film

1.56

–b

0.01

0.075 0.008 0.297 5.3

195.0

VTEC PI 80–851

3.56

0.20

0.06

0.19

0.03

0.48

7.4

121.6

VTEC PI 1388

3.97

0.06

0.04

0.17

0.05

0.53

7.5

79.4

PBI VTEC blend PI 851 w/CNT

2.17 4.6

b

0.017 0.075 0.012 0.300 7.2

180.8

b

0.12

35.4

– –

b

2.050 –

VTEC blend PI 1388 w/POSS 6.5

–

CO2

H2/ CH4

H2

VTEC PI 851

CH4

H2/ CO2

Polymer

b

–

b

–

b

0.29

0.13

0.85

5.4

–

b

0.280 0.630 3.3

7.3

–

b

0.8

3.1

8.1

2.1

VTEC blend 80-051 w/PEG

1.392 0.017 0.009 0.07

0.006 0.25

5.6

232

VTEC blend PI 80–851/PBI

3.45

0.05

0.02

0.17

0.02

0.61

5.7

172.5

VTEC blend PI 1388/PBI

3.12

0.05

0.04

0.15

0.03

0.49

6.4

104

a

2

Permeabilities measured in mol m Not tested.

b

s Pa.

304

Inorganic, Polymeric and Composite Membranes

VTEC polymers in Table 13.3 show neither improved permeability nor improved selectivity. However, pure gas testing with carbon nanotubes (CNT) and VTEC PI 851 (Table 13.3) observes slightly better CO2 gas permeability, but the selectivity did not improve over the parent polymer’s native selectivity. However, the VTEC-PEG blend showed a reduction in permeability for all of the gases tested. Table 13.4 contains data that were taken at 250  C using a gas mixture of H2, CO2, and CH4. The gas permeabilities are higher at 250  C; however, the differences in permeability are similar for most of the HP polymer blends. The most visible differences in gas permeability at elevated temperatures is VTEC PI 851 blended with bis(trimethylsilylmethyl)polybenzimidazole (PBI–TMS). Remarkably, this polymer blend gave better permeabilities and selectivities over the native VTEC polymer. Overall, the most intriguing component of the ambient temperature and high temperature data is the gas pair selectivity for H2/CO2, which remains fairly constant for the pure and mixed gas data. This behavior is unusual, since many polymeric membranes typically lose their selectivities as temperature increases. A critical variable when working with the glassy polymer membranes is their moisture content [9,25–27]. When the gas permeation rates are low as in these cases, moisture content plays a major role with the gas selectivity. It has been observed that water trapped within the polymer matrix (either as left over solvent, or physisorbed) can cause the polymer’s performance to change dramatically. At ambient temperatures, moisture in the polymer generally decreases the gas selectivity, but a thoroughly dried membrane gives steady performance. Therefore, additional effort was made to dry the films prior to testing by drying them at 150  C for longer periods (24–48 h). After drying, the

TABLE 13.4 Mixed Gas Data Collected at 250  C Permeabilitya H2

Polymer

Ar

Selectivity a

CH4

CO2

H2/CO2

H2/CH4

b

4.1

11.8

3.2

9.2

Kapton 30 mm film

37.5

–

VTEC PI 80-851

83.0

3.1

2.3

9.3

8.9

36.1

57.9

–

b

6.6

6.2

9.4

8.7

b

1.69

4.7

7.1

19.5

VTEC PI 851 PBI

33.1

–

VTEC blend PI 1388/PBI

30.5

1.1

0.8

4.0

7.6

38.1

99.9

b

6.9

11.5

8.7

14.5

VTEC blend PI 851/PBI–TMS a

2

Permeabilities measured in mol m Not tested.

b

s Pa.

–

Chapter

13

305

High Temperature Gas Separations

membrane was promptly transferred to the gas testing apparatus and placed under vacuum for the gas permeation evaluation. The presence of molecular water in the polymers’ void volume has been validated using Positron Annihilation Lifetime (PAL) spectroscopy for VTEC 80-051 and 1388 (Table 13.5). The PALS measurements give small differences between the lifetimes and intensities of the spectra for the unheated and heated samples. Since the microporosity of these samples is low compared to other open volume polymers, their longest lifetime and intensity values are relatively small, thus their values have relatively high uncertainty. In the case of spectra with less than 1 million counts in the spectrum, the error bars are relatively large. Therefore, the best indicator of any possible change in the structure is the product of both intensity and lifetime. Since the free volume fraction is calculated based on both—lifetime (pore size) and intensity (pore frequency, rate)—their product is the most reliable value to consider while comparing the unheated and heated spectra of polymers. It is interesting to note that the free volume fraction for VTEC polymers increased after the polymer was heated. It can be postulated that this may be caused by more positrons being trapped inside the larger open volume of the polymers due to release of bound water. In support of this conclusion, the intensity (I3) was also observed to increase, suggesting greater open pore frequency (less water) as well.

TABLE 13.5 Estimation of Free Volume from PAL Measurements Polymer

tL

tL error IL%

IL Pore Pore Free Volume Error Radius [A˚] Vol. [A˚3] Fractionc

VTEC PI 80-051a 2.15 0.67

2.6

1.6

3.01 114 0.53 (0.10–1.42) (2.33–3.55) (53–187)

VTEC PI 80-051b 1.8

0.26

4.42

0.86

2.69 (2.4–2.93)

VTEC PI 1388a

1.5

0.11

3.75

0.31

2.35 54 (2.23–2.48) (46–64)

0.37 (0.29–0.47)

VTEC PI 1388b

1.74 0.14

3.07

0.34

2.62 75 (2.45–2.76) (62–88)

0.42(0.30–0.54)

a

82 0.65 (0.37–1.00) (58–105)

Normal atmospheric humidity exposure (20–40% relative humidity). Heated at 150  C for 1 h. Free volume fraction calculated according to the formula: fv(%)¼0.0018  I3(%)  h4/3pR3i, where R ¼ f(tL). b c

306

Inorganic, Polymeric and Composite Membranes

CONCLUSIONS Several HP glassy polymers have been shown to be good candidates for high temperature gas separations. N-substitution of PBI provides a straightforward synthetic method for increasing PBI’s solubility in common organic solvents. However, no advantage for high temperature gas separations was demonstrated. Instead, a series of polyimide (VTEC PI series) membrane materials show attractive gas separations at higher temperatures (250  C). The VTEC polymers may be fully processed into free-standing films under normal membrane formation conditions. The VTEC polyimides have been blended with other materials and polymers, and the resulting blends show different gas permeation behavior than the parent VTEC films. Interestingly, the parent VTEC membranes have similar H2/CO2 factors at both 30 and 250  C. Finally, it has been found that water entrapped within the polymer matrix (left over from film processing or physisorbed from the atmosphere) can also cause the polymer’s fluxes and selectivities to change dramatically. PAL spectroscopy was used to detect the presence of molecular water in the polymer’s void volume. Overall, several HP polymers, including VTEC, show potential for performing gas separations at high temperatures.

ACKNOWLEDGMENTS The work described in this paper was supported by Battelle Energy Alliance, LLC and Battelle Memorial Institute (BMI), Department of Energy’s National Energy Technology Laboratory Fossil Energy Program (DOE-NETL-FE) through Contract DE-AC07-05ID14517. The work was supported also by National Science Foundation’s Internal Research and Development (IR&D) Program and Laboratory Directed Research and Development (LDRD) Program at the Idaho National Laboratory.

REFERENCES [1] A. Brunetti, F. Scuraa, G. Barbieri, E. Drioli, Membrane technologies for CO2 separation, J. Membr. Sci. 359 (2010) 115. [2] C.E. Powell, G.G. Qiao, Polymeric CO2/N2 gas separation membranes for the capture of carbon dioxide from power plant flue gases, J. Membr. Sci. 279 (2006) 1. [3] B.D. Freeman, I. Pinnau, Gas and liquid separations using membranes: an overview, Advanced Materials for Membrane Separations, ACS Symposium Series, vol. 876, (2004) Chapter 1, p. 1. [4] D. Ayala, A.E. Lozano, J. de Abajo, C. Garcı´a-Perez, J.G. de la Campa, K.-V. Peinemann, et al., Gas separation properties of aromatic polyimides, J. Membr. Sci. 215 (2003) 61. [5] B.D. Freeman, I. Pinnau, Polymeric materials for gas separations, Polymer Membranes for Gas and Vapor Separation, ACS Symposium Series, vol. 733, (1999), Chapter 1, p. 1. [6] B.D. Freeman, A.J. Hill, Free Volume and Transport Properties of Barrier and Membrane Polymers, Structure and Properties of Glassy Polymers, ACS Symposium Series, vol. 710, (1999) Chapter 21, p. 306. [7] Y. Hirayama, T. Yoshinaga, Y. Kusuki, K. Ninomiya, T. Sakakibara, T. Tamari, Relation of gas permeability with structure of aromatic polyimides I, J. Membr. Sci. 111 (1996) 169.

Chapter

13

High Temperature Gas Separations

307

[8] A. Morisato, K. Ghosal, B.D. Freeman, R.T. Chern, J.C. Alvarez, J.G. de la Campa, et al., Gas separation properties of aromatic polyamides containing hexafluoroisopropylidene groups, J. Membr. Sci. 104 (1995) 231. [9] L.M. Robeson, W.F. Burgoyne, M. Langsam, A.C. Savoca, C.F. Tien, High performance polymers for membrane separation, Polymer 35 (1994) 4970. [10] W.J. Koros, G.K. Fleming, Membrane-based gas separation, J. Membr. Sci. 83 (1993) 1. [11] K. Tanaka, H. Kita, M. Okano, K. Okamoto, Permeability and permselectivity of gases in fluorinated and nonfluorinated polyimides, Polymer 33 (1992) 585. [12] T.H. Kim, W.J. Koros, G.R. Husk, K.C. O’Brien, Relationship between gas separation properties and chemical structure in a series of aromatic polyimides, J. Membr. Sci. 37 (1988) 45. [13] V.T. Stannett, W.J. Koros, D.R. Paul, H.K. Lonsdale, R.W. Baker, Recent advances in membrane science and technology, Adv. Poly. Sci. 32 (1979) 69. [14] L.M. Robeson, Correlation of separation factor versus permeability for polymeric membranes, J. Membr. Sci. 62 (1991) 165. [15] B.D. Freeman, Basis of permeability/selectivity tradeoff relations in polymeric gas separation membranes, Macromolecules 32 (1999) 375. [16] L.M. Robeson, B.D. Freeman, D.R. Paul, B.W. Rowe, An empirical correlation of gas permeability and permselectivity in polymers and its theoretical basis, J. Membr. Sci. 341 (2009) 178. [17] R.M. Barrer, E.K. Rideal, Permeation, diffusion and solution of gases in organic polymers, Trans. Faraday Soc. 35 (1939) 628. [18] J.G. Wijmans, R.W. Baker, The solution-diffusion model: a review, J. Membr. Sci. 107 (1995) 1. [19] C. Rogers, J.A. Meyer, V. Stannett, M. Szwarc, Studies in the gas and vapor permeability of plastic films and coated papers, Parts I And II, TAPPI 39 (1956) 737. [20] G.J. van Amerongen, Permeation of gases through solids, J. Appl. Phys. 17 (1946) 972. [21] F.P. McCandless, Separation of binary mixtures of CO and H2 by permeation through polymeric films, Ind. Eng. Chem. Proc. Des. Dev. 11 (1972) 470. [22] J.R. Klaehn, C.J. Orme, T.A. Luther, M.G. Jones, A.K. Wertsching, E.S. Peterson, Soluble N-substituted organosilane polybenzimidazoles, Macromolecules 40 (2007) 7487. [23] M.B. Gieselman, J.R. Reynolds, Water-soluble polybenzimidazole-based polyelectrolytes, Macromolecules 25 (1992) 4832. [24] M.B. Gieselman, J.R. Reynolds, Aramid and imidazole based polyelectrolytes: physical properties and ternary phase behavior with poly(benzobisthiazole) in methanesulfonic acid, Macromolecules 26 (1993) 5633. [25] W.J. Koros, R.T. Chern, V. Stannett, H.B. Hopfenberg, A model for permeation of mixed gases and vapors in glassy polymers, J. Polym. Sci. Polym. Phys. Ed. 19 (1981) 1513. [26] R.T. Chern, W.J. Koros, E.S. Sanders, R. Yui, Second component effects in sorption and permeation of gases in glassy polymers, J. Membr. Sci. 15 (1983) 157. [27] Y. Hayashi, S. Sugiyama, T. Kawanishi, N. Shimizu, Kinetics of sorption and permeation of water in glassy polyimide, J. Membr. Sci. 156 (1999) 11.

This page intentionally left blank

Chapter 14

Using First-principles Models to Advance Development of Metal Membranes for High Temperature Hydrogen Purification Sunggu Kang, Shiqiang Hao and David S. Sholl* School of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA * Corresponding author: E-mail address: [email protected]

INTRODUCTION Hydrogen is a promising fuel source that is attractive as an alternative to fossil fuels. Hydrogen is an abundant elementary resource with huge energy capacity [1]. Pure hydrogen does not exist in nature, so it must come from other resources. Broadly, there are two possible hydrogen production processes: water electrolysis and extraction from fossil fuels. Since water splitting is energy intensive, hydrogen is most economically obtained by reforming of hydrocarbon sources with steam or partial oxidation of hydrocarbons from fossil fuels. By one estimate, the cost of hydrogen produced by electrolysis of water is three to five times higher than the hydrogen obtained from fossil fuels [2]. When hydrogen is produced from fossil fuels, it is necessary to purify hydrogen from mixed gas streams containing components such as CO, CO2, CH4, and H2S [3]. One important challenge related to hydrogen purification in this context is the development of membranes that can operate at elevated temperatures and pressures. Membranes should provide high fluxes, resistance to poisoning, and long operational lifetime with effective cost. Metal membranes are well suited for high-temperature applications [4]. Pd-based metal membranes have received much attention for H2 purification because, in principal, they have perfect selectivity for H2 over other gas species. By separating H from a CO2-rich Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

309

310

Inorganic, Polymeric and Composite Membranes

stream, Pd membranes could be helpful in carbon sequestration from gasification processes [5]. Pure Pd membranes, however, are prone to H2-induced embrittlement at temperatures below 300
C and to sulfur poisoning. To improve the performance of pure metal membranes, metal alloys can be considered. Pd has been the core material for binary alloy membranes, in combination with elements including boron, cerium, copper, gold, iron, nickel, platinum, rhodium, silver, yttrium, among others [5]. Alloying Pd with other metals such as Ag, Au, or Cu reduces the embrittlement problem and in some cases improves resistance to poisoning [6]. PdCu alloys have attracted particular attention as sulfur tolerant alloys [7–9]. Using alloys can also increase the permeability of a membrane relative to pure Pd. Gryaznov has listed a number of alloys which exhibit hydrogen permeability larger than pure Pd, including Pd90Ag10 (Palloy/PPd ¼ 1.5), Pd80Ag20 (Palloy/ PPd ¼ 1.6), Pd70Ag30 (Palloy/PPd ¼ 1.8), Pd60Ag40 (Palloy/PPd ¼ 1.8), Pd95Au5 (Palloy/PPd ¼ 2.0), Pd90Au10 (Palloy/PPd ¼ 2.2), Pd85Au15 (Palloy/PPd ¼ 2.1), Pd80Au20 (Palloy/PPd ¼ 2.0), Pd90Pt10 (Palloy/PPd ¼ 1.2), Pd95Rh5 (Palloy/ PPd ¼ 1.4), Pd95.5Ru4.5 (Palloy/PPd ¼ 1.4) [10]. Alloying PdAg with Y and Gd showed higher permeability than pure Pd as a ternary alloy. A range of alloys, PdYxAgy, PdGdxAgy with x < 7 and y < 20 showed better permeabilities than Pd76Ag24 alloys [11]. Yukawa et al. found that hydrogen diffusion is faster in fcc Pd–Ag alloy than in bcc pure niobium at T ¼ 773 K [12]. Way et al. demonstrated that adding Ru leads to membranes with high hydrogen flux at 673 K [13]. Dense metal membranes are highly selective for hydrogen because hydrogen is transported across the membrane as atomic hydrogen in the metal’s interstitial sites. For net transport through a membrane, H2 must dissociate on the membrane surface, atomic hydrogen must hop from the membrane surface to a subsurface interstitial site and subsequently diffuse through the bulk of the membrane. This process is then reversed on the downstream side, with subsurface hydrogen hopping onto the membrane surface followed by recombinative desorption of H2. This mechanism for hydrogen permeation indicates that a good membrane must have high catalytic activity for H2 dissociation, a significant level of solubility for atomic H in the metal’s interstitial sites, and rapid diffusion of interstitial H in the bulk metal. The ability of a membrane to transport hydrogen can be quantified in terms of either flux or permeability. The H flux through a membrane is obtained from Fick’s first law as the product of the diffusion coefficient and the concentration gradient across the membrane. For thick membranes, the rate-limiting step is the transport of H atoms through the membrane [14]. In this case, the surface reaction is sufficiently rapid for the dissolved H atoms at the surface to be in equilibrium with gaseous H2 on either side of the membrane. The permeability k through a membrane can be obtained from [5,8] k¼

JL 1=2 ðPfeed

1=2

 Pperm Þ

:

ð14:1Þ

Chapter

14

First-principles Models of Metal Membranes

311

Here, Pfeed (Pperm) is the H2 pressure on the feed side (permeate side) of the membrane, J is the H2 flux, and L is the thickness of the membrane. In many crystalline materials, the solubility of H under conditions of practical interest for high temperature separations satisfies Sieverts’ law. In this case, the interstitial concentration, c, satisfies c ¼ KsP1/2, where Ks is the Sieverts’ constant. When this is true, and diffusion of H through the bulk of the membrane is the rate-limiting step of the process, k is independent of the feed and permeate pressures and can be written as [15] 1 k ¼ DKs ; 2

ð14:2Þ

where D is the diffusion coefficient of interstitial H. Experimentally, the development and characterization of new metal membranes requires significant investments of resources and time. Thus, it has proved useful to develop theoretical methods to identify alloys and other materials with promising properties as membranes. Theoretical predictions have provided an effective complement to experiment in the development of practical metal membranes for H2 purification. In order for these approaches to be useful, they must provide quantitatively reliable information without requiring parameterization from experimental data. Density Functional Theory (DFT), a first principles quantum chemistry method, provides a useful basis for calculations of this kind. DFT calculations have been shown to be useful for predicting H diffusion and permeability in numerous dense materials, including pure metals [8], metal alloys [9], intermetallic compounds [16], and Pd4S [17]. DFT calculations have also been used to successfully describe the adsorption and diffusion behavior of H on the surfaces and in subsurface layers of metals and metal alloys without any experimental input [9,18]. For a general introduction to DFT methods, the reader is referred to the book by Sholl and Steckel [19]. In the remainder of this chapter, we review work over recent years to use DFT-based methods to predict properties of a variety of dense metal materials as H-selective membranes. Work in this field originated with Kamakoti and Sholl, who developed a theoretical framework to predict macroscopic properties of crystalline metals as H-selective membranes using a hierarchical approach based on DFT calculations and coarse grained lattice gas modeling [8,9,20]. First, we summarize work based on this general approach that has examined a range of Pd-based alloys in crystalline form. We then describe how these methods have recently been extended to predict the properties of amorphous metals as membranes.

DFT-BASED MODELING OF CRYSTALLINE METAL MEMBRANES DFT calculations can give accurate information about the energies of systems with small numbers of atoms (10s to 100s of atoms). It is therefore critical when using DFT calculations to describe metal membranes that these calculations be

312

Inorganic, Polymeric and Composite Membranes

combined with a coarse-graining approach that leads to a meaningful description of macroscopic permeation. Below, we outline the ideas required to achieve this goal. In all of the calculations described below, plane wave DFT calculations [19] using VASP [21] were used. These calculations represent a bulk material using a computational supercell replicated in three dimensions with periodic boundary conditions. Technical details of these calculations are available in the references cited below. Briefly, the calculations use the PW91 generalized gradient approximation functional [22], a plane wave expansion, and are well converged in k space. For a crystalline metal material, it is not difficult to locate the interstitial sites that can be occupied by H. The binding energy of H in each site is defined by [7] 1 1 ZP Eb ¼ Ehost=H  Ehost  EH2 þ EZP host=H  EH2 ; 2 2

ð14:3Þ

where Ehost(Ehost/H) is the DFT-calculated energy of the system without (with) atomic H in the host lattice, EH2 is the energy of a free H2 molecule, and EZP host=H (EZP H2 ) is the zero point energy contribution from H in the host lattice (in a free molecule). Zero point energies were computed in the harmonic approximation, and for interstitial H, we assume that vibrations of H were decoupled from lattice phonons [20]. Once the binding energies of each interstitial site are available, the net solubility of H in the alloy can be calculated. As mentioned above, the solubility of dilute amounts of atomic H can be found using Sieverts’ law [14], which relates the interstitial concentration of atomic H to the gas phase H2 pressure by c ¼ KsP1/2 where Ks is the Sieverts’ constant [23]. The Sieverts’ constant for an individual site with a known binding energy Ks,ind can be calculated by considering the zero point energy corrected binding energy for the interstitial site and the translational and rotational effects in the partition function of gas phase H2 [20]. The net Ks is then found by summing over the Ks,ind for each interstitial site in the material. To describe local hopping of H atoms between interstitial sites, it is necessary to locate the transition states (TSs) that control this process. TSs for diffusion of H between two adjacent interstitial sites can be determined within DFT calculations by using the Nudged Elastic Band (NEB) method [24] for simple structures. The vibrational frequencies of local minima and TSs were calculated in the harmonic approximation by assuming that localized vibrations of H atoms are decoupled from vibrations of the metal atoms. This procedure gives three real frequencies at a local energy minimum and two real frequencies and one imaginary frequency at a transition. Once a TS between two neighbor binding sites has been characterized, the H hopping rate between the sites can be computed using quantum corrected harmonic TS theory, [25] giving Q3 nO;i f ðhnO;i =2kB TÞ kOT ¼ 2i¼1 expðEa =kB TÞ: ð14:4Þ Pj¼1 nTS;j f ðhnTS;j =2kB TÞ

Chapter

14

First-principles Models of Metal Membranes

313

Here, f(x) ¼ sinh(x)/x, ni are the real vibrational frequencies of the binding site or TS and Ea is the activation energy for the particular hop. This expression explicitly includes contributions from the multiple vibrational energy levels available to the interstitial H at the temperature of interest [20]. The description above defines the hopping rate between adjacent interstitial sites, but this is not sufficient to describe the net diffusivity of interstitial H in disordered alloys. To calculate this diffusivity, we use Kinetic Monte Carlo (KMC) simulations of hopping dynamics within a lattice model with local hopping rates defined as described above. KMC is ideally suited to model stochastic systems defined by a succession of events with known rates [9,25]. Briefly, we simulate the hopping of a large number of noninteracting H atoms inside a simulation volume with periodic boundary conditions using an algorithm that correctly defines the absolute rate for each local hop. After observing a large number of hops, the mean square displacement of each H atom is calculated, and the diffusivity Ds is determined using an Einstein expression [9,26]: E 1D! j r ðtÞ  ! r ð0Þj2 : ð14:5Þ Ds ¼ lim t!1 6t Here, ! r ðtÞ is the position of the tagged particle at time t and h    i represents an average over all particles of the diffusing species. It is straightforward to calculate the self-diffusivity from these trajectories using Equation (14.5). If the diffusion of H in an ordered structure such as an intermetallic is being considered, an analytic method is available to describe the net diffusivity once the local hopping rates are known [27]. This theory was developed by Braun and Sholl [28] for calculating the net diffusion coefficient of a species within a periodic geometry. The methods just outlined make it possible to predict the solubility and diffusivity of H at dilute concentrations in the bulk of a metal alloy. Once these quantities have been found, the net permeability of H through a membrane of the alloy can be predicted using Equation (14.2).

Cluster Expansion Methods For binary or ternary alloys that exhibit substitutional disorder, a key challenge in using DFT calculations to describe interstitial H is that these materials have a plethora of structurally distinct binding sites. Kamakoti and Sholl [9] (and subsequently Sonwane et al [6].) approached this challenge by performing DFT calculations for H in a collection of distinct interstitial sites and then fitting the observed results for binding energies and TS energies to a simple lattice model in terms of parameters that describe the local environment of each site. Once a lattice model of this kind is defined, the net solubility, diffusivity, and permeability of H through the bulk alloy can be calculated using a combination of statistical mechanical calculations and KMC simulations. Because the macroscopic quantities defined by a lattice model for interstitial H can be calculated to high precision with minimal computational effort, the quality of the

314

Inorganic, Polymeric and Composite Membranes

agreement between the lattice model and the DFT data set upon which it is based is the key to the success of the approach defined above. An important difficulty with the model fitting methods used by Kamakoti and Sholl is that they do not provide a reliable way to verify the precision of the lattice model with respect to a DFT-based description of the full range of interstitial sites that can exist in a substitutionally disordered material. Semidey-Flecha and Sholl developed a more general model based on the concept of a cluster expansions (CEs) to define the energies of interstitial H atoms in crystalline metals to overcome this difficulty [20]. CEs offer a mathematical framework based on pairs, triplets, four-body terms, etc. to describe multiple body interactions that sum together to identify the energy of a configuration [29,30]. The CE model assumes that the total energy of a given configuration can be written as a linear combination of the energy of special clusters as EbðZPEÞ ¼ E0 þ SJi ð1Þ si ð1Þ þ SJi ð2Þ si ð2Þ þ SJi ð3Þ si ð3Þ þ   

ð14:6Þ

where each Ji(n) is a parameter defining the interaction energy of an H atom with clusters of n metal atoms. This infinite expansion must be truncated to determine which truncated model offers the most accurate prediction. If a particular truncation is selected, then it is straightforward to determine the interaction parameters Ji(n) by least squares minimization of the deviation between the truncated CE and the data that are available. The “leave one out (LOO)” method [31] is used to determine truncated form of the CE. In every case, the model with the lowest LOO error was chosen to define a lattice model. In the application of the CE approach to crystalline metal alloys, there are several stages. First, a set of DFT calculations is used to define binding energies in the two different kinds of interstitial sites such as octahedral (O) sites, tetrahedral (T) sites, and TSs of the alloys of interest. The LOO method is then used to fit a CE model for each site to this data. This level of comparison establishes that the CE model is able to accurately fit the available DFT data. A limitation of this comparison is that it only observes the performance of CE model with sites for which DFT data is currently available. It is also critical to compare the distribution of site binding energies observed in DFT calculations with the distribution of energies predicted in a large volume of a substitutionally random material treated with the CE models. Semidey-Flecha and Sholl have described methods to make this kind of comparison that allow CE models to be iteratively refined, if necessary, by computing additional DFT data [32].

Applications of DFT Calculations to Crystalline Membrane Materials The methods outlined above have been applied to a variety of crystalline metals to predict the permeability of these materials when used as H2-selective membranes. Here, we summarize these studies, providing references to the original literature in

Chapter

14

First-principles Models of Metal Membranes

315

each case. Kamakoti and Sholl considered fcc Pd, PdCu binary alloys, and Pd74Cu22M4 (M ¼ Nb, Pt, Ru, Rh, Ta, Ti, V) ternary alloys [7–9,25]. They considered PdCu alloys since there are experimental observations of increased contamination resistance in some PdCu alloys when it is compared to pure Pd membranes [33]. Kamakoti and Sholl’s calculation showed that the predictions of the permeability of H in PdCu alloys were in good agreement with experimental data [8]. This validation of model predictions against independent experiments was an important early step in this area indicating that DFT-based methods can be used to give quantitatively useful information. Sonwane et al. employed the same methods of Kamakoti and Sholl to investigate PdAg and PdAu alloys. They found that the H permeability in Pd100xAgx (Au) increases as Ag (Au) increases up to approximately 20 (12) (%) [34]. Ling and Sholl examined the permeation of hydrogen through metal sulfides by performing DFT calculation, and found that permeation rates of H through metal sulfides were low, primarily because of slow diffusion of H in these materials [27]. Semidey-Flecha and Sholl examined Pd-rich binary alloys and PdCu-based ternary alloys by performing DFT calculations [20,32]. They considered Pd96M4 (M ¼ Au, Ag, Pt, Rh, Cu, Ni) binary alloys, and Pd70Cu26M4 (M ¼ Au, Ag, Pt, Ni) ternary alloys. Among these alloys, Pd96Ag4 is the only binary alloy they examined that was predicted to have a higher permeability than Pd. Their predictions for binary alloys with composition Pd96M4 agree with the wellknown observation that adding small amounts of Ag to Pd increases the alloy permeability relative to pure Pd. They also found that Pd70Cu26Au4 is the alloy showing the largest hydrogen permeability among the ternary alloys they examined (Pd70Cu26M4 (M ¼ Au, Ag, Pt)) [32]. Ling and Sholl examined seven different PdCuAg ternary alloys, with compositions such as Pd92.6Cu3.7Ag3.7, Pd85.2Cu3.7Ag11.1, Pd85.2Cu11.1Ag3.7, Pd81.5Cu11.1Ag7.4, Pd70.4Cu11.1Ag18.5, Pd70.4Cu25.9Ag3.7, and Pd66.7Cu25.9Ag7.4 by using DFT calculations. They found that the net permeability of H increases with Ag concentration for alloys with the same amount of Cu. However, the strength of this trend was diminished at high Cu or Ag content. Semidey-Flecha, Hao, and Sholl studied isotopic selectivity of dense metal membranes for high temperature hydrogen purification by performing DFT calculations [35]. Hydrogen isotope separation is traditionally based on principles of adsorption [36,37], or cryogenic distillation [37,38]. Bhatia and Sholl have demonstrated methods for accurately predicting the rates of activated hopping and quantum tunneling for H in metals and on metal surfaces by using first-principles DFT to compute the potential energy surface for H, [16] and these methods make it straightforward to predict the permeance of hydrogen isotopes through metal membranes once the properties of interstitial H have been described for the same materials.

316

Inorganic, Polymeric and Composite Membranes

AMORPHOUS METAL MEMBRANES Using thin films of amorphous metals, that is, metals without long-range structural order, offers significant opportunities for fabrication of membranes that outperform existing Pd-based metal membranes based on crystalline alloys. The use of amorphous metal films as membranes can address several of the difficulties associated with crystalline membranes. In general, concerns about hydrogen embrittlement and sintering are reduced in amorphous materials when compared to crystalline materials [39]. The cost of the membrane materials in amorphous films can potentially be far less than in Pd-based alloys. Many of the amorphous films studied to date have been based on low-cost materials such as Zr, Ni, Cu, and Al. Another positive feature of these films is that the glassforming requirements of materials for amorphous films are much less stringent than for bulk samples of the same materials. Dolan et al. have recently provided a detailed review of hydrogen-selective amorphous alloy membranes [39], and Ockwig and Nenoff highlighted recent progress with these materials as part of their comprehensive review of membranes for hydrogen production [40]. Perhaps the single most important observation from these two reviews is that amorphous metal membranes have been tested by multiple groups, so the methods required for creating pinholefree membranes are now well developed. Ockwig and Nenoff noted that development of these membranes is “still an entirely open field”, and this is in large part due to the huge number of materials that could be used [40]. Among the limited number of materials that have been tested as membranes, a number have shown promising permeabilities relative to well-known crystalline membranes. Inoue and coworkers have reported several films involving Zr, Ni, and other metals that have H2 permeabilities comparable to pure Pd [41,42]. In this section, we summarize our recent efforts to accelerate the development of amorphous metal membranes by making quantitative theoretical predictions about the performance of these membranes [43–46]. Our current work on amorphous metal films is motivated by our previous theoretical treatments of crystalline metal films [8]. Although the properties of interstitial H in amorphous metals has been studied for at least 30 years [47–49], previous theoretical work on these materials has been phenomenological rather than quantitative. We have introduced a new approach based on quantum chemistry calculations that gives quantitative information about the flux of hydrogen through amorphous metal films. This work required development of new methods for modeling both solubility and diffusivity of H in amorphous metals. The ingredients of a continuum-level description of H2 transport through crystalline metal alloys are well known and were summarized above [5,14,50]. In the case of amorphous membranes, the flux as a function of the membrane’s operating conditions can be predicted once the solubility and diffusion coefficients of interstitial H are known provided we assume that it is H

Chapter

14

First-principles Models of Metal Membranes

317

transport through the bulk film that is rate limiting. As already indicated in our discussion of crystalline materials, these tasks can be performed effectively by appropriate use of DFT calculations.

Computational Approaches for Amorphous Metals In order for first-principles methods to be applied to H transport through amorphous metals, several technical challenges must be overcome that do not exist for crystalline alloys. First, a detailed model of the atomic coordinates of the amorphous material must be generated. Once this model is available, the sites that can be occupied by interstitial H and the TSs for hopping of H atoms between these sites must be identified. These tasks can be simplified by symmetry considerations in crystalline materials, but these simplifications do not exist for amorphous materials. Because the solubility of H in amorphous materials is frequently considerably stronger than in crystalline materials [39,49,51], modeling of H solubility in amorphous films must consider the effects of H concentration. Concentration effects can also have important implications for interstitial diffusion in amorphous films. Experiments and empirical models of H diffusion in amorphous films have shown that H diffusivities can increase greatly as the concentration of interstitial H increases [47,49]. As a result, models of these films must consider not only the dilute concentrations of H that are all needed in describing crystalline membranes but also the effects of interstitial H concentration.

Amorphous Structures The first task necessary for a DFT-based description of an amorphous material is to prepare an atomically detailed model of the amorphous alloy. We achieve this task using ab initio molecular dynamics [52] at temperatures higher than the melting points of the alloys of interest. Typically, calculations were performed with supercells containing 108 atoms in a liquid-like state that was then quenched using conjugate gradient relaxation. The liquid state volumes are about 7% greater than the volume of the quenched amorphous samples. In all calculations, reciprocal space was sampled with the G-point only. The reliability of the resulting amorphous structure can be assessed by comparison with experimental structural data. One convenient approach to describe the local structure for amorphous structures is the radial distribution function (RDF). We considered Zr30(Ni0.6Nb0.4)70 as an example because of the availability of experimental results for this material. Our calculated RDF before hydrogenation for Zr30(Ni0.6Nb0.4)70 is compared with the experimental RDF in Figure 14.1. Good agreement can be seen in both the overall shape and the peak positions of this RDF. These comparisons suggest that our ab initio MD calculations were successful in producing a realistic sample of amorphous Zr30(Ni0.6Nb0.4)70.

318

Radial distribution (arb. units)

Inorganic, Polymeric and Composite Membranes

Exp. Zr30(Ni0.6b0.4)70 Exp. hydrogenated Zr30(Ni0.6Nb0.4)70 Host - Hydrogenated H + host -Hydrogenated

2

3

4

5

6

r (Å) FIGURE 14.1 Comparison of calculated and experimental radial distribution functions for a-Zr30(Ni0.6Nb0.4)70 before and after hydrogenation. The curves in the lower (upper) part of the figure are calculated (from the experiments of Yamaura et al [42].). The curve-labeled “Host–Hydrogenated” shows the results for the hydrogenated material when H atoms are not included in calculating the radial distribution function. Reprinted with permission from Ref. [46].

The RDF for a hydrogenated sample of Zr30(Ni0.6Nb0.4)70 is compared with experimental data in Figure 14.1. The first peak in this RDF describes the average distance between H and host atoms, which is of course shorter than the average distance between host atoms. Figure 14.1 also shows the partial RDF that does not include pairs of atoms with one or more H atoms; this is the appropriate quantity for comparison with the experimental data. Our calculated result shows a slight rightward shift in the hydrogenated sample relative to the original alloy due to lattice expansion, but this effect is not evident in the experimental data. Both our calculations and the experiments show a distinct ˚ in the hydrogenated material. peak at around 3.4 A

Binding Energy of Interstitial H in Amorphous Alloys To predict the solubility of interstitial H in a metal, the binding energy of H atoms in each interstitial site in the material must be known. Finding these sites is usually straightforward in crystalline materials, but this is not the case for amorphous materials. We developed a method to identify all interstitial sites in an amorphous metal [44]. To avoid introducing bias, it is important that methods be developed to locate all interstitial sites without making assumptions about the structure of these sites. We approached this task in two stages. First, we used empirical pair potentials for H/M interactions to define the energy of individual H atoms with the positions of all host M atoms fixed. Using empirical pair potentials, the energy of H atoms was minimized from all points defining a grid with resolution 0.02  0.02  0.02 nm in each supercell. Any minimum

Chapter

14

First-principles Models of Metal Membranes

319

˚ from other preidentified by these calculations that was located more than 0.9 A viously identified minima was considered to be a distinct minimum within the structure. The distinct local minima found in this way as starting points for DFT calculations in which all atoms are allowed to relax. The DFT calculations performed starting from the minima found from the empirical potentials gave 351 structurally distinct interstitial sites in a 108-atom supercell of Zr30(Ni0.6Nb0.4)70 and similar numbers of interstitial sites in the other amorphous materials we examined.

H–H Interactions When the binding energies of interstitial H atoms in a material are unfavorable with respect to gaseous H2 (or are only marginally favorable), the solubility of H in this material will be low, even under relatively high pressures of H2. This situation occurs in many of the crystalline metals that are used as membranes for H2 purification, and it means that these membrane materials can be analyzed in terms of individual interstitial H atoms [8,9,40]. When individual binding sites exist with binding energies that are highly favorable relative to gaseous H2, however, appreciable concentrations of interstitial H will exist even for moderate pressures of H2. In this situation, it is important to understand whether interactions among interstitial atoms in the material will play a role in the material’s net solubility. To probe the effects of H–H interactions in amorphous materials, we used DFT to compute the interaction energy of a large number of H–H pairs with the H atoms located in nearby interstitial sites. In these calculations, the two H atoms were placed in the positions determined to be energy minima for the individual interstitial sites, then all atoms in the supercell were allowed to relax during energy minimization. After relaxation, the interaction energy for a pair of H atoms was defined by DEHH ¼ Ehost=2H þ Ehost  Ehost=H;1  Ehost=H;2 :

ð14:7Þ

Here, Ehost/2H(Ehost/H1) is the energy of the supercell including both H atoms, and Ehost/H,1 (Ehost/H,2) is the energy of the optimized supercell containing only the first (second) H atom. Zero point energies were not included in this analysis. With this definition, values of DEH – H > 0 correspond to repulsive H–H interactions. For each material, calculations were performed for multiple distinct pairs. Our DFT results are empirically described with reasonable accuracy with a simple repulsive exponential function, as shown by the solid curve in Figure 14.2 with data from an amorphous sample of Fe3B. H–H interactions in metals are often characterized via the Westlake criterion, an empirical prediction that two nearby interstitial sites cannot be simultaneously occupied by H atoms if the distance between the sites is less than 0.21 nm [53]. The results in Figure 14.2 allow the precision of this criterion to be considered. Our results are reasonably consistent with the Westlake criterion if the

320

Inorganic, Polymeric and Composite Membranes

Interaction energy (eV)

1 DFT for favorable sites DFT for other sites Westlake criterion a0exp(−a1x)

0.8

0.6

0.4

0.2

0

1

2

3 H–H distance (Å)

4

FIGURE 14.2 Calculated H–H interaction energies for 35 distinct H–H pairs in a-Fe3B. ˚  1. Reprinted with permission The parameters for fitted solid line are a0 ¼ 5.68 eV and a1 ¼ 1.61 A from Ref. [43].

distribution of a fixed number of H atoms in the host is considered, since in this case the energy penalty for placing H atoms in two sites that are closer than 0.21 nm is 0.2 eV or more.

H Solubility in Amorphous Alloys The DFT calculations outlined above provide detailed information on the binding energy of H atoms in the amorphous materials under the assumption that the interstitial H was present at dilute levels. A general property of amorphous metals, however, is that the solubility of H in these materials can be significant under many conditions of practical interest [47,54]. Once the energy of each interstitial site and a description of H–H repulsion are available, the net solubility of H in our amorphous samples can be calculated as a function of temperature and H2 pressure using Grand Canonical Monte Carlo (GCMC) simulations [50,55]. These calculations equate the chemical potential of interstitial H and the gaseous phase, assuming that dissociation of H2 occurs in order for the interstitial H to reach equilibrium. In all of our GCMC calculations, the gaseous phase was treated as an ideal gas. The binding energy for interstitial H in each site in the amorphous material during a GCMC simulation is defined by binding energies calculated in the DFT calculations described above. The solubility in Zr30(Ni0.6Nb0.4)70 calculated in this way for different H2 pressures is shown in Figure 14.3, together with experimental results, in terms of the ratio of H atoms to metal atoms in the material, H/M [56]. The temperatures at which experimental data is available for solubility are quite low relative to the temperatures at which metal membranes would be used in many applications, which implies that the levels of H solubility seen in these experiments are higher than what

Chapter

14

321

First-principles Models of Metal Membranes

Solubility (H/M)

(a) GCMC, 573 K GCMC (ΔE), 573 K exp. 573 K

0.5

0.4

0.3

10-1

100 Pressure (atm)

101

(b)

Solubility (H/M)

100

10-1 Zr30(Ni0.6Nb0.4)70, (ΔE),1 atm Zr30(Ni0.6Nb0.4)70, 1 atm Zr30(Ni0.6Nb0.4)70, exp. 1 atm

10-2

Zr36Ni64, (ΔE), 1 atm Zr36Ni64, 1 atm Zr36Ni64, exp. 1 atm

300

400

500 600 Temperature (K)

700

800

FIGURE 14.3 Calculated H solubility with binding energy corrections induced by lattice expansion (dashed curves). Solid curves indicate the predicted H solubility without these energy corrections. Experimental data [42,57] is also shown using a dotted curve in (a) and as individual symbols in (b). Reprinted with permission from Ref. [46].

would be relevant in many membrane applications. It can be seen from Figure 14.3 that there are many conditions where the concentration of H in a-Zr36Ni64 is high compared to crystalline Pd. For example, the H solubility in a-Zr36Ni64 is around 0.2 at 473 K and 1 atm, while the solubility of H in Pd is only about 0.04 under the same conditions [8]. At the lowest H concentrations seen for the binary alloy, the theoretical predictions are in good agreement with the experimental data. At higher H concentrations, however, our calculations systematically underestimate the solubility of H. One obvious origin of this underestimation is that the presence of large amounts of H can expand the volume of the amorphous material. This expansion affects that binding energy of H in each interstitial site, which in turn has an effect on the overall solubility.

322

Inorganic, Polymeric and Composite Membranes

By comparing the energy of a hydrogenated sample after relaxation with DFT to the energy predicted by our GCMC calculation, we can estimate the change in binding energy of each H atom due to the alloy’s expansion [46]. It is clear that this effect is dominated by lattice expansion rather than effects from long-range H–H repulsion that were not included in our GCMC simulation because the latter effect would make the H atom binding energies less favorable. We used calculations at two different values of H/M together with the fact that the lattice expansion must by definition be zero when H/M ¼ 0, to define a quadratic expression for the average change in binding energy of H in a-Zr30(Ni0.6Nb0.4)70. Using this expression, we can self-consistently predict the solubility of H including the effect of H-induced expansion in an iterative way. First, a GCMC simulation is performed at a temperature and pressure of interest as described above. The value of H/M predicted by this calculation is then used to define the average binding energy change due to expansion. Another GCMC simulation is then performed in which the binding energy of every interstitial site is modified by this average energy change. This process is repeated iteratively until the concentration used to define the average energy change and the concentration predicted by GCMC are self-consistent. These calculations require no experimental input. Figure 14.3a shows the results of this iterative calculation applied to aZr30(Ni0.6Nb0.4)70, as well as the GCMC results without including expansion effects and the experimental data. It is clear that including the effects of Hinduced expansion in this approximate manner gives results that are in close agreement with the experimental data. This result strongly suggests that most of the difference between our GCMC results without including expansion effect and the experimental data is due to neglecting this effect. Figure 14.3b shows how the T-dependent solubility of H is affected by H-induced expansion. In this figure, the solid curves are the results from the original GCMC calculations, while the dashed curves show the result from the iterative method we have just defined. It can be seen from Figure 14.3b that the impact of H-induced expansion on the solubility of H is mild at the elevated temperatures of greatest interest for membrane applications because the amount of interstitial H present at these conditions is smaller than at lower temperatures.

Hydrogen Diffusion in Amorphous Alloys For describing net transport of hydrogen through an amorphous membrane, the relevant description is Fick’s law, which relates the flux of hydrogen, J, to the concentration gradient in the hydrogen concentration, c, by [58] J ¼ Dt ðcÞrc

ð14:8Þ

Fick’s law introduces the transport diffusion coefficient, Dt, which we have written in a way that explicitly emphasizes the dependence of this diffusivity on the local concentration. The transport diffusion coefficient is also known as

Chapter

14

First-principles Models of Metal Membranes

323

the Fickian or chemical diffusion coefficient [59]. The self-diffusivity, which characterizes the motion of individual tracer particles, and the transport diffusivity are in general not equal; they are only equal in the limit of dilute concentrations of the diffusing species [60]. In many crystalline metal alloys that are used as metal membranes, the concentration of interstitial H is small, with 1% or fewer of the interstitial sites being filled with H atoms. As a result, the self-diffusivity, Ds, can be used to replace Dt in Fick’s law while still getting accurate results for the net flux of H [59]. As shown above, H solubilities in amorphous phases can be high at practical conditions. This implies that the concentration dependence of Dt must be considered to reliably describe diffusion of H in these materials. Although it is not calculated as commonly as the self-diffusivity, methods exist to determine the transport diffusivity once the trajectories of individual atoms are known. One approach is to rewrite the transport diffusivity without approximation as [61] Dt ðcÞ ¼ D0 ðcÞ

@ lnf @ lnc

ð14:9Þ

Here, D0(c) is the corrected diffusivity, and f is the fugacity of H in the bulk gas phase in equilibrium with the solid. The corrected diffusivity is also known as the single-component Maxwell–Stefan diffusivity [62,63], and the derivative term is called the thermodynamic correction factor. The advantage of Equation (14.9) is that the corrected diffusivity and the thermodynamic correction factor can be computed independently from the model of interstitial H we have defined above [45].

Corrected Diffusivities Corrected diffusivities in the supercells that model amorphous materials can be calculated using KMC simulations closely related to the methods used for crystalline materials [45]. To locate the TSs between adjacent sites, we used an efficient heuristic method developed by Chen and Sholl for crystalline membranes [64]. The DFT-calculated energies and vibrational frequencies of H at each interstitial site and TS were then used to define site to site hopping rates by applying quantum corrected transition state theory (QC-TST). This approach accurately describes the local hopping rates in the temperature regime we examine below. We examined diffusion of interstitial H by replicating the supercells used in our DFT calculations using periodic boundary conditions. At the beginning of each KMC simulation, H atoms were distributed randomly among the simulation volume’s interstitial sites according to a Boltzmann weighted distribution. This allowed us to equilibrate the initial state with a relatively short KMC simulation, typically 5000 MC events per H atom. In all of our KMC simulations, H–H interactions were described using the Westlake criterion. That is, if the destination site for an attempted hop was

324

Inorganic, Polymeric and Composite Membranes

within 0.21 nm of another interstitial site that was already occupied by H, the attempted hop was automatically rejected (although time is still incremented as it would be for the same event occurring without H–H interactions). The Westlake criterion was also enforced during the initial placement of H atoms in the simulation volume. We did not explicitly consider the effects of H-induced expansion in our KMC simulations. Because the binding energy of every site is modified by the same energy when lattice expansion is described as above, the only effect of H-induced expansion on diffusion is that the hopping distance associated with a site to site hop changes slightly due to expansion. This effect would increase our calculated diffusivities by at most a few percent under the highest H/M concentrations we considered. The KMC simulations outlined above generate dynamically correct trajectories for each H atom in the system as a function of time. The corrected diffusivity, D0(c), can be calculated from these trajectories with methods that have been used extensively in studies of molecular diffusion in nanoporous materials [59,61,65–67]. The resulting corrected diffusion coefficients at various interstitial concentrations for two amorphous materials we considered are shown in Figure 14.4. The most obvious feature is that the corrected diffusivity of H in these amorphous metals is strongly concentration dependent. At 600 K, for example, the corrected diffusivity in both materials increases by almost an order of magnitude as the H concentration is increased from H/M ¼ 0.01 to 0.2. This result is consistent with earlier experiments with other amorphous metals [48] and phenomenological models [47]. This behavior occurs because of the broad distribution of interstitial site energies that exist for H in these materials. At dilute concentrations, H atoms are essentially always found in the sites with the most favorable binding energies, and net diffusion only occurs when the large barriers that exist to moving H away from these sites can be overcome. At higher concentrations, the most favorable sites are populated by a subset of the total collection of H atoms, allowing the remaining H atoms to occupy and hop among the less favorable sites. The motions of this second subset of H atoms dominate the net diffusion of H in the material, causing the overall diffusivity to increase as the concentration is increased from dilute to moderate values of H/M.

The Thermodynamic Correction Factor The thermodynamic correction factor defined in Equation 14.9 can be computed directly from the same GCMC simulations we used to assess H solubility via a fluctuation formula [68,69]: @ ln f hN i ¼ @ lnc ðhN 2 i  hN i2 Þ

ð14:10Þ

Here N is the number of H atoms in the simulation cell. The calculated thermodynamic correction factors normally included the effects of H-induced

Chapter

14

325

First-principles Models of Metal Membranes

(a) c = 0.2 c = 0.1 c = 0.01

D0 (m2s-1)

10-9

10-10

10-11

10-12 1

1.2

1.4 1.6 1000/T (K−1)

1.8

2

1.4 1.6 1000/T (K−1)

1.8

2

(b) 10-8

D0 (m2s-1)

10-9

10-10 0.5 0.2 0.1 0.01

10-11

10-12

1

1.2

FIGURE 14.4 Corrected diffusivities of H in (a) a-Zr36Ni64 and (b) a-Zr30(Ni0.6Nb0.4)70 from the KMC simulations described in the text. The value of H/M for each set of calculations is indicated in the legend. Reprinted with permission from Ref. [46].

expansion as described above. The thermodynamic correction approaches unity at high temperature (e.g., 800 K) and low pressure (e.g., 10 4 atm). At lower temperatures where the solubility of H in the material is high, the thermodynamic correction factor monotonically increases with pressure. Once we had calculated the corrected diffusivities and thermodynamic correction factors separately, the transport diffusivities can be calculated using Equation 14.9. In the calculations below, corrected diffusivities were specified by fitting continuous curves to the results shown in Figure 14.4. The thermodynamic correction factors for each concentration of interest were evaluated using GCMC by varying the bulk H2 pressure until the observed value of H/M matched the value used in the KMC simulation within a small tolerance.

326

Inorganic, Polymeric and Composite Membranes

Hydrogen Permeability Through Amorphous Alloys Once the solubility and diffusivity of H in a metal film are known, it is straightforward to predict the permeability of the film when it is used as a membrane. Similar to the crystalline membranes, we consider amorphous membranes that are thick enough that the kinetics of surface processes can be neglected. In practical studies of amorphous metal membranes, a thin coating of Pd on the membrane surfaces is often used to ensure that the kinetics of surface processes are rapid [14,50]. By assuming that transport of H through the bulk of the membrane is the dominant transport resistance, the net hydrogen flux through a membrane can be written as 1 J¼ L

cfeed ð

cperm

1 Dt ðcÞdc ffi ðcfeed  cperm ÞDt ð cÞ L

ð14:11Þ

Here, cfeed (cperm) is the concentration of interstitial H on the feed (permeate) side, L is the membrane thickness, and c ¼ ðcfeed þ cpermeate Þ=2. The first equality in this expression is a direct consequence of Equation (14.1). In order to use a quantity that removes the dependence on the membrane thickness and the operating pressures, membrane performance is typically reported in terms of permeability k¼

JL 1=2 ðPfeed

1=2

 Pperm Þ

ð14:12Þ

When solubility in the membrane material obeys Sieverts’ law, the permeability is independent of the feed and permeate pressure. In amorphous films, however, Sieverts’ law may not be valid, so the membrane permeability must be calculated for specific feed and permeate pressures. To compare with the available experimental results, we used the same temperatures and membrane thickness, L ¼ 30 mm, as in the experiments by Hara et al. [57]. The predicted H2 permeabilities for amorphous Zr36Ni64 and Zr30(Ni0.6Nb0.4)70 using a feed (permeate) pressure of 3 atm (0.01 atm) are shown in Figure 14.5. The symbols connected by solid lines in this figure show results using the interstitial concentrations and transport diffusivities calculated as described above. The corresponding symbols connected by dotted lines are experimental reports [42]. For both of these amorphous materials, our calculations predict permeabilities that are a factor of 2–4 lower than the experimental results. Considering that our calculations required no input or fitting of parameters from experiment, we feel that this level of agreement is very good. Figure 14.5 also shows two independent experimental reports of H permeation through crystalline Pd. These two sets of data highlight the observation that variations in permeability by factors of 2–4 are not unusual when comparing experiments performed with nominally identical materials. The Pd data also make it clear that amorphous alloys such as the Zr–Ni–Nb material shown in

14

H2 permeability (mol m-1 s-1 Pa0.5)

Chapter

327

First-principles Models of Metal Membranes

10-8

10-9 Pd, exp. [A] Pd, exp. [B] a-Zr0.3(Ni0.6Nb0.4)0.7, exp. a-Zr0.3(Ni0.6Nb0.4)0.7, calc. a-Zr Ni , exp.

10-10

0.36

0.64

a-Zr0.36Ni0.64, calc.

10-11 500

600

700 800 Temperature (K)

900

FIGURE 14.5 Comparison of calculated and experimental permeability of H2 in a-Zr36Ni64 and a-Zr30(Ni0.6Nb0.4)70 as a function of temperature. The permeation and feed pressure are set as 0.01 and 3 atm. The permeabilities of Pd are also plotted, where exp. [A] and [B] refer to Ref. [8] and [42], respectively. The experimental data for a-Zr36Ni64 and a-Zr30(Ni0.6Nb0.4)70 is from Ref. [42] and [57], respectively. Reprinted with permission from Ref. [46].

Figure 14.5 have potential for creating membranes with permeability in the same range that is possible with crystalline Pd.

CONCLUSION In this chapter, we have reviewed work over the past few years that uses DFTbased calculations to make predictions about the permeability of hydrogen through metal membranes. Predictions of this type have now been made for a wide range of crystalline metal alloys and also for a number of amorphous metals. An important feature of the methods we have used here is that they can be applied to any alloy of interest. That is, these methods do not rely on any special chemical or physical properties of the specific example we have studied, or on the availability of experimental data. This means that these methods are useful for searching for new materials with promising properties. It is important to note, however, that permeability of H is only one factor that determines the viability of a membrane material in practical applications. As we have noted elsewhere [43], our calculations cannot predict other relevant physical properties such as the robustness of an alloy to chemical contaminants in the feed stream or the crystallization temperature of an amorphous material. These issues, which are crucial in the use of membranes in practical environments, will need to be addressed through experiments. Despite this caveat, it seems likely that the modeling methods we have described will play an important role in future efforts to develop new membranes by focusing experimental attention on novel compositions that have potential to have high permeability for hydrogen.

328

Inorganic, Polymeric and Composite Membranes

ACKNOWLEDGMENTS Financial support for this work came from the DOE-BES and the DOE’s National Energy Technology Laboratory. We acknowledge Preeti Kamakoti, Lymarie Semidey-Flecha, and Chen Ling for their work with us on the topics reviewed here. Our work has also benefited from a long term collaboration with experimental efforts by Kent Coulter (SWRI) and Douglas Way (Colorado School of Mines).

REFERENCES [1] L. Schlapbach, A. Zuttel, Hydrogen-storage materials for mobile applications, Nature 414 (2001) 353. [2] T.I. Sigfusson, Pathways to hydrogen as an energy carrier, Philos. Trans. R. Soc. Math. Phys. Eng. Sci. 365 (2007) 1025. [3] B.C.H. Steele, A. Heinzel, Materials for fuel-cell technologies, Nature 414 (2001) 345. [4] D.S. Sholl, Y.H. Ma, Dense metal membranes for the production of high-purity hydrogen, MRS Bull. 31 (2006) 770. [5] S.N. Paglieri, J.D. Way, Innovations in palladium membrane research, Sep. Purif. Methods 31 (2002) 1. [6] C.G. Sonwane, J. Wilcox, Y.H. Ma, Solubility of hydrogen in PdAg and PdAu binary alloys using density functional theory, J. Phys. Chem. B 110 (2006) 24549. [7] P. Kamakoti, D.S. Sholl, A comparison of hydrogen diffusivities in Pd and CuPd alloys using density functional theory, J. Membr. Sci. 225 (2003) 145. [8] P. Kamakoti, B.D. Morreale, M.V. Ciocco, B.H. Howard, R.P. Killmeyer, A.V. Cugini, et al., Prediction of hydrogen flux through sulfur-tolerant binary alloy membranes, Science 307 (2005) 569. [9] P. Kamakoti, D.S. Sholl, Ab initio lattice-gas modeling of interstitial hydrogen diffusion in CuPd alloys, Phys. Rev. B 71 (2005) 014301. [10] V. Gryaznov, Metal containing membranes for the production of ultrapure hydrogen and the recovery of hydrogen isotopes, Sep. Purif. Methods 29 (2000) 171. [11] Y. Sakamoto, F.L. Chen, M. Furukawa, M. Noguchi, Permeability and diffusivity of hydrogen in palladium-rich Pd-Y(Gd)-Ag ternary alloys, J. Alloy. Comp. 185 (1992) 191. [12] H. Yukawa, G.X. Zhang, N. Watanabe, M. Morinaga, T. Nambu, Y. Matsumoto, Analysis of hydrogen diffusion coefficient during hydrogen permeation through niobium and its alloys, J. Alloy. Comp. 476 (2009) 102. [13] S.K. Gade, M.K. Keeling, A.P. Davidson, O. Hatlevik, J.D. Way, Palladium-ruthenium membranes for hydrogen separation fabricated by electroless co-deposition, Int. J. Hydrogen Energy 34 (2009) 6484. [14] T.L. Ward, T. Dao, Model of hydrogen permeation behavior in palladium membranes, J. Membr. Sci. 153 (1999) 211. [15] B.D. Morreale, M.V. Ciocco, R.M. Enick, B.I. Morsi, B.H. Howard, A.V. Cugini, et al., The permeability of hydrogen in bulk palladium at elevated temperatures and pressures, J. Membr. Sci. 212 (2003) 87. [16] B. Bhatia, D.S. Sholl, Quantitative assessment of hydrogen diffusion by activated hopping and quantum tunneling in ordered intermetallics, Phys. Rev. B 72 (2005) 224302. [17] B.D. Morreale, B.H. Howard, O. Iyoha, R.M. Enick, C. Ling, D.S. Sholl, Experimental and computational prediction of the hydrogen transport properties of Pd4S, Ind. Eng. Chem. Res. 46 (2007) 6313.

Chapter

14

First-principles Models of Metal Membranes

329

[18] J. Greeley, M. Mavrikakis, Surface and subsurface hydrogen: adsorption properties on transition metals and near-surface alloys, J. Phys. Chem. B 109 (2005) 3460. [19] D.S. Sholl, J. Steckel, Density Functional Theory: A Practical Introduction, John Wiley &Sons 2009. [20] L. Semidey-Flecha, D.S. Sholl, Combining density functional theory and cluster expansion methods to predict H-2 permeance through Pd-based binary alloy membranes, J. Chem. Phys. 128 (2008) 144701. [21] G. Kresse, J. Furthmuller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set, Comput. Mater. Sci. 6 (1996) 15. [22] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh, et al., Atoms, molecules, solids, and surfaces—applications of the generalized gradient approximation for exchange and correlation, Phys. Rev. B 46 (1992) 6671. [23] B.D. Morreale, M.V. Ciocco, B.H. Howard, K.S. Rothenberger, A.V. Cugini, R.M. Enick, Effect of hydrogen-sulfide on the hydrogen permeance of palladium-copper alloys at elevated temperatures, J. Membr. Sci. 241 (2004) 219. [24] G. Henkelman, B.P. Uberuaga, H. Jonsson, A climbing image nudged elastic band method for finding saddle points and minimum energy paths, J. Chem. Phys. 113 (2000) 9901. [25] P. Kamakoti, D.S. Sholl, Towards first principles-based identification of ternary alloys for hydrogen purification membranes, J. Membr. Sci. 279 (2006) 94. [26] M.P. Allen, D.J. Tildesley, Computer Simulation of Liquids, Claredon Press, Oxford, 1987. [27] C. Ling, D.S. Sholl, First principles investigation of metal sulfides as membranes in hydrogen purification, J. Membr. Sci. 329 (2009) 153. [28] O.M. Braun, C.A. Sholl, Diffusion in generalized lattice-gas models, Phys. Rev. B 58 (1998) 14870. [29] M.H.F. Sluiter, Y. Kawazoe, Invariance of truncated cluster expansions for first-principles alloy thermodynamics, Phys. Rev. B 71 (2005) 212201. [30] A. Seko, K. Yuge, F. Oba, A. Kuwabara, I. Tanaka, T. Yamamoto, First-principles study of cation disordering in MgAl2O4 spinel with cluster expansion and Monte Carlo simulation, Phys. Rev. B 73 (2006) 094116. [31] R. Drautz, A. Diaz-Ortiz, Obtaining cluster expansion coefficients in ab initio thermodynamics of multicomponent lattice-gas systems, Phys. Rev. B 73 (2006) 224207. [32] L. Semidey-Flecha, C. Ling, D.S. Sholl, Detailed first-principles models of hydrogen permeation through PdCu-based ternary alloys, J. Membr. Sci. 362 (2010) 384. [33] F. Roa, M.J. Block, J.D. Way, The influence of alloy composition on the H(2)flux of composite Pd-Cu membranes, Desalination 147 (2002) 411. [34] C.G. Sonwane, J. Wilcox, Y.H. Ma, Achieving optimum hydrogen permeability in PdAg and PdAu alloys, J. Chem. Phys. 125 (2006) 184714. [35] L. Semidey-Flecha, S.Q. Hao, D.S. Sholl, Predictions of H isotope separation using crystalline and amorphous metal membranes: a computational approach, J. Taiwan Inst. Chem. Eng. 40 (2009) 246. [36] K. Aoki, Y. Ogata, K. Kusakabe, S. Morooka, Applicability of palladium membrane for the separation of protium and deuterium, Int. J. Hydrogen Energy 23 (1998) 325. [37] Q.Y. Wang, S.R. Challa, D.S. Sholl, J.K. Johnson, Quantum sieving in carbon nanotubes and zeolites, Phys. Rev. Lett. 82 (1999) 956. [38] J. Evans, I.R. Harris, D.K. Ross, A proposed method of hydrogen isotope-separation using palladium alloy membranes, J. Less-Common Met. 89 (1983) 407. [39] M.D. Dolan, N.C. Dave, A.Y. Ilyushechkin, L.D. Morpeth, K.G. McLennan, Composition and operation of hydrogen-selective amorphous alloy membranes, J. Membr. Sci. 285 (2006) 30.

330

Inorganic, Polymeric and Composite Membranes

[40] N.W. Ockwig, T.M. Nenoff, Membranes for hydrogen separation, Chem. Rev. 107 (2007) 4078. [41] S. Hara, N. Hatakeyama, N. Itoh, H.M. Kimura, A. Inoue, Hydrogen permeation through amorphous-Zr36-xHfxNi64-alloy membranes, J. Membr. Sci. 211 (2003) 149. [42] S.I. Yamaura, M. Sakurai, M. Hasegawa, K. Wakoh, Y. Shimpo, M. Nishida, et al., Hydrogen permeation and structural features of melt-spun Ni-Nb-Zr amorphous alloys, Acta Mater. 53 (2005) 3703. [43] S.Q. Hao, D.S. Sholl, Using first-principles calculations to accelerate materials discovery for hydrogen purification membranes by modeling amorphous metals, Energy Environ. Sci. 1 (2008) 175. [44] S.Q. Hao, M. Widom, D.S. Sholl, Probing hydrogen interactions with amorphous metals using first-principles calculations, J. Phys. Condens. Matter 21 (2009) 115402. [45] S.Q. Hao, D.S. Sholl, Self-diffusion and macroscopic diffusion of hydrogen in amorphous metals from first-principles calculations, J. Chem. Phys. 130 (2009) 11106. [46] S.Q. Hao, D.S. Sholl, Comparison of first principles calculations and experiments for hydrogen permeation through amorphous ZrNi and ZrNiNb films, J. Membr. Sci. 350 (2010) 402. [47] R. Kirchheim, Hydrogen solubility and diffusivity in defective and amorphous metals, Prog. Mater. Sci. 32 (1988) 261. [48] R. Kirchheim, T.M. Otschele, W. Kiennger, H. Gleiter, R. Birringer, T.D. Koble, Hydrogen in amorphous and nanocrystalline metals, Mater. Sci. Eng. 99 (1988) 457. [49] R. Kirchheim, Solubility and diffusivity of hydrogen in complex materials, Phys. Scr. 94 (2001) 58. [50] C. Ling, D.S. Sholl, Using first-principles calculations to predict surface resistances to H-2 transport through metal alloy membranes, J. Membr. Sci. 303 (2007) 162. [51] D.S. dos Santos, P.E.V. Miranda, Hydrogen diffusivity and solubility in crystalline and amorphous alloys, J. Mater. Sci. 32 (1997) 6311. [52] M. Mihalkovic, M. Widom, Ab initio calculations of cohesive energies of Fe-based glass-forming alloys, Phys. Rev. B 70 (2004) 144107. [53] D.G. Westlake, Hydrides of intermetallic compounds—a review of stabilities, stoichiometries and preferred hydrogen sites, J. Less-Common Met. 91 (1983) 1. [54] N. Eliaz, D. Eliezer, An overview of hydrogen interaction with amorphous alloys, Adv. Perform. Mater. 6 (1999) 5. [55] B.P. Uberuaga, A.F. Voter, K.K. Sieber, D.S. Sholl, Mechanisms and rates of interstitial H2 diffusion in crystalline C-60, Phys. Rev. Lett. 91 (2003) 105901. [56] K. Aoki, M. Kamachi, T. Masumoto, Thermodynamics of hydrogen absorption in amorphous Zr-Ni alloys, J. NonCryst. Solids 61–62 (1984) 679. [57] S. Hara, K. Sakaki, N. Itoh, H.-M. Kimura, K. Asami, A. Inoue, An amorphous alloy membrane without noble metals for gaseous hydrogen separation, J. Membr. Sci. 164 (2000) 289. [58] D.S. Sholl, Predicting single-component permeance through macroscopic zeolite membranes from atomistic simulations, Ind. Eng. Chem. Res. 39 (2000) 3737. [59] D.S. Sholl, Understanding macroscopic diffusion of adsorbed molecules in crystalline nanoporous materials via atomistic simulations, Acc. Chem. Res. 39 (2006) 403. [60] F.J. Keil, R. Krishna, M.O. Coppens, Modeling of diffusion in zeolites, Rev. Chem. Eng. 16 (2000) 71. [61] C. Uebing, R. Gomer, A Monte Carlo study of surface diffusion coefficients in the presence of adsorbate–adsorbate interactions. I. Repulsive interactions, J. Chem. Phys. 95 (1991) 7626. [62] R. Krishna, J.M. van Baten, Diffusion of alkane mixtures in zeolites: validating the Maxwell-Stefan formulation using MD simulations, J. Phys. Chem. B 109 (2005) 6386.

Chapter

14

First-principles Models of Metal Membranes

331

[63] A.I. Skoulidas, D.S. Sholl, R. Krishna, Correlation effects in diffusion of CH4/CF4 mixtures in MFI zeolite. A study linking MD simulations with the Maxwell-Stefan formulation, Langmuir 19 (2003) 7977. [64] C. Ling, D.S. Sholl, First-principles screening of PdCuAg ternary alloys as H2 purification membranes, J. Membr. Sci. 371 (2011) 189. [65] S. Keskin, D.S. Sholl, Screening metal-organic framework materials for membrane-based methane/carbon dioxide separations, J. Phys. Chem. C 111 (2007) 14055. [66] A.I. Skoulidas, D.S. Sholl, Direct tests of the Darken approximation for molecular diffusion in zeolites using equilibrium molecular dynamics, J. Phys. Chem. B 105 (2001) 3151. [67] C.H. Mak, H.C. Andersen, S.M. George, Monte Carlo studies of diffusion on inhomogeneous surfaces, J. Chem. Phys. 88 (1988) 4052. [68] H. Chen, D.S. Sholl, Efficient simulation of binary adsorption isotherms using transition matrix Monte Carlo, Langmuir 22 (2006) 709. [69] T. Vuong, P.A. Monson, Monte Carlo simulation studies of heats of adsorption in heterogeneous solids, Langmuir 12 (1996) 5425.

This page intentionally left blank

Chapter 15

High Performance Ultrafiltration Membranes: Pore Geometry and Charge Effects Andrew L. Zydney* Department of Chemical Engineering, The Pennsylvania State University, University Park, Pennsylvania, USA * Corresponding author: E-mail address: [email protected]

INTRODUCTION Ultrafiltration (UF) is now a well-established process for concentration, buffer exchange, and purification of a wide range of feed streams, including both native and recombinant proteins [1]. The key parameters describing the performance of these ultrafiltration processes are the filtrate flux (Jv), which is directly related to the membrane permeability (Lp): Lp ¼

Jv DP

ð15:1Þ

and the selectivity (c), which is determined by the degree of protein retention: c¼

1 Sa

ð15:2Þ

where Sa is the protein sieving coefficient, defined as the ratio of the protein concentration in the filtrate solution (Cf) to the protein concentration in the solution immediately upstream of the membrane (Cw). Equation (15.1) is valid for traditional ultrafiltration processes involving the separation of the desired protein product from small buffer components and impurities, both of which should have a sieving coefficient of Sa  1. In general, there is a tradeoff between the permeability and selectivity as shown in Figure 15.1 for data obtained using a series of polysulfone and cellulosic UF membranes using bovine serum albumin as the target protein [2]. Membranes with high permeability have low selectivity and vice versa. Membrane Science and Technology, Vol. 14. # 2011, Elsevier B.V. All rights reserved.

333

334

Inorganic, Polymeric and Composite Membranes

10000 Cellulosics Polysulfones

Selectivity, y

1000

100 Theory

10

1

0

6 8 2 4 Permeability, Lp (× 109 m s-1 Pa-1)

10

FIGURE 15.1 Selectivity–permeability tradeoff for a variety of cellulosic and polysulfone ultrafiltration membranes using bovine serum albumin as a model protein. Solid curve is a model calculation accounting for thermodynamic (partitioning) and hydrodynamic interactions for a membrane having a log-normal pore size distribution with coefficient of variation s=R ¼ 0:2. Adapted from Ref. [2].

This behavior is a direct result of the physics governing fluid and solute transport in ultrafiltration; membranes with large pore size have both high permeabilities (due to the lower resistance to fluid transport) and high sieving coefficients (due to the minimal steric exclusion of the protein from the pores). Although the performance characteristics of commercial ultrafiltration membranes have improved over the past 40 years, the overall level of improvement has been relatively small [3]. This is a particular concern in high performance applications encountered in bioprocessing where very high selectivities (low sieving coefficients) are required to achieve the high levels of protein recovery required during ultrafiltration processes [1]. For example, the protein yield in constant volume diafiltration process used for buffer exchange and impurity removal is given as [1] Y ¼ expðSa ND Þ

ð15:3Þ

A typical buffer exchange process requires ND ¼ 10 diavolumes (cumulative filtrate volume divided by the constant retentate volume); thus, a membrane selectivity of c  195 is needed to achieve at least 95% product yield. Most efforts to improve the performance characteristics of UF membranes have focused on trying to narrow the pore size distribution, typically by changing the casting conditions or by the formation of composite membrane structures. However, this approach has met with relatively little success, in part

Chapter

15

High Performance Ultrafiltration Membranes

335

because the pore size distribution in existing UF membranes has a fairly small effect on the permeability–selectivity tradeoff [2]. For example, the transport properties of current UF membranes are well described by a log-normal pore size distribution with coefficient of variation (ratio of standard deviation to mean pore size) of approximately 0.2 [4]. These membranes give a permeability of 1.3  10 9 m s 1 Pa 1 (corresponding to 32 L m 2 h 1 psi 1) for a selectivity of c ¼ 100 [2]. Completely eliminating the pore size distribution, that is, having a membrane with perfectly uniform pores of a single size, only provides a 60% increase in permeability [2]. Recent studies have demonstrated that it is possible to significantly increase the performance of UF membranes by changing the membrane pore geometry or by using electrically charged membranes to enhance the retention of likecharged proteins [5,6]. This chapter examines the theoretical basis for this improved performance, with model calculations compared with experimental results for protein transport through nanoporous silicon membranes with well-defined slit-shaped pores and with charge-modified cellulosic membranes. Calculations are also provided for the effects of concentration polarization on both the selectivity and the effective permeability. These results provide new insights into the behavior of ultrafiltration membranes with enhanced performance characteristics.

PORE GEOMETRY EFFECTS Fluid Flow The hydraulic permeability for a membrane with uniform cylindrical pores can be evaluated using the Hagen–Poiseuille equation as Lp ¼

eR2 8mdm

ð15:4Þ

where e is the membrane porosity (pore area per total membrane area), R is the radius of the cylindrical pores, m is the solution viscosity, and dm is the membrane thickness (pore length). The corresponding equation for a membrane with uniform slit-shaped pores is Lp ¼

eh2 3mdm

ð15:5Þ

where h is the slit half-width (Figure 15.2). Thus, a membrane that has slitshaped pores that are just small enough to completely exclude a given protein (h ¼ rprotein) will have a permeability that is a factor of 8/3 larger than that for a membrane with cylindrical pores (with R ¼ rprotein).The increase in permeability for a membrane with slit-shaped pores is a direct result of the different hydrodynamics in the two pore geometries. The hydraulic diameter of a membrane with slit-shaped pores is

336

Inorganic, Polymeric and Composite Membranes

FIGURE 15.2 Schematic diagram showing partitioning of a spherical solute into either a cylindrical pore of radius R or a slit-shaped pore with halfwidth h.

w

2R

dh ¼

2h

4wh w þ 2h

ð15:6Þ

where w is the width of the pore (see Figure 15.2). Equation (15.6) reduces to dh ¼ 4h for a membrane with very wide slits (aspect ratio h/w  1) compared to dh ¼ 2R for a membrane with cylindrical pores. This twofold difference in the hydraulic diameter leads to the much higher permeability for a slit-pore membrane with the same critical dimension (h ¼ R); the actual difference in permeability (a factor of 8/3) reflects the change in velocity profiles associated with the slit versus cylindrical pore geometry.

Solute Transport Solute transport through ultrafiltration membranes is due to the combination of solute partitioning between the bulk solution and the pore (thermodynamics) and hydrodynamic interactions between the solute and the pore wall. Both of these vary with the pore size, giving rise to a gradual variation in the sieving coefficient with increasing pore size instead of a sharp cut-off. The sieving coefficient is typically expressed as [7] Sa ¼ fKc

ð15:7Þ

where f is the equilibrium partition coefficient, and Kc is the hydrodynamic hindrance factor for convective transport. For a hard-sphere solute in cylindrical pores in the absence of any long-range interactions the partition coefficient is given as [7]
rprotein 2 ð15:8Þ f¼ 1 R while the corresponding expression for a slit-shaped pore is [7] rprotein f¼1 h

ð15:9Þ

Equations (15.8) and (15.9) both give f ¼ 0 when the critical pore dimension (R or h) is equal to the protein radius. However, when the protein radius is equal to one-half the critical pore dimension, the partition coefficient in the cylindrical

Chapter

15

High Performance Ultrafiltration Membranes

337

pore is only half as large as that in a slit-shaped pore, with a corresponding twofold increase in the selectivity (neglecting differences in the hydrodynamic hindrance factor in the two pore geometries). Thus, the greater permeability of the slit-shaped pore is at least partially compensated for by the increase in partition coefficient due to the different steric effects in the two pore geometries. Theoretical expressions for the hydrodynamic hindrance factors for convection are available in the literature [7]. As discussed elsewhere, the hindrance factor for convection in a cylindrical pore varies between 1.0 and 1.47 [8]; thus, the dominant contribution to the selectivity is the partition coefficient. Equations (15.4) and (15.8) can be combined to give the following very simple expression for the selectivity–permeability tradeoff for a membrane with uniform cylindrical pores: !2 rprotein ð15:10Þ c ¼ 1  bR pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi mLp =dm where bR ¼


e 1=2 8

ð15:11Þ

A similar expression can be developed for a membrane with uniform slit-shaped pores using Equations (15.5) and (15.9): !1 rprotein ð15:12Þ c ¼ 1  bh pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi mLp =dm where bh ¼


e 1=2 3

ð15:13Þ

Figure 15.3 shows a comparison of the selectivity–permeability tradeoff for ultrafiltration membranes with cylindrical and slit-shaped pores. The results have been plotted as a function of the scaled permeability: Lp ¼

mLp dm e

ð15:14Þ

to account for differences in membrane porosity and/or thickness for the different ultrafiltration membranes. The experimental data were taken from Figure 15.1 with the ratio of the porosity to skin thickness of the asymmetric membranes set equal to 106 m (consistent with e ¼ 0.5 and dm ¼ 0.5 mm). The solid curves are the model calculations for membranes with uniform pore size as given by Equations (15.10) and (15.12). More detailed calculations that include the contribution from the hindrance factor for convection show very similar behavior [9]. The results for the membrane with slit-shaped pores

338

Inorganic, Polymeric and Composite Membranes

1000 Cellulosics Polysulfones Slit-shaped Selectivity, y

100

10

1

0

2

4

6

8

10

Scaled permeability, L*p = mLpdm/e (m2 × 1018) FIGURE 15.3 Selectivity–permeability tradeoff for a variety of cellulosic and polysulfone ultrafiltration membranes and for a nanoporous silicon membrane with slit-shaped pores using bovine serum albumin as a model protein. Solid curves are model calculations based on the partition coefficient of a spherical solute into a membrane with uniform cylindrical or slit-shaped pores. Adapted from Ref. [9].

clearly lie above and to the right of those for the membrane with cylindrical pores; the change from cylindrical to slit-shaped pores provides a significant improvement in membrane performance due to the impact of pore geometry on solute and solvent transport. At high selectivities, that is, under conditions where the protein is nearly the same size as the critical pore dimension, the permeability of the slit-pore membrane is a factor of 8/3 larger than that of the cylindrical pore membrane, consistent with Equations (15.1) and (15.2). The filled triangles in Figure 15.3 represent data obtained with silicon nanopore membranes having slit-shaped pores [9,10]. These membranes were produced from a 400-mm-thick silicon wafer that was coated with a 500 nm-thick layer of low-stress silicon nitride followed by a 5 mm-thick of polycrystalline silicon (polysilicon). The polysilicon layer could be patterned by photolithography and reactive ion etching. A thin conformal film of SiO2 was grown on the polysilicon and subsequently etched with concentrated hydrofluoric acid to produce the slit-shaped pores. The resulting nanoporous membranes had pores that were approximately 45 mm in length and between 5 and 30 nm in half-width, with a total membrane thickness of 4 mm. The silica surface was modified by covalent attachment of poly(ethylene glycol) (PEG) to minimize protein adsorption and fouling. Additional details on the formation of the silicon membranes are provided elsewhere [11]. The porosity of the nanoporous silicon membranes was calculated directly from the measured pore width (determined from SEM) and the known pore density. The permeability–selectivity data for the silicon membranes are in good

Chapter

15

High Performance Ultrafiltration Membranes

339

agreement with the model calculations, with performance characteristics that are significantly better than those for polymeric membranes having largely cylindrical pores [9]. However, it is important to note that the porosity of the membranes examined in Figure 15.3 was less than 0.0006; this very low porosity would be impractical for most commercial applications of ultrafiltration. Recent efforts in optimizing the photolithography have resulted in membranes with much greater pore density and somewhat smaller membrane thickness, providing more than a two orders of magnitude increase in the hydraulic permeability.

Pore Size Distribution Effects All the model calculations presented in the previous sections were for membranes having perfectly uniform pores, that is, for a membrane with cylindrical pores all having radius R. Actual ultrafiltration membranes have a pore size distribution, which is typically described using a log-normal density function [4]: 8
2 9 2 1=2 > > R  = < h i ln R ½1 þ ðs=RÞ  1=2 n0  2Þ exp  nðRÞ ¼ pffiffiffiffiffiffi lnð1 þ ðs=RÞ  2 > > R 2p 2 ln½1 þ ðs=RÞ ; : ð15:15Þ where R is the mean and s2 the variance of the distribution, respectively. The log-normal density function has been used extensively in the past to describe the pore size distribution of a variety of membranes [4]. The effects of a pore size distribution on solute and solvent transport can be evaluated theoretically by appropriate integration of the expressions for the fluid velocity and solute flux over the pore size distribution. For example, the sieving coefficient for a membrane with a distribution of cylindrical pores is evaluated as [12] 1 Ð

Sa ¼

0

nðRÞSa ðRÞR4 dR 1 Ð

nðRÞR4 dR

ð15:16Þ

0

where the R4 dependence comes from the dependence of the fluid velocity on R at a fixed transmembrane pressure drop. The solid curve in Figure 15.1 shows the predicted permeability–selectivity tradeoff for a membrane composed of a parallel array of cylindrical pores having a log-normal distribution with coefficient of variation s=R ¼ 0:2. The model is in very good agreement with the experimental results for the cellulosic and polysulfone membranes, indicating that the performance characteristics of these polymeric ultrafiltration membranes are well described using the log-normal pore size distribution with coefficient of variation equal to 0.2.

340

Inorganic, Polymeric and Composite Membranes

Mochizuki and Zydney [13] showed that the sieving coefficients for a range of commercial ultrafiltration membranes could be well described using a simple theoretical expression originally developed by Giddings et al. [14] to describe solute partitioning into a porous media formed by the intersection of a random array of planes:
r  protein Sa ¼ exp  ð15:17Þ s where s is the specific area of the pore, equal to the pore volume divided by the pore surface area. Equation (15.17) has been shown to be in good agreement with Monte Carlo calculations of the partitioning of rigid nonspherical particles in pores of several different geometries [14]. Opong and Zydney [15] evaluated the specific pore area in terms of the membrane permeability using the Kozeny–Carman equation as   2mdm Lp 1=2 s¼ ð15:18Þ e Equations (15.17) and (15.18) can be combined to develop a very simple expression for the relationship between the selectivity and permeability of an ultrafiltration membrane with a distribution of irregularly shaped pores: ! 2bR rprotein c ¼ exp pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð15:19Þ mLp =dm Experimental data for BSA along with limited data for cytochrome c are plotted in Figure 15.4 in the form suggested by Equation (15.19), with the y-axis showing the selectivity (on a logarithmic scale) and the x-axis showing the permeability (plotted as the reciprocal of the square root of 1/Lp). The data for BSA were obtained with both cellulosic and polysulfone membranes (results from Figure 15.1), while the data for cytochrome c were from a series of cellulosic membranes [6]. The results for both proteins are highly linear when plotted in this manner, consistent with predictions of Equation (15.19), although the best fit line does not go through the origin. The reason for this discrepancy is unclear, although it could be related to differences in porosity or thickness for the different pore size membranes. The slope of the best fit line for BSA is 0.26 0.03 nm1/2 while that for cytochrome c is 0.15 0.02 nm1/2. The greater slope for BSA is consistent with its greater size: rBSA ¼ 3.6 nm versus rcyt ¼ 2.0 nm. The ratio of the measured slopes is 1.7 0.2, which is in very good agreement with the ratio of the protein radii, rBSA/rcyt ¼ 1.8, providing further support for the simple theoretical expression given by Equation (15.19). It is also possible to directly compare the best fit values of the slope with the theoretical model assuming a membrane porosity of e ¼ 0.5 and a membrane thickness of dm ¼ 0.5 mm. This gives 2bRrBSA ¼ 1.8 nm, which is about a factor

Chapter

15

341

High Performance Ultrafiltration Membranes

1000 Cellulosics Polysulfones Cytochrome c

Selectivity, y

100

10

1

0

10 20 30 40 Permeability function, (mLp)-1/2 (nm-1/2)

50

FIGURE 15.4 Selectivity as a function of the reciprocal of the square root of the product of the solution viscosity and the membrane permeability. Shaded symbols are for bovine serum albumin. Filled triangles are for cytochrome c. Solid lines are linear regression fits to the data.

of 3 smaller than that found experimentally. This discrepancy could be due to errors in the estimated values of the membrane porosity and thickness or to the oversimplifications in the use of the partitioning model from Giddings et al. [14] to describe solute transport for these ultrafiltration membranes.

ELECTROSTATIC INTERACTIONS Fluid Flow Fluid flow through a charged pore is reduced from that in an uncharged system due to the electrical stresses arising from the forces on the charged ions. The surface charge on the pore wall causes an unequal partitioning of the charged ions into the pore. Thus, the pressure-driven fluid flow causes a greater convective flux of the counterions, that is, the negatively-charged Cl ions in a positively charged pore, through the membrane [16]. In the absence of any other effects, this excess convective flux would generate a filtrate solution enriched in Cl compared to Naþ during filtration of a feed containing NaCl, which would violate the conditions of electroneutrality in the filtrate. In reality, the initial accumulation of Cl in the filtrate solution generates a net electrical potential across the membrane that is typically referred to as the streaming potential. The electrophoretic ion flux caused by the induced streaming potential exactly balances the excess convective ion flux associated with the fluid flow, with the net result that the filtrate solution remains electrically neutral.

342

Inorganic, Polymeric and Composite Membranes

The value of the streaming potential, Ez, for flow through a membrane can be evaluated directly from the constraint that there is no net current flow yielding the following expression for a membrane composed of a parallel array of uniform cylindrical pores [16]:    qp I2 ðkRÞ DP Ez ¼  ð15:20Þ mk I1 ðkRÞ Leff dm where Leff is the effective conductivity of the electrolyte solution in the cylindrical pores, qp is the surface charge density of the pore, and I1(kR) and I2(kR) are modified Bessel functions of the first kind or order 1 and 2, respectively. The effective solution conductivity is a function of the solvent properties and the overall salt concentration as discussed by Newman [16]. k is the reciprocal of the Debye length (or double-layer thickness) and is given as P  2 1=2 e NAv z2i coi ð15:21Þ k¼ eo er kB T where e is the electron charge, NAv is Avogadro’s number, zi and cio are the valence and bulk concentration of ion i, respectively, eo is the permittivity of free space, er is the dielectric constant of the fluid, kB is Boltzmann’s constant, and T is the absolute temperature. The induced streaming potential reduces the net rate of fluid flow through the membrane due to the effects of the electrical forces on the charged ions. This phenomenon is known as counter-electroosmosis since the solvent flow that is generated by the streaming potential is always in the opposite direction of the pressure-driven flow. The magnitude of this counter-electroosmotic flow can be evaluated by solving the Navier–Stokes equation including the electrical force terms, with the ion concentration profiles in the pore described by a Boltzmann distribution. The results for a membrane composed of a parallel array of uniform cylindrical pores can be written as [16]  2  qp I2 ðkRÞ Jv R þ Ez ¼ DP ð15:22Þ kmI1 ðkRÞ e 8mdm The first term on the right-hand side of Equation (15.22) represents the pressuredriven (Poiseuille) flow, while the second term describes the back flow associated with counter-electroosmosis. Equations (15.20) and (15.22) can be combined to evaluate the membrane permeability:   qp I2 ðkRÞ 2 eR2 e  ð15:23Þ Lp ¼ 8mdm Leff dm mkI1 ðkRÞ The permeability decreases with the square of the surface charge density of the pore; the quadratic dependence on qp arises from the combination of the linear dependence of both the streaming potential and the electroosmotically driven flow on qp.

Chapter

15

343

High Performance Ultrafiltration Membranes

1 100 mM

Permeability ratio, Lp/Lpo

0.8 10 mM 0.6

5 mM

0.4

0.2

0 0

2

4

6

8

10

Surface charge density, qp (mC m-2) FIGURE 15.5 Effect of counter-electroosmosis on the ratio of the membrane permeability to the permeability in the absence of electrostatic interactions. Shaded squares are experimental data for a series of charge-modified cellulose membranes using a 10 mM KCl solution [5]. Solid curves are model calculations based on Equation (15.23).

The effect of the membrane charge on the hydraulic permeability is examined theoretically in Figure 15.5 for several values of the solution ionic strength. Calculations were performed for a membrane with R ¼ 4 nm, e ¼ 0.5, and dm ¼ 0.5 mm, which is typical of the asymmetric ultrafiltration membranes examined previously in Figure 15.1. The symbols are experimental data for the normalized permeability through a series of charge-modified cellulose membranes obtained using a 10-mM KCl solution [5]. The data are in very good agreement with the model calculations, even though the model neglects the effects of a pore size distribution or the change in solution conductivity due to ion partitioning into the pore. The membrane charge has essentially no effect on the fluid flow at high ionic strength (100 mM KCl) since the electrical double layer is very thin under these conditions. Thus, the region of the pore that has an excess of counterions is located very close to the pore wall where the fluid velocity is small. In addition, the high electrical conductivity of the solution reduces the magnitude of the streaming potential; a very low potential is needed to generate a significant electrophoretic ion flux when the conductivity is high. The net result is that there is minimal counter-electroosmosis under these conditions. The effects of counter-electroosmosis become much more pronounced at low ionic strength, although the permeability in the 10 mM ionic strength solution remains more than 65% of the zero charge value even for the highest surface charge density examined in Figure 15.5.

344

Inorganic, Polymeric and Composite Membranes

Solute Transport Although ultrafiltration was originally viewed as a purely size-based separation process, it is now well established that the rate of protein transmission through a semipermeable membrane can be strongly influenced by long-range electrostatic interactions [17,18]. Electrostatic interactions can be exploited to develop high performance ultrafiltration membranes with significantly better selectivity (for a given value of the permeability) than conventional UF membranes due to the strong electrostatic repulsion of like-charged proteins [1]. Electrostatic effects can also be used to achieve high-resolution protein separations by exploiting differences in electrical charge of the product and impurity [1,19]. The equilibrium partition coefficient for a charged protein in a charged pore is determined by the total interaction potential ctotal(r) as [7] Cpore 2 ¼ f¼ Cbulk R2

ðR

  ctotal ðrÞ exp  rdr kB T

ð15:24Þ

0

where r is the radial coordinate within the cylindrical pore, kB is the Boltzmann constant, and T is the absolute temperate. The total interaction potential has contributions from steric (hard-sphere), electrostatic, and van der Waals forces. For purely hard-sphere interactions, ctotal ! 1 when the solute overlaps the pore wall and is zero in the pore interior, with the integral in Equation (15.1) reducing to Equation (15.8). Smith and Deen [20] developed the first rigorous analytical expressions for the electrostatic potential for a spherical solute in a cylindrical pore by solving the linearized Poisson–Boltzmann equation using matched asymptotic expansions in cylindrical and spherical coordinates. The results for interactions at constant surface charge density are conveniently expressed as cE ¼ ðAs s2s þ Asp ss sp þ Ap s2p Þ=Aden kB T

ð15:25Þ

where ss and sp are the dimensionless surface charge densities of the solute (protein) and pore: ss ¼

eRqs eo er kB T

ð15:26Þ

with eo the permittivity of free space, er the dielectric constant of the solution, e the charge on an electron, and qs the protein surface charge density. The coefficients As, Asp, Ap, and Aden are all positive functions of the solution ionic strength, solute radius, and pore radius: h i 1 1=2 4 tl ð K1 ðt2 þ y2 Þ 4ptl e h i dy As ¼ ð15:27Þ 1 þ tl I1 ðt2 þ y2 Þ1=2 0

Chapter

15

345

High Performance Ultrafiltration Membranes

Ap ¼



p2 ð1 þ tlÞetl  ð1  tlÞetl t2 I12 ðtÞ

ð15:28Þ

4p2 l2 I1 ðtÞ

ð15:29Þ

Asp ¼



Aden ¼ ptð1 þ tlÞetl  ð1 þ tlÞetl  ð1  tlÞetl

1 ð

0

h i K1 ðt2 þ y2 Þ1=2 h i dy I1 ðt2 þ y2 Þ1=2 ð15:30Þ

where t ¼ kR is the dimensionless pore radius, k is the inverse Debye length, and I1 and K1 are modified Bessel functions. Equations (15.27) to (15.30) are valid for a solute located at the pore axis. Corresponding results are available for arbitrary radial positions as well as for interactions at constant surface potential instead of constant surface charge density [20,21]. Pujar and Zydney [22] subsequently extended this analysis to include the effects of charge regulation using a linearized form of the charge regulation boundary condition. The resulting equations account for the change in surface charge/potential of the protein and pore wall associated with the alteration in the local electrical potential field when the charged protein enters the pore. Figure 15.6 shows experimental data for the selectivity determined using cytochrome c at an ionic strength of 2 mM for a series of positively charged

Selectivity, y

100

10

1

0

0.4

0.8 1.2 1.6 Surface charge density, qp (mC m−2)

2

FIGURE 15.6 Selectivity for cytochrome c as a function of the membrane surface charge density. Solid line is linear regression fit to the data.

346

Inorganic, Polymeric and Composite Membranes

ultrafiltration membranes generated by covalent attachment of different amine functionalities to a base cellulose membrane [23]. The ligands all had similar effective length, with the different charge densities arising from the different structures of the ligands, including differences in the number of amine groups per ligand. In each case, the membrane surface charge density was evaluated from measurements of the membrane streaming potential using Equation (15.20). The data point near the origin is for an unmodified cellulose membrane; this membrane had a slightly larger pore size than the charge-modified membranes due to the pore constriction associated with the small ligands used for the surface modification. The selectivity varies linearly with the membrane surface charge density on the semi-log plot, consistent with the theoretical expression for the partition coefficient given by Equations (15.24) and (15.25). The effect of electrostatic interactions on the permeability–selectivity tradeoff for cytochrome c is examined in Figure 15.7. The shaded squares are data for several unmodified composite regenerated cellulose membranes at high ionic strength, that is, under conditions where electrostatic interactions are negligible. The filled symbols represent data for three charge-modified membranes generated by covalent attachment of a quaternary amine functionality to the membrane. The different surface charge densities were produced by controlling the reaction time [5]. These charge-modified membranes provide much higher selectivity than the unmodified membranes; the slightly smaller permeability is associated with the constriction of the pores by the charged 10000 Un-modified qp = 0.2 mC/m2 qp = 3.5 mC/m2 qp = 7.0 mC/m2

Selectivity, y

1000

100

qp = 3.5 mC/m2

10 Uncharged 1

0

1 2 3 Permeability, Lp (× 109 m s-1 Pa-1)

4

FIGURE 15.7 Selectivity–permeability tradeoff for both unmodified and several charge-modified ultrafiltration membranes using cytochrome c. Data for charged membranes were obtained in 10 mM ionic strength solutions. Black curve is model calculation in the absence of electrostatic interactions. Gray curve is model calculation accounting for the effects of electrostatic interactions on both solute partitioning and fluid flow.

Chapter

15

High Performance Ultrafiltration Membranes

347

ligands. The solid curves are model calculations for membranes consisting of a parallel array of uniform cylindrical pores. The black curve represents the predicted behavior in the absence of any electrostatic interactions, while the gray curve represents the model calculations using Equations (15.24)–(15.30) for a membrane with a uniform surface charge density of 3.5  10 3 C/m2. The calculations for the charged membranes lie somewhat above the experimental value for qp ¼ 3.5  10 3 C/m2, with this discrepancy likely due to the effects of the pore size distribution on solute and solvent transport through the charged membranes.

CONCENTRATION POLARIZATION EFFECTS All of the experimental and theoretical results presented in the last two sections were developed assuming that the effects of concentration polarization were negligible. Concentration polarization refers to the accumulation of a concentrated layer of retained protein at the upstream surface of the membrane. Concentration polarization can affect both the selectivity and effective permeability. The accumulation of retained protein at the membrane surface reduces the filtrate flux due to osmotic pressure limitations and/or by providing an additional hydraulic resistance to fluid flow, while the increase in local protein concentration leads to an increase in protein transmission and a corresponding reduction in the selectivity. The effects of concentration polarization are typically described using a simple stagnant film model [26]:   Cw  Cf ð15:31Þ Jv ¼ km ln Cb  Cf where km is the bulk mass transfer coefficient in the particular membrane module, and Cb, Cf, and Cw are the protein concentrations in the bulk solution, the filtrate, and the solution immediately upstream of the membrane (the “wall”), respectively. Equation (15.31) can be rearranged to solve for the observed sieving coefficient in terms of the actual sieving coefficient (Sa ¼ Cf/Cw) yielding: So ¼

Cf Sa ¼ Cb ð1  Sa Þ expðJv =km Þ þ Sa

ð15:32Þ

At low filtrate flux, So ¼ Sa since there is minimal polarization and Cw ¼ Cb. However, at high flux, the observed sieving coefficient approaches a value of one irrespective of the value of Sa as long as the membrane is at least partially permeable to the protein of interest. Thus, the observed selectivity, given by Equation (15.2) but with Sa replaced by So, decreases to a value of one (i.e., no selectivity) as the filtrate flux increases. Figure 15.8 shows a plot of the selectivity–permeability tradeoff over a range of Jv/km. Model calculations were performed accounting for the effects

348

Inorganic, Polymeric and Composite Membranes

10000

Selectivity, y = 1/sο

1000

Jv/km = 0

Jv/km = 1 100 Jv/km = 2 10 Jv/km = 5 1

0

3 1 2 Permeability, Lp (× 109 m s−1 Pa−1)

4

FIGURE 15.8 Effect of concentration polarization on the selectivity–permeability tradeoff for very dilute protein solution. Model calculations are for a membrane with uniform cylindrical pores.

of concentration polarization on the selectivity, with the permeability assumed to be independent of the degree of polarization. This would be consistent with the behavior of a very dilute feed solution in which Cw remains well below the concentration at which the protein osmotic pressure becomes significant (corresponding to Cw much less than the gel concentration). The selectivity becomes infinite at a permeability of Lp ¼ 1.6  10 9 m s 1 Pa 1 (40 L m 2 h 1 psi 1) irrespective of the degree of polarization since the observed sieving coefficient is equal to zero whenever Sa ¼ 0. The selectivity decreases significantly with increasing polarization, that is, with larger Jv/km. For example, the permeability corresponding to c ¼ 10 decreases from Lp ¼ 3.5  10 9 m s 1 Pa 1 (86 L m 2 h 1 psi 1) in the absence of any polarization (Jv/km ¼ 0) to Lp ¼ 2.5  10 9 m s 1 Pa 1 (62 L m 2 h 1 psi 1) at Jv/km ¼ 1 and then to 1.7  10 9 m s 1 Pa 1 (42 L m 2 h 1 psi 1) at Jv/km ¼ 5. For more concentrated feed solutions, concentration polarization effects will lead to a reduction in the filtrate flux due to osmotic pressure effects [8]: Jv ¼ Lp ½DP  sDP

ð15:33Þ

where s is the osmotic (Staverman) reflection coefficient and DP is the osmotic pressure difference between the solution immediately adjacent to the membrane and that in the filtrate solution. The osmotic pressure is typically evaluated using a virial expansion in the protein concentration [24]: Pw ¼

RT Cw þ B2 C2w þ B3 C3w M

ð15:34Þ

Chapter

15

High Performance Ultrafiltration Membranes

349

where R is the ideal gas constant, T is the absolute temperature, and M is the protein molecular weight. The filtrate flux can be evaluated as a function of the applied transmembrane pressure drop by simultaneous solution of Equations (15.31), (15.33), and (15.34) using Sa ¼ Cf/Cw along with the assumption that: s ¼ 1  Sa

ð15:35Þ

based on Onsager reciprocity [8]. Several investigators have shown that the resulting model calculations are in good agreement with experimental data for the ultrafiltration of BSA using independent measurements for the osmotic virial coefficients [15]. Figure 15.9 shows experimental data for the effective permeability (bottom panel) and the effective selectivity (top panel) as a function of the transmembrane pressure drop for the ultrafiltration of a 25 g/L BSA solution through a 50 kDa nominal molecular weight cut-off membrane [15]. The effective permeability was defined by Equation (15.1) using the filtrate flux obtained with the FIGURE 15.9 Effects of concentration polarization on the selectivity (top panel) and effective permeability (bottom panel). Experimental data are for BSA ultrafiltration through a 50-kDa membrane [15]. Solid curves are model calculations developed by simultaneous solution to Equations (15.31), (15.33), and (15.34) as described in the text.

Selectivity, y

1000

100

10

1

Effective permeability, Jv/ΔP (× 109 m s−1 Pa−1)

0.8

0.6

0.4

0.2

0

0

40 80 120 160 Transmembrane pressure, ΔP (kPa)

350

Inorganic, Polymeric and Composite Membranes

protein solution (instead of the flux obtained with protein-free buffer), while the effective selectivity was defined by Equation (15.2) using the observed sieving coefficient (So) in place of the actual sieving coefficient (Sa). The solid curves are model calculations developed from simultaneous solution to Equations (15.31), (15.33), and (15.34) with Lp ¼ 0.93  10 9 m s 1 Pa 1 (23 L m 2 h 1 psi 1) and Sa ¼ 0.001 as determined experimentally. The osmotic virial coefficients for BSA were taken from Vilker et al. [24] as B2 ¼ 9.22 10 3 L g 1 and B3 ¼ 3.01 10 5 L2 g 2 at p. 7.4. The bulk mass transfer coefficient in the stirred cell was km ¼ 5.2  10 6 m s 1 [15]; this value is in good agreement with available correlations for mass transfer coefficients in a stirred cell [25]. The selectivity decreases monotonically with increasing transmembrane pressure drop due to the reduction in the observed sieving coefficient associated with concentration polarization effects. The effective permeability initially increases with increasing transmembrane pressure at very low DP because of the significant osmotic pressure associated with the 25 g/L bulk protein concentration; the filtrate flux and thus the effective permeability given by Equation (15.33) goes to zero at a small positive value of the transmembrane pressure where DP ¼ sDP. The maximum value of the effective permeability under these conditions is only 0.54  10 9 m s 1 Pa 1 (13 L m 2 h 1 psi 1) compared to the actual permeability of the 50 kDa membrane of Lp ¼ 0.93  10 9 m s 1 Pa 1 (23 L m 2 h 1 psi 1). The effective permeability decreases at higher transmembrane pressures due to the large increase in the osmotic pressure difference associated with the increase in Cw at high degrees of concentration polarization. The model calculations are in good agreement with the experimental data, although the experimental results do not show the maximum in the effective permeability due to the absence of any measurements at very low DP. The results in Figures 15.8 and 15.9 clearly demonstrate that concentration polarization effects can significantly degrade the performance characteristics of membranes that exploit either electrostatic interactions and/or pore geometry effects to achieve high-resolution separations. This requires the use of membrane modules with high bulk mass transfer coefficients and/or the use of low filtrate flux. The latter occurs naturally with very tight (low molecular weight cut-off) membranes that are used to achieve high retention of small proteins like insulin (MW ¼ 5.8 kDa) and human growth hormone (MW ¼ 22 kDa). The ultrafiltration behavior for larger proteins like monoclonal antibodies (MW  155 kDa) tends to be dominated by concentration polarization effects, with the filtrate flux approaching its pressure-independent value even at relatively low transmembrane pressures.

CONCLUSIONS The experimental and theoretical results presented in this chapter clearly demonstrate that ultrafiltration membranes with electrically charged surfaces or with slit-shaped pore geometry can have significantly enhanced performance,

Chapter

15

High Performance Ultrafiltration Membranes

351

as defined by the tradeoff between the selectivity and permeability, compared to traditional ultrafiltration membranes that have cylindrical pores. This improved performance is a direct result of the change in solute and solvent transport arising from the hydrodynamic and electrostatic interactions in narrow pore geometries. Concentration polarization can also have a significant effect on the effective permeability and selectivity due to the alteration in the protein sieving coefficient and the change in protein osmotic pressure associated with the highly concentrated protein solution that accumulates at the upstream surface of the membrane. An improved understanding of these phenomena can be used to develop high performance ultrafiltration membranes and processes. These effects will be of particular interest in applications of ultrafiltration for the concentration and purification of high-value recombinant protein products in the biotechnology industry where very high selectivities are needed to obtain the desired product yield.

ACKNOWLEDGMENT This work was supported in part by a grant from the National Institutes of Health (R01 EB008049-01) and by the Walter L. Robb Family Endowed Chair.

REFERENCES [1] R. van Reis, A.L. Zydney, Bioprocess membrane technology, J. Membr. Sci. 297 (2007) 16–50. [2] A. Mehta, A.L. Zydney, Permeability and selectivity analysis for ultrafiltration membranes, J. Membr. Sci. 249 (2005) 245. [3] A.L. Zydney, Membrane technology for purification of therapeutic proteins, Biotechnol. Bioeng. 103 (2009) 227–230. [4] A.L. Zydney, P. Aimar, M. Meireles, J.M. Pimbley, G. Belfort, Use of the log-normal probability density function to analyze membrane pore size distributions: functional forms and discrepancies, J. Membr. Sci. 91 (1994) 293. [5] A. Mehta, A.L. Zydney, Effect of membrane charge on flow and protein transport during ultrafiltration, Biotechnol. Prog. 22 (2006) 484. [6] A. Mehta, A.L. Zydney, Effect of spacer arm length on the performance of charge-modified ultrafiltration membranes, J. Membr. Sci. 313 (2008) 304–314. [7] W.M. Deen, Hindered transport of large molecules in liquid-filled pores, AIChE J. 33 (1987) 1409. [8] L.J. Zeman, A.L. Zydney, Microfiltration and Ultrafiltration: Principles and Applications, Marcel Dekker, New York, 1996. [9] D.M. Kanani, W.H. Fissell, S. Roy, A. Dubnisheva, A. Fleischman, A.L. Zydney, Permeability–selectivity analysis for ultrafiltration: Effect of pore geometry, J. Membr. Sci. 349 (2010) 405–410. [10] W.H. Fissell, A. Dubnisheva, A.N. Eldridge, A.J. Fleischman, A.L. Zydney, S. Roy, Highperformance silicon nanopore hemofiltration membranes, J. Membr. Sci. 326 (2008) 58. [11] C.A. Lopez, A.J. Fleischman, S. Roy, T.A. Desai, Evaluation of silicon nanoporous membranes and ECM-based microenvironments on neurosecretory cells, Biomaterials 27 (2006) 3075.

352

Inorganic, Polymeric and Composite Membranes

[12] S. Mochizuki, A.L. Zydney, Theoretical analysis of pore size distribution effects on membrane transport, J. Membr. Sci. 82 (1993) 211. [13] S. Mochizuki, A.L. Zydney, Dextran transport through asymmetric ultrafiltration membranes: comparison with hydrodynamic models, J. Membr. Sci. 68 (1992) 21. [14] J.C. Giddings, E. Kucera, C.P. Russell, M.N. Myers, Statistical theory for the equilibrium distribution of rigid molecules in inert porous networks, J. Phys. Chem. 72 (1968) 4397. [15] W.S. Opong, A.L. Zydney, Diffusive and convective protein transport through asymmetric membranes, AIChE J. 37 (1991) 1497. [16] J.S. Newman, Electrochemical Systems, Prentice Hall, Englewood Cliffs, 1991. [17] N.S. Pujar, A.L. Zydney, Electrostatic and electrokinetic interactions during protein transport through narrow pore membranes, Ind. Eng. Chem. Res. 33 (1994) 2473. [18] S. Saksena, A.L. Zydney, Effect of solution pH and ionic strength on the separation of albumin from immunoglobulins (IgG) by selective filtration, Biotech. Bioeng. 43 (1994) 960. [19] R. van Reis, J.M. Brake, J. Charkoudian, D.B. Burns, A.L. Zydney, High-performance tangential flow filtration using charged membranes, J. Membr. Sci. 159 (1999) 133. [20] F.G. Smith, W.M. Deen, Electrostatic double-layer interactions for spherical colloids in cylindrical pores, J. Colloid Interface Sci. 78 (1980) 444. [21] F.G. Smith, W.M. Deen, Electrostatic effects on the partitioning of spherical colloids between dilute bulk solution and cylindrical pores, J. Colloid Interface Sci. 91 (1983) 571. [22] N.S. Pujar, A.L. Zydney, Charge regulation and electrostatic interactions for a spherical particle in a cylindrical pore, J. Colloid Interface Sci. 192 (1997) 338. [23] M.M. Rohani, A. Mehta, A.L. Zydney, Development of high performance charged ligands to control protein transport through charge-modified ultrafiltration membranes, J. Membr. Sci, 362 (2010) 434. [24] V. Vilker, C. Colton, K. Smith, The osmotic pressure of concentrated protein solutions: effect of concentration and pH in saline solutions of bovine serum albumin, J. Colloid Interface Sci. 79 (1981) 548. [25] K.A. Smith, C.K. Colton, E.W. Merrill, L.B. Evans, Convective transport in a batch dialyzer: determination of the true membrane permeability from a single measurement, Chem. Eng. Prog. Symp. Ser. 64 (1968) 45. [26] W.F. Blatt, A. Dravid, A.S. Michaels, L. Nelsen, Solute polarization and cake formation in membrane ultrafiltration. Causes, consequences, and control techniques, in: J.E. Flinn (Ed.), Membrane Science and Technology, Plenum Press, New York, 1970, pp. 47–97.

Index

Page numbers followed by f indicate figures and t indicate tables.

A Acetic acid (AcOH) removal from biomass hydrolysates, 215–217, 215t materials and methods for, 219–221, 220t results of, 221–225, 222f, 223f, 225t, 226t water separation from, 176–179, 177f AcOH. See Acetic acid Activated diffusion, 118 mechanisms of, 118–119 Activation energy of permeation of He v. H2, 126–127, 127f He/H2 permeance ratio and, 125, 126f selectivity and, 126 of silica membranes, 81–82 calculation of, 82 of siloxane ring models, 84–85, 85f evolution of, 86–87, 86f ranking of, 87 Ag–SCF membrane, 246 Al2O3. See Alumina Alcohol, de-watering of, 184–186, 185f, 186f Algorithm, 3 AlOOH. See Boehmite sols Alternating reactant deposition CVD, 35 on Vycor glass, 41 a-Alumina carbon molecular sieves on, 97 in HF membranes, 65–66 MFI zeolite membranes synthesis on, 198–199, 199f zeolite membranes on, 93, 95–96, 181 g-Alumina, in HF membranes, 65–66, 68t, 72–73 gas separation mechanism, 74, 75f pore defect correction, 70–71, 70f Alumina (Al2O3)

for CVD of silica membrane, 26, 30, 48–56, 49f, 53f CO2 separation and, 54–55 with evacuation, 48–50, 49f extended counterdiffusion, 55 g-alumina and, 52–53, 53f in H2-H2O-HI mixture, 51–52 with Nanosil processes, 52 with new hybrid processing method, 54 with PTES and DPDES, 50 with steam pretreatment, 54 with TEOS, 48–51, 49f with TEOS and TMOS, 53–54 with TMOS, PTMS, and DMDPS, 55–56 Vycor glass v., 48 in CVD on Vycor glass, 38–39, 43 for CVI of silica membrane, with SiAc4, 51 pore size of, 48 in SCF membranes, 246 silica membranes with, 79 Alumina-silica membranes hydrogen permeance of, 80 permeation performance of, 87–88, 88t Aluminum isopropoxide, in boehmite sol preparation, 65 Aluminum tri-sec-butoxide (ATSB), for CVD on Vycor glass, with TEOS, 43 Ammonia (NH3), silica membranes and, 131–132 Amorphous metal membranes, DFT-based modeling of, 316–327 amorphous structures, 317–318, 318f binding energy of interstitial hydrogen, 318–319, 321–322, 321f computational approaches, 317 corrected diffusivities, 323–324, 325f H–H interactions, 319–320, 320f hydrogen diffusion, 317, 322–323 hydrogen permeability, 326–327, 327f

353

354 Amorphous metal membranes, DFT-based modeling of (cont.) hydrogen solubility, 317, 320–322, 321f thermodynamic correction factor, 324–325 XAmorphous silica membranes CVD of, for hydrogen separation, 61, 62–63t helium permeation through, 125–127, 126f, 127f on HF membranes g-alumina layers and, 68t, 72–73 concluding remarks on, 75–76 development of, 61 gas separation mechanism in, 74, 75f hydrogen permeation of, 61, 62–63t mesoporous silica layer of, 68t, 69–72, 70f, 71f, 72f preparation of, 65–66, 66f, 67f silica precursor and carrier gas flow rate, 73, 74f synthesis results for, 66–74, 68t hydrogen permeation through, 125–127, 126f, 127f for hydrogen separation, 61, 62–63t hydrothermal stability of, 98 improvement of, 120–125, 121f, 122f, 123f, 124f limitations of, 61 polar molecule permeation properties of, 127–132 temperature dependence of, 127–128, 128f time courses for, 128–129, 129f synthesis of, 98 Approximations, 4f, 11, 12t Aristotle, 1 Asymmetric oxygen-conducting MIEC ceramic membranes, 253–255, 254f ATSB. See Aluminum tri-sec-butoxide A-type zeolites, 180 for alcohol de-watering, 184–186, 185f, 186f Azeotropic mixtures, membrane separation for, 178–179

B B2O3, CVD on Vycor glass, 38–39 Ba0.5Sr0.5Co0.8Fe0.2O3–d, 247, 258, 261 Ba0.5Sr0.5Fe0.8 Zn0.2O3–d, 239 Ba1–xSrxCo0.8Fe0.2O3–d, 239 BaCe0.1Co0.4 Fe0.5O3–d, 262, 262f BaCo0.4Fe0.4Zr0.2O3–d, 259 BaCo0.7Fe0.2Ta0.1O3–d, 259–262, 260f

Index

Bacon, Francis, 2 Bi2O3, in oxygen-conducting MIEC ceramic membranes, 240–241 BIMEVOX, in oxygen-conducting MIEC ceramic membranes, 240–241 Binding energy of hydrogen in amorphous metals, 318–319, 321–322, 321f of hydrogen in metal membranes, 312 Biodiesel, glycerol extraction for production of, 218 material and methods for, 220t, 221 results of, 225t, 226t, 228–229, 229f, 230f Bioethanol membrane extraction for production of, 214 removal of acetic acid for production of, 215–217, 215t materials and methods for, 219–221, 220t results of, 221–225, 222f, 223f, 225t, 226t Biofuel production, membrane-based solvent extraction for, 213–231 acetic acid removal from biomass hydrolysates, 215–217, 215t, 219–225, 220t, 222f, 223f, 225t, 226t glycerol extraction, 218, 220t, 221, 225t, 226t, 228–229, 229f, 230f HMF extraction, 217, 220t, 221, 225t, 226t, 227–228, 227f, 228f materials and methods for, 218–221, 219f, 220t results of, 221–230, 222f, 223f, 225t, 226t, 227f, 228f, 229f, 230f US energy supply, 213, 213f Biomass hydrolysates, acetic acid removal from, 215–217, 215t materials and methods for, 219–221, 220t results of, 221–225, 222f, 223f, 225t, 226t Bis(trimethylsilylmethyl)polybenzimidazole (PBI–TMS), VTEC blended with, 304 Boehmite (AlOOH) sols, preparation of, 65 Bond angles, of siloxane ring models, 83f, 84 Bond lengths, of siloxane ring models, 81f, 84 Boria-silica membrane, permeation performance of, 87–88, 88t Bovine serum albumin (BSA), ultrafiltration of, 340, 341f, 349–350, 349f Brownmillerite, 243 BSA. See Bovine serum albumin

Index BSCF membranes. See also Ba0.5Sr0.5Co0.8Fe0.2O3–d asymmetric, 254, 254f surface modifications to, 252 BTDA-DAPI (Matrimid) chemical structure and characteristics of, 143t in CMS formation, 142 gas separation, pyrolysis temperature and, 160–161, 160f oxygen exposure and, 150–152 FTIR of, 151, 151f gas separation performance and, 152, 152f structure of “molecular-ruler” for, 156t, 157–160, 158t, 159f sorption isotherms, 154–157, 155f, 156f, 156t, 157t TGA-FTIR for, 153–154, 153f, 154f 1-Butyl-3-methylimidazolium hexafluorophosphate, for CO2/CH4 separation, 107 1-Butyl-3-methylimidazolium tetrafluoroborate, for CO2/CH4 separation, 107

C C1 chemistry, de-watering for, 187–188, 187f, 189f Carbon barriers, in CVD on Vycor glass, 41–42 Carbon dioxide (CO2), activation energy of permeation for, in silica membranes, 86f, 87 Carbon dioxide (CO2) permeance BTDA-DAPI, 152, 152f of carbon molecular sieves, 94–95t, 96–97, 108f 6FDA/BPDA-DAM, 149, 149f of mixed-matrix membranes, 103, 104t, 108f of polymeric membranes, 100–101t, 102, 108f of SILMs, 103, 105–106t, 107, 108f of zeolite membranes, 93, 94–95t, 95, 108f Carbon dioxide (CO2) reforming, 256t Carbon dioxide (CO2) separation, 91–110 CVD of silica membrane on alumina for, 54–55 with HP polymers, 295, 303t, 304, 304t mixed-matrix membranes for, 102–103, 104t

355 polymeric membranes for, 98–99, 100–101t, 102, 108f silica membranes for, 97–98, 99t SILMs for, 103, 105–106t, 107 traditional methods for, 91 zeolite membranes and carbon molecular sieves for, 93–97, 94–95t Carbon dioxide/methane (CO2/CH4) selectivity BTDA-DAPI, 152, 152f of carbon molecular sieves, 94–95t, 96–97, 108f 6FDA/BPDA-DAM, 149, 149f of mixed-matrix membranes, 103, 104t, 108f of polymeric membranes, 100–101t, 102, 108f of SILMs, 103, 105–106t, 107, 108f of zeolite membranes, 93, 94–95t, 95, 108f Carbon molecular sieve (CMS) membranes, 92–93 characterization methods for, 145–146 permeation tests, 145–146 sorption tests, 146 for CO2/CH4 separation, 93–97, 94–95t concluding remarks on, 170 formation of, 137–138 polymer precursor films, 142, 143t, 144 pyrolysis for, 144–145, 145f gas separation properties of, 94–95t, 96–97 oxygen concentration and, 161–168 oxygen exposure and, 139, 141–142, 142f, 146–161 pyrolysis atmosphere and, 140–142 gas transport in, 138 limitations of, 97 oxygen doping process for, 169–170 performance of, 107, 108f, 109 structure of, 139–140, 139f, 140f Carbon monoxide (CO) activation energy of permeation for, in silica membranes, 86f, 87 H2 selectivity v., with sol-gel technique, 80 oxygen-conducting MIEC ceramic membrane selectivity for, 260–261 Carbonization, for CMS membrane formation, 153–154 oxygen doping process during, 169–170 Carrier gas flow rate, for CVD on HF membranes, 73, 74f CaTi0.8Fe0.2O3–d, 239 Causality, 2 Cause, 1–2 CE methods. See Cluster expansion methods

356 CeO2, in oxygen-conducting MIEC ceramic membranes, 241 Ceramic membranes. See Dual-phase membranes; Mixed ionic–electronic conducting ceramic membranes CH4. See Methane Charged ultrafiltration membranes, electrostatic interactions in, 341–347 fluid flow, 341–343, 343f solute transport, 344–347, 345f, 346f Chemical coprecipitation (CP) method, for preparing dual-phase membranes, 282–283, 285–286, 286f Chemical diffusion coefficient. See Transport diffusion coefficient Chemical expansivity, of oxygen-conducting MIEC ceramic membranes, 238–239 Chemical vapor deposition (CVD) alternating reactant deposition, 35 on Vycor glass, 41 components of, 26 control of, 29 counterdiffusion, 30–31, 34–35, 34f apparatus for, 34, 34f deposition rate in, 34–35 mechanism of, 31, 34, 34f deposition of coatings by, 26 on HF membranes g-alumina layers and, 68t, 72–73 concluding remarks on, 75–76 development of, 61 gas separation mechanism in, 74, 75f hydrogen permeation of, 61, 62–63t mesoporous silica layer of, 68t, 69–72, 70f, 71f, 72f preparation of, 65–66, 66f, 67f silica precursor and carrier gas flow rate, 73, 74f synthesis results for, 66–74, 68t mechanisms of, 27, 28f one-sided, 30–31, 34–35, 34f apparatus for, 34, 34f deposition rate in, 34–35 mechanism of, 30, 34f pore structure formation in, 28–29 principles of, 26–29, 28f reaction sequence of, 27–28 for silica membrane deposition, 25–26, 29–37 on alumina, 48–56, 49f, 53f hydrogen permeation mechanism, 36–37, 36f

Index

for hydrogen separation, 61, 62–63t objective of, 29–30 reactant diffusion geometry, 30–31, 34–35, 34f reaction conditions in, 35 silica precursors for, 30, 31t, 32–33t stability studies of, 35 supports for, 30 on Vycor glass, 26, 30, 37–48, 40f, 42f, 46f sol-gel technique v., 26 gas permeance, 98 types of, 26 Chemical vapor infiltration (CVI), for silica membrane deposition, on alumina with SiAc4, 51 Chemisorption, in hydrogen permeation, 36 Cluster expansion (CE) methods, for DFT calculations of hydrogen interaction with metal membranes, 313–314 CMS. See Carbon molecular sieve membranes CO. See Carbon monoxide Co. See Cobalt CO2. See Carbon dioxide Coatings, deposition of, by CVD, 26 Cobalt (Co)-doped silica membrane fabrication of, 119–120 gas permeance through, 122–123, 123f molecular size and, 132, 133f partial pressure of water on, 129–130, 130f temperature dependency of, 130, 131f time courses for, 128–129, 129f hydrogen permeance of, 123–124, 124f hydrothermal stability and, 118, 124–125 imaging of, 122, 122f Colligative properties, 6 Colloidal silica sols, preparation of, 65, 119–120 Combined reforming, of hydrocarbons, 257 Composite membranes, 92 Concentration polarization effects, in high performance ultrafiltration membranes, 347–350, 348f, 349f Concentration profiles, in counterdiffusion CVD, 39, 40f Conductivity, of dual-phase membranes, 275–306, 276f high-stability membranes and, 278–280, 280f other factors affecting, 289–290 phase ratio and, 288–289

Index

Conductivity, of dual-phase membranes (cont.) preparation methods for powders and, 282–287, 284f, 285f, 286f pure electronic conductors v. mixed conductors, 280–281 sintering temperature and, 287–288, 287f surface exchange and, 281–282, 282f Convective flow, in silica membranes, 36–37, 36f Corrected diffusivities, for modeling of amorphous metal membranes, 323–324, 325f Correlations, 13–14 in membrane separation, 16–22, 17f, 18f, 19f, 20f, 21f, 22f properties in membrane science, 14–16 scientific laws and, 3–14, 4f summary for, 21–22 Counterdiffusion CVD, 30–31, 34–35, 34f apparatus for, 34, 34f concentration profiles in, 39, 40f deposition rate in, 34–35 extended, 55 mechanism of, 31, 34, 34f Counter-electroosmosis, 342–343, 343f Covariance, 13 CP method. See Chemical coprecipitation method b-Cristobalite, structure of, 80–81, 81f Criteria, 4f, 11–12, 12t Crystalline metal membranes, DFT modeling of, 311–313 applications of, 314–315 Curves, 4f, 11–13, 12t CVD. See Chemical vapor deposition CVI. See Chemical vapor infiltration Cylindrical pores fluid flow through, 335–336, 336f pore size distribution effects in, 339 solute transport in, 336–338, 338f Cytochrome c, ultrafiltration of, 340, 341f, 345–346, 345f, 346f

D Data, 3 Deduction, 2 Defect structures, dual-phase membrane oxygen permeation and, 289–290 Dehydration. See De-watering Dehydrogenation, for CMS membrane formation, 153–154 oxygen doping process during, 169–170

357 Dense ceramic membranes. See Dual-phase membranes; Mixed ionic–electronic conducting ceramic membranes Dense metal membranes. See Metal membranes Densification permeability and, 38 selectivity and, 38 of silica membranes, 35 hydrogen permeation and, 118 overcoming, 38–39 water vapor and, 79 Density Functional Theory (DFT) calculations, 311 for modeling amorphous metal membranes, 316–327, 318f, 320f, 321f, 325f, 327f amorphous structures, 317–318, 318f binding energy of interstitial hydrogen, 318–319, 321–322, 321f computational approaches, 317 corrected diffusivities, 323–324, 325f H–H interactions, 319–320, 320f hydrogen diffusion, 317, 322–323 hydrogen permeability, 326–327, 327f hydrogen solubility, 317, 320–322, 321f thermodynamic correction factor, 324–325 for modeling crystalline metal membranes, 311–313 applications for, 314–315 for silica membrane structure, 80, 82 DES. See Diethylsilane De-watering of alcohol, 184–186, 185f, 186f for C1 chemistry, 187–188, 187f, 189f distillation-membrane hybrid separation system for, 178 of ethanol, 178, 186 of organic acids, 186–187 with zeolite membranes, 184–188 DFT calculations. See Density Functional Theory calculations Diethylsilane (DES), in CVD on Vycor glass, with N2O, 44 Differential scanning calorimetry (DSC), 144 Diffusion. See also specific types of diffusion activated, 118–119 of hydrogen, in amorphous metal membranes, 317, 322–323 in hydrogen permeation, 36–37, 36f with carbon barriers, 42 in mixed-matrix membranes, 102–103

358 Diffusion. See also specific types of diffusion (cont.) in Nanosil, 46 in polymeric membranes, 102 Diffusion coefficient for acetic acid extraction, 224–225, 226t in CMS membranes, 138 BTDA-DAPI, 157–159, 158t, 159f 6FDA/BPDA-DAM, 157–159, 158t, 159f for glycerol extraction, 226t for HMF extraction, 226t of hydrogen in amorphous metal membranes, 322–323 Diffusivity, of hydrogen in metal membranes, 313, 323–324, 325f Dimensionless numbers, 10–11, 10t Dimethoxydiphenylsilane (DMDPS), for CVD on alumina, with TMOS and PTMS, 55–56 Dimethylfuran (DMF), HMF extraction for production of, 217 Diphenyldiethoxysilane (DPDES), for CVD on alumina, 50 Dispersion-free membrane-based extraction. See Membrane-based solvent extraction Distillation processes energy consumption of, 176–178, 177f membranes for, 175–189 Distillation-membrane hybrid separation system, 176–178, 177f DMA. See Dynamic mechanical analysis DMDPS. See Dimethoxydiphenylsilane DMF. See Dimethylfuran DPDES. See Diphenyldiethoxysilane DSC. See Differential scanning calorimetry Dual mode mechanism, of polymeric membranes, 295 Dual-phase membranes composite materials in, 248–249, 250–251t oxygen permeation through, 275–306, 276f high-stability membranes and, 278–280, 280f other factors affecting, 289–290 phase ratio and, 288–289 preparation methods for powders and, 282–287, 284f, 285f, 286f pure electronic conductors v. mixed conductors, 280–281 sintering temperature and, 287–288, 287f surface exchange and, 281–282, 282f Dynamic mechanical analysis (DMA), of VTEC polyimides, 301, 302t, 303f

Index

E

EC process. See EDTA-citric acid process EDTA-citric acid (EC) process, for preparing dual-phase membranes, 282–283, 285–286, 286f Effects, 4f, 6–7, 7t Electronic conducting ceramic membranes. See Mixed ionic–electronic conducting ceramic membranes Electronic conducting phase, of dual-phase membranes with high stability and permeability, 278–280, 280f Electrostatic interactions, in high performance ultrafiltration membranes, 341–347, 343f, 345f, 346f fluid flow, 341–343, 343f solute transport, 344–347, 345f, 346f Elemental diffusion, dual-phase membrane oxygen permeation and, 289 Energy consumption, of distillation processes, 176–178, 177f Energy supply, US, 213, 213f Equations, 4f, 7, 10, 8t–9t Ethanol de-watering of, 178, 186 membrane extraction for production of, 214 removal of acetic acid for production of, 215–217, 215t materials and methods for, 219–221, 220t results of, 221–225, 222f, 223f, 225t, 226t steam reforming of, 26, 26f Evacuation in CVD on alumina, 48–50, 49f in CVD on HF membranes, 64 pore plugging with, 49, 49f Expansivity, of oxygen-conducting MIEC ceramic membranes, 238–239 Explicit function, 7 Extended counterdiffusion CVD method, on alumina, 55 Extensive properties, 6 Extraction. See Membrane-based solvent extraction

F Factors, 4f, 11–13, 12t 6FDA/BPDA-DAM chemical structure and characteristics of, 143t in CMS formation, 142–146

359

Index

6FDA/BPDA-DAM (cont.) gas separation, pyrolysis temperature and, 160–161, 160f oxygen exposure consumption and, 146–148, 147f, 148f gas separation performance and, 148–150, 149f, 150f structure of “molecular-ruler” for, 156t, 157–160, 158t, 159f sorption isotherms, 154–157, 155f, 156f, 156t, 157t TGA-FTIR for, 153–154, 153f, 154f synthesis of, 142, 144 FFV. See Fractional free volume Fickian coefficient. See Transport diffusion coefficient Fick’s law, 14–15, 322 Film deposition technology, for silica membrane production, 26 Filtrate flux, 333 First-principles models, for developing metal membranes for high temperature hydrogen purification, 309–327 applications of DFT calculations to crystalline membranes, 314–315 cluster expansion methods, 313–314 DFT-based modeling of amorphous metal membranes, 316–327, 318f, 320f, 321f, 325f, 327f DFT-based modeling of crystalline metal membranes, 311–313 Fischer-Tropsch (FT) syntheses, de-watering for, 187–188 Fluid flow, in high performance ultrafiltration membranes electrostatic effects on, 341–343, 343f pore geometry effects on, 335–336, 336f Fluorites in dual-phase membranes, 275, 276f in oxygen-conducting MIEC ceramic membranes, 240–241, 242t Fourier transform infrared (FTIR) spectroscopy, of BTDA-DAPI after oxygen exposure, 151, 151f Fractional free volume (FFV), permeability v., 20–21, 20f FT syntheses. See Fischer-Tropsch syntheses FTIR. See Fourier transform infrared spectroscopy

G Gas mixture for CVD, 26 membrane separation for, 178–179 Gas permeability. See also Activation energy of permeation; Hydrogen permeation alumina and, 48 of amorphous metal membranes, 326–327, 327f in CMS membranes BTDA-DAPI, 157–159, 158t 6FDA/BPDA-DAM, 157–159, 158t of CMS membranes, 138 densification and, 38 FFV v., 19–20, 20f of HF membranes, pure, 67, 69f of high performance ultrafiltration membranes, 333 electrostatic effects on, 341–343, 343f pore geometry effects on, 335 of HP polymers, 296–297 of metal membranes, 310–311 of Nanosil, 46 Gas permeance, 92 through Co-doped silica membranes, 122–123, 123f molecular size and, 132, 133f partial pressure of water on, 129–130, 130f temperature dependency of, 130, 131f time courses of, 128–129, 129f in CVD v. sol-gel technique, 98 of polymeric membranes, 102 through silica membranes, 120–121, 121f, 125 kinetic diameter and, 120–121, 121f molecular size and, 132, 133f types of, 36–37, 36f, 118–119 Gas phase, of CVD, 27–29, 28f Gas separation BTDA-DAPI and inert flow rate and, 166–167, 166f, 166t, 167f oxygen exposure, 152, 152f precursor film thickness, 167–168, 168f, 168t, 169f thermal soak time and, 162–166, 163f, 163t, 164f, 165f, 165t CMS membranes for, 94–95t, 96–97 oxygen concentration and, 161–168 oxygen exposure and, 139, 141–142, 142f, 146–161

360 Gas separation (cont.) pyrolysis atmosphere and, 140–142 6FDA/BPDA-DAM inert flow rate and, 166–167, 166f, 166t, 167f oxygen exposure and, 148–150, 149f, 150f precursor film thickness, 167–168, 168f, 168t, 169f thermal soak time and, 162–166, 163f, 163t, 164f, 165f, 165t with HF membranes, 74, 75f HP high temperature, 295–306 experimental setup, 297–298 instrumentation for, 297 permeability gas testing, 297–298 positron annihilation lifetime spectroscopy of, 298 results and discussion, 298–305, 302t, 303f, 303t, 304f, 305t membrane technology for, 91–92, 137 silica membranes for, 120 in zeolite membranes, 93 Gas transport in CMS membranes, 138 in glassy polymers, 137 Gasification processes, metal membranes for, 309–310 Gas-to-liquid (GTL) processes, 256 GCMC simulations. See Grand Canonical Monte Carlo simulations Glassy polymers, 98 gas sorption in, 98 gas transport properties of, 137 Glassy polymer membranes. See High performance polymers Glycerol, membrane-based extraction of, 218 materials and methods for, 220t, 221 results of, 225t, 226t, 228–229, 229f, 230f Glycine-nitrate combustion process (GNP), for preparing dual-phase membranes, 282–283 GNP. See Glycine-nitrate combustion process Grain boundaries of dual-phase membranes, 286–288 of oxygen-conducting MIEC ceramic membranes, 239–240 Grain size, dual-phase membrane oxygen permeation and, 289 Grand Canonical Monte Carlo (GCMC) simulations, of hydrogen solubility in

Index

amorphous metal membranes, 320–322, 321f GTL processes. See Gas-to-liquid processes

H

H2. See Hydrogen H2O. See Water Hagen-Poiseuille equation, 335 Hagen-Pouisselle mechanism, 15 He. See Helium Helium (He), activation energy of permeation of of hydrogen v., 126–127, 127f in silica membranes, 86f, 87 Helium (He) permeation, through amorphous silica membranes, 125–127, 126f, 127f Helium/hydrogen (He/H2) permeance ratio activation energy of hydrogen permeation and, 125, 126f pore size distribution and, 125–126 Herschel, John, 2 1-Hexyl-3-methylpyridinium bis (trifluoromethylsulfonyl)imide, for CO2/ CH4 separation, 107 HF membranes. See Hollow fiber membranes H–H interactions, in amorphous metal membranes, 319–320, 320f High performance (HP) polymers high temperature gas separations using, 295–306 experimental setup, 297–298 instrumentation for, 297 permeability gas testing, 297–298 positron annihilation lifetime spectroscopy of, 298 results and discussion, 298–305, 302t, 303f, 303t, 304f, 305t moisture within, 304–305, 305t structures of, 296f High performance ultrafiltration membranes, 333–351, 334f concentration polarization effects on, 347–350, 348f, 349f electrostatic interactions in, 341–347, 343f, 345f, 346f fluid flow, 341–343, 343f solute transport, 344–347, 345f, 346f pore geometry effects on, 335–341, 336f, 338f, 341f fluid flow, 335–336, 336f pore size distribution effects, 339–341, 341f solute transport, 336–339, 338f

Index

High temperature gas separations, with HP polymers, 295–306 experimental setup, 297–298 instrumentation for, 297 permeability gas testing, 297–298 positron annihilation lifetime spectroscopy of, 298 results and discussion, 298–305, 302t, 303f, 303t, 304f, 305t High temperature hydrogen purification, developing metal membranes for, 309–327 applications of DFT calculations to crystalline membranes, 314–315 cluster expansion methods, 313–314 DFT-based modeling of amorphous metal membranes, 316–327, 318f, 320f, 321f, 325f, 327f DFT-based modeling of crystalline metal membranes, 311–313 HMF. See 5-Hydroxymethylfurfural Hollow fiber (HF) membranes g-alumina layers in, 65–66, 68t, 72–73 concluding remarks on, 75–76 development of, 61 gas separation mechanism in, 74, 75f hydrogen permeation of, 61, 62–63t for membrane-based solvent extraction, 214 mesoporous silica layer of, 68t, 69–72, 70f, 71f, 72f H2 permeance and, 69–72, 70f, 72f nomenclature for, 68t, 71, 72f pore defect correction in, 70–71, 70f oxygen-conducting MIEC ceramic membranes as, 254–255 preparation of, 65–66, 66f, 67f properties of pure, 67, 69f silica precursor and carrier gas flow rate, 73, 74f synthesis results for, 66–74, 68t Homogeneity, of dual-phase membranes, 282–288, 284f, 285f, 286f HP polymers. See High performance polymers Hydraulic permeability, of high performance ultrafiltration membranes, 333, 335, 341–343, 343f Hydrocarbon reforming carbon dioxide reforming, 256t, 257 oxygen-conducting MIEC ceramic membranes in, 236, 255–263, 256f, 256t, 259t, 260f, 262f benefits of, 257–258

361 overview of, 255–257, 256t reaction environment effects on membrane performance, 261–263, 262f reaction temperature effects on membrane performance, 260–261, 260f work to date on, 258–260, 259t steam reforming, 256t, 257 Hydrogen (H2) activation energy of permeation of He/H2 permeance ratio v., 125–126, 126f of helium v., 126–127, 127f in silica membranes, 86f, 87 CO selectivity v., with sol-gel technique, 80 HP polymer separation of, 295, 303t, 304, 304t within metal membranes, 310–313 methane production of, 54 production of, 309 uses of, 25 Hydrogen (H2) diffusion, in amorphous metal membranes, 317, 322–323 Hydrogen (H2) flux, in metal membranes, 310, 326 Hydrogen (H2) permeation alumina and CO2 separation and, 54–55 of CVI with SiAc4, 51 with evacuation, 49–50 in H2-H2O-HI mixture, 51–52 with Nanosil processes, 52 with new hybrid processing method, 54 with steam pretreatment, 54 with TEOS and TMOS, 53 of alumina-silica membranes, 80, 87–88, 88t of amorphous metal membranes, 326–327, 327f of amorphous silica membranes, 125–127, 126f, 127f of boria-silica membranes, 87–88, 88t on HF supports, 61, 62–63t g-alumina layers and, 68t, 72–73 evacuation and, 64 gas separation mechanism, 74, 75f mesoporous silica layer and, 69–72, 70f, 72f silica precursor and carrier gas flow rate, 73, 74f under hydrothermal conditions, 118 mechanism of, 36–37

362 Hydrogen (H2) permeation (cont.) diffusion, 36–37, 36f sorption, 36 of metal membranes, 310–311 of metal-doped silica membranes, 123–124, 124f oxidant reagents and, 64 of silica membranes, 87–88, 88t of titania-silica membranes, 80, 87–88, 88t Vycor glass and, 30 with alternating reactant deposition, 41 with carbon barriers, 41–42 with DES and N2O, 44 Nanosil, 47 with SiCl4, 39–41 with SiH4, 38 with TEOS, 37–38 with TEOS and ATSB, 43 with TEOS and TIPT, 43–44 with TPS, 42 of zirconia-silica membranes, 79–80, 87–88, 88t Hydrogen purification. See High temperature hydrogen purification Hydrogen (H2) separation applications of, 61 current methods for, 61 palladium for, 25 polymeric membranes for, 25 silica membranes for, 25–27, 29, 61, 62–63t Co-doped, 118 Hydrogen (H2) solubility, in metal membranes, 311–312 amorphous membranes, 317, 320–322, 321f Hydrolysates, acetic acid removal from, 215–217, 215t materials and methods for, 219–221, 220t results of, 221–225, 222f, 223f, 225t, 226t Hydrophilicity, organic acids de-watering and, 186–187 Hydrothermal stability, of silica membranes amorphous, 98 Co-doped, 118, 124–125 composite membrane formation for, 79 control of, 117, 121–122 in humidified air, 118 improvement of, 120–125, 121f, 122f, 123f, 124f measurements of, 120 5-Hydroxymethylfurfural (HMF), membrane-based extraction of, 217

Index

materials and methods for, 220t, 221 results of, 225t, 226t, 227–228, 227f, 228f

I Identity relation, 5–6 Implicit function, 11 Induction, 2 Inert flow rate, in CMS membrane gas separation and oxygen concentration, 166–167, 166f, 166t, 167f Inorganic membranes, 92 for CO2/CH4 separation, 93–97, 94–95t development of, 179 for hydrogen separation, 25, 61, 64 limitations of, 61 types of, 179–180 Intensive properties, 6 Interaction energies of hydrogen in amorphous metal membranes, 319–320, 320f of siloxane ring models, 85–86 Intergrowth phase, in SFC oxygen-conducting MIEC ceramic membranes, 248 Intergrowth supporting substances (ISS), for zeolite synthesis, 184 Interstitial hydrogen in amorphous metals, 318–319 in metal membranes, 312–313 Ionic conducting ceramic membranes. See Mixed ionic–electronic conducting ceramic membranes Ionic conducting phase, of dual-phase membranes with high stability and permeability, 278–280, 280f Ionic liquids. See Room temperature ionic liquids Island formation, in CVD, 28 ISS. See Intergrowth supporting substances

J Jump frequency, of silica membranes, 81

K Kinetic diameter determination of, 119, 132 gas permeance of silica membranes and, 120–121, 121f in molecular sieving, 119 Kinetic Monte Carlo (KMC) simulations of hydrogen in amorphous metal membranes, 323–324, 325f of hydrogen in metal membranes, 313

Index KMC simulations. See Kinetic Monte Carlo simulations Knudsen diffusion membrane permeance in, 67 permselectivity and, 48 in silica membranes, 36–37, 36f on Vycor glass, 48 temperature and, 38 Knudsen flow, 118 Kozeny–Carman equation, 340

L La0.2Ba0.8Fe0.8Co0.2O3–d, 258 La0.5Sr0.5FeO3–d, 239 La0.6Sr0.4Co0.2Fe0.8O3–d (LSCF membrane), 246, 255 La0.6Sr0.4Co0.8Fe0.2O3–d, 258 La0.7Sr0.3Ga0.6Fe0.4O3–d, 247 La0.8Sr0.2Co0.1Fe0.8Cr0.1O3–d, 259 LaCoO3–d membranes, 239 Langmuir affinity constants of BTDA-DAPI, 155–157, 156t, 157t of 6FDA/BPDA-DAM, 155–157, 156t, 157t Laws, 4–6, 4f, 5t Layer-by-layer growth, in CVD, 28 Leave one out (LOO) method, for cluster expansion, 314 Lennard-Jones collision diameter, in molecular sieving, 119 Lennard-Jones potential, for kinetic diameter, 119 Lignocellulosic biomass, acetic acid removal from, 215–217, 215t materials and methods for, 219–221, 220t results of, 221–225, 222f, 223f, 225t, 226t Liquid alkanes, HMF extraction for production of, 217 LOO method. See Leave one out method LSCF membrane. See La0.6Sr0.4Co0.2Fe0.8O3–d

M M-1A-3S-1 silica precursor and carrier gas flow rate for, 73, 74f synthesis details for, 68t M-1A-3S-2 g-alumina layers in, 68t, 72–73 silica precursor and carrier gas flow rate for, 73, 74f synthesis details for, 68t

363 M-2A-2S-1, synthesis details for, 68t M-2A-2S-2 mesoporous silica layers in, 71–72, 72f synthesis details for, 68t M-2A-2S-3 mesoporous silica layers in, 71–72, 72f synthesis details for, 68t M-2A-3S-1, synthesis details for, 68t M-2A-3S-2 gas permeance and selectivity of, 75f synthesis details for, 68t M-2A-3S-3 g-alumina layers in, 68t, 72–73 silica precursor and carrier gas flow rate for, 73, 74f synthesis details for, 68t M-2A-3S-4 gas permeance and selectivity of, 74f, 75f silica precursor and carrier gas flow rate for, 73, 74f synthesis details for, 68t Macroporous, 179 Magnesia, silica membranes with, 79 Masking technique, for zeolite membrane synthesis, 182–183 Mass transfer coefficient, 214 for acetic acid removal from biomass hydrolysates, 223–225, 223f, 225t for glycerol extraction, 225t, 228–229, 230f for HMF extraction, 225t, 227–228, 228f Material screening, for developing metal membranes for high temperature hydrogen purification, 309–327 applications of DFT calculations to crystalline membranes, 314–315 cluster expansion methods, 313–314 DFT-based modeling of amorphous metal membranes, 316–327, 318f, 320f, 321f, 325f, 327f DFT-based modeling of crystalline metal membranes, 311–313 Matrimid. See BTDA-DAPI Membrane contactors, for membrane-based solvent extraction, 218–219 Membrane gas separation, 91–92 advantages of, 91 limitations of, 91 performance of, 92 types of, 92 Membrane matrix vibration, in activated diffusion, 118–119

364 Membrane reactors for FT syntheses, 188 for methanol, 188 silica membranes in, 26, 61 Membrane science, properties in, 14–16 Membrane separation correlation samples in, 16–21, 17f, 18f, 19f, 20f, 21f, 22f with distillation, 176–178, 177f history of, 179 other applications of, 178–179 Membrane-based solvent extraction, for biofuel production, 213–231 acetic acid removal from biomass hydrolysates, 215–217, 215t, 219–225, 220t, 222f, 223f, 225t, 226t glycerol extraction, 218, 220t, 221, 225t, 226t, 228–229, 229f, 230f HMF extraction, 217, 220t, 221, 225t, 226t, 227–228, 227f, 228f materials and methods for, 218–221, 219f, 220t results of, 221–230, 222f, 223f, 225t, 226t, 227f, 228f, 229f, 230f US energy supply, 213, 213f Mesoporous, 179 Mesoporous silica layer in HF membranes, 65–66, 68t, 69–72, 70f, 71f, 72f H2 permeance and, 69–72, 70f, 72f nomenclature for, 68t, 71, 72f pore defect correction in, 70–71, 70f in zeolite membrane synthesis, 183 Metal alloys, for high temperature hydrogen purification, 310 Metal membranes, first-principles models for development of, 309–327 applications of DFT calculations to crystalline membranes, 314–315 cluster expansion methods, 313–314 DFT-based modeling of amorphous metal membranes, 316–327, 318f, 320f, 321f, 325f, 327f DFT-based modeling of crystalline metal membranes, 311–313 Metal-doped silica membranes, hydrogen permeance of, 123–124, 124f Methane (CH4). See also Partial oxidation of methane activation energy of permeation for, in silica membranes, 86f, 87 hydrogen production with, 54

Index

Methane (CH4) separation, 91–110 mixed-matrix membranes for, 102–103, 104t polymeric membranes for, 98–99, 100–101t, 102, 108f silica membranes for, 97–98, 99t SILMs for, 103, 105–106t, 107 zeolite membranes and carbon molecular sieves for, 93–97, 94–95t Methanol vapor separation, CVD on Vycor glass for, 44–45 Methyl groups, for silica membrane stability, 35 MFI zeolite membranes, 180 bilayer membranes, characteristics of, 201–202, 201t, 202f, 230f characterization of, 198–200 synthesis of, 184, 198–200 xylene separation with, 195–210, 197f, 199f bilayer structure reversal, 207–209, 208f, 209t higher PX feed concentrations, 209–210, 210t pervaporation experiments, 200–201 through single and bilayer membranes, 202–206, 203f, 205f, 206f, 207f Microporous, 179 Microstructure, of oxygen-conducting MIEC ceramic membranes, 239–240 MIEC ceramic membranes. See Mixed ionic–electronic conducting ceramic membranes Mixed conductors, in dual-phase membranes, 279–281 Mixed ionic–electronic conducting (MIEC) ceramic membranes, 235–264 common materials used in, 240–255, 242t, 244–245t, 250–251t, 253f, 254f dual-phase composite materials, 248–249, 250–251t fluorites, 240–241, 242t membrane modifications improving oxygen flux, 249–252 membrane thickness reduction, 253–255, 254f perovskites, 241–248, 244–245t SCF perovskite materials, 243–248 surface modifications, 252, 253f general attributes of, 236–240, 237f chemical expansivity, 238–239 microstructure, 239–240 oxygen nonstoichiometry, 236–237, 237f

365

Index

Mixed ionic–electronic conducting (MIEC) ceramic membranes (cont.) self-adjusting phase equilibria, 237–238 for syngas production, 255–263, 256f, 256t, 259t, 260f, 262f benefits of, 257–258 overview of, 255–257, 256t reaction environment effects on membrane performance, 261–263, 262f reaction temperature effects on membrane performance, 260–261, 260f work to date on, 258–260, 259t Mixed-matrix membranes, 92 for CO2/CH4 separation, 102–103, 104t performance of, 107, 108f, 109 Mixtures with close boiling points, membrane separation for, 178–179 Moisture within HP polymers, 304–305, 305t silica membrane permeability and, 35 Molecular sieving in activated diffusion, 119 in silica membranes, 36–37, 36f on Vycor glass, 48 in zeolite membranes, 93 Molecular size, gas permeance through silica membranes and, 132, 133f Molecular weight, Knudsen diffusion and, 67 MOR. See Mordenite Mordenite (MOR), 180 Multiphase SFC, in oxygen-conducting MIEC ceramic membranes, 247–248 MX. See m-Xylene

N N2O, in CVD on Vycor glass, with DES, 44 Nanosil, 45–47, 46f Natural gas production, CO2/CH4 separation for, 91 Navier–Stokes equation, 342 Ne. See Neon NEB method. See Nudged Elastic Band method Neon (Ne), activation energy of permeation for, in silica membranes, 86f, 87 “New hybrid processing method,” for CVD on alumina, 54 Newton, Isaac, 2 NH3. See Ammonia Ni. See Nickel Nickel (Ni)-doped silica membrane, 35

hydrogen permeance of, 123–124, 124f Nonporous membranes, 179 N-substituted PBIs, 297, 299–300 Nudged Elastic Band (NEB) method, 312

O

O2. See Oxygen O3. See Ozone Occam’s razor, 2 OLC. See Operability level coefficient One-sided CVD, 30–31, 34–35, 34f apparatus for, 34, 34f deposition rate in, 34–35 mechanism of, 30, 34f stability studies of, 39–41 Operability level coefficient (OLC), 20–21, 21f Opposing reactants geometry. See Counterdiffusion CVD Organic acids, de-watering of, 186–187 Organic-inorganic hybrid alkoxides, pore sizes and, 117–118 Organic-inorganic membranes, 92 for CO2/CH4 separation, 102–103, 104t Overdeposition, with chlorosilanes in CVD, 41 OX. See o-Xylene Oxidant reagent, H2 permeance and, 64 Oxygen (O2) within dual-phase membranes, 275, 276f within oxygen-conducting MIEC ceramic membranes, 236–237, 237f Oxygen (O2) concentration, CMS membrane gas separation and, 161–168 inert flow rate and, 166–167, 166f, 166t, 167f oxygen-carbon reaction mechanism, 168 precursor film thickness, 167–168, 168f, 168t, 169f review, 161–162 thermal soak time and, 162–166, 163f, 163t, 164f, 165f, 165t Oxygen (O2)-conducting MIEC ceramic membranes, 235–264 common materials used in, 240–255, 242t, 244–245t, 250–251t, 253f, 254f dual-phase composite materials, 248–249, 250–251t fluorites, 240–241, 242t membrane modifications improving oxygen flux, 249–252 membrane thickness reduction, 253–255, 254f perovskites, 241–248, 244–245t

366 Oxygen (O2)-conducting MIEC ceramic membranes (cont.) SCF perovskite materials, 243–248 surface modifications, 252, 253f general attributes of, 236–240, 237f chemical expansivity, 238–239 microstructure, 239–240 oxygen nonstoichiometry, 236–237, 237f self-adjusting phase equilibria, 237–238 for syngas production, 255–263, 256f, 256t, 259t, 260f, 262f benefits of, 257–258 overview of, 255–257, 256t reaction environment effects on membrane performance, 261–263, 262f reaction temperature effects on membrane performance, 260–261, 260f work to date on, 258–260, 259t Oxygen (O2) doping process, for CMS membranes, 169–170 Oxygen (O2) exposure BTDA-DAPI and, 150–152 FTIR of, 151, 151f gas separation performance and, 152, 152f CMS membrane gas separation and, 139, 141–142, 142f 6FDA/BPDA-DAM and, 146–150 consumption and, 146–148, 147f, 148f gas separation performance and, 148–150, 149f, 150f Oxygen (O2) flux of dual-phase composite membranes, 250–251t of fluorite membranes, 242t membrane modifications improving, 249–252 of perovskite membranes, 244–245t under POM conditions, 259t, 260–263, 260f, 262f sintering conditions effect on, 239–240 Oxygen (O2) nonstoichiometry, of oxygenconducting MIEC ceramic membranes, 236–237, 237f Oxygen (O2)-permeable membranes. See Dual-phase membranes; Mixed ionic–electronic conducting ceramic membranes

Index

Oxygen (O2) permeation, through dual-phase membranes, 275–306, 276f high-stability membranes and, 278–280, 280f other factors affecting, 289–290 phase ratio and, 288–289 preparation methods for powders and, 282–287, 284f, 285f, 286f pure electronic conductors v. mixed conductors, 280–281 sintering temperature and, 287–288, 287f surface exchange and, 281–282, 282f Ozone (O3), for CVD on Vycor glass, with TEOS, 45

P

PAL. See Positron annihilation lifetime spectroscopy Palladium (Pd)-based membranes alloys, 310, 315 for high temperature hydrogen purification, 309–310, 315 history of, 179 for hydrogen separation, 25 PdAg alloys, 315 PdAu alloys, 315 PdCu-based ternary alloys, 315 Pd-rich binary alloys, 315 Partial oxidation of methane (POM), oxygenconducting MIEC ceramic membranes in, 236, 255–263, 256f, 256t, 259t, 260f, 262f benefits of, 257–258 overview of, 255–257, 256t reaction environment effects on membrane performance, 261–263, 262f reaction temperature effects on membrane performance, 260–261, 260f work to date on, 258–260, 259t Partition coefficient, of high performance ultrafiltration membranes, 336–337, 344 PBI. See Polybenzimidazole PBI-TMS. See Bis(trimethylsilylmethyl) polybenzimidazole Pd-based membranes. See Palladium-based membranes Pearson correlation coefficient, 13 Permeability. See also Activation energy of permeation; Hydrogen permeation alumina and, 48 of amorphous metal membranes, 326–327, 327f

Index Permeability. See also Activation energy of permeation; Hydrogen permeation (cont.) of CMS membranes, 138 BTDA-DAPI, 157–159, 158t 6FDA/BPDA-DAM, 157–159, 158t densification and, 38 FFV v., 19–20, 20f of HF membranes, pure, 67, 69f of high performance ultrafiltration membranes, 333 electrostatic effects on, 341–343, 343f pore geometry effects on, 335 of HP polymers, 296–297 of metal membranes, 310–311 of Nanosil, 46 Permeability coefficient, of gases, 17–18 Permeability gas testing, of HP polymers, 297–298 N-substituted PBIs, 299 VTEC polyimides, 301–304, 303t, 304t Permeability–selectivity tradeoff, of high performance ultrafiltration membranes, 333–335, 334f concentration polarization effects on, 347–350, 348f, 349f electrostatic interactions affecting, 344–347, 345f, 346f pore geometry effects on, 337–338, 338f pore size distribution effects on, 340, 341f Permeance, selectivity v., 18–19, 18f, 19f Permeance order, of silica membranes, 80–81, 81f Permeation. See Hydrogen permeation; Oxygen permeation Permeation performances, of silica membranes, 87–88, 88t Permeation rates, of silica membranes, 25–26, 35 Permeation tests, for CMS membrane characterization, 145–146 Permselectivity Knudsen diffusion and, 48 Vycor glass and, 30 with alternating reactant deposition, 41 with carbon barriers, 41–42 with DES and N2O, 44 Nanosil, 47 with SiCl4, 39–41 with SiH4, 38 with TEOS, 37–38 with TEOS and ATSB, 43

367 with TEOS and TIPT, 43–44 with TPS, 42 Perovskites in dual-phase membranes, 275, 276f membrane stability and, 276–280, 280f in oxygen-conducting MIEC ceramic membranes, 241–243, 244–245t SCF-based materials, 243–248 in POM reactions, 258 Pervaporation distillation processes v., 176, 177f xylene separation, 195, 200–201 bilayer structure reversal, 207–209, 208f, 209t binary feeds for, 200 higher PX feed concentrations, 209–210, 210t sample processing, 200–201 selectivity with, 196, 198 through single and bilayer membranes, 202–206, 203f, 205f, 206f, 207f stability with, 196, 198 Phase assemblages, of oxygen-conducting MIEC ceramic membranes, 238, 247–248 Phase equilibria, of oxygen-conducting MIEC ceramic membranes, 237–238 Phase ratio, dual-phase membrane oxygen permeation and, 288–289 Phenyl-substituted ethoxysilanes, for CVD on alumina, 50 Phenyltriethoxysilane (PTES), for CVD on alumina, 50 Phenyltrimethoxysilane (PTMS) for CVD on alumina, with TMOS and DMDPS, 55–56 for CVD on HF membranes, 64 Phosphazenes, 296, 296f Physisorption, in hydrogen permeation, 36 Platinum film, on oxygen-conducting MIEC ceramic membranes, 252, 253f Poison resistance, of silica membranes, 25–26 Polar molecule permeation, through amorphous silica membranes, 127–132 temperature dependence of, 127–128, 128f time courses for, 128–129, 129f Polyamides, 296, 296f Polyazoles, 296, 296f Polybenzazoles, 296, 296f Polybenzimidazole (PBI), 296–299, 296f N-substituted, 297, 299–300 PBI–TMS and VTEC blend, 304

368 Polyimides, 296, 296f. See also BTDA-DAPI; 6FDA/BPDA-DAM VTEC, 297, 300–301 gas permeation of, 301–304, 303t, 304t moisture within, 304–305, 305t PAL spectroscopy of, 305, 305t PBI–TMS and VTEC blend, 304 thermal analysis of, 301–302, 302t Polymer precursor films, for CMS membrane formation, 142, 143t, 144 thickness of, 167–168, 168f, 168t, 169f Polymer synthetic substitution, 297, 299–300 Polymeric membranes, 92. See also High performance polymers for CO2/CH4 separation, 98–99, 100–101t, 102, 108f performance of, 102, 108f development of, 179 dual mode mechanism of, 295 for hydrogen separation, 25 performance of, 102, 107, 108f, 109 prevalence of, 137 types of, 98 Polysulfones, 296, 296f POM. See Partial oxidation of methane Pore geometry effects, in high performance ultrafiltration membranes, 335–341, 336f, 338f, 341f fluid flow, 335–336, 336f pore size distribution effects, 339–341, 341f solute transport, 336–339, 338f Pore plugging in counterdiffusion CVD, 34–35 with evacuation, 49, 49f in HF membrane CVD, 70–71, 70f in one-sided CVD, 34–35, 39 Pore size of alumina, 48 BTDA-DAPI, 157 pyrolysis temperature and, 160–161, 160f in carbon molecular sieves, 93 6FDA/BPDA-DAM, 157 pyrolysis temperature and, 160–161, 160f gas molecule size v., 37 He/H2 permeance ratio and, 125–126 Knudsen diffusion and, 67, 69f in Nanosil, 47 organic-inorganic hybrid alkoxides and, 117–118

Index

in silica membranes, control of, 117, 121–122 silica precursors and, 50 sol-gel technique and, 117 of Vycor glass, 48 of zeolite membranes, 180, 180f Pore size distribution effects, in high performance ultrafiltration membranes, 339–341, 341f Pore structure CVD formation of, 28–29 with sol-gel technique, 98 Porous materials, classification of, 179 Porous membranes, 179–180 Positron annihilation lifetime (PAL) spectroscopy of HP polymers, 298 of VTEC polymers, 305, 305t Precarbonation, for CMS membrane formation, 153–154 oxygen doping process during, 169–170 Principles, 3, 4f Promoted SCF materials, 246 Properties, 4f, 6, 6t in membrane science, 14–16 Protein purification, with high performance ultrafiltration membranes, 333–351 concentration polarization effects on, 347–350, 348f, 349f electrostatic interactions in, 341–347, 343f, 345f, 346f pore geometry effects on, 335–341, 336f, 338f, 341f selectivity–permeability tradeoff in, 333–335, 334f PTES. See Phenyltriethoxysilane PTMS. See Phenyltrimethoxysilane Pure electronic conductors, in dual-phase membranes, 279–281 Pure HF membranes, 67, 69f PX. See p-Xylene Pyrolysis atmosphere, for CMS formation, 145 and gas separation, 140–142, 160–161, 160f Pyrolysis method, for CMS membrane formation, 137–138 oxygen doping process for, 169–170 stages of, 153–154

R Radial distribution function (RDF), for describing amorphous structures, 317–318, 318f

Index RDF. See Radial distribution function Reactant diffusion geometry, for CVD of silica membranes, 30–31, 34–35, 34f Reaction rate coefficient, of CVD, with carbon barriers, 42 Reaction temperature, oxygen-conducting MIEC ceramic membrane performance and, 260–261, 260f Renewable energy, 213–215 Representations, 2–3 Reversible absorption, for CO2 separation, 91 Robeson limit, 17f, 44 Robeson’s selectivity/permeability correlation plot, 296–297 Room temperature ionic liquids (RTILs), 92, 103 performance of, 107, 108f, 109 RTILs. See Room temperature ionic liquids Rubbery polymers, 98

S

SCF perovskite materials. See Sr–Co–Fe perovskite materials Scientific laws, correlations and, 3–14, 4f Scientific method, 4 SDAs. See Structure directing agents Seeding technique, for zeolite membrane synthesis, 181–182 MFI-type, 199–200 Selectivity activation energy of permeation and, 126 alumina and, 48 of CMS membranes, 138 densification and, 38 of high performance ultrafiltration membranes, 333 membranes and, 92 permeance v., 18–19, 18f, 19f with pervaporation separation of xylene, 196, 198 reversal of, in xylene separation, 204 of silica membranes, 25–26, 37 Selectivity–permeability tradeoff, of high performance ultrafiltration membranes, 333–335, 334f concentration polarization effects on, 347–350, 348f, 349f electrostatic interactions affecting, 344–347, 345f, 346f pore geometry effects on, 337–338, 338f pore size distribution effects on, 340, 341f

369 Self-adjusting phase equilibria, of oxygenconducting MIEC ceramic membranes, 237–238 SFC. See SrFeCo0.5O3–d SH4. See Silane Si-1 membrane, temperature dependence of, 127–128, 128f Si-3 membrane, temperature dependence of, 127–128, 128f SiAc4, in CVI on alumina, 51 SiCl4, in CVD on Vycor glass, 38–39 stability studies of, 39–41 TEOS v., 43 Sievert’s law, 15, 311–312, 326 Sieving coefficient, of high performance ultrafiltration membranes, 336, 339–340, 347 Silane (SH4), in CVD on Vycor glass, 38 Silica membranes, 92. See also Amorphous silica membranes activation energy of permeation of, 81–82 calculation of, 82 on alumina, 48–56, 49f, 53f CO2 separation and, 54–55 with CVI and SiAc4, 51 with evacuation, 48–50, 49f extended counterdiffusion, 55 g-alumina and, 52–53, 53f in H2-H2O-HI mixture, 51–52 with Nanosil processes, 52 with new hybrid processing method, 54 with PTES and DPDES, 50 with steam pretreatment, 54 with TEOS, 48–51, 49f with TEOS and TMOS, 53–54 with TMOS, PTMS, and DMDPS, 55–56 Vycor glass v., 48 ammonia and, 131–132 application of, 25–26 for CO2/CH4 separation, 97–98, 99t Co-doped fabrication of, 119–120 gas permeance through, 122–123, 123f, 128–130, 129f, 130f, 131f hydrogen permeance of, 123–124, 124f hydrothermal stability and, 124–125 imaging of, 122, 122f CVD of, 25–26, 29–37 hydrogen permeation mechanism, 36–37, 36f objective of, 29–30

370 Silica membranes (cont.) reactant diffusion geometry, 30–31, 34–35, 34f reaction conditions in, 35 silica precursors for, 30, 31t, 32–33t stability studies of, 35 supports for, 30 diffusion in, 36–37, 36f gas permeation of kinetic diameter and, 120–121, 121f measurement of, 120 molecular size and, 132, 133f gas separation with, measurement of, 120 for hydrogen separation, 25–27, 29 hydrothermal stability of composite membrane formation for, 79 control of, 117, 121–122 in humidified air, 118 metal-doped, hydrogen permeance of, 123–124, 124f Ni-doped, 35 hydrogen permeance of, 123–124, 124f permeation performance of, 87–88, 88t pore size control in, 117, 121–122 previous theoretical studies on, 80–82, 81f production of, 26 siloxane ring as model of, 82 activation energy of permeation of, 84–87, 85f, 86f bond angles, 83f, 84 critical openings of, 87 energies and bond lengths, 81f, 84 interaction energies of, 85–86 permeation performances of, 87–88, 88t structure of, 83–84, 83f sol-gel method for, 119–120 solubility sites of, 80–81, 81f structure of, 80 types of, 97–98 on Vycor glass, 26, 30, 37–48, 40f, 42f, 46f alternating reactant deposition, 41 alumina substrate v., 48 binary systems, 43–44 with carbon barriers, 41–42 densification in, 38–39 with DES and N2O, 44 for methanol vapor separation, 44–45 Nanosil, 45–47, 46f with SH4, 38 with SiCl4, 38–39 with TEOS, 37–38, 43 with TEOS/O3, 45

Index

with TPS, 42–43, 42f water vapor and, 79, 131–132 Silica precursors for CVD of silica membranes, 30, 31t, 32–33t for CVD on HF membranes, 64, 73, 74f pore size and, 50 Silicalite membranes, 180 characteristics of, 201–202, 201t, 202f, 230f synthesis of, 199–200 xylene separation with, 198 bilayer structure reversal, 207–209, 208f, 209t higher PX feed concentrations, 209–210, 210t through single and bilayer membranes, 202–206, 203f, 205f, 206f, 207f Silicon nanopore membranes, 338, 338f SILMs. See Supported ionic liquid membranes Siloxane ring activation energy of permeation of, 84–85, 85f as silica membrane model, 82 activation energy of permeation of, 84–87, 85f, 86f bond angles, 83f, 84 critical openings of, 87 energies and bond lengths, 81f, 84 interaction energies of, 85–86 permeation performances of, 87–88, 88t structure of, 83–84, 83f Silver coating, on oxygen-conducting MIEC ceramic membranes, 252 Silver doping, for carbon molecular sieves, 97 Sintering conditions dual-phase membrane oxygen permeation and, 287–288, 287f oxygen flux and, 239–240 SiO2, CVD on Vycor glass, 38–39 Slit-shaped pores fluid flow through, 335–336, 336f solute transport in, 336–338, 338f Sol-gel technique CVD v., 26 gas permeance, 98 H2 v. CO selectivity, 80 for silica membrane deposition, 26, 119–120 composite membranes, 79 pore size and, 117 pore structures of, 98 stability of, 35 for zeolite membrane synthesis, 183

Index

Solid solutions, of oxygen-conducting MIEC ceramic membranes, 238 Solid state reaction (SSR) method, for preparing dual-phase membranes, 282–283, 285–287, 286f Solubility, of hydrogen in metal membranes, 311–312 amorphous membranes, 317, 320–322, 321f Solubility sites, of silica membranes, 80–81, 81f siloxane ring as model of, 82 Solute transport, in high performance ultrafiltration membranes electrostatic interactions affecting, 344–347, 345f, 346f pore geometry effects on, 336–339, 338f Solution-based synthesis methods, for preparing dual-phase membranes, 283–285, 284f, 285f Solution-diffusion in polymeric membranes, 102 properties and, 17–18, 18f in silica membranes, 36–37, 36f of H2, 38 Solvent extraction, 214. See also Membrane-based solvent extraction Sorption of gases in glassy polymers, 98 in hydrogen permeation, 36 Sorption coefficients, in CMS membranes BTDA-DAPI, 157–159, 158t 6FDA/BPDA-DAM, 157–159, 158t Sorption isotherms of BTDA-DAPI, 154–157, 155f, 156f, 156t, 157t of 6FDA/BPDA-DAM, 154–157, 155f, 156f, 156t, 157t Sorption tests, for CMS membrane characterization, 146 Sorption-diffusion mechanism, gas transport in CMS, 138 Sr0.97Ti0.6Fe0.4O3–d, 252 SrCo0.8Fe0.2O3–d, 243, 246 Sr–Co–Fe (SCF) perovskite materials in oxygen-conducting MIEC ceramic membranes, 243–248 multiphase SFC, 247–248 promoted SCF materials, 246 substituted SCF materials, 246–247 in POM reactions, 258 SrFeCo0.5O3–d (SFC), 238

371 surface modifications to, 252 SrSnO3-doped SCF, 246 SSR method. See Solid state reaction method Stability of dual-phase membranes, oxygen permeation and, 278–280, 280f of perovskites, 241 Stability studies, of silica membranes, 35 with one-sided CVD, 39–41 with sol-gel technique, 35 Stabilized ZrO2, in oxygen-conducting MIEC ceramic membranes, 240 Stagnant film model, 347 Stainless steel, zeolite membranes on, 93, 95–96 Standard deviations, 13 Standard low-pressure CVD. See One-sided CVD Steam pretreatment, for CVD on alumina, 54 Steam reforming of ethanol, 26, 26f of hydrocarbons, 256t Stockmayer potential, for kinetic diameter, 119 Streaming potential, 341–342 Structure directing agents (SDAs), for zeolite membrane synthesis, 183–184 Substituted SCF materials, 246–247 Supported ionic liquid membranes (SILMs), 92 for CO2/CH4 separation, 103, 105–106t, 107 performance of, 107, 108f, 109 Supports, for CVD of silica membranes, 30 Surface area, of inorganic membranes, 61 Surface diffusion, 118 in zeolite membranes, 93 Surface exchange, dual-phase membrane oxygen permeation and, 281–282, 282f Surface modifications, on oxygen-conducting MIEC ceramic membranes, 252, 253f Syngas. See Synthesis gas Synthesis gas (Syngas), oxygen-conducting MIEC ceramic membranes in production of, 236, 255–263, 256f, 256t, 259t, 260f, 262f benefits of, 257–258 overview of, 255–257, 256t reaction environment effects on membrane performance, 261–263, 262f reaction temperature effects on membrane performance, 260–261, 260f work to date on, 258–260, 259t

372 Synthesis methods, for dual-phase membranes, 282–287, 284f, 285f, 286f Synthetic substitution, of HP polymers, 297, 299–300

T Temperature, Knudsen diffusion and, 38, 67 TEOS. See Tetraethoxysilane Tetraethoxysilane (TEOS) in colloidal silica sol preparation, 65, 119–120 for CVD on alumina evacuation and, 48–51, 49f with TMOS, 53–54 for CVD on HF membranes, 64–66 molar flow rate of, 73, 74f for CVD on Vycor glass, 37–38, 43 with ATSB, 43 with O3, 45 SiCl4 v., 43 with TIPT, 43–44 in Nanosil, 46 Tetraisopropyl titanate (TIPT), for CVD on Vycor glass, with TEOS, 43–44 Tetramethoxysilane (TMOS) for CVD on alumina, with PTMS and DMDPS, 55–56 for CVD on HF membranes, 64 Tetramethylorthosilicate (TMOS), for CVD on alumina, with TEOS, 53–54 Tetrapropylammonium hydroxide (TPAHO), for zeolite synthesis, 183 TGA. See Thermogravimetric analysis TGA-FTIR of BTDA-DAPI, 153–154, 153f, 154f of 6FDA/BPDA-DAM, 153–154, 153f, 154f Theories, 3–4, 4f Thermal analysis of HP polymers, 297, 302t of VTEC polyimides, 301–302, 302t Thermal gas separation. See High temperature gas separations Thermal soak time, in CMS membrane gas separation and oxygen concentration, 162–166, 163f, 163t, 164f, 165f, 165t Thermodynamic correction factor, for modeling of amorphous metal membranes, 324–325 Thermogravimetric analysis (TGA) of HP polymers, 302t of VTEC polyimides, 301, 302t Thermomechanical analysis (TMA), of VTEC polyimides, 301, 302t

Index

Thickness reduction, of oxygen-conducting MIEC ceramic membranes, 253–255, 254f TiO2. See Titania TIPT. See Tetraisopropyl titanate Titania (TiO2) in CVD on Vycor glass, 38–39, 43–44 silica membranes with, 79 Titania-silica membranes hydrogen permeation of, 80 permeation performance of, 87–88, 88t TMA. See Thermomechanical analysis TMOS. See Tetramethoxysilane; Tetramethylorthosilicate Total interaction potential, of high performance ultrafiltration membranes, 344 TPAHO. See Tetrapropylammonium hydroxide TPS. See Triisopropylsilane Transport diffusion coefficient, of hydrogen in amorphous metal membranes, 322–323 Triisopropylsilane (TPS), in CVD on Vycor glass, 42–43, 42f Two-step masking, for zeolite membrane synthesis, 183

U

UF. See Ultrafiltration Ultrafiltration (UF), 333 Ultrafiltration membranes. See High performance ultrafiltration membranes Unsteady state, of dual-phase membrane oxygen permeation, 281–282, 282f

V Vapor permeation (VP) distillation processes v., 177f xylene separation, 195 Vera causa, 2 Viscous flow, 118 VP. See Vapor permeation VTEC polyimides, 297, 300–301 gas permeation of, 301–304, 303t, 304t moisture within, 304–305, 305t PAL spectroscopy of, 305, 305t PBI–TMS and VTEC blend, 304 thermal analysis of, 301–302, 302t Vycor glass for CVD of silica membrane, 26, 30, 37–48, 40f, 42f, 46f alternating reactant deposition, 41 alumina substrate v., 48

373

Index

Vycor glass (cont.) binary systems, 43–44 with carbon barriers, 41–42 densification in, 38–39 with DES and N2O, 44 for methanol vapor separation, 44–45 Nanosil, 45–47, 46f with SH4, 38 with SiCl4, 38–39 stability studies of, 39–41 with TEOS, 37–38, 43 with TEOS/O3, 45 with TPS, 42–43, 42f hydrogen permeation with, 30 pore size of, 48

W Wagner equation, 253 Water (H2O). See also De-watering AcOH separation from, 176–179, 177f within HP polymers, 304–305, 305t silica membrane permeability and, 35 Water (H2O) vapor carbon molecular sieves and, 97 silica membranes and, 79, 131–132 gas permeance through Co-doped, 129–130, 130f hydrothermal stability of, 118 Westlake criterion, for H–H interaction in metals, 319, 320f, 323–324 Wilke–Chang equation, 224–225 William of Ockham, 1–2

X

m-Xylene (MX), MFI zeolite membrane separation of, 195 bilayer structure reversal, 207–209, 208f, 209t higher PX feed concentrations, 209–210, 210t with single and bilayer and membranes, 202–206, 203f, 205f, 206f, 207f o-Xylene (OX), MFI zeolite membrane separation of, 195 bilayer structure reversal, 207–209, 208f, 209t higher PX feed concentrations, 209–210, 210t with single and bilayer and membranes, 202–206, 203f, 205f, 206f, 207f p-Xylene (PX), MFI zeolite membrane separation of, 195 bilayer structure reversal, 207–209, 208f, 209t

higher PX feed concentrations, 209–210, 210t with single and bilayer and membranes, 202–206, 203f, 205f, 206f, 207f Xylene separation, with MFI zeolite membranes, 195–210, 197f, 199f pervaporation of, 200–201 bilayer structure reversal, 207–209, 208f, 209t higher PX feed concentrations, 209–210, 210t through single and bilayer membranes, 202–206, 203f, 205f, 206f, 207f

Z Zeolite membranes, 92–93, 180–181. See also A-type zeolites; MFI zeolite membranes; Mordenite; Silicalite; ZSM-5 for CO2/CH4 separation, 93–97, 94–95t de-watering with, 184–188 of alcohol, 184–186, 185f, 186f for C1 chemistry, 187–188, 187f, 189f of organic acids, 186–187 gas separation properties of, 94–95t, 95–96 hydrophilic/hydrophobic nature of, 181 limitations of, 96 performance of, 107, 108f, 109 pore dimensions of, 180, 180f promise of, 179–181, 180f structure of, 180 supports for, 93, 95–96 synthesis of, 181–184 masking technique, 182–183 SDA for, 183–184 seeding technique, 181–182 Zirconia (ZrO2), in oxygen-conducting MIEC ceramic membranes, 240–241 Zirconia-silica membranes, 79 hydrogen permeance of, 79–80 permeation performance of, 87–88, 88t ZrO2. See Zirconia ZSM-5 membrane, 180 characteristics of, 201–202, 201t, 202f, 230f membrane orientation in, 183 synthesis of, 199–200 xylene separation with, 198 bilayer structure reversal, 207–209, 208f, 209t higher PX feed concentrations, 209–210, 210t through single and bilayer membranes, 202–206, 203f, 205f, 206f, 207f