New Insights into Membrane Science and Technology: Polymeric and Biofunctional Membranes

NEW INSIGHTS INTO MEMBRANE SCIENCE AND TECHNOLOGY: POLYMERIC AND BIOFUNCTIONAL MEMBRANES Membrane Science and Technol...

Author: Dibakar Bhattacharyya | Allan DA Butterfield

34 downloads 919 Views 19MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

NEW INSIGHTS INTO MEMBRANE SCIENCE AND TECHNOLOGY: POLYMERIC AND BIOFUNCTIONAL MEMBRANES

Membrane Science and Technology Series Volume 1 : Pervaporation Membrane Separation Processes Edited by R.Y.M. Huang (1991) Volume 2 : Membrane Separations Technology, Principles and Applications Edited by R.D. Noble and S.A. Stern (1995) Volume 3 : Inorganc Membranes for Separation and Reaction By H.R Hsieh (1996) Volume 4 : Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot (1996) Volume 5 : Membrane Biophysics Edited by H. Ti Tien and A. Ottova-Leitmannova (2000) Volume 6 : Recent Advances in Gas Separation by Microporous Ceramic Membranes Edited by N.K. Kanellopoulos (2000) Volume 7 : Planar Lipid Bilayers (BLMs) and their Applications Edited by H.T. Tien and A. Ottova-Leitmannova (2003) Volume 8 : New Insights into Membrane Science and Technology: Polymeric and Biofunctional Membranes Edited by D. Bhattacharyya and D.A. Butterfield

Membrane Science and Technology Series, 8

NEW INSIGHTS INTO MEMBRANE SCIENCE AND TECHNOLOGY: POLYMERIC AND BIOFUNCTIONAL M E M B I ~ N E S Edited

by

Dibakar Bhattacharyya Department of Chemical and Materials Engineering and Center of Membrane Sciences University of Kentucky, Lexington, KY 40506-0046, USA and

D. Allan Butterfield Department of Chemistry and Center of Membrane Sciences University of Kentucky, Lexington, KY 40506-0059, USA

2003 ELSEVIER Amsterdam

- B o s t o n - L o n d o n - N e w Y o r k - O x f o r d - Paris

San Diego - San Francisco.

Singapore - Sydney - Tokyo

ELSEVIER SCIENCE B.V. Sara Burgerhartstraat 25 P.O. Box 211,1000 AE Amsterdam, The Netherlands © 2003 Elsevier Science B.V. All rights reserved.

This work is protected under copyright by Elsevier Science, and the following terms and conditions apply to its use: Photocopying Single photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permission of the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying, copying for advertising or promotional purposes, resale, and all forms of document delivery. Special rates are available for educational institutions that wish to make photocopies for non-profit educational classroom use. Permissions may be sought directly from Elsevier's Science & Technology Rights Department in Oxford, UK: phone: (+44) 1865 843830, fax: (+44) 1865 853333, e-mail: [email protected]. You may also complete your request on-line via the Elsevier Science homepage (http://www.elsevier.com), by selecting 'Customer Support' and then 'Obtaining Permissions'. In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; phone: (+1) (978) 7508400, fax: (+1) (978) 7504744, and in the UK through the Copyright Licensing Agency Rapid Clearance Service (CLARCS), 90 Tottenham Court Road, London W1P 0LP, UK; phone: (+44) 207 631 5555; fax: (+44) 207 631 5500. Other countries may have a local reprographic rights agency for payments. Derivative Works Tables of contents may be reproduced for internal circulation, but permission of Elsevier Science is required for external resale or distribution of such material. Permission of the Publisher is required for all other derivative works, including compilations and translations. Electronic Storage or Usage Permission of the Publisher is required to store or use electronically any material contained in this work, including any chapter or part of a chapter. Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher. Address permissions requests to: Elsevier's Science & Technology Rights Department, at the phone, fax and e-mail addresses noted above. 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. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made.

First edition 2003 Library of Congress Cataloging in Publication Data A catalog record from the Library of Congress has been applied for. British Library Cataloguing in Publication Data A catalogue record from the British Library has been applied for.

ISBN: 0-444-51175-X ISSN: 0927-5193 ~ T h e paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper). Printed in Hungary.

TO:

Bhattacharyya: Gloria and Anita, my wife and daughter, for their support and understanding," my graduate students, for making teaching and research life to be very exciting. Butterfield: Marcia and Nyasha, my wife and daughter, who have been so inspirational, encouraging, and loving.

This page intentionally left blank

Preface Membrane techniques provide a broad science and technology base with applications ranging from water purification to chemical/biomolecule synthesis, material recovery, medical devices, to nano-domain interaction- based sensors and highly selective separations. Although there are several books in the traditional membrane field, there is a great need for a highly comprehensive book containing advances in both synthetic and biofunctional/bimimetic membranes including non-invasive characterization to biomedical devices. Over the years the membrane field has advanced from the development of thinfilm composite membranes for desalination to recognition-based separation and reactions by taking advantage of biomolecular interactions in the nano-domain. This refereed book covers materials from highly respected researchers with topics ranging from membrane transport models to non-invasive characterization, functionalized material, biomedical devices, to sensors and environmental applications. Nineteen chapters in this special book are invited, refereed papers, and mostly based on the 2001 North American Membrane Society Annual Meeting held in Lexington, KY, organized and hosted by the University of Kentucky Center of Membrane Sciences. The book is divided into two sections. Section I, is subdivided into three areas, Advances in Membrane Transport~ouling, Imaging Techniques, and Contacting Devices. Section II deals with Functional Membranes and Materials for Biocatalysis, Separation, and Analysis and is further subdivided into three areas ranging from biofunctional membranes to sensors. Authors and co-authors are from various fields including chemistry, chemical engineering, mechanical engineering, biomedical engineering, biotechnology, and environmental engineering. This book is highly multidisciplinary in nature and should be highly valuable to scientists and engineers involved in activities ranging from separations/reactions, to advanced biofunctional materials, to contactor designs, to the general field of membrane science and technology. Students and faculty members around the world should find this to be an excellent reference book for courses ranging from traditional separation, to bioscience/engineering, to formal membrane courses. Each chapter of this book was peer-reviewed, and we would like to give special thanks to the reviewers for this book, including various authors and co-authors of the chapters, and Drs. Kloos (Osmonics Corporation) and Hestekin (Argonne National Lab). Thanks also go to the NAMS Board of Directors, and in particular to NAMS presidents Drs. Pushpinder Puri and Glenn Lipscomb for their encouragement of the book. Mollie Fraim of the University of Kentucky °. Vll

Center of Membrane Sciences did an extraordinary amount of work with us to bring this book to fruition, and we would like to express our sincere appreciation and thanks to Ms. Fraim. The editors would also like to thank Mike Lundin (chemical engineering student) for his hard work in putting the book in right format.

Dibakar Bhattacharyya, Ph.D. Department of Chemical and Materials Engineering and Center of Membrane Sciences University of Kentucky Lexington, KY D. Allan Butteriield, Ph.D. Department of Chemistry and Center of Membrane Sciences University of Kentucky Lexington, KY

viii

About the Editors Dibakar Bhattacharyya (DB) is the University of Kentucky Alumni Professor of Chemical Engineering and a Fellow of the American Institute of Chemical Engineers. He obtained his B.S. (Jadavpur University) and M.S. ( Northwestern University, Evanston, IL) in Chemical Engineering, and his Ph.D. in Environmental Engineering from the Illinois Institute of Technology. He is the Co-Founder of the Center for Membrane Sciences at the University of Kentucky. He and his students have advanced passive membrane applications to functionalized poly-ligand membranes for high capacity metal capture to site-specific biocatalysis for organic degradation to membrane-based nanoparticle synthesis for dechlorination reactions. He has published over 140 refereed journal articles and book chapters, and has recently received three U.S. Patents (two on Functionalized Membranes for Toxic Metal Capture, and one on hazardous waste destruction technology). Dr. Bhattacharyya has mentored many graduate and undergraduate students in the area of water and wastewater related research and membrane separation. For his research, Dr. Bhattacharyya has received funding from U.S. EPA, DoD, NSF, NIEHS-SBRP, Dow Chemical, Glaxo SmithKline, Eastman Chemical Co., Daramic, Inc., etc. He has received a number of awards for his research and educational accomplishments, including the Larry K. Cecil AIChE Environmental Division Award, the Kentucky Academy of Sciences Distinguished Scientist Award, Henry M. Lutes Award for Outstanding Undergraduate Engineering Educator, AIChE Outstanding Student Chapter Counselor Awards, and the University of Kentucky Great Teacher (1984 and 1996) Awards. D. Allan Butterfield received his B.A. in Chemistry from the University of Maine in 1968. Following three years of teaching Africans chemistry and mathematics in Zimbabwe, he entered Duke University, receiving the Ph.D. in Physical Chemistry in 1974. A NIH Postdoctoral Fellowship at the Duke University Neuroscience Program followed. Dr. Butterfield joined the Chemistry Department at the University of Kentucky soon after and was quickly promoted to Full Professor in 1983. In 1986, he and several others, but principally Professor Dibakar Bhattacharyya, formed the University of Kentucky Center of Membrane Sciences, and Professor Butterfield has been Director from its inception. Continuous federal funding from NIH, NSF, and DoD has supported his research on membrane structure and function in neurodegenerative disorders and on applications of biofunctional membranes to important societal problems. He and his students have published over 260 refereed papers. Professor Butterfield received the Southern Chemist Award from the American Chemical Society. He has directed the graduate careers of over 50 Ph.D. and M.S. students, and in 1998 Professor Butterfield received the Presidential Award for Excellence for Science, Mathematics, and Engineering Mentoring from President Clinton in the White House. This award was his efforts to increase the number of female and Appalachian Ph.D. students in Chemistry, both groups being highly under represented in this discipline. In 2002, the University of Kentucky Board of Trustees appointed Dr. Butterfield the Alumni Professor of Chemistry.

ix


Contents Preface ................................................................................................

vii

About the Editors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

List of Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xv

A d v a n c e s in M e m b r a n e T r a n s p o r t / F o u l i n g , I m a g i n g T e c h n i q u e s , a n d C o n t a c t i n g D e v i c e s .............................................................................. 1

Membrane Transport Models, Fouling, and Formation ..................... 3

Mass Transfer in Axial Flows through Randomly Packed Fiber Bundles ......................................................................... 5 L. Bao, G.G. Lipscomb Fouling Phenomena during Microfiltration: Effects of Pore Blockage, Cake Filtration, and Membrane Morphology ....... 27 A.L. Zydney, C. Ho, W. Yuan Differential Scanning Calorimetry and rheological experiments to study membrane formation via thermallyinduced phase-separation .................................................................. 45 P.C. van der Heijden, M.H.V. Mulder, M. Wessling Non-lnvasive Characterizations o f Membrane Fluid Transport and Fouling ........................................................................................ 63

Study of Membrane Fouling and Cleaning in Spiral Wound Modules Using Ultrasonic TimeDomain Reflectometry ...................................................................... 65 Zh.-X. Zhang, A.R. Greenberg, W.B. Krantz, G.-Y. Chai Nonintrusive Characterization of Fluid Transport Phenomena in Hollow-Fiber Membrane Modules Using MRI: An Innovative Experimental Approach ......................... 89 C.K. Poh, P.A. Hardy, W.R. Clark, D. Gao Membrane Contactors and Environmental Applications ................ 123

Industrial applications and opportunities for membrane contactors ....................................................................................... 125 R. Klaassen, P.H.M. Feron, R. van der Vaart, A.E. Jansen xi

Membrane Contactors: Recent Developments ................................ 147 A.S. Kovvali, K.K. Sirkar

Membrane Aromatic Recovery System (MARS)A New Process for Recovering Phenols and Aromatic Amines from Aqueous Streams ....................................... 165 F.C. Ferreira, A. Livingston, S. Han, A. Boam, S. Zhang Functional Membranes and Materials for Biocatalysis, Separation, and Analysis ............................................................................................ 183

Biofunctional Membranes and Biomedical Devices ........................ 185 Membrane Bioreactors for Biotechnology and Medical Applications ................................................................... 187 L. Giomo, L. De Bartolo, E. Drioli Structural and Performance Characteristics of Hemodialysis Membranes ............................................................ 219 D. Gao, W.R. Clark Biofunctional Membranes: Site-Specifically Immobilized Enzyme Arrays ........................................................ 233 D.A. Buttertield, D. Bhattacharyya Biocatalytic Membrane Reactor with Continuous Removal of Organic Acids by Electrodialysis .............................. 241 H.C. Ferraz, T.L.M. Alves, C.P. Borges Use of Micro-Porous Affinity Membranes for Protein Purification: A Case Study ............................................... 263 F. Cattoli, G.C. Sarti Economic Production of Biopharmaceuticals by High-Speed Membrane Adsorbers ............................................... 283 W. Demmer, S. Fischer-Frtihholz, D. NuBbaumer, D. Melzner Functionalized Membranes f o r Separations and Reactions ............ 297 Polymer Grafted Membranes ....................................................... 299 S.M.C. Ritchie

xii

Functionalized Membranes for Tunable Separations and Toxic Metal Capture ............................................................... 329 A.M. Hollman, D. Bhattacharyya The Design of High Performance, Gel-Filled Nanofiltration Membranes ........................................................... 353 R.F. Childs, A.M. Mika Sensors ............................................................................................. 377 Membranes for the Development of Biosensors ........................... 379 V.G. Gavalas, J. Wang, L.G. Bachas

Ion-Partitioning Membranes as Electroactive Elements for the Development of a Novel Cation-Selective CHEMFET Sensor System ................................ 393 E.A. Moschou, N.A. Chaniotakis

Index ........................................................................................................ 415

xiii


List of Contributors

T.L.M. Alves Chemical Engineering Program- COPPE Federal University of Rio de Janeiro, P.O. Box: 68502 21945-970, Rio de Janeiro R J - Brazil Leonidas G. Bachas Department of Chemistry and Center of Membrane Sciences University of Kentucky Lexington, Kentucky 40506-0055 USA E-mail: bachas@pop, uky. edu Lihong Bao Chemical & Environmental Engineering Department Mail Stop 305 The University of Toledo Toledo, Ohio 43606-3390 USA Dibakar Bhattacharyya Department of Chemical and Materials Engineering and Center of Membrane Sciences University of Kentucky Lexington, KY 40506-0046 USA E-mail: db@engr, uky. edu Andrew Boam Membrane Extraction Technology Ltd Imperial College London SW7 2BY UK C.P. Borges Chemical Engineering Program - COPPE Federal University of Rio de Janeiro, P.O.Box: 68502 21945-970, Rio de Janeiro R J - Brazil E-mail: cristiano@peq,coppe, ufrj.br

XV

D. Allan Butterfield Department of Chemistry and Center of Membrane Sciences University of Kentucky Lexington, KY 40506-0059 USA E-mail: dabcns@uky, edu Francesca Cattoli DICMA - Dipartimento di Ingegneria Chimica, Mineraria e delle Tecnologie Ambientali Alma Mater Studiorum - Universit~t di Bologna, viale Risorgimento 2 40136 Bologna Italy G.-Y. Chai Department of Chemical Engineering University of Colorado Boulder, Boulder CO 80309-0424 USA Nikolas A. Chaniotakis Laboratory of Analytical Chemistry Department of Chemistry University of Crete, 71409 Iraklion, Crete Greece Ronald F. Childs Department of Chemistry McMaster University Hamilton, ON, L8S 4M1 Canada E-mail: rchilds@mcmaster, ca William R. Clark, M.D. Renal Division Baxter Healthcare Corporation McGaw Park, IL Nephrology Division Indiana University School of Medicine Indianapolis, IN USA E-mail: bill_clar~axter, com

xvi

L. De Bartolo Research Institute on Membranes and Modelling of Chemical Reactors, IRMERC-CNR c/o University of Calabria via P. Bucci cubo 17/C, 87030 Rende (CS) Italy W. Demmer Sartorius AG G/Sttingen Germany E. Drioli Research Institute on Membranes and Modelling of Chemical Reactors IRMERC-CNR c/o University of Calabria via P. Bucci cubo 17/C, 87030 Rende (CS) Italy E-mail: e. drioli@irmerc, cs. cnr. it

P.H.M. Feron TNO Institute of Environmental Sciences Energy Research and Process Innovation Department Chemical Engineering PO Box 342 7300 AH Apeldoorn The Netherlands H.C. Ferraz Chemical Engineering Program - COPPE Federal University of Rio de Janeiro, P.O.Box: 68502 21945-970, Rio de Janeiro R J - Brazil S. Fischer-Frtihholz Sartorius AG Gfttingen Germany Frederico Castelo Ferreira Department of Chemical Engineering and Chemical Technology Imperial College of Science Technology and Medicine Prince Consort Road London SW7 2BY UK

xvii

Dayong Gao Department of Mechanical Engineering Center for Biomedical Engineering University of Kentucky Lexington, KY USA E-mail: dgao@engr, uky. edu Vasilis G. Gavalas Department of Chemistry University of Kentucky Lexington, Kentucky 40506-0055 USA L. Giomo Research Institute on Membranes and Modelling of Chemical Reactors, IRMERC-CNR c/o University of Calabria via P. Bucci cubo 17/C, 87030 Rende (CS) Italy A.R. Greenberg Department of Mechanical Engineering University of Colorado Boulder, Boulder CO 80309-0427 USA E-mail: greenbea@,spot, colorado, edu Shejiao Han Department of Chemical Engineering and Chemical Technology Imperial College of Science Technology and Medicine Prince Consort Road London SW7 2BY UK Peter A. Hardy Center for Biomedical Engineering University of Kentucky Lexington, KY USA Chia-Chi Ho Department of Chemical Engineering University of Cincinnati Cincinnati, OH 45221 USA

xviii

Aaron M. Hollman Department of Chemical & Materials Engineering University of Kentucky Lexington, Kentucky 40506-0046 USA A.E. Jansen TNO Institute of Environmental Sciences Energy Research and Process Innovation Department Chemical Engineering PO Box 342 7300 AH Apeldoom The Netherlands R. Klaassen TNO Institute of Environmental Sciences Energy Research and Process Innovation Department Chemical Engineering PO Box 342 7300 AH Apeldoorn The Netherlands E-mail: r. klaassen@mep, tno.nl A. Sarma Kovvali Department of Chemical Engineering New Jersey Institute of Technology Newark, NJ 07102 USA W.B. Krantz Department of Chemical Engineering University of Cincinnati Cincinnati, Ohio 45221-0171 USA G. Glenn Lipscomb Chemical & Environmental Engineering Department Mail Stop 305 The University of Toledo Toledo, Ohio 43606-3390 E-mail: [email protected]

xix

Andrew Livingston Membrane Extraction Technology Ltd. Imperial College London SW7 2BY UK and

Department of Chemical Engineering and Chemical Technology Imperial College of Science, Technology and Medicine Prince Consort Road London SW7 2BY UK E-mail: a. livingston@ic, ac. uk

D. Melzner Sartorius AG Gfttingen Germany E-mail: dieter.melzner@sartorius, com

Alicja M. Mika Department of Chemistry McMaster University Hamilton, ON, L8S 4M1 Canada Elizabeth A. Moschou Laboratory of Analytical Chemistry Department of Chemistry University of Crete, 71409 Iraklion Greece Current address: Department of Chemistry University of Kentucky Lexington KY 40506-0055 USA E-mail: l_moschou@yahoo, com

M.H.V. Mulder University of Twente PO Box 217 7500 AE Enschede The Netherlands D. NuBbaumer Sartorius AG Grttingen Germany

XX

Chum K. Poh Department of Mechanical Engineering University of Kentucky Lexington, KY USA S.M.C. Ritchie Department of Chemical Engineering University of Alabama Tuscaloosa, AL, 35487-0203 USA E-mail: sritchie@bama, ua. edu Giulio C. Sarti DICMA - Dipartimento di Ingegneria Chimica Mineraria e delle Tecnologie Ambientali Alma Mater Studiorum - Universith di Bologna viale Risorgimento 2 40136 Bologna Italy E-mail: giulio.sarti@,mail, ing. unibo, it Kamalesh K. Sirkar Department of Chemical Engineering New Jersey Institute of Technology Newark, NJ 07102 USA E-mail: sirkar@adm, nj it. edu P.C. van der Heijden University of Twente PO Box 217 7500 AE Enschede The Netherlands E-mail: pcvdheijden@hetnet, nl R. van der Vaart TNO Institute of Environmental Sciences Energy Research and Process Innovation Department Chemical Engineering PO Box 342 7300 AH Apeldoom The Netherlands Jianquan Wang Department of Chemistry University of Kentucky Lexington, Kentucky 40506-0055 USA xxi

M. Wessling University of Twente PO Box 217 7500 AE Enschede The Netherlands Wei Yuan Innovasep Technologies 420 Maple Street Marlborough, MA 01752 USA Shengfu Zhang Membrane Extraction Technology Ltd Imperial College London SW7 2BY UK Zh.-X. Zhang Department of Chemical Engineering University of Colorado Boulder, Boulder CO 80309-0424 USA Andrew L. Zydney Department of Chemical Engineering The Pennsylvania State University University Park, PA 16802 USA E-mail: [email protected], edu

xxii

Advances in Membrane Transport/Fouling, Imaging Techniques, and Contacting Devices


Membrane Transport Models, Fouling, and Formation


New Insights into Membrane Science and Technology: Polymeric and Biofunctional Membranes D. Bhattacharyya and D.A. Butterfield (Editors) © 2003 Elsevier Science B.V. All rights reserved.

Chapter 1

Mass transfer in axial flows through randomly packed fiber bundles L. Bao and G, G. Lipscomb* Chemical & Environmental Engineering Department, Mail Stop 305, The University of Toledo, Toledo, Ohio 43606-3390 *Corresponding Author, e-mail: [email protected], Phone: (419) 530-8088, Fax: (419) 530-8086 1.

ABSTRACT

The literature contains numerous correlations of mass transfer coefficients for axial flows through randomly packed fiber bundles. The predictions of these correlations can differ by an order of magnitude. This large variation severely limits the usefulness of the correlations for design purposes and confounds an understanding of the underlying physics. While the literature suggests randomness in fiber packing may contribute to the variation, a rigorous analysis of mass transfer in randomly packed bundles has not been reported. The results of such analyses are summarized here for uniform wall concentration and uniform wall mass flux boundary conditions. The results indicate that channeling through randomly packed bundles can dramatically reduce mass transfer coefficients relative to regularly packed bundles, especially m the welldeveloped limit. However, experimental mass transfer coefficients are significantly different from predicted values for regularly and randomly packed bundles; the literature contains correlations that predict much higher and much lower values. This suggests that other factors can control performance such as cross-flow regions that exist near shell ports or poor fluid distribution from the shell inlet port. 2.

INTRODUCTION

Hollow fiber membrane modules are used for a wide range of applications including dialysis [1], filtration [2], eontaetors [3], and gas separations [4]. Typically, the modules consist of a bundle of randomly packed hollow fibers. The ends of the fiber bundle are potted to form tubesheets - slicing the tubesheets provides access to the fiber interior or lumens. The bundle is inserted into a case and a seal formed between the tubesheet and ease. Ports on the

Mars TranafcrIn Axial Flm Tlmugb Rmdaoly Packed Fiber B d e d - Lipscomb

periphery of the case permit fluid introduction and removal from the shell space (the region outside the fibers) while headers on either end permit fluid introduction and removal from the lumen space (the region inside the fibers). The tubesheet prevents the direct mixing of the shell and lumen fluids. Figure 1 illustrates a typical module. In operation, mass transfer occurs between the fluids flowing through the lumen and shell and the module functions llke the mass exchange equivalent of a shell-and-tube heat exchanger. The material used to produce the hollow fiber membranes can facilitate the desired separation - ideally, the membrane would selectively permeate one or more components of the feed streams at high rates. Through proper choice of membrane material and operating conditions, one can separate the components of a feed stream into hgh-purity product streams. Unfortunately, the efficiency and throughput of these devices can be limited not by the membrane material but by concentration boundary layers that develop in the lumen and shell spaces. The literature [ 5 ] indlcates that the effect

Figure 1. Schematic of a typical hollow fiber membrane module and the case that holds it.

of lumen boundary layers is described well by the Graetz solution [6] for mass transfer coefficients in pipe flows. However, there is much greater uncertainty about the effect of shell boundary layers. Theoretical analyses and experimental measurements of shell-side mass transfer coefficients have been confounded by the complex geometry of the shell and the flows within it. This complexity arises primarily from four factors: 1) fibers are packed randomly in the module whch leads to irregular flow domains that can be described only statistically (i.e., one can speclfy a packing fraction but not the exact location of the fibers) in contrast to the single, well-defined domain that exists in regularly packed bundles; 2) the fibers may not lie parallel to each other which can lead to a transverse flow (cross-flow) component and a velocity field with three components that varies in all three coordinate duections in contrast to a uniaxial velocity field that varies in two coordinate directions in regularly packed bundles; 3) the need to introduce and remove fluid from the shell leads to cross-flow regions and axial-flow regions - concentration boundary layer growth depends strongly on the relative size of these regions but most analyses of heat and mass transfer assume only axial flow; and 4) fluid

6

Mass Transfer In Axial Flows Through Randomly Packed Fiber Bundles - Lipscomb

distribution into the bundle may not be uniform if the entrance and exit ports are not well designed- locally flows may be higher or lower than expected based on the assumption of a uniform axial pressure gradient in the bundle (This has been seen experimentally in hemodialyzers [7] where flows appear higher in that part of the bundle closest to the inlet port.). The last two factors can affect performance in both randomly and regularly packed fiber bundles. All four are illustrated in Figure 2. a) cross-sections of bundles with random (left) and regular square (fight) fiber packings o• -N • • o•i O %

/_

. I o og

/oo

•

/000000% ~ooooooo% Ioooooooo Iooooooool

;o_O

i• • OoOp I gO • o o oa,/ gO | ooa= % o - g o • ,=3" \ o • o_O • • ~o % oe-o o //

!

ooooeo

\go•o••/ %~ooo/

b) transverse flows that arise from flow past non-parallel fibers (left) that do not exist when fibers are parallel (right) I

,

''

'

,

'

....

,'

,'

,

, , l

I'

""

, , ,

'

I

i

,

I

I

c) cross-flow regions near shell entrance and exit ports due to fluid distribution from the shell port across the fiber bundle

I

i

"

OOi' I

,

"' gOl

I

,

~:>

got

I

,

,

i

i

i

,,i,,!

, i

i i

ii

ii

ii

d) more fluid may flow through that portion of the bundle closest to the shell inlet port than in other regions .Shell

Ftber

• "

oundle

~

: " ~ = ~ '

inlet

port

hell fluid distribution collar

Figure 2. Four complexities associated with shell flow through hollow fiber membrane modules: a) random fiber packing, b) transverse flows arising from non-parallel fibers, c) cross-flow regions near shell inlet port, and d) poor distribution from shell inlet port.

Mass Transfer In Axial Flows Through Randomly Packed Fiber Bundles -

Lipscomb

Despite these challenges, the literature offers a number of mass transfer coefficient correlations nominally for axial flows through randomly packed hollow fiber bundles [8-13]. Lipnizki and Field [14] provide an excellent summary and discussion of the reported expressions. The proposed correlations are summarized in Table 1 in terms of the mean Sherwood number: Shr,,=2Rkm/D where R is the fiber radius, km the mean mass transfer coefficient and D the diffusion coefficient. The Sherwood number is a function of the Reynolds number (Re=2RVb,O/lt), Schmidt number (Sc=ldp/D), fiber length (L), and fiber packing fraction (¢~= total fiber cross-sectional area / module cross-sectional area). Here, V5 is the bulk fluid velocity in the shell, ,o the fluid density, and/t the fluid viscosity. Table 1. Literature mass transfer coefficient correlations for axial flows through hollow fiber bundles and the reported range of conditions used to obtain the correlations. In all cases the Schmidt number is approximately 500-1000. Note [15] that the correlation for Prasad and Sirkar [9] has been modified to provide a better fit to their data. Additionally, (1-¢)/~ is the ratio of the hydraulic diameter to the fiber diameter. Correlation

[8]

/, ~'xo.s6/,,,,,~o.93 Sh,,, = 1.25~2-~ l ~ ~ Re°93Sc °'33

[9]

Shin = 11.5

ll-',/[

(1_ +t_._~. j

[10] Shr,, = (0.53-0.58~tL~) [11] Sh,,, = 8.71

ReO.6ScO.33

0001

0

0001

,('1- ~'~-°'2-°16~ Re°'S-°'16~Sc °'33

Sh,,, = (0.31¢2 -0.34~+0.10

Re°'9S¢°'33

0.003

0.004 30

Re°'74Sc °'33

2R/L 0.0020.006

0 00

Re°SaSc°'33

[12] Sh. : 0.090-¢~-7J [13]

100~ Re(1-~)/ # 3-26 0.5-500

0-0.2

0.002

35-80 0-0.5

0.0030.005

10-70 -~0-200

-~0.006

Figure 3 compares the predictions of the mean Sherwood number as a function of Graetz number (defined here as Gz = Re*Sc*(rcR/2L)*(1/~-1) to facilitate comparison with the theory described in the next section) for correlations valid over the range of conditions shown or that require limited extrapolation. The predictions for a regular square packing (constant wall mass flux boundary condition) are given as well; values for triangular packing are just

Lipscomb

Mass Transfer In Axial Flows Through Randomly Packcxl Fiber Bundles -

slightly higher. Clearly, there is significant variation between the correlations. Additionally, the correlations predict values that are b o t h much high and much lower than that for a regular packing. The literature attributes the large differences in mass transfer coefficients to the above four factors. However, a rigorous analysis of the effect of each is not available, and only the first (the effect of random fiber packing) has received much attention. Rogers and Long [16], Wu and Chen [13], and Lipnizki and Field [14] describe similar ad hoe approaches to predicting the effect of random fiber packing. Each set of authors introduces randomness by surrounding a representative set of fibers with an annular region of varying cross-sectional area; the first two pairs of authors use Voronoi tessellation to generate polygonal regions while the last assumes concentric circular regions, see Figure 4. The 10 2

101

Sh~ 10 0

1 0 -1 10

.

.

.

.

.

.

.

.

. 10 2

.

.

.

.

.

. 10 3

Gz

Figure 3. Comparison of literature correlations for mean shell-side mass transfer coefficients in randomly packed bundles with theoretical predictions for regular square packing for []=0.3, Sc=1000, and 2R/L=0.003. The sources ofthe correlations are: o Yang and Cussler [8], + Prasad and Sirkar [9], * Costello et al. [10], x Viegas et al. [11], 0 Gawronski and Wrzesinska [12], and small dot Wu and Chen [13]. The [] represent theoretical predictions for a square packing with uniform wall mass flux from Miyatake and Iwashita [23]. flow through each region is calculated from an appropriate friction factorReynolds number relationship using the region's hydraulic diameter and assuming the pressure drop across each region is the same. Each pair of authors predict a reduction in mass transfer coefficient due to randomness, but the effect of their ad hoe approach on the results is unclear. Results from rigorous analyses of the effect of random packing on mass transfer coefficients for axial flows through fiber bundles are summarized here. As found previously, randomness reduces mass transfer coefficients due to

Mass Transfer In Axial Flows Through Randomly Packed Fiber Bundles - Lip~omb

channeling in the fiber bundle. The results provide a baseline for evaluation of experimental results and suggest most experimental correlations do not reflect mass transfer in axial flows - either the shell-side flow possessed a significant cross-flow component or flow distribution from the shell ports into the bundle was poor. b) Surrounding each fiber with concentric circular regions

a) Using Voronoi tessellation to surround individual fibers with polygons

Figure 4. Approximate approaches for evaluating the effect of random packing on module performance - mass transfer rates are calculated for each fiber by using the hydraulic diameter of the surrounding shell region to calculate a flow rate and mass transfer coefficient for that region.

3. THEORY Commercial modules contain from 10,000 to over 1 million fibers. Unfortunately, finite computational resources limit the ability to calculate velocity and concentration fields within such large bundles. To reduce computational complexity, we use a unit cell containing a finite number of fibers to represent an infinite, spatially periodic bundle. This unit cell is translated in each co-ordinate direction to generate the bundle; the fibers lie parallel to the zaxis and the unit cell is translated along the x- and y-axes as illustrated in Figure 5. The value of a field variable at a point (x, y) in the unit cell is the same as the value at a point (x + ilx, y + fly) in the infinite bundle where i a n d j are integers and lx and ly represent the unit vectors in the x and y co-ordinate directions, respectively. Fiber packing fraction and type of packing (i.e., a regular or random packing) determine unit cell dimensions. Fiber centers are placed on a regular grid to produce regular packings while fibers centers are placed in a random, sequential fashion to produce random packings. The size and properties of all fibers are assumed to be identical. The random media produced by this method are representative of infinite random media only if the unit cell contains a sufficient number of fibers. 10


Calculations with increasing fiber numbers are used to verify that the calculated results do not depend on fiber number- up to 100 fibers were used in the work

lb-

Figure 5. Use of a unit cell to represent an infinite, randomly packed fiber bundle. The unit cell is indicated by the bold square and contains nine solid fibers. The solid lines emanating from the unit cell indicate the required translation to produce an infinite medium. The numbers 1 and 2 correspond to a pair of periodic boundaries: a denotes the top and bottom while b denotes the left and right.

reported here. However, given sufficient computational resources, the algorithms could be used for arbitrarily large fiber numbers. Additionally, one may use this approach to generate bundles containing non-uniform fibers if the unit cell contains a sufficiently large number of fibers to accurately represent the statistical distribution. This would further increase the required computational resources. The governing conservation of momentum and mass equations are solved given the following assumptions: 1. Fluid flow through the fiber bundle is axial, laminar, isothermal and welldeveloped; velocity field entrance lengths are more than an order of magnitude smaller than concentration field entrance lengths [14]. 2. Mass transfer rates are sufficiently low that shell flow rates are constant along the fiber bundle. 3. The wall concentration or wall mass flux along the surface of each fiber in the mass transfer region is constant. 4. Mass transfer by axial diffusion is negligible relative to convective mass transfer. 5. Diffusion is Fickian. 44


Lipscomb

6. Physical properties of the fluid are constant. These assumptions are equivalent to those used to obtain the Graetz solution for mass transfer in pipes and allow one to write the conservation of momentum and mass equations as"

02V 02V

--+

ax 7

=1

(1)

0¢ 02¢ 02¢

v-- = --+--

Oz

Ox2

(2)

03,2

where v is the dimensionless axial velocity, c the dimensionless concentration, and x, y, and z the dimensionless coordinates. The dimensionless variables are defined as

V V= ~R2 , 0p

X

Y

x=-~-,

y=-~,

/t0Z

ZD Z=R40p

(3 /

p0Z

where P is the pressure. For a constant wall concentration boundary condition, c is given b y C -C w Co - C w

(4a)

where Co is the inlet fluid concentration and Cw the specified wall concentration. For a constant wall flux boundary condition, c is given by: C -C O c=~

(4b)

J~R/D

where Jw is the specified wall mass flux. The boundary conditions for Eq. (1) and (2) are: c-1 v = 0,

or

c=0

c = 0,

(5)

z=0, f o r a l l x , y or

ac/an = 1 along fiber surface

19

(6)


=q=, cl, =cl=

(7a) (7b)

1

2

1

2

where n is the unit outward normal direction along the boundary of the solution domain. The first of the two values for the concentration boundary condition given in (5) and (6) is for the constant wall concentration case while the second is for the constant wall mass flux case. The periodicity of the unit cell is captured by Eq. (7). The velocity and concentration are equal along a pair of periodic boundaries (the subscripts 1 and 2 refer to a pair of periodic boundaries as indicated in Figure 5) while the gradients of velocity and concentration are equal in magnitude but opposite in sign because of the difference in unit outward normal directions for the two boundaries. The axial velocity field, v, is independent of axial distance, z, because of the well-developed velocity field and low mass transfer rate assumptions. Consequently, one can solve for v, substitute it into Eq. (2), and avoid solving the two equations simultaneously. Since an analytical solution cannot be found, a numerical approximation to the solution for v is obtained using the Finite Element method as described in the next section. If shell flow rate changes significantly due to flow across the fiber wall, one may still use the theoretical approach described here but must solve the conservation of momentum equation simultaneously with the conservation of mass equation - this greatly increases the computational complexity. Solving the conservation of mass equation, Eq. (2), for the concentration field allows one to calculate mass transfer coefficients. The effective, overall mass transfer coefficient, k, is given by: N

Z I (Cb -cw),Rao k(2n'RN ) i=1 2xRN

N

= ~, I2x D ( O C / O r ) ~ , i R d O i=1

(8)

where N is the number of fibers in the unit cell, Cb the mixing cup average bulk concentration given by: CVdA Cb = A

(9)

IvdA

A

and A the area available for flow. Eq. (8) can be written in terms of a local Sherwood number as: 13

Mass TransferIn Axial Flows ThroughRandomlyPacked Fiber

Bundles - L i p s c o m b

N

2Rk

Sh = ~ =

~,f2o~(a¢/a¢)w,iao

2 i=l

D

iv y, f2~r(c b _ c w l i d O

(10)

i=1

where ~ is the dimensionless radial coordinate (r/R). For a constant wall concentration boundary condition, Eq. (10) reduces to:

Sh e = 2 * 0¢/0.....~ Co

(11)

while for a constant wall mass flux boundary condition it reduces to: 2 Shf = (cb _cw )

(12)

where the overbar indicates the average value along the surface of each fiber in the unit cell. Values for the Sherwood number will be computed as a function of the Graetz number: (13)

Gz = Q/DL = Re*Sc*(xR/2L)*(1/~- 1)

where Q is the flow rate per fiber. In the large Graetz number limit (either high flow rates or short mass transfer regions), concentration boundary layers between adjacent fibers are much smaller than the fiber spacing. This limit is referred to as the entry mass transfer limit. In the small Graetz number limit (either low flow rates or long mass transfer regions), concentration boundary layers between adjacent fibers completely overlap. This limit is referred to as the well-developed mass transfer limit. 4.

NUMERICAL APPROXIMATIONS The Finite Element method [17] is used to obtain a numerical approximation to the governing conservation equations. To discretize the shell domain, V oronoi tessellation is used to surround each fiber in a unit cell with a polygon. The external boundary of these regions forms the boundary of the unit cell as illustrated in Figure 6. The region within each polygon is further discretized to form the finite element mesh. Note that the unit cell boundary consists of a series of line segments, instead of the four sides of the square unit cell in Figure 5, thus the periodic boundary consists of a sequence of line 14


Lipscomb

segment pairs (see Figure 6). The use of this boundary facilitates the calculations, because one does not have to identify where fibers cross the boundary. The velocity field is calculated first from the Finite Element discretization of Eq. (1) and substituted into Eq. (2). Although the velocity field is independent of axial position, z, the concentration field is not and the Finite Element discretization of Eq. (2) leads to the following set of order differential equations: M -dc - + Kc = F dz

(14)

where c is now a vector containing the concentration at each node in the Finite Element mesh, M and K are coefficient matrices and F is a coefficient vector. lb

1~ \''~ S.

Figure 6. Use of Voronoi tessellation to generate the boundary of the unit cell and discretize the shell domain. Two pairs of periodic boundaries are indicated: (la, 2a) and (lb, 2b). A small portion of the actual mesh is also shown. These equations are solved by approximating dc/dz with a second-order, backward Finite Difference and marching along the z-direction to calculate the concentration field from the start to the end of the mass transfer region. A variable axial step size (Az) is used to speed up the calculations - the smallest values are used at the start of the mass transfer region where large concentration gradients exist. The calculations are stopped when the difference between the last two mass transfer coefficients is less than 0.001 in magnitude (approximately 0.1% change) or the bulk concentration is less than 10-6. One can avoid this computationally intensive solution for the concentration field as a function of z in the entry mass transfer limit. For large Graetz numbers, the relationship between Sh and Gz is given by [18]:

15


Lipscomb

(15)

Sh = a G z 1/3

where the constant a depends on ~b and the wall boundary condition. As demonstrated previously [19], a can be calculated from the velocity field alone for both constant wall concentration and mass flux boundary conditions, so one does not have to obtain a numerical approximation to Eq. (2). In the welldeveloped mass transfer limit, the relationship between Sh and Gz is of the form

[2o]: (16)

Sh = fl

where the constant fl also depends on ¢~and the wall boundary condition. As demonstrated previously [21], flf (the subscript f indicates a constant wall mass flux boundary condition) can be calculated from that portion of the welldeveloped concentration field that is independent of z. One must obtain a numerical approximation to the well-developed portion of c, but marching along the z-axis is not required. Unfortunately, a similar solution does not exist for tic (the subscript c indicates a constant wall concentration boundary condition) and one must solve for the concentration field as a function of x, y, and z to evaluate Values for the Sherwood number calculated using each of these approaches are summarized in the next section. Note that one may use the marching solution procedure to obtain values for a and fl for both boundary conditions. The results obtained by marching are in excellent agreement with the results for ac, o?, and flfobtamed using the simplifications just described. Mesh refinement analyses are performed to evaluate discretization error. The number of nodes is increased by 50% until the calculated mass transfer coefficients change by less than 5% over the range of Graetz numbers considered. Additionally, mass transfer coefficients are calculated for regular arrays to validate the computational approach. 5.

RESULTS AND DISCUSSION Figure 7 illustrates the calculated variation of the local Sherwood number with Graetz number for triangular packings and a constant wall concentration boundary condition using the marching solution approach. The results are compared to those reported by Miyatake and Iwashita [22] for the equivalent heat transfer problem. The agreement between the two is very good and is comparable to that observed for square packings. We believe the minor differences that exist are due primarily to differences between Miyatake and Iwashita's actual results and the correlation they propose to represent the results; the correlation was used to generate Figure 7. 46

Mass Transfer In Axial Flows Through Randomly Packed Fiber Bundles Lipscomb -

Table 2 contains the values of a and fl calculated for triangular and square packings as a function of packing fraction and wall boundary condition. The computational approach used to evaluate each is indicated. Miyatake and Iwashita's results [22, 23] are also given; note that these values correspond to specific results they report and are not generated from the correlation they propose. The agreement between the two is excellent which helps validate the work summarized here. .

.

.

.

.

.

.

.

i

.

.

.

.

.

.

.

.

i

.

.

.

.

.

.

.

.

,

.

.

.

.

.

.

.

.

,

.

.

.

.

.

.

.

.

0.75 102

O.e3 0.40 0.23

Sh 10

.......

o

101

10 2

10 3

10 4

10 s

Gz

Figure 7. Variation of local Sherwood number with Graetz number for triangular fiber packings and constant wall concentration boundary condition. The solid lines are the results obtained using the marching algorithm described here and the dashed lines are the correlation proposed by Miyatake and Iwashita [22]. The values of ~ are indicated on the curves.

In-between the entry and well-developed mass transfer limits, the following expression provides a simple correlation of the relationship between Sh and Gz [24]:

Sh = (fl3 + ot3Gz~/3

(17)

Figure 8 compares the Sherwood numbers generated by this correlation and the actual calculated values for regular triangular arrays and a constant wall concentration boundary condition. Differences between the two are small except for the highest fiber packing, ~b= 0.75, which is outside the range of commercial interest; note that fiber packing ranges from 0.4 to 0.6 in most commercial modules. Consequently, mass transfer coefficients calculated for random packings will be reported in terms of a and fl and Eq. (17) will be used to calculate Sh. Figure 9 illustrates the differences between velocity fields calculated for a square packing and typical random packing with 0~=0.50. Regions of high flow where fiber packing is lowest are dearly visible in the randomly packed bundle. 17


This flow channeling is undesirable and reduces mass transfer coefficients. Where fibers are dose, only a few contours are visible indicating low velocities and flows. Such regions are nearly stagnant and also undesirable. Table 2. Values of a and fl calculated for triangular and square packings. Values for a~ and af were taken from [ 19]. Values for/~ were taken from [21]. Values for tic were obtained using the marching algorithm described in the text. Triangular army - Constant wall concentration boundary condition ot~ This Work Miyatake-Iwashita [22] This Work 0.823 3.15 3.15 9.49 0.750 2.61 2.56 9.94 0.630 1.93 1.88 9.89 0.403 1.12 1.11 6.92 0.227 0.70 0.71 4.21 0.057 0.31 0.32 1.98

Miyatake-Iwashita [22] 9.52 9.97 9.89 6.86 4.17 1.96

Triangular army - Constant wall mass flux boundary condition a? This Work Miyatake-Iwashita [23] This Work 0.823 3.74 3.64 4.89 0.750 3.14 3.02 8.78 0.630 2.33 2.28 11.8 0.403 1.35 1.39 7.59 0.227 0.84 0.87 4.48 0.057 0.37 0.37 2.05


Square array - Constant wall concentration boundary condition o~ This Work Miyatake-Iwashita [22] This Work 0.712 1.80 1.81 4.05 0.649 1.67 1.68 4.13 0.545 1.41 1.41 4.27 0.349 0.95 0.94 4.19 0.196 0.62 0.62 3.34 0.049 0.28 0.28 1.81


Square array - Constant wall mass flux boundary condition o? ¢~ This Work Miyatake-Iwashita [23] This Work 0.712 2.07 2.06 2.29 0.649 1.97 1.92 3.12 0.545 1.70 1.63 4.42 0.349 1.14 1.14 4.98 0.196 0.75 0.77 3.68 0.049 0.34 0.35 1.89


18

Mass Transfer In Axial Flows

Through

Randomly

Packed

Fiber Bundles

Lipscomb

-

Figure 10 illustrates the concentration field in the well-developed limit for the same randomly packed unit cell as shown in Figure 9. The concentration is nearly zero everywhere except in the lowest fiber packing region which corresponds to the highest flow region in Figure 9. Fluid residence time is lower in the low packing regions than in the high packing regions resulting in smaller concentration changes - one may envision a low packing region as a channel that allows fluid to "by-pass" the mass transfer region. The low packing regions dominate overall mass transfer performance because flow rates are much higher through them than high packing regions. ........

,

........

|

........

|

........

|

........ 0.75

102 0.63 0.40

0.23

Sh 101

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

........

i

104

.......

105

Gz

Figure 8. Comparison of the proposed local Sherwood number correlation (dashed lines) with calculated results (solid lines) for triangular packings and constant wall concentration boundary condition. The values of # are indicated on the curves.

F4 (a) (b) Figure 9. Contours of constant velocity in (a) regular square and (b) random fiber packings.

19

Mass Transfer In Axial Hows Through Randomly Packed Fiber Bundles - Lipscomb

Figure 10: Contours of constant concentration in a randomly packed bundle in the welldeveloped mass transfer limit for a constant wall concentration boundary condition. Table 3. Values of a and fl calculated for random packings. Random array - Constant wall concentration boundary condition a~ [19] Fiber Min. Max. Ave. Fiber Number Number 0.30 9 0.580 0.709 0.643 32 0.40 9 0.799 0.889 0.860 32 0.50 9 0.971 1.19 1.08 9 0.50 16 1.01 1.21 1.08 16 0.50 32 32 0.50 64 64 0.50 100 1.12 1.15 1.14 100 0.55 9 1.09 1.33 1.22 32 0.575 9 1.17 1.46 1.32 32 Random array - Constant wall mass flux af [19] Fiber Min. Max. Number 0.30 9 0.604 0.808 0.40 9 0.50 9 1.08 1.43 0.50 16 1.09 1.31 0.50 32 0.50 64 0.50 100 1.23 1.28 0.55 9 1.28 1.40 0.575 9 1.11 1.53

Min.

Max.

Ave.

0.484 0.480 1.11 0.914 0.694 0.758 0.676 0.713 0.615

0.840

0.623 0.918 1.71

1.23 2.83 2.08 1.27

1.27

1.19

0.968 0.987 0.800 0.928 0.904

]~ [21] Max. Min.

Ave.

0.185 0.329 0.225 0.153 0.165 0.142 0.175 0.154 0.098

0.213 0.355 0.460 0.342 0.331 0.247 0.252 0.197 0.144

1.20 0.908

1.20

boundary condition Ave.

Fiber Number 64 64 9 16 32 64 100 64 64

0.742 1.25 1.20 1.27 1.34 1.35

20

0.275 0.387 0.657 0.504 0.452 0.373 0.415 0.256 0.199


Values calculated for a and fl for randomly packed bundles are summarized in Table 3. The maximum, minimum, and average values for at least five different unit cells are given. Table 3 also indicates the effect of unit cell fiber number. For 50% fiber packing, a and fl were calculated for unit cells containing 9, 16, 32, 64, or 100 fibers. An analysis of the variance (ANOVA [25]) indicates there is no significant difference between the results at the 95% confidence level if a sufficiently large number of fibers is used: at least 9 for c~c and o~, 32 for tic, and 64 for fl~ The results of this analysis were used to specify the fiber number for calculations with other ~ values. A comparison of Tables 2 and 3 indicates the detrimental effect of random fiber packing on mass transfer performance. Relative to a square packing, the random packing values for ac and af are-~20% lower. The decrease in the welldeveloped limit is much more dramatic: tic is 80% lower while flf is 90% lower. This detrimental effect is graphically illustrated in Figure 11 for ~=0.40 where the rapid decline in performance with decreasing Gz is readily apparent. 102

....................................

• .......

101

J J

Sh

J J J 10 0 ___.

1 0110_ 1

J

__-.-

10 °

101

10 2

10 3

04

GZ

Figure 11. Variation of local Sherwood number with Graetz number for regular square (solid line) and random (dashed line) fiber packings with ~=0.40 and constant wall concentration boundary condition. The following correlations accurately capture the dependence of a and fl on fiber packing fraction for randomly packed bundles - for a constant wall concentration boundary condition:

(18a) ( 8b)

ac = 2.44~- 0.097 tic = - 1 0 . 2 g + 9.92~- 1.43 and for a constant wall mass flux bounda~ condition:

21

Mass Transfer In Axial

Flows T h r o u g h

R a n d o m l y Packed Fiber Bundles -

Lipscomb

(19a)

off= 2.30~ + 0.061 f l f = - 8 . 1 5 g + 6 . 8 2 ~ - 1.09

(19b)

Substituting Eq. (18) and (19) into Eq. (17) gives a correlation that can be used to predict local shell-side mass transfer coefficients for randomly packed bundles as a function of G z and ~k (0.30< ~ 85 mL, consistent with the dominance of the cake filtration mechanism at long times. The slope for the composite membrane beings around 6 and decreases monotonically during the start of the filtration process. This reduction in slope reflects the shunting of the fluid flow towards the open regions of the membrane. There is again a sharp decrease m slope around V = 100 mL due to the transition between a pore blockage dominated fouling to a cake filtration mechanism. The initial slope for the isotropie PVDF membrane is around 9 and decreases in value throughout the filtration run. Similar extreme values of the slope have also been reported previously [7,8], but they never have been explained fundamentally. Our new theoretical framework provides a good qualitative prediction of the observed behavior,

40

Fouling Phenomena Dunng ~.~crofiitration: Effects Of Pore Blockage, Cake Filtration, And Membrane Morphology -Zydney

10

,~~,

8 c:

6

a-

4

.=-

2

O

0

..O

.,--.,.

u,.,

DF

O

O PCTE r l Pvog ~ PCTE/PVDF

Z~ A

%~'~A E]

-2

PCTE/PVDF

,

0

O

E!

-4.

-8

~,n ,-,

PCTEt~:~ #~)O O

II--

O

,.~ '

I

50

,

,

100

I

150

,

I

200

n ,

250

Cumulative FiltrateVolume, V (mL) Figure 6: Slope determined from the plot of d2t/dV 2 versus dt/dV for the PCTE, PVDF, and composite PCTE/PVDF membranes. Experimental data are from Figure 3. Solid curves are model calculations. with the quantitative discrepancies seen in Figure 6 arising from the multiple derivatives that have to be taken to evaluate the slope. Thus, there is no need to invoke any new or complex fouling phenomena to explain these experimental observations. The key is the proper inclusion of both the pore blockage and cake filtration behavior and the proper analysis of the flow properties of membranes with different underlying pore morphologies. Although the experimental data presented in Figures 3 - 6 were all obtained with bovine serum albumin, the pore blockage- cake filtration model can also explain the flux decline behavior seen in a wide variety of filtration systems. For example, Figure 7 shows results for the constant pressure filtration of BSA, lysozyme (a 14 kD molecular weight protein), and a soil-based humic acid through the 0.2 gm pore size polycarbonate track-etched membranes Additional details on the protein and humic acid experiments are provided elsewhere [14,18], respectively. In each case, the derivatives were evaluated numerically directly from data for the filtrate flux as a function of the cumulative filtrate volume. The solid curves are calculations using the simple approximate solution (Eq. 6) with the best fit values of the parameters determined by comparison of the model and data. Jbackwas set equal to zero for both BSA and lysozyme, while the best fit value of Jbackwas 5.3 x 104 m s1 for

41

Fouling Phenomena During Mierofiltration: Effects Of'Pore Blockage, Cake Filtration, And Membrane Morphology - Zydney

lo 3

A ,,f,

.

.

.

.

.

.

.

.

i

.

.

.

.

.

.

.

.

i

.

.

.

.

.

.

.

.

,~

1012

E m

"o ol

"o 1011

LJ 10lo 10 6

113_ t.yso me

~

..~.. Hum.ic~.Ac.id

10 7

10 8

10 9

dt/dV (s m "3) Figure 7: Flux decline analysis for the constant pressure filtration of solutions of BSA (2 g/L), lysozyme (2 g/L), and a soil-based humic acid (2 ppm) through polycarbonate track-etch membranes. Solid curves are model calculations.

the 2 ppm humic acid. The initial slope for all three systems was approximately equal to two, which is again consistent with the dominance of the pore blockage phenomenon at the start of the filtration. The flux decline data for the humic acid solution show a distinct maximum in d2t/dV2 reflecting the transition between the pore blockage and cake filtration mechanisms. At long filtration times, dZt/dV2 for the humic acid solution approaches zero as the flux approaches its steady-state value J = Jb,ck. This sharp decline in d2t/dV2 provides a convenient diagnostic for determining the presence of a steady-state flux in any given filtration process. The lysozyme data show a much smoother transition from the pore blockage to cake filtration mechanisms, with no distinct maximum observed in the dZt/dV2 data. This occurs because of the greater rate of cake growth relative to pore blockage for the lysozyme [ 18]. 4.

SUMMARY

Membrane fouling remains a major problem in applications of microfiltration for particle removal, sterile filtration, and clarification. The theoretical framework presented in this chapter provides a completely new approach for the analysis and interpretation of flux decline data obtained in these membrane systems. The model explicitly accounts for fouling due to surface (pore) blockage and cake filtration, with the cake forming over those

42

Fouling Phenomena During Microfiltration: Effects Of Pore Blockage, Cake Filtration, And Membrane Morphology - Zydney

regions of the membrane that have first been covered by large aggregates. Just as importantly, this new theoretical framework provides the first quantitative analysis of the effects of the pore morphology on the rate of flux decline. Membranes with straight-through pores show the greatest rate of flux decline since the particles completely cover (block) the non-interconnected pores in these membranes. In contrast, particle deposition on the surface of a homogeneous membrane with highly interconnected pores will disturb the filtrate flow only over a relatively small penetration distance into the membrane pore structure. This implies that very thick membranes should exhibit a slower rate of flux decline due to the smaller relative disturbance in the flow, an effect that has been confirmed experimentally [ 16,17]. The flux decline for composite membrane structures shows a more complex behavior. The surface fouling completely blocks the pores in the upper skin layer, but the fluid rapidly redistributes itself through the very highly interconnected pores within the membrane substructure. This causes a shunting of the fluid flow towards the remaining open pores, leading to a reduction in the rate of flux decline compared to that for a single membrane layer. This theoretical analysis also provides a framework that can be used for the design and development of new membrane structures having reduced rates of fouling. This would include the proper choice of pore connectivity, membrane thickness, and the specific properties of the individual layers in more complex composite structures. These phenomena have often been neglected in membrane design/development, although they can clearly have a dramatic effect on the flow distribution, fouling, and overall selectivity for the particular separation. REFERENCES [1]

P. H. Hermans and H. L. Bred6e, "Principles of the Mathematical Treatment of Constant-Pressure Filtration," J. Soc. Chem. Intl. 55T (1936) 1.

[21

V. E. Gonsalves, "A Critical Investigation on the Viscose Filtration Process," Ree. Trav. Chim. Des Pays-Bas, 69 (1950) 873.

[3]

H. P. Grace, "Structure and Peformance of Filter Media," AIChE J., 2 (1956) 307.

[4]

J. Hermia, "Constant Pressure Blocking Filtration Laws - Application to Power Law Non-Newtonian Fluids," Trans. Inst. Chem. Eng., 60 (1982) 183.

[5]

K. Luckert, "Model Selection Based on Analysis of Residue Dispersion Using SolidLiquid Filtration as an Example," Int. Chem. Eng., 34 (1994) 213.

[6]

W. R. Bowen, J. I. Calvo, and A. Hemandez, "Steps of Membrane Blocking in Flux

Decline During Microfiltration," d. Membrane Sei., 101 (1995) 153. [7]

E. Iritani, Y. Mukai, Y. Tanaka, and T. Murase, "Flux Decline Behavior in Dead-End Microfiltration of Protein Solutions," J. Membrane Sci., 103 (1995) 181.

43

Fouling Phenomena During Microfiltration: Effects Of Pore Blockage, Cake Filtration, And Membrane Morphology - Zydney

[8]

J-Y. Wang, K-S. Chou, and C-J. Lee, "Dead-End Filtration of Solid Suspension in Polymer Fluid through an Active Kaolin Dynamic Membrane," Sep. Sei. Tech., 33 (1998) 2513.

[9]

S. T. Kelly, W. S. Opong, and A. L. Zydney, "The Influence of Protein Aggregates on the Fouling of Microfiltration Membranes During Stirred Cell Filtration," J. Membrane Sei., 80 (1993) 175.

[10]

W. Yuan and A. L. Zydney, "Humic Acid Fouling During Microfiltration," J. Membrane Sci., 157 (1999) 1.

[11]

G. Belfort, R. H. Davis, and A. L. Zydney, "The Behavior of Suspensions and Macromolecular Solutions in Crossflow Microfiltration," J. Membrane Sei., 96 (1994) 1.

[12]

C. C. Ho and A. L. Zydney, "A Combined Pore Blockage and Cake Filtration Model for Protein Fouling during Microfiltration." J. Coll. Interf. Sei. 232 (2000) 389.

[13]

C. C. Ho and A. L. Zydney, "Measurement of Membrane Pore Interconnectivity," J. Membrane Sci., 170 (2000) 101.

[141

W. Yuan, A. Kocic, and A. L. Zydney, "Analysis of Humic Acid Fouling using a Combined Pore Blockage - Cake Filtration Model," J. Membrane Sei. (in press 2002).

[15]

C. C. Ho and A. L. Zydney, "Protein Fouling of Asymmetric and Composite Microfiltration Membranes," Ind. Eng. Chem. Res., 40 (2001) 1412.

[16]

C. C. Ho and A. L. Zydney, "Effect of Membrane Morphology on Protein Fouling During Microfiltration," d. Membrane Sei., 155 (1999) 261.

[17]

C. C. Ho and A. L. Zydney, "Theoretical Analysis of the Effect of Membrane Morphology on Fouling during Microfiltration." Sep. gel. Teeh. 34 (1999) 2461.

[18]

L. Palacio, C. C. Ho, and A. L. Zydney, "Application of a Pore Blockage - Cake Filtration Model to Protein Fouling during Microfiltration," Bioteeh. Bioen~ (submitted 2002).

44


Chapter 3

Differential scanning calorimetry and rheological experiments to study membrane formation via thermallyinduced phase-separation P.C. van der Heijden, M.H.V. Mulder, M. Wessling University of Twente, PO Box 217, 7500 AE Enschede, The Netherlands, Fax: xx31-53-4894611 1.

INTRODUCTION

1.1

Thermally-Induced Phase-Separation Porous polymer membranes can be prepared by different techniques: track-etching, stretching and phase-separation [ 1]. The phase-separation method is very common to prepare polymeric membranes and much attention has been paid to prepare membranes from various polymers and diluents [1,2]. The preparation of a porous structure consists of two steps. First, the homogeneous polymer solution has to undergo liquid-liquid demixing to obtain a polymer-rich continuous matrix and a dispersed polymer-lean phase of almost pure solvent. The second step is a fixation step to give the structure mechanical stability. Liquid-liquid demixing takes place when the solvent power is not sufficient anymore to keep the polymer dissolved. Two major methods can be distinguished to decrease the solvent power, either by adding a non-solvent which is called diffusion-induced phase-separation (DIPS) or by changing the temperature, defined as thermally-induced phase-separation (TIPS). The latter method is studied in this paper. Fig. 1 shows a schematic representation of the formation process of a porous structure with the TIPS method. The dashed arrow represents a typical cooling trajectory. When passing the binodal, the homogenous polymer solution liquid-liquid demixes in a polymer-rich phase with a polymer concentration given by the binodal and a polymer-lean phase of almost pure diluent. After liquid-liquid demixing of the binary polymer-diluent system, the structure has to be fixed. This can be achieved by crystallization, vitrification (glass transition), or gelation of the polymer-rich phase. After the structure fixation, it is possible to remove the polymer-lean phase by evaporation or extraction. An advantage of the TIPS method over the DIPS method is the ability to form porous structures with polymers which do not form a solution at room temperature, for example polyethylene and polypropylene.

45

Differential Scanning Calorimetry And Rheological Experiments To Study Membrane Formation Via Thermally-Induced Phase-Separation Van Der Heijden

,soo

c

H I binodal

. _ _

•

400 1

E(L) 300 -F-

"~pi~n°dalT

200

' 0.0

I 0.2

S

'

I 0.4

i

I 0.6

i

I 0.8

i 1.0

Volume fraction polymer (-) Figure 1. Example of a phase diagram of a binary polymer- diluent solution. The black arrow represents a cooling trajectory to obtain a porous structure with the TIPS method. It goes from the homogeneous solution (I-I) via the liquid-liquid demixing gap (L-L) into the structure fixation area (S). ~c is the critical concentration of the solution. The porous structures obtained with TIPS can be used for more applications than membranes, for example: separators m electrochemical cells, synthetic leather, 'breathable' rainwear, diapers, surgical dressings and bandages [3], inertial confmed fusion targets [4,5], biodegradable implants for cell transplantation [6], and scaffolds for tissue engineering [7]. Typical cell sizes obtained with this method are in the order of micrometers.

1.2

Historical Background

The formation of porous structures via TIPS was firstly reported by Castro [8]. In this patent, he investigated hundreds of polymer-diluent systems, both amorphous and crystalline polymers, on their ability to form a porous structure. Later on, Shipman [9] described a procedure to prepare porous structures with a certain shape, and Josefiak et al. [10] introduced a second liquid prior to the cooling step to adjust the morphology of the porous structure. The first papers on the TIPS method were published in 1984 [ 11 ]. Since then, the interest in the formation of porous structures of various polymerdiluent systems with the TIPS method has increased rapidly. Table 1 gives a

46


summary of crystalline polymer-diluent systems used in the TIPS method and Table 2 gives an overview of the formation of amorphous porous structures via TIPS. Besides the studies on new polymer-diluent systems, studies were performed on variations of the TIPS method by introducing, for example, an evaporation step to obtain an asymmetric structure [ 12-16].

Table 1. Crystalline polymer-diluent systems which have been used to form a porous structure with TIPS. Polymer Diluent Isotactic polypropylene (iPP) Diphenylether [ 17-19] n,n-Bis(2-hydroxylethyl)tallowamine [11,20-231 Polyvinylidene fluoride Cyclohexanone [ 11 ] (PVDF) Butrolactone Propyl enecarbonate Carbitolacetate Isotactic polystyrene (iPS) Nitrobenzene [24] Cyclohexane [25] Dioxane / isopropanol Poly(tetrafluorethylene-co-per- Chlorotrifluorethylene [26] fluoro-(propyl vinylether)) (Teflon®PFA, NeoflonT~FA) Poly(2,6-dimethyl- 1,4Cyclohexanol [ 12] phenylene ether) (PPE) Nylon 12 Poly(ethylene~lycol) [271 Dioxane / water [28] Poly(7-benzyl-l-glutamate) Benzene (PBIG) Dichloroethane Polyethylene (HDPE) Diphenylether [29] 1,2-Ditridecylphtalate / hexadecane[30] Polysilastyrene (PSS) Cyclohexane [31 ] Benzene 4-Methyl- 1-pentene (PMP) Diisopropylbenzene [5] Polylactide 1,4-Dioxane [6] As mentioned before, the formation of a porous structure with the TIPS method consists of two steps: liquid-liquid demixing to obtain a porous structure and the fixation step to obtain mechanical stability. Both steps have been studied separately for quite a time. The first papers on liquid-liquid demixing of polymer solutions were already published in 1950's (for references see [43,44]).

47


The existence of crystalline polymers has been known from the mid 1940's (a historic overview can be found m [45]) and the vitrification temperature depression (or glass transition temperature depression), for example, was already quantified in 1961 [46]. Table 2. Amorphous polymer-diluent systems which have been used to form a porous structure with TIPS. Polymer Diluent Atactic poly(methylmethacrylate) Cyclohexanol [32,33] (aPMMA) Tetramethylene sulfone (sulfolane) [34,35] 1-Butanol [32] Tert-butylalcohol [36] Atactic polystyrene (aPS) Cyclohexane [37] Diethylmalonate Cyclohexanol [38-41] Trans-decahy dronaphtalene (decalin) [42]

Liquid-liquid demixing of polymer solutions is extensively studied since the last half of the twentieth century (for references see [43,44]). Phase diagrams of many polymer-diluent systems have been determined visually or with other optical techniques such as optical microscopy and light scattering [21,23,27,28,35,38,43]. These techniques are also used to follow the time dependency of the demixing process to study the kinetics of demixing [17,27,33,47-50]. Also other experimental techniques have been occasionally used to compose phase diagrams or to follow the demixing process like viscometry [51 ], NMR [34], X-ray scattering [52] and DSC [32,42,53,54]. 1.3

Aim of this Paper In this paper, DSC is used to study liquid-liquid demixing and the structure fixation step of an amorphous polymer solution. In addition, rheological experiments are carried out to study liquid-liquid demixing as well. Using DSC and theology, new insights may be obtained in the membrane formation process. However, practical reasons can be mentioned to use these techniques instead of the frequently used light based techniques. For example, with light based techniques the refractive index between both the diluent and polymer has to differ significantly, and transparency of the experimental setup is required to observe a signal. This is not a problem with DSC and rheological experiments. Furthermore, with DSC and rheology it is possible to detect both the liquid-liquid demixing step and the structure fixation step in one experiment. 48


With optical microscopy for example, liquid-liquid demixing can be observed but it is impossible to observe the glass transition. Finally, high temperatures are involved in the TIPS process and to exclude evaporation of diluent, closed sample holders are necessary to use, and this is very easy to achieve with DSC. Q

DIFFERENTIAL SCANNING CALORIMETRY [55]

2.1

Introduction With DSC, the heat flow is measured as a function of time or temperature to observe physical transitions. After the introduction of the first differential scanning calorimetry (DSC) apparatus polymers have been studied extensively. See Refs. [56,57] for a historical overview about the development of thermal analysis devices in general and for polymers in particular. DSC is a well-known technique to study solid-liquid demixing (crystallization) [20,23,24,29] and vitrification [32,38,54] of liquid-liquid demixed polymer solutions. Berghmans and co-workers [32,42,53,54] used DSC as well to determine the liquid-liquid demixing temperature. They carried out DSC experiments for the systems atactic polystyrene / decalin and atactic polymethylmethacrylate / 1-butanol and cyclohexanol respectively. Upon cooling (cooling rate 5 K.min-1), an exothermic heat flow shift was observed and the onset was taken as the liquid-liquid demixing temperature. This signal agreed very well with optical observations. With one DSC run they could determine both the liquid-liquid demixing temperature and the glass transition temperature of the polymer-rich phase. However, DSC is hardly used for the determination of liquid-liquid demixing because the heat effect is very small and disappears easily in the base line drift. Recently, a rather new technique has been developed: temperature modulated differential scanning calorimetry (TMDSC) [58,59]. In spite of some discussion about the interpretation of the measured signals [60-67], TMDSC is a very useful device to measure small heat effects and to separate overlapping thermal events. In the next sections, TMDSC experiments are used to examine the cooling trajectory of binary polymer solutions in the concentrated region (at the right side of the critical point in Fig.l). The experimental work described in this work is carried out with the polymer-diluent system atactic polystyrene in 1dodecanol. A number of reasons can be mentioned to choose this system. This system forms a porous polymer structure with the TIPS method at room temperature [8], 1-dodecanol is a non-toxic solvent, polystyrene has solvents at room temperature and therefore easy to use, and many physical data are available for both components.

49


2.2

Temperature Modulated Differential Scanning Calorimetry

An extensive description of this technique can be found in Refs. [58,62,68]. Here, only a short description is given. With TMDSC, a second function (for example a sine wave) is superimposed onto the conventional linear or isothermal temperature ramps (see Fig. 2).

200

-

0 o v

"~ 1 9 5 O.

E I--

"

190 I

0

'

I

1

'

I

'

2

I

3

'

I

4

'

5

Time (min)

Figure 2: Linear cooling trajectory (thick black line) with superimposed temperature modulations.

The temperature ramp can then be described as: (1)

T = To + bt + A sin cot

where To is the initial temperature, b the underlying scanning rate, A the temperature amplitude, co the angular frequency, and t the time. The TAinstruments user manual [68] suggests a value of the underlying scanning rate (b) below 5 K.min "1 and typical values for A and co are 1 K and 60 s. The resulting heat flow consists of two contributions. The first part is caused by rapid processes and is proportional to the scanning rate. The second part is caused by kinetically hindered or irreversible processes and hence independent of the scanning rate: dQ dt

= Cp

dT -~

(2)

+ f (t,T) "

50


dQ/dt is the heat flow, Cp the heat capacity, and fit, T) a contribution to kinetic effects. The resulting heat flow is a wave function as well. With the help of discrete Fourier transformation software which resides in the DSC module, the heat flow signal is deconvoluted in a cyclic signal and an underlying signal (which is equivalent to the conventional DSC). In this work the modulus of the complex heat capacity is used to present the TMDSC results. This quantity is calculated from only the amplitude of both the temperature and the heat flow modulation. It is thus completely derived from the cyclic signal. Physical processes, which take place at time scales smaller or comparable to the modulation period, contribute to this complex heat capacity. Examples of such physical processes are vibrations and rotations of atoms. Slow physical processes, like the mobility of vitrified polymers with time scales much larger than a modulation period does not contribute to the complex heat capacity.

(Icp*l)

(IcjI)

2.3

Experimental

2.3.1 Materials Atactic polystyrene (aPS) (Styron* 686E) was kindly supplied by Dow Benelux NV (Mw and Mw/M,: 2.3.105 g.mol 1 and 2.1 respectively, determined with GPC). The diluent, 1-dodecanol (purity > 98%, Merck-Schuchardt), was used without further purification.

2.3.2 Sample preparation A homogeneous solution of aPS and 1-dodecanol was prepared in a threeneck bottle under nitrogen at 200°C. 1-Dodecanol vapor was allowed to evaporate during stirring with a mechanical stirrer. Small amounts of various polymer concentrations were poured in Petri-dishes and cooled in air. The compositions of the samples were determined by thermogravimetric analysis. About 20 mg of the sample was inserted in a platinum sample pan of a TGA 2950 Thermogravic Analyzer of TA Instruments and heated up to 200°C with a heating rate of 10 K-min-1. Afterwards the temperature was kept constant at 200°C for maximum 2 h to evaporate all the 1-dodecanol. From the ultimate weight loss the polymer concentration was determined.

2.3.3 Temperature (TMDSC)

Modulated Differential

Scanning Calorimetry

The TMDSC used is a DSC 2920 of TA Instruments. Calibration with indium and high density polyethylene (HDPE) (for calibration of the heat capacity) was carried out. About 5 mg of the sample was put in the aluminum closed sample pan. The TMDSC was heated to 200°C and kept isothermally for

51


30 min to ensure homogeneity. The cooling rate was set to 2 K-min 1 to 0°C and after an isothermal step of 5 min the sample was heated again with 2 K.min -1. The amplitude of the superimposed sine wave was 1 K with a period of 60 s [68]. The glass transition temperature Tg and the liquid-liquid demixing temperature TL_Las well as the heat capacity shift at TL_Lwas determined with the TA Universal Analysis software.

2.3.4 Optical Microscopy (OM) To compare the TMDSC results with a well-known technique for studying liquid-liquid demixing, optical microscopy experiments are carried out. The polymer sample was placed on an object glass within an aluminum ring (thickness 0.1 mm, inner diameter 5 mm) and covered by a second glass. To prevent diluent loss caused by evaporation, laboratory grease was used to stick the aluminum spacer to the object glasses [23]. The sample was placed in a hot stage (Linkam THMS 600) which was controlled by the Linkam TMS92 hot stage controller. The sample was heated and cooled with a rate of 2 K.min -1 and demixing was observed visually with an Olympus BH2 microscope (magnification 200x).

2.4

Results

Cooling and subsequent heating curves of aPS - 1-dodecanol are plotted in Fig. 3 for two polymer concentrations (weight fractions of 0.38 and 0.69). The modulus of the complex heat capacity (Icp'l) is plotted at the y-axis. Two transitions can be observed: the glass transition and a small baseline shift at higher temperatures, which is assumed to be the liquid-liquid demixing temperature. In the following, the onset of this signal upon cooling is defined as the liquid-liquid demixing temperature (TL-L) comparable with the observations of Amauts et al. and Vandeweerdt et al. with the conventional DSC [32,42,53,54]. The glass transition temperature (Tg) is chosen as the onset upon cooling because below this temperature influences can be expected of vitrification on the liquid-liquid demixing behavior. The difference in heat capacities ([Cp*l) between the results for the two polymer concentrations can be explained by the heat capacity difference between both single components. Literature values of the heat capacity for pure polystyrene and 1-dodecanol at T = 180°C are 2.1 J.gl.K-1 [69] and 3.1 J.gl.K-1 [70], respectively. Therefore, a polymer solution that contains the higher content of polystyrene should have a smaller heat capacity than a diluted polymer solution. At least, when it is assumed that the heat capacity of a solution of aPS in 1-dodecanol is between the values of both pure components.

52


3.0

'7, '7

--

2.5-ETJ

-..3

"

0

0.

2.0

--

..,.=,_.-

69 wt-%

1.5

;

I

;

40

I

80

:T,, ;

I'

120

'

I

160

'

200

Temperature (°C) Figure 3. Cooling and subsequent heating curves (weight fractions of polymer: 0.38 and 0.69). Gray lines" heating curves, black lines" cooling curves.

With conventional DSC experiments, an optimum has to be found between a high scanning rate which results in a high intensity and a low scanning rate which results in a high resolution of a thermal signal. Consequently, a small thermal transition such as liquid-liquid demixing is very difficult to observe with conventional DSC. With TMDSC, both a low scanning rate (the underlying scanning rate) and higher scanning rates (the temperature modulations) are present within one experiment and therefore it is very suitable to detect liquid-liquid demixing. In the cooling curves the L-L phase transitions at TL_Lare represented by a steep heat capacity shift. The heating curves have the same slopes as the cooling curves only at the liquid-liquid demixing temperatures the transition is not as distinct. This difference is discussed later in this section. In the following, the details of the cooling curves are further used and discussed only. Performing such TMDSC cooling experiments over a large concentration range allows the construction of the phase diagram of the polymer-diluent system. To support the assumption of the base line shift to stem from the L-L demixing, the TMDSC results are compared with optical microscopy (OM) data indicating visually the phenomenon of L-L demixing. In Fig. 4, the TL-Land Tg determined with TMDSC and OM are plotted. The open circles represent the TMDSC liquid-liquid demixing data whereas the filled black squares are the OM data. The closed circles are the glass-transition temperature data points. The liquid-liquid demixing data obtained from TMDSC

53

Differeraial Scanning Calorimetry And Rheological Experiments To Study Membrane Formation Via Thermally-Induced Phase-Separation Van Der Heijden

and OM agree well and therefore it can be concluded that the observed TMDSC signals are indeed caused by liquid-liquid demixing. 200

~,

160

o o v

::}

120

(1.) (:}.

E

80

!--

40

0.0

0.2 Weight

0.4 fraction

0.6

0.8

polymer

1.0 (-)

Figure 4. Phase diagram a P S - 1-dodecanol. Open circles: TMDSC data L-L demixing. Closed squares: OM data L-L demixing (cloud points). Closed circles: TMDSC data glass transition. Lines are drawn to guide the eye.

The shift of the TMDSC curve at the liquid-liquid temperature is assumed to be completely caused by the enthalpy of demixing. The difference in values in the heat capacity shift (Acp* at Tt.t, defined in Fig. 3) between the different concentrations is caused by the interaction between polymer and diluent. This can be quantified by calculating the enthalpy of mixing with the help of the Flory-Huggins theory [54]. To minimize the error in the calculation of the modulus of the complex heat capacity with the TMDSC software, it is recommended that at least 4 complete superimposed cycles fit within a phase transition [68]. This requirement is satisfied for the glass transition because this transition covers a temperature range of at least 10 K. However, in case of liquid-liquid demixing the heat capacity shift only covers a temperature interval of 2 K, so only one modulated cycle fits within this transition. By lowering the underlying cooling rate, the number of cycles within the transition can be increased; the resulting TMDSC curves are shown in Fig. 5. From Fig. 5 it can be concluded that cooling rates of 2 K.min -1 and lower have no significant influence on the measured modulus of the complex heat capacity. As already mentioned the measured complex heat capacity is only measured from the amplitudes of the modulated heat flow and modulated temperature, hence it is not influenced by the cooling rate. However, the shape of the curve can be influenced by time-dependent effects like the growth of the

54


demixing domains and the increasing viscosity of the polymer solution (see section 3) upon cooling. The growth of the demixed domains can be the reason for the difference between the cooling and heating curve (see Fig. 3 for a 38 wt.% polymer solution). In the demixed polymer solution regions of almost pure diluent grow when the temperature of the solution is in between the TL-L and the Tg. The TMDSC heating curve was measured after the cooling trajectory, so, the demixed domains already had a long time to grow. Therefore, it is more difficult for the polymer solution to follow the temperature modulation because of larger distances the diluent molecules must diffuse. ,,

,

,,

0.2 K.min~ 2 . 8

,,

,

,

2 K-min"~

-"

"7 "7

2.4 - -

v

-1

•

..

(3 J

2.0--

~/

5 K.min

1.6 J

50

!

I 100

'

I

'

150

I 200

Temperature (°C) Figure 5. Influence of cooling rate on modulus of the complex heat capacity. Weight fraction of polymer is 0.48.

The shape of the curves at low cooling rates (between 0.2 K.min 1 and 2 K.min1) are comparable, this means that influence of time depending effects play a negligible role in the cooling curves at these cooling rates. 2.5

Conclusions and Outlook With Temperature Modulated DSC, liquid-liquid demixing of polymerdiluent systems can be determined as well as the glass transition of the polymerrich phase in one run at a relatively low scanning rate. Liquid-liquid demixing observed with TMDSC agrees well with visually observed cloud points.

55


Upon quantifying the heat capacity shift at the liquid-liquid demixing temperature additional information with respect to the thermodynamics of the polymer solutions can be obtained. Furthermore, it offers the possibility to follow the heat capacity shift as a function of temperature both theoretically and experimentally to study the kinetics of liquid-liquid demixing. Especially this is interesting in the region of the glass transition temperature in which the structure fixation step can be studied. 3.

VISCOMETRY AND RHEOLOGY

3.1

Introduction

During the formation of porous structures with the TIPS process, the homogenous solution is demixed in two liquid phases and afterwards the polymer-rich phase is vitrified. It is expected that these phase transitions can be observed with rheological experiments. The possibility of studying liquid-liquid demixing with viscometry and rheology experiments is briefly discussed in this section. The experiments are carried out for the polymer-diluent system aPS diisodecylphthalate (DIDP). The reason to study this system is the following. The high vapor pressure of many common diluents in relation to the liquidliquid demixing temperature is often a problem studying the thermally-induced phase-separation method experimentally. Special care has to be taken to prevent evaporation of the diluent during an experiment. The diluent DIDP shows a cloud point with the polymer of about 50°C and has a very low vapor pressure. (For comparison, the vapor pressure of water at T = 20°C is 2.10 3 Pa and the vapor pressure of DIDP at T = 100°C is 0.1 Pa [70].) Therefore, for this system no special care has to be taken to avoid evaporation of diluent.

3.2

Experimental Homogeneous solutions of aPS in DIDP (Merck-Schuchardt, purity>99%) were prepared in the same way as described in section 2.3.2. The viscosity of the polymer diluent system aPS-DIDP was measured with a Brabender Viscotron, Mod. Nr 8024. Starting at high temperatures (in the homogeneous solution), the viscosity was measured for different shear rates at a fixed temperature. Subsequently the temperature was lowered and the complete procedure was repeated. Furthermore, polymer solutions of aPS in DIDP were studied in a Bohlin Rheometer CS50. Upon cooling with a cooling rate of 2 K.min -1, the complex viscosity was measured for different strains at a constant frequency of 0.1 Hz.

56


3.3

Results 200

150 --t~ ¢~

13_ >"100 0 ¢o

>

50-

~'~1 I 20

i

40

I 60

s~ = 80

T e m p e r a t u r e (°C) Figure 6. Viscosity of a 30 wt.% solution of aPS in DIDP as a function of the temperature upon cooling for different shear rates (indicated in the figure). Results of the viscosity measurements are visualized in Fig. 6 for a polymer concentration of 30 wt.%. Experiments in the concentration range of polymer between 20 and 40 wt.% show a comparable trend'as plotted in Fig. 6. The observation of an increasing viscosity upon cooling as the temperature approaches the liquid-liquid demixing temperature is in agreement with the work of Wolf et al. [51]. In fact, the location of the sharp decrease in viscosity (between T = 45 and 50°C) is in good agreement with cloud point experiments of this polymer-diluent system (T = 49°C). The observation of the sudden decrease at lower temperatures inside the liquid-liquid demixing gap is not in agreement with the expectation. Upon cooling a polymer solution, the viscosity increases. At the liquid-liquid demixing temperature, a polymer-rich matrix is formed enclosing the polymerlean phase. The viscosity is expected to be dominated by the viscosity of the continuous, polymer-rich, phase of the system. Therefore, it should be expected that upon liquid-liquid demixing the viscosity increases further. An explanation of the sharp decrease in viscosity could be the loss of contact between the polymer rich matrix and the rotating cone of the viscometer. Upon storage below the cloud point temperature the demixed polymer solution displays synereses; DIDP is expelled out of the continuous phase. This process is probably enhanced by the applied shear rate resulting into the formation of a slip

57

Differential Scanning Calorimetry And Rheological Experiments To Study Membrane Formation Via Thermally-Induced Phase-Separation Van I)er Heijden

layer of DIDP between the polymer rich matrix and the rotating cone. To apply less mechanical force on the polymer solution, an oscillating rheometer was used instead of the rotating viscometer. 2 D1

?

t

Q... >,

.~_ f/)

o o

U)

•~

0.5

•

x

1

Q.,

E o

20

30

40

50

60

70

Temperature (*C) Figure 7. Complex viscosity for a 30 wt.% solution of aPS in DIDP. The frequency is 0.1 Hz. The strain varies from 0.1 to 0.001. The cooling rate is 2 K-min-1. In Fig. 7, a small plateau can be observed upon cooling at the liquid-liquid demixing temperature in the complex viscosity at a low strain (0.001). However, as expected, the complex viscosity increases upon further cooling. With increasing strains the plateau becomes longer and the complex viscosity even shows a depression at a strain of 0.1. The reason of this depression is probably the same as in the viscosity experiments. Diluent is pushed out of the demixed solution and forms a slip layer. When using rheology or viscometry experiments to determine the liquidliquid demixing temperature, sufficient mechanical force should be applied to achieve immediate synereses of the demixed solution. The synereses results into an apparent viscosity drop that could be used as a marker for the onset of demixing. However, for studying the mechanical properties of the demixed solution, very small strains are recommended in order to avoid shear enhanced synereses. Otherwise, the measurement of the mechanical properties of the demixed solution is unreliable.

58


3.4

Conclusions and Outlook Liquid-liquid demixing temperatures can be observed with rheology and viscometry by using high shear rates. To study the mechanical behavior of demixing polymer solutions, care has to be taken to avoid altering of the liquid structure due to shear forces. It should be of much interest when rheological experiments can be carried out with a polymer solution at different frequencies. These results can be compared with results of emulsions to obtain information about the structure belonging to the demixed solution. Furthermore, rheological experiments can be very suitable to study the mechanical behavior of the polymer-diluent system at a temperature in the region of the structure fixation step. For example, the system aPS - DIDP forms a gel at room temperature because of the high viscosity of the polymer solution. After the gelation of the solution, no growth of the demixed domains is observed. Hence, gelation is sufficient to stop the liquid-liquid demixing process. The polymer-diluent system aPS - 1-dodecanol, which was studied in section 2, shows an onset of the glass transition temperature at about 65°C. The glass transition temperature is known to be sufficient to stop the liquid-liquid demixing process. However, it is very interesting to study whether at higher temperatures than the glass transition temperature, the liquid-liquid demixing process is influenced by or even stopped by gelation. Rheological experiments have the potential to clarify the structure fixation step of a polymer solution.

4.

ACKNOWLEDGEMENTS

E. Schomaker, J. Bos en R. Lammers of Akzo Nobel in Amhem are acknowledged for discussions with respect to the TMDSC work, and for giving us the possibility to carry out the TMDSC experiments. B. Reuvers (Akzo Nobel in Arnhem) and M. van Egmond (University of Twente) are acknowledged for carrying out the rheological experiments and for the discussions. REFERENCES

[1] [2] [3] [4] [5] [6]

M. Mulder, Basic principles of membrane technology, Kluwer Acedemic Publishers, Dordrecht, (1996). L.J. Zeman and A.L. Zydney, Microfiltration and Ultrafiltration, Marcel Dekker, Inc., New York, (1996). D.R. Lloyd, K.E. Kinzer and H.S. Tseng, J. Membrane Sci., 52 (1990) 239. A.T. Young, J. Vac. Sci. Techn., A 4 (1985) 1128. J.M. Williams and J.E. Moore, Polymer, 28 (1987) 1950. C. Schugens, V. Maquet, C. Grandfils, R. Jerome and P. Teyssie, Polymer, 37 (1996) 1027.

59


[7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17]

[18] [19] [201 [21] [22] [23]

[24] [25] [26] [27]

[28] [29] [30] [31] [32] [33] [34]

[35] [36] [37]

[38] [39]

[40] [41]

[42] [43] [44] [45]

Y.S. Nam and T.G. Park, J. Biomed. Mat. Res., 47 (1999) 8. A.J. Castro, US-patent 4247498 (1981). G.H. Shipman, US-patent 4539256 (1985). C. Josefiak and F. Wechs, GB-patent 2115425B (1985). W.C. Hiatt, G.H. Vitzthum, W.K.B., K. Gerlach and C. Josefiak, in Material science of synthetic membranes, Lloyd, D.R.(Ed.) American Chemical Society, Missouri (1984) 229. S. Berghmans, H. Berghmans and H.E.H. Meijer, J. Membrane Sci., 116 (1996) 171. H. Matsuyama, S. Berghmans and D.R. Lloyd, J. Membrane Sci., 142 (1998) 213. H. Matsuyama, S. Berghmans and D.R. Lloyd, Polymer, 40 (1999) 2289. P.M. Atkinson and D.R. Lloyd, J. Membrane Sci., 175 (2000) 225. P.M. Atkinson and D.R. Lloyd, J. Membrane Sci., 171 (2000) 1. A. Laxminarayan, K.S. McGuire, S.S. Kim and D.R. Lloyd, Polymer, 35 (1994) 3060. K.S. McGuire, A. Laxminarayan and D.R. Lloyd, Polymer, 35 (1994) 4404. K.S. McGuire, K.W. Lawson and D.R. Lloyd, J. Membrane Sci., 99 (1995) 127. K.E. Kinzer and D.R. Lloyd, Polym. Mater. Sci. Eng., 61 (1989) 794. D.R. Lloyd, S.S. Kim and K.E. Kinzer, J. Membrane Sci., 64 (1991) 1. S.S. Kim and D.R. Lloyd, J. Membrane Sci., 64 (1991) 13. S.S. Kim and D.R. Lloyd, Polymer, 33 (1992) 1047. J.H. Aubert, Macromolecules, 21 (1988) 3468. J.H. Aubert and C.R.L., Polymer, 26 (1985) 2047. M.R. Caplan, C.-Y. Chiang, D.R. Lloyd and L.Y. Yen, J. Membrane Sci., 130 (1997) 219. B.J. Cha, K. Char, J.-J. Kim, S.S. Kim and C.K. Kim, J. Membrane Sci., 108 (1995) 219. C.L. Jackson and M.T. Shaw, Polymer, 31 (1990) 1070. L. Aerts, M. Kunz, H. Berghmans and R. Koningsveld, Makromolek. Chem., 194 (1993) 2697. H.C. Vadalia, H.K. Lee, A.S. Myerson and K. Levon, J. Membrane Sci., 89 (1994) 37. L.L. Whinnery, W.R. Even, J.V. Beach and D.A. Loy, J. Polym. Sci. Polym. Chem., 34 (1996) 1623. P. Vandeweerdt, H. Berghmans and Y. Tervoort, Macromolecules, 24 (1991) 3547. P.D. Graham, A.J. Pervan and A.J. McHugh, Macromolecules, 30 (1997) 1651. G.T. Caneba and D.S. Soong, Macromolecules, 18 (1985) 2538. F.J. Tsai and J.M. Torkelson, Macromolecules, 23 (1989) 775. F.J. Tsai and J.M. Torkelson, Macromolecules, 23 (1990) 4983. S.-W. Song and J.M. Torkelson, Macromolecules, 27 (1994) 6389. R.M. Hikmet, S. Callister and A. Keller, Polymer, 29 (1988) 1378. J.H. Aubert, Macromolecules, 23 (1990) 1446. S.-W. Song and J.M. Torkelson, Polym. preprints, 34 (1993) 496. S.-W. Song and J.M. Torkelson, J. Membrane Sci., 98 (1995) 209. J. Amauts, H. Berghmans and R. Koningsveld, Makromolek. Chem., 194 (1993) 77. R. Koningsveld and A.J. Staverman, J. Polym. Sci., Part A-2, 6 (1968) 349. R. Koningsveld, W.H. Stockmayer and E. Nies, Polymer phase diagrams, Oxford University Press, Oxford, (2001). F. Khoury and E. Passaglia, in Treatise on solid state chemistry, Plenum Press, Hannay, N.B. (Ed.), New York-London, (1976) 335.

60


[46] [47]

[48] [49]

[50] [51] [52]

[53] [54] [55] [56] [57]

[58] [59]

[60] [61] [62] [63] [64] [65] [66] [67]

[68] [69] [70]

F.N. Kelley and F. Bueche, J. Polym. Sci., 50 (1961) 549. S. Nojima, K. Shiroshita and T. Nose, Polym. J., 14 (1982) 289. J. Lal and R. Bansil, Macromolecules, 24 (1991) 290. K.S. McGuire, A. Laxminarayan and D.R. Lloyd, Polymer, 36 (1995) 4951. J. Szydlowski and W.A. Van Hook, Macromolecules, 31 (1998) 3255. B.A. Wolf and M.C. Sezen, Macromolecules, 10 (1977) 1010. Y. Xie, K.F.J. Ludwig, R. Basnil, P.D. Gallagher, C. Kon~ik and G. Morales, Macromolecules, 29 (1996) 6150. J. Arnauts and H. Berghmans, Polym. Comm., 28 (1987) 66. J. Arnauts, R. De Cooman, P. Vandeweerdt, R. Koningsveld and H. Berghmans, Thermochim. Acta, 238 (1994) 1. P.C. van der Heijden, M.H.V. Mulder and W. M., Thermochim. Acta, 378 (2001) 27. B. Wunderlich, Thermal Analysis, Academix Press, Inc., London, (1990). V.B.F. Mathot, Calorimetry and thermal analysis of polymers, Hanser Publishers, Munich, (1994). M. Reading, Trends Polym. Sci., 1 (1993) 248. M. Reading, B.K. Hahn and B.S. Crowe, US-patent 5224775 (1993). J. Schawe, Thermochim. Acta, 260 (1995) 1. T. Ozawa and K. Kanari, Thermochim. Acta, 253 (1995) 183. K.J. Jones, I. Kinshott, M. Reading, A.A. Lacey, C. Nikolopoulos and H.M. Pollock, Thermochim. Acta, 304/305 (1997) 187. G.W.H. H/Shne, Thermochim. Acta, 304/305 (1997) 121. J.E.K. Schawe, Thermochim. Acta, 304/305 (1997. B. Wunderlich, A. Boiler, I. Okazaki and K. Ishikiriyama, Thermochim. Acta, 304/305 (1997) 125. R. Scherrenberg, V. Mathot and A. Van Hemelrijck, Thermochim. Acta, 330 (1999) 3. R. Scherrenberg, V. Mathot and P. Steeman, J. Therm. Anal., 54 (1998) 477. TA Instruments, Modulated DSC (MDSC) How does it work ? (1998). B. Wunderlich, ATHAS Databank, http://web.utk.edu/~athas/databank/intro.html (2000). T.E. Daubert, R.P. Danner, H.M. Sibul and C.C. Stebbins, Physical and thermodynamic properties of pure chemicals. Data compilation, Taylor&Francis, Pennsylvania, (1989).

61


Non-Invasive Characterizations of Membrane Fluid Transport and Fouling

63


New Insights into Membrane Science and Technology: Polymeric and Biofunctional Membranes

D. Bhattacharyya and D.A. Butterfield (Editors) © 2003 Elsevier Science B.V. All rights reserved.

Chapter 4

Study of membrane fouling and cleaning in spiral wound modules using ultrasonic time-domain retlectometry Zh.-X. Zhang a, A.R. Greenberg b, W.B. Krantz ¢, and G.-Y. Chai" aDepartment of Chemical Engineering, University of Colorado at Boulder, Boulder CO 80309-0424, USA bDepartment of Mechanical Engineering, University of Colorado at Boulder, Boulder CO 80309-0427, USA CDepartment of Chemical Engineering, University of Cincinnati, Cincinnati, Ohio 45221-0171, USA NSF Membrane Applied Science and Technology Center, University of Colorado, Boulder, CO 80309-0424, USA ABSTRACT Better detection and control of fouling in liquid-separation processes is essential if membranes are to find increased use in a variety of industrial applications. Ultrasonic time-domain reflectometry (UTDR) is an in situ, noninvasive real-time technique that has been successfully utilized to quantify fouling and cleaning of fiat sheet membranes. This study describes the extension of UTDR for the measurement of fouling and cleaning in commercial membrane modules employing spiral-wound elements. Experiments were conducted using a high-pressure separation system incorporating a commercial spiral-wound module. Calcium sulfate was utilized as the primary scalant in concentrations ranging from 1.2-1.8 g/L; ferric hydroxide colloidal fouling also occurred during the tests. A multi-cycle protocol was adopted whereby each cycle consisted of three stages: pure water equilibration, fouling and cleaning. Permeation rate as well as acoustic amplitude and arrival time measurements were made at regular intervals during three complete cycles on two modules. A new signal analysis protocol was developed such that systematic changes in the entire acoustic spectrum as a function of module operation time could be represented in terms of shift factors. In addition, gravimetric, microscopic and spectroscopic measurements were made at the end of the experiments so that the extent of the membrane fouling and cleaning could independently assessed. Results indicate that the overall decline in permeation rate is reasonably correlated with changes

65

Study Of Membrane Fouling And Cleaning In Spiral Wound Modules Using Ultrasonic Time-Domain Reflectometry - Greenberg

in both the amplitude and arrival time shift factors. Overall, this work confirms the effectiveness of the UTDR technique and provides a sound basis for continued development with an ultimate goal of providing on-line sensors for the timely detection of fouling and cleaning of commercial liquid-separation membrane systems. 1.

INTRODUCTION

As generally def'med, membrane fouling involves the deposition of retained particulates, colloids, macromolecules, crystallization of dissolved inorganic salts on or within the membrane. Fouling is the most critical problem limiting further growth and wider application of membrane-based liquidseparation processes [ 1]. Although the occurrence of fouling can be traditionally inferred from a marked decline in permeate flux or product quality, such an inference is unreliable since other phenomena such as concentration polarization, membrane compaction and degradation may also lead to similar changes in permeate flux or quality [2]. Moreover, using currently available techniques, it is not possible to quantify when a commercial membrane module has been adequately cleaned via chemical or mechanical de-fouling. Consequently, there is a compelling rationale for the development of an in situ, noninvasive technique that can monitor the growth and removal of a membrane-fouling layer in real time under realistic operating conditions. Such a technique would not only provide for identification of the specific phenomena underlying the flux decline, but also enable the improvement of fouling remediation strategies. Ultrasonic time-domain reflectometry (UTDR) has been used in a number of applications that involve the measurement of the location and topographical characteristics of a moving or stationary interface [3]. When the operating requirements of commercial high-pressure separation modules are considered, UTDR appears to be more suitable for real-time monitoring of membrane fouling than other noninvasive techniques, such as nuclear magnetic resonance microimaging [3]. Indeed, previous studies have demonstrated that UTDR can be successfully utilized to quantify fouling in a flat-sheet membrane module [2-4]. Recently, the UTDR technique was extended to the much more complex situation that occurs with the multiple layers that comprise spiral-wound reverse-osmosis (RO) modules. As in the flat-sheet studies, this work utilized calcium sulfate as the primary scalant due to its importance as a membrane-fouling agent [5-8]. Whereas the initial efforts to interpret the complex acoustic spectra utilized only a few of the many signal peaks [9,10], the present study employed a signal

66

Study Of Membrane Fouling And Cleaning In Spiral Wound Modules Using Ultrasonic Time-Domain Refleetometry - Greenberg

analysis protocol that accounts for systematic shifts in the entire acoustic spectrum as a function of module operating time. This paper also reports the first results obtained for UTDR detection of deposits formed by multiple foulants. 0

2.1

ULTRASONIC TIME-DOMAIN REFLECTOMETRY (UTDR)

Basic Concepts

The basic principles of UTDR and their application to the measurement of fouling on a fiat-sheet membrane have previously been described [2,3]. As shown in Figure 1, the ultrasonic wave pulses are transmitted by the transducer and propagate through the module. When the pulses encounter an interface, such as the one formed by the feed solution and top surface of the membrane, they will be reflected back to the transducer. The amplitude of the reflected waves depends on the acoustic impedance difference between the media on either side of the interface and the topography of the interface. The acoustic impedance is a function of the physical characteristics of the medium, and is given by the product of the density and the acoustic velocity within the medium. In addition, the time required for the transmission and reflection of a wave from the interface (i.e. arrival time) is a function of the path length and wave velocity: 2As AT = ~

(1)

where AT is the arrival time, As is the distance between the transducer and the interface and c is the acoustic velocity in the medium. Since the impedance, interface properties and path length may change with the growth of a fouling layer on the membrane surface, the change in the amplitude (AV) and the shift in the arrival time (AT) of the interface echoes can be analyzed and used to monitor fouling quantitatively in real-time.

2.2

Adaptation for Spiral-Wound Membrane Modules When the UTDR technique is applied to a spiral-wound RO membrane module contained in a cylindrical fiber-reinforced polymer housing, the signal analysis is complicated by two factors. First, a much more complex signal pattern will be obtained due to multiple reflections from the outer and inner surfaces of each layer of the multiple windings in the module assembly (Figure 2) [ 10]. Second, a significant loss of acoustic information will be caused by signal attenuation through these layers as well as through the module shell. The various reflections from these interfaces correspond to the many peaks in the waveform

67

Figure 1. Schematics that represent the principles of UTDR measurement: (Left) cross-sectional view of the separation cell, the externally mounted ultrasonic transducer, and the primary reflections, 1, 2, and 3 from the cell plate/fluid, fluidmembrane, and fluidfouling layer interfaces, respectively; (Right) the general ultrasonic response in the time-domain for each of these reflections.


obtained by plotting the signal amplitude as a function of the arrival time (Figure 3). Here the peaks in the waveform represent very complex superposed reflections from the many close-packed interfaces within the spiral-wound module instead of the simpler patterns from the largely separated interfaces that are characteristic of the fiat-sheet geometry. Within this complex spectrum, certain peaks are more responsive to the growth and subsequent removal of fouling layers than others. In addition, the shape of the waveforms depends on the position and mounting mode of the transducer on the external surface of the module shell. In the previous studies [10], considerable effort was required to select and analyze those peaks in the amplitude versus arrival time spectra that appeared most sensitive to fouling. The technique employed was time-intensive and made real time analysis difficult. In addition, all of the acoustic information obtained could not be fully used. In order to overcome these limitations, a more sophisticated signal acquisition and analysis protocol was developed. This procedure accounts for systematic shifts in the entire acoustic spectrum as a function of module operating time and enables information about the extent of fouling to be obtained in real time. The specific steps of the signal analysis protocol are summarized below: • All acoustic spectra are digitized. • A reference spectrum is obtained for a particular set of operating conditions with the spiral-wound module of interest. The reference spectrum corresponds to a "zero" operating time at which fouling has not occurred. Changes in the acoustic spectra as a function of operating time are determined based upon a systematic departure or shift from the reference spectrum. Such changes are shown in Figure 3 for the RO module employed in this study. Here, two waveforms corresponding to the reference spectrum and an updated spectrum, i.e. a spectrum obtained at a later time, are superposed to demonstrate the changes in the signal amplitude and arrival time. • Two shift factors are computed using a root-mean-square calculation such that values are always positive. One shift factor, QA, is based upon changes in the signal amplitude and the other factor, QT, is based upon changes in the signal arrival time. The two relationships are: N

"QT = '

~(Tiu_Tir) 2 1/2 N

(2)

Here, Viu and Vir are the amplitude of peak i in the updated spectrum and the reference spectrum, respectively; Tiu and Tir are the arrival time of peak i in the updated spectrum and the reference spectrum respectively; and N is the total number of peaks in the spectrum under consideration.

69


Transducer

Shell Feed Channel Spacer Membrane Permeate Collection Film

II II II [Layer1

~l ~]

Layern

-~ [

[

Figure 2. Schematic showing a cross-sectional view of a spiral-wound RO module assembly, including the fiberglass shell, the position of an externally mounted ultrasonic transducer, and reflections from the multiple interfaces. The width and length of the arrows are proportional to the reflected amplitude and arrival time, respectively.

2.50E-02

;~ 0.00E+00 3.7:E-0

J

E-05

t

AT (s)

°~,,i

-2.50E-02

Figure 3. Representative results showing a portion of the acoustic response in terms of signal amplitude versus arrival time (AT) from a commercial spiral-wound RO module. Changes in the UTDR signal peaks that occur as function of operating time are observed as a shift from the reference spectrum (heavy line) to the updated spectrum (light line).

The effectiveness of the new methodology was assessed by quantifying the signal fluctuations obtained during 30-hr runs using pure-water, and then representing

70


the effects of module fouling in terms of the shift factors based upon anticipated changes in the signal amplitude and arrival time. An "ideal" response is shown in Figure 4 and indicates the expected relationship between the permeation rate and the amplitude shift factors. Here, the shift factors and the permeation rate are plotted as a function of operating time for a pure water equilibration step, a period of CaSO4 fouling, and a pure-water cleaning step. The signal analysis protocol is designed to distinguish the relatively small-scale fluctuations due to normal operating condition variations from the large-scale changes that are due to fouling. The dashed lines represent a "small-scale fluctuation" confidence interval for each parameter. From a practical point of view, a successful technique should indicate when the signal has moved out of the normal operating window, i.e. the confidence interval, as a result of module fouling. The new methodology also allowed instantaneous calculation of the shift factors so that the occurrence and removal of fouling in the RO spiral-wound module could be detected and monitored quantitatively in real-time. Moreover, the portion of the spectrum that provides the basis for the shift factor calculations can be easily altered. For example, Figure 5 shows a representative waveform in which the spectrum is divided into three regions, each designated by a different rectangle. The waveform within each rectangle contains information from different layers within the module. Since fouling does not occur to the same extent in all regions of the module at the same time, particular portions of the spectrum will be most responsive to its occurrence and thus provide the most sensitive detection of the growth and removal of fouling layers. 3. EXPERIMENTAL 3.1. RO System and UTDR Instrumentation Based on previous work [9,10], the experiments were conducted on a bench-scale RO flow system that allowed for the control and measurement of important system operating parameters including temperature, pressure, flow rate and calcium-sulfate feed concentration. The major components of the RO system consisted of a Koch 2521 spiral-wound RO membrane module, a feed tank with temperature control, a low-pressure pump used to transport the feed through a 2micron prefilter, and a high-pressure booster pump to drive the RO membrane separation process (Figure 6). The system was designed to enable both the retentate and permeate from the membrane module to be returned to the feed tank to enable stable operation and conservation of chemicals. The acoustic hardware included two specially designed high-frequency (3.5 MHz) transducers (Research Institute of Acoustics, Chinese Academy of Science), a pulsar-receiver (Panametrics 5052 PRX) and a digital oscilloscope (Nicolet Pro 50) that enabled a 20-ns sampling rate. A custom mounting assembly was utilized to allow the transducers to be securely attached at various

71

Operating Time (hr) Figure 4. Schematic showing the anticipated change in the UTDR shift factors (QA and QT)and the permeation rate using the new signal analysis protocol during the water equilibration (WE), fouling (F) and cleaning (C) phases of module operation. The solid lines represent the overall shift factor and permeation rate behavior as determined from values obtained at discrete time intervals (solid circles). The dashed lines indicate a confidence interval for normal operation. Fluctuations within the confidence interval are acceptable but departures outside the dotted lines are taken to indicate module fouling.


positions on the external surface of the module. Secure attachment is essential to maximize changes in the acoustic waveforms that correspond to membrane fouling and cleaning. In the current experiments the transducers were mounted on the midstream and downstream portion of the module. Such multiple-point sampling can provide important information about the relative uniformity of the fouling and cleaning processes as a function of position. The UTDR signals were continuously monitored and recorded (waveform records containing 1000 data points) using a data-acquisition system that consisted of a personal computer (Pentium II 300 MHz) with a GPIB board (CyberResearch). Custom dataacquisition software enabled interfacing with the measurement instrumentation, real-time analysis of the UTDR signals, calculation of the two shift factors, and graphical display of both the shifts in the acoustic spectrum and changes in the values of the two shift factors. The acoustic hardware included two specially designed high-frequency (3.5 MHz) transducers (Research Institute of Acoustics, Chinese Academy of Science), a pulsar-receiver (Panametrics 5052 PRX) and a digital oscilloscope (Nicolet Pro 50) that enabled a 20-ns sampling rate. A custom mounting assembly was utilized to allow the transducers to be securely attached at various positions on the external surface of the module. Secure attachment is essential to maximize changes in the acoustic waveforms that correspond to membrane fouling and cleaning. In the current experiments the transducers were mounted on the midstream and downstream portion of the module. Such multiple-point sampling can provide important information about the relative uniformity of the fouling and cleaning processes as a function of position. The UTDR signals were continuously monitored and recorded (waveform records containing 1000 data points) using a data-acquisition system that consisted of a personal computer (Pentium II 300 MHz) with a GPIB board (CyberResearch). Custom dataacquisition software enabled interfacing with the measurement instrumentation, real-time analysis of the UTDR signals, calculation of the two shift factors, and graphical display of both the shifts in the acoustic spectrum and changes in the values of the two shift factors.

3.2. Experimental Procedure As indicated in Table 1, the testing protocol utilized three cycles, each of which generally contained water equilibration, fouling and cleaning phases. During the water-equilibration phase (phase A), the membrane module was operated using pure deionized water until the membrane module performance (permeation rate) and the UTDR signals (amplitude and arrival time) reached steady-state values. Then the feed was switched to an aqueous calcium sulfate solution that was prepared by adding anhydrous calcium sulfate powder (Aldrich, 325 mesh) to deionized water (equilibrium calcium sulfate solubility: 2 g/L).

73

0.30

0.20

0.10

0.00

-

0 0

0.10

c :

a

-0.20

0

~ o . Y o o o o . . . o o . o o o o o o o o o o o o o o o o o o o o o o o ~ o o o . o o o ~

-0.30

~

1

-0.40

I

Arrival Time (ps) Figure 5. The regions (A, B, and C) of the representative waveform each correspond to reflections from different locations within the spiral-wound module: The portion of the signal indicated by A (within square) is contained within the larger region given by B (within rectangle); both A and B correspond to portions of C. The methodology employed in this study enables selection of the particular region of the spectrum from which the shift factors are calculated. This provides the means to enhance the sensitivity for fouling detection.

v1

Control Valve

L

Y

s

fP

I

d I FeedTank

I

Filter

F

Flow Feed Pump 2

E

Transducers

5’

00

L

‘

Feed Pump 1

1

:

System Electronics

0

c

.

:

5’

m

F

t-

Transducer

s5’

.............................................................................................

110

f

R’

4

I’

Pulser Receiver Digital Oscilloscope Computer ............................................................................................

B

I

Figure 6. Schematic of the high-pressure RO separation system that shows the major components including the integrated capabilities for UTDR measurement. The arrows indicate the directions for the recycled retentate and permeate flows. The system electronics for acoustic measurements is detailed in the enlarged view at the bottom.

fff.

I B P

d

Study Of Membrane Fouling And Cleaning In Spiral Wound Modules Using Ultrasonic Time-Domain Refleetometry - Greenberg

This initiated phase B in which concentration polarization and fouling occurred. This phase was generally allowed to continue until the ultrasonic responsed permeation rate approached steady-state behavior. In the subsequent cleaning phase (phase C), the feed solution was switched back to pure deionized water in order to remove the fouling layer from the membrane surfaces. Overall, long-term experiments were conducted on two modules in order to accommodate the multiple fouling and cleaning cycles; the modules were run for a total of 654 and 1108 hours, respectively. In each experiment the fouling cycles generally utilized systematically higher concentrations of calcium sulfate. Table 1. Details of the Fouling and Cleaning Protocol

Module Cycle/Phase I Cycle 1:CaSO4 (1.2 g/L) Phase A: Water Equilibration Phase B: Fouling Phase C: Cleaning Cycle 2:CaSO4 (1.6 g/L) Phase A: Water Equilibration Phase B: Fouling Phase C: Cleaning Cycle 3:CaSO4 (1.6 g/L) Phase A: Water Equilibration Phase B: Fouling Total Operating Time (hr) II Cycle 1:CaSO4 (1.2 g~) Phase A: Water Equilibration Phase B: Fouling Phase C: Cleaning Cycle 2:CaSO4 (1.6 g/L) Phase A: Water Equilibration Phase B: Fouling Phase C: Cleaning Cycle 3:CaSO4 (1.8 g/L) Phase A: Water Equilibration Phase B: Fouling Phase C: Cleaning Total Operating Time (hr)

Duration (hr) 37 164 82 30 100 71 70 100 654 72 102 150 75 126 110 33 296 144 1108

All experiments were conducted under the same operating conditions. The upstream pressure of the module was maintained as 0.7MPa, and the feed

76


solution flow rate was maintained as 0.9 L/min. The temperature of the feedsolution in the tank was controlled at 20°C and the inlet temperature of the module was maintained at 22°C. The CaSO4 concentration of the feed, permeate and retentate were measured at regular intervals using a conductivity meter (Yellow Spring Instruments), and the membrane rejection was calculated via standard methods. In addition, the permeate and retentate flow rates were determined via a digital balance and timer. In order to confirm independently the extent of membrane fouling indicated by the values of the acoustic shift, each of the two RO modules was opened after the appropriate phase in the operating protocol for visual inspection, morphological studies, and gravimetric and elemental analysis. Module I was opened at the end of the Cycle 3 fouling phase to observe the extent of coverage as well as the distribution of the fouling layer on the membrane surfaces, while Module II was opened at the end of Cycle 3 cleaning phase to determine the degree of fouling deposit removal via pure water cleaning (Table 1). The morphological features of the fouled and water-cleaned membrane samples were characterized using light microscopy (Nikon Labophot) and scanning electron microscopy (Aspex Instruments). Gravimetric analysis was performed o n sections of the unrolled membrane using a microbalance, and the elemental composition and distribution of the deposits were determined via energy dispersive spectroscopy (EDS). 4. RESULTS AND DISCUSSION The important aspects of the UTDR response are demonstrated in the data obtained for module I after the completion of each of the three cycles. Results for the first cycle are shown in Figure 7 for the two shift factors, QA and QT, from the downstream transducer as well as the permeation rate as a function of module operating time. In order to compare directly the sensitivity and effectiveness of the UTDR technique with the new signal analysis protocol and the traditional fouling detection mode, changes in both the shift factors and the permeation rate are represented as percentage changes from their respective reference values, i.e. the values at the start of phase A. During the water-equilibration phase, the permeation rate is relatively constant, whereas there is a gradual but distinct increase in QA. During the fouling phase, the permeation rate declines by approximately 12% and QA increases by over 1500%. By completion of the cleaning phase, the permeation rate has returned to its reference value; however, QA remains almost unchanged. In addition, no apparent change in QT is observed over the course of cycle 1. Note that the spikes in QT at the end of the fouling run and the beginning of the cleaning run are believed to be due to a problem with the power supply. Similar results for the shift factors were obtained from the midstream transducer. Overall, the observed behavior in QA and QT was not as expected. Nevertheless, we continued with the protocol indicated in Table 1.

77


Results for the second cycle are shown in Figure 8. Here each of the shift factors is represented by three different curves, each of which corresponds to the different regions of the signal spectrum shown in Figure 5. These data demonstrate that the shift factors from each region of the spectrum lead to a similar response although the differences suggest that particular regions of the spectrum may indeed be more sensitive to the presence and removal of a fouling layer. After establishing steady-state behavior during the equilibration phase, there is a marked decline in the permeation rate (~20%) and an increase in QA and QT during the fouling phase. During the subsequent cleaning phase, the permeation rate and QT return to their reference values. Although QA clearly trends in the same direction, this factor has not reached the reference value at the completion of the cleaning cycle. The results for cycle 2 obtained from the midstream transducer, however, were considerably different than those from the downstream transducer. Indeed, the signals from midstream transducer provided shift factor responses that were rather similar to those observed during cycle 1 (Figure 7). This emphasizes a fundamental difference between the permeation and ultrasonic measurements whereby the former represents an average response for the entire module while the shift factors can reflect the condition of the module on a local basis. A third operating cycle was then conducted under the same conditions as for cycle 2. However, in this case the protocol was modified so that instead of a cleaning phase, the membrane module would be opened at the end of the fouling phase so that microscopic, spectroscopic and gravimetric analyses could be performed. Results for the permeation rate and shift factor responses during cycle 3 are shown in Figure 9. As expected, the three response parameters are reasonably constant during the water equilibration phase. With the initiation of the fouling phase, there is a relatively rapid decrease in the permeation rate (~20%) over the first 10 hours to an approximately steady-state value that is maintained for the next 90 hours until the end of the run. In contrast, both shift factors increase throughout the duration of the fouling phase. Whereas QA increases at a relatively constant rate, the initial rate of increase of QT becomes more modest after the first 30 hours of fouling. Overall, these results confirm the trends observed during cycle 2 (Figure 8). At the end of the fouling phase in cycle 3, the module was opened and unwound. Visual examination indicated the presence of a reddish-brown deposit on the membrane surface, with the color turning progressively lighter (whiter) from the entrance to the exit end of the module. Microscopic examination confirmed the coexistence of the reddish-brown and white deposits where the latter evidenced the same characteristics previously reported for crystalline CaSO4 (Figure 10) [2,3]. The CaSO4 crystals were mainly deposited along the regularly distributed convex lines formed on the membrane sheet during

78

Study O f Membrane Fouling And Cleaning In Spiral Wound Modules Using Ultrasonic Time-Domain Reflectometry - Greenberg

800 2500 600

z

o~'1500

< O "O 500 N t~ E O z -500

3 ~°

,~

i

0

.

.

.

.

.

.

.

.

.

.

.

. vl.........................................................t

..................

,I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ............................................................................................... 0

•

i

..................

1..............................................................................................................

1 400N -I

4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

i ~.

,,

200~

~

I

-1500

o "ql-WE ~'q

F

" "

C

i1,'- -'ql

"

120 v

"~ 110 n,"

P,

&l& •

100

E • 90

••

• ••~4t.-AA• •

•••~••

• N

~

0 I I I

8o

I

E L_

0

z

I I

70

|

|

|

|

|.

50

|

| .

|

|

.

|

100

i

. i

i

i

i

I

I

I

i

Ii

150 200 Operating Time (hr)

i

i

i

i

~

250

.

.

.

.

300

Figure 7. Results obtained for cycle 1, module I (Table 1) including the water equilibration (WE), fouling (F) and cleaning (C) phases and shown as percent changes from the respective reference values. Top: Amplitude and arrival-time shift factors (downstream transducer); bottom: permeation rate. Permeation rate values in F and C phases based upon reference value that incorporated the membrane compaction that occurred during the WE phase.

79


800 I

1500

I I I I

!

"

500

.,L__

~

...........................................................................................

Fe ÷3) and the subsequent formation of ferric hydroxide. Previous work has shown that insoluble ferric hydroxides (colloidal) can produce fouling of RO systems [12]. The presence of this mixed foulant provided an unanticipated degree of complexity to the experiments and also accounts for the unexpected shift factor response previously described. Formation and deposition of the iron-containing foulant on the membrane surface undoubtedly occurred during the initial exposure of the new section of pipe to water (cycle 1). Fouling due to the deposition of iron-containing layers most likely became more extensive during phase B because the introduction of CaSO4 would facilitate electrolytic corrosion of the iron pipe [ 13]. Correspondingly, QA evidenced only a slight increase during the cycle 1 water-equilibration phase but a much more pronounced increase during the subsequent fouling phase (Figure 7). In contrast, QT was relatively unchanged throughout cycle 1. This behavior would be expected from the deposition of a very thin layer with a relatively loose structure since QA is sensitive to the topographical characteristics whereas QT is most responsive to the thickness of a sufficiently thick deposit. In the present case this suggests that QA and QT respond preferentially to the deposition of the ironcontaining and CaSO4 layers, respectively. If correct, this indicates that the deposition of the iron-containing colloidal material accounted for a significant portion of the overall fouling that occurred during phase B of cycle. This hypothesis is supported by the results of the retentate concentration measurements that indicated values of only 1.6 g/L at the end of phase B. As previously described, the permeation rate and QA responses during the cleaning phase of cycle 1 were puzzling since the former returned to its reference value but the latter decreased only slightly. However, such behavior would be consistent with the fact that while the pure water quickly cleaned the CaSO4, the cleaning protocol was not effective in removing the iron-containing layer. These results also suggest that the iron-containing layer had only a relatively small effect on the permeate flux due to its non-dense colloidal structure. During cycle 2, QA remained relatively stable during phase A in contrast to the increase observed during cycle 1 (Figure 7). The behavior of QA in conjunction with the relatively constant values of QT during the waterequilibration phase of both cycles suggests that any additional iron-based foulinglayer deposition during phase A was not extensive. Results obtained during the cycle 2 fouling phase, when a higher concentration of CaSO4 (1.6 g/l) was used in the feed solution, are consistent with a fouling layer that consisted of more significant amounts crystalline CaSO4 as well as the iron-containing colloidal material. In this case the permeation rate decreased and both shift factors increased. During the subsequent pure-water cleaning phase, the permeation rate and QT return to their respective reference values. 81


2500

800

1500

600 Z O

'-'1

-0.5

-0.5 ~

-1

!;~iiilziiii:i~i~ili ~ii~,iiiii;~ii

~ii~iiii!iii!!~;i

i 150

~

0

0

......:-~'--'--" SO

Y-Position In Space

150 100

II

-1.5

X-Position In Space

Figure 9. 3-D representation of velocity profile acquired from the mid-length of the flow phantom. Reprinted with permission from Poh [8].

5.2.2. 2DPC Technique Figure 8 shows the raw phase image acquired by the 2DPC technique from the flow phantom. The image artifact was removed during image processing, and the image processing methodology has been discussed in Poh [8]. The processed phase image was then converted into a velocity image using Eq. (13). A 3-D representation of the velocity profile (Figure 9) acquired from the mid-length of the flow phantom shows that the countercurrent flows had the characteristics of a fully developed laminar pipe flow, and the maximum positive and negative velocities were approximately equal in magnitude. These countercurrent flows were fully developed laminar flows because at these cross sections of the tubes, the ratio of the tubes' length to their inner diameters (L/d) was 56. Generally, a laminar pipe flow is fully developed when the L/d ratio [35] is greater than 0.06Re, Re being the Reynolds number determined from the inner diameter of the pipe. In our flow phantom, its 0.06Re had a value of eight,

109

NonintrusiveCharacterizationOf FluidTransportPhenomenaIn Hollow-FiberMembraneModulesUsingMRI:An Innovative ExperimentalApproach- Poh T a b l e 1. Comparison of experimental data measured from the 2DPC technique and rotameter. Reprinted with permission from Poh [8] Positive-Flow Tube in a Flow Phantom

Flow Rate (mL/min) Maximum Velocity a (cm/s)

2DPC

Rotameter

% Relative Error b

97.27

100.2

-2.9

1.65

1.8

-8.3

Negative-Flow Tube in a Row Phantom

Flow Rate (mLJmin) Maximum Velocity a (cm/s)

2DPC

Rotameter

% Relative Error b

94.52

100.2

-5.7

1.68

1.8

-6.7

a Calculated from the Poiseuille's equation based on the flow rate measured by the calibrated rotameter. b Determined from the following equation" % Relative Error =

2DPC Data - Rotameter Data

x 100 %

Rotameter Data

which was much less than its L/d ratio; therefore, the countercurrent flows at the mid-length of the flow phantom were fully developed laminar flows. In addition to the parabolic velocity profile of the flow phantom, the velocities measured from the 2DPC technique were in good agreement with the analytical velocities determined from the Poiseuille's equation based on the flow rate measured from the calibrated rotameter (Table 1). To further validate the accuracy of the 2DPC technique, we quantified the flow rates and determined the flow areas in each tube from the velocity image. The flow rates in each tube quantified from the 2DPC technique were calculated by multiplying the area of a single pixel with the sum of all velocities in each tube, and they were in good agreement with that measured from the calibrated rotameter (Table 1). The flow areas in each tube, on the other hand, were determined from the velocity image by multiplying the area of a single pixel with the sum of all pixels in each tube; they were 2.04 and 1.96 cm 2 in the tubes with positive and negative flows, respectively, compared to the actual flow area of each tube, 1.96 c m 2. Therefore, the accuracy of the 2DPC technique, with the imaging parameters used in this study, has a six to eight percent relative error in the measured velocity, a three to six percent relative error in the quantified flow rate, and a four percent relative area in the measured flow area of the flow phantom. The accuracy of these measurements may be further improved by increasing the spatial resolution of the velocity image using a smaller pixel size

110

Nonintrusive Characterization Of Fluid Transport Phenomena In Hollow-Fiber Membrane Modules Using MRI: An Innovative Experimental Approach - Poh

F l o w R a t e = 100 m L / m i n

R e = 135

O ..........

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

,.,., .

.

.

.

.

i .io............, .....ii ,

Oi

iiil,i~i4i i,!~i~i~!:ii.i!~i~ !i~:

O

e ~:~,,

12 mm/s

= + 14.:~

Figure 10. Sixteen of the twenty-four raw Fourier-velocity images of countercurrent flows acquired from the mid-length of the flow phantom. Reprinted with permission from Poh [8]. (the pixel size used in this study was 0.34 mm2). The readers should take note, however, this accuracy does not necessary represent the accuracy in measuring flow in hollow-fiber membrane modules, which depends upon the inner diameter of the hollow-fiber membrane and the spatial resolution of the MR images. 5.2.1.2DFT Technique We acquired twenty-four Fourier-velocity images, each with a velocity bin size of two mm/s, across the mid-length of the flow phantom. Figure 10 shows sixteen of the twenty-four raw Fourier-velocity images acquired. The image artifact was caused by the Gibbs ringing [36-38] and noise [39,40] generated from the receiver coil during image acquisition and was removed during image processing [8]. The percent flow rates in each velocity bin relative to the total flow rate in each tube determined from these Fourier-velocity images were in good agreement with the analytical values derived from the Poiseuille's equation for a fully developed laminar pipe flow for mean velocities ranging from 2 to 14 mm/s (Figure 11). For a fully developed laminar pipe flow, this percent flow rate relative to the total flow rate is predicted from the Poiseuille's 111


FI0w Rate = 100 mLJmin

Re = 135

25 2DFT Data (Positive Flow) 2DFT Data (Negative Data) Analytical Data

20-

i]

;)i i

(1)

rr"

15

\ \ i:~:i \

o LL ¢..

\ i:i:i

a~ 10 o !,_

fl_

\

0

!

!

i

!

!

2

4

6

8

10

12

!

i

!

14

16

18

20

Mean Velocity (mm/s)

Figure 11. Comparison of the percent flow rate as a function of mean velocity at the mid-length of the flow phantom. Reprinted with permission from Poh [8].

equation to increase linearly with increasing mean velocity (Appendix). The discrepancies between the experimental and analytical values at the mean velocities of 16 and 18 mm/s may have been caused by experimental errors such as inhomogeneity in sensitivity profile of the radio frequency coil, finite velocity resolution in the Fourier-velocity images (each image covers a specific velocity bin size, and the velocity resolution decreases with increasing velocity bin size), Gibbs artifact [36-38], noise [39,40] generated from the receiver coil, and variations in flow velocity during the image acquisition period. @

EXAMPLES OF EXPERIMENTAL RESULTS FROM HOLLOWFIBER HEMODIALYZERS

All experiments were performed on a Siemens 1.5T MAGNETOM Vision whole-body MR imager (Siemens AG, Erlangen, Germany) with a maximum gradient strength of 25 mT/m and a rise time of 600 ItS. All image processing was performed using user-written MATLAB programs together with the MATLAB Image Processing Toolbox 2.2 software (The MathWorks Inc., Natick, MA, USA) running on a Linux workstation.

112

Nonintrusive Characterization Of Fluid Transport PhenomenaIn Hollow-Fiber Membrane Modules Using MRI: An Innovative Experimental Approach - Poh

POLYAMIDE HEMODIALYZER Flow Rate: 400 mL/min

Blood Compartment

..

o ?,-

.

150 0 Y-Po~

in Space

X - P o ~ n in Space

Figure 12. Blood-side velocity distribution at the middle cross section of a polyamide hemodialyzer. Reprinted with permission from Poh [8].

We characterized the flow distribution in the blood compartment of hemodialyzers by introducing a three-millimolar cupric sulfate solution, prepared from distilled water, through the blood compartment using a nonpulsatile pump, while the dialysate compartment was filled with the same solution and sealed. Similarly, we characterized the flow distribution in the dialysate compartment by introducing the solution through the dialysate compartment, while the blood compartment was filled with the solution and sealed. In addition, we characterized the flow profile in each compartment and quantified the local ultrafiltration rates by introducing the solution through the blood compartment; we recycled the outflow from the blood compartment to the dialysate compartment countercurrently. The flow rates were between 200 to 1000 mL/min, and five imaging planes were acquired from different transverse cross sections of each hemodialyzer. Detailed experimental methods and image processing methodology have been reported in Poh [8]. Figures 12 and 13 show the velocity distributions in the blood and dialysate compartments of a polyamide hemodialyzer. These velocity images were acquired from the middle cross section (mid-length) of the hemodialyzer using the 2DPC technique with nominal blood and dialysate flow rates of 400 and 600 mL/min, respectively. The polyamide hemodialyzer contains 10,000 hollow fibers, each with an inner diameter of 215 ~m and a wall thickness of 50 113


POLYAMIDE

HEMODIALYZER

Dialysate Compartment

Flow Rate: 600 mL/min

....... •.

...; ..-:"i;~i!~.' ~'

i!, '

"

'

"

:

.

.

.

4

"}i. .

.

2 o

._~

1

0

150 100

50

> '~

150 0 Y-Position in Space

X-Positionin Space

t/" Figure 13. Dialysate-side velocity distribution at the middle cross section of a polyamide hemodialyzer. Reprinted with permission from Poh [8].

~tm. It has an ultrafiltration coefficient of 71 mL/Hr/mmHg and a total membrane surface area of 1.7 m 2. Figures 12 and 13 show that the flow distribution in the blood compartment of the polyamide hemodialyzer was uniform and that in the dialysate compartment was nonuniform. Figure 14 shows the dialysate-side axial velocity distributions acquired from the imaging planes right across the dialysate ports of the polyamide hemodialyzer at nominal flow rates between 200 and 1000 mL/min. Figure 14 shows that the axial dialysate-side flow near to the dialysate ports moved preferentially at the top region of the cross section of the hemodialyzer. Figure 15 shows that the percent flow rate increased linearly with increasing mean velocity at the mid-length of a cellulose triacetate hemodialyzer, indicating that the flow in the blood compartment of the hemodialyzer had a fully developed laminar flow profile (Appendix). On the other hand, the flow in the dialysate compartment of the hemodialyzer at the same imaging plane was slightly skewed towards higher mean velocities (positive skewness in mean velocities), indicating that slightly more flow moved at lower mean velocities than at higher mean velocities. The data in Figure 15 were obtained from the Fourier-velocity images reconstructed by the 2DFT technique at a nominal flow rate of 280 mL/min. The cellulose triacetate hemodialyzer contains 12,000 hollow fibers, each with an inner diameter of 200

114


CELLULOSE TRIACETATE HEMODIALYZER Dialysate C o m p a r t m e n t

Dialysate In

Dialysate Out

200 mLtrniN

400 mLtmir~

600 mLlmir;

800 mL!min

t000 mLimin

~i~#iii~i!iiii~i~i~i~i~ii~i~i~i~i~i~i~!i~!iii!!!i !i!~i!ii!?__ 0

5

10 Axial Velocity in cml$

15

20

Figure 14. Dialysate-side axial velocity distributions acquired from the imaging planes

right across the dialysate ports of a polyamide hemodialyzer at nominal flow rates between 200 and 1000 mL/min. Reprinted with permission from Poh [8]. lam and a wall thickness of 15 pm. It has an ultrafiltration coefficient of 36 mL/hr/mmHg and a total membrane surface area of 1.9 m 2. Figure 16 shows the local maximum velocities and ultrafiltration rates quantified from a polyethersulfone hemodialyzer using the 2DFT technique at nominal flow rates of 300 and 600 mL/min, respectively. The local maximum velocities and ultrafiltration rates decreased monotonically along the length of the hemodialyzer, indicating that the transmembrane pressure decreased along the length of the hemodialyzer. The polyethersulfone hemodialyzer contains 11,200 hollow fibers, each with an inner diameter of 200 ~tm and a wall thickness of 30 ~tm. It has an ultrafiltration coefficient of 84 mL/hr/mmHg and a total membrane surface area of 1.9 m 2. These experimental results served as an illustration of the capabilities of the 2DPC and 2DFT techniques. Detailed experimental results and discussions have been published elsewhere [8-14].

115

N o n i n t r u s i v e Characterization O f F l u i d Transport P h e n o m e n a In H o l l o w - F i b e r Membrane Modules U s i n g E x p e r i m e n t a l A p p r o a c h - Poh

Flow Rate = 280 mL/min 20

MRI: An Innovative

Re(single hollow fiber) = 2

,

,

,

,

,

Hemodialyze

Cellulose Triacetate

0

0 '

4

8

12

16 20 Mean Velocity (minis)

24

'

20

'

:

:

28

'

.

32

'

;

I"

I Cellulose Triacetate Hemodialyzer N 15

I1

................... ~..................... i ..................... !..................... :.................... D i a l y s a t e C o m p a r t m e n t

°

•

!

rrlO

.

"E

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

!

::

:

4

8

12

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

5

o

0

16 20 Mean Velocity (ram/s)

24

28

32

Figure 15. Percent flow as a function of mean velocities at the mid-lengthof a cellulose triacetate hemodialyzer. Reprinted with permission from Poh [8]. 7.

LIMITATIONS OF THE 2DPC AND 2DFT TECHNIQUES

Like all other experimental tools, these flow-imaging techniques have their own limitations. Both flow-imaging techniques are only sensitive to flow moving in the flow-encoding direction. In the examples of experimental results presented earlier, the flow-encoding direction was in the axial direction of the hemodialyzer. In reality, the flow in the dialysate compartment of a hemodialyzer moves in all other directions in addition to the axial direction due to the complex flow pathway created by the presence of several thousands of hollow fibers in the hemodialyzer. Moreover, these flow-imaging techniques are only applicable to flow with steady velocity due to their long acquisition times (approximately one minute in the 2DPC and ten minutes in the 2DFT). Any changes in flow velocity during the acquisition period can lead to artifacts in the acquired MR images and subsequently can lead to inaccuracy in the recorded velocities. However, these flow-imaging techniques can be modified by using rapid imaging techniques to account for pulsatile flow seen in all hemodialysis therapies.

116


120

100

-r-

UltrafUtration Rate I - - - 3 Ultrafiltration Rate ~ Maximum Velocity -0Maximum Velocity

-r-

|

t°~

E E

80

r'F

60

I I !'

at Q=300 at Q=600 at e=300 at Q=600

mL/min mL/min mL/min mL/min

(~

5 ~" E 4 o,~

v

I

"g ~_.

40

!

20 -

~i~;

0

.

. 0.00 (Blood Inlet)

o

l

3 2

.~

1

. 0.25

.

. 0.50

0.75

0 1.00 (Blood Outlet)

Normalized Distance Along the Length of Hemodialyzer

Figure 16.

Local blood-side maximum velocities and ultrafiltration rates along the length of a polyethersulfone hemodialyzer at nominal flow rates of 300 and 600 mL/min. Reprinted with modification from Hardy et al. [14].

It is also worthwhile pointing out that the 2DPC technique is only capable of measuring unidirectional flow in a hollow-fiber membrane module. Any countercurrent velocities contained in a single pixel are averaged over the whole pixel. Moreover, if the flow is oblique instead of perpendicular to the imaging plane, the velocity can be underestimated [34] by a factor of cos(B), where fl is the oblique angle between the flow direction and the imaging plane. For a small r, the error caused by the obliqueness of flow is small since cos(r) is approximately one for a small ft. If fl is known and is smaller than 45 °, the underestimated velocity can be corrected by dividing it by cos(r). In addition to the obliqueness of flow, eddy-current [41 ] induced by the changes in gradient waveforms during data acquisition from the 2DPC technique causes some bias in the velocity determined from the measured phase shift. The eddy current causes a small variation in the phase shift and hence a small variation in the velocity. The eddy-current effect is often seen as a variation in phase shift in the phase images with stationary flow. The eddycurrent effect can be corrected after image acquisition by examining the variation in phase shift of a stationary flow, and it has been corrected in the

117


experimental results shown in this chapter. The correction for the eddy-current effect has been discussed in Walker et al. [41] and Poh [8]. In the 2DPC technique, the measured velocities are also distorted by the partial volume effect [42-45]. The measured velocity in each pixel is determined by the phase of the resultant magnetization, which is affected by all MR-visible protons within the voxel (volume element). In the case of the hemodialyzers, each voxel contains a mixture of protons in three environments: inside the hollow fibers, within the wall of the hollow fibers, and outside the hollow fibers. Different voxels will have different percentages of protons in each environment because of the random distribution of the hollow fibers relative to the imaging grid. In the characterization of the blood-side flow distribution, the fluid inside the hollow fibers was moving, while that outside the hollow fibers was stationary; the opposite occurred in the characterization of the dialysate-side flow distribution. Thus, in each voxel, the signal arises from both stationary and moving protons; the measured velocity in each pixel is therefore distorted and is generally lower than the true velocity. Nevertheless, the region-to-region variation in velocity can be estimated by measuring the mean velocities in different region-of-interest (ROI) large enough to include sufficient number of pixels to average out the variation arising from differences in the number of hollow fibers in each voxel. That is, by determining the mean velocity in one region, we will have derived a quantity that is proportional to the total flow rate of that region. This means velocity can then be used to compare the differences in flow in different ROI of the hemodialyzers. On the other hand, in the 2DFT technique, the measurable velocity resolution is limited by the velocity bin size selected for each Fourier-velocity image. Theoretically, we can improve the measurable velocity resolution by decreasing the velocity bin size. However, there is a tradeoff between velocity resolution, acquisition time, and signal-to-noise ratio for a given range of velocities. For a given range of velocities, a higher velocity resolution causes a lower signal-to-noise ratio and requires a longer acquisition time due to the increase in the number of Fourier-velocity images required for a smaller velocity bin size. 8.

CONCLUSIONS

We introduced two innovative, nonintrusive flow imaging techniques using MRI called the 2DPC and 2DFT techniques with the hope that these flowimaging techniques will provide a greater insight into the fluid transport phenomena in hollow-fiber membrane modules. These flow-imaging techniques provide a valuable nonintrusive experimental tool for studying flow distribution in hollow-fiber membrane modules. The experimental results obtained from these flow-imaging techniques give the actual fluid transport behavior in the 118


hollow-fiber membrane modules that cannot be predicted from numerical models and provide us a better physical understanding of the fluid transport phenomena in these modules. Moreover, an improved understanding of fluid transport phenomena in these modules may lead to the development of more efficient hollow-fiber membrane modules. These flow-imaging techniques can also be used to validate, improve, or develop numerical models capable of predicting separation performance of hollow-fiber membrane modules under different engineering designs and operating conditions. Some examples of experimental results obtained from hollow-fiber hemodialyzers are presented to illustrate the capabilities of these flow-imaging techniques. The readers are encouraged to refer to the references cited throughout this chapter for a full appreciation of these and other MRIbased flow-imaging techniques.

REFERENCES [ 1] [2] [3] [4]

[5] [6] [7] [8] [9] [10] [11] [12] [13]

B.E. Hammer, C.A. Heath, S.C. Mirer, G. Belfort, Quantitative flow measurements in bioreactors by nuclear magnetic resonance imaging, Biotechnology, 8 (1990) 327-30. C.A. Heath, G. Belfort, B.E. Hammer, S.D. Mirer, J.M. Pimbley, MRI and modeling of flow in hollow-fibre bioreactors, AIChE J, 36 (1990) 547-58. C. Donoghue, M. Brideau, P. Newcomer, B. Pangrle, D. BiBiasio, E. Walsh, Use of magnetic resonance imaging to analyze the performance of hollow-fiber bioreactors, Ann N Y Acad Sci, 665 (1992) 285-300. B.J. Pangrle, E.G. Walsh, S. Moore, D. DiBiasio, Investigation of fluid flow patterns in a hollow fiber module using magnetic resonance imaging, Biotech Tech, 3 (1989) 67-72. J. Zhang, D.L. Parker, J.K. Leypoldt, Flow distributions in hollow fiber hemodialyzers using magnetic resonance fourier velocity imaging, ASAIO J, 41 (1995) M678-82. S. Laukemper-Ostendorf, H.D. Lenke, P. Blumler, B. Blumich, NMR imaging of flow in hollow fiber hemodialyzers, J Membr Sci, 138 (1998) 287-95. T. Osuga, T. Obata, H. Ikehira, S. Tanada, Y. Sasaki, H. Naito, Dialysate pressure isobars in a hollow-fiber dialyzer determined from magnetic resonance imaging and numerical simulation of dialysate flow. Artif Organs, 22 (1998) 907-9. C.K. Poh, Characterization of fluid transport in artificial kidneys - A MR/approach, MS Thesis, University of Kentucky, Lexington, KY, USA, 2001. C.K. Poh, P.A. Hardy, W.R. Clark, D. Gao, Characterization of flow in hemodialyzers using MRI, Proc Intl Soc Mag Reson Med, 9 (2001) 1984. C.K. Poh, P.A. Hardy, D. Gao, W.R. Clark, Measurement of local ultrafiltration rate in a hemodialyzer: a novel MRI approach (abstract), ASAIO J, 47 (2001) 160. C.K. Poh, P.A. Hardy, Z. Liao, Z. Huang, D. Gao, W.R. Clark, Characterization of flow distribution in hemodialyzers using MR/(abstract), ASAIO J, 48 (2002) 179. C.K. Poh, P.A. Hardy, Z. Liao, Z. Huang, W.R. Clark, D. Gao, Effect of spacer yams on the dialysate flow distribution of hemodialyzers: A MRI study, ASAIO J, in press. C.K. Poh, P.A. Hardy, Z. Liao, Z. Huang, W.R. Clark, D. Gao, Characterization of flow distribution in hemodialyzers - A MR/approach, in preparation.

119


[14]

[151 [16] [17] [18] [19] [20] [21] [22] [23]

[24] [25] [26] [27] [28] [29]

[30] [31]

[32] [331

P.A. Hardy, C.K. Poh, Z. Liao, D. Gao, W.R. Clark, The use of magnetic resonance imaging to measure local ultrafiltration rate in hemodialyzers, J Membr Sci, 204 (2002) 195-205. Abragam, The Principle of Nuclear Magnetism, Oxford University Press Inc., New York, USA, 1983. B.R. Friedman, A.P. Salmon, G. Chaves-Munoz, C.R. Merritt, J.P. Jones, Principles of MR/, McGraw-Hill Co., USA, 1989. P.T. Callaghan, Principles of Nuclear Magnetic Resonance Microscopy, Oxford University Press Inc., New York, USA, 1995. M.A. Brown, R.C. Semelka, MR/: Basic Principles and Applications, 2nd ed., Wiley, John & Sons Inc., USA, 1999. T.J. Mosher, M.B. Smith, A DANTE tagging sequence for the evaluation of translational sample motion, Magn Reson Med, 15 (1990) 334-9. D.D. Blatter, D.L. Parke, R.O. Robinson, Cerebral MR angiography with multiple overlapping thin slab acquisition. Part I. Quantitative analysis of vessel visibility, Radiology, 179 (1991) 805-11. E.M. Haacke, A.S. Smith, W. Lin, J.S. Lewin, D.A. Finelli, J.L. Duerk, Velocity quantification in magnetic resonance imaging, Top Magn Reson Imaging, 3 (1991) 34-49. P. Mansfield, M. Bencsik, Fluid flow in porous systems, Magn Reson Imag, 16 (1998) 451-4. R.B. Miles, W.R. Lempert, Quantitative flow visualization in unseeded flows, Annu Rev Fluid Mech, 29 (1997) 285-326. F.W. Wehrli, Time-of-flight effects in MR imaging of flow, Magn Reson Med, 14 (1990) 187-93. A. Constantinesco, J.J. Mallet, A. Bonmartin, C. Lallot, A. Briguet, Spatial or flow velocity phase encoding gradients in NMR imaging, Magn Reson Imag, 2 (1984) 335340. P.M. Pattany, J.J. Phillips, L.C. Chiu, J.D. Lipcamon, J.L. Duerk, J.M. McNally, S.N. Mohapatra, Motion artefact suppression technique (MAST) for MR imaging, J Comput Assist Tomogr, 11 (1987) 369-377. P.J. Keller, F.W. Wehrli, Gradient moment nulling through the Nth moment. Application of binomial expansion coefficients to gradient amplitudes, J Magn Reson, 78 (1988) 145-149. J.G. Pipe, T.L. Chenevert, A progressive gradient moment nulling design technique, Magn Reson Med, 19 (1991) 175-179. P.R. Moran, A flow velocity zeugmatographic interlace for NMR imaging in human, Magn Reson Imag, 1 (1982) 197-203. F. St~hlberg, B. Nordell, A. Ericsson, T. Greitz, B. Persson, G. Sperber, Quantitative study of flow dependence in NMR images at low flow velocities, J Comput Assist Tomogr, 10 (1986) 1006-1015. G.T. Gullberg, S. Margaret, and F.W. Wehrli, A mathematical model for signal from spins flow during the application of spin echo pulse sequences, Magn Reson Imag, 6 (1988) 437-461. A. Caprihan, and E. Fukushima, Flow measurements by NMR, Physics Reports (Review Section of Physics Letters), 198 (1990) 195-235. J.M. Pope, S. Yao, Quantitative NMR imaging of flow, Concepts in Magnetic Resonance, 5 (1993) 281-302.

120


[34]

[351 [36] [37] [38] [39] [40]

[41] [42] [431 [44] [45]

N.J. Pelc, G. Sommer, K.C.P. Li, T.J. Brosnan, R.J. Herfkens, D.R. Enzmann, Quantitative magnetic resonance flow imaging, Magnetic Resonance Quarterly 10 (1994) 125-147. D.C. Wilcox, Basic Fluid Mechanics. DCW Industries Inc., California, USA, 1997, pp. 214. R.M. Henkelman, and M.J. Bronskill, Artifacts in magnetic resonance imaging, Rev Magn Reson Med, 2 (1987) 1-126. D. L. Parker, G.T. Gullberg, P.R. Frederick, Gibbs artifact removal in magnetic resonance imaging, Med Phys, 14 (1987) 640-5. S. Amartur, E.M. Haacke, Modified iterative model based on data extrapolation method to reduce Gibbs ringing, J Magn Reson Imaging, 1 (1991) 307-17. C.N. Chen, V.J. Sank, S.M. Cohen, D.I. Hoult, The field dependence of NMR imaging I. Laboratory assessment of signal-to-noise ratio and power deposition, Magn Reson Med, 3 (1996) 722-9. A. Macovski, Noise in MRI, Magn Reson Med, 36 (1996) 494-7. P.G. Walker, G.B. Cranney, M.B. Scheidegger, G. Waseleski, G.M. Pohost, A.P. Yoganathan, Semiautomated method for noise reduction and background phase error correction in MR phase velocity data, J Magn Reson Imaging, 3 (1993) 521-30. R.L. Wolf, R.L. Ehman, S.J. Riederer, P.J. Rossman, Analysis of systematic and random error in MR volumetric flow measurements, Magn Reson Med, 30 (1993) 8291. C. Tang, D.D. Blatter, D.L. Parker, Accuracy of phase-contrast flow measurements in the presence of partial-volume effects, J Magn Reson Imaging, 3 (1993) 377-85. C. Tang, D.D. Blatter, D.L. Parker, Correction of partial-volume effects in phasecontrast flow measurements, J Magn Reson Imaging, 5 (1995) 175-80. R.M. Hoogeveen, C.J. Bakker, M.A. Viergever, MR phase-contrast flow measurement with limited spatial resolution in small vessels: value of model-based image analysis, Magn Reson Med, 41 (1999) 520-8.

APPENDIX

This appendix shows the derivation of the linear relationship between the percent flow rate in each velocity bin and the mean velocity for a fully developed laminar, circular tube flow from the Poiseuille's equation. From the Poiseuille's equation, we have:

vi

2Q

=~(R n.R4

2 -r/2)

(A.1)

where v~ is the axial velocity at radius, r~ of the tube, Q is the total flow rate in the tube, and R is the inner radius of the tube. Rearranging Eq. (A. 1), we have:

121


/

V i ;r~ 4

t~ = aiR 2 V

(A.2)

2Q

The concentric flow area, Ai of a laminar tube flow at radii, r~ and Fi+1 is:

Ai --;rg'(ri21--ri 2 )

(h.3)

Substituting Eq. (A.2) into Eq. (A.3) and simplifying, we have:

;rt.2R4 A i "- ~ ( v

20

i

-Vi+l)

(m.4)

From our 2DFT technique, the velocity bin size, ( v i - Vi+l) is a constant. Therefore, A; is independent of (vg- V;+l) and is a constant for a given R and Q from Eq. (A.4). The flow rate, Q; at a given concentric flow area, A; is given by:

Qi =(v)iAi

(A.5)

where ; is the mean velocity corresponding to A; and is given by: 1

(v)i = --~(vi + vi+, )

(A.6)

Eq. (A.5) shows that Qi is linearly proportional to ;. The percent flow rate, %Q; in each velocity bin, (v;- V~+l) is determined by dividing Eq. (A.5) with Q and then multiplying by 100 percent:

=~

Q

x 100%

(A.7)

Therefore, from Eq. (A.7), the percent flow rate in each velocity bin, %Q is linearly proportional to the mean velocity, ; where A~ is a constant for our 2DFT technique, and Q is a constant for a fully developed laminar, circular tube flow.

122

Membrane Contactors and Environmental Applications

123



Chapter 6

Industrial applications and opportunities for membrane contactors R. Klaassen, P.H.M. Feron, R. van der Vaart, A.E. Jansen

TNO Institute of Environmental Sciences, Energy Research and Process Innovation, Department Chemical Engineering PO Box 342, 7300 AH Apeldoorn, The Netherlands phone: +31 55 493196 fax: +31 55 5493410 email: [email protected] SUMMARY

In a membrane contactor membrane separation is fully integrated with another separation technology, like extraction or absorption, in order to exploit the benefits of both technologies fully. Membrane contactor applications can be found both in water and gas treatment. Several recently developed applications that have been introduced in industry - aromatics recovery from a process water, selective removal of heavy metals from a galvanic process bath and ammonia recovery from an off gas s t r e a m - and new applications under developmentCO2 membrane gas absorption and membrane stripping- will be discussed technically and economically to show the wide scope of the technology. 1.

INTODUCTION

Stringent demands are put onto new separation technologies due to stricter product quality requirements, environmental legislation, and energy efficiency demands and needs for cost reduction. In order to meet these ever increasing needs there is a tendency to combine processes to a hybrid process. In a membrane contactor a membrane separation is not only combined with an extraction or absorption process but both processes are fully integrated and incorporated into one piece of equipment. In this way advantages of both processes can be fully exploited.The membrane offers a flexible modular energy efficient device with a high specific surface area. The extraction or absorption process can offer a very high selectivity and a high driving force for transport even at very low concentrations. Applications of membrane contactors can be found in gas and water

125

Industrial Applications And Opportunities For Membrane Contactors - Klaassen

treatment. In the membrane gas absorption process the gas stream to be treated is brought very efficiently into contact with an absorption liquid in the membrane contactor. Typical examples are CO2 removal from flue gas or indoor air, flue gas desulphurisation, indoor air conditioning, ammonia recovery from off gasses or mercury removal from natural gas. In the pertraction process water is treated in a membrane contactor with an extraction liquid. Typical applications are hydrocarbons (e.g. aromatics, phenol, haloginated hydrocarbons) removal and recovery from waste and process water and selective recovery of heavy metals from e.g. galvanic bath liquids or waste water. In a membrane contactor also ammonia can be removed and recovered from waste water in a combined stripping and absorption process. Many different developments in the field of membrane contactor technology are going on worldwide. Development of a specific application normally starts with a proof of principle followed by a feasibility study, further development, pilot plant tests and finally full-scale demonstration on site. After a general introduction on membrane contactors some examples of successful developed full-scale industrial membrane contactor installations will be discussed in this paper. These industrial reference installations help to gain confidence of end users in the membrane contactor technology. Thus further market introduction will be facilitated as well as development of new applications for membrane contactors. 2.

MEMBRANE GAS ABSORPTION

2.1

Principle

Membrane gas absorption is a gas-liquid contacting operation [ 1,2]. The key element in the process is a microporous hydrophobic hollow fibre membrane. The process is illustrated in figure 1 for removal of component X from a gas stream. The gas stream is fed along one side of the membrane where an absorption liquid is flowing at the other side of the membrane. The hydrophobic membrane wall keeps gas phase and absorption liquid separated from each other. The absorption liquid is chosen in such a way that it has a high affinity for component X. Component X will now diffuse through the gas filled pores of the membrane to the other side of the membrane where it is absorbed in a liquid phase. Absorption in the liquid phase takes place either by physical absorption or by a chemical reaction. This determines the selectivity of the process. The membrane used gives no contribution to the selectivity: the membrane's role is to keep the two phases separated and to provide a large contacting surface area.

126


Hollow fiber

gas

=,.,ou,v,u,, liquid ~

gas

Membrane

Figure 1: Membrane gas absorption For operation of membrane gas absorbers it is essential that liquid phase and gaseous phase do not mix. This means that the absorption liquid is not allowed to enter the pores. This is influenced by pore size, pressure difference across the membrane and the interaction of the absorption liquid with the membrane material. This is described by the Laplace equation: d P = - 2 (y/r) . c o s (~o)

( 1)

where: d P - pressure difference between liquid and gas [Pa] 7 - surface tension of the absorption liquid [N/m] r - pore radius [m] (p - contact angle for liquid and membrane material The membrane pores will not be wetted if the contact angle is greater than 90 ° and the pressure difference is limited for a given pore size. Suitable membrane materials for aqueous absorption liquids are non-polar polymers such as polypropylene, polyethylene and Teflon. 2.2

Membrane module The design of a membrane absorber cannot be based on conventional

127


available hollow fibre membrane modules. These module designs have historically been based on the use of filtration duties and consist of a bundle of hollow fibres in a tubular housing. This gives ill-defined flow conditions on the outside of the fibre bundle. When membranes are used for contacting duties it is necessary that flow conditions are well def'med on both sides of the membrane to achieve good mass transfer. Furthermore it is essential that, given the often large volume flows of flue or off gas streams, the module design gives low pressure drops and is easy to scale up. TNO has patented a new membrane module design to overcome the problems with conventional membrane modules. Figure 2 shows this new membrane module design [3] for contacting duties. The module consists of a rectangular housing with fibres at well-def'lned positions. The gas flow is perpendicular to the fibres. The absorption liquid is fed through the fibres. Elements of the module can be coupled parallel and in series. In this way the module can be easily adjusted to the flow rate of the gas stream to be treated and the desired removal efficiency.

inlet

outlet

inlet

Figure 2: New membrane module

Important features of the module are: well defined flow conditions on both sides of the membrane, easy scale up, high mass transfer, low pressure drop and high specific surface area. Specific surface area's of well over 1000 rn2/m3 can be reached. 2.3

Membrane gas absorption versus conventional contactors

Conventional G-L contactors, such as packed towers, preferably operate

128


with relatively large gas flows under steady-state conditions. Membrane gas absorption can be an eminently suitable alternative for 'smaller' gas flows and operation at variable loads. Especially by using hollow fibre membranes it is possible to develop a very compact gas-liquid contactor. Hollow fibre membrane modules combine a high specific surface area, a well-def'med gas flow regime and short diffusion ways in a modular piece of equipment. The membrane gas absorption technology is able to treat gas streams with extreme variations in gas flow and or concentrations of the components to be removed. This is possible at low costs in very compact equipment. The reduction in volume and weight compared to a conventional absorber can be more than a factor 10. Through the use of hollow fibre membranes the membrane gas absorption process can offer operational and economical advantages over conventional spray towers or packed columns. Operational advantages include independent Gas/Liquid control, flexible operation, optimal load of the absorption liquid, no entrainment, flooding or foaming and modular low weight and very compact equipment. Economical advantages include low investment costs, low pumping power for the absorption liquid and the fact that no expensive civil engineering work is required. That gas absorption membranes can result in very compact equipment can be illustrated by the following example: For a coal fired power plant of 645 MWe a design has been made for a conventional absorber (spray tower) and a membrane gas absorption unit. A flue gas flow of 2"106 nm3/hr has to be treated in this case. The calculated overall volumes of the absorbers are given below. Conventional SO2 absorber Calculated membrane gas absorption

2.4

- volume 9000 m 3 - volume 250 m 3

Applications

Membrane gas absorption can be applied for removal of those components from a gas stream where a suitable absorption liquid is available. A suitable absorption liquid has a high affinity for the components to be removed and does not wet the membrane. The majority of absorbents used in conventional gas absorption processes can also be used for membrane gas absorption. Applications [4,5] for the use of membrane gas absorption are listed below. Flue-gas and off-gas containing e.g. SO2, HC1, NH3 or HES can be treated to meet emission standards in many industrial situations. Recovery and re-use of CO2 from flue gas, biogas and off-gases in horticultural industry, beverage production and other industrial

129


applications. Upgrading and desulphurisation of biogas from anaerobic digesters and landfills. Acid gas removal and dehydration of natural gas. Acid gas removal from fuel gas mixtures. Mercury removal from natural gas, flue gas or glycol overheads. Alkene-alkane separation in petrochemical industry. Indoor air treatment for e.g. tobacco smoke components. Due to the modular equipment used in membrane processes application of gas absorption membranes is also possible on a small scale. This is particularly valuable for applications in the field of environmental technology, as it enables an efficient and economical treatment of small-scale emission sources. If regeneration of the absorbent cannot be done one site, central regeneration is another possibility. In paragraph 2.5 the recovery of ammonia with membrane gas absorption is discussed. In paragraph 2.6 developments with respect to CO2 absorption are shown. Another industrial application of membrane gas absorption is the transfer of unstable C102 to potable water [27]. Also the recovery of VOC's from a paint booth exhaust is recently shown under minipilotplant conditions [28].

2.5.

Membrane gas absorption installation for ammonia recovery An industrial membrane gas absorption unit for ammonia recovery has been installed at a dyes intermediates production plant of Aliachem in Pardubice the Czech Republic (Figure 3). In this plant dyes intermediates are produced batch wise in pressurised reactors. Ammonia is used as a reactant in the reactor. Depressurisation of the reactor results in an ammonia containing off gas stream, which gives a significant loss of ammonia. In the membrane gas absorption unit the total off gas stream is treated with water as absorption liquid. The ammonia is recovered as an aqueous ammonia solution of > 20 wt% which can be re-used in the dyes intermediates production process. The membrane gas absorption unit is successfully in operation since June 1999; aqueous ammonia solutions of 27 wt% ammonia have been produced and the emission of ammonia to the environment is reduced with 99.9%. The installation has a capacity to absorb 50 kg/hr ammonia and has proved to be very easy in operation. Variation in gas flow rate or ammonia concentration can be handled without any problems.

130


Figure 3: MGA unit on site in Pardubice, Czech Republic 2.6

C02 removal

2.6.1 Introduction CO2 membrane gas absorption has been under development at TNO for a number of years. The potential of this technology has been identified in a number of assessment studies [6,7]. The ensuing customer-driven development has been focusing on the production of carbon dioxide from flue gas with the aim of supplying high purity carbon dioxide to the horticultural industry to increase crop yield [8,9]. Other commercial applications are in the field of life support (spacecrafts, submarines, operating theatres, consumer products, biogas treatment). In recent years there has also been an increased interest in CO2-removal and subsequent storage in view of the enhanced greenhouse effect. Figure 4 shows an overview of the atmospheric pressure applications currently being investigated. The process is based on novel membrane absorption method patented by TNO [10], which employs dedicated aqueous absorption liquids, called CORAL, based on amino-acids and alkaline salts. Benefits of these liquids over conventional amines are: - Stable operation with cheapest membranes available viz. polyolefin membranes. - Oxygen stability. - Improved corrosion resistance. - No vapour pressure, hence no loss of absorption liquids through evaporation.

131


CO~-MGA application range 100 --0-

x "

~

• O~ (J

~

--

Greenhouse Blogas Submarine Space Medical Consumer

0.1 --o--

0.01 0.001

1

1000

Capacity [kg C021h ]

Figure 4:CO2 membrane gas absorption applications (atmospheric pressure) currently under development.

2.6.2 C02-production for greenhouses In the Netherlands carbon dioxide is used to promote plant growth in greenhouses. Depending on the crop, the production can be increased by 25% by increasing the CO2-concentration to 500 ppm. In temperate climates it is necessary to heat these greenhouses, which makes the horticultural industry in the Netherlands a large consumer of natural gas. Future heat supply systems will be largely based on cogeneration plants (gas engines and gas turbines) as significant energy savings can be achieved. Energy savings can be increased if the heat supply is coupled to carbon dioxide supply. This will incite growers to buy heat from the cogeneration plant rather than to use their own boilers. These boilers are needed to cover peak demand and emergencies. Also the amount of available carbon dioxide per unit heat will be larger in case of combined production of heat and electricity, thus allowing for an improved production. The heat demand and CO2 demand are generally anti-cyclic. During periods of high heat demand (during winter and at night) little or no CO2 is needed to maintain the desired CO2 levels in the greenhouse. During periods of high CO2 demand (during summer and during the day) little or no heat is needed. The thermal energy of the cogeneration plant can either be used for greenhouse heating or for the production of carbon dioxide. This will also extend the operating time of the cogeneration plant to daytime hours, which is an additional benefit in terms of capacity credits. A schematic drawing of a cogeneration plant supplying heat and carbon dioxide to greenhouses is shown in Figure 5.

132

IndustrialApplicationsAndOpportunitiesForMembraneContactors- Klaassen

Greenhouses Flue gas

I

ico2 (loo°/o) ,

l

H.t

N,ura,

Electricity 1

T Natural gas

Grid

Figure 5: Cogeneration plant supplying heat and carbon dioxide to greenhouses

2.6.3 Pilot plant Pilot plant experiments were carried out to achieve the following objectives: - Assessment of mass transfer under a variety of conditions with several liquids, - Assessment of heat consumption, - Assessment of CO2-quality. During the feasibility study a pilot plant has been designed and built for testing in the lab as well as for testing on location with a real flue gas. The basic specifications of the pilot plant are shown in table 1. Table 1: Basic specifications of COz-removal pilot plant CO2-concentration 3.5 % Gas flow rate 25 m3/hr CO2-removal 80% Liquid flow rate 20-1001/hr The pilot plant flow sheet is shown in Figure 6. In case of the lab experiments CO2 and air are mixed via mass flow controllers and fed to a prototype cross-flow membrane absorber (DAM-module). In this absorber is CO2 is transferred from the gas stream to a liquid stream in a counter current fashion. On location a small stream has been taken from the chimney, cooled down and pumped directly to the membrane absorber. The rich liquid is fed to the top of a stripper via a plate heat exchanger. The stripper temperature is maintained at a temperature of around 120°C by an electric heating element. As a result of the temperature increase of the absorption liquid CO2 is liberated from the solution. Water is recovered in the condensor and trickles back into the stripper. The lean liquid exits the stripper at the bottom and is fed back to the membrane absorber via the heat exchanger. Temperatures, pressures, dew

133

IndustrialApplicationsAndOpportunitiesForMembraneContactors- Klaassen

points, flow rates and heat consumption have been measured and are used for the performance assessment. C02

Steam Gas in

Air r v

~Mi i

mo Ju

Rich I exchanger I =' I Heat =I

I "~

! Stripper

Lean

Cooler

~

Heatero

20 wt%) to be reused in the dyes intermediate production process. The unit in operation since June 1999 has reduced NH3 emission to the environment by 99% and currently has a capacity to absorb 50 kg/hr. The volume and weight of the membrane absorbers were more than 10 times smaller than conventional absorbers. Additional applications being exploited include SO2 absorption [38]. The above applications involve liquids which are primarily aqueous and under ordinary circumstances many not wet the hydrophobic microporous membrane. Organic liquids (e.g. silicone oil) which are substantially nonvolatile and spontaneously wet the pores of such fibers have been

154

Membrane Contactors:

Recent Developments -

Sirkar

successfully employed as an absorbent for the removal of VOCs from air streams in LiquiCel Extraflow ® membrane contactors under minipilot plant conditions using an actual paint booth exhaust stream [39]. This effort followed the laboratory scale studies of Poddar et al. [40] where they had used both porous membrane as well as a porous membrane with a nonporous silicone coating for absorption of VOCs in an oil. Stripping of absorbed gases from the liquid absorbent at a higher temperature with or without vacuum or an inert gas (e.g. N2) / vapor (steam) can also be implemented using membrane contactors. If a higher temperature is involved, special attention has to be paid to the materials of the membrane and the device. Efforts are ongoing with the microporous Goretex PTFE microporous membranes to strip the CO2 from the amine-absorption solution at a higher temperature [41]. Stripping of absorbed VOCs from the silicone oil absorbent was successfully implemented in a minipilot plant scale using vacuum and hydrophobic microporous polypropylene membranes having an ultrathin nonporous plasmapolymerized silicone membrane coating facing the silicone oil [39]. The coating was highly permeable to the absorbed VOCs but was impermeable to the silicone oil absorbent. Laboratory studies [9] carried out similar stripping successfully at a much higher temperature of 50-70 °C. The first successful laboratory studies of membrane distillation were by Gore [42]. Schneider et al. [43] studied the process of direct contact membrane distillation (DCMD) using larger units made from hydrophobic PP hollow fibers. Lawson and Lloyd [44] had demonstrated using very small fiat membranes that very high water vapor flux is possible in desalination of water by DCMD as much as 2-3 times that in reverse osmosis. However, this DCMD process is plagued by water vapor flux reduction and distillate contamination by brine with time due to wetting of the pores of the hydrophobic porous PP membranes normally employed. Further, high water vapor fluxes have to be achieved in a compact but high surface area contactor device. Vacuum membrane distillation process is not favored as such due to the need for a separate condenser. In DCMD, a hot saline solution produces water vapor which diffuses through the gas filled pores of a porous hydrophobic membrane and then condenses in the cold distillate water stream on the other side of the membrane. The water vapor pressure difference on the two sides of the membrane is the driving force through the gas in the pores, the so called "gas membrane" or the "supported gas membrane" (SGM). When there is a volatile species in an aqueous solution and an aqueous solution without the volatile species on the other side, then a porous hydrophobic membrane may be used to desorb the volatile species from the feed solution and get it absorbed in the receiving solution (for earlier references, especially the studies by Edward Cussler, see [2]). Chlorine dioxide (C102) has been stripped from an aqueous electrolytic 155

Membrane Contactors: Recent Developments - Sirkar

solution of an electrochemical reactor through a highly chemically resistant porous hydrophobic membrane stack and absorbed in the potable water to be disinfected and treated; large-scale plant operation has successfully taken place [45]. Osmotic Distillation is said to occur when the receiving solution on one side of the gas membrane is a highly concentrated saline solution having a vapor pressure of water which is much less than that of an aqueous feed solution on the other side of the membrane. Both solutions are at the same temperature just as in the C102-separation example above and unlike that in DCMD. Consequently, there is a partial pressure-based driving force for water transport from the feed to the receiving solution. Large-scale tests using two 19.2 m 2 LiquiCel ® modules and continuous production have been carried out for fruit juice concentration using such a technique [46]. Wetting of the pores due to hydrophobic components in the feed solution can be prevented by having a hydrogel layer on the surface of the hydrophobic membrane facing the feed fruit juice solution [ 11 ].

4.2

Liquid-Liquid Contactors Membrane-based liquid contactors are primarily employed to extract a nonionic or ionic solute selectively from one liquid phase into another immiscible liquid phase. The phases often encountered in an immiscible twophase system are aqueous and organic. (References to immiscible phase pairs which are organic-organic or biphasic aqueous are available in [1]). A large number of studies have been conducted exploring membrane solvent extraction of metals, organic acids, organic pollutants, antibiotics etc. from an aqueous solution into an organic extractant. Back extraction into an aqueous phase has also been conducted especially for metals, organic acids, antibiotics etc. A number of pilot plant studies were implemented. Almost all such studies employ Celgard hydrophobic microporous PP hollow fibers/Liquicel ® modules with the organic solvent phase in the pores and the aqueous phase outside at a higher pressure [47]. The Liquicel ® modules have the tube sheet made out of a lower Tg olefinic resin and have considerable solvent resistance. However, a number of solvents including chlorinated solvents, dimethyl formamide (DMF), dimethylsulfoxide (DMSO) etc. are not to be used. Back extraction studies of the solute into an aqueous phase also employ the same type of hydrophobic membranes/modules even though hydrophilic fibers with the aqueous phase in the pores is much more efficient from a mass transfer perspective [ 1]. A full-scale membrane-based solvent extraction plant has been operating since 1998 at Kosa, Vlissingen, Netherlands [37]. It employs Liquicel ® modules and an organic feedstock to extract an aromatic compound formed in a reactor

156


and being discharged in an aqueous stream. The organic feed stock recycles the product back into the reactor. The waste water volume treated is 15 m3/hr. The waste water incineration facility was eliminated resulting in an annual saving of 5 million m 3 of natural gas. An additional example of large-scale hydrophobic microporous membrane-based solvent-extraction involves extraction of Cr 6+ from an aqueous solution into an organic extractant containing a secondary amine present in the pores of a Celgard fiber. The aqueous phase is maintained at a higher pressure than the organic extractant phase to immobilize the aqueous-organic interface in this plant operated by Commodore Separation Technologies, Kennesaw, GA [48,49]. However, there is a novelty in this process wherein the organic extractant phase has dispersed in it drops of the aqueous back extraction liquid which strips the Cr6÷ from the complex into a highly basic stripping solution. A similar technique is being used in large scale for removing zinc and iron from a passivating bath at Rogal, Enschede, The Netherlands [37]. A large enantiomer resolution plant at Tanabe Seiyuku, Osaka, Japan employs lipase enzyme-based resolution of the enantiomers present in an organic stream; the enzymes are reversibly immobilized in the pores of substrate of a hydrophobic asymmetric polyacrylonitrile (PAN)-based hollow fiber ultrafiltration membrane. The skin of this fiber is in the I.D.; an aqueous solution flows through the fiber bore. The aqueous-organic interface is immobilized on the fiber O.D. with the shell-side organic phase outside the pores maintained at a pressure higher than that of the aqueous phase. Total membrane area employed is 1440 m2. This technology is from Sepracor, Marlborough, MA. [50]; however the microporous membrane-based solvent extraction aspect of it was licensed from Hoechst Celanese Inc. (Current successor is Celgard Inc., Charlotte, NC). Additional techniques are being developed wherein liquid-liquid extraction takes place at one side of a porous membrane whereas a sweep gas/vacuum is maintained on the other side. It is called supported liquid membrane pervaporation (SLMPV) [51]. Highly stable technologies are currently being developed in our laboratory since conventional SLMs are generally considered unstable.

4.3

Supereritieal Fluid-based Contactors PoroCrit LLC, Berkeley, CA has started using Celgard fiber-based Liquicel ® contactors and other porous PP fiber-based devices to achieve species transfer from a liquid phase into supercritical CO2 or vice versa. They claim that "the process is conducted with the pressure on both sides of the membrane in the module being essentially the same" [52]. As the two phases flow in a countercurrent fashion in the PP hollow fiber module, they will undergo considerable pressure drops. So there will be considerable pressure difference 157

M e m b r a n e Contactors: Recent D e v e l o p m e n t s - Sirkar

between the two phases along the module length. Such an operation has been successfully used to extract flavors from wines and juices, recover aromas from vegetable and marine oils and nuts and fermentation broths and for industrial organic solvent recovery and recycle [53]. 4.4

Adsorption-based Contactors

When a gas or liquid phase containing a solute is contacted with a packed bed of adsorbents/ion exchange resin beads, adsorption or ion exchange-based transfer of solutes occur to the bed particles. There is significant~igh pressure drop in flow through packed beds, especially of fine particles. Further there is considerable resistance to diffusion of solute molecules (especially larger ones) through the porous resin beads; as a result, the utilization of the ligands etc. in the porous resin beads is low. Porous membranes used in the microfiltration separations and having pores in the range of 0.02-10 gm have been used as a substrate: ligands/biofunctional groups of appropriate types have been grafted onto the surfaces of pores in microfiltration membranes. The liquid phase containing solutes has been passed through the pores in the microfiltration mode of operation. The liquid convection through the membrane pores conveniently brings the solutes directly to the ligands on the pore surfaces eliminating the diffusional resistance encountered in the conventional resin beads. The process of adsorption/complexation takes place throughout the pore volume. Fractional ligand utilization is very high. The solute uptake process takes place rapidly. After some time, the feed flow is stopped and the ligands on the pore surfaces are regenerated by passing an eluent solution since they get saturated. Such devices are called membrane adsorbers but are in effect, liquidsolid contactors. Their application to protein separation is commercially practiced, albeit in a smaller scale. An earlier reference to the literature is available in Thommes and Kula [5]. Such a technique has now been adopted to heavy metal adsorption from waste waters with or without disposable membrane materials [54]. Disposable and cheap membrane materials (e.g. of cellulose) are useful especially if regeneration efficiency for the heavy metals is poor. 4.5

Contactors as Reactors

Whenever a reactor employs two immiscible phases which have to be contacted intimately, a membrane contactor may be employed. The two immiscible phases will flow/be on the two sides of the membrane and appropriate phase pressure differences have to be maintained for nondispersive operation. The reaction may take place in the bulk of one or both phases, at the interfaces on the pore mouths and/or in the fluid layers immediately adjacent to the phase interface [20]. Membrane-based ozonators developed by W.L.Gore and Pall Inc. are commercialized examples of such reactors: ozone-containing oxygen/air flows

158

Membrane Contactors: RecentDevelopments- Sirkar

on one side of the membrane and the water to be treated flows on the other side. The materials of the device are chemically resistant perfluoropolymers. The membranes provide very little resistance to gas transport but do not usually allow liquid to come to the other side of the membrane. This area has an enormous potential for growth in the near future. The limitations are: selection of an appropriate example amenable to reaction in a membrane contactor; chemical resistance of the membrane material; controlling the thermal effects; flow maldistribution etc. Additional examples of membrane contactors employed as reactors are identified below: - Destruction of VOCs in a gas phase by their absorption in a fluorocarbon phase which is supplied with ozone from a gas phase via a separate membrane [55]. - Elimination of mercury from a gas phase by oxidative membrane absorption in an oxidizing aqueous liquid solution of H202/H2SO4 or K2Cr207 o r K 2 S 2 0 8 , K M n O 4 etc. [56]. The review paper by Sirkar et al. [57] provides additional examples especially in the production of chemicals/biochemicals. Amongst these examples are cases of hydrogenation of organic liquids via a porous ceramic membrane reactor which is essentially a membrane contactor. Porous ceramic membranes hold special promise especially due to their excellent thermal resistance. Membrane contactors have also been employed by Porocrit Co. with supercritical CO2 to sterilize juices and liquids by destroying microorganisms and inactivating enzymes [52]. 5.

MASS TRANSFER CORRELATIONS AND CONSIDERATIONS

The resistance to mass transfer of a species in a hollow fiber-based membrane contactor is the sum of three resistances: lumen side, pore and shell side. This can be represented as 1

=

1

1

1

~ + ~ + ~ (2) gtotat kl .... kpore kshett where Ktotal is the overall mass transfer coefficient, klumen, kpore, kshell are the local mass transfer coefficients on the lumen side, membrane pore, and shell side respectively. Depending on the application, one or more of these resistances may become negligible compared to the others. For example, for gas-liquid systems, gas phase resistance is usually negligible compared to liquid phase resistance. So, the side where the gas flows will not offer much resistance to mass transfer and almost all of the resistance is offered by the liquid side. For membranes wetted by the liquid phase, resistance offered by the liquid in the pore is constant and does not change with the hydrodynamics.

159


In hollow fiber membrane contactors, lumen side mass transfer is usually well defined whereas the shell side mass transfer coefficient is a strong function of module geometry, fiber distribution and the resulting hydrodynamics. Recent efforts to understand the effect of shell side flows on the membrane contactor performance [31,58-60] highlight the complex nature of shell side hydrodynamics. In addition, these efforts may lead to a scaleup parameter for designing membrane contactors for different conditions. While membrane contactors are commercially available with different membrane areas [22] for handling a range of flow volumes, the blood oxygenation devices come in a limited number of sizes. Each device is presumably optimized for the required shear rates, pressure drops etc. 6.

FUTURE DIRECTIONS

Membrane contactors are finding increasing commercial acceptance in recent years. This trend is expected to continue with new applications attempted using membrane contactors. The successful use of membrane contactors will depend on one or more of its unique advantages (small footprint, flexible capacity, broad range of phase flow rates/ratios without associated problems, chemical resistance of materials etc.) which are not offered by conventional contacting equipment. The main class of applications where membrane contactors are used still appears to be for effective gas-liquid contacting in a nondispersive manner using aqueous streams. New applications in gas-liquid contacting and other areas are gaining ground. New materials and geometries are being used for making microporous substrates and membranes. The new materials are pursued mainly for their non-wetting properties and chemical resistance. Characterization of membrane contactors for their mass transfer is still empirical. A better understanding of the hydrodynamics (particularly shell side) and ways to quantify it in terms of mass transfer correlations is needed for effective scaleup of these contactors. Similarly, designs of membrane contactors for conventional applications are limited in number. As the range of applications grows, more designs are to be expected to cater to various aspects of contacting applications. 7.

ACKNOWLEDGEMENT

One of the authors (ASK) wishes to acknowledge many discussions with Dr. P.V. Shanbhag during the preparation of this manuscript.

160


REFERENCES [1]

Prasad, R. and K. K. Sirkar, Membrane-based Solvent Extraction, in Membrane Handbook, W.S.W. Ho and K.K. Sirkar (Eds.), Chapman and Hall, New York (1992) Chap 41.

[2]

Sirkar, K.K., Other New Membrane Processes, in Membrane Handbook, W.S.W. Ho and K.K. Sirkar (Eds.), Chapman and Hall, New York (1992) Chap 46.

[3]

Reed, B.W.; Semmens, M.J.; and Cussler, E.L. Membrane Contactors in Membrane Separations Technology, Principles and Applications, R.D. Noble and S.A. Stem (Eds.), Elsevier Sci., (1995) 468-498.

[4]

Gableman, A.; Hwang, S.-T. Hollow Fiber membrane Contactors, J. Membr. Sci., 159 (1999) 61-106.

[5]

Thommes, J. and M.R. Kula, Membrane Chromatography- An Integrative Concept in the Downstream Processing of Proteins, Biotechnol. Prog. 11 (1995) 357.

[6]

Myers, D., Surfaces, Interfaces and Colloids. Principles and Applications, VCH Publishers, New York, (1991) 349-356.

[7]

Prasad, R. and K.K. Sirkar, Hollow Fiber Solvent Extraction of Pharmaceutical Products: A Case Study, J. Membr. Sci., 47 (1989) 235-259.

[8]

Kim, B.-S. and P. Harriott, Critical Entry Pressure for Liquids in Hydrophobic Membranes, J. Colloid and Interface. Sci. 115 (1987) 1-8.

[9]

Xia, B., S. Majumdar and K.K. Sirkar, Regenerative Oil Scrubbing of Volatile Organic Compounds from a Gas Stream in Hollow Fiber Membrane Devices, Ind. Eng. Chem. Res. 38 (1999) 3462-3472.

[10]

Mansouri, J. and A.G. Fane, Membrane Development for Processing of Oily Feeds in IMD, Presented at the Workshop on Membrane Distillation, Osmotic Distillation and Membrane Contactors, Cetraro, Italy, July 2-4, 1998.

[11]

Mansouri, J. and A.G. Fane, Osmotic Distillation of Oily Feeds, J. Membr. Sci., 153 (1999) 103-120.

[12]

Zisman, W.A., "Relation of the Equilibrium Contact Angle to Liquid and Solid Constitution", in "Contact Angle, Wettability and Adhesion", Advances in Chemistry Series 43, ACS, Washington, D. C. (1964) 1-51. p 32.

[13]

Falk-Pedersen, O. and H. Dannstrom, Method for Removing Carbon Dioxide from Gases, US Patent 6,228,145, May 8, 2001.

[]4]

Shanbhag, P.V., J.Cissel, M. Gramerb and S. M. Nemser, Bioreactor Oxygenation Applications with CMS Membrane Contactors, paper presented at American Institute of Chemical Engineers, 2000 Spring National Meeting, Atlanta, GA, March 5-9, 2000.

[15]

Stutman, M.B., P.V. Shanbhag, J.J. Bowser and S.M. Nemser, Membrane Contactors Featuring Non-wetting Perfluorinated Membranes for Efficient Gas/Degasification of Process Fluids, paper Presentated at the AIChE Annual Meeting, Reno, NV, November 8 (2001).

[16]

Guha, A.K. and K.K. Sirkar, A Novel Separation-Reaction-Separation System, Paper Presented at the AIChE Annual Meeting, Chicago, IL, Nov. 13 (1990) paper 58g.

161

M e m b r a n e Contactors: Recent D e v e l o p m e n t s - Sirkar

[17]

Guha, A.K., P.V. Shanbhag K.K. Sirkar, C.H. Yun, D.H. Trivedi and D.A. Vaccari, Novel Membrane-Based Separation and Oxidation Technologies, Waste Management 13 (1995) 395.

[18]

Shanbhag, P.V., A. K. Guha and K. K. Sirkar, Single-phase Membrane Ozonation of Hazardous Organic Compounds in Aqueous Streams, J. Hazard. Mater. 41 (1995) 95104.

[19]

Shanbhag, P.V. and K.K. Sirkar, Ozone and Oxygen Permeation Behavior of Silicone Capillary Membranes Employed in Membrane Ozonators, J. Appl. Polymer. Sci. 69 (1998) 1263-1273.

[201

Shanbhag, P.V., A.K. Guha and K.K. Sirkar, Membrane-Based Ozonation of Organic Compounds, Ind. Eng. Chem. Res. 37 (1998) 4388-4398.

[21] [22] [23] [241

Pall Corporation website, www.pall.com (2001). Celgard website, www.celgard.com (2001). Killer, A. and C. Bier, Tubing Unit, WO Patent 01/70375 Voorhees, M.E. and B.F. Brian, Blood-gas Exchange Devices, International Anesthesiology Clinics, 34(2) Spring 1996, 29-45.

[25]

Mueller, X.M., B. Marty, H.T. Teveaerai, P. Tozzi, D. Jigger and L.K. von Segesser, A Siliconized Hollow Fiber Membrane Oxygenator, ASAIO J., (2000) 38-41.

[26]

Kawahito, S., T. Maeda, T. Motomura, T. Takano, K. Nonaka, J. Linneweber, M. Mikami, S. Ichikawa, M. Kawamura, J. Glueck, K. Sato and Y. Nose, Development of a New Hollow Fiber Silicone Membrane Oxygenator: In Vitro Study, Artif. Organs, 25 (2001), 494-502.

[27]

Majumdar, S., A. Sengupta, J.S. Cha and K.K. Sirkar, Simultaneous SO2/NO Separation from Flue Gas in a Contained Liquid Membrane Permeator, Ind. Eng. Chem. Res., 33 (1994) 667-675.

[28]

Guha, A.K., C.H. Yun, R. Basu and K.K. Sirkar, Heavy Metal Removal and Recovery by Contained Liquid Membrane Permeator, AIChE J. , 40 (1994) 12231237.

[29]

Mandal, D., A.K. Guha and K.K. Sirkar, Isomer Separation by a Hollow Fiber Contained Liquid Membrane Permeator, J. Membr. Sci. 144 (1998) 13-24.

[30]

Sirkar, K.K., Hollow Fiber-Contained Liquid Membranes for Separations: An Overview, in Chemical Separations with Liquid Membranes, R.A. Bartsch and J.D. Way (Eds.), ACS Symp. Ser. 642, American Chemical Society, Washington, D.C., Chapter 16.

[311

Sengupta, A., P.A.Peterson, B.D.Miller, J.Schneider, C.W. Fulk Jr., Large-scale Application of Membrane Contactors for Gas Transfer from or to Ultrapure Water, Separation and Purification Technology, 14 (1998) 189-200.

[32]

Mahmud, H., A. Kumar, R.M. Narbaitz and T.Matsuura, A Study of Mass Transfer in the Membrane Air-Stripping Process using Microporous Polypropylene Hollow Fibers, J. Membr. Sci., 179 (2000) 29-41.

[33]

Hattler, B.G., G.D. Reeder, P.J. Sawzik, L.W. Lund, F.R. Waiters, A. Shah, J. Rawleigh, J.S. Goode, M. Klain and H. Borovetz, Development of an Intraveneous Membrane Oxygenator: Enhanced Intraveneous Gas Exchange through Convective

162


Mixing of Blood Around Hollow Fiber Membranes, Artif. Organs, 18(11) (1994) 806812. [34]

Sirkar, K.K. Membrane Separations: Newer Concepts and Applications for the Food Industry, in Bioseparation Processes in Foods, R.K.Singh and S.S.H. Rizvi (Eds.), Marcel Dekker, New York (1995) pp 351-374.

[35]

Peinemann, K.V., K. Ludtke, D. Fritsch and K. Kleine, Water Purification with Catalytically Active Membranes, paper I1-04-IN presented at the International Conference on Materials for Advanced Technologies, Singapore, July 1-6, 2001.

[36]

Qi, Z., and E.L. Cussler, Microporous Hollow Fibers for Gas Absorption. I Mass Transfer in the Liquid II. Mass Transfer Across the Membrane, J. Membr.Sci. 23 (1985) 321-340.

[37]

Klaassen, R. Jansen, A.E. The Membrane Contactor: Environmental Applications and Possibilities. Environ. Prog. 20(1) April (2001) 37-43.

[38]

Klaassen, R. , P.H.M. Feron and A.E. Jansen, Membrane Gas Absorption: Environmental Applications, Proceedings ICOM'96, pp 916, 1996.

[39]

Majumdar, S., D. Bhaumik, K.K. Sirkar and G. Simes, A Pilot-scale Demonstration of a Membrane-based Absorption-Stripping Process for Removal and Recovery of Volatile Organic Compounds, Environ. Prog., 20 (1) April (2001) 27-35.

[40]

Poddar, T.K., S. Majumdar and K.K. Sirkar, Membrane-based Absorption of VOCs from a Gas Stream, AIChE J., 42 (1996) 3267-3282.

[41]

Herzog, H. and O. Falk-Pedersen, The Kvaerner Membrane Contactor: Lessons from a case study in how to reduce captm'e costs, paper presented in 5th International Conference on Greenhouse Gas Control Technologies, Australia, 13-16 August, 2000.

[42]

Gore, D.W., Gore-Tex Membrane Distillation, in Proc. 10th Annual Convention of Water Supply Improvement Association, Honolulu, Hawaii, July 25-29 (1982).

[43]

Schneider, K., W. Holtz, R. Wollbeck and S. Ripperger, Membranes and Modules for Transmembrane Distillation, J. Membr. Sci. 39 (1998) 25.

[44]

Lawson, K. and D.R. Lloyd, Membrane Distillation. II Direct Contact MD, J. Membr. Sci. 120 (1996) 123.

[45]

Muttoo, T. and D. Dean, Development of Chlorine Dioxide Gas Transport Contactor (ERCO R 101 Generator) for Use in Drinking Water Disinfection, Paper 1-2 presented at the 12th Annual NAMS Meeting, Lexington, KY, May 16, 2001.

[46]

Hogan, P.A., R.P. Canning, P.A. Peterson, R.A. Johnson and A.S. Michaels, A New Option: Osmotic Distillation, Chem. Eng. Prog., July (1998) 49-61.

[47]

Kiani, A.K., R.R. Bhave and K.K. Sirkar, Solvent Extraction with Immobilized Interfaces in a Microporous Hydrophobic Membrane, J. Membr. Sci. 20 (1984) 125.

[48]

Ho, W.S.W. and T.K. Poddar, New Membrane Technology for Removal and Recovery of Chromium from Waste Waters, Environ. Prog., 20(1) April (2001) 44-52.

[49]

Ho., W.S.W. and T.K. Poddar, New Membrane Technology for Removal and Recovery of Metals from Waste Waters and Process Streams, Paper Presented at the Topical Conference on Membrane and Extraction Science and Technologies for Environemental Applications, AIChE 2000 Spring National Meeting, Atlanta, GA, March 5-9 (2000).

163


[50]

Lopez, J.L. and S.L. Matson, A Multi-phase Extractive Enzyme Membrane Reactor for Production of Diltiazem Chiral Intermediate, J.Membr. Sci., 125 (1997) 189.

[51]

Qin, Y., J. P. Sheth and K. K. Sirkar, Supported Liquid Membrane-Based Pervaporation for VOC Removal from Water, Ind. Eng. Chem. Res. 41 (2002) 34133428.

[52]

Robinson, J.R. and M.J. Sims, Method and System for Extracting a Solute from a Fluid Using Dense Gas and a Porous Membrane, U.S. Patent, 5,490,884, February 13 (1996).

[53]

Towsley, R.W., J.A.Turpin, M. J. Sims and J. R. Robinson, "Porocritical Fluid Extraction using COE for Industrial Organic Solvent Recovery and Recycle", Paper Presented at CISF-99, Fifth Conference on Supercritical Fluids and their Applications, June 13-19, Garda (Verona), Italy (1999).

[54]

Hestekin, J.A., L.G. Bachas and D. Bhattacharyya, Poly(amino acid) Functionalized Cellulosic Membranes: Metal Sorption Mechanisms and Results, Ind. Eng. Chem. Res. 40 (2001) 2668-2678.

[55]

Shanbhag, P.V., A.K.Guha and K.K. Sirkar, Membrane-based Integrated AbsorptionOxidation Reactor for Destroying VOCs in Air, Env. Sci. Tech. 30 (1996) 3435.

[56]

van der Vaart, R., J. Akkerhuis, P. Feron and B. Jansen, Removal of Mercury from Gas Streams by Oxidative Membrane Gas Absorption, J. Membr. Sci., 187 (2001) 151-157.

[57]

Sirkar, K.K., P.V. Shanbhag and A.S. Kovvali, Membrane in a Reactor: A Functional Perspective, Ind. Eng. Chem. Res., 38(10) (1999) 3715-3737.

[58]

Lemanski, J. and G.G. Lipscomb, Effect of Shell-side Flows on Hollow-Fiber Membrane Device Performance, AIChE J. 41 (1995) 2322-2326.

[59]

Lipnizki, F. and R.W. Field, Mass Transfer Performance for Hollow Fiber Modules with Shell-side Axial Feed Flow: Using an Engineering Approach to Develop a Framework, J. Membr. Sci. 193 (2001) 195-208.

[601

Schoner, P., P. Plucinski, W. Nitsch and U. Daiminger, Mass Transfer in the Shell Side of Cross Flow Hollow Fiber Modules, Chem. Eng. Sci., 53 (1998) 2319-2326.

164


Chapter 8

Membrane Aromatic Recovery System (MARS) - A new process for recovering phenols and aromatic amines from aqueous streams .

•

F. C. Ferreira*, A. Llvlnt[ston

+q¢

, S. Han*, A. Boam +, S. Zhang

+

*Department of Chemical Engineering and Chemical Technology, Imperial College of Science, Technology and Medicine, Prince Consort Road, London SW7 2BY, Unity Kingdom; +Membrane Extraction Technology Ltd, Imperial College, London SW7 2BY, UK Corresponding author: Tel +44-20-75945582; fax +44-20-75945629, E-mail address: [email protected] 1.

INTRODUCTION

Phenolic compounds (chemicals such as phenol and its derivatives) are used in phenolic resins, polycarbonates, biocides and agrochemicals. Aromatic amines (chemicals such as aniline and its derivatives) are used in a wide range of consumer products, including polyurethane foam, dyes, rubber chemicals and pharmaceuticals. The factories that manufacture and/or use these types of chemicals often create aqueous waste streams containing significant concentrations (0.1-10wt%) of these chemicals. Both aromatic amines and phenolic compounds are toxic and many of them are also carcinogenic. Tightening legislation in many countries calls for dramatic reductions in emissions of these species. A variety of processes have been proposed for treatment of these aromatic amine or phenolic compound containing wastewaters. Off-site disposal (landfill, deepwell injection) and biodegradation result in the compounds, which have typical values in the range US$0.75-US$20 per kg, being lost. These compounds have high boiling points and low vapour pressures. Hence, processes that rely on liquid-gas phase transition, such as distillation and pervaporation, have high-energy requirements. The use of adsorbents, such as activated carbon [1-4] or resins [58], is usually expensive due to difficulties and complexity in the regeneration stage. Problems associated with the use of solvent extraction [9-13] arise with phase separation [14] and contamination of the wastewater with solvent [15] due to the intermediate polarity of the compounds, which require moderately water soluble solvents. 165

Membrane Aromatic Recovery System (MARS) - A New Process For Recovering Phenols And Aromatic Amines From Aqueous Streams Livingston

Emulsion liquid membranes [16-19] and surfactant liquid emulsion membranes [20-22]systems were used to extract aromatic acids and bases from wastewaters. In particular Li and co-workers [23] used an emulsion liquid membrane containing caustic as reactive agent to increase phenol mass transfer. The main drawback of the use of liquid membranes is its inherently instability due to the leakage of the internal phase to the wastewater and the easy swelling of the liquid membrane by the wastewater phase. To overcome this problem, porous membranes have been used to support liquid membranes [24]. To avoid breakthrough of the immobilized phase in the pores to the aqueous phase, Ho and et al. [25-29] report development of a supported polymeric liquid membrane, in which the pores of a microfiltration or ultrafiltration membrane are filled with polyamphiphilic polymeric (oligomeric) liquids. These membranes were successful used, for phenolic compounds extraction from wastewater to an alkaline stripping solution at lab scale. However the use of liquid membranes, including supported polymeric liquid membrane, always requires an additional step for membrane phase regeneration, which often requires difficult demulsifications and/or the use of additional chemical agents [16]. The use of a non porous solid dense membrane avoid this step, hence aromatic acids and bases extraction from wastewaters have been tried in systems such as pervaporation [30,31 ], and aqueous/membrane/organic perstraetion [32]. The use of an acid-base reaction to maintain the driving force for phenol [33] or aniline [34] dialysis through several membranes was demonstrated using, sodium hydroxide or sulphurie acid respectively. In these studies, polymethylsilane-polyearbonate eopolymer appears to be the more promising membrane, although the values for permeability appear lower than those for reported for PDMS (phenol 4.5x10 -11 mE.s-1 [35] o r 2 . 5 x 1 0 -11 mE.s-l[36] against lxl0 "11 mE.s-1 [33] and aniline 20.9x10 -11 mE.s-1 [37] against 2.5x10 "11 mE.s 1134]). In the work reported by Klein et al. [33,34], the molar concentration of phenolate or anilinium is kept much lower than the concentration of base or acid used as the stripping solution in order to ensure a high driving force, and the stated intention is to dispose of the extracted aromatics. In this present report, working under conditions where the molar concentration of the phenolate or anilinium is carefully maintained near to the concentration of acid or base, we are able to show that the recovery of relatively pure phases of phenol and aniline is possible. This chapter describes the Membrane Aromatic Recovery Systems (MARS), a stable membrane process able to recover valuable aromatic amines [37] and phenolic compounds [38] from aqueous streams and its successful application at pilot scale for aniline recovery. The aromatic compound is extracted from the k,wastewater to a stripping solution, through a nonporous membrane selectively permeable to aromatic in its non ionic form and

166


impermeable to charged species. The driving force is maintained using an acidbase reaction in the stripping solution, where the aromatic is accumulated in its ionic form until a high enough concentration to allowed aromatic recovery by neutralization and consequentially phase separation. The MARS process is shown schematically in Figure 1.

1.1

MARS Operating Principle Wastewater out stripped of dissolved Amines (phenols)

HCI (NaOH)

NaOH (HCl)

R-NH4 +

(R-O-)

,~!~i ~:~:, ,,i~,

R-NH4+ (R-O-)

Nonporous Membrane

~: ~

R-NH 3 recovered amines (ROH recovered phenols)

Saline aqueous phase Wastewater containing dissolved amines (phenols) Figure 1" Schematic diagram of MARS process showing operating principles. Diagram show configuration for aniline recovery- figures in brackets show phenol recovery configuration.

1.1.1 Extraction Stage The aromatic compound is continuously extracted from the wastewater through a nonporous membrane separating layer into a stripping solution, where pH is controlled. The stripping solution is acidic for aromatic amines, and alkaline for phenolic compounds. In the stripping solution, acid-base reaction takes place and the aromatic molecule is converted into an ionized form (anilinium chloride or sodium phenolate). The membrane separating layer is hydrophobic; therefore charged molecules (such as ionized aromatic rings) cannot cross back through the membrane into the wastewater. The driving force for mass transfer of the aromatic acid or base through the membrane from the wastewater is maintained because the aromatic reacts to form its ionised form, a chemically distinct species. Note that the pH control operates such that each mole of aniline or phenol diffusing across the membrane requires one mole of either acid or base for neutralisation in the stripping solution. 167


1.1.2 Recovery Stage The stripping solution is periodically collected and pH is adjusted in order to recover the nonionic form of the aromatic molecule by acid-base reaction (i.e. alkaline conditions for amines and acid conditions for phenolics). Anilinium and phenate (the ionized forms of the aromatic) have virtually infinite solubility in water. However the solubility of nonionic forms of aromatic amines and phenols are usually less than 5 wt%. Hence, when the aromatic ion (highly concentrated in the stripping solution) is neutralized to the nonionic form, the resulting concentration greatly exceeds the aqueous solubility limit, and the solution separates into two phases: an aromatic-rich phase and a saline aqueous underlayer. NaC1 is a by-product in both cases, but this salty underlayer can be simply recycled to the wastewater feed as shown in Figure 1. In many of the industrial wastewater samples studied by Membrane Extraction Technology Ltd, we have found that the salt created by the acid-base recovery mechanism is negligible compared to the concentration of salt already in the waste stream. For example, the MARS process typically produces around 0.5 g of salt per g of organic recovered. In a wastewater containing 10 g L -1 of aromatic, the process will remove the aromatic and add around 5 g L -~ of salt. Many chemical industry wastes contain in excess of 100 g L 1 inorganic salts which arise in reaction processes, so the additional salt is not usually an environmental issue. 2.

MATERIALS AND METHODS

Analytical techniques employed for concentration determinations are described elsewhere[37,38]. The membrane used was silicone rubber tube (70 controlle Pressure gauge(~ I

pH / probe

Acid or caustic for pH control stewater out ~

Wastewater container

Stripping Membrane tank with solution stirring and heating overflow

Membrane coil

Figure 2: Laboratory and pilot plant continuous extraction configuration, with wastewater inside membrane tubes.

168


wt% poly(dimethylsiloxane) or PDMS and 30 wt % fumed silica), supplied by Silex Ltd. The membrane tubes used in all the laboratory experiments and pilot trials were 3 mm internal diameter and 0.5 mm wall thickness, with the exception of results shown in Figure 4, where the membrane tube had 3mm internal diameter but 0.35 mm wall thickness. The continuous flow configuration used for extraction stage laboratory experiments and continuous extraction trials at Solutia, UK is shown in Figure 2. The batch extraction configuration used for trials at Solutia UK is shown in Figure 3. The membrane tank solution temperature was kept at 50°C in all laboratory experiments and pilot trials. pH controller pH probe

Stripping solution recirculation

Wastewater Membrane coil

Acid for pH control

Stripping solution Membrane tank with vessel, level rises stirring and heating as aniline transfers Wastewater outside and acid added membrane Figure 3" Pilot plant batch extraction configuration, with wastewater outside membrane tubes. 3.

LABORATORY RESULTS

3.1

Phenol and Aniline Extraction

In phenol experiments a 28 m length membrane tube was used, a synthetic wastewater containing 10g.L -1 phenol was fed to the process, and a 12.5wt% NaOH solution was used to control the pH of a stripping solution between pH 11-13. Figure 4 shows phenol inlet and outlet concentrations from the membrane tube and total phenol (phenol plus phenate) concentration in the stripping solution over time. It can be seen that total phenol concentration in the stripping solution increases from 0 g.L -1 to 200 g.L ~. Operation at steady state is shown in Figure 5, in which the total phenol concentration (phenol and phenate) in the stripping solution was 242 g.L-1, three times higher than the phenol solubility in water and twenty five times higher than phenol concentration in the wastewater. The measured phenol concentration corresponds well to the value

169


oo tO

~

30

i

25 =

i

8,--. =o

-

160~ Q-

i

O

-,~ "~15 o

200

~10

~

,,

;H-111iI RH- !

80 ~ .-

8e- ~-~40

5

E

~ o 120 8 ~

8 ~

u

(1)

e-

-

0

10

0

I

-o- inlet ~

20 30 Time (day)

40

outlet -,,- stripping solution I

Figure 4: Evolution of phenol concentrations in the inlet, outlet and stripping solution over time in the MARS process operated continuously at laboratory scale: non steady state (reprinted with permission from J.Membr.Sci[38], page 224).

calculated for steady state assuming 12.5 wt% NaOH is neutralised by phenol to pH 13 of 227 g L 1. Figure 4 shows outlet concentration increasing with increasing phenol concentration in the stripping solution. From calculations based on the pKa of phenol, at pH 11, 9% of the phenol in the stripping solution is present as nonionic phenol and 91% as phenate ion. Therefore, when the concentration in the stripping solution increases the nonionic phenol u} to

300

15-

t-

_J

Q.

t--

250 ==

=o310-

200

"5 .o_

8 ~

-150 pH=13

~

_

5-

-

_.¢ _c

100

50 0

|

0 o inlet

5

i

|

10

~

i

,

==

G) O

Q. 0.

~ o

"r.

~

""

I--

c

|

15 20 Time (day)

r~ outlet

~E

25

× stripping solution

I

I

Figure 5: Evolution of phenol concentrations in the inlet, outlet and stripping solution over time in the MARS process operated continuously at laboratory scale: steady state (reprinted with permission from J.Membr.Sci [38] page 228).

170


concentration becomes high enough to have a significant negative effect on the driving force. At pH 13 only 0.1% of the phenol is in the nonionic form, the driving force is restored, and outlet concentrations drop to lower values, as shown in Figure 4. In aniline experiments, performed at laboratory scale using a 10.5 m membrane tube length, a synthetic wastewater containing 5 g.L-1 aniline was fed to the process, and 10.5 wt% HC1 solution was used to control pH in the stripping solution at pH 1. - 250

108

-

-

(D "7 200 ~._ ~.-J 0

L

15o

~6

loo 2-

...._~ ~._..; 50

'

"~. 0 ~ ' =Q..

~

2.4 2.0

0

0 0

2O

40

60

80

100

I Time (day) o inlet "-" outlet ~stripping solution - - --Predicted value at steady sate based on HCI concentration fed Figure 6: Evolution of aniline concentrations in the inlet, outlet and stripping solution over time in the MARS process operated continuously at laboratory scale.

Figure 6 shows the aniline inlet and outlet concentrations in the membrane tube, and total aniline (anilinium and aniline) concentration in the stripping solution, increasing over time. At steady state total aniline concentration in the stripping solution was 218.5 g.L-1, six times higher than the aniline solubility in water and forty four times higher than aniline concentration in the wastewater. These values illustrate the potential of MARS as a concentrating process. The steady state aniline concentration also compares well to the value obtained from stoichiometric neutralisation of 10.5 wt% HC1, 212 g L -1.

After total concentration in the stripping solution reached a steady state, the pH set point was varied in order to study the effect of pH on mass transfer flux (after day 85). The higher the pH in the stripping solution, the higher the percentage of nonionized aniline, and therefore the lower the driving force, resulting in lower mass transfer rates. It can be seen in Figure 6 that an increase in pH of the stripping solution clearly corresponds to an increase in outlet aniline concentration.

171


In the laboratory experiments a short membrane tube was used in order to quantify the outlet concentrations and estimate the overall mass transfer coefficient. Temperature was used to enhance mass transfer. The overall mass transfer coefficient obtained at 50°C across the 0.5 mm thickness membrane, was 1.4x10-7 m.s-1 for phenol and 4.8x10 -7 m.s-1 for aniline. Each mole of aromatic extracted from the wastewater requires one mole of acid or base to be fed to the stripping solution for neutralisation, ie each mole of aniline extracted requires one mole of HC1 and each mole of phenol extract requires one mole of NaOH. These ratios were tracked during the experiments and were usually close (within + 10%) to 1.0.

3.2 Recovery Stage Aromatic recovery was performed in batch. The aromatic compound accumulated in the stripping solution in ionized form. The overflow of the stripping solution, enriched with aromatic in the ionized form, was fed to the recovery stage and neutralized in order to obtain the neutral form of the aromatic. Neutralization was performed by adding 37% HC1 solution to the sodium phenate solution and 50% NaOH solution to the anilinium chloride solution. The aqueous solubility of the aromatic in the neutral form is low (8.2 wt% for phenol and 3.4wt% for aniline), hence when neutralization is completed the solution separates into two layers: a top phase that is the final product, and a underlayer aqueous phase that contains resultant NaC1 and traces of aromatic.

80

n Phenol in the aqueous saline phase • .,.Phenol,in the orqanic phase /otat pnenot Ted-to recovery stage

,° t

60

50 -~

40

a.

20

30 10 0

1

Overflow period (day)

Figure 7: Mass balances for phenol batch recoveries from laboratory data (reprinted with permission from J.Membr.Sci. [38], page 225).

172


Mass balances for recovery of phenol are shown in Figure 7. The organic phases were composed of aromatic compound and water. With phenol, the organic phase consists of 86.5% phenol and 13.5% water and, with aniline, 96.5% aniline and 3.5% water. No other organic was detected by GC analysis in the laboratory experiments. Recovery efficiency is defined as the ratio of mass of aromatic in the organic phase over the mass initially present in the stripping solution. Recovery efficiencies of over 92% were achieved for both systems. The balance of aniline or phenol is present dissolved in the aqueous layer resulting from the recovery process. The NaC1 in this saline aqueous underlayer is a by-product, which is responsible for a salting out effect that limits the solubility of the aniline or phenol in the aqueous layer. Hence, the higher the NaC1 concentration in the aqueous layer, the lower the aromatic concentration in the saline underlayer and the lower the water concentrations in the organic phase. Therefore the higher the NaC1 concentration, the higher the aromatic concentration in the final product and the higher the recovery efficiencies become. Figure 8 shows the salting out effect, with phenol aqueous solubility decreasing from 83 to 15.7 g.L1 as NaC1 concentration increases from 0 to 200 g.L 1 and phenol concentration in the organic layer increasing from 70.5 to 87.5 wt% for the same range of NaC1 concentration. Clearly there is a benefit to operating by adding as high a concentration of acid or base to the stripping solution as possible. However, this is balanced by the overall concentration of phenol or aniline rising and increasing the concentration of unionised aromatic in the stripping solution, leading to lower mass transfer fluxes. .~"~ 90 80 ; 70 c

k \

e Solubility o Phenol in organic _ phase

1 100 | | ~"

60

90

50

Q..

.~

~. 4o

"6

80 o

30

_c-

~= 20 10 0

0 ¢-

•c--

,

,

,

100

200

300

70 n 400

NaCI in water (g.L -~)

Figure 8: Phenol concentration in the aqueous phase, water and in the organic-rich phase as a function of NaC1 concentrations in the aqueous phase (reprinted with permission from J.Membr.Sci. [38], page 227). 173


3.3

Other Aromatics

In order to illustrate the applicability of MARS to recover compounds other than aniline and phenol, the overall mass transfer coefficients of a range of compounds were measured and results are shown in Table 1. Three independent non-coiled (randomly spaced) membrane tubes were submerged in a single stripping solution vessel, pH was kept below 1 using HCI solution for aromatic bases and over 14 using NaOH for aromatic acids. Under these conditions the unionised aromatic compound concentration in the stripping solution was assumed to be zero. In each run, three aromatic compounds were tested. The overall mass transfer coefficient for aniline and phenol in Table 1 are around twice (respectively 1.7 and 2.2 times higher than) the average value obtained in the laboratory MARS extraction. This is probably since in the mass transfer test the membrane was loosely arranged whereas the membrane in the MARS process was tightly rolled around a support, reducing the effective membrane-stripping solution interfacial area. Dimethylamine,4-nitrophenol,and 2,4,6 tris(dimethylaminomethyl)phenol exhibit very low overall mass transfer coefficients and their extraction from wastewaters will be difficult using silicone rubber tubes as membranes. Mass transfer rates for phenol, 4-nitroaniline, hydroquinone and pyridine have Table 1" Overall mass transfers for different compounds at 50°C across a 0.5 mm wall thickness silicone robber membrane Compounds

Kov x 107 (m.s-1)

Compounds

Kov x 107 (m.s"1)

Aniline 4-chloroaniline 2,4-chloroaniline 4-nitroaniline 4-fluoroaniline 2,4-fluoroaniline triethylamine dimethylamine benzyldimethylamin dicyclohexylamine

8.2 11.6 6.36 4.34 10.27 9.33 20 0.72 18 16.5

phenol 4-chlorophenol 2,4-dichlorophenol 4-nitrophenol 4-cresol hydroquinone Pyridine 2,4,6- tris (dimethy laminomethyl) phenol

3.1 9.3 14.7 0.48 7.04 2.37 4.4 0.64

intermediate values, and the extraction of these three compounds from wastewaters by MARS technology is possible using silicone rubber as membrane, but will use relatively large membrane areas. All other compounds tested have higher overall mass transfer coefficients, and so they can be

174


relatively easily removed from wastewaters by MARS technology using silicone rubber tubes as membranes. 4. PILOT PLANT RESULTS FOR ANILINE The MARS process has been scaled-up for recovery of aniline at Solutia, UK by Membrane Extraction Technology Ltd (UK). Further pilot trials are currently underway with phenol. Production of 4-nitrodiphenylamine (4NDPA) at Solutia UK results in a wastewater containing an aniline concentration that varies from 4.1 to 9.5 g.L -1, with an average value of 6 g.L 1. A pilot scale MARS unit was operated on site and recovered a total of 120 L of aniline from 29,500 L of wastewater over a period of some two months intermittent operation. 4.1

Continuous Extraction

Initially the pilot plant was configured for continuous operation, as for laboratory scale work (Figure 2), in which the wastewater flowed inside the membrane tube and aniline was accumulated out side of the tube, in an acidic stripping solution. At lab scale a membrane area of 0.099 m 2 (a single 10.5 m length tube) was used at a flow rate of 5.5 L.day -1, immersed in a membrane tank holding 1.5 L of stripping solution with pH controlled at 1. The pilot plant used a total area of 45 m" (48 tubes with 100m length), immersed in a 1000 L membrane tank and a pH in the stripping solution of 1.5. Flowrate was between 200-500 L.day 1. The temperature of the stripping solution was kept at 50°C and an 11% HC1 solution was fed to the stripping solution to control pH. Quantities of the stripping solution (80 L in the pilot plant against 0.25L in the lab scale) were periodically sent to batch recovery, which was achieved by addition of NaOH (33 wt% in the pilot plant, against 50 wt% in the lab scale). 12 11 10

60 - * - - i n l e t - - e - o u t l e t -e-stripping solution I

50

a) .-.. 9

=-~'

~

= o

6~

40 "~ ~, 88

8

"-"

7

o

15

30

4

20

2

10 ~-';

o ._ -~. m

~

1 0

.

.

.

.

,

5

.

.

.

.

.

.

10

.

.

.

.

.

15

.

.

.

.

.

.

.

.

,

20 25 Time (days)

.

.

.

.

.

.

30

.

.

.

.

.

.

35

.

.

i

0

40

Figure 9: Evolution of aniline concentrations in the inlet, outlet and stripping solution over time in the MARS process operated continuously at pilot plant scale.

175


Figure 9 shows aniline inlet and outlet concentrations, and total aniline concentration in the stripping solution, over time during continuous pilot plant operation. The average remo~cal achieved was 95% over the whole operating period, which corresponds to an overall mass transfer coefficient of 2.7 x 10-7 m.s 1, somewhat lower then the 4.8x10 -7 m.s -1value obtained at lab scale. 4.2

Interruption of Continuous Extraction The continuous operating period ended after day 36 when the backpressure of the wastewater being pumped down the membrane modules rose suddenly to over 0.5 bar. This caused a relief valve in the wastewater line to open and flow through the membranes ceased. Examination revealed that the membranes, and pipework downstream from the membranes, had blocked. The membrane tubes were removed from the membrane tank and some of them were opened up to reveal crystals of a black material, which were obviously blocking the flow. This black material dissolved readily in dichloromethane. When injected onto the GC, it was identified as 84% (by peak area) 4NDPA. The blockage corresponded to a dark-coloured organic phase being present in the wastewater due to insufficient upstream phase separation. This organic phase is probably organic product phase from the 4NDPA reaction. This mixture of aniline, 4NDPA and other compounds was initially liquid (aniline acts as a solvent for 4NDPA). In passing through the membranes, the aniline was stripped leaving the 4NDPA originally present in the liquid mixture as crystals in the membrane tubes. Filtration methods to avoid entry of this organic phase in the membrane tube were attempted, however frequent cleaning and filter changes are required and the filters were unable to prevent the passage of liquid solvents, and so were not sufficiently reliable or effective. 4.3

Batch Extraction To overcome these practical challenges, the pilot plant was reconfigured to extract aniline from wastewater batches, with the acidic stripping solution flowing inside the membrane tube and the wastewater in the membrane tank as shown in Figure 3. 770 L of wastewater was treated in each run. The same membrane area was used for batch and continuous extraction, and the membrane tank was kept at the same temperature (50°C). pH in the stripping solution was controlled at 1.5 using a 11% HC1 solution. A re-circulating flow rate of 80 L.h -1 was used in the stripping solution. In this configuration the wastewater, with the tar or organic phases when they are present, were pumped directly to the membrane tank without filtration. Membrane blockage was avoided since the solution inside membrane tube is the stripping solution in which no blockage material was present. Figure 10 shows aniline concentration initial and final

176


100

~'6o

- BO

°ii iI

r~

4"~

!

I

20

0

1 2 3

4

5 6

7 8

9 10 11 12 13 14 15 16 17 18 19 20 21

/ A n i l i n e inlet D Aniline outlet--*-Aniline removal

Batch Number

Figure 10" Batch operation pilot plant data: aniline initial and final concentration in each batch, aniline removal in each batch. concentration for each of the batch runs and the respective removal efficiency. Batch extraction was on average 22 hours in duration. An average of 90% removal over all the 21 batches was achieved. Aniline removal can easily be increased by increasing the membrane area/ membrane tank volume ratio (values of 100 m2m3 are easily obtainable). The stripping solution aniline concentration increases from 80g.L -1 to 160 g.L -1. The theoretical steady state of 240 g.L -1 which corresponds to 11% HC1 solution used was not reached in the time span of the trial. An average value for overall mass transfer coefficient of 4.2 xl0 -7 m.s 1 was calculated during the pilot plant batch extractions. This value is comparable to the average value obtained in laboratory scale continuous extraction (4.8x107 m.s-1), and higher than 2.7 x 10-7 m.s 1, the value obtained during continuous operation of the pilot plant. The mass transfer rate in the continuous pilot plant most probably decreased due to partial blockage of some tubes by tarry solids.

4.4

Recovery Stage

Aniline recovery from the acidic stripping solution enriched in anilinium chloride was achieved by neutralization using a 33% NaOH solution. A total of 128 L of aniline rich phase was recovered with a composition of 95 wt% aniline, 2.4 wt% Toluidine and 2.6 wt% water. The average aniline concentration in the salt under layer is 4 g.L -1. The purity of aniline obtained was high enough to allow it to be re-used in 4NDPA production.

177


5. PROCESS ECONOMICS Aniline price at time of writing is about $0.75 Kg -~, and hence it is not one of the more valuable aromatics in the market. Nevertheless due its Item

Comment

MARS Plant

Capital Charge Factor = 0.3 $45,000

Annual Cost (Benefit)

Membrane Replacement 2 year lifetime

$7,500

Acid (33% HC1)

17 ty-l@ $105 t-1

$1,785

Base (50 %)

12 t y -1 @ $210 t -1

$2,520

Power ( 3 kW)

8000 h y-~ @ $0.075

$1,800

Steam

82 t y-1 @ $15 t-1

$1,230

Labour

10% of one staff

$11,250

TOTAL

BENEFIT

($41,415)

Table 2: Case S t u d y - Process economics of MARS process using fluoroaniline as an example.

environmental impact, aniline removal is necessary and MARS provides a considerable economic advantage to alternative processes, although details cannot be provided here for commercial reasons. However, an example of MARS to recover a more valuable chemical is given in the Case Study contained in Table 2. The commercial value of fluoroaniline is ten times the aniline value (around $7.5 kg-1). In the example given the treatment of 10 m3 d-1 wastewater with a 5 g L 1 fluoroaniline concentration is considered. The MARS process in this case is able to deliver a net benefit through the value of the recovered fluoroaniline. 6. CONCLUSIONS MARS is a novel process coupling detoxification and recovery. It is capable of achieving high recovery efficiencies, and producing a relatively pure stream of recovered organics. At the laboratory scale, MARS has proven to be a successful process for removing and recovering aniline and phenol, with a recovered purities of 95% and 86.5% respectively, pH is an important parameter which controls the driving force across the membrane and consequently the removal efficiency of the aromatic from the wastewater. Anilinium and phenate salts were accumulated in the stripping solution to high concentrations, allowing further recovery by neutralization. The strip feed solution (HC1 or NaOH) concentration is a key parameter that controls the aromatic concentration at steady state in the stripping

178


solution. The salting out effect, resulting from high concentrations of NaC1 in the aqueous phase after neutralization, has a positive effect on the phase separation, decreasing the water concentration in the organic phase and the aromatic concentration in the aqueous phase. At pilot plant scale MARS was proven able to recover aniline from an industrial wastewater in a good purity. MARS was shown to be easily scaled up based on membrane area. Using different configurations, MARS was adapted to deal with a common problem in membrane technology, that is membrane blockage by tarry solids or organics precipitation. The MARS process can utilise very simple nonporous rubber tubes as membranes. Mass transfer is relatively low, but these membranes are relatively cheap. This is a key area where we expect to improve the process over the next months. MARS has low energy requirements because it exploits the acid-base functionality of aromatic acids and bases to produce a driving force based on the chemical energy contained in NaOH or HC1. The process can be carried out at conditions of pressure and temperature which are near ambient throughout all items of equipment. MARS does not rely on volatility of organics (or any phase transition), so can recover organics which membrane technologies such as pervaporation cannot reach. Finally, MARS has been shown to be promising for industrial application to recovery of a larger range of aromatics than phenol and aniline. Examples include amines, phenolics and pyridines. REFERENCES [ 1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

ChiangP.C., Chang E.E., Wu J.S.: Comparison of chemical and thermal regeneration of aromatic compounds on exhausted activated carbon. Wat. Sci. Tech. 35-7 (1997) 279. OrshanskyF., Narkis N.: Characteristics of organic removal by PACT simultaneous adsorption and biodegradation. Wat. Res. 31-3 (1997) 391. Cha T., Glasgow: Sorptive reclamation of phenol from coal conversion wastewater. Ind. Eng. Chem. Res 22 (1983)198. Pahari P.K., Sharma M.M.: Recovery of heterocyclic amines from dilute aqueous waste streams. Ind. Eng. Chem. Res.30 (1991) 1980. Fox C.R. Plants uses prove phenol recovery with resins. Hydrocarbonprocessing 11 (1978) 269. Fox C.R. Remove and recover phenol. Hydrocarbon processing 7 (1975) 109. Fox C.R.: Removing toxic organics from waste water. Chem. Eng. Process. 8 (1979) 70. FergusonG.U.: Phenol recovery using polymeric adsorbent resins. Chem. Ind.8 (1982) 567. TeramotoM., Takihana H., Shibutani M.Yuasa T., et al: Extraction of amine by W/O/W emulsion system. J. Chem. Eng. Jap. 14-2 (1981) 122. Jagirdar G.C., Sharma M.M.: Separation of close boiling substituted Anilines: Gasliquid vs Conventional liquid-liquid dissociation extraction. J. Separ. Proc. Technol.24(1981)7.

179


[11]

[12] [13]

[14] [15]

[16] [17]

[18] [19]

[201 [21]

[22] [23]

[24] [25] [26] [27] [28] [29]

[301 [31]

Wadekar V.V., Sharma M.M.: Separation of close boiling organic acids/bases by dissociation extraction: Binary and Ternary Systems: Substituted anilines; Binary system with thermally regenerative extractant: Chlorophenols. J. Sep Proc. Tech.2-2 (1981)28. Medir M., Arriola A., Mackay D., et al: Phenol recovery from water effluents with mixed solvents. J. Chem. Eng. Data 30 (1985) 157. Jagirdar G.C., Sharma M.M.: Recovery and separation of mixtures of organic acids from dilute aqueous solutions. J. Separ. Proc. Technol. 1-2, (1980) 40. Netke S.A., Pangarkar V.G.: Extraction of naphthenic acid kerosene using porous and nonporous polymeric membranes. Sep. Sci. Technol. 31 (1996) 63. Krishnakumar V.K., Sharma M.M.: A novel method of recovering phenolic substances from aqueous alkaline waste streams. Ind. Eng. Chem. Process Des. Dev. 23(1984)410. Devulapalli R., Jones F: Separation of aniline from aqueous solutions using emulsion liquid membranes. J. Hazard. Mater. B.70 (1999) 157. Boyadzhiev L., Besenshek E., Lazarova Z.: Removal of phenol from waste water by double emulsion membranes and creeping film pertraction. J. Memb. Sci.21 (1984) 137. Thien M.P., Hatton T.A.: Liquid emulsion membranes and their applications in biochemical processing. (2001) 819. Gadekar P.T., Mukkolath A.V., Tiwari K.K.: Recovery of nitrophenols from aqueous solution by a liquid emulsion membrane system. Sep. Sci. Technol. 27 (1992) 427. Kataoka T., Nishiki T., Kimura S.: Phenol permeation through liquid surfactant membrane-Permeation model and effective diffusivity. J. Memb. Sci. 41 (1989) 197. Kakoi T., Goto M., Natsukawa S., et al: Recovery of phenols using liquid surfactant membranes prepared with newly synthesized surfactants. Sep. Sci. Technol.31-1 (1996) 107. Ulbrich M., Marr R., Draxler J.: Selective separation of organic solutes by aqueous liquid surfactant membranes. J. Memb. Sci.59 (1991) 189. Terry R.E., Li N.N., HO W.S.: Extraction of phenolic compounds and organic acids by liquid membranes. J. Memb. Sci. 10 (1982) 305. Kiani A., Bhave R.R, Sirkar K.K.: Solvent extraction with immobilized interfaces in a microporous hydrophobic membrane. J. Memb. Sci.;20 (1984) 125. Ho S.V.: A supported Polymeric Liquid Membrane Process for Removal of Carboxylic Acids from a Waste Stream. Environ. Prog. 18-4 (1999) 273. Ho S.V. Extracting organic compounds from aqueous solutions. (1996) US. Patent 5512180. Ho S.V. Solid poly-amphiphilic polymer having use in a separation process. (1996) US. Patent 5,552,053. Ho S.V. Supported liquid membrane and separation process employing some. (1996) US. Patent 5507949. Ho S.V., Sheridan P.W., Krupetsky E.: Supported polymeric liquid membranes for removing organics from aqueous solutions I. Transport characteeristics of polyglycol liquid membranes. J. Memb. Sci. 112 (1996) 13. Meckl K., Lichtenthaler R.N.: Hybrid process using pervaporation for the removal of organics from process and waste water. J. Memb. Sci. 113 (1996) 81. Lipnizki F., Hausmanns S., Ten P.K., et al: Organophilic pervaporation: prospects and performance. Chem. Eng. J. 73 (1999) 113.

180


[32] [33] [34]

[351 [36]

[37]

[38]

Ray S.K., Sawant S.B., Joshi J.B., et al: Perstraction pf phenolic compounds from aqueous solution using a nonporous membrane. Sep. Sci. Technol. 32-16 (1997) 2669. Klein E., Smith J.K., Weaver R.E.C., et al: Solute Separation from Water by dialysis II. The separation of Phenol by Downstream Conjugation. Separ. Sci. 8-5 (1973) 592 Klein E., Smith J.K., Wendt R.P., et al: Solute Separation from Water by dialysis I. The separation of Aniline. Separ. Sci. 7-3 (1972) 285. Brookes P.R, Livingston A.G.: Aqueous-aqueous extraction of organic pollutants through tubular silicone rubber membranes. J. Memb. Sci. 104 (1995) 119. Bennett M., Brisdon B.J., England R., et al: Performance of PDMS and organofunctionalised PDMS membranes for the pervaporative recovery of organics frm aqueous streams. J. Memb. Sci.137 (1997) 63. Castelo-Ferreira F., Han S., Livingston A.G.: Recovery of Aniline from Aqueous Solution using the Membrane Aromatic Recovery System (MARS). Ind. Eng. Chem. Res. (2002) In Press. Han S, Castelo-Ferreira F., Livingston A.G.: Membrane aromatic recovery system (MARS) -a new membrane process for the recovery of phenols from wastewaters. J. Memb. Sci. 188-2 (2001) 219.

181


Functional Membranes and Materials for Biocatalysis, Separation, and Analysis

183


Biofunctional Membranes and Biomedical Devices

185



Chapter 9

Membrane bioreactors for biotechnology and medical applications L. Giorno, L. De Bartolo, E. Drioli* Research Institute on Membranes and Modelling of Chemical Reactors, IRMERC-CNR, c/o University of Calabria, via P. Bucci cubo 17/C, 87030 Rende (CS); *Corresponding author, Tel: +39 0984 492039; Fax. +39 0984 402103 e-mail: [email protected] ABSTRACT In this chapter we will discuss the properties and applications of bioartificial hybrid systems used in biotechnological and medical applications. In the first part we will focused on membrane bioreactors using biocatalysts such as enzymes, microrganisms and cells for the production of biological molecules or for separation of pharmaceutical products. In the second part we will discuss on membrane bioreactors using isolated mammalian cells (i.e., liver cells, pancreatic cells) as bioartificial organs in temporary or continuous substitution of injured organ. The properties of membranes to be used in these devices will be reported including mass transport, morphological and physico-chemical properties that influence their performance. The discussion of the biotechnological part will consider the effect of immobilization on functional stability and activity of biocatalysts, including membrane material and morphology, physico-chemical properties of reaction environment. The successful examples of bioreactors running at large scale, of which the authors are aware, will be also presented. In the medical application part, the development of membrane bioartificial organs, e.g. bioartificial pancreas and bioartificial liver, as well as the properties of membranes for these systems will be discussed. Particular attention will be given to the recent achievements in cytocompatibility and biocompatibility of membranes in bioartificial organs. 1.

INTRODUCTION

The combination of artificial membranes with macromolecules of biological origin allow to obtain hybrid systems capable to imitate what nature

187

Membrane

Bioreactors For Biotechnology And Medical Applications -

Drioli

has fine tune through aeons of evolution. In fact, these systems, based on their selective transport and catalytic properties, are able to carry out chemical reactions, selective separation, waste removal, etc. The use of membranes and biological tools for improving traditional production systems is a possible approach for maintaining a sustainable growth. Typical examples include novel pharmaceutical products with well-defined enantiomeric composition; new foods with improved nutritive properties; medical devices for the care of crucial diseases; and the treatment of wastes. The performance of the hybrid systems are governed by the properties of specific biomolecules and membranes and in particular by their reciprocal interactions [ 1]. When the biomolecule is an enzyme immobilized in a membrane, the overall performance of the biocatalytic membrane reactor is due to the preserved functional conformation of the enzyme as well as to the transport of reagent and product, through the membrane, to and from the enzyme, respectively. When mammalian cells are to be adsorbed on membrane to prepare artificial organs, the membrane properties are of crucial importance to keep them viable. The biomolecule can either be loaded to the membrane or just compartmentalized by the membrane. Reactors with segregated cells are suitable for therapeutic applications; for example, the bioartificial pancreas, the bioartificial liver, and the extracorporeal detoxification device [2]. 2.

M E M B R A N E BIOREACTORS IN B I O T E C H N O L O G Y

Biocatalytic membrane reactors can combine selective mass transport with chemical reactions and the selective removal of products from the reaction site increases the conversion of product-inhibited or thermodynamically unfavorable reactions. Many enzymes and micro-organisms have been used in membrane reactors to catalyze bioconversions for various application [4-24]. The possibility of integrating biotransformation into productive reaction cycles makes reliable the use of biocatalysts at large scale. The main parameter that influenced the development of biocatalytic processes at productive level is the stability of the biocatalyst. Immobilization has proved to increase stability of biocatalysts, although it may cause changes in the catalytic activity and enantioselectivity [25-27]. This is a common observation, nevertheless, it cannot be considered a general rule of the inverse relationship between stability and activity and enantioselectivity of immobilized enzymes. Most probably, these effects are related to the interactions between the chemical groups of enzyme and membrane: due to these interactions the molecule becomes more rigid, and therefore more stable, but the loss in flexibility damages the specificity and activity. 188

Membrane

Bioreactors For Biotechnology And Medical Applications - Drioli

Recently, studies on the structure of enzymes present in extremophiles (thermo-stable) and mesophiles (thermo-labile) micro-organisms have been reported [28-31 ]. Referring to the activity, it is observed that the enzyme that are resistant at high temperature are much less active (especially at low temperature). For immobilized enzyme, the challenge is to obtain the proper interactions to guarantee a balanced efficiency of the various properties (stability, enantioselectivity and activity). The efficiency of the overall immobilized system, depend on the biochemical (catalytic activity; half-life time; productivity; reaction kinetics; catalyst inhibition effects; concentration, viscosity, solubility and purity of substrate and product; sensitivity to pH and temperature; immobilization stability, etc.) mechanical (membrane configuration, structure and strength; size, morphology and pore size distribution; swelling and compaction effects; shear stress, etc.) and hydrodynamics parameters (flow regime; pressure drop; transmembrane pressure; flow velocity; residence time; easy to clean, etc). A comparative analysis of the performance of enzymes free and immobilized on membrane using various substrates is reported in Table 1. Hydrolases/esterases (such as lipases) are among the most studied biocatalysts [34-50]. Lipases (triacylglycerol-hydrolases, E.C. 3.1.1.3) act efficiently on water-insoluble substrates by binding to the water-organic interface. The natural organic interface consists of triglycerides. This binding places the lipase close to the substrate, and in particular increases the catalytic efficiency by interfacial activation. Due to their natural function at the oil-water interface, lipase exhibit high stability in contact with organic solvents, and beside triglycerides they are also active with a variety of non-natural substrates, showing generally a high stereo- and/or regioselectivity [41, 53]. Some author reported that the immobilized lipases used in organic solvent to catalyze esterification reactions, show higher activity and stability than lipase free in suspension [54, 55]. 2.1

Influence of operating conditions

2.1.1. Effect of immobilization

Lipases have on their surface hydrophobic and hydrophilic patches, therefore they can be immobilized by ionic interactions on ionic hydrophilic membrane support, or by van der Waals forces on hydrophobic membrane. Hydrophobic materials have shown to increase the performance of lipase [63, 64]; nevertheless, the support has to be chosen in agreement with the properties of the overall reaction system. For example, when using organic solvents, the resistance of the membrane to the solvent has to be taken into the account. In this case, hydrophilic membranes (such as polyamide,

189

MembraneBioreactorsFor BiotechnologyAnd MedicalApplications- Drioli

polyacrilonitrile) have shown to have higher stability than hydrophobie ones (e.g., polypropylene). Table 1" A comparative analysis of the performance of enzyme immobilized in various

membrane Enzyme (substrate)

Lipase (Olive oil)

Type of immobilization

Type of reactor

(Enzyme Free) a

STR E-EMR

Observed reaction rate (retool.1 "l h q)

Enzyme stability

ef.

-

6.95

Low

2

1.6

6.80

Medium

4.50

High

Axial Flow rate (ml-min "t)

3, 34

Entrapment by cross-flow filtration

TSP-EMR (organic) 1.6 (aqueous) 160 Cross-linkingby TSP-EMR (organic) 1.3 glutharaldehyde (aqueous) 167 Lipase (Cyano ester of (Enzyme free) b STR ibuprofen) Entrapment by TSP-EMR (organic) 160 cross-flow (aqueous) 400 filtration Lipase (Methyl ester (Enzyme free) STR of naproxen) Entrapment by TSP-EMR (organic) 300 cross-flow (aqueous) 300 filtration Fumarase STR (fumaric acid) (Enzyme free) ¢

4.50

1.12

Low

1.2x10 ~

High

2

Low 5.3x10 "3 8x10 "4

High

290

Low

Entrapment by UF-EMR 140 2375 High cross-flow filtration d aKm = 9 m m o l ' l "1, V m a x = 20 k + 2 - 6.86 s'l; aKm = 47 m m o l ' l l , V m a x = 4 mmol.1-1 h -1, k+2 = 1.37 s l ; aKin = 212 m M retool.l-l; V m a x = 9000 retool.1 -l h l ; k+2 = 5x1012 s "l, dk+2 = 8x1012 s -1

In addition to the membrane material, it is very important also the amount of enzyme immobilized [26, 55, 65, 66]. It is generally observed that the performance of immobilized enzyme as a function of density of enzyme on membrane (gprotein/Cm3) has an optimum. Increasing the gprotein/Cm3 the catalytic activity can be stable or increase until it reaches a maximum, after which, for further increase of protein loading no increase of activity is achieved. The initial increase of activity with the enzyme loading, indicates that the interaction with the support destabilizes the protein, and as the protein amount is increased it is more protected, and then more active. The constant activity is mainly due to mass transport limitations. It some cases, further increase of protein can even lead to decrease of activity. This happens when the protein-protein interactions

190

Membrane BioreactorsFor BiotechnologyAnd MedicalApplications- Drioli

at high concentration cause denaturation of protein conformation; either because of bound with the active site or because of proteolitic activity present in the crude enzyme preparation. The behavior of some enzyme as a function of loading amount is summarized in Figure 1, from which it can be seen that the same type of immobilized enzyme can show different behavior with different substrates. For

Activity

h.._

Enzyme loading

v

Figure 1. Qualitative behaviour of activity of immobilized enzymes as a function of enzyme loading: •

kipase C. rugosa in polyamide membrane (hydrolysis activity with olive oil) [26]

II

Lipase C. rugosa in polyamide membrane (hydrolysis activity with naproxen ester) [35] Lipase C. antarctica on Accurel0a)EP100 (Esterification activity) [55]

O

Lipase Humicola sp on Accurelc.)EP100 (Esterification activity) [55]

r]

Fumarase from porcine heart (hydrolysis activity)[ 19]

example, lipase from C. rugosa when used with triglycerides as substrate, shows constant behavior for initial increase of immobilized enzyme and after a decrease; while, when used with naproxen ester for initial increase of immobilized enzyme the activity increases and after, for further increase of enzyme, the activity decreases. This implies that the same enzyme conformation, achieved with the same type of immobilization, has different affinity for the different substrates. The use of proper conditions may allow to reduce the amount of enzyme with the result of reducing the mass transport limitations and cost.

2.1.2. The influence of hydrodynamics conditions In order to measure the kinetic properties of immobilized systems, it is necessary to work in conditions of reaction limited regime, where the mass transport rate is much higher compared to the reaction rate [67].

191

Membrane Bioreactors For Biotechnology And Medical Applications - Drioli

Among the various methods to control the mass transfer coefficient, the fluid dynamics conditions, such as axial velocity and transmembrane pressure, may help in achieving the proper mass transfer properties. In two-separate phase (or biphasic) membrane reactors (TSP-EMR), where the mass transport occurs mainly by diffusion, the effect of the flow rate of organic phase (which contains the substrate) can be very different from that of the aqueous phase (which extracts the product) on the reactor performance. An example of the mentioned behavior is reported in Figure 2 and refers to the lipase using olive oil and cyano ester of ibuprofen as substrate. For the studied values, the observed reaction rate decreases with increase of the organic flow rate, for both substrates. Vice versa, the observed reaction rate increases with increasing the aqueous flow rate. This behavior is due to the fact that the higher transport of substrate (achieved at higher axial velocity of the organic phase) a higher reaction rate is obtained. On the other hand, the fixed axial velocity of the aqueous phase was not able to promote the extraction of the reaction product, which therefore, inhibited the reactor performance. This show the importance of both transport of substrate and product through the enzyme-loaded membrane 0.14 0.12 .,,.... 0.1

.,=

0.08 E g 0.06

l e Olive oil &CNE

0.04 0.02 0 0

1O0

200

300

Qa (mmol/I h)

Figure 2. Effect of axial flow rate of organic and aqueous phases on the performance of two-separate phase enzyme membrane reactor

reactor. In ultrafiltration enzyme membrane reactors (UF-EMR), where the mass transport occurs by convection, the permeate velocity is the parameter that mostly affects the reactor performance, as it influences the residence time of reagents within the enzyme-loaded membrane. The residence time, T, is calculated as the ratio between the membrane thickness, L, and the permeate velocity, L/T. Figure 3 shows the influence of residence time on the conversion of fumaric acid into L-malic acid using fumarase immobilized in a ultrafiltration membrane reactor [19]. The effect of the amount of enzyme loading is also illustrated.

192

Membrane

Bioreactors For B i o t e c h n o l o g y

And

Medical

Applications

- Drioli

0.9 0.8 .-. 0.7

~

n-n

~o 0.6 •E

0.5 0.4 0.3 • 0.3 mg/cm 3

0.2 0.1

• 1.14 mg/cm3

0 0

i

i

i

~

i

100

200

300

400

500

600

Residence time (s) Figure 3. Influence of residence time on the efficiency of and ultrafiltration enzyme membrane reactors

2.1.3. The influence of reaction microenvironment on enzyme enantioselectivity Thanks to their properties, lipases have found applications in various fields of biotransformations. One of the most recent and attractive field is the synthesis of chiral compounds. The importance of using optical pure enantiomers became clear in the recent years thanks to the understanding of the different effects of enantiomers, where often one has the positive properties whilst the other can cause serious side effects [68]. The chiral market represents about the 25% of the drug market; on 1993 only the 3% of this was represented by pure enantiomers while the rest was sealed as racemic mixtures; on 1997 more than 75% of chiral drugs were marketed as pure enantiomers. This picture shows the rapid change towards the use of pure enantiomers in the pharmaceutical field, that had his first push in 1992 when FDA issued a policy statement. The same trend is expected to happen for food and agro-chemicals. On the basis of this context, many efforts are focused on the development of technologies to produce optically pure isomers. Although the most common method to produce enantiomers at industrial level is the chemical synthesis of a racemic mixtures followed by separation of the two enantiomers by diastereomeric crystallization, new resolution processes are being considered. Among others, membrane processes are today investigated for this purpose. On the basis of the current literature, there are two schemes on the use of membrane technology to produce enantiomers [69]. In one case, the membrane itself is intrinsically enantioselective and separates the wonted isomer on the basis of the pore spatial conformation. In the other case, a kinetic resolution catalyzed by an enantiospecific catalyst is combined with a membrane 193


separation process: the chiral system is represented by the catalyst, the membrane only separates the product from the substrate on the basis of their physical chemical properties. Studies on the use of lipase-catalyzed enantioselective biotransformations in multiphasic membrane reactors, for the production of pure (S)-naproxen have been carried out at our laboratory. Particular attention was focused on the parameters that influence the catalytic properties of the lipase immobilized on polymeric membrane. The performance of lipase immobilized with the same method on different type of membranes, which were then used in different reactor systems, has been investigated. The lipase was immobilized by cross-flow ultrafiltration, so that to obtain a stable irreversible lipase fouling within the sponge layer or on the lumen surface of asymmetric polymeric membranes. The lipase-loaded membranes were used to carry out the kinetic resolution of (R,S)-naproxen methyl ester to obtain the pure (S)-naproxen acid, in twoseparate phase enzyme membrane reactor (TSP-EMR) and in emulsion enzyme membrane reactor (E-EMR) systems. In the TSP-EMR the organic and aqueous phases are separated by the enzyme-loaded membrane and the mass transport through the membrane occurs by diffusion. In this system, it is difficult to obtain the interface at the level where the enzyme is immobilized. Therefore, the observed catalytic activity and enantioselectivity resulted lower compared to the free native lipase, because they were negatively affected by the fact that some enzyme worked in organic phase, some in aqueous phase and only part worked at the organic/water interface. Polyamide membranes allowed to obtain higher enantioselectivity compared to polysulphone membrane, and in general, for both types of membranes, the sponge layer resulted a better immobilization site, compared to the thin selective layer, due to its higher loading capacity. The catalytic stability of immobilized lipase was very good for both type of membranes [70]. In the E-EMR an organic in water (O/W) microemulsion, prepared by membrane technology, with a size distribution from 1 to 4 ~m, was fed to the lipase-loaded membrane from shell to lumen, so that pores into the sponge layer were filled with the microemulsion. The thin selective layer allowed only water to permeate through the membrane whilst the organic phase was retained. Therefore, while the substrate is kept in contact with the immobilized lipase, the product is removed in the permeated water. In this system mass transfer through the membrane occurs by convection, and between the two phases it occurs by diffusion. The great advantage of the E-EMR is that all the immobilized enzyme is able to work at the organic/water interface thanks to the presence of the microemulsion within the pores. This allows to achieve catalytic activity and enantioselectivity for the immobilized lipase comparable to the free native lipase, and much higher stability, with overcome of the product inhibition

194


effects. In this system, each finger of the sponge layer works as a continuous tank emulsion microreactor. The experiments with E-EMR demonstrated that the lower enantioselectivity of immobilized lipase observed in TSP-EMR is due to microenvironment reaction conditions rather than to deactivation of the enzyme due to immobilization. In fact, for comparable amount of enzyme loading, immobilized with the same procedure on the same type of membrane support, the E-EMR showed about 30% higher enantioexcess compared to the TSPEMR.

2.2

Applications Table 2 summarizes some of the applications of immobilized enzymes at large scale. Table 2. Examples of industrial application Purpose Ref. Biocatalyst Immobilization F,scherichia coli Entrapped in Production of L-aspartic acid 56 cells polyacrylamide Entrapped into the Hydrolysis of the milk and whey lactose 57 Lactase (13galactosidase) and fibres of cellulose acetate used for the Synthesis of the dipeptide AspartameTM 58 Termolysin Pseudomonas Immobilized with Production of L-alanine 59 dacunahe glutaraldehyde Glucose-isomerase Reticulate with Production of fructose concentrated 60 glutaraldehide syrups Amino acylase Immobilized on Production of L-amino acids from 61 DEAE-Sephadex racemic mixtures Brevibacterium Entrapped in Production of L-malic acid 62 ammoniagenes polyacrylamide Lipase Entrapped in hollow- Production of (2R,3S)-trans isomer of 49 fiber asymmetric methyl ester of 4-methoxyphenylglycidic membranes acid The major technological difficulties in using biological immobilized systems on an industrial level are related with the life-time of the enzyme; the availability of pure enzyme at an acceptable cost; the necessity for biocatalysts to operate at low substrate concentrations; microbial contamination; the lack of general assumption and the consequent need for individual development on a trial and error basis. Nevertheless, nowadays some new conditions (such as the higher attention of governments towards environment and health care; the need of innovative technologies for a sustainable growth, etc.) may force the industry to develop new, safe, clean, and low energy consumption technologies such as the membrane bioreactors. .....

195


Membrane technologies used in the production of optical pure isomers have been recently described [69, 71]. The major problems in the production of these compounds on a large scale concern: the requirements for expensive cofactors; the low water-solubility of the substrates; the separation and purification of the products from complex solutions. Studies carried out mainly on a laboratory scale indicate that in many cases the use of the appropriate type of membrane and membrane reactor design can overcome these difficulties. For example, UF-charged membrane reactors can be applied to retain cofactors [72]; two separate phase membrane reactors can be used for the bioconversion of low water-soluble substrates [49]. In addition to the use of enzymes in organicaqueous systems, their use in pure organic solvent is increasing. Enzymes in organic media are able to work in microenvironments that contain very little quantifies of water (usually less than the solubility limit). Immobilized enzymes operating in organic media show novel properties, such as enhanced stability and altered substrate specificity. An investigation on transesterification reaction in anhydrous tetrahydrofuran carried out with free and immobilized lipase demonstrated that the enzyme immobilized on zirconia membranes is more active compared with the enzyme in suspension and that the selectivity towards a reaction intermediate changed, resulting in the production of a non-naturally available flavanoid [54]. In many instances, when the product is obtained by fermentation, it is present as a component of a complex solution from which it needs to be separated and purified. In these cases, integrated membrane systems can be used for continuous production and downstream separation. For example, the production of L-lactic acid is obtained by continuous fermentation in a membrane fermentor. This consists of a traditional fermentor combined with an ultrafiltration unit. The solution recovered as permeate contains the product, Llactic acid, together with other small molecules that are not retained by the membrane, whilst the cells and macromolecules are recycled back to the bioreactor. The purification of the product can be achieved by membrane-based solvent extraction carried out through two membrane contactors. Electrodyalisis [73], nanofiltration and reverse osmosis [74, 75], membrane crystallization [76] are other separation processes that can be used in the downstream processing. In recent years, membrane technology has also attracted considerable attention for the treatment of waste-water [77,78]. The reason of this lies on the increase of potable water demand and on the increase of domestic and industrial wastewater discharges linked to the population growth. In addition, more restricted legislation towards environmental care, forces industries to minimize the input of energy and water. Two different membrane bioreactor configurations have been mainly investigated: MBRs with the membrane module outside the activated sludge tank, and MBRs with membrane submerged in the biological reactors. Due to

196


lower energy requirements, the submerged membrane bioreactors have found more success than the external ones. Membranes and membrane bioreactors have been studied for the removal of nitrogen from wastewater [79, 80]. Zoh et al. at University of California, combined a membrane bioreactor to a ceramic cross-flow ultrafiltration module for treating a synthetic wastewater containing hydrolysis byproducts of high explosive RDX (Hexahydro-l,3,5trinitro- 1,3,5-triazine) compound [81 ]. The state-of-the-art of submerged membranes has been recently depicted by R. Ben Aim who has discussed the potentialities (low energy input, long term operation without cleaning, less dependence on variation of rheology behavior with concentration) and limits (high membrane area due to low fluxes linked to the small transmembrane pressure) of the technology. Particular attention was devoted to the use of submerged membranes associated with aeration and on the effect of air bubble size on fouling [82]. Many studies in the field are oriented to the investigation of operating conditions and optimization, such as the effect of air sparging on cell viability; the nitrification rate and COD removal; the use of low cost membranes such as non-woven polypropylene to reduce capital costs; the conversion of organic wastes to energy and fertilizers. Others aim at showing the application at industrial level: Tazi-Pain reported the results of the first year operation of the BIOSEP (R) process, a full scale sequenced aeration MBR for municipal wastewater treatment. He claimed that this is the first plant in the world capable to achieve, in a single tank, the biodegradation of the organic matter, the nitrification-denitrification reactions and the physicochemical elimination of the phosphates [83]. The plant uses an immersed membrane system equipped with MF/UF hollow fibers supplied by Zenon Environmental Inc. Some examples of membrane bioreactors for waste water treatment at large scale are summarized in Table 3. 3.

MEMBRANE BIOREACTORS IN MEDICAL APPLICATIONS

Polymeric semipermeable membranes and membrane processes play a pivotal role in replacement therapy for acute and chronic organ failure and in the management of immunological disease. All clinical blood purification methods employ membrane devices [2].The next generation of artificial organs and tissue therapies is almost certain to be similarly grounded in membrane technology. Membranes of suitable molecular weight cut-off are used in bioartificial organs (e.g., pancreas, liver) using isolated cells, as selective barriers to prevent immune system components from getting into contact with the implant, while allowing nutrients and metabolites to permeate freely to and from cells. The use of membranes in bioartificial substitutes and tissue engineering date back to the 197


Table 3. Membrane bioreactors in water treatment at large scale application

Type of reactor

Type of wastewater

Application

BIOSEP (Submerged membrane bioreactor)

Municipal and industrial

Reduce BOD Removal of solids, nitrogen, phosphorous, etc.

Aerated MBR (Biofilms attached to membrane)

Industrial

COD and ammonia removal

Moving bed biofilm MBR (Submerged membrane)

Municipal

Separate biomass, colloids, soluble organic matter

Zeewed (Zenonsubmerged MBR)

Domestic and pharmaceutical effluents Urban

Reduce COD

Kubota membrane activated sludge

Reduce BOD

year 1933, when Vincenzo Bisceglie in Bail Italy encased mouse tumor cells in a nitro-cellulose membrane and inserted them into the abdominal cavity of guinea pig, to show that cells were not killed by an immune reaction in the pig [84]. Subsequently many researchers focused on the development of immunoprotective membranes to prolong the life of transplant. Currently two important areas of interest are the bioartificial pancreas for the treatment of insulin-dependent diabetes and the liver assist device for the temporary treatment of acute liver failure. Membrane capsules containing dopamine secreting cells also are being explored for treating of Parkinson's disease, a progressive brain disorder characterized by a deficiency of the neurotransmitter dopamine [85]. Immunoprotective membrane cell transplants are being investigated to treat other nervous system disorders. Polymer membranes also are being explored to block cell adhesion or scar tissue formation, for example after surgery, and thus improve wound healing. In addition, the membranes are being investigated for prevention of restenosis (coronary artery narrowing) after angioplasty [84]. In membrane bioartificial organs, cells are compartmentalized by means of semipermeable membranes that permit the transport of nutrients and metabolites to cells and the transport of catabolites and specific metabolic products to blood. The membrane must avoid the contact between xenogenic cells and patient's blood to prevent immunological response and rejection of xenograft. Membranes act as means for cell oxygenation and in the case of

198


anchorage-dependent cells as substrata for cell attachment and culture. As a result, the type of membrane to use in a bioartificial organ must be chosen on the basis of its permeability characteristics as well as on its physico-chemical properties related to the separation process. 3.1.

Membrane bioartificial pancreas

In patients with insulin-dependent diabetes, the cells that normally produce insulin don't function, so patients must receive insulin injections to regulate their blood glucose level. This approach in the long term result in the so-called "diabetic complications" such as kidney failure, neuropathy, which lead to the patient's severe incapacity or death. Transplantation of the whole pancreas, pancreatic tissue fragments or islets of Langerhans would assure the required glucose-related insulin secretion, but the rejection of implants and scarce availability of donor organs limits this therapeutic approach [86]. An alternative approach is the development of membrane bioartificial pancreas using isolated islets of Langerhans or single beta-cells, which are able of sensing the plasmatic glucose concentration and produce insulin amounts related to the actual glycemia, entrapped by means of membranes. In 1970 W.L. Chick and colleagues transplanted isolated islets protected by hollow-fibre ultrafiltration membrane (an acrylonitrile-vinyl chloride copolymer) into dogs made diabetic by surgically removing the pancreas. The device consists of a chamber through which passes a copolymer membrane connected to standard vascular grafts. Islets are placed inside the chamber, through ports in the housing into the cavity, but are outside of the blood stream. Nominal molecular porosity of 80,000 Dalton permits free diffusion of nutrients and insulin across the membrane but inhibits the entry of immunoglobulins and immunocytes from the blood stream into the chamber (Figure 4a) [87]. Since 1970 research efforts were devoted to the development of a hybrid bioartificial membrane pancreas [84-87]. Different bioartificial pancreas were designed in four physical types: hollow fibers, capsule, coatings and sheet. The many different types of prosthetic devices proposed so far can be grouped in three main categories" extravascular devices, intravascular devices and microencapsulated islets of Langerhans [89-98]. In the first case, the tissue is enclosed between membranes, if in a flat sheet configuration, or in the lumen of hollow fibre membranes, and then implanted in an extravascular site (see Figure 4). Other researchers [88] proposed hollow fiber device as extravascular bioartificial pancreas. This device consists of hollow fiber with 0.5 mm diameter, 20 mm length, containing 80,000 islets/ml (Figure 4b). The extravascular systems generally suffer from an intrinsically slow insulin response following changes of blood glucose concentration, limited by the purely diffusive mass transport and by the fibroblastic response of the host.

199

MembraneBioreactorsForBiotechnologyAndMedicalApplications- Drioli membrane i

m

membrane

~w

cavity

m

blood

housing ~x artificialvasculargraft

,ml "

m

m i

a)

m

~

i,m

mu

'~ ~

m l

I

i l

i i

i

mw i

-

pancreaticislets

am,

i

b)

membrane

~cis

u

Microencapsu pancreaticisle

c)

pancreatic islets

d)

~IB 0

~~ ....

blood

~ m

eassist *,,~ ,I~m r @

m e m b r a n e

qr Im ~

4' dmllllk~'$

:, I L Y ~ m Z ¢ . ~

~

17" abrane

insulin

~ ~'~1~,, "" • +

N~=,~

e)

200

J V

Pancreaticislets

M e m b r a n e B i o r e a c t o r s F o r B i o t e c h n o l o g y A n d M e d i c a l A p p l i c a t i o n s - Drioli

Corewithislets

.

,.

.

Meshforphysicalstr

Coating

. . . . . . . . . .

...........................................................................i............................................................................................................................................................................................................................. ................. Outer

layer

g)

h)

Figure 4 Schematicof bioartificial pancreas configurations: a) device developed by Chick et al., pancreatic cells are loaded outside the tubular membrane; b) pancreatic cells are loaded inside of hollow fiber in lumen compartment; c) microencapsulated cells are loaded in the lumen compartment of hollow fiber membrane; d) pancreatic cells loaded outside of hollow fibres in extracapillary compartment; e) cells loaded outside of tubular membrane; f) microencapsulate pancreatic cells; g) pancreatic cells cultured in islet sheets between fiat sheet membranes; h) pancreatic cells inside coating. Intravascular membrane devices are designed so that the membrane separate the graft directly from blood stream of the host. In the Figure 4d cells are cultured outside hollow fiber membranes arranged in a housing in a shelland-tube configuration [89]. These devices suffer from blood clotting at the interface between blood and the synthetic material of the membranes or the point of access, but they are extremely attractive in terms of flexibility of design and use. Additionally, the implant site can be chosen on the basis of reducing the response time of the prosthesis following an increase of blood glucose concentration. A device currently in preclinical development licensed from Circe Biomedical is the PancreAssist Bioartificial Pancreas System [90]. This device consists of a single tubular membrane surrounded by insulin-producing porcine islets, which are in turn, enclosed within a disk-shaped housing (Figure 4e). The porous tubular membrane permit the transport of nutrients and glucose to cells and the transport of insulin from cells to blood. Membrane prevents also the contact between immunological species present in the patient's blood and islets. This device should be implanted near the kidney and surgically connected directly to circulatory's system using vascular graft. In the case of microencapsulated islets, the membrane in the form of alginate gel is formed around the islets of Langerhans, obtaining thus microcapsules with diameters of 300-400 ~tm (Figure 4f) [91, 92]. Microcapsules may make better implants than hollow fibres because they offer better conditions for diffusion of nutrients to the insulin-producing cells and

201


waste products from them. Normoglycemia has been reported after intraperitoneal transplantation of alginate-polylysine microencapsulated allogenic and xenogenic islets in diabetic animal models and recently also in men. However, graft survival is always limited to several weeks. Graft failure is interpreted as non-specific immune response i.e., foreign body reaction against the microcapsules resulting in a progressive overgrowth of the capsule and subsequent necrosis of islets. Researchers focused study on highly purified alginate and other biochemicals more biocompatible [91] and on novel membranes able to prevent permeation of low molecular weight humoral molecules released by xenogenic islets. The Islet Sheet is a thin planar bioartificial pancreas, licensed from Islet Sheet Medical Company, contains live, functional islets in an artificial polymer matrix (Figure 4h) [93]. Each sheet is several cm in diameter and contains 2 to 3 million cells and contains islets microencapsulated within in a mesh to increase the physical strength between two layer of semipermeable alginate membrane. A semipermeable membrane such as 0.2 ~tm cellulose ester filter membrane is saturated with crosslinking solution. The cells 4 to 6 sheets contain enough islet tissue to cure diabetes in an adult. The sheet is so thin (the overall thickness is 250 ~tm) that diffusion alone allows sufficient nutrients to reach the center of the sheet. A coat on the exterior of the sheet prevents contact between the cells inside and immune effector cells of the host as well as inhibiting diffusion of antibody and complement. The alginate membranes show high permeability to different solutes and excluded immunocompetent species. No immune suppression drugs are needed. The sheet may be removed or replaced at any time. Experiments on islet sheet began at the University of Chicago in September 1998, and moved to the University of Cincinnati and the University of Alberta in 2000. Large animal studies given encouraging results. The implantation of islet sheet in omentum of pancreatectomised dog permitted to return blood sugar to normal and at 60 days the blood sugar was lower at every measurement time [94]. 3.1.2. Membrane bioartificial liver

Each year in the US, approximately 150,000 people are hospitalized with liver disease and over 43,000 people die from it [95]. Transplantation, the only effective means of treating liver failure, is not an option for many patients. Ironically, the liver is a highly regenerative organ [96]. Supportive therapies as dialysis, hemofiltration, hemoperfusion for patients with acute liver failure have been proven ineffective because physical methods are not sufficient for the management of severe biochemical disorders. The development of an extracorporeal liver assist device, using isolated liver cells, to which a patient would be temporarily connected until he/she recovered or received a liver transplant could be a promising approach for 202

Membrane

Bioreactors

For

Biotechnology

And

Medical

Applications

-

Drioli

patients with liver failure. Since fulminant liver failure is potentially reversible, the extracorporeal bridging of liver function would also be beneficial until the patient's own liver resumes functional activity.

hepatocytes

membrane

m

I i m

blood

m

-

J

mlaam mm ,m~

am*

*am *D .m am am

am

b)

a) microcamer-attached am

blood-

m

_ n

-.,m

I,m

. _ --

.m

m~

blo

he~atc~es

blood " - _ -

;

-.d)

c) _

med

_

~

_

_

membrane

blo

/

g:ltrapme

ocytes blood

_

-_---

m

-_-----_-

e)

203

_-

-

M e m b r a n e B i o r e a c t o r s F o r B i o t e c h n o l o g y A n d M e d i c a l Applications - Drioli

membrane for • oxy

non-woven polyester matrix

hepatocytes

0 !. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

"1

hepatocytes

g) Figure 5: Schematic of liver support device configurations: a) microencapsulated hepatocytes; b) hepatocytes loaded outside of hollow fibres in extracapillary compartment; c) hepatocytes loaded between flat-sheet membranes; d) microcarrier-attached hepatocytes in the extracapillary compartment; e) hepatocytes loaded in a spirally wound non-woven polyester matrix; f) hepatocytes entrapped in a three-dimensional contracted gel matrix inside of hollow fibres; g) hepatocytes loaded in a network formed by four separate membrane capillaries with different functions.

The first clinical report of bioartificial membrane liver was released in 1987 [97]. This device consisted of a hepatocyte suspension that was separated from the patient's blood by a cellulose acetate dialysis membrane. Since different devices have been proposed (Figure 5). These devices differ from one other for the type of material used to construct the bioreactor: membrane, glass or non-woven fabrics; for configuration: fiat or hollow fiber; for coating used: matrigel, collagen; for cell culture technique" microcarrier, spheroids, aggregates, etc, and for cell capacity. In devices using hollow-fibre membranes, isolated hepatocytes are usually loaded outside of the hollow fibres in the extracapillary compartment, while blood, plasma, or culture medium flow through the lumen of the hollow fibres (Figure 5b). In devices using fiat-sheet membranes, hepatocytes are loaded between fiat membranes in a sandwich way,

204


while blood or culture medium flow outside of the membranes as shown in Figure 5c. Cells may be free in suspension, attached to walls or attached to microcarrier (Figure 5d). A key issue concerning the development of bioartificial liver is the maintenance of long-term viability and functions of hepatocytes: oxygen transport resistance and catabolite accumulation may limit the hepatocyte viability and metabolism [98-99]. To overcome the problem of maintenance of viability and liver specific functions in vitro of liver cells, Sussmann [100] used immortalized liver cell line in the extracorporeal hollow fibre device as biological element. This device, developed as Hepatix/Baylor unit, consisted of 10,000 individual hollow fibres with cut-off of 70 kDa. This has been tested on 11 patients with fulminant hepatic failure: four patients were sustained until transplant and two survived without the need of transplant and five died for complications. A disadvantage of this unit is the potential of seeding tumor cells into patients. Thus, other researcher proposed new culture models to address the problem of long-term maintenance of liver specific functions of liver cells inside of the bioreactor. Flendrig et al. [101], developed a bioreactor based on a spirally wound non-woven polyester matrix, three-dimensional framework for hepatocyte immobilization and hydrophobic polypropylene membranes for decentralized oxygen supply and CO2 removal (Figure 5e). Medium or plasma is perfused through the extrafibre space and in direct hepatocyte contact. Generally many investigators inoculate cells in the extrafibre space. In Nyberg's design hepatocytes are cultured in the hollow fibre lumens [102]. In this device hepatocytes are entrapped in a three-dimensional gel matrix and the extrafibre space of the bioreactor is perfused for 24 h with recirculating medium, as is shown in Figure 5f. The blood is passing through the extrafibre compartment and the gel-entrapped cells are perfused by means of medium to improve the supply of oxygen, nutrients, and the removal of catabolites. This device is scaled-up by using hepatocyte spheroids inside polysulfone membranes at University of Minnesota and recently was licensed to Algenix Inc. and it was approved by FDA for Phase I human clinical trials. An interesting device was designed by Gerlach et al. [103] by mixing different membrane fibres in various directions, thus improving the oxygen and nutrients supply to hepatocytes located in between such a network. The Figure 5g shows the scheme of such device licensed to Hybrid Organ GmbH. Four separate capillary membranes with different functions are utilized: plasma inflow by polyamide membrane; plasma outflow by polysulfone membranes; decentralized oxygen supply and carbon dioxide removal with low gradients by polypropylene membranes; sinusoidal endothelial coculture by hydrophilic membrane. This device was used for preliminary clinical studies. Based on dialysis model, the liver support system developed by Demetriou et al. [ 104] of Cedar Sinai Medical Centre of Los Angeles, consists 205

Membrane

Bioreactors For Biotechnology And Medical A p p l i c a t i o n s - Drioli

of a bioartificial liver cartridge that contains billions of pig liver cells and a machine that controls the flow of blood through the cartridge (Figure 5d). The commercial name of the device is "HepatAssist System" of Circe Biomedical and includes cellulose nitrate/cellulose acetate hollow fibre bioreactor containing porcine hepatocytes attached to a collagen-coated microcarriers, two charcoal columns, a membrane oxygenator, and a pump. Plasma from a patient flows through the cartridge. Pig liver cells on the outer surface of the fibres are in contact with the plasma through large pores on the fibre surface. This device was used in USA for a treatment of 6-7 h of 25 patients with severe acute liver failure. These reports encourage the further studies to address the problems related to the development of a bioartificial liver. In fact, although several groups worldwide are creating systems there is not still an ideal device. This is not surprising, given the complexity of liver functions. Liver support system must supply various liver specific functions including synthetic and detoxification activities for a time sufficient to allow recovery of a patient or maintenance of patients until transplantation. Recently, we evaluated in vitro the performance of a full-scale fiat membrane bioreactor (FMB) developed by Bader et al., that permits the culture of liver cells under in vivo-like conditions and at high-density culture [105-109]. In such bioreactor porcine hepatocytes are cultured within extracellular matrix between oxygen permeable fiat-sheet membranes. Isolated liver cells are located at a distance of 10-20 ~tm of extracellular matrix. This bioreactor provides culture conditions that improve liver specific functions of liver cells in vitro. The FMB is able to provide an in vivo-like microenvironment for liver cells: hepatocytes are arranged as a plate in 3-D coculture with intermingled nonparenchymal cells. In contrast to other bioreactors the FMB is based on the organization of liver cells as a plate within extracellular matrix in which each individual hepatocyte has its own membrane support and thereby its own oxygen supply position. In vitro studies demonstrated that the performance of a scale-up FMB using porcine hepatocytes is stable over a period of about 3 weeks and compares well with that of other systems present in literature. Isolated hepatocytes cultured in the FMB reconstitute many of the features of the liver in vivo. The cell concentration inside of the FMB increased in the first days of culture and then remained constant until 18 days. Specific metabolic functions of hepatocytes in terms of albumin synthesis, ammonia elimination and urea synthesis are sustained for the investigated culture time demonstrating thus the long-term maintenance of functional integrity of hepatocytes cultured in the FMB.

206


3.1.3.~Cytocompatibility of membranes in bioartificial organs In membrane bioartificial organs using isolated cells as biological component, semipermeable membranes play more functions: they act as immunoselective barriers, as means for cell oxygenation and provide a large area for cell attachment. All these functions are important for the maintenance of cell viability and specific functions. In our experimental study we demonstrated that isolated rat liver cells cultured on oxygen-permeable membranes reconstitute many feature of the in vivo [98]. Cells cultured on polythetrafluoroethylene oxygen-permeable membranes maintained a morphological appearance of hepatocytes similar to their in vivo appearance and exhibited at high levels tissue specific functions in vitro in terms of albumin secretion, urea synthesis and drug biotransformation functions. This finding is of relevance for oxygen supply to cells in vitro both in batch and in large-scale bioreactor systems. In a membrane bioartificial organ, cells come into contact with the membrane surface. Therefore, the response of the biological components depends on surface properties of the used membrane. Physico-chemical properties including surface composition, surface charge, surface energy, and surface morphology, may affect cell adhesion and behavior. Recently, studies performed on isolated hepatocytes cultured on semipermeable polymeric membranes have indicated that wettable and rougher surfaces enhance adhesion and metabolism of isolated hepatocytes [ 110-112]. In particular, measurements of the wettability of membrane, expressed by the contact angle in the presence of different liquids permitted to evaluate and to compare surface free energy components of membranes with different physicochemical properties. Such measurements before and after modification of native membranes in culture medium resulted to be predictive index of their cytocompatibility and/or tissue biocompatibility. Therefore a material surface treatment might enable the adaptation of its surface free energy to biological requirements. The adhesion and activity of liver cells are affected by hydrophilic properties of membranes: surface free energy and its components. Isolated hepatocytes formed on membranes with high base parameters of surface free energy, such as cellulose acetate, polycarbonayte membrane, large cellular aggregates while on hydrophobic membranes such as polypropylene membranes cells spread to a large extent and the edge of cells were often indistinguishable in a network formed by fibrous proteins of extracellular matrix. For each membrane we observed that cell adhesion increased with increasing base parameter of membrane surface tension. In Figure 6 are reported the cell adhesion after 48 h of culture on membranes with different base parameter of surface tension. Liver cells adhere strongly on membranes with high value of base parameter. The absolute value of cell adhesion is higher in the presence of

207


serum proteins adsorbed on the membrane surface, which change the wettability by increasing base parameter of surface tension. Adsorption of proteins from surrounding liquids is one of the key steps for the cytocompatibility of the membrane. We observed that the extent of serum protein adsorption is relatively large on hydrophobic membranes and relatively small on more hydrophilic membranes [ 112].

e~

E 0

~

76-

(l) r,.)

,.n

5-

li

.". 4 o

•~

3-

.

2 I

0,5

I

2,9

I

7,4

12,4

30

y- [m Jim 2] Figure 6: Cell density after 48 h of culture on various membranes with different surface tension base parameter calculated according to Good van Oss's equation. The number of cells adhered on various membranes was estimated by counting at the light inverted microscopy after their removing from membrane surface. The cell adhesion values are the mean of eight experiments: + standard deviation.

The base parameter of surface free energy affected the metabolic functions of liver cells. We found a correlation between urea synthesis and base parameter of membrane surface tension. For each investigated membrane cells synthesise urea with a rate that increases with increasing of base parameter value of membrane surface tension (Figure 7). The metabolic activity is particularly expressed at high levels when cells are cultured on PC and CA membranes. Figure 7 shows that on CA and PC membranes the ability of liver cells to synthesize urea increases linearly with increasing the value of base parameter of surface tension and reached the value of about 120 mg/h in correspondence of the surfaces with respectively 46 mJ/m 2 for CA membrane and 54 mJ/m 2 for PC membrane. On such membrane surfaces cells exhibited rates of urea synthesis that significantly increase with time. These results suggest that there is a marked

208

MembraneBioreactorsForBiotechnologyAndMedicalApplications- Drioli

effect of the physico-chemical properties of membranes on cell adhesion and functions. Protein adsorption modified the surface tension components of the membrane and the consequent cell adhesion and functions. As a result, 100 r-""'l

""

90-

E

80-

.-~ 7 0 ¢~ 6 0 e-, t-

>, 5 0 t~ 4 0 -

= o

3020-

~

10-

(D

O-

r 0,5

r 2,9

T 7,4

12,4

-30

?-[m Jim 21 Figure 7: Urea synthesis rate of liver cells cultured on membranes with different base parameter of surface tension after 24h (full bar) and 48 h (shade bar) of culture. The metabolic rates were determined by culturing cells in the presence of 1 mM ammonia in the medium. The reported values are the mean of eight experiments' + standard deviation.

independently on the type of native polymeric membranes it is possible to improve cell adhesion and specific functions by changing their surface tension and their components. On the basis of our results, CA and PC membranes resulted to be good substrata for liver cell culture in vitro promoting cell adhesion and specific functions in term of urea synthesis. To this purpose, researches on interactions of cells with novel membranes made in our laboratory are running in order to develop more cytocompatible and biocompatible membranes.

3.1.4. Membrane biocompatibility When blood comes in contact with foreign substances by extracorporeal circulation, reaction to foreign body occurs in the blood. Thus, to evaluate the safety of the material is used a reaction to foreign body as an indicator of "biocompatibility". A biocompatible membrane is a membrane more affinous is to living body and that gives less reaction to foreign body.

209

Membrane Bioreactors For Biotechnology And Medical Applications -

Drioli

Materials in blood contact often initiate coagulation processes, affect platelet morphology and function and lead to an immunologic answer of the body. The first event of a complicated sequence that leads to thrombus formation and immunoreactions is the protein adsorption, which plays a mediating role in bioaccumulation, systemic foreign body reactions, tissue regeneration and is, therefore, crucial for biocompatibility and performance of medical devices. Thus, the development of biomedical polymers has to comprise the detailed investigation of protein adsorption. Several studies indicated that platelet activation increases in the positively charged membrane and decreases in the membrane having a microdomain structure in which hydrophilic groups and hydrophobic groups coexist randomly as molecules. Also the complement activation, which in the case of biomaterials proceeds via alternative pathway, is manly noted with cellulose membranes; free hydroxyl groups on the membrane surface is bonded with C3b and further with factor B which promotes the activation. Improvements of biocompatibility of cellulose membrane can be obtained by surface modification devoted to graft for example alkyl group to the hydroxyl group. Considering that functions of biointerfaces are mainly based on principles of molecular recognition, several powerful methods of surface functionalization including plasma treatments, grafting of reactive chains, coating of functional layers and others can be used to engineering news modified surfaces [113-114]. Furthermore, in addition to original cellulose membrane, a variety of synthetic polymers less potent in complement activation, can be used clinically. 30 ,.--, tN

E

25

t-

20

0 o) ::L ' - -

.O ~

2offl

15

c

10

0 c

"=

..1o ~ U_

5 0
13) in presence of NaBH4 as catalyst. The second step is the immobilization of the ligand to the matrix via the spacer arm. The reaction is conducted in a mixture containing a high concentration of amylose, pH>13, in presence of NaBH4 at 37 °C. Finally, the unreacted epoxy groups after treatment with amylose solution, are blocked by soaking the functionalized membranes in a solution of mono ethanolamine, pH-9.5, at 25 °C. Amylose affinity membranes can be stored in water or in standard buffer solution at 4 °C until use. No degradation of the support has been observed over a period of 6 months [12-14].

3.2

Protein Concentration Measurement MBP-intein-CBD (97 kDa) concentration was measured performing the Bradford assay. Bradford reagent from Sigma was used. A calibration curve was

268

Use Of Micro-Porous Affinity Membranes For Protein Purification: A Case

S t u d y - Sarti

prepared by using BSA (Sigma, fraction V, 96-99%) as standard protein. The absorbance of the dye-protein complex was measured at 595 nm [ 15-16]. 3.3

Purification of MBP-Intein-CBD from Cell Lysate The fusion protein MBP-intein-CBD was obtained from genetically modified E. Coli strains ("IMPACTTM-CN '' system, from New England Biolabs Inc. ) [9, 17]. A p-TYB vector allows the fusion of the self-cleavable intcin tag to the C-terminal of the MBP group from one side and to the chitin binding domain to the other side. This procedure can be generalized by substituting the CBD domain with a different target protein. Isopropyl-[3-D-thiogalactopyranosidc (IPTG) is recognized as inductor for the expression of the fusion protein. E. Coli was grown following the conditions indicated by the manufacturer. IPTG was added during the last growing stage to stimulate the production of the recombinant protein inside the bacteria. The fusion protein is expressed as an intracellular protein. The cell culture, therefore, is harvested, the cells resuspended in buffer solution then sonicated a T = 4 °C. The cell extract is recovered, centrifuged for the removal of cell walls and debris. The sumatant, containing the protein of interest, is collected and stored at-20 °C until use. MBP-intcin-CBD is purified by a separation process conducted using amylosc affinity membranes. Flat sheet affinity membranes were located inside a membrane holder, realized in order to offer a high binding area. The total membrane area can be varied depending on the binding capacity needed to perform a specific separation process [ 18]. In particular, the membrane holder is realized with a series of discs arranged in a column configuration. On each disc a number of membranes can be located. In this work, a total membrane area of 1000 cm 2 has been used, corresponding to stacks of 3 membranes for 17 stages. The single discs arc designed in order to offer a uniform flow distribution over the cross section all along the membrane stack (Figure 1). The purification procedure is similar to the one adopted to perform the purification of other MBP-fusion proteins [19-20]. A volume of cell lysatc is diluted with buffer solution (0.5 M NaC1, 10.3 M EDTA, 0.02 M Tris HC1, pH=7.4). After a conditioning step performed on the stationary phase, the protein mixture to be purified is loaded into the circuit. The feed solution is rccirculatcd maintaining a flow rate of 12 ml/min for 80 minutes. In this condition, the specific interaction occurring between the MBP domain and the ligand present on the membrane surface allowed for the immobilization of the desired product on the membrane surface. When the adsorption step is accomplished, the stationary phase is washed with an a-specific buffer solution (0.2 M NaC1, 10.3 M EDTA, 0.02 M Tris HC1, pH=7.4) in order to remove all the fragments trapped on the membrane surface and not specifically bound. Finally, the adsorbed material is clutcd applying a buffer solution containing maltose as competing substratc for the MBP active binding site and a high 269

Use Of Micro-Porous Aft'mity Membranes For Protein Purification: A Case Study - Sarti

concentration of NaC1 (0.1 M Maltose, 0.3 M NaC1, 103 M EDTA, 0.02 M Tris HC1, pH=7.4) [21]. The product is recovered in several subsequent fractions. The protein fractions collected are then analyzed through an electrophoresis technique, using the CRITERION system from B io-Rad. Protein bands are stained using the Blue Coomassie method.

Figure 1. Large scale membrane holder, suitable for flat sheet membranes. A series of discs are arranged in a column configuration. On each discs a stack of membrane is located.

The solutions recovered during the elution stage contain an excess of unbound maltose and NaC1, besides maltose bound to the active site of MBP. A dialysis step is then performed in order to remove the excess of NaC1 as well as of maltose either free in solution or bound to the MBP domain. The conditions adopted are the same as in a previous work [22]. Several dialysis stages have been performed, lasting 24 hours each for a total time of 5 days. The system was kept at 4 °C, mechanically stirred. Pure protein solutions are thus obtained, with all the MBP active sites free of maltose.

3.4

Adsorption of Pure MBP-Intein-CBD in Batch System Pure MBP-intein-CBD solutions have been used in adsorption experiments conducted in batch system using amylose affinity membranes. In a volume of pure protein solution, variable between 10 and 14 ml, a number of amylose membranes are immersed (dm=4,7 cm). The system is kept at 4 °C and is mechanically stirred. The adsorption reaction is followed until an equilibrium condition is reached. The concentration of the protein in the liquid solution is monitored at regular time intervals, via Bradford assay and the adsorption curves are thus obtained. 270

Use Of Micro-Porous AtTmity Membranes For Protein Purification: A Case Study - Sarti

A second set of experiments has been performed loading pure MBP-intein-CBD solution, in a continuos flow mode, on a stack of 5 membranes located inside a stainless steel Millipore® membrane holder (din=4.7 cm, cross section diameter 3.8 cm). The permeate is recirculated to the stationary phase, at a flow rate of 4 ml/min, until a steady state condition is reached. The concentration in the liquid solution is monitored at regular time intervals in order to measure the kinetic of the adsorption reaction. Once a steady state condition is reached, the membranes are removed from the holder, and soaked in the recovered protein solution until a further equilibrium condition is reached. The two subsequent adsorption stages are performed at 4 °C. 3.5

Elution of Immobilized MBP-Intein-CBD in Batch System First of all, adsorbed membranes are washed with an a-specific buffer and then elution is performed by applying 20 ml of maltose buffer solution. Both stages are carried out in batch system, at 4 °C under mechanical agitation. The concentration of the released protein is monitored until a constant value is reached. 4.

RESULTS

4.1

Purification of MBP-Intein-CBD from Cell Lysate The complete saturation is achieved after 6 to 10 passages of the feed solution on the stationary phase corresponding to an adsorption time of about 80 minutes. The washing step is performed at a flow rate variable between 25 and 35 ml/min. The liquid solution is forced throughout the stationary phase without any recirculation. The efficiency of the washing step in the removal of the nonspecifically adsorbed materials is crucial in order to get a pure product at the end of the separation process. UV analysis performed on the washed-out fractions, collected at regular time intervals, shows a rapid decrease of the amount of materials dragged by the washing solution flowing out from the stationary phase. The washing step is completed in a time between 30 and 45 minutes. Electrophoresis analysis confirmed that in the last fractions collected during washing, no traces of impurities are present [20]. The elution of the specifically adsorbed material is obtained applying maltose buffer. Experimental data confirm the effectiveness of the combined action of maltose, as competing substrate, and of the ionic strength of the solution in the removal of the desired product. The recovered fractions are characterized by high purity (Fig. 2) and high protein concentrations, as showed by the elution profiles reported in Fig. 3. The two curves in Fig.3 refer to experiments performed maintaining different flow rates during the elution step. At higher flow rate, the peak of the elution profile is shifted towards higher elution volumes and the maximum concentration value is lower. This behavior can be seen as an indication of the 271

Use of Micro-Porous Affinity Membranes For Protein Purification: A Case Study - Sarti

F

E1

M

E2

E3

MBP-intein-CBD M W 97 k D a

! . . . .

Figure 2. Electrophoresis analysis of protein samples obtained from the purification process of MBP-intein-CBD using amylose affinity membranes. (F = feed, Ei = i-th eluted fraction, M -- Molecular weight Markers). 0.7 flowrate 48 ml/min ~ Exp. 1

0.6

"_ flowrate 6 ml/min [] flowrate 6ml/min ~

0.5

~

8

Exp. 2 Exp.

0.3 0.2

0.1

0

50

100

150

volume (ml)

Figure 3. Elution profiles obtained in the separation process of MBP-intein-CBD, at two different flow rates: flowrate 6 ml/min (I--!), flowrate 48 ml/min ( . ) continuing with flowrate 6 ml/min ( • ) . combined effect of the mass transfer resistance within the liquid phase and the kinetics of the desorption reaction on the overall elution rate. It must be noticed, on the other hand, that an overall elution time of almost 15 minutes is required to complete the recovery of the adsorbed product, while several hours are necessary when affinity resins are used as stationary phases.

272

Use Of Micro-Porous Aff'mity Membranes For Protein Purification: A Case Study - Sarti

Thus, the use of micro-porous membranes, characterized by an open porous structure, strongly reduces the limitation due to the diffusion of the biomolecules within the liquid phase, allowing shorter processing times for each single stage of the entire separation process. This aspect is one of the main advantages associated with the use of modified membranes as stationary phase in chromatographic processes. Furthermore, higher flow rate can be applied with reduced pressure drops along the column, preserving the mechanical stability of the stationary phase and the biological activity of the immobilized product [19,

20]. The binding capacity has been determined from the amount of recovered product; the value obtained per unit total external area is equal to 0.015 mg/cm 2 or equivalently 1.6 10v mmol/cm 2 considering a molecular weight of 97 kDa; the capacity per unit volume of the membranes is 1.05 mg/ml. This value is in excellent agreement with what obtained by using the same stationary phase to purify MBP-fusion proteins of different molecular weights [19, 20]. Interestingly, the capacity observed is also comparable with the typical values encountered for resins characterized by the same specific interaction towards the MBP domain.

4.2

Equilibrium Isotherm Adsorption experiments of pure protein on amylose affinity membranes, performed in a batch system, offer equilibrium data, reported in terms of the pair c and q* in Fig. 4; the curve corresponding to the interpolation of the equilibrium trend with the Langmuir isotherm is also shown. 0.018 0.016 0.014 0.012 ft,, 0.01 0.008 experimental data

•

0.006

m

Langmuir isotherm

0.004 0.002 0

r

i

i

i

i

r

0

0.05

0.1

0.15

0.2

0.25

0.3

C* (m~/ml)

Figure 4. Experimental equilibrium data interpolated by the Langrnuir isotherm. The data are relative to adsorption step of pure MBP-intein-CBD on amylose membranes in batch system.

273

Use Of Micro-Porous Aff'mity Membranes F o r Protein Purification: A Case Study - Sarti

From the fitting of the experimental data one determines the values of the parameters qm and Ka equal to 0.018 mg/cm2 and 0.023 mg/ml, respectively. Steady state data have been also obtained from the series of adsorption experiments performed in the continuos flow configuration. The steady state points (Cinf~ qinf) are reported in Fig. 5 and compared with the equilibrium Langmuir isotherm. This comparison is presented in order to show the effect of convection on the amount of protein adsorbed at steady state. A flow rate equal to 4 ml/min has been applied, in order to realize a surface velocity comparable with that maintained during the adsorption step of the separation process, performed in the membrane column. 0.018 0.016 0.014

f

-~ 0.012 0.01 0.008 0.006 ~X

0.004 0.002

i

0

0.05

i

0.1 c* (m g/m i)

i

0.15

0.2

Figure 5. Comparison between equilibrium and steady state isotherm in flowing systems. Continuous line is the equilibrium Langmuir isotherm. In each pair, the point in the rightmost position (empty marker) refers to steady state conditions, the leflmost point (gray background) refers to the subsequent equilibrium conditions in batch mode.

As clearly evident, there is always a good agreement between the steady state data and the equilibrium curves, even though a certain scatter of data can be observed in the region of low concentration. The effect of the flux of the liquid solution and consequently of its contact time within the stationary phase is negligible when the concentration in the bulk is large enough to reduce the effect of the concentration gradient between the bulk and the solid interface. It is interesting to point out that the equilibrium adsorption isotherm obtained in batch experiments is exactly the same both for new membranes and for membranes previously used in continuous flow adsorption.

4.3

Adsorption Kinetics The concentration of the protein in the liquid solution has been recorded versus time during adsorption. A common experimental behavior is reported in 274

Use Of Micro-Porous Affmity Membranes For Protein Purification: A Case Study - Sarti

Fig. 6. The experimental trend can be interpolated with a suitable kinetic model in order to determine the value of the kinetic constant of the binding reaction k~. In this work we have used a short times analysis, valid under the hypothesis that in the initial adsorption times the desorption reaction is negligible and the adsorption rate is a function of the feed concentration Co and of the total membrane capacity qm (Equation (7)). 0.25

0.2

o.15

~ o.1

0.05

0

0

50

100

150

200

time (min)

Figure 6. Experimental adsorption curve of pure MBP-intein-CBD on amylose affinity membranes performed in batch system, T=4°C. The values of kl relative to the three different series of adsorption experiments are reported in Fig. 7. Apart from some inevitable scatter due to protein concentration measurements, the results obtained are quite consistent with each other. The values of the kinetic constant of the desorption reaction k2 can be then estimated once the kinetic constant kl and the equilibrium constant K,~ are known. In order to test the consistency of the analysis, the characteristic rate of the forward reaction, kl*Co, is compared with the rate of the backward reaction k2 in Fig.8, at all the feed concentrations inspected. As it is apparent, at low concentration, say below 0.05 mg/ml, the rates of the adsorption and desorption reactions are of the same order of magnitude, while at higher concentrations the rate of the adsorption reaction is indeed much larger. Thus the short times analysis performed is valid only for protein concentrations exceeding the value of 0.05 mg/ml; in that range the scatter in kl values is indeed rather reduced (Fig.7). On the other hand, the low concentration range is less interesting for the adsorption stage.

275

Use O f M i c r o - P o r o u s A t T m i t y M e m b r a n e s F o r P r o t e i n P u r i f i c a t i o n : A C a s e S t u d y - S a r t i

0,025 • Series 1

x Series 2 z~Series 3

X

0,020

.-., 0,015

.~ 0,010

*# x

x~

•

0,005

0,000

0

i

i

i

i

i

0,05

O,1

O,15

0,2

0,25

0,3

C0 (mg/ml)

Figure 7. Estimated values of kl using the short times analysis model. Series 1 ( • ) refers to adsorption of MBP-intein-CBD in batch system; series 2 (*) refers to adsorption performed on continuos flow conditions, and series 3 (A) refers to adsorption performed in batch system using membranes previously used in series 2.

The conditions maintained during adsorption are not favorable for the release of adsorbed material, independently of the concentration of product immobilized on the solid surface. That explains why for the recovery of a specifically adsorbed protein, a competing substrate must be added to the elution buffer and its ionic strength must be enhanced. 1,0E-02

.

.

.

.

.

...-

_~_ 1,0E-03

I_':l" •

P

El

1,0E-04

o k l *cO I [ ] 11,3

1,0E-05 0

0,05

0,1

0,15

0,2

0,25

0,3

c o (mg/rnl)

Figure 8. Comparison between the characteristic rates of the forward and backward reactions during adsorption of MBP-intein-CBD on amylose affinity membranes in batch system.

276

Use Of Micro-Porous AtTmity Membranes For Protein Purification: A Case Study - Sarti

The estimation of kl allows, also, the determination of the characteristic time of the binding reaction tR defined from the following relation:

tR =

V/,t

Ak, qmO-O )

where Vi,/A represents the ratio between the volume of the liquid within the pores, V/,t, and the total area of the membrane sheets; 0 represents the surface fraction already occupied by adsorbed molecules [2]. Considering never used stationary phases, 0 is equal to 0; for convenience the relevant parameter values are reported in Table 2. It is interesting to compare the characteristic time of the binding reaction tR with the characteristic time of protein diffusion within the pores tD, and with the residence time within the pores, to [2]. The corresponding values are reported in Table 3 together with the total interaction time, tg, defined as a sum of tD and tR. The apparatuses used in this work operate at a total interaction time larger (membrane column) or much larger (Millipore holder) than the residence time tc; thus the operating conditions are definitely not optimized. Indeed, to ensure efficient capture of the ligand the condition ti" easier upscaling )~ separation of large biomolecules like endotoxins and virus Especially the capability of binding of large molecules is an advantage in comparison to gel chromatography. Typical gel media have pores in the range of about 30 nm in comparison to 3 - 5 ~tm in membrane adsorbers. Therefore membrane adsorbers can be used for molecules where gel media are not working because of size exclusion effects.

Figure 1: Comparison of a Membrane Adsorber membrane Sartobind Q with Q-Sepharose FF The larger pores of the membrane compared to the beads are clearly visible allowing access of large biomolecules like virus particles to the ion exchange sites located on the inside of the pores.

284

Economic Production Of Biopharmaceuticals By High-Speed Membrane Adsorbers - Melzner

Figure 1 shows a typical 3 ~tm microporous membrane structure in contrast to a commercially available beaded chromatographic material i.e. Sepharose Fast Flow Q-Type. 2.

MODULES AND PROCESS DESIGN

For production and large-scale application the Sartobind ® Factor-Two Family of membrane adsorber modules has been developed. The modules consist of a membrane adsorber reeled up like a paper roll to form a cylindrical module sealed at both ends with POM (polyoxymethylene) caps. For scaling up, the modules have areas between 0.12 m2and 8 m 2. The different module heights can be purchased with 15, 30 or 60 layers of membrane. In combination with the different heights a variety of 15 large-scale modules are available. Since the direction of flow is from inside to the outside of the Membrane Adsorber cylinder, a solid core of the appropriate size is inserted into the module to keep hold-up volume as small as possible. The solid POM cores are as well available in lengths of 3, 6, 12, 25 and 50 cm and the thickness varies with the number of membrane layers used. The module is inserted in a specific housing, which consists of a top and base plate, the housing tube and the solid core. For operating the system the unit is filled first with starting buffer. The feed solution enters the unit at the top (see Fig 2) as indicated by the red arrow.

Figure 2: Schematic cross section through a Factor Two Family module and appropriate housing. The direction of the flow is indicated by the arrows. For details see text.

285


The central cylindrical core distributes the fluid to the inside of the module. The flow is directed from the inner channel radial through the module to the outer channel (see arrows in Fig.2). The permeate leaves the housing at the bottom plate (blue arrow). For large-scale protein isolation the adsorber modules of different sizes can be combined to achieve a desired yield and productivity. The modules offer unprecedented high flow rates and short cycle times in the range of minutes. For process plants, parallel running modular units can be combined with modules in series. A system for the extraction of hemoglobin from bovine blood is shown in Fig 3. Bovine blood as starting material was used as a model system because it is relative safe and easy accessible in large quantities and is representative for numerous biological substances.

ixt ens i on Co re

2 x S-8OK-60-fiO

S-40K-30-50

Set ioi Connectors

S-10K-15-25

Figure 3" Cross section of 21 m2 3 stage pilot plant. Not true to scale (reduced in length to 1/5). 1, venting valves; 2, pressurized air connectors for the actuation of the pressure plates. All components are made from AISI 316L (1.4435) stainless steel, the pressure plate is made from POM.

286


The details of the experiments are recently published [21 ][23] so we will focus here on the special features of parallel and serial connection of modules which makes the system so versatile and attractive. For small multistage plants a serial connector is available, which can be mounted between the housings and leads the flow through from the outer channel of the preceding (upper) stage to the inner channel of the following one. A 21 m 2 three- stage pilot plant based on this concept was assembled in this way (Fig.3) and has been operated successfully. The module equipment is listed in Table 1. Table 1" Module equipment (sulfonic acid type) of the pilot plant used for break through curve experiments.

Stage module type

# of modules

membrane

in Fig. 3

% of total

a r e a [m 2]

area

1

S-80K-60-50

146 + 147

16

76

2

S-40K-30-50

150

4

19

3

S-10K-15-25

152

1

5.0

3.

BREAK THROUGH CURVES

Break through curves for purified hemoglobin were measured at the individual modules (Fig. 4) in separate housings.

2.0

............................

#152 S 10K-15-25

1.8

E ,4 1.2

#14;

1.0 ~e.O.8

#146 S 8013.2.

293

Economic Production Of Biopharmaceutieals By High-Speed Membrane Adsorbers - Melzner

10.

CONCLUSIONS

The demonstrated membrane adsorber systems represent versatile and economic tools for applications in the biopharmaceutical field and show the possibilities to work in areas where conventional purification systems fail. This is especially important in new or growing fields in the pharmaceutical and biotechnological industry, i.e., isolation of viral vectors or even particles, antibodies or nucleic acid molecules. Additional benefits can be defined as easy set up and handling of a fully validated system which can be used in a "single shot" mode reducing validation cost to a minimum. REFERENCES

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23]

C. Charcosset, J. Chem. Technol. Biotechnol. 71 (1998)95-110. I. Adisaputro, Y. Wu, M. Etzel, J. Liq. Chrom. and Rel. Tech. 19(9) (1996) 1437-50. M. Belanich, B. Cummings, D. Grob, J. Klein, A. O'Connor, D. Yarosh, Pharmaceutical Technology 20(3), (1996) 142-150. S. Broverman, G. Prestwich, BioTechniques 19, (1995) 874-875. B. Champluvier, R.-M. Kula, Bioseparation 2, (1992) 343-351. W. Demmer, H. Hoed, A. Weiss, E. Wuenn, D. Nussbaumer, in: Proceedings of the 5th European Congress on Biotechnology, Vol 2 pp 766-769. C. Christiansen et al, (eds.) Munksgaard, Copenhagen, July (1990). K. Gebauer, J. Thoemmes, R.-M. Kula, Biotech. and Bioengin. 53(3), (1997) 181-189. A. Karger, B. Bettin, H. Granzow, T. Mettenleiter, Journal of Virological Methods 70, (1998) 219-224. D. Luetkemeyer, M. Bretschneider, H. Buentemeyer, J. Lehmann, J. Chromatography 639, (1993) 57-66. O. Reif, R. Freitag, J. Chromatography A 654, (1993) 29-41. O. Reif, R. Freitag, Bioseparations 4, (1994) 369-81. X. Santarelli, F. Domergue, G. Clofent-Sanchez, M. Dabadie, R. Grissely, C. Cassagne, J. Chromatography B 706, (1998) 13-22. T. Sellati, M. Burns, M. Ficazzola, M. Furie, Infection and Immunity 63(11), (1995) 4439-4447. Sartorius AG U.S. Patent 5,215,692 (1993). Sartorius AG U.S. Patent 5,556,708 (1996). Sartorius AG U.S. Patent 5,739,316 (1998). Sartorius AG U.S. Patent 5,547,575 (1996). Sartorius AG U.S. Patent 5,618,418 (1997). W. Wang, S. Lei, H. Monbouquette, W. McGregor, BioPharm. 8(5) (1995) 52-59. Sartorius AG International Application WO 98/41300 (1998). W. Demmer, D. Nussbaumer, presented at BIOEUROPE '98, Bioseparation and Bioprocessing of Biomolecules, Sept. 1998 Cambridge UK. Sartorius AG International Application WO 98/41301 (1998). W. Demmer, D. Nussbaumer, J. Chromatography A, 852 No. (1999) 73-81.

294


[24]

J. K. Walter, in: Bioseparation and Bioprocessing; G. Subramanian, (Ed.), Processing, Quality and Characterization, Economics, Safety and Hygiene, Wiley VCH, 1998, vol. II, pp. 447-460.

295


Functionalized Membranes for Separations and Reactions

297


New Insights into Membrane Science and Technology: Polymeric and Biofunctional Membranes D. Bhattacharyya and D.A. Butterfield (Editors) © 2003 Published by Elsevier Science B.V.

Chapter 15

Polymer grafted membranes S.M.C. Ritchie*

Department of Chemical Engineering, University of Alabama, Tuscaloosa, AL, 35487-0203 * Corresponding author, Phone: 205-348-2712, Fax: 205-348-7558,e-mail: [email protected] 1.

INTRODUCTION

The modification of porous microfiltration and ultrafiltration membranes for control of intrinsic properties is a field of considerable interest. Although these modifications have in general been accomplished through alteration of membrane surface chemistries, considerable interest has also been shown toward the incorporation of polymeric groups and structures into the membrane pores. In the case of dense materials, the porous structure of the membrane behaves as a stabilizing support for grafted polymers that may be too fragile in a homogeneous setting creating a more stable composite material. For example, although hydrogels are very useful for drug release applications [ 1], their utility in industrial applications is limited due to their inability to resist mechanical and osmotic forces [2]. By immobilization of the hydrogel in a microfiltration membrane, mechanical stability is added without significant additional resistance to transport processes in the base hydrogel. Incorporating polymeric groups to form dense and porous polymer grafted membranes can be accomplished either by polymerization in-situ or by grafting of large chain polymers. In either case, the addition of polymeric groups to a membrane can significantly alter properties of the membrane, including: permeability, molecular weight cut-off (MWCO), fouling resistance, separation capabilities, biocompatibility, conductivity, responsiveness to environment, and sorption capacity. The purpose of this chapter will be to describe the membranes used for this purpose, the techniques utilized for modification, the behavior of the modified membranes including some basic models for transport, and finally to give some example applications where this technology is currently used, as well as some future possible applications. 2.

MEMBRANES

The field of membranes is extremely broad, and the applicability of different types of membranes for polymer grafting is equally diverse. Polymeric 299

Polymer Grafted Membranes - Ritchie

membranes of nearly every variety have been utilized in the literature as substrates for polymer addition.. The membranes are generally microfiltration (0.02- 10 ~m) or ultrafiltration (1,000- 100,000 MWCO) varieties and may be configured as flat sheets or hollow fibers. Simple non-woven materials may also be employed. Many different types of inorganic membranes (tubular and flat sheet varieties) have also been utilized, including some inorganic/organic hybrids. In general, polymers may be functionalized in any matrix where the porous structure provides sufficient interaction between the membrane and the grafted polymer in the integrated material.

2.1

Polymeric Membranes Membrane materials utilized for polymer grafting may be of nearly any variety. There are examples in the literature for polymer grafting to fluoropolymers [2-11], polyolefins [12-27], cellulosics [28-32], polysulfones [33], polycarbonates [34], and polyacrylates [35,36]. A summary of these different materials, as well as chemical structures, is given in Table 1. Notice that these are all fairly stable chemically, since they only behave as substrates for the grafted polymer. Thus, only a select few contain reactive groups, including the cellulosics and PAN. In most cases they serve as inert backbone polymers for the more fragile grafted polymer, and thus reactivity is not desired. In addition, the polymers are relatively radiation stable so that material integrity is not overly compromised during pretreatment [37]. The physical properties of the materials add further credence to this philosophy. Since most applications involve interaction with aqueous solutions, the solubility parameters are well away from that of water (47.9 MPal/2). This helps to prevent swelling of the membrane, thus providing a stable constraining matrix for the polymer gel. The polymers are also generally temperature stable. This is particularly true in the case of PES and PC. Although most membrane operations are performed at or near room temperature, emerging applications may benefit from added temperature stability. A summary of these physical properties is given in Table 2. The porous structure of polymeric membranes is conducive to a role as a support structure for polymer grafts. Representative scanning electron micrographs for cellulose acetate and polyethersulfone membranes are shown in Figures 1 and 2. Notice that a network of interconnecting pores is formed during the process of phase-inversion. This network is irregular and there are no straight through pores. This ensures excellent interaction between the membrane substrate and the grafted polymer, such that there is effective transfer of mechanical stability. In addition, there is less chance for the formation of pinhole defects that can nullify separation in dense membranes. Careful control of phase-inversion conditions allows regulation of pore size, porosity, and mesh

300

P o l y m e r Grafted Membranes - Ritchie

size. This network can be utilized as a skeletal support for any subsequently formed hydrogel or polymer network. Table 1" Chemical Structures of Various Membrane Materials for Polymer Grafting Polymer Structure

Fluoropolymers: Poly(tetrafluoroethylene) Poly(vinylidene fluoride)

(PTFE) (PVDF)

-.Ec F2 --CF2 ~n ..~CH2__CF2 ~." n

Polyolefins: Poly(ethylene) Poly(propylene) Cellulosics: Cellulose triacetate

-ECH¢--CH2~ -ECH,--~;H~n CH3

(PE)

n

(PP)

(CTA)

/ Ac

C.Hz

-Eo o~. Ac OH CH=

Cellulose

~oo~..

(C)

OH

Polysulfones: Poly(ether sulfone)

0

(PES)

II

0

Polycarbonates: Polycarbonate

(PC)

Polyacrylates: Poly(acrylonitrile)

(PAN)

CH3

-Ec.,-~. 3-

c.

301

Table 2: Physical Properties of Membrane Materials Utilized for Polymer Grafting (data from [38,39]) Polymer PTFE PVDF PE PP PES CTA C PC PAN

T~~(K) ~5b (MPa1/2) 390 23.2 223-371 12.7 148-275 16-17 170-370 18-19 498 22.9 473 25-28 243-433 18-32 420 383 25.3

Tg = glass transition temperature solubility parameter

51.tin Figure 1: SEM of a cellulose acetate MF membrane.

Figure 2: SEM of a polyethersulfone MF (0.22 pm pore size) membrane filter (Courtesy of Millipore Corporation).

If a more open support is required, non-woven supports (lower half of Figure 3) may be utilized. Since the support has a much coarser structure, local interactions between the support and the grafted polymer would be fewer, and thus there may be less mass transfer resistance. The same would be true for track-etched polycarbonate membranes. Notice in Figure 4 that the pores pass in a linear fashion through the membrane. This would be particularly useful from a fundamental point-of-view, since modeling the flow behavior in these membranes would be easier than comparable (pore size) phase-inversion membranes.

302

P o l y m e r Grafted M e m b r a n e s - Ritchie

100 ~m

Figure 4: SEM of a track-etched (uniform pores of 0.05 - 12 ~tm), Polycarbonate, ISOPORETM membrane filter (Courtesy of Millipore Corporation).

Figure 3: Cross-sectional view of a thin film composite membrane (lower section is the non-woven support).

2.2

Inorganic Membranes The use of inorganic membranes has generally been relegated to systems where superior membrane strength is required. The physical strength of these membranes is unmatched, particularly for higher temperature applications. In addition, inorganic membranes (notably silica) possess excellent acid stability. Surface functionalization of inorganic and organic/inorganic composite membranes has been achieved with negligible loss of mechanical stability through the addition of functional silanes [31,40]. This is in contrast to cellulose acetate membranes (Figure 5), where surface functionalization is achieved via hydrolysis and oxidation. Notice the ragged structure of the membrane compared to the precursor. Mediation of membrane degradation by using milder hydrolysis conditions must be balanced with the potential loss of aldehyde sites used for polymer grafting.

Figure 5: Surface views of a cellulose acetate MF membrane before and after hydrolysis (used with permission, [30]). Inorganic membranes that have been utilized in the literature for polymer grafting include tubular membranes composed of silica [41] or

303

Polymer Grafted Membranes Ritchie -

zirconia [42,43]. Other alternatives include sintered glass [44], sol-gel membranes [45], and inorganic/organic composites [32]. In the latter case, membranes are extruded films of up to 80% silica in polyethylene. The polyethylene is used as a binder, and allows for excellent material flexibility while maintaining acid stability and ease of functionalization [31]. A representative scanning electron micrograph of this latter material is shown in Figure 6.

•

~

~3 ,) ,3 : 3 / :

~

;, ~

~

.5 .3..)

~

~

.

~. c ~ ' , ; :.

Figure 6: Surface view of a silica-polyethylenecomposite MF membrane. 3.

POLYMERIZATION TECHNIQUES

There are many techniques for polymer grafting. Some of these techniques, such as electrochemical and ,/-irradiation, require continuous application of a current or radiation throughout the polymerization process. Others, such as UV-initiation, electron beam, plasma, and silanization, create reactive sites for subsequent formation of polymers in-situ, in general by free radical polymerization. Whole polymers can also be incorporated, either by cross-linking of a polymer solution in the pores, or by specific attachment of polymer chains to the membrane pore surfaces. The basic principles of each technique will be described in detail, as well as the benefits and drawbacks to utilization.

3.1

Electrochemical Polymerization The technique of electrochemical polymerization is based on application of a current in a standard three-electrode cell where a membrane has been affixed to the working electrode [45]. When the current is applied, the polymer forms at the exposed working electrode surface and propagates until the void spaces in the membrane are filled. A schematic of this process is shown in Figure 7. Polymers formed in this manner are commonly polyaniline and polypyrrole

304


Reference Electrode

1

Counter Electrode

1

Aqueous AnilineElectrolyte Solution

Membrane

Working Electrode

Polyaniline

Figure 7" Schematic of electrochemical polymerization process. (Figure 8). Incorporation of nitrogen on the chain permits quatemization and thus charging of the chain. Counterions may be monoanions or polyelectrolytes for higher or lower electrochemical activity, respectively [46].

H

I

H (a)

(b)

Figure 8: Structures of (a) polyaniline, and (b) polypyrrole. The greatest advantage of electrochemical polymerization is that the resulting polymer, and by association the membrane, is conductive. Polymers made in this fashion are generally more conductive than comparable polymers made chemically [47]. In addition, the response of the conductive polymer is enhanced in the membrane phase [48]. The use of more structured membranes, such as tetraethyl orthosilicate (TEOS) sol-gel films, may be 305

Polymer Gratted Membranes - Ritchie

useful for MEMS applications, for example the production of an array of nanoelectrodes [45]. It should be noted that electropolymerization produces a fairly dense polymer, and thus even monoanion transport is restricted. Consequently, different techniques for polymer grafting are required to make more open structures. 3.2

y-irradiation Another technique for polymer grafting is 7-irradiation. This form of radiation is generally more powerful than UV-radiation, and thus is more able to activate the entire material [18]. During polymerization, a capsule containing monomer and the membrane undergoes exposure to y-irradiation from a Co-60 source. Free radicals formed on the membrane act as initiation sites for polymerization. However, the monomer is also activated in solution, thereby promoting homopolymerization [17]. Since the monomer solution is both outside and inside the membrane (excess monomer solution), pore blockage and transport of additional monomer into the membrane matrix may be an issue. Cross-linking the structure during formation has been found to mediate this effect and create a more homogeneous graft across the membrane thickness [7]. Another disadvantage to this technique is that the source is more expensive and incurs more hazards than UV-sources. However, the extensive use of 7-irradiation for sterilization does present an opportunity for wider usage in polymer grafting. 3.3

UV-Initiation Inter-connected polymer networks (IPNs) are generally formed using radiation-grafting techniques. A crude schematic of an IPN is shown in Figure 9. Notice how the fibers of the support material are interwoven with the grafted polymer network. The network itself is composed of polymer chains that have been stabilized by the introduction of a cross-linking agent in the reaction mixture. Photo-initiators have been utilized in the literature for these reactions, and thus UV-light is required. This technique can be utilized on large areas of membrane, is suitable for continuous processes, and generally activates sites deeper into the membrane than surface activation techniques (e.g., plasma). Polymerization takes place primarily by a free radical mechanism with sites on the membrane material or by activation of photo-initiators. Since the membrane has been thoroughly wetted before initiation, polymerization takes place only in the membrane porous structure [2]. This technique is particularly useful for vinyl-containing monomers. Specific examples of polymers grafted in membranes by UV-initiation and other radiation techniques are shown in Figure 10. It should be noted that these membranes

306


have been primarily examined in the literature for pH- and ionic strengthresponse, and thus the presence of ionizable groups is imperative.

Crosslinked

Membrane

Network of Grafted Polymer

Support Fiber

Figure 9: Schematic of inter-connected polymer network in a porous support.

Polymer

Radiation Technique

Polystyrene

"Ec"'-~~n

Poly(4-vinylpyridine) Poly(acrylic acid)

~/ UV, y, e-

•

n

UV, y, e-

COOH

Poly(acrylamide)

-ECH=--~H ~n

UV

C~-O

I

NH=

Poly(glycidyl

-Ec.,-~. ~n

methacrylate)

c=o i 0

I

/CH

°:,cl., Figure 1O:Typicalpolymersradiation-graftedinto membranes. 307

e

.


3.4

Electron Beam

The wide availability of electron beam sources has lead to greater implementation in industry, most notably for curing applications, compared to T-irradiation. Electron beam processing for polymer grafting has generally been used for preirradiation of the membrane. This helps to avoid issues with homopolymerization observed with T-irradiation. Decay of radicals on the membrane surface, however, does necessitate the immediate immersion of the activated material into a monomer solution after pretreatment. Electron beam activation does provide a uniform distribution of active sites throughout the starting material. Although specialized equipment is required, the process is amenable to use in a continuous process. The added capital expense and operating cost for high accelerating potentials (-200 kV) does detract from this technique when compared to UV-initiation. 3.5

Plasma

Electron beam and other radiation sources can alter the intrinsic properties of the membrane, and thus alternatives, such as plasma activation, have been examined. Plasma activation is accomplished by formation of an ionized gas (plasma) at the surface of the membrane. A schematic depicting this operation is shown in Figure 11. The plasma will only activate exposed areas, so it is generally useful more for surface treatment (see Figure 12). However, researchers have utilized it successfully for track-etched membranes, where there is good accessibility to the porous areas of the membrane (recall Figure 4). It should be noted that plasma treatment can result in a wide variety of surface chemistries and chemistries for polymerization. For this reason, it is often difficult to characterize what will occur upon plasma treatment. However, its lower cost compared to Tirradiation and electron beam techniques may make its use in niche areas appropriate. "

Gas Inlet

mY

" ~

~

;,

,,

,~i

: ,|

i,:

: , :~,~i

i~.i

,,",, ,,~ i

MEMBRANE

,:::,

,

:;,

"= V a c u u m

~

RF Power Source

]

+ Figure 11" Plasma treatment system for activation of polymeric membranes. 308


% % = m =

%

,

..

Figure 12: Radical groups formed by plasma treatment located near the membrane surface, with much less affect on interior. 3.6

In-situ Crosslink

When in-situ polymerization is difficult, for example if the temperature/ pressure required cannot be accomplished in the membrane material, in-situ crosslinking of an existing polymer may be appropriate. In this technique, a solution of the polymer, as well as cross-linking agents, is allowed to fully wet the membrane. Once excess solution is removed, the solution is allowed to gel by the application of heat or UV [27,33]. One example where this is done extensively is with water-soluble polymers, such as polyethylene glycol (PEG) and polyvinyl alcohol (PVA). Either through the use of silanes (PEG) or glutaraldehyde (PVA), very specific molecular weight chains of these polymers may be cross-linked in-situ. The use of precisely controlled molecular weight distributions for the precursors allows for excellent control of gel mesh size [28,29]. This is crucial for control of drug delivery and other end-uses of the modified material. 3.7

Silanization and Polymerization

While silanes can be utilized for cross-linking of PEG, they may also be used to apply initiating groups to inorganic membranes. This is particularly effective for silica membranes, as hydrated surfaces readily react with chloroand alkoxy-silanes. When these silanes also contain functional groups, such as vinyl and amine groups, the membrane surface can be primed for subsequent polymerization (Figure 1 3 ) . Cohen and co-workers have effectively utilized this technique to modify a number of inorganic membranes for use in oil-water emulsion separations. Vinyl-functional silanes permit the subsequent polymerization of vinyl pyrrolidone, producing a layer of 309


cH3

OH OH OH

+

C H 3 0 - Si - (CH2) 3 - O C H 2 - CH - CH 2

i

\o /

CH30

S i - (CH2) 3 O C H 2- C H - CH 2

"l-

H2N- R

Si - (CH2) 3 - O C H 2 - CH - CH 2 - N H - R

I OH Figure 13: Preparation of silica composite membranes by silane derivatization followed by polyamino acid functionalization. covalently attached chains. These "f'mgers" extend into the membrane pores, inhibiting the passage of oil droplets [42,43]. Previous studies by researchers in Japan have demonstrated the viability of this technique for attachment of solvent-responsive acrylate polymers [50]. 3.8

Whole Polymer Grafting

When polymerization in-situ is not possible, whole polymers may be grafted. Many of the same techniques for surface modification may be used, but the polymer is then passed in convective flow through the membrane, with single point attachment via a single end group. This technique has been used extensively by Bhattacharyya and co-workers [30-32] for the attachment of amino acid homopolymers. This technique is a natural extension of enzyme immobilization techniques, and methodologies are similar. For example, cellulosics are utilized by surface modification (via hydrolysis and oxidation) to create aldehyde groups. A Schiff-base reaction may then be utilized for attachment of the polyamino acids via the terminal primary amine group. The formed double bond is reduced to form an acid-stable imine bond. In the case of inorganic membranes, introduction of a glycidyl-functional silane permits direct formation of the imine bond. In both cases, regular polymeric structures can be introduced into the structure while preserving the intramolecular

310


interactions that impart special functionality to the membrane. An example of this is the ability to form helical structures, such that exceptional selectivity of Pb(II) over Cd(II) is observed (nearly irreversible sorption of Pb(II)) [31]. In addition, more selective side groups may be utilized, for example thiol groups (cysteine) which are effective for Hg(II) sorption [32]. One caveat to this technique is that immobilization becomes more complicated as molecular weight increases since the sole end group (used for attachment) may get hidden in a globular conformation. 4.

RESPONSIVE MEMBRANES

In all of the above cases, the overall goal is to impart selectivity or to change the properties of an essentially inert and non-selective membrane. Ideally, this selectivity would be tunable to the needs of the user. For example, permeability controlled by temperature, pH, ionic strength, and other environmental conditions may be especially useful for biomedical applications. The goal of this section will be to detail membranes that have been developed in the literature for controlled responses in relation to environmental conditions.

4.1

pH

Control of membrane properties by precise pH control is the most easily visualized concept. The definite pKa for each functional group incorporated into the membrane structure allows for a very precise break in the permeability of a polymer grafted membrane. For example, the pKa for the carboxylic acid

/

[H ÷] [salt] [Me x÷]

/ Figure 14: Environmentaleffects on immobilizedpolyelectrolytes in a membranepore. groups of polyaspartic acid is around 3.9. This is on a logarithmic scale, and thus small changes in the pH result in massive changes in the fraction of

311


charged and uncharged groups. As the number of charged groups increases, repulsion along the polymer backbone increases such that the attached polymers extend into the membrane pores. This effectively decreases the pore size of the membrane, thereby decreasing the permeability of the membrane. This is shown schematically in Figure 14. This phenomenon has been observed experimentally by a number of researchers, and a specific example is given in Figure 15 [12]. The presence of a step change in the permeability is critical as this imparts a switching capability to the membrane. It should be noted that changing the charge of the fixed groups would invert the behavior displayed in Figure 15.

12 A

10

=,

ohm

I

~_er

ed

EE

6 i

2 .....

2

.b-

""

n

n

4

6

8

pH Figure 15: Permeability differences with pH for poly(4-vinyl pyridine) grafted membranes (used with permission, [12]).

4.2

Ionic Strength A similar response is observed with changes in ionic strength. In this case we are specifically limiting discussion to monovalent ions that possess essentially no affinity for the fixed charge. Whereas hydrogen ions possess high affinity for the charged groups below the pKa, changes in ionic strength affect the permeability by charge shielding. That is, once a significant number of charged groups in solution surround the fixed charge, it is less able to interact with neighboring fixed charges. Hence, the repulsive forces generated by neighboring charged species are mediated by "masking" of these charges at excessive solution ionic strength. The source of ionic strength may be a buffer or simply an excess of sodium or chloride ions. In either case, an effect 312


similar to protein salting out occurs with corresponding collapse of extended chains to the pore walls, as shown in Figure 14. A reversible switching ability would be observed in this situation simply by the permeation of pure water.

4.3

Heavy Metals

When heavy metals are present in the solution, a different possible scenario exists. This is due to specific interactions between the metals and the charged species, such that neutralization takes place by either ion exchange or chelation. In either case, repulsion among charged groups on the attached polymer chains is reduced, and permeability of the membrane will generally increase. Since heavy metals can bind quite strongly well beyond the pKa of the charged groups, the potential exists for creating an irreversible change in the membrane. This may particularly be the case when chelation occurs, since this mechanism is the result of multiple interactions between the sorbed metal and the polymer (very stable complex). Since the metals cannot be desorbed at the operating conditions, the membrane response to the environment would at least decrease and may become totally inactive. 4.4

Temperature

The use of temperature responsive membranes is of particular interest in the biomedical field. Since the polymers utilized generally have a shift at or near body temperature, they are ideally suited for implantable devices, for example in drug delivery modules. The mechanism of operation is governed

(a) A

~ LCST

V

F-

UCST,~ (b)

Critical Concentration Figure 16: Phase behavior of thermo-responsive graftedpolymer chains.

313

Polymer Grafted Membranes - Ritehie

by the lower critical solution temperature (LCST) of the attached polymer. Notice in Figure 16 that the normal upper critical solution temperature is present (higher solubility as temperature increases). However, there is also a temperature above which solvation of the polymer chain decreases again. At this temperature, there is a phase transition that results in shrinkage of the polymer. ~In the case of polyisopropylacrylamide, this transition occurs at around 32 °C and results in shrinkage of the polymer above this temperature. The transition is relatively sharp, and experimentally determined pure water permeabilities have changed an order of magnitude with as little as a degree change in temperature [51 ]. 5. MODELS FOR TRANSPORT IN POLYMER GRAFTED MEMBRANES

Although transport in polymer grafted membranes has been altered by changing feed conditions (i.e., pH, ionic strength, etc.), models for transport have been lax on the explicit inclusion of these effects. In some cases, this is justified by cross-linking of the polymer matrix to inhibit permeability changes due to changes in solution conditions. For example, Anderson and co-workers have examined protein transport through charged and uncharged polyacrylamide grafted membranes and observed negligible change in permeability [2,3]. In other cases, models are adequate for one state, but fail upon transition of the membrane properties [12,52]. In this section, the general equations used to model this transport will be presented, as well as some insights into inclusion of transition effects. 5.1

General Equations Three basic equations have been utilized to model transport in polymer filled membranes. Since the membranes are generally still porous, equations for flow in pipes and packed beds have been adapted. First, the membrane permeability, k, can be calculated from bulk properties using Darcy's Law, k= Q~/ .~ap

(1)

where Q is the bulk flow rate,/z is the fluid viscosity, l is the membrane thickness, A is the membrane cross-sectional area, and Ap is the pressure drop. Notice that since the bulk flow rate and the cross-sectional area are utilized, knowledge of membrane pore size is not required. In addition, all of the parameters may be determined quite easily. Unfortunately, there is little

314


predictive value, and no information regarding the membrane (besides thickness) is utilized. The permeability can be utilized in the second basic equation by combination with a material balance to obtain the molar flux of solute by a modified form of Fick's Law, N, =~Ap

(2)

where ks is the solute permeability (based on the membrane permeability and the permeate solute concentration), and Cf can Cp are the feed and permeate solute concentrations, respectively. Again, however, there is little explicit information about the membrane in the equation. The pore size can be elicited via a third basic equation, the Hagen-Poiseuille equation,

321ult

(3)

Ap= D2

where D is the membrane pore diameter and u is the actual fluid velocity (accounting for porosity and tortuosity). Equations (1-3) do an adequate job of predicting properties in static systems, but do not capture the transient nature of responsive membranes. That is, they cannot predict the changes in membrane permeability, solute flux, and effective membrane pore radius at different conditions. Rather, experimental data must be taken at each set of conditions. In addition, no gel characteristics are inherent in the equations. Some progress toward including these parameters has been achieved via empirical models [2,3] and models based on first principles [52].

5.2

Accounting for Gel Properties The permeability of the polymer filled membranes will change as the membrane pore diameter changes. Since the presence of polymer in the pores is the cause for the change in pore diameter, the membrane permeability should be a function of how much polymer is in the membrane. This quantity may be expressed as the gel volume fraction, ~, in the pore. Anderson and coworkers have found good agreement between measured and calculated membrane permeabilities based on a power law relation as in Eq. (4) [2]. k = 4.35 x 10-'8~-33' (cm 2)

(4)

Figure 17 shows how Eq. (4) fits permeability data for uncharged, negatively charged, and positively charged polymer filled membranes. However, this 315

P o l y m e r Grafted Membranes - Ritchie

100

A

•~

10

-

1

-

**X•

• Neutral Gel

O Qsstfenr;aArYe~mGi;S

Gel

E

E,r, L O IX

× I

0.1

I

'

'

'

I

II

I

I

I

'

'

'

'

'

0.1

0.01

Polymer Gel Fraction

Figure 17: Power law fit of permeability date for polyacrylamide gels (used with permission, [2]). comes with the caveat that the gels are sufficiently cross-linked such that there is no change in the structure over a range of pH and ionic strength. In addition, accurate prediction of the gel volume fraction is critical. In Figure 17, gel volume fractions were determined experimentally, and thus the calculated permeability accurately reflected what was observed by experiment. To truly make the model predictive, however, the gel volume fraction should be calculated as well. Childs and co-workers have attempted to calculate the gel volume fraction from f'trst principles for dynamic systems. That is, the immobilized gel is allowed to change conformation as dictated by electrostatic repulsions, hydrogen bonding, and van der Waal's forces. Mika and Childs have proposed calculation of a correlation length based on polymer persistence length (Lp), contour distance (distance between adjacent charged groups, A), and the concentration of charge-beating polymer segments (c) [52]. The resulting correlation length, ~, is then assumed to be the diameter of a hard sphere (2R) about which exists a solvent film. A packed bed composed of

316

Polymer Grained Membranes Ritchie -

these solvated spheres is then assumed to represent the polymer filled pores of the membrane as shown in Figure 18. Since the solvated spheres take up the

Re~ I~~) % ~S

~R

Figure 18: Happel model for packed bed (one pore) of solvated polymer chains as representative gel structure where R is the polymer hard sphere radius and Re is the radius of the spherical solvent envelope. entire volume of the bed, the polymer gel fraction can be related to the correlation length by the following equation,

I 1 / -~/4 = Lp + 167r~bAC) (4m4c)'/8(Ac) -3/4

(5)

where lb is the Bjerrum length. The permeability is then calculated using the equation developed by Happel [56],

k = 9¢

3 + 2¢ 5/3

(6)

'

Mika and Childs found excellent agreement between calculated and measured permeabilities for charged gels. However, the calculated permeability varied by 4 orders of magnitude for the neutralized gel. This indicates the need for inclusion of solution properties such as pH and ionic strength into the model such that neutralization from interactions with hydrogen ions or charge shielding by excessive ionic strength and the resulting compression of the solvent films can be adequately accounted for in the model. This last area has great potential additional research, particularly for inclusion of temperature effects.

317

PolymerG r a f t e d

6.

M e m b r a n e s - Ritchie

APPLICATIONS

Polymer filled membranes can be either static or dynamic with regards to separations. Static polymer filled membranes are generally dense or sufficiently cross-linked such that they are non-responsive to environmental conditions. They are used in applications where long-term stability is required for continuous processing. Dynamic membranes are designed for systems where periodic changes in the environmental conditions trigger a response in the membrane. In this section, applications for static and dynamic membranes will be examined.

6.1

Static Membrane Applications Although there are a wide variety of separations for which polymer filled membranes are applicable, five primary areas will be addressed in this review. These areas include bioseparations and artificial organs, metal sorption and water treatment, fuel cells, nanofiltration, and gas separations/pervaporation. These various applications will be examined starting from porous membranes to more dense membranes. 6.1.1 Bioseparations and Artificial Organs There are two sub-categories in this area of applications: protein separations and immunoisolation. First, protein separations may be performed with polymer filled membranes in a manner very similar to column separation, for example by electrostatic interactions. Depending on the chemistry of the polymer gel, positively or negatively charged proteins may be separated from each other, with the interacting species being retained on the membrane. For example, bovine serum albumin (BSA) has been separating using polyethylene membranes filled with an acrylic polymer that has been modified with amine groups [23]. Interaction of the negatively charged BSA with the membrane allows for its separation. In a similar fashion, by introduction of a sulfonic acid containing polymer, a positively charged species, such as lysozyme, has been successfully separated [22]. It should be noted that to add selectivity to the membrane, more specific interactions would be required, for example the use of ELISA-type groups. However, in comparison to packed bed separations, the polymer filled membranes do offer the advantage of removing larger proteins in convective flow, which eliminates the diffusive mass transfer resistances encountered with ion exchange resins and nanoporous column packings. More intriguing is the use of these membranes for immunoisolation. In that case, the membrane is used to regulate the passage of nutrients, therapeutic agents, and immunological molecules. The object here is the creation of an implantable device, containing living cells, that permits 318

Polymer Gratted Membranes - Ritchie

production and delivery of a therapeutic agent. Since live cells are implanted, there is a requirement for nutrient transport to the cells and for release of the produced therapeutic agent to the affected region. In addition, immunological agents that would destroy the foreign cells must be kept away. A schematic of this process is shown in Figure 19.

•

Immunological Molecules

Therapeutic Molecules Nutrients

Live

Figure 19: Schematicof implanted device containinglive cells (e.g., islets of Langerhans for insulin production). Barbari and co-workers have developed membranes with a cross-linker gradient across the filled polymer to regulate the molecular weight cut-off of the membranes used in this process [28,29]. Glutaraldehyde was used to cross-link a polyvinyl alcohol gel in cellulose ester membranes. In their work, the representative nutrient molecule passed relatively uninhibited. Therapeutic molecules, represented by a 50 kD protein passed through the membrane, though much more slowly than the nutrient. Very large molecules (> 150 kD) were rejected by the membrane to a greater degree, but only with a selectivity of about 2.5 (compared to 50 kD). Similar results have been observed by Baker and co-workers [33]. In that case, polyethersulfone membranes containing crosslinked polyvinyl alcohol were developed for the immunoisolation of pancreatic cells. In that case, in vivo studies of the membranes showed negligible change in membrane performance after 6 months, thereby showing significant promise for their use in implantable devices containing living cells. 6.1.2 Metal Sorption and Water Treatment Another large application for polymer filled membranes is heavy metal sorption for the treatment of wastewater. Work in this area was begun in Japan by radiation grafting of glycidyl methacrylate containing polymers in 319


polyolefin membranes. Post-modification of these membranes to get iminodiacetic acid (IDA) groups permits the utilization of chelation mechanisms for sorption of metal ions [21]. The presence of IDA groups is particularly useful since they allow for much greater selectivity over ions such as calcium and magnesium. Additional work where membranes are utilized to mimic the behavior of ion exchange and chelation resins has been performed by Choi and Nho [18]. They polymerized styrene in a membrane, followed by material sulfonation. In that case, the membrane is useful over a wider pH range since the membrane mimics a strong acid ion exchange resin. Further enhancements have including copolymerization to include carboxylic acid and pyridine groups for the production of mixed ion exchange and chelation membranes [19,20]. It should be noted that in all of these cases, the polymer is formed in-situ, thereby limiting the use of more complex polymers that possess improved selectivity or stability. In work done by Bhattacharyya and co-workers [30-32], homopolymers of various amino acids have been immobilized in a variety of organic (cellulosic) and inorganic-organic hybrid (silica-polyethylene) microfiltration membranes. Work with polyglutamic and polyaspartic acids has revealed an additional mechanism of metal sorption besides ion exchange and chelation. This mechanism is counterion condensation, which allows non-specific sorption of metal ions by attraction to a negatively charged electrostatic field caused by close residence of charged groups on the polymer [31 ]. Since this is a regular structure, the spacing between charged groups is regular, and superposition of the electrostatic fields in the membrane pores leads to a large overall field effect. The result is exceptionally high sorption of heavy metals with greater than 1 g heavy metal sorbed per gram of dry membrane sorbent [30]. The use of polyamino acids also permits the use of more exotic functional groups, such as thiol groups that have been shown to be particularly effective for the removal of soft metals, such as mercury [32]. 6.1.3 Fuel Cells The use of membranes containing sulfonic acid groups has also found utility for fuel cell membranes. Currently, the benchmark for new fuel cell membranes is Nation® [59]. One of the critical parameters for their use in fuel cell applications is the presence of water in the membrane. Water is critical to maintain conductivity across the membrane. For Nation® there must be at least 6-7 water molecules for each sulfonic acid group in the membrane [4]. Operating conditions usually involve 24 water molecules for each sulfonic acid group. The use of polystyrene that has been subsequently sulfonated is an excellent technique for raising the number of sulfonic acid groups in the membrane, as well as increasing its hydrophilicity and conductivity [4]. Membranes developed by the incorporation of sulfonated

320

Polymer Grained Membranes - Ritchie

polystyrene have led to the inclusion of around 60 water molecules for each sulfonic acid group [57]. The resulting conductivity is about double that of Nation®. Considering the expense of Nation®, this represents an attractive alternative. One caveat that must be considered is that for application in direct methanol fuel cells, the passage of methanol must not be too high, and the porous structure of these modified membranes need to be considered [58].

6.1.4 Nanofiltration Another use for membranes containing charged polymers is nanofiltration. This is a particularly intriguing application, since a microfiltration membrane ( 0 . 0 2 - 10 ~tm pore size) is used to reject ions. Childs and co-workers [27] have incorporated poly(4-vinyl pyridine) (PVP) into microfiltration membranes and achieved 70% rejection of a NaC1 solution at only 300 kPa (< 3 bar). When membranes containing an acrylic polymer gel are utilized, greater than 90% rejection of Na2SO4 was observed. Higher rejection of divalent cationic species should also be observed with the PVP membranes due to their positive charge, making them attractive for water softening applications. It should be noted that since these membranes are fairly open structures, excessively large fluxes will lead to loss of solute rejection. In addition, increases in ionic strength, as with regular nanofiltration membranes, will lead to charge shielding and a subsequent drop in solute rejection. 6.1.5 Gas Separation and Pervaporation In this last application of static membranes, alteration of solute solubility is obtained by changing the membrane chemistry. These membranes are also much more dense than previous examples as pinhole defects ruin membrane selectivity. One example in the area of gas separations is the recovery of CO2 from flue gas streams. Mechanical and thermal stability is provided by the inert membrane, in this case a polyacrylonitrile membrane [36]. A polymer that has excellent solubility for CO2, poly(ethylene glycol) acrylate (PEGA), was then incorporated as a gel in the membrane. Since it is technically challenging to make consistent, thin, stable films of PEGA, this is an ideal combination. When these materials were utilized for CO2/N2 separation at 30 °C, a selectivity of 32.4 was observed [36]. Pervaporation membranes are also dense and transport is governed by the solubility and diffusivity of solutes in the membrane material. For example, Ihm and Ihm [10] have incorporated sulfonated polystyrene in polyvinylidene (PVDF) membranes for pervaporation of ethanol/alcohol mixtures. When utilized as part of a membrane reactor, water formed during esterification reactions at sulfonic acid membrane surface sites (not in the 321

Polymer Grafted Membranes - Ritehie

porous structure) may be subsequently removed from the reacting phase by transport through the membrane. This is illustrated schematically in Figure 20.

Ethyl Acetate

Water

Ben:ene -I),

Cyclohexan

+

Benzene

Ethanol

(a)

(b)

Figure 20: Polymer grafted membranes for (a) water extraction in a membrane reactor, and (b) organic/organic separation by pervaporation.

The selectivity of the membrane was not especially high (only around 21), but used in this manner it can help to limit unwanted side reactions and can maximize reaction rates. A more important application for pervaporation may be in organic/organic separations. Classically, membrane separations have been limited in this capacity, as the formation of mainly polymeric membranes is performed in organic solvents. Swelling on the membranes is thus a major concern, in regards to the loss of selectivity, and in the lifetime of the membranes. The use of polymer filled membranes shows great promise in this area, as an inert support can add the required mechanical strength to resist or control swelling of the separating polymer. Kai et al. [25] made hollow fiber and fiat sheet polyethylene membranes containing polymethacrylate. Flat sheet membranes were found to have a lower flux but higher separation factor, while hollow fibers provided a higher flux with less selectivity. This lead to the conclusion that fiat sheet membranes were better equipped to add mechanical strength and thus control the swelling of the filled polymer. Earlier work by the same group for benzene/cyclohexane separations demonstrated a selectivity of 7-8 due to the enhanced solubility of the aromatic in polymethacrylate [26]. This is significant since distillation is especially costly for the separation of this mixture due to low relative volatility (- 1.02) and the formation of an azeotrope.

322


6.2

Dynamic Membrane Applications The potential of adding dynamic properties is an exciting development of polymer grafted membranes. Switching capabilities provides great promise for the creation of biomimetic structures, "lab-on-a-chip", and other structures for emerging applications. Not surprisingly, biomedical applications are near the forefront of this technology, with some applications in chemical valves and environmentally stimulated drug delivery. For each of these applications, however, very simple responses, such as pore expansion with shear [41], to more complex operations, where membrane permeability is altered by differences in solution pH, ionic strength, and temperature are utilized.

6.2.1 Oil/Water Separations Cohen and co-workers have examined the use of polymer filled membranes for oil/water separations. In their work, poly(vinylpyrrolidone) has been grafted onto ceramic membranes. Originally this work was performed to decrease the pore size and permeability of ceramic membranes [40]. Due to their method of formation (sintering of f'me ceramic powders), the formation of very narrow pore size distribution ceramic membranes is extremely challenging. This work presented a method of post-treatment that could sufficiently decrease the pore size to enable effect oil/water separation. Since the filled polymer was charged, they found that pH, ionic strength, and solvation of the chains could all be utilized to decrease the membrane permeability and effective pore size. Although they have not been utilized specifically as dynamic membranes, this work does indicate the possibility, in this case for ceramic membranes. 6.2.2 Chemical Valves Dynamic membranes have been used more specifically in the literature as chemical valves. A chemical valve, as shown in Figure 14, has been defined as the reversible expansion and contraction of membrane pores in response to environmental conditions of pH, ionic strength, and divalent/trivalent metals. In all cases, when charged groups interact with ions in solution, the end effect is the same: shrinkage of the immobilized chains and an increase in membrane permeability up to three orders of magnitude [53]. Changes in the membrane permeability are then a function of metal concentration, ionic strength, and pH. If the ionic strength of the solution is held constant, then the permeability has been found to change linearly with pH [53]. Thus, a very simple model,

k=a-b(pH)

(7) 323


is sufficient to predict the membrane permeability. Unfortunately, this relation does not hold if the ionic strength is not held constant, and the behavior will be more reminiscent of Figure 15. That is, there will be a step jump in the permeability above the pKa [34]. Comparisons to biological membranes are apparent, and the biomimetic characteristics of these membranes makes them excellent candidates for uses in the biomedical industry, in nanosized reactors, for "lab-on-a-chip", and other novel structures.

6.2.3 Thermo-responsive Membranes and Drug Delivery Hydrogels have been investigated extensively for drug delivery applications, and their extension to inclusion in membranes should thus not be unexpected. In this way, some negative aspects regarding hydrogels as drug delivery media, most notably issues with stability, are addressed. Hydrogels consisting of N-isopropylacrylamide are well known to have a volume-phase transition at around 32 °C. Early work with this gel in a porous glass membrane found a dramatic shift in the molecular weight cut-off (MWCO) of the membrane of about an order of magnitude [54]. Once a change in the permeability was established, research began examining the transport of drugs. Conventional thermo-responsive drug delivery systems only provide slight or gradual changes in drug transport rates as temperature changes due to increases in diffusivity and solubility. Polymer filled membranes were used in an attempt to make rate changes more dramatic, thus obtaining a switching mechanism. In this case, the transport of salbutamol sulfate, a pulmonary drug, was examined across hydrophilic and hydrophobic membranes containing a thermo-responsive polymer [55]. In their study, there was some change in the diffusion rate across the membrane at different temperatures. However, there was only a distinct difference for cellulose nitrate membranes. Similar work for BSA transport performed by Li and D'Emanuele [44] gave similar results. Some interesting ideas of better controlling the switching point have been put forward by Iwata et al [51]. They have indicated that by copolymerization of a thermo-responsive polymer with a second monomer that the inflection point may be increased or decreased. 7.

CONCLUDING REMARKS

Polymer grafting offers an excellent opportunity for controlling and tuning the properties of membranes. For example, properties such as permeability, separation capability, conductivity, and biocompatibility can be altered vastly, simply by the attachment of different polymer chains in the membrane porous structure. An additional feature is that in some cases, these changes can be dynamic, with stimuli-response to various environmental 324


conditions, such as pH, ionic strength, metal concentration, and temperature. Membranes are available in a wide range of compositions and configurations, and thus there are many different techniques for grafting the polymers. In most cases this is by in-situ polymerization, but immobilization of large chain polymers to utilized more exotic moieties has also been accomplished. Many researchers have attempted to model the behavior of polymer grafted membranes, and though they have been successful for static membranes, more work is needed to truly account for stimuli explicitly for dynamic membranes. This is particularly true in the case of thermo-responsive membranes. In any case, the many different areas where these membranes can and are being applied, from filtration to biomedical devices, should ensure that this emerging area in polymer and membrane science will be of interest for years to come. REFERENCES

[1] [2] [3] [4]

[5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15]

[16] [17]

[18] [19] [20] [21] [22] [23]

N.A. Peppas, P. Bures, W. Leobandung, H. Ichikawa, Eur. J. Pharma. Biopharma., 50 (2000) 27. V. Kapur, J.C. Charkoudian, S.B. Kessler, J.L. Anderson, Ind. Eng. Chem. Res. 35 (1996) 3179. V. Kapur, J. Charkoudian, J.L. Anderson, J. Membr. Sci, 131 (1997) 143. D.I. Ostrovskii, L.M. Torell, M. Paronen, S. Hietala, F. Sundholm, Solid State Ionics, 97 (1997) 315. J. Hautojarvi, K. Kontturi, J.H. Nasman, B.L. Svarfvar, P. Viinikka, M. Vuoristo, Ind. Eng. Chem. Res., 35 (1996) 450. R. Teleranta, J.A. Manzanares, K. Kontturi, J. Electroanal. Chem., 464 (1999) 222. B. Gupta, F.N. Buchi, G.G. Scherer, A. Chapiro, J. Membr. Sci., 118 (1996) 231. K. Kontturi, S. Mafe, J.A. Manzanares, B.L. Svarfvar, P. Viinikka, Macromolecules, 29 (1996) 5740. G. Ellinghorst, A. Niemoller, D. Vierkotten, Radiat. Phys. Chem., 22 (1983) 635. C.-D. Ihm, S.-K. Ihm, J. Membr. Sci., 98 (1995) 89. H. Matsuyama, M. Teramoto, H. Sakakura, K. Iwai, J. Membr. Sci., 117 (1996) 251. A.M. Mika, R.F. Childs, J.M. Dickson, J. Membr. Sci., in press 2002. M. Ulbricht, React. Func. Polym., 31 (1996) 165. A.M. Mika, R.F. Childs, J.M. Dickson, J. Membr. Sci., 153 (1999) 45. A.M. Mika, R.F. Childs, J.M. Dickson, B.E. McCarry, D.R. Gagnon, J. Membr. Sci., 108 (1995) 37. N.M. E1-Sawy, A.Z.A. Elassar, Eur. Polym. J., 34 (1998) 1073. Y.C. Nho, J.L. Garnett, P.A. Dworjanyn, J. Polym. Sci.: Part A: Polym. Chem., 31 (1993) 1621. S.-H. Choi, Y.C. Nho, J. App. Polym. Sci., 71 (1999) 2227. E.-S.A. Hegazy, H. Kamal, N. Maziad, A.M. Dessouki, Nuc. Instr.Meth. Phys. Res. B, 151 (1999) 386. E.-S.A. Hegazy, H.A. Abd E1-Rehim, H.A. Shawky, Rad. Phys. Chem., 57 (2000) 85. S. Konishi, K. Saito, S. Fumsaki, T. Sugo, Ind. Eng. Chem. Res., 31 (1992) 2722. M. Kim, M. Sasaki, K. Saito, K. Sugita, T. Sugo, Biotechnol. Prog., 14 (1998) 661. K. Sunaga, M. Kim, K. Saito, K. Sugita, T. Sugo, Chem. Mater., 11 (1999) 1986.

325

Polymer Grained Membranes - Ritchie

[24] [25] [26] [27] [28] [29]

[30] [31] [32] [331

[34] [35] [36] [37]

[38] [39] [401 [411 [42] [43] [44] [45]

[46] [47]

[48] [49]

[50] [51] [52]

[53] [54]

[55] [56] [57]

S. Tsuneda, T. Endo, K. Saito, K. Sugita, K. Horie, T. Yamashita, T. Sugo, Macromolecules, 31 (1998) 366. T. Kai, T. Tsuru, S. Nakao, S. Kimura, J. Membr. Sci., 170 (2000) 61. T. Yamaguchi, S. Nakao, S. Kimura, Macromolecules, 24 (1991) 5522. R.F. Childs, A.M. Mika, A.K. Pandey, C. McCrory, S. Mouton, J.M. Dickson, Sep. Purif. Tech., 22-23 (2001) 507. W.S. Dai, T.A. Barbari, J. Membr. Sci., 171 (2000) 79. W.S. Dai, T.A. Barbari, Biomaterials, 21 (2000) 1363. D. Bhattacharyya, J.A. Hestekin, P. Brushaber, L. Cullen, L.G. Bachas, S.K. Sikdar, J. Membr. Sci., 141 (1998) 121. S.M.C. Ritchie, L.G. Bachas, T. Olin, S.K. Sikdar, D. Bhattacharyya, Langmuir, 15 (1999) 6346. S.M.C. Ritchie, K.E. Kissick, L.G. Bachas, S.K. Sikdar, C. Parikh, D. Bhattacharyya, Env. Sci. Tech., 35 (2001) 3252. A.R. Baker, R.L. Foumier, J.G. Sarver, J.L. Long, P.J. Goldblatt, J.M. Homer, S.H. Selman, Cell Transplant., 6 (1997) 585. Y. Ito, M. Inaba, D.-J. Chung, Y. Imanishi, Macromolecules, 25 (1992) 7313. M. Ulbricht, K. Richau, H. Kamusewitz, Coll. Surf. A: Phys. Eng. Asp., 138 (1998) 353. J.H. Kim, S.Y. Ha, S.Y. Nam, J.W. Rhim, K.H. Baek, Y.M. Lee, J. Membr. Sci., 186 (2001) 97. A. Dawood, K. Miura, Polym. Degrad. Stab., 73 (2001) 347. Encyclopedia of Polymer Science and Engineering, 2n°Edition, Ed.: H.F. Mark, N.M. Bikales, C.G. Overberger, G. Menges, Wiley: New York, 1985. Polymer Handbook, 4th Edition, Ed.: J. Brandup, E.H. Immergut, E.A. Grulke, Wiley: New York, 1999. R.P. Castro, Y. Cohen, H.G. Monbouquette, J. Membr. Sci., 84 (1993) 151. R.P. Castro, H.G. Monbouquette, Y. Cohen, J. Membr. Sci., 179 (2000) 207. R.S. Faibish, Y. Cohen, Coll. Surf. A: Phys. Eng. Asp., 191 (2001) 27. R.S. Faibish, Y. Cohen, J. Membr. Sci., 185 (2001) 129. S.K. Li, A. D'Emanuele, J. Contr. Rel., 75 (2001) 55. M.M. Verghese, K. Ramanathan, S.M. Ashraf, M.N. Kamalasanan, B.D. Malhotra, Chem. Mater., 8 (1996) 822. W.S. Shim, Y.H. Lee, I.-H. Yeo, J.Y. Lee, D.S. Lee, Synth. Met., 104 (1999) 119. A.F. Diaz, K.K. Kanazawa, G.P. Gardini, J. Chem. Soc. Chem. Commun., (1979) p. 635. S. Das Neves, M.-A. De Paoli, Synth. Met., 96 (1998) 49. S. Machi, Radiat. Phys. Chem., 14 (1979) 155. K. Otake, T. Tsuji, M. Konno, S. Saito, J. Chem. Eng. Japan, 21 (1988) 443. H. Iwata, M. Oodate, Y. Uyama, H. Amemiya, Y. Ikada,Y., J. Membr. Sei., 55 (1991) 119. A.M. Mika, R.F. Childs, Ind. Eng. Chem. Res., 40 (2001) 1694. Y. Osada, K. Honda, M. Ohta, J. Membr. Sci., 27 (1986) 327. T. Tsuji, M. Konno, S. Sato, J. Chem. Eng. Japan, 23 (1990) 447. S.-Y. Lin, Y.-Y. Lin, K.-S. Chen, Pharma. Res., 13 (1996) 914. J. Happel, H. Brenner, Low Reynolds Number Hydrodynamics with Special Applications to Particulate Media, Prentice-Hall:Englewood Cliffs, NJ, 1965. S. Holmberg, T. Lehtinen, J. Nasman, D. Ostrovskii, M. Paronen, R. Serimaa, F. Sundholm, G. Sundholm, L. Torell, M. Torkkeli, J. Mater. Chem., 6 (1996) 1309.

326


[58] [59]

K. Scott, W.M. Taama, P. Argyropoulos, J. Membr. Sci., 171 (2000) 119. H.-G. Haubold, Th. Vad, H. Jungbluth, P. Hiller, Electrochim. Acta, 46 (2001) 1559.

327


New Insights into Membrane Science and Technology: Polymeric and Biofunctional Membranes D. Bhattacharyya and D.A. Butterfield (Editors) O 2003 Elsevier Science B.V. All rights reserved.

Chapter 16

Functionalized membranes for tunable separations and toxic metal capture A. M. Hollman and D. Bhattacharyya

Department of Chemical & Materials Engineering, University of Kentucky, Lexington, Kentucky 40506-0046 * Corresponding Author, e-mail: [email protected], Phone: 859-257-2794, Fax: 859-323-1929

1.

INTRODUCTION

The design of sustainable processes for separations, chemical reaction and heavy metal sorption will require the development of systems with reduced energy consumption and minimal environmental impact. Membrane processes have gained wide acceptance due to their unique properties such as their compactness, ease of fabrication, low-cost operation and modular design. For these reasons, membrane processes are finding numerous applications in fields ranging from water treatment to reactor design to advanced bioseparations. Conventional separations involving pressure-driven membrane processes are classified as: reverse osmosis or RO (for very high degree of salt separation including NaC1, and soluble organics); nanofiltration or NF (for separation of divalent salts and organics while permeating monovalent salts such as NaC1); ultrafiltration or UF (for separation of organics above 5000 MW); and microfiltration or MF (for separation of colloidal to micron particles and microorganisms) [1-2]. For UF and MF type membranes the mechanism of separation is based on size exclusion or steric hindrance factors. For more dense NF membranes this mechanism of exclusion is coupled with the electrostatic repulsive forces resulting from the presence of ionizable functionalities incorporated within the membrane matrix. However through appropriate modification (i.e. with charged macro-molecules), the applicability of porous microfiltration membranes can be extended to molecular level separations at applied pressures well below those required by traditional RO and NF. The development of membrane materials containing immobilized biomolecules (i.e. poly(amino acids), proteins, enzymes etc.) provide added

329

Functionalized Membranes For Tunable Separations and Toxic Metal Capture - Bhattaeharyya

opportunities for high performance selective separations and recognition-based chemical reactions. Traditionally, microfiltration (MF) membrane separations have been limited to the filtration of suspended solids, bacteria, viruses, etc. However, the separation capabilities of MF-type membranes (i.e. cellulose, silica, polysulfone) can be enhanced through functionalization with a variety of reagents. Depending on the chemical nature of the attached molecules (such as its chain length, acidic or basic properties, catalytic reactive site, etc.), these types of low resistance membranes could be used in processes like dissolved metal ion (or oxyanions) separation, chlorinated organic detoxification or biocatalysis. The primary focus of this chapter will be the use of microporous support membranes containing immobilized multi-functionalities for applications involving tunable charged-based separations or the development of high capacity sorbent materials. In particular, the fabrication of these novel membrane materials has been achieved by attachment of various polypeptides (MW 2,500-100,000) directly within the pore structure. Both the configurational and electrostatic properties as well as the binding affinity of these polyligands have been exploited to yield highly specific separations involving aqueous systems. 20

TUNABLE SEPARATIONS

Growing interest, in recent years, regarding the chemical modification of synthetic membranes has resulted in the development of materials that are sensitive to changes in their surrounding micro-environment. The properties of these new generation membranes, such as reversible water permeability [3-5], ion-selectivity [6-8] and surface hydrophobicity [9], have been shown to respond to slight variations in temperature [10], ionic strength [11], electric current [12], or pH. These stimuli-responsive membranes could have extended applications in catalysis, recognition-based separations, biomimetic device development, and controlled-drug release. 2.1

Stimuli-Responsive Macromolecules The most common methodology used in preparing an environmentally sensitive material is to alter the properties of a base support through the immobilization of an ionizable polymer. Monomeric repeat structures of some synthetic and biological polymers used in the development of tunable membranes are shown in Figure 1. Polyelectrolytes, comprised of these repeat units, are particularly well suited for modulating chemical signals because their secondary structure is dependent upon the electrostatic interactions between side group constituents. In solution, these intramolecular forces are regulated by the

330

FunctionalizedMembranesFor TunableSeparationsand ToxicMetalCapture- Bhattaeharyya

o II

O I! O II H2c - - CH - C - OH

acrylic acid

H2C-C-C-OH I CH 3

methacrylic acid

O II H 2 C : CH - C - NH 2

H2C - CH -- C -- NH CHCH 3 1 CH 3

acrylamide

N-isopropylacrylamide

NH H NH20 II ~ ~ II H2N - - C - NHCH2CH2CH2 C .... C - O H

CH - CH 2

4-vinylpyridine

O II HO-C-CH

L-arginine

vinylbenzyl chloride

H2N t r i O °~ II 2 CH 2 C - - C - O H

L-glutamic acid

H2N . H O H2NCH2(CH2)2CH2C -- C -- OH

L-lysine

H2N H O II HSCH2C--C-OH

L-cysteine

Figure 1. Repeat unit structures of some polymeric materials used for the development of stimuli-responsive membranes.

properties of the contacting solution, such as pH, ionic strength and temperature. A typical morphological response to variations in pH for a macromolecule containing acidic side groups (i.e. poly(acrylic acid), poly(methacrylic acid), poly(L- glutamic acid), etc.) is given in Figure 2. At low pH, the carboxyl side groups are protonated alleviating the repulsive force between neighboring constituents. This causes the polymer chain to contract, thus increasing its entropy. Conversely at high pH, these carboxylic acid groups are fully dissociated. Under these conditions, the electrostatic energy overcomes the entropic contribution to the free energy resulting in an extension of the polymer chain [13]. Upon immobilization, this response can be utilized to tune the performance characteristics of membrane materials. 2.2

Immobilization Techniques Preparation of these 'intelligent' or 'smart' materials can be achieved through both covalent and non-covalent immobilization techniques. Typically, these chemical modifications involve either in-situ polymerization [14-16] or the grafting of pre-existing macromolecules [7, 17-18]. 2.2.1 Cross-Linked Hydrogels An example of the former case is the incorporation of cross-linked hydrogels

331

Functionalized Membranes

For Tunable Separations and Toxic Metal Capture - Bhattacharyya

100 % 80

.X

60 \ 40

20

%

% ~

Low

pH

~

mm - - -

mm am=

High

Figure 2. Typical morphological response to variations in the pH of the contacting solution for a negatively-charged macromolecule (i.e. poly(L-glutamic acid), polyacrylic acid, etc.). within porous media [5, 15]. Hydrogels are defined by their ability to undergo large volume changes when exposed to an external stimulus. The usefulness of gels as separation devices is dependent upon the ability to maintain their integrity against mechanical stresses. To circumvent problems associated with stability, hydrogels can be enmeshed within the pores of a rigid polymer matrix by cross-linking [15]. In-situ polymerization is carried out by wetting the support material in an aqueous solution containing monomer, cross-linking agent and initiator. The membrane-supported gels are then formed through photo-initiation [19] or by heating [20]. The cross-linked polymer chains become entangled within the support matrix, thus stabilizing their morphology with respect to mechanical and osmotic forces [15]. This non-covalent technique results in membrane materials with a high effective concentration of charged groups. An entire chapter of this book has been devoted to the detailed description and applications of gel-filled nanofiltration membranes. 2.2.2

Covalent Attachment Biomolecules

of

Linear

Macromolecules

and

Although cross-linking provides mechanical stability and uniform coverage of the pore cross-section, it also severely restricts chain mobility resulting in poor molecular diffusivity [21]. Sensing devices characterized by

332

Functionalized Membranes For Tunable Separations and Toxic Metal Capture - Bhattacharyya

much faster response times have been developed through surface grafting of linear polymer chains [8]. The immobilization of polymer chains by covalent coupling usually leads to very stable materials with extended life when compared with other coupling methods, namely, physical adsorption and ionic binding. Covalent attachment of linear polymers can be achieved through a variety of functionalization techniques. These methodologies include plasmainduced grafting [8, 22], radiation-induced grafting [13, 23] and chemical grafting [7, 17]. There are significant advantages and drawbacks associated with each technique. 2.2.2.1 Plasma-Induced Polymerization Plasma-induced polymerization is a well-established surface modification technique that can be utilized to form polymer chains within porous supports. The attached polymer is linear because it propagates from a radical formed on the polymer backbone of the support matrix. Radicals formed within the pores of the substrate can be utilized as an initiator for subsequent polymerization [24]. The disadvantage associated with this technique is that plasma species (electrons, ions, etc.) interact strongly with the polymer matrix and with the monomer vapor through various undesirable side reactions [23]. This leads to difficulties in characterization due to the formation of an array of different functionalities. An example of the possible uses of plasma-induced polymerization was the development of a novel, molecular recognition gating membrane using Nisopropylacrylamide (NIPAM, see Figure 1) containing a crown ether receptor [8]. This particular polymer is appropriate for sensing applications because it can recognize a chemical signal, in this case Ba 2÷ ions, and respond through a conformational change exhibiting highly responsive gating properties. Comprehensive reviews on the properties and applications of NIPAM as a stimuli-responsive material can be found in the literature [25-26].

2.2.2.2 Radiation-Induced Polymerization Radiation-induced grafting is a highly versatile functionalization technique. It allows for the attachment of virtually any macromolecule, which can be polymerized through radical addition, with any polymer backbone forming radicals via irradiation (UV radiation, y-radiation or electron beam) [23]. The basis of this method is that ionizing radiation excites radicals in the polymeric backbone of the support material. Upon immersion in monomer containing solution, the polymerization reaction is initiated. The extent of grafting is regulated by the reaction conditions (time, radiation dose, and

333


concentration) [23]. In addition, the monomer solution is never exposed to irradiation alleviating problems typically associated with plasma-induced polymerization. One drawback associated with radiation-induced polymerization is that the solubility of the substrate material can be altered by this technique [24]. Stimuli-responsive membranes have been prepared through graft polymerization of acrylic acid (see Figure 1) onto a porous poly(vinylidene fluoride) support via radiation induced grafting [23]. These tunable membranes showed variable permeability upon changes in the pH and ionic strength of the permeate solution. The response or permeability change associated with this particular support was found to be highly dependent on the degree of polymer grafting. The concept of introducing a pH-sensitive response via grafted poly(acrylic acid) (PAA) has been used in the development of novel drug delivery systems [27-28]. In these particular studies, glucose oxidase was coimmobilized through amide bond formation with the grafted carboxyl side chains of PAA. The introduction of glucose into the system resulted in its conversion to gluconic acid resulting in a decrease in local pH. The subsequent protonation of the grafted PAA chains induces a conformational change causing the membrane pores to open and the release of insulin into the permeate. 2.2.2.3

Direct Covalent Attachment

Common problems associated with in-situ polymerization and other functionalization techniques are controlling the degree of polymerization (chain length) and the handling of expensive, poisonous initiators. To overcome these deficiencies, a great deal of research has been directed towards the development of functionalization techniques involving pre-existing, uniform macromolecules. In our work, we have been examining the use of polypeptide functionalized membranes (cellulose, silica) as high capacity sorbent materials [17, 29] and for variable permeability membranes [7]. These membranes were prepared via single-point, covalent bonding of poly(amino acids) to functionalized (aldehyde or epoxide), microporous supports. Synthesis consists of two steps, derivatization of surface functionalities followed by polypeptide attachment under convective flow conditions. The method of derivatization is dependent upon the type of membrane used as the support material. For cellulosic membranes, the formation of surface aldehyde groups was achieved through periodate oxidation [17] or by ozonation [7]. For silica-based membranes, epoxide formation was obtained by silane (3-glycidoxypropyltrimethoxysilane) attachment [ 17]. This reaction involves the methoxy groups of the silane with the silanol groups present on the silica support. Single-point attachment was then performed by permeating aqueous poly(amino acid) solutions through the

334


derivatized support under basic conditions. This is done to ensure that the amine terminus of the polypeptide is not protonated and will participate in the nucleophilic attack of the aldehyde/epoxide moities present within the pores of the support. This simple functionalization technique allows for attachment of macromolecules with well-characterized molecular weight distributions. Furthermore, it avoids the use of toxic initiators common to alternative methodologies.

2.2.2.4 Gold-Thiol Chemistry Attachment Techniques The development of comprehensive models describing transport in microporous membranes functionalized with linear polymers has been somewhat limited due to the non-uniform pore distributions common to most support materials (polyethylene, polypropylcne, cellulose, etc.). The fabrication of well-defined stimuli-responsive materials of this type can be achieved through simple gold-thiol chemistry involving track-etched supports [4, 6]. The chemisorption of thiols onto gold-plated surfaces is a well-established phenomena [30]. Deposition of Au nanolayers can be achieved through electrodeposition or by an electroless plating method [4, 30-31]. The latter technique, although more experimentally intensive, has shown more uniform metal deposition. In the electroless method, gold deposition is initiated at the pore walls yielding hollow cylindrical channels at short deposition times [32]. Self-assembled structures of thiol-containing compounds (poly(glutamic acid) with a terminal disulfide group [4], cysteine [6], etc.) spontaneously chemisorb on these gold surfaces. Drawbacks to this method include the formation of "bottleneck" pore structures that would introduce irregularities to the welldefined geometry of the support [30]. Also, practical applications are limited due to the low porosity associated with track-etched membranes and the number of processing steps required in the gold plating procedure. A number of research groups have employed gold-thiol chemistry in the development of environmentally-sensitive membrane materials [4, 6, 30, 32]. One such application has been the chemisorption of L-cysteine onto a goldcoated track-etched polycarbonate membrane having molecular dimensions (pore size as small as 0.9 nm) [6]. The amphoteric properties of chemisorbcd Lcysteine allows for adjustable ion-selectivity of organic anions and cations based on the pH of the feed solution.

335


2.3

PROPERTIES OF MEMBRANE MATERIALS CONTAINING TERMINALLY-GRAFTED LINEAR POLY(AMINO ACIDS)

The primary focus of our research group, regarding tunable separations, has been the functionalization of microporous membranes with charged (acidic or basic) poly(amino acid) homopolymers. It has been shown that the morphological and electrostatic properties of these biomolecules can be utilized for selective separations and controlled transport applications [7]. As described in the previous section, attachment involves the reaction of the amine terminus of the polypeptide with functional groups present within the pore structure of the support material.

2.3.1

pH-Sensitive Permeability and Ion-Selectivity

Various polypeptides, such as poly(L-glutamic acid) (PLGA), have been shown to undergo a conformational transition from a random coil structure to a compact helix based on variations in the properties of its contacting solution [4, 7, 33]. These properties include pH, salt concentration and counterion valency [33-35]. In particular, the effects of solution pH on the charge density and morphology of PLGA (see Figure 2) can be used to tune the separation properties of microporous support membranes. The conformation of this macromolecule is highly dependent on the degree of proton binding exhibited by the functional groups present along its backbone. The degree of protonation (X) of the attached polypeptide is given by equation 1 [ 17],

x =

K' exp(-eog/ / kT)[H] +

(1)

1 + K' exp(-eo¢," / kr)[H] + where K' is the intrinsic binding constant (K' = 1/K where K is the intrinsic dissociation constant), eo is the elementary charge, ~g is the electrostatic potential produced by the charged side group constituents, k is the Boltzmann constant, T is the temperature, and [H+] is the hydrogen ion concentration. At pH conditions above 5.5, the degree of proton binding is low and PLGA is fully ionized. Electrostatic repulsive forces between neighboring side groups cause the molecule to unwind and extend into a random coil formation [33-34]. In this conformation, PLGA molecules will extend into the membrane pore in a highly charged state, as shown in Figure 3a. This will allow for electrostatic interactions with ions present in the bulk solution in regions far removed from the pore surface. Also, the presence of these charged macromolecules creates an obstruction to solvent (water) transport. Conversely

336

Functionalized M e m b r a n e s For Tunable Separations and Toxic Metal Capture - Bhattacharyya

at low pH (< 4), a high degree of protonation alleviates the repulsive forces between side group constituents inducing a compact helical morphology as can be seen in Figure 3b. This structure is stabilized primarily by intramolecular hydrogen bond formation. Under these conditions, PLGA offers much less resistance to solvent transport. Furthermore, the reduction of the charge density of the attached macromolecule associated with this conformational shift will result in a dramatic change in the interaction between ionic species and the membrane.

e ChargedSide-GroupConstituent I

a)

|

rp Top View

Side View

b) Core Region Expands i

,,

Figure 3. Schematic representing the helix-coil transitions within the pore of a poly(Lglutamic acid) functionalized membrane; a) random-coil formation at pH > 5.5, b) helix formation at pH (< 4) (rp= pore radius, rc = pore radius - polymer chain length).

2.3.2

Permeation Studies The effects of poly(L-glutamic acid) and poly(L-arginine) functionalization on the pure water transport through a cellulosic-based support membrane are shown in Figure 4. Flux measurements, Jv, for both types of membranes were normalized using the maximum flux, Jv, max. At high pH, the PLGA functionalized membrane displayed a marked decline in the observed water flux. As shown in Figure 3a, PLGA exists in a fully extended random-coil

337


formation under these conditions. These extended polymer chains mitigate water transport resulting in the observed decline in permeability [5]. The permeability determined at high pH was 2.31 x 10-4 cm3/cm2 s bar. In its ionized form, this would correspond to greater than an order of magnitude increase in permeability over gel-filled MF membranes found in the literature [3]. At low pH, the flux of the PLGA functionalized membrane increased significantly •

........

...........

•

. . . . . . . .

%°

%

0.8

o

X

I~

E o.6

°o °

"t. . . ........... ~ . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.~

> 0.4 • Poly(L-glutamic acid)

0.2

A Poly(L-arginine)

2

i

i

i

r

i

i

3

4

5

6

7

8

pH

9

Figure 4. The pH dependence of water permeation through a cellulosic membrane functionalized with poly(L-glutamic acid) [7] and poly(L-arginine).

indicating the morphological shift of the attached polypeptide. On the other hand, the membrane functionalized with positively-charged poly(L-arginine) showed virtually no flux response to variations in pH. The guanadino side group of this residue has a pKa of about 12.5. Thus, poly(L-arginine) does not undergo a conformational transition under normal solution conditions. This clearly indicates that the environmentally (pH) sensitive properties of PLGA functionalized membranes is derived from the helix-coil transitions of the attached biomolecule. Results reported in the literature for a gold-coated polyearbonate membrane that had been functionalized with PLGA containing a terminal disulfide group are shown in Figure 5 [4]. The observed flux trend was

338


similar for both functionalization techniques indicating that the stimuliresponsive nature of PLGA is independent of the base support membrane. i

i

i

.......... 4, .......... ~.. °. ° % °o

0.8

•o

-,.

X °o

0.8

""'Oo

-)

.e

....~. >

"3

0.4

0.2

1

2

3

4

pH

5

6

7

8

Figure 5. The pH dependence of water flux through a poly(L-glutamic acid) functionalized gold-coated polycarbonate membrane (data taken from [4]).

2.3.3

Charged-based Separations In addition to controlled permeability applications, these membrane platforms can be used for ultra-low pressure charged-based separations. Conventional nanofiltration membranes separate electrolytes based on solute size and charge. NF membranes are generally comprised of monomeric acidic (COO-, SO3) and/or basic (NH3+) functionalities incorporated within a relatively dense (pore diameter < 2 nm) polymer matrix. The charged pore surface establishes an electric potential over its cross-section enabling Donnan-type exclusion of electrolytes. The separation principles of NF can be applied to microporous membranes through extension of the electric potential into regions far removed from the pore surface. This can be achieved through immobilization of charged poly(amino acids) [7]. 2.3.3.1

Feed Concentration Effects on Solute Rejection The dependence of feed concentration on solute rejection, R,

339


Cp Cf

R = 1-~

(2)

(where Cp and Cf are the solute concentration in the feed and permeate, respectively) for dilute divalent co-ion solutions (Ca 2+ for poly(lysine), SO42 for PLGA) is shown in Figure 6. The decline in solute rejection at higher electrolyte concentrations is consistent with established transport theory for charged media. It is brought about due to the saturation or shielding effect of counterions on the fixed membrane charge [7]. ~~

~

Positively Charged Membrane 12 solutions)

C 0.8 0

= m

dl=.l

@

1~ "3'

0.6

t¢ 0.4

A-BC (MW 36,400)

m

0

0.2

Negatively Charged Membrane (Na2SO4 solutions) ,

0

i

0.5

,

~

,

1

i

1.5

,

2

mM Divalent Coion Figure 6. Relationship between the rejection of divalent co-ions and feed concentration for cellulosic membranes functionalized with poly(amino acids).

2.3.3.2

Examination of this Non-Steric Mechanism of Ion Exclusion

The separation of ionic solutes with a microporous membrane (0.21 ~tm) can be considered a significant achievement when considering that the hydrated radii of calcium and sulfate ions are 0.412 and 0.379 nm, respectively [36]. Similar experiments performed using neutral 482,000 MW dextran with a molecular diameter greater than 36 nm [37] showed less than 10 % rejection for poly(L-lysine) (75,900 Da) functionalized membranes. Thus, the mechanism of

340


separation is obviously non-steric and can be attributed solely to the extension of the membrane charge into the inner portion of pore cross-section. This statement is further supported by the fact that poly(L-lysine) (75,900 MW) showed much greater performance than PLGA (36,400 MW) with respect to solute separation.

A

v e~ 0 in

0.8"

@ 0.6-

I'Y 0.4J je e

~,

0

0.2

e •

,oi,,ee, j,l,l,I I I oleDDI°~=e~lel=jjlq=°°~ee • | 18 B J D BB I J I ~T

I

e

e

i

0

0.2

0.4

0.6

0.8

Chain Length /Ave. Pore Radius Figure 7. Dependence of the solute rejection on the ratio between the poly(amino acid) (PLGA) chain length to the average pore radius for dilute Na2SO4 solutions (0.2 mM) with a permeate flux of 2.5 x 104 cm3/cm2 s (some data taken from [7]).

The chain length, associated with poly(L-lysine) extends much further into the membrane pore structure alleviating the effects of solute leakage in the core region (see Figure 3). To further examine the effects of solute leakage in the core region, rejection studies were performed on PLGA functionalized membranes with varying chain length to pore radius ratios at constant pH (- 6) and flux (2.5 x 10 -4 cm3/cm 2 s). The results of these experiments using dilute (0.2 mM) Na2SO4 solutions are shown in Figure 7. This figure clearly indicates that membrane performance is highly dependent upon the limitation of the core region for this type of functionalized platform [7].

341


2.3.3.3

Ion Transport in Poly(amino acid) Functionalized Membranes: Effect of Amino Acid Chain Length Electrolyte transport in charged media is generally described by the extended Nernst-Planck (N-P) equation, Jx,i =

~C i

-Di --~ +Vm Ci

-

O i z i C i F c3(I)

RT

(3)

~x

where Jx,i is the ionic flux in the direction of flow (x), Di is the diffusivity of ion i, Vm is the solution velocity across the membrane, zi is the ion valency, R is the gas constant, Ci is the ion concentration, F is the Faraday constant, and • is the total electrostatic potential within the membrane pore [38-40]. Models based on the N-P equations have been well accepted because they account for the superposition of all three modes of electrolyte transport (namely diffusion, convection and electromigration). It is well established that rejection, R, of electrolytes increases to asymptotic values at higher applied pressures for dense membranes (RO, NF) as is shown in Figure 8. For NF type membranes at low flux, the contribution of diffusion (1 st term of eqn 3) to the overall ionic flux is much more prevalent resulting in lower solute rejection. However, due to its 1.0

__ •

• mm= emmwnmmmnmmllmmmmmmnmmummmmm|

A--.~.4

C 0 0.8 mn "~"~ 0.6

~

R

........ Nanofiltration

0.4

Reverse Osmosis

m

0 01~ 0.2

"--

PP Functionalized MF (Long Chain)

=

PP Functionalized MF (Short Chain)

0.0

Low

AppliedP r e s s u r e

High

Figure 8. Schematiccomparison of polypeptide functionalized microporous membranes with conventional nanofiltration and reverse osmosis.

342


narrow pore structure, the electrostatic exclusion (3 rd term of eqn 3) of charged species is maintained by a relatively constant electric potential over the pore cross section. Thus, at high applied pressures, where convection is dominant (2 nd term of eqn 3), sustained electrostatic exclusion is coupled with enhanced water flux resulting in an increase in the observed solute rejection (see Figure 8) to an asymptotic value. In contrast, microporous membranes containing terminally anchored charged poly(amino acids) display these characteristics only over a certain portion of the pore cross section. In regions of the pore structure devoid of charge, the core or inner region (see Figure 3), electrostatic interactions can be assumed negligible. Solute transport in these regions will be unabated by the electric potential established by the grafted polymer. This solute leakage effect will be enhanced at higher volume flux resulting in a decline in solute rejection at higher applied pressures as is indicated in Figure 8. More effective membranes of this type can be synthesized through limitation of this core region. This can be accomplished by using polypeptides with a higher degree of polymerization (chain length, see Figure 8), by enhanced polymer loading or by reducing the pore diameter. The effects of poly(amino acid) chain length on the 1

© E 0.8 O

m e

@ A

0.6

t¢ ¢) 0.4

,4,,,I i

:3 O

P-Lys (75.9 kDa) with CaCI 2 solutions

0.2

/~ PLGA (36.4 kDa) with Na2SO 4 solutions 2.0

i

~

i

i

i

3.0

4.0

5.0

6.0

7.0

P e r m e a t e Flux x 10 4

8.0

(cm31cm 2 s)

Figure 9. The dependence of solute rejection on the permeate flux for polypeptide functionalized microporous cellulosic membranes using dilute (0.15 mM) divalent coion solutions.

343


relationship between solute rejection and permeate flux are shown in Figure 9. These experiments were performed using dilute, divalent coion (CaC12 or Na2SO4) solutions of approximately equivalent molar concentration. The cellulosic membrane functionalized with positively-charged poly(L-lysine),PLys (MW = 75,900) showed much greater solute rejection than the shorter poly(L-glutamic acid) (MW = 36,400). In addition, the rejection of CaC12 by the poly(L-lysine) functionalized membrane was much more stable over broad ranges of permeate flux (applied pressure). This can be attributed to the overall size of the attached poly(amino acid) and its ability to occupy larger portions of the pore structure.

2.3.3.4

An Example Application of Poly(amino acid) Functionalized Microporous Membranes for Arsenic Removal A viable application of charged-based separations involving porous supports functionalized with polypeptides is the removal of arsenic containing compounds from contaminated drinking waters. Arsenic is a known carcinogen that is linked to various forms of skin and internal cancers [41 ]. It is commonly 1.0 ¢~ 0.8

0

"~

0.6

monolvalent (HzAsO4-)) causes the observed solute rejection at pH 4 to decrease to approximately 29.3 percent.

0

POLYLIGAND FUNCTIONALIZED MEMBRANES FOR METAL CAPTURE

The functionalization of MF membranes with appropriate chelation groups allows for dissolved heavy metal (and anions like nitrate and arsenate) removal from drinking water or high quality water production at low pressure and high throughput rates. It should be noted that the tunable membrane studies presented before dealt with separations based on ion exclusion rather than multivalent (Pb 2+, Fe 3+, etc.) metal capture. For example, to exclude ions such as, HAsO42-, one needs carboxylic functionality (such as, PLGA) where as for capturing these ions requires positively-charged side groups (such as those

345

FunctionalizedMembranesForTunableSeparationsandToxicMetalCapture- Bhattacharyya

associated with poly(lysine) or poly(arginine)) The advantage of MF membranebased adsorbents is that functional groups can be attached to membrane pores

00"

+~

SH

SH

1 Convective flow

Figure 11. Schematicof poly-functionalized membrane sorbents with three different types of functional groups (carboxyl, amine, and thiol).

as polymeric ligands with multiple metal binding sites, rather than as monomeric surface functional groups. Hence, interactions between these groups and the heavy metal ions are very rapid, as sorption proceeds under convective flow and transport resistance is minimized. The overall objective of this research with regards to ion exchange is to create inexpensive sorbents for the entrapment of heavy metal ions at high capacity/rate over conventional ion exchange resins. Our system involves the use of inexpensive, commercially available cellulosic and silica membrane materials that are easily functionalized with polymeric ligands (such as, polyamino acids). A schematic showing various ligands inside a MF type membrane pore is shown in Figure 11. A wide variety of microfiltration membranes can be functionalized with polyamino acids for heavy metal sorption. Cellulosic membranes are generally derivatized with aldehyde groups, and polyamino acids are grafted via a Schiff

346

Functionalized Membranes For Tunable Separations and Toxic Metal Capture

-

Bhattacharyya

base reaction [44]. For a cellulose acetate (CA) base support, the derivatization involved two steps: partial hydrolysis followed by periodate oxidation to create

Cd sorption (0.5 - 1.4 g/g) I I Pb sorption (1.3 - 1.9 g/g) -¢¢ -

4

E

•

0.74

1.14

3.64

6.3

CA Composite CA PE-Silica Pure Cellulose mmoles x 104 Aldehyde / cnf External Membrane Area Figure 12. Metal Sorption Capacities for Cellulosic and Silica Based MF Membrane Sorbents. aldehyde groups. Silica-based membranes have also been functionalized (using silane chemistry) with polyamino acids [17]. Figure 12 shows metal sorption capacities of Cd and Pb (from nitrate salts, 1000 mg/L metals) for polyamino acid (polyglutamic and polyaspartic acids) functionalized cellulosic (Cellulose Acetate, CA Composite and CA, and Pure Cellulose) and silica based (Polyethylene,PE-Silica) MF membranes. Metal sorption (pH 5-5.5, no precipitate formation) is extremely high (>10 meq/g for Pb) compared to conventional ion exchange (1-3 meq/g). This high capacity is due to the availability of a large number of ion exchange groups and the role of counterion condensation [17] in addition to ion exchange. Helix formation (induced by ionic strength, pH, and by sorbed metals) also affects sorption capacity by compressing the chains and further enhancing electrostatic field strength. Anion sorption (nitrate, chromate, etc.) has also been demonstrated with positively charged polyarginine (with amine functionality) functionalized MF membranes, such as 0.9 moles Cr2072 per mole exchange group. The mode of metal capture with carboxyl and amine based poly-amino acids is primarily by ion exchange.

347


On the other hand, the use of chelating polyamino acids (such as, polycysteine) containing side-chain thiols (SH groups) provide selective and high capture efficiency for mercury salts [44]. Chelated ion exchangers [45] have wide industrial applications. Various researchers in Japan have also reported the incorporation of IDA (iminodiacetate) functionality and diethylamine groups on membrane supports [46,47]. Denzili et al [48,49] have shown the impregnation of membrane structures with multifunctional dyes for the capture of toxic metals, such as Alkali Blue 6B and Cibacron blue F3GA. Functionalized membrane sorbents offer a wide array of advantages for removing toxic metal cations and anions over conventional ion exchange and other pressure driven membrane processes [ 17, 50-52]. First, MF sorbents offer very low-pressure operation at comparable fluxes for RO/NF membrane separation. Production rates are also competitive with ion exchange, which must contend with pressure drop and channeling across the column. Kinetics for metal sorption is also very high. In competitive sorption technologies like ion exchange and activated alumina, high internal surface areas are provided by microporous (1-10 nm) structures, and hence metals must diffuse to the surface. Convective flow is achieved in MF sorbents, and diffusion mass transfer barriers are virtually eliminated. It should be noted that MF sorbents do have a fixed capacity, and hence will become saturated as any other sorbent and must be regenerated, unlike ion exclusion separation processes. For the regeneration of polyamino (carboxyl) a c i d s - based membrane sorbents, helix-coil transition phenomena also provide an added benefit of selective regeneration of metals because of strong retention of high acidity metals in the helix region [52]. For example, at pH 3 one can selectively recover Cd from a high acidity metal such as, Pb.

4.

ACKNOWLEDGEMENTS

The authors would like to acknowledge NSF-IGERT, and the US EPA, and NIEHS for the financial support of this project. The authors would like to thank Dr. J. Hestekin, Dr. S. Ritchie, Dr. L. Bachas, and Dr. S. K. Sikdar for their highly significant contributions on the metal sorption work. Noah Scherrer (NSF/REU support) conducted some of the experiments involving tunable separations. The authors also acknowledge Osmonics Corporation (for cellulose acetate membranes), and Daramic Corporation (for PE-Silica membranes) for providing some of the base membrane supports used in our work.

348


REFERENCES

[1]

[2] [3]

[4]

[5] [6]

[7]

[8] [9]

[10] [11]

[12]

[13]

[14]

D. Bhattacharyya, W.C. Mangum, and M.E. Williams, "Reverse Osmosis." in Encyclopedia of Environmental Analysis and Remediation ( Robert A. Myers, Editor), John Wiley & Sons, Inc., (1998)4149-4166. S. P. Nunes, and K.V. Peinemann (editors), Membrane Technology in the Chemical Industry, Wiley-VCH (2001). M. Ulbricht, Photograft-polymer-modified microporous membranes with environment-sensiti~'e permeabilities, Reactive & Functional Polymers 31 (1996) 165. Y. Ito, Y. S. Park, Y. Imanishi, Nanometer-sized channel gating by self-assembled polypeptide brush, Langmuir 16 (2000) 5376. A. M. Mika, R. F. Childs, J. M. Dickson, Salt separation and hydrodynamic permeability of a porous membrane filled with pH-sensitive gel, Journal of Membrane Science 191 (2001) 225. S. B. Lee, C. R. Martin, pH-switchable, ion-permselective gold nanotubule membrane based on chemisorbed cysteine, Analytical Chemistry 73 (2001) 768. A. M. Hollman, D. Bhattacharyya, Controlled permeability and Ion Exclusion in microporous membranes functionalized with poly(L-glutamic acid), Langmuir 18 (2002) 5946. T. Yamaguchi, T. Ito, T. Sato, T. Shinbo, S. Nakao, Development of a fast reponse molecular recognition ion gating membrane, Journal of American Chemical Society 121 (1999) 4078. Y. Choi, T. Yamaguchi, S. Nakao, A novel separation system using porous thermosensitive membranes, Industrial Engineering and Chemistry Research, 39 (2000) 2491. A. S. Hoffman, A. Afrassiabi, L. C. Dong, Thermally reversible hydrogels. II Delivery and selective removal of substances from aqueous solutions, Journal of Controlled Release 4 (1986) 213. Y. Ito, M. Inaba, D. J. Chung, Y. Imanishi, Oxidoreduction-sensitive control of water permeation through a polymer brushes-grafted porous membrane, Macromolecules 25 (1992) 7313. S. P. Schwendeman, G. L. Amidon, M. E. Meyerhoff, R. J. Levy, Modulated drug release using iontophoresis through heterogeneous cation exchange membranes: membrane preparation and influence of resin crosslinkage, Macromolecules 25 (1992) 2531. K. Konturri, S. Mafe, J. A. Manzanares, B. L. Svarfvar, P. Viinikka, Modeling of the salt and pH effects on the permeability of grafted porous membranes, Macromolecules 29 (1996) 5740. R. F. Childs, A. M. Mika, A. K. Pandey, C. McCrory, S. Mouton, J. M. Dickson, Nanofiltration using pore-filled membranes: effect of polyelectrolyte composition on performance, Separation and Purification Technology 22-23 (2001) 507-517.

349


[15]

[161

[17]

[18] [19]

[201 [21] [221 [23] [241

[25] [26]

[27] [28] [291

[3o] [31]

[32] [33]

V. Kapur, J. C. Charkoudian, S. B. Kessler, J. L. Anderson, Hydrodynamic permeability ofhydrogels stabilized within porous membranes, Ind. Eng. Chem. Res. 35 (1996) 3179. Y. Ito, Y. Ochiai, Y. S. Park, Y. Imanishi, pH-sensitive gating by conformational change of a polypeptide brush grafted onto a porous polymer membrane, J. Am. Chem. Soc. 119 (1997) 1619. S. M. C. Ritchie, L. G. Bachas, T. Olin, S. K. Sikdar, D. Bhattacharyya, Surface modification of silica- and cellulose-based microfiltration membranes with functional polyamino acids for heavy metal sorption, Langmuir 15 (1999) 6346. Y. Ito, Signal-responsive gating by a polyelectrolyte pelage on a nanoporous membrane, Nanotechnology 9 (1998) 205. A. M. Mika, R. F. Childs, J. M. Dickson, B. E. McCrary, D. R. Gagnon, A new class of polyelectrolyte-filled microfiltration membranes with environmentally controlled porosity, Journal of Membrane Science 108 (1995) 37. K. L. Buehler, J. L. Anderson, Solvent effects on the permeability of membrane supported gels, Ind. Eng. Chem. Res. 41 (2002) 464. T. Yamaguchi, A. Tominaga, S. Nakao, S. Kimura, (LOOK UP Title ) AIChE Journal 42 (1996) 892. Y. Choi, T. Yamaguchi, S. Nakao, A novel separation system using porous thermosensitive membranes, Ind. Eng. Chem. Res. 39 (2000) 2491. J. Hautojarvi, K. Konturri, J. Nasman, B. Svarfvar, P. Viinikka, M. Vuoristo, Ind. Eng. Chem. Res. 35 (1996) 450. T. Yamaguchi, S. Nakao, S. Kimura, Plasma-grail filling polymerization: preparation of a new type of pervaporation membrane for organic liquid mixtures, Macromolecules 24 (1991) 5522. B. Jeong, A. Gutowska, Lessons from nature: stimuli-responsive polymers and their biomedical applications, TRENDS in Biotechnology 20 (2002) 305. A. Kikuchi, T. Okano, Intelligent thermoresponsive polymeric stationary phases for aqueous chromatography of biological compounds, Prog. Polym. Sci. 27 (2002) 1165. Y. Ito, M. Casolaro, D. J. Chung, Y. Imanishi, Design and synthesis of glucosesensitive insulin-releasing systems, Drug. Deliv. Syst. 3 (1998) 391. Y. Ito, M. Casolaro, K. Kono, Y. Imanishi, An insulin-releasing system that is responsive to glucose, J. Control Release 10 (1989) 195. J. A. Hestekin, L. G. Bachas, D. Bhattacharyya, Poly(amino acid)-functionalized cellulosic membranes: metal sorption mechanisms and results, Ind. Eng. Chem. Res. 40 (2001) 2668. C. R. Martin, M. Nishizawa, K. Jirage, M. Kang, Investigations of the transport properties of gold nanotubule membranes, J. Phys. Chem. B 105 (2001) 1925. K. B. Jirage, J. C. Hulteen, C. R. Martin, Effect of thiol chemisorption on the transport properties of gold nanotubule membranes, Anal. Chem 71 (1999) 4913. Z. Hou, N. L. Abbott, P. Stroeve, Self-assembled monolayers on electroless gold impart pH-responsive transport of ions in porous membranes, Langmuir 16 (2000) 2401. W. Zhang, S. Nilsson, Helix-coil transition of a titrating polyelectrolyte analyzed within the Poisson-Boltzmann cell model: effects of pH and counterion valency, Macromolecules 26 (1993) 2866.

350


[34]

[35] [36] [37]

[38] [39]

[40]

[41] [42] [43] [44]

[45] [46]

[47]

[481 [49]

[50]

S. Nilsson, W. Zhang, Helix-coil transition of a titrating polyelectrolyte analyzed within the Poisson-Boltzmann cell model: effects of pH and salt concentration, Macromolecules 23 (1990) 5234. Y. Huh, T. Inamura, M. Satoh, J. Komiyama, Counterion-specific helix formation of poly(L-glutamic acid) in a crosslinked membrane, Polymer 33 (1992) 3262. E. R. Nightingale, Jr., Phenomenological theory of ion solvation: effective radii of hydrated ions, J. Am. Chem. Soc. 73 (1959) 1381. L. Hagel, Aqueous-size-exclusion chromatography, J. Chrom. Library 40 (1988) 119. X. Wang, T. Tsuru, S. Nakao, S. Kimura, Electrolyte transport through nanofiltration membranes by the space-charge model and the comparison with Teorell-Meyer-Sievers model, Journal of Membrane Science 103 (1995) 117. J. M. M. Peeters, J. P. Boom, M. H. V. Mulder, H. Strathmann, Retention measurements of nanofiltration membranes with electrolyte solutions, Journal of Membrane Science 145 (1998) 199. J. Schaep, C. Vandecasteele, A. W. Mohammad, W. R. Bowen, Modelling the retention of ionic components for different nanofiltration membranes, Separation and Purification Technology 22-23 (2001) 169. W. P. Tseng, Effects and dose-response relationships of skin cancer and blackfoot disease with arsenic, Environ. Health Perspect. 6 (1977) 109. Y. Sato, M. Kang, T. Kamei, Y. Magara, Performance of nanofiltration for arsenic removal, Water Research 36 (2002) 3371. E. M. Vrijenhoek, J. J. Waypa, Arsenic removal from drinking water by a "loose" nanofiltration membrane, Desalination 130 (2000) 265. S.M.C. Ritchie, K.E. Kissick, L.G. Bachas, S.K. Sikdar, C. Parikh, D. Bhattacharyya, Polycysteine and other polyamino acid functionalized microfiltration membranes for heavy metal capture, Environmental Science & Technology 35 (2001) 3252. P.S. Myers, How chelating resins behave, Metals Adsorption Workshop, U.S. EPA, Cincinnati, OH, May 5-6, 1998. K. Sunaga, M. Kim, K. Saito, K. Sugita, T. Sugo, Characteristics of porous anionexchange membranes prepared by cogra~ing of glycidyl methacrylate with divinylbenzene, Chemistry of Materials 11 (1999) 1986 G. Q. Li, S. Konishi, K. Saito, T. Sugo, High collection rate of Pd in hydrochloric acid medium using chelating microporous membrane, Journal of Membrane Science 95 (1994) 63. A. Denizli, K. Kesenci, M. Y. Arica, B. Salih, V. Hasirci, E. Piskin, Novel dyeattached macroporous films for cadmium, zinc and lead sorption: Alkali Blue 6Battached macroporous poly(2-hydroxyethyl methacrylate), Talanta 46 (1998) 551. A. Denizli, D. Tanyolac, B. Salih, E. Aydinlar, A. Ozdural, E. Piskin, Adsorption of heavy-metal ions on Cibacron Blue F3GA-immobilized microporous polyvinylbutyral-based affinity membranes, Journal of Membrane Science 137 (1997) 1. D. Bhattacharyya, J. A. Hestekin, P. Brushaber, L. Cullen, L. G. Bachas, S. K. Sikdar, Novel poly-glutamic acid functionalized microfiltration membranes for sorption of heavy metals at high capacity, Journal of Membrane Science 141 (1998) 121.

351


[51]

[52]

S. M. C. Ritchie, D. Bhattacharyya, Membrane-based hybrid processes for high water recovery and selective inorganic pollutant separation, Journal of Hazardous Materials 92 (2002) 21. S. M. C. Ritchie, D. Bhattacharyya, "Polymeric Ligand-Based Functionalized Materials and Membranes for Ion Exchange," in Ion Exchange and Solvent Extraction Series (A.K. SenGupta, Editor), Vol. 14, (2001), 81-118.

352


Chapter 17

The design of high performance, gel-filled nanofiltration membranes R. F. Childs* and A. M. Mika

Department of Chemistry, McMaster University, Hamilton, ON, L8S 4M 1, Canada SUMMARY This chapter is concerned with the underlying factors that control the performance of polyelectrolyte gel-filled membranes in nanofiltration (NF) applications. It shows how an understanding of these factors can be used to optimize the performance of this new type of NF membrane. 1.

INTRODUCTION

There is a considerable amount of work currently underway on the development of high performance NF membranes. Typically these membranes are based on a supported, dense, thin-film construct, Fig. l a [ 1]. Major strides have recently been made in understanding the fundamental parameters controlling transport through dense thin-films and this is leading to substantial improvements in membrane performance [2-12].

a)

b)

thin dense separatinglayer-~

Vu,

u u o

ouvJ~U,

t

.......

r

. . . . . . . . . .

Fig. 1. Nanofiltration membrane constructs, a) conventional, b) gel-filled.

353

The Design Of High Performance, Gel-Filled Nanofiltration Membranes - Childs

In our work we have been examining the use of relatively thick, but low density, charged gels anchored within a microporous support membrane as an alternative architecture to thin-films, Fig lb. Membranes with this type of construct can have performances that at least match those of the best NF membranes currently available commercially [ 13,14]. Pore-filled membranes are not new p e r se. For example, there is a considerable amount of work reported on supported liquid membranes [15-18]. These membranes can be effective in a variety of applications; however, they typically lose the incorporated liquid over time and are not stable when subjected to a trans-membrane pressure difference [ 19]. Approaches to overcoming this stability problem include skinning the two surfaces of the membranes thereby physically encapsulating the liquid [20-22]. An alternative and effective approach involves anchoring the separating component within the pores of the microporous host and thus preventing leakage under operating conditions. Typically the separating component is a functionalized, solvent swollen polymer that is either grafted to the pore-walls of the support [23-26] or entangled with the structural elements of the support by cross-linking of the polymeric gels [27-31 ]. In this latter case there need be no chemical bond between the supporting membrane materials and the incorporated gel. Table 1. Water-softening capability of pore-filled membranes obtained with untreated municipal tap water a at 100 kPa. Membrane

Flux

Rejection (%)

kg/m2h

Na +

Mg 2+

Ca 2+

Poly(vinylbenzyl ammonium) (Gel 2)b

12

35

79

80

Poly(4-vinylpyridinium) (Gel 1)b

8

20

74

61

60

Sucrose

a Typical composition: Na+: 12 ppm; K+: 0.8 ppm; Mg2+: 8 ppm; Ca2+: 38 ppm, F: 1 ppm; CI: 25 ppm; NO3: 2.8 p p m ; 8 0 4 2 : 26 ppm; HCO3" and CO32: unmeasured; pH: 7.8. b For structures of the gels see Fig. 3.

The pore-filled membranes used in NF applications contain water-swollen polyelectrolyte gels. The potential performances of this type of membrane are evident when compared with some of the best commercial, water-softening membranes. Comparison of several pore-filled membranes is made with Desal-51, a high performance NF membrane produced by Osmonics, Table 1. As can be seen, pore-filled membranes containing a crosslinked poly(vinylbenzyl ammonium salt)

354


gel have fluxes that are almost twice that of Desal-51 at 100 kPa with somewhat higher bivalent cation rejections. These pore-filled membranes exhibit reasonably linear pressure-flux relationship up to at least 500 kPa and there is no loss of gel or irreversible changes observed at these higher pressures and high flux conditions. Separation changes with flux in a fairly typical manner [14]. The obvious question that arises from these results is what is the NF performance limit of polyelectrolyte filled membranes? In other words, can their performances be substantially enhanced from that shown in Table 1 to the point that they would represent a significant improvement in ultra-low pressure water softening applications? In order to answer such questions it is necessary to fully understand the factors affecting the performance of these pore-filled membranes. 1.1.

Some Basic Considerations

Pore-filled NF membranes consist of two components, namely a support component and a water-swollen polyelectrolyte gel, Fig. 2. The support membrane provides containment and protection of the gel, preventing or restricting its osmotic swelling. The host also largely prevents hydraulic pressure and/or flow induced changes in gel distribution and morphology when in use. Solvent and solute transport through the membrane only occurs through the gel phase.

polyelectrolyte chains

Fig. 2. Schematic representation of the components of a polyelectrolyte-filledmembrane. This simple picture of a polyelectrolyte-gel filled membrane, Fig. 2, already points to an inherent trade off that must be made in membrane design, namely the balance between the porosity of the support membrane and its mechanical strength. In principle, the highest flux would be achieved with no support membrane, i.e.,

355


A100% porosity". This is an untenable situation, however, due to the unrestricted swelling or compression of the gel. On the other hand, a low porosity substrate would afford a membrane with excellent mechanical properties but a low flux as the gel-volume fraction is reduced. A second point to come out of this simple picture is that the nature of the gel itself will be the primary determinant of membrane performance. We can view the gel phase as a two-component system with a gel-polymer and water, Figure 2. Water/solute transport will occur through the water and not the gel-polymer itself. In other words, we would expect the volume ratio of the gel-polymer to water to be a major determinant of flux. Polyelectrolyte gels bear a charge and as such a Donnan potential will be established between the membrane and the surrounding solution. The magnitude of this potential, and hence separation ability of ionic solutes, will depend on the relative charge concentration in the membrane and solute concentration in the contacting solution. Clearly there will be a trade offbetween the requirement for low gel-polymer volume fractions for high flux and higher volume fractions needed to achieve high charge concentrations for good ionic solute rejection. These are very simple considerations of performance. In the following sections we take a more quantitative look at these factors and show how they can be used to obtain polyelectrolyte filled membranes with significantly better performances than those in Table 1. 2.

FACTORS AFFECTING THE PERFORMANCE OF POREFILLED MEMBRANES

2.1

Hydraulic Flux/Permeability Determinants: the Properties of the Gel As noted above, transport of solvent through gel-filled membranes occurs only through the gel phase. Kaput and Anderson [27] have shown that the Darcy permeability ofpoly(acrylarnide)-filled membranes is related to the permeability of the pore-filing gel, as shown in Eq. (1) k m =kp £

(1)

where km is membrane permeability, kp is gel permeability, e is the porosity of the substrate and r, tortuosity. The tortuosities of typical microporous membranes do not differ greatly, however, their porosities are an important determinant of

356


membrane permeability. As noted above, the substrate should have as high porosity as is practicable and consistent with membrane strength. The intrinsic permeability of the gel is the dominant control parameter. There are a large number of studies on gel permeability including gel sedimentation rates. In essence, gel permeability is a function of the gel-polymer volume fraction, ~. The same relationship holds when the gels are constrained from swelling by incorporation into a microporous substrate, however, one must be careful to base the gel-polymer volume fraction on the pore volume of the support membrane and not that of an unconstrained gel.

kp = f ( ~ )

(2)

We have studied the relationship between membrane permeability (or gelpermeability through Eq. (1)) and the polymer volume fraction of a variety of polyelectrolyte gels incorporated into the same poly(propylene) host membrane.

"

1

2

HO~~,.OH

3

H HO= ~O,H''~" "~o.. 4

5

..~-o

me~O.,H E

Fig 3. Structures of polyelectrolyte gels inserted into microporous hosts: 1 - poly(4vinylpyridinium salt), 2 - poly(vinylbenzylammonium salt), 3 - poly(ethyleneimine), 4 poly(acrylic acid), 5 - poly(styrenesulfonic acid), 6 - poly(2-acrylamido-2methylpropanesulfonic acid).

As shown in Fig. 3, these gels include positively charged gels (anion exchange membranes) and negatively charged gels (cation exchange membranes). The relationship between gel permeability and gel polymer volume fraction for membranes containing polymer 1; (R = N-benzyl) is shown in Fig. 4. Over the polymer volume fraction studied, there is an inverse relationship between ~ and kp 357


that follows the relationship shown in Eq. (3). kp = 6.10 x 10- 2 2 ~ - 3 . 5 5 (m z)

(3)

There is a very strong dependence of permeability on polymer volume fraction for this gel polymer and it is only at low values of ~bthat membranes will have high permeabilities. This relationship shows why typical ion-exchange membranes, which have low water contents and consequently high polymer volume fractions, are not suited for pressure driven applications.

100

A

10

om

.Q t~

E

'-Q I1.

1

"~ (.9

Calculated

0.1

.....

0~

011

Gel-Polymer Volume Fraction ~

(-)

Fig. 4. Relationship between gel Darcy permeability and gel-polymervolume fraction for a membrane made using gel 1. The solid line is derived using the model described below. (Reproduced with permission from Ind. Eng. Chem. Res. 40 (2001) 1694-1705. Copyright 2001 Am. Chem. Soc.). We have recently described a model that can be used to calculate gel permeability from first principles [32]. The gel is modeled as an assemblage of spheres, Fig. 5, with the sphere diameter being equal to the correlation length, (, of an equivalent semidilute polymer solution. The correlation length is calculated according to the Schaeffer relationship [33], Eq. (4), where n is the number of

358


bonds of the length a in the polymer persistence length, X is the Flory-Huggins interaction parameter, and w is the three-body excluded volume parameter.

n2a ~=[n(l_2z~+wa_6¢2In

(4)

The effect of charge on the persistence length of the polyelectrolyte is included. The permeability of the sphere assemblage is calculated using Happel=s cell model [34].

Fig. 5. Gel modeled as an assemblage of spheres. It can be seen from the solid line shown in Fig. 4, this sphere model for gel permeability predicts the experimental data exceptionally well. Other models for gel permeability, such as the fiber model [35,36], cannot be used to fit data obtained over a wide range of polymer volume fractions without the artificial device of adjusting the thickness of the fiber. We have examined the importance of the various terms in the Asphere@ model. The gel-polymer concentration is a dominant factor. In principle chain stiffness, n, is important and it would be expected that this would be greatly affected by the high nominal charge on the polyelectrolyte-gels. However, this effect is not of major importance due to charge shielding and counter-ion condensation. The Flory-Huggins solvent interaction parameter is also an important term. This is exemplified in Fig. 6 where the permeabilities of two gels with 359


different solvent interaction parameters are calculated as a function of ~b. It is interesting that the more hydrophilic the gel polymer, the lower the permeability of the gel for a given value of ~b. A

E

100

"7,

•~-

lO

"'''"

Q

0.45

E

a.Q

1

e-

.m

~0.3s_J ~

.....

,10

E Q ~

0.1

O.O4

0.08

0.12

0.16

Gel-Polymer Volume Fraction, ~ (-)

Fig. 6. Effect of the Flory Huggins polymer-solventinteraction parameter of the gel on membrane permeability. These effects were verified by measurement of the permeabilities of a series of membranes containing the gel polymers 1-4 and 6, Fig. 7. A series of correlations is seen for these membranes with the permeabilities being dependent on the gel polymer. The calculated permeabilities of the membranes, shown as the lines in Fig. 7, closely match the experimental results. Poly(vinylpyridine) and poly(vinylbenzyl ammonium) based systems ( gels I and 2, respectively) have very similar permeability relationships. Poly(acrylic acid) (gel 4) filled membranes have permeabilities slightly lower than those with gels 1 and 2, consistent with higher hydrophilicity of 4. Poly(ethyleneimine)-based membranes (gel 3) show somewhat higher permeability caused by the higher rigidity of the branched polymer. The results obtained with poly(2-acrylamido-2-methyl-propanesulfonic acid) gels, 6, show the largest difference from the other gel-polymers and membranes made with this gel-polymer have significantly lower permeabilities. This is in large part due to the very hydrophilic nature of this polymer. At this point we cannot model this polymer precisely as we are in process of determining some of its basic parameters to insert into the model. It can be seen, however, that the shape of the relationship between membrane permeability and polymer-volume fraction parallels those of the other systems. The assumption made in the Asphere@ model presented above is that the 360


incorporated gels are homogeneous and evenly distributed through the thickness of the membrane. For all the membranes shown in Fig. 7, we carefully checked that when the gels were made in test-tubes they were perfectly clear and showed no solvent syneresis. If this is not the case then deviations from these correlations would be expected to occur. This can be seen in Fig. 8 where the permeabilities of a series of membranes with a constant polymer fraction of a poly(vinypyridinium salt) gel but increasing degree of cross-linking are shown. ,,,10000

E ~" "7. O

~e

1000

>~

100

m ... "Q t~

10

:" ",~: ~,

4)

'~'~,.'.:,., .

E e~

[] [] ....... [] .......... • ©

D[]

Gel Gel Gel Gel Gel Gel Gel Gel

1- Model 1- Experiment 2 - Experiment 3 - Model 3 - Experiment 4 - Model 4 - Experiment 6 - Experiment

I

o.1

E O.Ol 0.001 0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

Gel-Polymer Volume Fraction, # (-) Fig 7. Membrane permeability as a function of ~bfor various gel polymers (for chemical structures of gel polymers see Fig. 3).

As can be seen, counter to one's intuition where one would expect a reduction in permeability as the net size of the gel decreases with increasing crosslinking, there is a sizeable increase in permeability. The origin of this effect would appear to lie in increasing heterogeneity being induced in the gel with increasing cross-linking. This will have the effect of forming polymer rich ~on-draining@ and polymer depleted Adraining@ regions within the gel. Given the large sensitivity of permeability to the polymer volume fraction, the effective reduction in pore volume due to the formation of the non-draining regions is more than compensated by the reduction in polymer volume fraction of the draining regions leading to the enhanced permeability. Similar effects are seen with poly(acrylic acid) gels where micro-heterogeneity in the gels can be induced by using a poorer solvent in the polymerization reactions used to make the membranes [37]. It is interesting to note these effects can also be dynamic. We have recently shown that reversible changes in permeability of membranes can be induced by the use of weak base polyelectrolyte gels such as poly(4-vinylpyridine) and feed solutions of different 361


pH [38]. The overall picture emerging from this work is that the permeability of these gel-filled membranes is very predictable if the properties of the gel are known. More work is required to fully define all the parameters, particularly the quantitative analysis of gel heterogeneity and the effect of counterion, and water of solvation on permeability. However, this analysis already provides the quantitative tools to optimize the properties of the gel. Clearly these ultimately must be combined with the separation properties associated with these gels and this will be analyzed later in this chapter. But first, how does the base membrane affect these membranes? A N

100

Degree

of Crosslinking

(mol-%):

0 t25 0,-20 e,..15

G. r,,

..5-10

1

.Q

E o :E

0.1 0.04

i ,

,.

, i 0.06

,

,

,

,

i

0.08

,

,

,

J i 0.1

,

,

,

J i 0.12

,

,

,

,

i

0.14

. . . . 0.16

Gel-Polymer Volume Fraction, ~ (-) Fig 8. Effect of increasing the degree of cross-linking on the permeability of a membrane containing a poly(4-vinylpyridinium salt), gel 1 (line: model, points: experiment).

2.2 Hydraulic Flux~ermeability Determinants: the Properties of the Supporting Membrane The above analysis of membrane permeability has largely focused on the properties of the incorporated gel and the importance of the base membrane parameters such as porosity and pore tortuosity. As hydraulic flux under pressure is a key operating parameter, it is important to know what other factors are important in selection of a suitable supporting membrane. A comparative study of two types of support membranes has been carried out. The first support is a poly(propylene) membrane produced by the 3M Company using a thermally induced phase separation process (TIPS), Fig. 9a, and the second one is a non-woven, ultra-high molecular weight poly(ethylene) membrane produced by DSM Solupor, Figure 9b. Examination of the SEM images 362

The Design Of High Performance, Gel-Filled Nanofdtration Membranes - Childs

in Figure 9 shows that these membranes have quite different morphologies. Despite this, their porosities are essentially identical (80%). The thicknesses of the starting membranes are given in Table 2. It should be noted that the 3M and first Solupor membranes have essentially the same thickness. The second Solupor membrane has a much smaller thickness.

•

Fig 9. SEM micrographs of different microporous support membranes: a) 3M Company poly(propylene) and b) DSM Solupor UHMW poly(ethylene) membranes.

The pure-water flux of a pore filled membrane, Jw, is related to the membrane permeability, kin, thickness, Ax, of the gel layer, and the pressure difference, AP, across the membrane, as shown in Eq. (5). AP J w oE k m ~

(5)

Ax

The gel layer occupies the entire thickness of the support membrane in the porefilled membranes currently under discussion and, as a result, the gel thickness corresponds to the total average thickness of the pore-filled membrane. Thus, if one continues to fill the entire thickness of the membrane with the gel then to enhance flux we need to use thinner, but still high porosity, support membranes. However, the issue is more complex than this as is shown by the results obtained with the two 363


hosts. The two supporting membranes were filled with either poly(N-benzyl-4vinylpyridinium salt) (1 R=Be) or poly(N,N,N-trimethyl-vinylbenzylammonium salt) (2) gels with very similar gel-polymer volume fractions. The thicknesses of the resulting gel-filled membranes were measured as well as their pure-water fluxes. The results are given in Table 2. Table 2. Comparisonof different host-membrane substrates. Basemembrane

Filling gela

3M

Thickness (,um)

Flux at 100 kPa (kg/m2h)

Starting

Filled

Measured

Calculated

2

79

80

7.3

7.1

Solupor

2

92

146

3.8

21.4

Solupor

2

47

66

8.2

13.7

Solupor 1 (R=Be) 47 aFor structures of the gels see Fig. 3.

160

5.2

18.5

Several key features stand out from these results. First, there is a substantial difference in the ability of the two different types of membrane to resist osmotic swelling of the gels. This is clearly evident from the first two entries where the nonwoven Solupor membrane undergoes a substantial increase in thickness on introduction of gel 2 while the 3M support remains the same thickness within error. The degree of thickness increase also depends on the properties of the gel-polymer chosen. This is evident in the last two membranes where the more hydrophilic poly(vinylpyridine) causes the base membrane to swell by more than a factor of 3! The pure-water fluxes (100 kPa) of the membranes were measured and compared to the values calculated using the permeability model referred to above, Table 2. As the exact masses of incorporated gels are known the polymer volume fractions, membrane permeabilities, and membrane fluxes can easily be calculated using the measured dimensions of the various gel-filled membranes. As would be expected based on the substantial thickness increase associated with the Solupor membranes, they are predicted to have substantially lower gel-polymer volume fractions and, consequently, higher fluxes than the 3M support-based membrane. The agreement between the measured and estimated fluxes of the 3M-based membrane is remarkable and points to the power of the permeability model outlined above. The results with the Solupor membranes are quite different, and in each case

364


the measured fluxes are substantially smaller than the estimated values. It would appear that just as the Solupor membranes lack the necessary mechanical strength to prevent osmotic swelling of the membrane, they also lack the strength to prevent compression of the membrane under hydraulic flow. Such a compression would increase the effective gel-polymer volume fraction and reduce flux. Evidence supporting such a compression comes from the pressure/flux relationships for the Solupor based membranes where, unlike the ones typically found for the 3M membranes, they are markedly non-linear [14]. It should be stressed that as membrane permeability is so strongly dependent on the polymer volume fraction of the gel, even small compressions of the membrane will lead to very significant reductions in flux. One further feature of the membranes summarized in Table 2 is the middle two entries where the membranes have identical gel compositions and concentrations but the support membranes have different thicknesses. Just as would be expected, the thinner membrane has a substantially higher flux. Overall, these results with the two different supports emphasize the importance of choosing a support with the appropriate properties. The Solupor membrane is a very interesting product, however, it is not well suited for use with polyelectrolyte gels in that it lacks strength in the thickness dimension. The 3M TIPS membrane couples a high porosity with high mechanical strength. We also note that care must be taken not to modify the properties of the base membrane during insertion of the gel. It would appear, for example, that with membranes made by in-situ radical polymerization of monomers there is frequently some modification of the support presumably resulting from either grafting or interpenetration of the gel polymer with the support. This can lead to fairly large thickness changes in the membranes on introduction of the gel. On the other hand, membranes made by in-situ cross-linking of a preformed polymer typically show greater dimensional stability. 2.3

Determinants of Separation: Neutral Solutes The results given in Table l show that there is a key difference in separation performance between polyelectrolyte gel-filled and conventional thin-film membranes. This difference lies in their basis of separation. For example, with sucrose, a relatively large organic molecule, a Desal-51 thin-film membrane exhibits a fairly large rejection (>60% at 100 kPa) while a poly(4vinylpyridinium salt) gel based membrane has 90% asymptotic rejection of PEG 4000.

1.0 .s"

0.9

i Em~O. 8

•"

0.7

s "S

..o.....

~'"

0 0.6

J •

.....

"""•"

/"

.o"*"

•.......

m 0.5

O:EE!:OOO

"~'rw 0.412

"""'°""

0.3

~.

0.2

E

o.1

]

o.o . . . . 0.02 0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

Gel Polymer Volume Fraction, ¢ (-)

Fig. 10. Asymptotic rejection of PEGs of different molecular weight as a function of poly(4vinylpyridinium salt) volume fraction.

366


1.0

'm 0.8

0.6

'0.4

n,

Legend

~ o . 2 tl

- -

•

Mode!

Experiment

0.0 0.0

1.0

2.0

Volume

3.0

4.0

5.0

6.0

Flux, Jv (104 m/s)

Fig 11. Calculated and measured rejection of PEG 4000 by a poly(N-methyl-4-vinylpyridinium chloride) membrane with ~b= 0.08V0.005.

This prediction of PEG 4000 rejection, based on the assumption that rp .2~, has been tested experimentally. The rejection of PEG 4000 obtained with a membrane with a poly(N-methyl-4-vinylpyridinium chloride) gel (polymer volume fraction ~b= 0.08V0.005) is shown in Fig. 11 as a function of volume flux. The real rejection of a PEG 4000 was calculated from the observed rejection using the mass transfer relations for a stirred cell [9]. As can be seen the real rejections reach a value of ca. 0.8. The steric hindrance model [9] was used to calculate the fluxrejection curve for a membrane having the thickness and porosity equal to those of the gel-filled membrane. As can be seen from Fig. 11, the experimental points fall very close to the simulated curve. This result further substantiates the modeling approach we have taken with these membranes.

2.4

Determinants of Separation: Charged Solutes The basis of salt separation with these charged, gel-filled membranes is Donnan exclusion rather than sieving. This is well exemplified by the data shown in Fig. 12 where the single salt rejections of NaC1 and Na2SO4 with a positively charged membrane based on a poly(vinylbenzyl ammonium salt) gel are plotted against solute concentration. Both salts have the same monovalent cation (Na+), the co-ion in this case, but different counter-ions. The Donnan potential with the divalent counter-ion (sulfate) is expected to be substantially lower than that with monovalent counter-ion (chloride) and decrease rapidly with increase in solute concentration. As a result, the rejection of sodium sulfate is markedly lower than that of sodium chloride and rapidly decreases with increase in concentration. The question naturally arises as to whether there are suitable models to account for such 367

The DesignOf High Performance, Gel-FilledNanofiltration Membranes- Childs

behavior and which can be used to optimize gel composition? ........................................................................... _ ~

100

90 so

•"

CI

30 a2SO4

20 10 0 0

1O0

200

300

400

500

600

700

800

Salt Concentration, c, (ppm)

Fig 12. Single salt rejection as a function of salt concentration in feed for a membrane with a poly(vinylbenzyl ammonium salt) gel (gel-polymer volume fraction ~ = 0.16; applied pressure: 500 kPa).

The fixed-charge or Teorell-Meyer-Sievers (TMS) model [40-43] predicts that the salt rejection by charged membranes depends on the ratio of the membrane effective charge density to the salt concentration in the feed and not on their absolute values [44,45]. The main assumption underlying the model is homogeneous distribution of aqueous interstices through which convection and diffusion take place. The application of the TMS model to gel-filled membranes is justifiable provided that the distribution of electrical potential and ion concentration across the gels are uniform. Clearly, for gels with large mesh sizes this assumption will be invalid as the concentration of ions will vary in the radial direction. What constitutes a large mesh or pore size? In fact, a radial variation would be expected when the pore radius is substantially larger than the Debye-Htickel screening length, x-:, at the ionic strength of the feed solutions [44]. For such cases, the socalled space charge (SC) model of electrolyte transport through charged capillaries proposed by Osterle and coworkers [46-48] would be a more realistic model to predict salt rejection by porous charged membranes. However, for the NF membranes, under discussion in this chapter, the TMS model is applicable. Thus uniform ion-concentrations and electrical potentials would be expected for membranes with gel-polymer volume fractions in the range of 0.08 to 0.12, provided the ionic strengths of the feeds do not exceed 15 to 30 mM, respectively. The total charge density in a unit membrane volume can be calculated based on the polymer volume fraction in the pore-filling gel, partial molar volume of the gel polyelectrolyte, and porosity of the substrate. The effective membrane charge density may, however, be substantially lower than the nominal densities calculated

368


from the concentrations of ionic sites on the polymer chains due to binding (association) of a fraction of counter-ions to the fixed charge of the polyion. Assuming that no specific binding of counter-ions to the fixed charge takes place, the effective charge density can be estimated from the Manning counter-ion condensation theory [49]. The theory postulates the existence of a limiting value of the charge density along a polyelectrolyte chain above which the system (polyion and counter-ions) is unstable. At this point a sufficient number of counter-ions must condense onto the polyion to reduce its charge density to the limiting value such that the distance between charges along the chain contour is equal to the Bjerrum length (0.713 nm in water at 25°C). For polyelectrolyte chains of vinyl monomers in water, this condensation should start at approximately 35 % ionization [50]. The asymptotic rejections of NaC1 of various concentrations were calculated as a function of the gel-polymer volume fraction using the TMS model [44,45]. The results obtained with positively charged gels ofpoly(4-vinyl-N-benzylpyridinium chloride) are shown in Fig. 13. The dotted lines indicate the regions of gel-polymer volume fractions where the assumptions of the TMS model may be invalid due to the pore-radii becoming larger than the Debye-Hiickel screening lengths. The data presented in this figure show clearly that monovalent salts rejections well exceeding 50 % are available even at the feed concentration as high as 25 mM ( -~ 1500 ppm NaC1) with the higher gel-polymer volume fractions. Rejection is a function of the charge density of the polyelectrolyte gel incorporated into the membranes. Increasing the charge density by increasing the gel-polymer volume fraction, as was done in the data recorded in Fig. 13, will inevitably lead to a reduction in flux at a constant pressure. An alternative way of increasing the charge density would be to use polyelectrolytes with lower molecular weight repeating units and low partial molar volumes. Such an approach should give higher charge densities without a major increase in gel-polymer volume fraction. The net result should be higher rejections without undue reduction in the membrane permeability or flux. We have examined the separation of sodium chloride with the gel-filled membranes containing different pore-filling gels and have found a remarkable similarity [51 ]. In Fig. 14 we show NaC1 rejections for a series of membranes as a function of membrane hydrodynamic permeability. Each point in the figure represents a different membrane. The line is a polynomial fit of the experimental points. The striking feature is that all the membranes, whether filled with strong base or strong acid polyelectrolyte-gels, show directly comparable performances. Indeed, the data can be correlated with a single line!

369


1.00

0.95

0.90 0.85 0.80

0

0.70 0.65

NaCl

0.55 0.50

,

,

,

0.04

,

i

,

,

,

0.06

,

L

,

,

,

0.08

,

i

. . . .

0.1

ncentration

ii i

,

,

,

0.12

mM

,

i

,

,

,

,

0.14

0.16

Polymer Volume Fraction, ~ (-)

Fig. 13. Asymptotic rejection of NaC1 as a function of gel-polymer volume fraction in poly(N-benzyl-4-vinylpyridinium chloride). The dotted parts of the curves are regions where the TMS model may be invalid as described in the text.

100

o~"

©

80

~

0

Poly(vinylpyridinlum) or poly(vinylbenzylammonium)

•

Poly(ac~lamido-2-methyl

©

~

Ifonic

a c i d }

O

60 (.1

G)

"3' m

40

m

z

20 0 0.1

,

~

,

,,

,~,1

,

,

1

,

, , ,,~v-,~G~

,

~ ~ ,

10

,,~

100

Membrane Permeability, k m (10 "18 m 2)

Fig. 14. NaC1 separation for a series of membranes as a function of membrane permeability. (The data was obtained with 300 ppm (-~ 5 mM) NaC1 feeds at 300 kPa). The data in Fig. 14 can be divided into two distinct regions with the boundary being at a permeability of about 2 -_-1018 m 2. This boundary point corresponds to a gel concentration of ~b- 0.08. In the region of lower permeability (~ 3 0.08), the salt rejection is very high (70-80%) but remarkably, this rejection

370


shows tittle sensitivity either to large changes in membrane permeability (over one order of magnitude) or the accompanying changes in charge density. This lack of sensitivity comes from high charge density as compared to the feed concentration. As shown in Figure 13, at 5 mM NaC1 concentration the asymptotic rejection shows little dependence on polymer volume fraction. This means that the ratio of the fixed charge concentration to the salt concentration is high over the whole range of gel concentration. In contrast, rejections in the region of higher permeability (~< 0.08), decrease very rapidly with increasing membrane permeability and approach zero at the permeability of about 10-17m 2. More work is required to fully characterize the behavior of these membranes; for example, the role of factors such as concentration polarization with membranes of very high fluxes and hydraulic instability of gels at low polymer volume fractions need to be probed. However, even at this point, it is clear that the membranes shown in Fig. 14 can readily be optimized in terms of flux and separation by adjustment of the gel-polymer volume fractions.

lOO

90 80

A

70 o

60

(.1 q)

50

"6' n,

40

0

30

Z

20

C

ly(acrylicacid)

~er

gels

lO

o o.1

i

,

,

i

,

,lhl

i

L

~

,

,

LL,I

1

Membrane Permeability,

~

,

,

,

10 k m (10 -18 m 2)

a

~

i,i

100

Fig. 15. NaC1 rejection as a function of membrane permeability of poly(acrylic acid) containing membranes (feed: 300 ppm NaC1, pressure: 300 kPa).

The separation behavior of membranes containing poly(acrylic acid) gels is quite different from the general trend found with both strong base and strong acid membranes, Fig. 15. The salt rejection of this weak polyacid is substantially lower than that found with other gels and is independent of membrane permeability (gelpolymer volume fraction) over an extended range. While we continue to study this and related systems, the lower rejection of NaC1 indicates that the effective charge

371


density in these membranes is substantially lower than in the other membranes. This is suggestive of a very strong polyelectrolyte effect and a large shift of its pKa to make it a weaker acid on incorporation into a microporous support. A large change of dissociation constant, amounting to more than two units ofpKa, has been found, for example, for membranes filled with poly(4-vinylpyridine) gels [38,52]. The work presented thus far has largely focused on single salt separations and understanding the behavior ofpolyelectrolyte filled membranes. The situation becomes more complex with mixtures of solutes particularly where specific interactions of solute ions with a gel can occur. For example, the performance of a series of different membranes with a municipal tap-water feed is given in Table 3. As would be expected with separation being based on Donnan exclusion the charge type of the filling gel is important. As can be seen, the negatively charged gels show substantially reduced rejections when compared to positively charged membranes. The lower rejection by the polyacid containing membranes is not only caused by the presence of divalent counter-ions Ca 2+ and Mg 2+ but also by strong specific binding of cations to carboxylic and sulfonic acid groups. This binding has an effect of reducing the effective charge density. The divalent anions SO42- and CO32, the counter-ions for the positively charged membranes, are also present in this feed but their binding to cationic polyelectrolytes is weak and does not significantly affect rejection. Further work is underway to probe these effects more deeply. Table 3. Softening of tap water a (test conditions: untreated municipal tap-water as a feed; stirred dead-end cells; pressure: 100 kPa). Membraneb

Charge

Flux

Rejection (%)

kg/m2h

Na +

Mg 2+

Ca 2+

Poly(vinylbenzyl ammonium) (Gel 2)

+

12

35

79

80

Poly(4-vinylpyridinium) (Gel 1)

+

8

20

74

61

Poly(AMPS) (Gel 6)

-

12

12

22

27

Poly(acrylic acid) (Gel 4)

-

5

10

29

38

7

21

71

54

DESAL-51 (Osmonics)

a For tap water composition see footnote at Table 1; b For structures of the gels see Fig. 3.

3.

CONCLUSIONS In this chapter we have described a new family of NF membranes consisting 372


of a polyelectrolyte gel anchored within the pores ofa microporous host membrane. At the outset of the chapter we pointed out that these Aion-exchange@ membranes have NF performances which match commercial NF membranes and asked the question as to whether their performance could be enhanced. In answering this question attention has been particularly focused on understanding the principles governing the performance of gel-filled membranes. It has been shown that the permeability and flux of these membranes can be calculated with excellent precision using an a-priori model for gel permeability that we have developed. The ability of this model to predict permeability and flux is such that experimentally observed deviations with certain membranes indicate the onset of factors such as gel microheterogeneity or pressure induced compression of the membranes in NF applications. The gel-mesh radii derived from the model can also be used to predict the separation of large neutral solutes. It has also been shown that the separation of ionic solutes is based on Donnan exclusion. Salt rejections can be calculated using the TMS model and a good fit between observed and calculated values are found. Table 4. Some partially optimized NF membranes (feed: municipal tap-water; stirred deadend cells). Membrane

Pressure (kPa)

Flux (kg/m2h)

Na÷

Rejection (%) Mg2+

I

50

18

17

63

61

II

100

35

16

63

61

III

100

38

18

61

64

7

21

71

54

DESAL-51 (Osmonics) 100 a For tap water composition see footnote at Table 1

Based on the principles presented here work is currently underway to fabricate higher-performance NF membranes that will function effectively at ultralow driving pressures. This work is bearing fruit and it is clear that substantial improvements in performance can be achieved. This is evident from the data given in Table 4 and comparing these data with those given in Table 1. The striking features of the various membranes listed in Table 4 are that not only can substantial improvements in performance be made but that it is possible to fabricate membranes capable of softening municipal tap-water with a trans-membrane pressure of only 50 kPa. The polyelectrolyte-filled membranes described in this chapter are not just

373


laboratory curiosities. They are robust membranes despite the intuitive impression given by a membrane construct in which a relatively soft, water-swollen gel is inserted into a microporous support. For example, several membranes have been continuously tested for well over a year in our laboratories with no loss in properties. They are easy to clean, and potentially both simple and cheap to fabricate on a continuous basis. They even come with a Abuilt-in@ self-repairing capability associated with the soft-gel separating component. The ability of the membranes to achieve comparable fluxes and separations but operate at lower pressures offers the prospect of significant reduction in membrane fouling and lower energy consumptions. This work clearly demonstrates that consideration be given to gel-filled membranes as an alternative to the conventional thin-film approaches to high performance membranes. ACKNOWLEDGMENTS The work described in this chapter has been ably carried out by a variety of people at McMaster University. These include Drs A. K. Pandey, H. Byun, graduate students W Jiang, C. McCrory, J. Zhou, and M. Kim, and technical support ofE. H. M. Lee. Financial support from the 3M Canada Corporation, the Natural Science and Engineering Research Council of Canada is gratefully acknowledged. Membrane supports were donated by 3M Canada Corporation and DSM Solupor.

REFERENCES

[1] [2] [3] [4] [5] [6] [7] [8] [9]

[10] [11] [12] [13] [14]

R.W. Baker, Membrane Technology and Applications, McGraw-Hill, New York, 2000. R.F. Childs, Membrane Quarterly, 16(2) (2001) 7. A.M. Mika and R.F. Childs, In: Membrane Separations, A. Noworyta and A. TrusekHolownia (Eds.), Agencja Wydawnicza "Argi", Wroclaw, 2001, p.23. S.-Y. Kwak, C.K. Kim, and J.J. KJm, J.Polym.Sci.:B.Polym.Phys. 34 (1996) 2201. S.-Y. Kwak, Polymer 40 (1999) 6361. S.-Y. Kwak and D.W. Ihm, J.Membr. Sci. 158 (1999) 143. S.-Y. Kwak, S.G. Jung, Y.S. Yoon, and D.W. Ihm, J.Polym.Sci.:B.Polym.Phys. 37 (1999) 1429. W.R. Bowen and H. Mukhtar, J.Membr. Sci. 112 (1996) 263. W.R. Bowen, A.W. Mohammad, and N. Hilal, J.Membr. Sci. 126 (1997) 91. W.R. Bowen and T.A. Doneva, J.Colloid Interface Sci. 229 (2000) 544. W.R. Bowen and T.A. Doneva, Desalination 129 (2000) 163. P. Fievet, A. Szymczyk, B. Aoubiza, and J. Pagetti, J.Membr. Sci. 168 (2000) 87. A.M. Mika, R.F. Childs, and J.M. Dickson, Desalination 121 (1999) 149. A.M. Mika, A.K. Pandey, and R.F. Childs, Desalination, 140 (2001) 265.

374


[15] [16] [17]

[18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29]

[3o] [31] [32] [33] [34] [35] [36] [37] [38] [39]

[40] [41] [42] [431 [44] [45] [46] [47]

[48] [49]

[50]

G. Schulz, Desalination 68 (1988) 191. F. Nakashio, M. Goto, and T. Kakoi, Solvent Extr.Res.Dev.,Jpn. 1 (1994) 53. F. de Jong and H.C. Visser, Comp.Supramol.Chem. 10 (1996) 13. M. de Gyves and E.R. de Miguel, Ind.Eng.Chem.Res. 38 (1999) 2182. A.J.B. Kemperman, D. Bargeman, Th. van den Boomgaard, and H. Strathmann, Sep.Sci.Technol. 31 (1996) 2733. M.C. Wijers, M. Jin, M. Wessling, and H. Strathmann, J.Membr. Sci. 147 (1998) 117. Y. Wang and F.M. Doyle, J.Membr. Sci. 159 (1999) 167. X.J. Yang, A.G. Fane, J. Bi, and H.J. Griesser, J.Membr. Sci. 168 (2000) 29. M. Ulbricht, React.Funct.Polym. 31 (1996) 165. T. Yamaguchi, S. Nakao, and S. Kimura, Macromolecules 24 (1991) 5522. T. Yamaguchi, S. Nakao, and S. Kimura, Ind.Eng.Chem.Res. 32 (1993) 848. A.M. Mika, R.F. Childs, J.M. Dickson, B.E. McCarry, and D.R. Gagnon, J.Membr. Sci. 108 (1995) 37. V. Kapur, J. Charkoudian, S.B. Kessler, and J.L. Anderson, Ind.Eng.Chem.Res. 35 (1996) 3179. A.M. Mika, R.F. Childs, J.M. Dickson, B.E. McCarry, and D.R. Gagnon, J.Membr. Sci. 135 (1997) 81. R.F. Childs, A. Pandey, B.E. McCarry, J.M. Dickson, M. West, and J.N.A. Lott, J.Polym.Sci.A:Polym.Chem. 39 (2001) 807. A.M. Mika, R.F. Childs, and J.M. Dickson, PCT/CA97/00770 (WO 98/17377), 1998, Canada. G.E. Gillberg-LaForce and E.M. Gabriel, U.S. Pat. 5,049,275, 1991, U.S.A. A.M. Mika and R.F. Childs, Ind.Eng.Chem.Res. 40 (2001) 1694. D.W. Schaefer, Polymer 25 (1984) 387. J. Happel, AIChE J. 4 (1958) 197. J. Happel, AIChE J. 5 (1959) 174. J.E. Drummond and M.I. Tahir, Int.J.Multiphase Flow 10 (1984) 515. C. McCrory, Impact of Gel Morphology on Pore-Filled Membranes, Ph.D. Thesis, McMaster University, Hamilton, Canada, 2001. A.M. Mika, R.F. Childs, and J.M. Dickson, J.Membr. Sci., In press. S. Singh, K.C. Khulbe, T. Matsuura, and P. Ramamurthy, J.Membr. Sci. 142 (1998) 111. T. Teorell, Progr.Biophysics Biophys.Chem. 3 (1953) 305. K.H. Meyer and J.F. Sievers, Helv.Chim.Acta 19 (1936) 649. K.H. Meyer and J.F. Sievers, Helv.Chim.Acta 19 (1936) 665. K.H. Meyer and J.F. Sievers, Helv.Chim.Acta 19 (1936) 987. E. Hoffer and O. Kedem, Desalination 2 (1967) 25. T. Tsuru, S. Nakao, and S. Kimura, J.Chem.Eng.Japan 24 (1991) 511. F.A. Morrison, Jr. and J.F. Osterle, J.Chem.Phys. 43 (1965) 2111. R.J. Gross and J.F. Osterle, J.Chem.Phys. 49 (1968) 228. J.C. Fair and J.F. Osterle, J.Chem.Phys. 54 (1971) 3307. G.S. Manning, Acc.Chem.Res. 12 (1979) 443. P. Molyneux, In: Water Soluble Synthetic Polymers: Properties and Behavior, CRC Press, Boca Raton, Florida, 1983, p.1.

375


[51]

[52]

R.F. Childs, A.M. Mika, A.K. Pandey, C. McCrory, S. Mouton, and J.M. Dickson, Sep.Purif.Technol. 23 (2001) 507. A.M. Mika and R.F. Childs, J.Membr. Sci. 152 (1999) 129.

376

Sensors

377


New Insights into Membrane Science and Technology:Polymeric and Biofunctional Membranes D. Bhattachatyyaand D.A. Butterfield (Editors) © 2003 Elsevier Science B.V. All rights reserved.

Chapter 18

Membranes for the development of biosensors V. G. Gavalas, J. Wang, L. G. Bachas Department of Chemistry and Center of Membrane Science, University of Kentucky, Lexington, Kentucky 40506-0055 1.

INTRODUCTION

According to the Intemational Union of Pure and Applied Chemistry (IUPAC), "a biosensor is a self-contained integrated device which is capable of providing quantitive or semi-quantitive analytical information using a biological recognition element (biochemical receptor) which is in direct contact with a transducer element" [1 ]. The biological reaction takes place in the interface of the sensor with the sample. The recognition of the analyte is based on its interaction with the biological element. This interaction can be either affinity type or of biocatalytic nature (Fig. 1). The efficient transduction of the biological recognition to analytical signal is crucial for the biosensor's successful operation. The selection of the transducer depends mainly on the type of the biological recognition and the changes that it induces on the system.

Biological Recognition

Affinity Antibodies ,

,,

,,

,

Catalysis Enzyme cofactors

RIgA Receptors ,

,

,

,

,,,

Cells Microorganisms ,

,

,

,

Transduction of biological recognition to signal

Electrochemical llllll

i

ii

i

Thermal i

i

i

Optical i

i

Figure 1. Componentsofbiosensors

379

Mass i

i

i

i i

i

i

i

,,

,,,

,

Membranes For The Development Of Biosensors

2.

-

Bachas

M E M B R A N E S AS COMPONENTS OF BIOSENSORS

Membranes play a key role in the design and development of biosensors. They have been used extensively for the immobilization of biomolecules and/or cofactors, to improve or even alter the selectivity of the biosensor, to control the enzyme kinetics, to improve the biocompatibility of the system, and to control the electron transfer properties of the system. In most cases, the membranes perform more than one of the above functions. These functions are discussed separately in the following sections. 2.1

Membranes for the Immobilization of Biomolecules The biological recognition element in a biosensor is usually immobilized in a layer directly adjacent to or on the surface of the transducer. In most cases, a membrane is used for the immobilization, whether it is a simple physical entrapment or adsorption or more elaborate covalent attachment of the biological recognition elements. In electrochemically mediated immobilization, biomolecules get entrapped within membranes formed by the electropolymerization of appropriate monomers. The final polymeric membrane may be conductive or not, depending on the monomer employed. This is a fast, one-step immobilization procedure, which can be applied essentially to transducers of any shape and size. The thickness of the membrane can be precisely controlled by the charge applied to the deposition of the film. Further details on the electrochemical immobilization of biomolecules can be found in two recently published reviews [2, 3]. For covalent immobilization, enzymes or other recognition elements are tethered to the membranes through chemical bonding between the functional groups of the enzyme and those on the surface of the membrane. However, if more than one such functional groups on the enzyme are available, attachment of the enzyme to the membrane surface through the various groups will inevitably lead to different orientations of the enzyme. The active site of some of the enzyme molecules can become partially or completely blocked by the membrane surface (Fig. 2A). This random immobilization results in reduced enzyme activity due to the limited accessibility of the active site of the enzyme to the sample. The same argument can be made for orienting the binding site of receptors and binding proteins. To minimize activity loss due to random immobilization, the enzyme can be immobilized in an orientation-specific fashion such that the active site points to the same direction that is facing away from the immobilization surface (Fig. 2B). This can be achieved by choosing a unique residue of the enzyme (eg, the SH of a cysteine) through which the protein becomes attached to the membrane. However, choosing such a unique residue may be difficult because (a) most

380

Membranes For The Development O f Biosensors - Bachas

A

Figure 2. Random (A) vs. site-specific (B) protein immobilization. Arrows point to the binding site of the protein.

proteins lack such residues, (b) even if such a residue is present, it may be inaccessible, or (c) the location of the residue may not be ideal for orientation specific immobilization. Recent advancements in molecular biology and other interdisciplinary research efforts have lead to new strategies for orientation specific immobilization of proteins. These strategies include introducing a unique reactive residue into a protein by site-directed mutagenesis, and attaching an affinity tag or a recognition sequence for post-translational modification by gene fusion technology. For example, Huang et al. in our laboratory utilized site-directed mutagenesis to introduce a single cysteine residue to the cysteine-free subtilisin, and this mutant subtilisin was site-specifically immobilized to agarose beads and membranes through the -SH group of the introduced cysteine [4, 5]. Their studies showed that, besides improved stability, an enhancement of activity was observed for the site-specifically immobilized enzyme. Affinity interactions between biomolecules and their ligands have been successfully used in affinity purification of proteins/cells. For example, protein A and protein G have high affinity toward the constant region (Fc) of immonoglobulins (IgG), and therefore, have been used to construct fusion proteins to facilitate purification of other proteins [6, 7]. In essence, the nucleotide sequence coding for the protein of interest is fused to the gene for protein A or protein G. The fused gene is expressed in an appropriate microorganism to get the fusion protein. The fusion protein is readily purified by passing through an affinity column consisting of immobilized IgG. The protein of interest can then be cleaved off the fusion protein and obtained in a purified form for further characterization or application. Other affinity interactions that have been utilized for protein purification include:

381


a. Interactions between binding proteins/domains and their ligand: biotin(strept)avidin [8], cellulose binding domain - cellulose [9, 10], maltose binding domain - amylose [ 11 ], calmodulin - phenothiazine [ 12]. b. Enzymes and their substrates: glutathione-S-transferase - glutathione [ 13], [3galactosidase - p-aminophenyl-13-D-thiogalactoside [ 14]. c. Interactions between affinity tags and their ligands: polyhistidine tag immobilized metal ions (Co 2+, Ni2+), Strep-tag - streptavidin [15, 16], FLAG -antiFLAG antibody [ 17, 18]. d. Others. For example, polyarginine - negatively charged mica surfaces [ 19]. These affinity interactions can be used to help site-specifically immobilize proteins [17, 19]. Such site-specific immobilization enables the active site of the protein to be fully accessible to its substrate even after immobilization. Therefore, high enzyme activity is retained. Site-specific protein immobilization should be advantageous when applied to the design and development of biosensors using immobilized enzymes and other biomolecules. 2.2

Selectivity Control Through Membranes

Interferences in the signal of biosensors can be introduced through two different mechanisms, the biological recognition and the transduction. The first case occurs when the biological element is class-selective (i.e., alcohol oxidase, tyrosinase, etc.) or it can catalyze the reaction of various compounds (i.e., if the biological element is a cell or a tissue). The control of the selectivity in this case is rather difficult since a multitude of compounds can induce a response. When the interference occurs through the transduction pathway, membranes can play a key role in avoiding or minimizing the interference. The most pronounced example of employing membranes to eliminate interference can be found in amperometric biosensors. Since several of the amperometric biosensors are based on the detection of enzymatically produced H202, which requires high potentials (>_ +0.6 V vs Ag/AgC1), electroactive species present in the sample can be oxidized on the electrode surface prohibiting accurate measurements. Even when the applied potential has been lowered by the use of suitable mediators, the signal is still affected since the mediator can promote the oxidation of interferents even at potentials lower than 0.3 V [20-23]. Thus, the use of membranes has been extensively employed to avoid interference from electroactive species such as ascorbic acid, uric acid and acetaminophen [24] by employing one of the following selective permeation mechanisms. Permselective membranes can be based either on size exclusion (cellulose acetate, polyaniline, polypyrrole) or on charge exclusion (Nation, poly(vinylpyridine), poly(ester-sulfonic acid)). The polymeric films used for this purpose are generally solvent-cast or electropolymerized. In general, most of the solvent-cast films are negatively charged and therefore exhibit good rejection only for the negatively charged ascorbic acid 382


and uric acid, but not for the neutral acetaminophen. To overcome this drawback, solvent-cast films based on composite membranes have been employed. Zang et al. [25] used a composite membrane of cellulose acetate and Nation to completely eliminate acetaminophen interference, but the sensor also lost 80% of the H202 sensitivity. Benmakroha et al. [26] demonstrated that by mixing plasticized poly(vinyl chloride) and sulfonated polyether-ether sulfone in suitable proportions, the permeability of hydrogen peroxide over acetaminophen was controlled. Recently, Mizutani et al. [27] used poly-L-lysine and poly(4styrenesulfonate) to form a polyion complex membrane on-top of a gold electrode to prevent electrochemical interferences. The level of selectivity that they obtained was sufficient to allow accurate detection of glucose in blood and beverages. With solvent-cast films it can be difficult to obtain uniform thickness coatings, especially on non-smooth surfaces. In contrast, the thickness of the electropolymerized films can be controlled, and the geometry of the electrode surface does not influence the polymer growth. There are many different monomers that have been used for the fabrication of electropolymerized membranes [2, 3]. Significant impact had the report from Centonze et al. [28] on the use of overoxidized polypyrrole films for the development of interference-free biosensors. The overoxidation of polypyrrole produces a permselective, antifouling film that rejects proteins in addition to the electroactive interferents. It is generally believed that oxygen-containing groups are introduced as a result of overoxidation [29, 30]. Although the detailed identification of these groups has not yet been achieved, these oxygencontaining groups include carbonyl, hydroxyl and carboxyl groups, and they are considered to be responsible for the improved selectivity. Murphy [31] studied a series of electropolymerized films based on derivatives of diaminonaphthalene (DAN) and compared their selectivity and stability with electropolymerized films prepared from o-, m-, or pphenylenediamine. His experiments showed that the positioning of the amino groups greatly influences the selectivity and stability of the sensor. The best results were obtained when 2,3-DAN, 1,5-DAN or 5-amino-l-napthol were employed. 2.3

Controlling enzyme kinetics

The kinetics of many enzymes can be described with the model that Leonor Michaelis and Maud Menten have proposed [32]" E+S

~

ES

It,. AL~

~ r

p

k_l

383

Membranes For The Development O f Biosensors - Baehas

The enzyme (E) reacts with the substrate (S) to form a complex (ES) with rate constant kl. The complex ES can either dissociate into E and S (with rate constant k_l) or form the product P (with rate constant k2). Introducing the Michaelis constant (KM) as the ratio (k.1 + k2)/kl and assuming that (a) the ES complex is in steady state, (b) under saturating conditions all the enzyme is converted to ES complex, and (c) the maximum velocity (Vmax) is achieved when all the active centers of all the enzyme molecules are saturated with substrate, then the velocity (v) is defined by the equation: V = Wmax

IS] [s] + KM

(1)

From this non-linear equation it is clear that the velocity, v, will have a linear relationship with the substrate concentration, if this concentration is low relative to KM (i.e. when KM >_0.1 IS]). The KM of most enzymes is in the low mM range (i.e., glucose oxidase has a KM value of 4.2 mM for glucose, L-lactate oxidase a KM of 1.4 mM for lactic acid, L-lactate dehydrogenase a KM of 1.6 mM for lactic acid, malate dehydrogenase a KM of 0.2 mM for maleic acid, etc.), while the concentration of the corresponding analytes in real samples is much higher (glucose up to 500 mM in food samples,-~5 mM in blood; lactate can exceed 7 mM in blood; malate concentrations are greater than 5 mM in food). Thus, in order to increase the linear range of response, membranes that limit the substrate diffusion can be used. Numerous models have been reported that describe the effect of diffusion processes in the response of enzyme biosensors of various configurations [3337]. Cambiaso et al. [38] calculated that decreasing the substrate diffusion coefficient from 1.1xl0 9 cm2/s to 5x10 -1° cm2/s, the KM is increased by 50%. Schulmeister and Pfeiffer [39] developed a model to describe the effect of using a semipermable membrane of variable thickness on the dynamic behavior of a glucose biosensor. In Fig. 3 the correlation of the linear range of response of the glucose biosensor with the thickness of the additional perforated membrane is presented. As it can be seen, by using a thicker membrane the linear range of response is extended, and this extension is obtained because the enzymatic process is controlled by the diffusion of the enzyme's substrate. However, by limiting the diffusion of the substrate, the sensitivity of the biosensor is decreased as well. Schulmeister and Pfeiffer [39] calculated that by increasing the membrane thickness from 5 to 40 ~tm the sensitivity of the biosensor is reduced by 85%. Thus, there is a trade off between linear range and sensitivity, which has to be considered when designing a biosensor.

384

Membranes For The Development Of Biosensors - Bachas

50 E "E m ~t ,m .--I

40

._. 30 u

~: w 20

7 o

~

;o

~

~'o

~

~o

~

4o

,~

Membrane Thickness (pm)

Figure 3. Correlation of the effective linearity of a glucose biosensor with the thickness of the semipermable membrane. Data from Schulmeister and Pfeiffer [39] Various membranes have been used as extemal diffusion control layers, including cellulose acetate/triacetate, polycarbonate, polyurethane, Nation, polyvinylchloride (PVC), etc. Yang et al. [40] used a PVC coating to increase the upper range of linear response of glucose and lactate biosensors, from 1 to 24 mM and from 2 to 16 mM, respectively. The sensitivities of the PVC-coated sensors were similar, despite the significant difference in the activities of the enzymes. Upon the use of the PVC membrane, the contribution of the enzyme activity to the sensor's signal is minimized, since the membrane provides sufficient diffusion limitation for both substrates. In a recent systematic study, Maines et al. [41] evaluated a series of modified membranes on the basis of physical and diffusion properties, by calculating the diffusion coefficients of the enzyme's substrate and various interferents. Further modification of the membranes allowed for the optimization of additional parameters (i.e., stability and selectivity) other than the linear range of response.

2.4

Membranes for Improving the Biocompatibility of Sensors In vivo monitoring of various analytes is important for many bioanalytical

and biomedical applications. The crucial challenge in this type of application is the interactions of the sensor with the host environment, which are qualitatively described by the term biocompatibility. The real issue of sensor biocompatibility should not only be whether there are adverse biological reactions to the sensor, 385


but whether the sensor has any negative effect on the body. The term biocompatible should suggest that the material described exhibits harmonious behavior when in contact with tissue and body fluids [42]. However, biocompatibility has been used occasionally in bioanalytical chemistry to describe the effect of the body on the sensor. Insertion of a biosensor in the body of humans or animals may stimulate various "foreign body response" processes, which can lead to the failure of the sensor. The most important reactions that follow the contact of a sensor with blood are protein adsorption, platelet adhesion, platelet aggregation, fibrin formation and thrombus formation [43]. Biodegradation of the membrane, breakdown/denaturation of the biological recognition element (e.g., the enzyme) or other components of the biosensor could also lead to its malfunction. The most common failing reasons of sensors in vivo are presented in Fig. 4.

Membrane biodegradation

Biofouling

Degradation of recognition element

Fibrous encapsulation

Figure 4. Schematic illustration of failing reasons ofbiosensor in vivo.

Several recent reviews discuss methods and materials used for the improvement of sensor biocompatibility [44-46]. One of the most important goals in the design of biocompatible membranes is to develop materials that resist biofouling. To avoid biofouling, the sensor's surfaces have been treated with anticoagulants (i.e., heparin) or surfactants. Brooks et al. [47] have covalently attached heparin on the surface of cellulose triacetate membranes without any significant alteration on the response characteristics of the sensor (potassium ISE based on valinomycin). The cellulose triacetate membranes, which were coated with heparin, inactivate blood coagulation factor Xa, increasing the blood compatibility of the sensor. Reddy and Vadgama [48] compared the performance of a series of commercial membranes with PVC membranes that have been modified with nonionic (Tween 82 or Triton X-100), anionic (bis(2-ethylhexyl)hydrogenphosphate) or cationic surfactants (tricaprylylmethyl-ammonium chloride). The cationic surfactant-modified PVC membrane exhibited the worst biocompatibility. The other modified PVC membranes had improved biocompatibility compared to commercial porous

386

Membranes For The DevelopmentOf Biosensors- Bachas

polycarbonate and Cuprophan hemodialysis membranes. However, a lactate biosensor lost between 35 to 80% of its response upon 6 h exposure to blood, regardless the type of membrane used. Berrocal et al. [49] introduced the use of Biospan-S, a new bloodcompatible polymeric material (Fig. 5A). Biospan-S is segmented polyurethane modified with poly(dimethylsiloxane) and exhibits higher in vivo stability than the unmodified segmented polyurethane [50]. The authors prepared three different types of cation-selective electrodes, which demonstrated comparable analytical characteristics to conventional PVC-based electrodes, verifying that the Biospan-S can be used as a polymeric matrix for the development of bloodcompatible sensors. Recently, new polymeric materials that mimic the phospholipid groups naturally occurring in membranes of biological cells have been used (Fig. 5B). Berrocal et al. [51] used poly(2-methacryloyloxyethyl phosphorylcholine-cobutylmethacrylate), a biocompatible hydrogel containing phosphorylcholine groups, as a coating on the membrane surface of ion-selective electrodes to improve their compatibility. They demonstrated that the hydrogel is compatible with several membrane materials, introducing slight alterations in the selectivity but leaving unaffected the sensitivity and detection limit of the sensors. In addition, immunostaining experiments verified that there was limited platelet adhesion to the coated PVC membranes. This hydrogel has also been used to improve the biocompatibility of a glucose biosensor [52].

CHa-Si-Ol"-'Si-OcH3 LCH3 Jp

(A)

"

'

-('CH,.CH,CH,CH,O),.,.,

-0-"

hI'-o-"

o-.-s,-lo-s,-ro-s,-c.

"

z ,

CH3 (B)

c. 3[

ca3Jp

Ca3

CH3

( I--C2 )a ( ? O-'-C O O~ I , + I O C H 2CH20~OCH 2CH2N(CH3)3 ocn

2 c n 2CH 2 c n 3

O"

Figure 5. Structure of (A) Biospan-S and (B) poly(2-methacryloyloxyethyl phosphorylcholine-co-butylmethacrylate) or poly(MPC-co-BMA) 387


Another approach is the use of materials that continuously release biological active molecules such as nitric oxide to prevent platelet adhesion and activation. Schoenfisch et al. [53] prepared a silicone rubber membrane containing NO-generating diazeniumdiolate compounds and applied it on the outer surface of an oxygen catheter. The NO-generating diazeniumdiolate compounds, when in contact with water, release NO for a substantial period of time, without altering the analytical characteristics of the sensor. The performance of the membrane-covered sensor was evaluated in vivo in dogs with very promising results. Despite the extensive research, the issue of sensor biocompatibility remains a significant challenge. More fundamental research on the sensor-tissue interactions will be needed to fully understand and eventually overcome the problems related to membrane biofouling. 2.5

Membranes for Electron Transfer Control

The function of amperometric biosensors is based on the electron transfer from the active site of the immobilized enzymes to the electrode surface. The enzyme during the catalysis stores the electrons either in a prosthetic group (i.e., FAD, PQQ, heme, etc.) or in a co-substrate (i.e., NAD(P)+). The amperometric biosensor's signal is based on the transfer of these electrons from the intermediate storage group to the electrode surface [54]. It is, thus, evident that the analytical characteristics of a biosensor can be optimized through the control of the electron-transfer process. Membranes that contain redox active centers, known as "redox polymers", have been extensively used for the development ofbiosensors [54, 55]. In Fig. 6 the electron pathway in a redox polymer-based biosensor is shown. The electrons, through an electron-hopping mechanism, are transferred from the enzyme to the electrode surface. One of the most successful examples of redox polymer is the osmium based poly(vinyl imidazole) polymer introduced by Heller's group [56, 57]. Glucose oxidase has been "wired" (the enzyme was connected to the electrode through cross-linked electron-conducting redox hydrogel) with polyelectrolytes having electron relaying redox centers [Os(bpy)2C1]+/2+ in their backbones. Hydrogels were formed upon cross-linking the enzyme with the polymer. The detection was achieved by applying +0.3 V potential to the electrode. Osmiumbased polymers have been used to promote electron transfer from other enzymes as well; e.g., horseradish peroxidase [58-60], lactate oxidase [61, 62], glutathione sulfhydryl oxidase [63]. Swann et al. [64] reported on the use of the conductive polypyrrole polymer as the electron transfer pathway. They observed electron transfer from the PQQ of fructose dehydrogenase to the electrode, which was affected by the polypyrrole growth conditions. More widespread is the use of redox-modified 388


monomers for the growth of polymeric film. There have been reports on osmium-complex-modified pyrrole [65-67], ferrocene-modified pyrrole [68], or aniline [69-71 ]. The use of such redox-polymers has been attractive since it involves a one-step preparation procedure that enables the control of the polymer thickness even when the electrode surface is rough, small sized, or highly porous.

Figure 6. Electron-hopping mechanism in a redox polymer-based biosensor (-- polymer backbone, • redox active center). Arrows denote possible paths of electron transfer. 3.

CONCLUSIONS AND FUTURE CONSIDERATIONS

Membranes play an indispensable role in the development of biosensors, not only because most electrodes surfaces themselves are membranes, but also because more and more polymeric membranes are developed/functionalized to improve the selectivity, sensitivity, and operational range of biosensors. Since sensor performance characteristics are largely determined by the biological recognition element of a biosensor, increasing efforts are devoted on how to gain a well-controlled and reproducible immobilization of those biomolecules. Though relatively simple, entrapment of biomolecules into polymeric films through electrochemical polymerization has the drawbacks of reduced biological activity. Site-specific immobilization, in which all the active sites of the immobilized biomolecules point to the direction toward the solution, provides a new way to a well-controlled sensing layer with high biological activity. The

389


drawback of the initial investment in modifying the biomolecules either chemically or genetically is outweighed by the gain of highly active immobilized biomolecules on the sensor surface. The need for smaller biosensors with faster response as well as array sensors for multi-analyte sensing makes the preparation of the sensor surfaces more challenging. Recent advancement in micro-patterning of biomolecules and cells may provide useful tools to gain well-defined immobilization of biomolecules with superior activity at a molecular scale. For example, self-assembled monolayers (SAMs) using microcontact printing allow patterned formation of (bio)molecules and cells on gold and silicon-based membrane surfaces [72, 73]. This technique has been recently extended to include microcontact printing on polymeric materials such as poly(ethylene teraphthalate) [74]. These techniques can be used to mass produce sensor surfaces in the micrometer scale with delicate biomolecules at a reasonable cost. Polymeric microcapsule arrays have been synthesized by Martin and coworkers, and subsequently used as cavities to immobilize enzymes [75, 76]. Recently, the same group has successfully produced membranes formed by orderly aligning nanotubules modified with HS-R, which can selectively transport hydrophobic/hydrophilic species depending on the nature of the R group [77, 78]. These nanotubules should be suitable for the immobilization of small enzymes and other biomolecules and lead to the development of biosensors at the nanoscale.

REFERENCES [1] [2] [3]

[4] [5] [6]

[7] [8] [9] [10] [ll] [12] [13] [14]

D.R. Thevenot, K. Toth, R.A. Durst, G.S Wilson, Biosens. Bioelectron. 16 (2001) 121. S. Cosnier, Biosens. Bioelectron. 14 (1999) 443. F. Palmisano, P.G. Zambonin, D. Centonze, Fresenius J. Anal. Chem. 366 (2000) 586. W. Huang, J. Wang, D. Bhattacharyya, L. G. Bachas, Anal. Chem. 69 (1997) 4601. S. Viswanath, J. Wang, L. G. Bachas, D. A. Butterfield, D. Bhattacharyya, Biotechnol. Bioeng. 60 (1998) 608. P.A. Nygren, M. Eliasson, L. Abrahmsen, M. Uhlen, E. Palmcrantz, J. Mol. Recognit. 1 (1988) 69. B. Nilsson, L. Abrahmsen, M. Uhlen, EMBO J. 4 (1985) 1075. D. A. Clare, V. W. Valentine, G. L. Catignani, H. E. Swaisgood, Enzyme Microb. Technol. 28 (2001) 483. A. A. Wang, A. Mulchandani, W. Chen, Biotechnol. Prog. 17 (2001) 407. I. Levy, O. Shoseyov, J. Pept. Sci. 7 (2001) 50. E. Kobatake, Y. Ikariyama, M. Aizawa, J. Biotechnol. 38 (1995) 263. J. C. Lewis, J. J. Lopez-Moya, S. Daunert, Bioconjug. Chem. 11 (2000) 65. H. Youssoufian Biotechniques 24 (1998) 198. R. Chinery, R. Roulsom, G. Elia, A. M. Hanby, N. A. Wright, Eur. J. Biochem. 212 (1993) 557.

390


[151 [16] [17]

[18] [191 [201 [21] [22] [23] [241 [25] [26] [27]

[28] [29] [30] [31] [32] [33] [34]

[35] [36] [37]

[38] [39]

[40] [41] [42] [43] [44] [45] [46] [47]

[48] [49]

W. H. Shao, X. E. Zhang, H. Liu, Z. P. Zhang, A. E. Cass, Bioconjug. Chem. 11 (2000) 822. K. J. Gregory, L. G. Bachas, Anal. Bioehem. 289 (2001) 82. J. Wang, D. Bhattacharyya, L. G. Baehas, Biomacromolecules 2 (2001) 700. J. Liu, J. Wang, L. G. Baehas, D. Bhattacharyya, Biotechnol. Prog. 17 (2001) 866. S. Noek, J. A. Spudich, P. Wagner, FEBS Lett. 414 (1997) 233. P.C. Pandey, S. Upadhyay, H.C. Pathak, C.M.D. Pandey, Anal. Lett. 31 (1998) 2327. A. P. Doherty, M. A. Stanley, J. G. Vos, Analyst 120 (1995) 2371. J. R. Woodward, A. Brunsman, T.D. Gibson, S. Parker, Biosensors for Food Analysis, A.O. Scott (Ed.), R. Soc. Chem., London (1998), 71. J. Kulys; W. Schuhmann; H.L. Schmidt, Anal. Lett. 25 (1992) 1011. J. Wang, Biosensors & chemical sensors: optimizing performance through polymeric materials, P.G. Edelman, J. Wang (Ed.), ACS, Washington DC, 1992, 125. Y. Zang, Y. Hu, G. Wilson, D. Moatti-Sirat, V. Poitout, G. Reach, Anal. Chem. 66 (1994) 1183. Y. Benmakroha, I. Christie, M. Desai, P. Vadgama, Analyst 121 (1996) 521. F. Mizutani, Y. Sato, Y. Hirata, S. Yabuki, Biosens. Bioelectron. 13 (1998) 809. D. Centonze, A. Guerrieri, C. Malitesta, F. Palmisano, P.G. Zambonin, Fresenius J. Anal. Chem. 342 (1992) 729. F. Palmisano, C. Malitesta, D. Centonze, P.G. Zambonin, Anal. Chem. 67 (1995) 2207. Z. Chen, Y. Takei, B.A. Deore, T. Nagaoka, Analyst 125 (2000) 2249. L.J. Murphy, Anal. Chem. 70 (1998) 2928. T. Keleti, Basic Enzyme Kinetic, Akad~miai Kiad6, 1986 W.J. Albery, P.N. Bartlett, J. Electroanal. Chem. 194 (1985) 211. P.N. Bartlett, K.F.E. Pratt, Biosens. Bioelectron. 8 (1993) 451. M.E.G. Lyons, C.H. Lyons, C. Fitzgerald, P.N. Bartlett, J. Electroanal. Chem. 365 (1995) 29. E.A.H. Hall, J.J. Gooding, C.E. Hall, Microchim. Acta 121 (1995) 119. P.N. Bartlett, K.F.E. Pratt, J. Electroanal. Chem. 397 (1995) 61. A. Cambiaso, L. Delfino, M. Grattarola, G. Verreschi, D. Ashworth, A. Maines, P. Vadgama, Sens. Actuators B 33 (1996) 203. T. Schulmeister, D. Pfeiffer, Biosens. Bioelectron. 8 (1993) 75. Q. Yang, P. Atanasov, E. Wilkins, Electroanalysis 10 (1998) 752. A. Maines, M.I. Prodromidis, S.M. Tzouwara-Karayanni, M.I. Karayannis, D. Ashworth, P. Vadgama, Anal. Chim. Acta 408 (2000) 217. J. Black, Biological performance of materials: fundamentals of biocompatibility, Marcel Dekker, New York, 1999 G. Harsfinyi, Sensors in biomedical applications. Fundamentals, technology & applications. Technomic Publishing Company Inc., Pennsylvania, 2000. N. Wisniewski, F. Moussy, W.M. Reichert, Fresenious J. Anal. Chem. 366 (2000) 611. G. S. Wilson, Y. Hu, Chem. Rev. 100 (2000) 2693. N. Wisniewski, W.M. Reichert, Col. & Surf. B 18 (2000) 197. K.A. Brooks, J.R. Allen, P.W. Feldholff, L.G. Bachas, Anal. Chem. 68 (1996) 1439. S. M Reddy, P. M. Vadgama, Anal. Chim. Acta 350 (1997), 77. M.J. Berrocal, I.H.A. Badr, D. Gao, L.G. Bachas, Anal. Chem. 73 (2001) 5328.

391


[50] [51] [52]

[53] [54]

[55] [56]

[57] [58] [59]

[60] [61] [62]

[63] [64] [65] [66] [67] [68] [69]

[70] [71] [72] [73] [74] [75] [76] [77] [78]

A.B. Mathur, T.O. Collier, W.J. Kao, M. Wiggins, M.A. Shubert, A. Hiltner, J.M. Anderson, J. Biomed. Mater. Res. 36 (1997) 246. M.J. Berrocal, R.D. Johnson, I.H.A. Badr, D. Gao, L.G. Bachas, in preparation. C.-Y. Chen,Y.-C. Su, K. Ishihara, N. Nakabayashi, E. Tamiya, I. Karube, Electroanalysis 5 (1993) 269. M.H. Schoenfisch, K.A. Mowery, N. Baliga, J.A. Wahr, M.E. Meyerhoff, Anal. Chem. 72 (2000) 1119. K. Habermtiller, M. Mosbach, W. Schuhmann, Fresenius J. Anal. Chem. 366 (2000) 560. G.G. Wallace, M. Smyth, H. Zhao, TrAC 18 (1999) 245. E. Cs6regi, C. P. Quinn, D. W. Schmidtke, S.E. Lindquist, M. V. Pishko, L. Ye, I. Katakis, J. A. Hubbell, A. Heller, Anal. Chem. 66 (1994) 3131. E. Cs~regi, D. W. Schmidtke, A. Heller, Anal. Chem. 67 (1995) 1240. N. Larsson, T. Ruzgas, L. Gorton, M. Kokaia, P. Kissinger, E. Cs~regi, Electrochim. Acta 43 (1998) 3541. L. Mao, K. Yamamoto, Anal. Chim. Acta 415 (2000) 143. A. Narvaez, G. Suarez, I. Popescu, I. Katakis, E. Dominguez, Biosens. Bioelectron. 15 (2000) 43. K. Sirkar, A. Revzin, M. V. Pishko, Anal. Chem. 72 (2000) 2930. G. Kenausis, C. Taylor, A. Heller, Polym. Mater. Sci. Eng. 76 (1997) 505. L. Mao, K. Yamamoto, Electroanalysis 12 (2000) 577. M. J. Swann, D. Bloor, T. Haruyama, M. Aizawa, Biosens. Bioelectron. 12 (1997) 1169. W. Schuhmann, C. Kranz, J. Huber, H. Wohlsch~iger, Synth. Met. 61 (1993) 31. K. Habermiiller, A. Ramanavicius, V. Laurinavicius, W. Schuhmann, Electroanalysis 12 (2000) 1383. S. Gaspar, K. HabermOller, E. Cs~regi, W. Schuhmann, Sens. Actuat. B 72 (2001) 63. N. C. Foulds, C. R. Lowe, Anal. Chem. 60 (1988) 2473. A. Mulchandani, C.-L. Wang, Electroanalysis 8 (1996) 414. A. Mulchandani, C.-L. Wang, H. H. Weetall, Anal. Chem. 67 (1995) 94. C.-L. Wang, A. Mulchandani, Anal. Chem. 67 (1995) 1109. M. Mrksich, G.M. Whitesides, TIBTECH 13 (1995) 228. C. Marzolin, A. Terfort, J. Tien, G. M. Whitesides, Thin Solid Film 315 (1998) 9. Z. Yang, A. Chilkoti, Adv. Mater. 12 (2000) 413. R. V. Parthasarathy, C. R. Martin, Nature 369 (1994) 298. Brinda B. Lakshmi, Charles R. Martin, Nature 388 (1997) 758. K. B. Jirage, J. C. Hulteen, C. R. Martin, Anal. Chem. 71 (1999) 4913. C. R. Martin, M. Nishizawa, K. Jirage, and M. Kang, J. Phys. Chem. B. 105 (2001) 1925.

392


Chapter 19

Ion-partitioning membranes as electroactive elements for the development of a novel cation-selective CHEMFET sensor system E. A. Moschou .1, N. A. Chaniotakis* Laboratory of Analytical Chemistry, Department of Chemistry, University of Crete, 71409 Iraklion, Crete, Greece 1Current address" Department of Chemistry, University of Kentucky, Lexington KY 40506-0055 1.

ABSTRACT

The development of a cation-selective sensor based on a new response mechanism is presented in this paper. The sensor is based on a bulk active ionpartitioning membrane as the chemical recognition element while the signal transduction is obtained by a pH sensitive transducer, such as the pH-ISFET. The ion-partitioning membrane is doped with two different ionophores, one selective to the analyte cation and the other to protons. The increase of the analyte cation activity in the sample results in the carrier mediated partition of the analyte into the bulk of the membrane phase. Due to the electroneutrality principle requirements, a simultaneous displacement of protons of equal charge out of the membrane must take place. This proton flux takes place towards both the membrane/sample and membrane/transducer interfaces. The monitoring of the proton flow in or out of the membrane by the pH-sensitive transducer is directly related to the analyte ion activity in the sample. Based on this mechanism, it is shown that there is a large effect of all membrane components, as well as the ionic composition of the sample, on the sensor selectivity, sensitivity and detection limit. It is also shown that the optimization of the primary cartier to the proton ionophore ratio in the membrane and that of the properties of the proton ionophore are vital in obtaining the desired sensor characteristics. It is shown that the use of a plasticizer and a polymer matrix with low dielectric constant favors the cationprotons exchange mechanism stabilizing the observed signal. Finally, the proper adjustment of the sample pH and the use of hydrophilic anionic counteranions in the sample contribute to the improvement of both the sensitivity and the detection limit of the sensor. The analytical characteristics of the CHEMFETs selective to the monovalent potassium ion and the divalent calcium ion with the 393

Ion-Partitioning Membranes As Electroactive Elements For The Development Of A Novel Cation-Selective CHEMFET Sensor System Moschou

optimized ion-partitioning membranes are evaluated by the analysis of blood serum samples. 2.

INTRODUCTION

Polymeric membranes have been extensively used as the chemical recognition element in the design of chemical sensors. They are water immiscible phases of high viscosity and are based on the polymer matrix, which plays the role of the supporting material; and the plasticizer, which serves as the organic solvent of the membrane phase. The chemical recognition of the analyte ion, Iz, is based on the partition of the hydrophilic ion from the aqueous sample into the lipophilic membrane phase. This partition is expressed by the partition coefficient of the ion, ki, and is affected by the dielectric constant of the medium, and the size and charge of the analyte ion [1,2]. The increase of the ion size and the decrease of the ion charge result in the increase of the ion partition in the lipophilic membrane. An experimental partitioning series of the ions according to their lipophilicity in the octanol-water system has been formulated, and is the so-called Hofmeister selectivity series [3]: Cs ÷ > Rb ÷ > K ÷ > NH4÷ > H ÷ > Na + > Li ÷ > Ba 2+ > Sr2÷ > Ca 2+ > Mg 2÷ (1) C104" >

SCN" > I" > NO3- > Br > NO2- > C1- > HCO3

>

5042- > F-

(2)

In order to deviate from the Hofmeister selectivity series and increase the relatively low partition of the hydrophilic ions into the lipophilic membrane phase, the ionophores are added in the polymeric membrane. These ionophores [4,5] are mobile electrically neutral or charged, lipophilic, organic or inorganic substances that can selectively interact [6,7] with the analyte ions through electrostatic interactions, wan der Waals or hydrogen bonding. This interaction facilitates the transport of the analyte ions into the lipophilic polymeric membrane. In addition, the polymeric membranes are also doped with electrically charged, organic substances called lipophilic ionic sites [8,9]. These charged compounds decrease the electrical resistance of the membrane, provide charged sites and labile counterions, and can thus increase the partitioning of the analyte ions in the lipophilic polymeric membrane. The response mechanism of the polymeric membrane as a chemical recognition element is determined by its chemical composition. If the membrane is doped with one ionophore that is selective to the analyte ion of interest, this membrane is called surface active membrane [10]. There is a direct chemical interaction between the ionophore and the analyte ion. This interaction mediates

394


the ion transport from the aqueous sample to the outer phase boundary of the membrane. The ion extraction is restricted to the membrane/aqueous solution Surface Active Polymeric Membrane AqueousSample L

R-

LI + +

I+ A-

LI +

I+ A"

L : Electrically neutral ionophore

L-

R+

L-: Electrically charged ionophore Figure 1. Response mechanism of a surface active membrane; where I+ is the analyte cation; A is the counteranion in the aqueous sample; L and L- stand for the electrically neutral and charged ionophore, respectively; and R, R+ are the lipophilic anionic and cationic sites, respectively. interface due to the electroneutrality principle. Therefore, according to the surface active membrane response mechanism (Figure 1) the analyte ions are extracted into the outer membrane phase boundary to a final concentration, which is related to the activity of the analyte ion in the aqueous sample. The one-carrier-doped surface active polymeric membrane can be further modified with the addition of a second ionophore, to produce the so-called bulk active membrane [10]. This ionophore can be selective to either an ion of opposite charged, or of the same charge, which is usually the proton. The increase of the analyte ion activity in the sample results in the ionophoremediated transport of the ion from the aqueous sample to the polymeric membrane. The charge accumulation alteration due to the primary ion partitioning into the membrane can now be compensated by the extraction of the secondary ion of analogous charge out of the membrane for electroneutrality reasons. For a specific application when the analyte ion is a cation, its partitioning into the polymeric membrane will promote the extraction of protons of equal charge out of the membrane. Due to the ion-exchange of the analyte cation with protons the response mechanism of this kind of bulk active membranes is called ion-exchange mechanism [11,12] (Figure 2). In the case of anionic analytes, protons of opposite charge must be co-extracted into the membrane to fulfill the electroneutrality requirements. Due to the simultaneous 395


Bulk Active Membranes Based the Ion-Exchange Mechanism

Based on an electrically neutral ionophore L /

I+ ~

LI +

L

I+ ~1~ LI +

L

H+

C H + R-

C

H + : = ~ CH

C"

Based on an electrically charged ionophore L/

/

I+ ; = ~ LI H + m : ~ CH

R+

I+ ~

L" C"

LI

L"

H + ~=~ CH + C

I+ ~

LI

H+

LH

Figure 2. Bulk active membranes based on the ion-exchange mechanism. [I÷ is the analyte cation; L, LI+ stand for the free and complexed ionophore, respectively; C, CH + are the unprotonated and protonated ionophore and R+, R the lipophilic ionic sites].

Bulk Active Membranes Based on the Co-Extraction Mechanism

Based on electrically neutral ionophore L /

X:;~ H+~ : ~

LX CH

R+ L C

X••

LX"

L

CH + C

Based on electrically charged ionophore L+ /

X"

LX

H÷

CH +

R"

L+

X" ~

LX

L+

C

H ÷~=~ CH

C"

Figure 3. Bulk active membranes based on the co-extraction mechanism. [X is the analyte anion; L, LX" stand for the free and complexed ionophore, respectively; C-, CH are the unprotonated and protonated ionophore and R-, R+ the lipophilic anionic and cationic sites, respectively].

396


co-extraction of anions and protons in the bulk of the polymeric membrane, these active membranes operate based on the so-called co-extraction mechanism [13,14] (Figure 3). In conclusion, in the case of the surface active membranes, the analyte ions are partitioned into the outer phase boundary of the membrane generating a potential difference [15]. For this reason this type of membranes are used as the chemical recognition element for the construction of potentiometric sensors. In contrast, in bulk active membranes, the ions of interest are partitioned into the bulk of the membrane phase, with the electroneutrality principle being fulfilled due to the simultaneous partition of the secondary ions [ 16]. Therefore, the bulk active membranes cannot be used as the chemical recognition element for the construction of potentiometric sensors, but they are rather used in optical sensors (i.e. the ionophore selective to the secondary ion is a chromoionophore and is changing its optical properties upon the analyte ion/secondary ion partition in the membrane). The most significant example of the optical sensors based on bulk active polymeric membranes are the optodes [ 17,18], while the main representatives of potentiometric sensors based on surface active polymeric membranes are the ion selective electrodes [19,20] (ISEs) and the chemically modified field-effect transistors [21,22] (CHEMFETs). In the ISEs the internal reference electrode (usually a Ag/AgC1 electrode) and the internal filling solution (containing a known activity of the analyte ion) are used for the signal transduction. In the CHEMFETs the reference electrode and the internal filling solution are substituted [23] by a pH sensitive field-effect transistor (pH-ISFET) [24,25]. This design allows for the elimination of the internal filling solution. For this reason CHEMFETs are attractive sensor systems since they can be easily miniaturized [26,27]. The CHEMFETs are thus constructed (Figure 4) ivy the application of an electroactive polymeric membrane onto the gate of the pH-ISFET. When the sensor is immersed in the sample solution, the target ions are partitioned into the outer phase boundary of the polymeric membrane developing a potential difference [28,29]. This potential difference is measured via field-effect of the pH-ISFET generating the sensor signal. The pH-sensitive metal oxide [30,31] gate of the pH-ISFET is thus active only in the case of a pH-sensitive sensor system [32]. Its electrochemical activity is subdued when the polymeric membrane is applied on the pH-ISFET's gate. The metal oxide gate does not participate in the mechanism of the signal generation of the CHEMFET any longer [33,34]. Recently it has been shown in the literature that water molecules diffuse into the bulk of the polymeric membrane and can thus form a thin aqueous layer at the interface between the polymeric membrane and the gate of the pH-ISFET 397


[35,36]. When the sensor is immersed in a sample containing volatile acids or

Reference Electrode

r"!

[[

PolymericMembrane

U

@1

; ~alyte ~aiion

'AIOH~

I S°urce I

= = = I Drain substrate Si

I

T Figure 4. Schematic representation of a CHEMFET based on a polymeric membrane. bases, these compounds can diffuse through the polymeric membrane and dissociate in the thin aqueous layer [37]. This dissociation alters the pH at the gate-membrane aqueous layer, producing a signal transduced [6,38]. This was the first report of the pH-ISFET's gate playing an active role on the signal generation of the CHEMFET even though the response to these species was considered as interference. This interference has been overcome by the introduction of a pH-buffered layer between the pH-ISFET's gate and the polymeric membrane [39,40]. This way the gate of the pH-ISFET is in contact with a constant pH environment, and is thus passivated. In this paper we present the development of a cationic electrochemical sensor based on a bulk active ion-partitioning membrane as the chemical recognition element while the signal transduction is obtained by a pH-ISFET. The ion-partitioning membrane is doped with the ionophore selective to the analyte cation as well as an ionophore selective to the proton as the secondary ion. Due to the chemical composition of the polymeric membrane, the ionexchange of the analyte cations with protons takes place in the bulk of the membrane. The displaced protons exit the membrane from both the outer and the inner phase boundaries towards the aqueous sample and the pH-ISFET's gate, respectively. The proton transducer monitors the membrane proton flux generating the CHEMFET's signal [41,42]. Therefore, the direct measurement of the membrane proton flow by the pH-ISFET is translated into the indirect determination of the analyte cation activity in the sample. The potentiometric response of the CHEMFET based on the ion-partitioning membrane is based on an original response mechanism where the chemical recognition element

398


interacts actively with the signal transducer. In addition, due to the direct determination of the displaced membrane protons by the pH-ISFET, the actual sensitivity of the sensor is 59.2mV/pH for every cation analyte determined, and is translated into the sensitivity of 59.2mV/[IZ÷]. Therefore, the main advantage of the CHEMFET based on an ion-partitioning membrane is the fact that its sensitivity is 59.2 mV/decade of the analyte cation activity and therefore is not proportional to the inverse of the analyte's charge as for the conventional potentiometric sensors. As a result the CHEMFETs based on an ion-partitioning membrane present the super-Nemstian sensitivity of 59.2 mV/decade [I"+] for the multivalent cations Iz+. In this paper we present the development of CHEMFETs based on ionpartitioning membranes that are selective towards monovalent and divalent cations. Potassium and Calcium were selected as the model cations. The effect of the composition of both the sample and the ion-partitioning membrane is evaluated as a function of the magnitude of the membrane proton flux, and thus the observed sensitivity of the resulting CHEMFETs. The analytical characteristics of the CHEMFETs based on the optimized ion-partitioning membranes are evaluated with the determination of the free potassium and calcium ion activity in blood serum samples. 3.

EXPERIMENTAL

3.1

Reagents and Membranes All reagents used were of puriss p.a. grade and were obtained from Fluka, while deionized water (Bamstead NAN-O-Pure) was used for the preparation of all solutions. All the polymeric membrane components were Selectophores and were purchased from Fluka. The proton carriers used were ETH 1907, 4nonadecylpyridine, ETH 2439, [9-Dimethylamino-5-[4-(16-butyl-2,14-dioxo3,5-dioxaeicosyl)phenylamino] benzo[a] phenoxazine and tridodecylamine (TDA). Valinomycin was incorporated as the K÷ ionophore and potassium tetrakis (4-chlorophenyl)borate served as the lipophilic anionic sites for the potassium selective ion-partitioning membranes. The calcium ionophore used in the calcium selective ion-partitioning membranes was the ETH 1001 [((-)-(R,R)N,N'-(Bis( 11-ethoxyc arbonyl)undecyl-N,N'-4,5-tetramethyl-3,6dioxaoctanediamide with the lipophilic anionic sites potassium tetrakis [3,5bis(trifluoromethyl)phenyl] borate. The plasticizers tested were bis(2ethylhexyl) sebacate (DOS), 2-nitrophenyl octyl ether (o-NPOE) and tributyl phosphate (TBP). The polymer matrixes examined were high molecular weight poly(vinyl chloride) (PVC), carboxylated PVC (PVC-COOH, kindly provided by Prof. Mark E. Meyerhoff, Department of Chemistry, University of Michigan, Ann Arbor), aminated PVC (PVC-NH2) and silicone rubber (constructed by the

399


combination of siloprene K1000 and siloprene crosslinking agent K-11 in CH2C12 directly before membrane application). The ion-partitioning membranes constructed contained valinomycin or ETH 1001 (1%wt), the proton carrier ETH 1907 at a molar ratio 1 unless otherwise stated, the lipophilic anionic sites (0.3 %wt), the plasticizer and polymer matrix (in 2:1 weight ratio) in either THF (for PVC and substituted PVC-based membranes) or CH2C12 (for the silicone rubber-based membranes) in 1.8 mL solvent per 100 mg of membrane constituents. 3.2

Sensor Construction Aluminum oxide pH-ISFETs (with internal Ag/AgC1 reference electrodes) were provided by Thermo Orion (Catalog No. 615700). The CHEMFETs were prepared by casting 8 ~tL portions of the ion-partitioning membrane cocktail onto the pH-ISFET's gate and were allowed to dry at room temperature for 30 minutes (estimated membrane thickness less than 10 ~tm). The optimized Ca-CHEMFETs incorporated a dextran layer between the pHISFET's gate and the ion-partitioning membrane. The role of this dextran layer was to stabilize the thin aqueous layer formed between the ion-partitioning membrane and the pH-ISFET's gate, where the membrane proton flow is measured. These Ca-CHEMFETs were constructed by casting a small amount (0.3~tl) of the diethylaminoethyl dextran (Sigma, D-9885) solution prepared by dissolving 10mg dextran in 0.5 mL deionized water, on the pH-ISFET's gate. After 90 minutes, the cocktail of the calcium ion-partitioning membrane was deposited over the dextran layer. The CHEMFETs were then soaked in the testing solution for 30 minutes. Solutions with 10~-10 -1 M KC1 and low initial pH values adjusted with 1 M HC1 were used for the pH-response studies. The pH was increased from low to high pH values using either 1 M Tris (Tris(hydroxymethyl)aminomethane) or 2.5 M NaOH. The calibration curves were made by standard additions of 1 M KC1 or CaC12 to 10-5 M solutions of the respective cation buffered with 0.01M Tris-HC1 to pH 7.5. 3.3 Electrochemical Measurements

The CHEMFET signal was transduced to potential changes by a model 605 preamplifier (Thermo Orion). A Thermo Orion ROSS pH electrode was used to monitor the sample pH during all experiments while the laboratory temperature was set to 25+0.5 ° C. The data were collected using a personal computer with software written in Basic.

400

Ion-Partitioning Membranes As Electroactive Elements For The Development Of A Novel Cation-SelectiveCHEMFET Sensor System Moschou

3.4

Blood Serum Analysis The flee potassium and calcium ion activity in blood serum samples was measured by the K-CHEMFETs and also a potassium [43] and calcium [44] selective ISE as the control method [45,46]. Both of the above sensors were inserted in a flow injection analysis (FIA) system. The FIA system consisted of one pulseless pump (Sage 362, Thermo Orion), and two units, the unit of the potassium determination with the KCHEMFET and the K-ISE in series and the second unit in parallel were the CaCHEMFET was followed in series by the Ca-ISE for the calcium measurement. So, the system was constructed by the use of two V-100D diagonal flow injection valves (Upchurch Scientific) with 230 gL loop volumes and two reaction coils (0.030 in. i.d., 100 cm length). The four potentiometric flow through cells used were constructed of Derlin (Goodfellow Cambridge Ltd.) and were used in series (the outlet of the K-CHEMFET flow through cell was connected to the inlet of the K-ISE flow through cell and in analogous way the Ca-CHEMFET and Ca-ISE were connected to each other). The stream carriers used was 10-2 M Tris-HC1 pH 7.5 for the potassium determination and a 102M potassium hydrogen phthalate (KHP) pH 3.0 for the calcium measurement under a flow rate of 2.5 mL/min. The sensors were calibrated by the use of artificial blood serum solutions containing the following electrolytes [47]: 5.0x10 "4 M NaHSO4, 8.5x10 "4 M MgCI2, 3.0xl 0 .2 M NaHCO3, 9.5xl 0 .4 M NaH2PO4, 1.4xl 0 -1 M NaC1 and either 2.5x10 3 M CaCI2 or 4.0x103M KC1 for the potassium and calcium standards, respectively. The blood serum samples and the artificial blood serum solutions were diluted five times (ratio 1:5) using the buffer Tris-HC1 pH 7.5; while the blood serum sample was either diluted five to eleven times (ratios 1:5 up to 1"11) for the potassium determination or standard calcium additions took place for the calcium measurement to provide blood serum samples of various potassium and calcium concentrations in the physiological range. 4.

RESULTS AND DISCUSSION

Due to the composition of the membranes employed, any increase of the analyte cation activity in the sample will result in the mediated transport of the analyte cations into the bulk of the polymeric membrane with the simultaneous transport out of the membrane of the secondary cation already present in the membrane.

I

z+

(aq) +

nL(org) +

zCH+(org)

New Insights into Membrane Science and Technology: Polymeric and Biofunctional Membranes

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

Membrane Biophysics (Membrane Science and Technology)

Inorganic Polymeric Nanocomposites And Membranes

Polymeric Gas Separation Membranes

Membrane Technology and Applications

Inorganic Polymeric Nanocomposites and Membranes (Advances in Polymer Science)

Fundamentals of inorganic membrane science and technology

Membrane Technology and Applications

Ion Exchange Membranes, Volume 12: Fundamentals and Applications (Membrane Science and Technology)

Synthetic Membranes and Membrane Separation Processes

Retroviruses and Insights into Cancer

Advanced membrane technology and applications

Comprehensive Membrane Science and Engineering

Membrane Technology: Volume 3: Membranes for Food Applications

Membrane Technology: Volume 3: Membranes for Food Applications

Membrane Technology: Volume 1: Membranes for Life Sciences

Membrane Contactors: Fundamentals, Applications and Potentialities, Volume 11 (Membrane Science and Technology)

Fibrocytes: New Insights into Tissue Repair And Systemic Fibrosis

New Insights Into Business Toeic Workbook

Fibrocytes: New Insights into Tissue Repair And Systemic Fibrosis

Food Science and Technology: New Research

Insights Into Hittite History and Archaeology

New Nanotechnology Developments (Nanotechnology Science and Technology)

Insights into game theory

Membrane Receptors (Current Topics in Membranes and Transport ; V. 18)

Biofunctional Textiles And the Skin

Insights Into Teachers' Thinking And Practice

The Postdigital Membrane: Imagination, Technology and Desire

Membrane Separations Technology: Principles and Applications

Nano and micro engineered membrane technology

New Insights into Membrane Science and Technology: Polymeric and Biofunctional Membranes

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

Membrane Biophysics (Membrane Science and Technology)

Inorganic Polymeric Nanocomposites And Membranes

Polymeric Gas Separation Membranes

Membrane Technology and Applications

Inorganic Polymeric Nanocomposites and Membranes (Advances in Polymer Science)

Fundamentals of inorganic membrane science and technology

Membrane Technology and Applications

Ion Exchange Membranes, Volume 12: Fundamentals and Applications (Membrane Science and Technology)

Synthetic Membranes and Membrane Separation Processes

Retroviruses and Insights into Cancer

Advanced membrane technology and applications

Comprehensive Membrane Science and Engineering

Membrane Technology: Volume 3: Membranes for Food Applications

Membrane Technology: Volume 3: Membranes for Food Applications

Membrane Technology: Volume 1: Membranes for Life Sciences

Membrane Contactors: Fundamentals, Applications and Potentialities, Volume 11 (Membrane Science and Technology)

Fibrocytes: New Insights into Tissue Repair And Systemic Fibrosis

New Insights Into Business Toeic Workbook

Fibrocytes: New Insights into Tissue Repair And Systemic Fibrosis

Food Science and Technology: New Research

Insights Into Hittite History and Archaeology

New Nanotechnology Developments (Nanotechnology Science and Technology)

Insights into game theory

Membrane Receptors (Current Topics in Membranes and Transport ; V. 18)

Biofunctional Textiles And the Skin

Insights Into Teachers' Thinking And Practice

The Postdigital Membrane: Imagination, Technology and Desire

Membrane Separations Technology: Principles and Applications

Nano and micro engineered membrane technology

Recommend Documents