Re Engineering the Chemical Processing Plant

RE-ENGINEERING THE CHEMICflL PROCESSING PMNT Process Intensification edited by

flndrzej ftankiewicz DSM Research Geleen, and Delft University of Technology Delft, The Netherlands

Jacob fl.Moulijn Delft University of Technology Delft, The Netherlands

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Although great care has been taken to provide accurate and current information, neither the author(s) nor the publisher, nor anyone else associated with this publication, shall be liable for any loss, damage, or liability directly or indirectly caused or alleged to be caused by this book. The material contained herein is not intended to provide specific advice or recommendations for any specific situation. Trademark notice: Product or corporate names may be trademarks or registered trademarks and are used only for identification and explanation without intent to infringe. Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress. ISBN: 0-8247-4302-4 This book is printed on acid-free paper. Headquarters Marcel Dekker, Inc., 270 Madison Avenue, New York, NY 10016, U.S.A. tel: 212-696-9000; fax: 212-685-4540 Distribution and Customer Service Marcel Dekker, Inc., Cimarron Road, Monticello, New York 12701, U.S.A. tel: 800-228-1160; fax: 845-796-1772 Eastern Hemisphere Distribution Marcel Dekker AG, Hutgasse 4, Postfach 812, CH-4001 Basel, Switzerland tel: 41-61-260-6300; fax: 41-61-260-6333 World Wide Web http://www.dekker.com The publisher offers discounts on this book when ordered in bulk quantities. For more information, write to Special Sales/Professional Marketing at the headquarters address above. Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved. Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, microfilming, and recording, or by any information storage and retrieval system, without permission in writing from the publisher. Current printing (last digit): 10 9 8 7 6 5 4 3 2 1 PRINTED IN THE UNITED STATES OF AMERICA


CHEMICAL INDUSTRIES A Series of Reference Books and Textbooks Founding Editor HEINZ HEINEMANN

1. Fluid Catalytic Cracking with Zeolite Catalysts, Paul B. Veriuto and E. Thomas Habib, Jr. 2. Ethylene: Keystone to the Petrochemical Industry, Ludwig Kniel, Olaf Winter, and Karl Stork 3. The Chemistry and Technology of Petroleum, James G. Speight 4. The Desulfurization of Heavy Oils and Residua, James G. Speight 5. Catalysis of Organic Reactions, edited by William R. Moser 6. Acetylene-Based Chemicals from Coal and Other Natural Resources, Robert J. Tedeschi 7. Chemically Resistant Masonry, Walter Lee Sheppard, Jr. 8. Compressors and Expanders: Selection and Application for the Process Industry, Heinz P. Bloch, Joseph A. Cameron, Frank M. Danowski, Jr., Ralph James, Jr., Judson S. Swearingen, and Marilyn E. Weightman 9. Metering Pumps: Selection and Application, James P. Poyntori 10. Hydrocarbons from Methanol, Clarence D. Chang 11. Form Flotation: Theory and Applications, Ann N. Clarke arid David J. Wilson 12. The Chemistry and Technology of Coal, James G. Speight 13. Pneumatic and Hydraulic Conveying of Solids, 0.A. Williams 14. Catalyst Manufacture: Laboratory and Commercial Preparations, Alvin B. Stiles 15. Characterization of Heterogeneous Catalysts, edited by Francis Delannay 16. BASIC Programs for Chemical Engineering Design, James H. Weber 17. Catalyst Poisoning, L. Louis Hegedus and Robert W. McCabe 18. Catalysis of Organic Reactions, edited by John R. Kosak 19. Adsorption Technology: A Step-by-step Approach to Process Evaluation and Application, edited by Frank L. Slejko 20. Deactivation and Poisoning of Catalysts, edited by Jacques Oudar and Henry Wise 21. Catalysis and Surface Science: Developments in Chemicals ,torn Methanol, Hydrotreating of Hydrocaribons, Catalyst Preparation, Monlomers and Polymers, Photocafalysis and Photovoltaics, edited by Heinz t-ieinemann and Gabor A. Somorjai 22. Catalysis of Organic Reactions, edited by Robert L. Augustine


23. Modern Control Techniques for the Processing Industries, T. H. Tsai, J. W. Lane, and C. S. Lin 24. Temperature-Programmed Reduction for Solid Materials Characterization, Alan Jones and Brian McNichol 25. Catalytic Cracking: Catalysts, Chemistry, and Kinetics, Bohdan W. Wojciechowski and Avelino Corma 26. Chemical Reaction and Reactor Engineering, edited by J. J. Carberry and A. Varma 27. Filtration: Principles and Practices, Second Edition, edited by Michael J. Matteson and Clyde Orr 28. Corrosion Mechanisms, edited by Florian Mansfeld 29. Catalysis and Surface Properties of Liquid Metals and Alloys, Yoshisada Ogino 30. Catalyst Deactivation, edited by Eugene E. Petersen and Alexis T. Bell 3 1. Hydrogen Effects in Catalysis: Fundamentals and Practical Applications, edited by Zoltan Paal and P. G. Menon 32. Flow Management for Engineers and Scientists, Nicholas P. Cheremisinoff and Paul N. Cheremisinoff 33. Catalysis of Organic Reactions, edited by Paul N. Rylander, Harold Greenfield, and Robert L. Augustine 34. Powder and Bulk Solids Handling Processes: Instrumentation and Control, Koichi linoya, Hiroaki Masuda, and Kinnosuke Watanabe 35. Reverse Osmosis Technology: Applications for High-Purity-Water Production, edited by Bipin S. Parekh 36. Shape Selective Catalysis in lndustrial Applications, N. Y. Chen, William E. Garwood, and Frank G. Dwyer 37. Alpha Olefins Applications Handbook, edited by George R. Lappin and Joseph L. Sauer 38. Process Modeling and Control in Chemical Industries, edited by Kaddour Najim 39. Clathrate Hydrates of Natural Gases, E. Dendy Sloan, Jr. 40. Catalysis of Organic Reactions, edited by Dale W. Blackburn 41. Fuel Science and Technology Handbook, edited by James G. Speight 42. Octane-Enhancing Zeolitic FCC Catalysts, Julius Schetzer 43. Oxygen in Catalysis, Adam Bielanski and Jerzy Haber 44. The Chemistry and Technology of Petroleum: Second Edition, Revised and Expanded, James G. Speight 45. Industrial Drying Equipment: Selection and Application, C. M. van't Land 46. Novel Production Methods for Ethylene, Light Hydrocarbons, and Aromatics, edited by Lyle F. Albright, Billy L. Crynes, and Siegfried Nowak 47. Catalysis of Organic Reactions, edited by William E. Pascoe 48. Synthetic Lubricants and High-Performance Functional Fluids, edited by Ronald L. Shubkin 49. Acetic Acid and Its Derivatives, edited by Victor H. Agreda and Joseph R. ZoelI er 50. Properties and Applications of Perovskite-Type Oxides, edited by L. G. Tejuca and J. L. G. Fierro


51. Computer-Aided Design of Catalysts, edited by E. Robert Becker and Carmo J. Pereira 52. Models for Thermodynamic and Phase Equilibria Calculations, edited by Stanley I. Sandler 53. Catalysis of Organic Reactions, edited by John R. Kosak and Thomas A. Johnson 54. Composition and Analysis of Heavy Petroleum fractions, Klaus H. Altgelt and Mieczyslaw M. Boduszynski 55. NMR Techniques in Catalysis, edited by Alexis T. Bell and Alexander Pines 56. Upgrading Petroleum Residues and Heavy Oils, Murray R. Gray 57. Methanol Production and Use, edited by Wu-Hsun Cheng and Harold H. Kung 58. Catalytic Hydroprocessing of Petroleum and Distillates, edited by Michael C. Oballah and Stuart S. Shih 59. The Chemistry and Technology of Coal: Second Edition, Revised and Expanded, James G. Speight 60. Lubricant Base Oil and Wax Processing, Avilino Sequeira, Jr. 61. Catalytic Naphtha Reforming: Science and Technology, edited by George J. Antos, Abdullah M. Aitani, and Jose M. Parera 62. Catalysis of Organic Reactions, edited by Mike G. Scaros and Michael L. Prunier 63. Catalyst Manufacture, Alvin B. Stiles and Theodore A. Koch 64. Handbook of Grignard Reagents, edited by Gary S. Silverman and Philip E. Rakita 65. Shape Selective Catalysis in lndustrial Applications: Second Edition, Revised and Expanded, N. Y. Chen, William E. Garwood, and Francis G. Dwyer 66. Hydrocracking Science and Technology, Julius Scherrer and A. J. Gruia 67. Hydrotreating Technology for Pollution Control: Catalysts, Catalysis, and Processes, edited by Mario L. Occelli and Russell Chianelli 68. Catalysis of Organic Reactions, edited by Russell E. Malz, Jr. 69. Synthesis of Porous Materials: Zeolites, Clays, and Nanostructures, edited by Mario L. Occelli and Henri Kessler 70. Methane and Its Derivatives, Sunggyu Lee 71. Structured Catalysts and Reactors, edited by Andrzei Cybulski and Jacob Moulijn 72. lndustrial Gases in Petrochemical Processing, Harold Gunardson 73. Clathrate Hydrates of Natural Gases: Second Edition, Revised and Expanded, E. Dendy Sloan, Jr. 74, fluid Cracking Catalysts, edited by Mario L. Occelli and Paul O’Connor 75 Catalysis of Organic Reactions, edited by Frank E. Herkes 76, The Chemistry and Technology of Petroleum, Third Edition, Revised and Expanded, James G. Speight 77. Synthetic Lubricants and High-P erformance functional Fluids, Second Edition: Revised and Expanded, Leslie R. Rudnick and Ronald L. Shubkin I


78. The Desulfurization of Heavy Oils and Residua, Second Edition, Revised and Expanded, James G. Speight 79. Reaction Kinetics and Reactor Design: Second Edition, Revised and Expanded, John B. Butt 80. Regulatory Chemicals Handbook, Jennifer M. Spero, Bella Devito, and Louis Theodore 8 1. Applied Parameter Esfimation for Chemical Engineers, Peter Englezos and Nicolas Kalogerakis 82. Catalysis of Organic Reacfions, edited by Michael E. Ford 83. The Chemical Process Industries Infrastructure: Function and Economics, James R. Couper, 0. Thomas Beasley, and W. Roy Penney 84. Transport Phenomena Fundamentals, Joel L. Plawsky 85. Petroleum Refining Processes, James G. Speight and Baki Ozum 86. Health, Safety, and Accident Management in the Chemical Process Industries, Ann Marie Flynn and Louis Theodore 87. Plantwide Dynamic Simulators in Chemical Processing and Clonfrol, William L. Luyben 88. Chemicial Reactor Design, Peter Harriott 89. Catalysis of Organic Reactions, edited by Dennis Morrell 90. Lubricant Additives: Chemistry and Applications, edited by Leslie R. Rudnick 91. Handbook of Fluidization and Fluid-Particle Systems, edited by Wenching Yang 92. Conservation Equations and Modeling of Chemical and Biochemical Processes, Said S. E. H. Elnashaie and Parag Garhyan 93. Batch Fermentation: Modeling, Monitoring, and Control, Ali Cinar, Satish J. Parulekar, Cenk Undey, and Gulnur Birol 94. lndustrial Solvents Handbook: Second Edition, Nicholas P. Cheremisinoff 95. Petroleum and Gas Field Processing, H. K. Abdel-Aal, Mohamed Aggour, and M. A. Fahim 96. Chemical Process Engineering: Design and Economics, Harry Sit la 97. Process Engineering Economics, James R. Couper 98. Re-Engineering the Chemical Processing Plant: Process lntensification, Andrzej Stankiewicz and Jacob A. Moulijn 99. Thermodynamic Cycles: Computer-Aided Design and Optimization, Chih Wu ADDITIONAL VOLUMES IN PREPARATION

Handbook of Methyl Tertiary Butyl Ether, edited by S. Halim Harriid and Mohammad Ashraf Ali Catalytic Naphtha Reforming, Second Edition, Revised and Expanded, edited by George J. Antos and Abdullah M. Aitani


Preface

The book you are about to read will introduce you to modern ways of reengineering the chemical processing plant by means of Process Intensification (PI). The story behind this book had begun with the paper Process Intensification: Transforming Chemical Engineering, which we published in the millennium issue of Chemical Engineering Progress (January 2000). After a pretty enthusiastic response to our paper by the chemical engineering community, Marcel Dekker proposed to us writing a book on that subject. After some discussions we came to the conclusion that it was not a good idea to write the entire book ourselves because, as you will see next, Process Intensification is a very broad discipline and includes many diverse expertise fields. So, instead of writing all chapters on our own, we have invited a number of prominent experts in various areas of Process Intensification, both from industry and from academia, to contribute to what now has become the world’s first book on that subject. The principal aim of this highly practice-oriented book is to illustrate the current developments and the frontline research in the area of Process Intensification. The book is primarily intended for engineers, technologists and researchers in chemical, biochemical and engineering companies, who are involved in process design and development and are interested in learning more about equipment and techniques that may bring quantum-leap improvements to their technologies. Also for others working in the forefront of process design and


development it is intended to be inspiring, in particular for the chemical engineering community in the universities and the National Laboratories. We hope that it will contribute to a better image of the chemical industry and even play a role in attracting more high-quality, motivated students to the discipline. The book may also be beneficial to R&D managerial personnel who wish to have a broader understanding of the principles and methodology of Process Intensification and gain the up-to-date knowledge of the emerging novel equipment and processing methods that could help to achieve technological breakthroughs in the processes at their companies. The book has a certain logical structure that can be inferred from scanning the individual chapter headings. Chapter 1 introduces the reader into the genesis, philosophy and principles of Process Intensification and discusses its dimension and structure. It provides general information on process-intensifying equipment and methods and gives some examples of their application on the commercial scale. The three subsequent chapters describe selected types of the PI-equipment. Most of that equipment have already been successfully implemented on the commercial scale or is ready for implementation. Chapters 2 and 3 are devoted to the rotating equipment, rotating packed beds and spinning disk reactors, in which the use of high gravity fields leads to spectacular miniaturization of the processing units. Chapter 4 in turn describes the technology, design and application of compact and multifunctional heat exchangers. The next three chapters show how bringing certain structures in various scales of chemical processing environment can boost process efficiency, by dramatically improving mixing, heat and mass transfer. Various types and scales of such structuring are presented: microreactors in Chapter 5, large-scale structured catalysts and reactors in Chapter 6 and inline mixing equipment in Chapter 7. Following that “hardware” part of the book, its next four chapters focus on some important methods that can be used for intensification of chemical processes. Chapter 8 presents the application aspects of functional integration of reaction and separation into reactive separation systems, or integration of different separative techniques into hybrid separations. In Chapter 9 the modeling issues of the reactive separation systems are discussed. Chapter 10 discusses some aspects of the integration of reaction and heat transfer in multifunctional reactors, while Chapter 11 focuses on the application of process synthesis principles to the optimal design of integrated chemical processing plants. The final three chapters of the book address more general issues of Process Intensification. Chapter 12, based on the experiences within DSM, shows how the PI-principles can be applied in the industrial environment for redesigning and development of process-intensive chemical plants, while Chapters 13 and 14 focus respectively on safety and sustainability aspects of PI. The chemical industry skyline in the 21st Century is changing. New highly efficient devices have already begun replacing the tens-of-meters high reactors and separation columns. In the still denser populated world inhabited by the still more


educated and environment-conscious society, there will be no room (literally and figuratively) for the huge, inefficient chemical factories of today, generating tens of tons of waste per each ton of the useful product. As a part of the society-driven changes miniaturization and, in general, intensification of chemical and biochemical plants, will become inevitable. We are well aware that the present book does not cover all developments in the field of Process Intensification. It has not had such ambitions. With this collection of contributions by the leading experts in the field, we have tried to focus on the main developments and main issues only, hoping that they will give the reader sufficient flavor of PI and will encourage him/her to further studies on how to re-engineer a chemical processing plant basing on the “smaller-cheaper-saferslicker” principles of Process Intensification. Both contributors and editors will be very glad to hear from the reader if we indeed have succeeded. Also suggestions for a possible next edition are welcome! Andrzej Stankiewicz Jacob A. Moulijn


Contents

Preface Contributors 1. Process Intensification: History, Philosophy, Principles Andrzej Stankiewicz and A. A. H. Drinkenburg 2. Chemical Processing in High-Gravity Fields David L. Trent 3. The Spinning Disc Reactor C. Ramshaw 4. Compact Multifunctional Heat Exchangers: A Pathway to Process Intensification B. Thonon and P. Tochon 5. Process Intensification Through Microreaction Technology Wolfgang Ehrfeld 6. Structured Catalysts and Reactors: A Contribution to Process Intensification Jacob A. Moulijn, Freek Kapteijn, and Andrzej Stankiewicz


7. Inline and High-Intensity Mixers Andrew Green 8. Reactive and Hybrid Separations: Incentives, Applications, Barriers Andrzej Stankiewicz 9. Reactive Separations in Fluid Systems E. Y. Kenig, A. Górak, and H.-J. Bart 10. Multifunctional Reactors: Integration of Reaction and Heat Transfer David W. Agar 11. Process Synthesis/Integration Patrick Linke, Antonis Kokossis, and Henk van den Berg 12. Process Intensification in Industrial Practice: Methodology and Application Remko A. Bakker 13. Process Intensification for Safety Dennis C. Hendershot 14. Process Intensification Contributions to Sustainable Development G. Jan Harmsen, Gijsbert Korevaar, and Saul M. Lemkowitz


Contributors

David W. Agar Lehrstuhl für Technische Chemie B, University of Dortmund, Dortmund, Germany Remko A. Bakker DSM Fine Chemicals Austria, Linz, Austria H.-J. Bart Institute of Thermal Process Engineering, University of Kaiserslautern, Kaiserslautern, Germany A. A. H. Drinkenberg DSM Research, Geleen, The Netherlands Wolfgang Erhfeld Ehrfeld Mikrotechnik AG, Wendelsheim, Germany A. Górak Lehrstuhl für Thermische Verfahrenstechnik, University of Dortmund, Germany Andrew Green BHR Group Limited, Cranfield, England G. Jan Harmsen Delft University of Technology, Delft, The Netherlands Dennis C. Hendershot Rohm and Haas Company, Bristol, Pennsylvania, U.S.A.


Freek Kapteijn Delft University of Technology, Delft, The Netherlands E. Y. Kenig Lehrstuhl für Thermische Verfahrenstechnik, University of Dortmund, Dortmund, Germany Antonis Kokossis University of Surrey, Surrey, England Gijsbert Korevaar Delft University of Technology, Delft, The Netherlands Saul M. Lemkowitz Delft University of Technology, Delft, The Netherlands Patrick Linke University of Surrey, Surrey, England Jacob A. Moulijn Delft University of Technology, Delft, The Netherlands C. Ramshaw Department of Chemical and Process Engineering, University of Newcastle upon Tyne, Newcastle upon Tyne, England Andrzej Stankiewicz DSM Research, Geleen, and Delft University of Technology, Delft, The Netherlands B. Thonon P. Tochon

Greth, CEA–Grenoble, Grenoble, France Greth, CEA–Grenoble, Grenoble, France

David L. Trent

The Dow Chemical Company, Freeport, Texas, U.S.A.

Henk van den Berg Faculty of Chemical Technology, University of Twente, Enschede, The Netherlands, and Ghent University, Ghent, Belgium


1 Process Intensification: History, Philosophy, Principles Andrzej Stankiewicz and A. A. H. Drinkenburg DSM Research, Geleen, The Netherlands

1.

INTRODUCTION

Process intensification (PI) is currently one of the most significant trends in chemical engineering and process technology. It is attracting more and more of the attention of the research world. Four international conferences, several smaller symposia/workshops every year, and a number of dedicated issues of professional journals are clear proof of it. A number of commercial-scale applications of the PI principles have already taken place. But how did it all begin? 2.

A BIT OF HISTORY

According to Miriam-Webster’s Collegiate Dictionary, the word intensive has probably its origins somewhere in 15th century. And it was not many years later, right at the peak of the Renaissance, when Georgius Agricola published his famous book De Re Metallica (1), the book that is commonly regarded as the first comprehensive textbook on the engineering of mining and metallurgy. De Re Metallica is richly illustrated with woodcuts showing equipment and processing methods used in the times of Agricola. In many of those woodcuts clear elements of process intensification–oriented thinking can be found. One example is shown in Figure 1,


FIGURE 1 Sixteenth century technology of gold retrieval from gold ore. (From Ref. 1.)


which illustrates the process of retrieving gold from gold ore. The technology is pretty simple. The ore is crushed by the stamp, “C,” ground in the mill, “F,” and mixed with mercury in vessels “O.” Gold is extracted from the ore by mercury and is later separated from it by pressing the mixture through a leather or cloth filter bag (not shown in the drawing). Taking a closer look at the woodcut, one notices that the stamp, the mill and the stirrers for mixing the ore with mercury are all driven by the same water wheel, “A,” via the common axle, “B,” and a number of various gears. Speaking the language of the 21st century, one could say, “A marvelous example of a green, energy-based, highly integrated processing plant!” (One dare not, however, call it a sustainable technology. Not only are the gold reserves unsustainable, but the operations involving mercury are not environmentally friendly, as we all know today.) Yet there is another aspect to Agricola’s woodcut. As one may have noticed, some of the equipment shown (“O”–“S”) exhibits a striking resemblance to the equipment used in the chemical process industry almost 450 years later (see Figure 2). Were the contemporaries of Agricola so ingenious, or are we merely satisfied with the inventions of past centuries? At the dawn of the third millennium, in-series stirred tanks still remain the most common chemical processing system. An attempt to break this domination of the stirred-tank technology by the invention and introduction of the static mixer (2,3), is one of the finest and earliest modern examples of process intensification. Here, the technological leap was achieved not by the improvement of the stirring itself but, quite the opposite, by abandoning the mechanical stirring as a method of mixing fluids! This reveals one of the most important features of PI—the changes it brings are drastic in nature, revolutionary rather than evolutionary. In the scientific literature, the term process intensification started to appear in the mid-1960s and early ’70s, mostly in East European publications concerning

FIGURE 2 Four and half centuries have passed, yet almost no fundamental differences can be seen between the technology of 1556 and that of 2002.


metallurgical processing (4 –7) (an interesting coincidence, considering that Agricola’s book was also on metallurgy, not on chemical processing). Of course, all of those papers understood process intensification as simply equivalent to process improvement. Also, in the first chemical industry–oriented articles (all of East European origin, by the way), the term process intensification had that same meaning (8–10). The birth of process intensification as a chemical engineering discipline came several years later in the United Kingdom and was marked by the paper published in 1983 by Colin Ramshaw from the ICI New Science Group, who described their studies on the application of centrifugal fields (so-called “HiGee”) in distillation processes (11). A few months later the Annual Research Meeting, entitled Process Intensification, was held at UMIST, Manchester (12). Interestingly, the first paper presented at that meeting concerned processing of gold ore using intensive methods—a strange coincidence, indeed. Both in the paper by Ramshaw and in the report from the UMIST conference, first definitions (or rather descriptions) of process intensification can be found. Ramshaw (11) describes PI as “devising an exceedingly compact plant which reduces both the ‘main plant item’ and the installation’s costs,” while according to Heggs (12) PI is concerned with order-of-magnitude reductions in process plant and equipment. In one of his subsequent papers, Ramshaw writes about typical equipment volume reduction by two or three orders of magnitude (13). Until the early 1990s, process intensification was mainly a British discipline and was focused primarily on four areas: the use of centrifugal forces, compact heat transfer, intensive mixing, and combined technologies (14). It was also the Brits who organized the first international conference on PI (15). By that time, however, process intensification had already become an international business, for many research centers in different countries had entered the field. In Holland, for instance, Delft University of Technology, together with DSM, carried out research on structured reactors (16). Another group in Delft investigated centrifugal adsorption technology (17). In France, Greth CEN institute in Grenoble carried out extensive studies on compact heat exchange equipment (18). In Germany, research on microtechnology flourished in the Institut für Mikrotechnik Mainz (19), while in China a special center at Beijing University was established to carry out R&D activities in the area of high-gravity processing (20). In the United States a number of research institutes started PI-related studies too, e.g., Pacific Northwest National Laboratory in the field of microchannel heat exchangers (21) and MIT in the field of microreactors (22). Also early on, a number of chemical companies got involved in process intensification. This involvement resulted in the first successful commercial-scale applications, such as the methyl acetate process of Eastman Chemical (23), the hydrogen peroxide distillation system of Sulzer (24), and the hypochlorous acid process of Dow Chemical (25).


The end of the 20th century and the beginning of the 21st have seen a fast growth in PI-related activities in both industry and academia. In the UK the Process Intensification Network was launched, gathering a large number of industrial and academic participants; a similar network has been established in the Netherlands. Four international conferences and several smaller symposia on PI have been organized so far. Process Intensification, traditionally tied up with the commodity chemicals sector, has begun entering new areas, such as bioprocessing and fermentation (26–28) and, quite recently, fine chemistry (29). The definition of process intensification has changed accordingly. It is no longer exclusively regarded as drastically smaller equipment/plants (although equipment compactness remains its most obvious feature). Process intensification, as it is widely understood nowadays, comprises novel equipment, processing techniques, and process development methods that, compared to conventional ones, offer substantial improvements in (bio)chemical manufacturing and processing (30). The question may arise why it took so long for PI to come into the picture. One possible answer is the enormous expansion of the process industry in the third quarter of the 20th century, expansion in market size but certainly also in plant scale. There were very few incentives at that time for the very risky introduction of new technologies. In the fourth quarter of the century, much effort was spent on modeling, optimization, and control, resulting, among other things, in the well-known onionskin methodologies of process development (e.g., Ref. 31). Although very worthwhile at the time, these sequential, onionskin methodologies (first the reactor, then separation/purification, then heat integration, then process control, safety, etc.) hindered the thinking in terms of integrated equipment. More incentives were found in environmental engineering, also a topic of the second half of the 20th century, that developed from the end-of-pipe solutions to problems to the integrated process solutions backsourcing the problems. 3. THE PHILOSOPHY AND OPPORTUNITIES OF PROCESS INTENSIFICATION The philosophy of process intensification has been traditionally characterized by four words: smaller, cheaper, safer, slicker. And indeed, equipment size, land use costs, and process safety are among the most important PI incentives. But process intensification can (and should) also be placed in a broader context—the context of sustainable technological development. Several years ago DSM published a picture symbolizing its own vision of process intensification (32), in which skyscraping distillation towers of the naphtha-cracking unit are replaced by a compact, clean, and tidy indoor plant (see Figure 3). The importance of PI for sustainable development and its role in the company’s responsible business has been further stressed in a recent publication by the company’s CEO, Peter Elverding (33). Here,


FIGURE 3 DSM’s vision of process intensification.

process intensification was the highest-rated activity of DSM within the known “Triple-P” (profit–planet–people) triangle, as shown in Table 1. From this general philosophy of process intensification follow concrete opportunities that PI offers to chemical enterprises, as shown in Figure 4. These opportunities exist primarily in four areas: costs, safety, time to market, and company image. TABLE 1 Process Intensification in the Profit–Planet–People Triangle of DSM Triple P

Process intensification Green routes Recycling Energy efficiency

Profit

Planet

People

•• • • •

••• ••• ••• •••

•• •• •• •

Source: Ref. 32.


FIGURE 4 Main benefits from process intensification.

3.1.

Costs

Process intensification leads to substantially cheaper processes, particularly in terms of: Land costs, resulting from much higher production capacity and/or number of products per unit of manufacturing area Other investment costs, resulting from cheaper, compact equipment, reduced piping, reduced civic works, integrated processing units, etc. Costs of raw materials, due to higher yields/selectivities Costs of utilities, in particular costs of energy, due to higher energy efficiency Costs of waste processing (less waste generated in process-intensive plants) Figure 5 shows the estimated savings in some DSM technologies, after applying the PI principles (grass-roots situation). 3.2.

Safety

Process intensification drastically increases the safety of chemical processes. It is obvious that smaller is safer. In Table 2 some of the more severe chemical disasters of the past century are listed. The table shows clearly how disastrous consequences may arise from the large inventories when something goes wrong. And of course, one may not claim that process intensification would have prevented all those tragedies. Yet a study done at AIChE showed that methyl isocyanate (MIC),


FIGURE 5 Estimated savings in some DSM technologies, achieved by applying PI principles to process and plant design (grass-roots situation).

the poisonous intermediate that was released at Bhopal, could be generated and immediately converted to final products in continuous reactors that contained a total inventory of less than 10 kg of MIC (34)! But process intensification offers not only smaller equipment but also much better possibilities for keeping processes under control, for instance, via extremely efficient heat removal from exothermic reactions (one speaks about heat transfer coefficients exceeding 20,000 W/m2K) or via fully controlled gas–liquid flow in structured catalysts that prevents liquid maldistribution and hot-spot formation. Furthermore, intensification of the processing plant often leads to elimination of one or more of its components, which also has a direct advantageous effect on process safety (“What you do not have cannot leak”). 3.3.

Time to Market

Process intensification also offers substantial improvements to those sectors of the chemical industry in which time to market plays a crucial role, e.g., the fine chemical and pharmaceutical sectors. Ramshaw (35) discussed how process intensification could shorten the time to market in case of a low-tonnage pharmaceutical process. The idea consists in developing a continuous lab-scale process and using it directly as the commercial-scale process. One must not forget that liquid flow of only 1 milliliter per second means, in continuous operation, circa 30 tons per year, which is quite a reasonable capacity for many pharmaceuticals.


TABLE 2 Some 20th Century Disasters in the Chemical Industry Place

Date

Chemicals

Estimated amount

Casualties

4,500 tons exploded

ca. 600 dead, 1500 injured 28 dead, 89 injured

Oppau/ Ludwigshafen Flixborough

September 21, 1921 June 1, 1974

Ammonium sulfate, ammonium nitrate Cyclohexane

Beek

Propylene (mainly)

Seveso

November 7, 1975 July 10, 1976

2,4,5 Trichlorophenol, dioxin

400-ton inventory, 40-ton escaped 10,000-m3 inventory, 5.5 tons escaped 7-ton inventory, 3 tons escaped

San Juan, Mexico City

November 19, 1984

LPG

10,000-m3 inventory

Bhopal

December 3, 1984 October 23, 1989

Methyl isocyanate

41 tons released

Ethylene, isobutane, hexene, hydrogen

33 tons escaped

Pasadena


14 dead, 107 injured No direct casualties, ca. 37,000 people exposed ca. 500 dead, 7000 injured (mainly outside the plant) ca. 3,800 dead, 2720 permanently disabled 23 dead, 130–300 injured

The advantages of such an approach over the conventional one, based on the scale-up philosophy of “stirred tank → bigger stirred tank → even bigger stirred tank, always batch” are twofold: Process development takes place only once, with no scale-up via a pilot plant to the industrial scale. The scale-up of a batch process in stirred tanks is not straightforward, especially in the case of reactions with a large heat effect or a strong viscosity effect, and therefore can be troublesome and time-consuming. All the administrative (FDA) procedures involved in the legal approval of the production technology of the drug take place only once: The labscale technology is the commercial-scale technology. In consequence, the start of commercial production can be greatly speeded up, in some cases even by several years. Time to market will be shortened and the patent lifetime of the drug will be much more effectively utilized (read: utilized longer). 3.4.

Company Image

More and more chemical companies do recognize the fact that their image, their reputation, plays a very important role in successful business. A proper image of the company is necessary to ensure public support for its activities. A study done in the United States showed that only the tobacco industry and the nuclear energy sector had a worse reputation than the chemical industry. The situation in Europe is probably not very much different. On the other hand, process intensification, deeply anchored in the philosophy of sustainable development, in safe and environmentally friendly processing, presents perhaps the simplest, the most obvious key to the improved image of the chemical industry. 4. TECHNOLOGICAL BREAKTHROUGHS AND CREATION OF SHAREHOLDER VALUE In the reality of the global markets of the 21st century, not only do chemical companies compete with each other, they also have to compete with other sectors of the economy by proving to their shareholders that the revenues they receive from the chemical business are as good as or better than from other fast-growing sectors, such as software and servicing. But the chemical process industry struggles to create value for its shareholders. As John Goldhill of Arthur D. Little Inc. wrote, the “chemical industry has lagged other industries in creating value for shareholders for at least the past 10 years” (36). And indeed, when looking at stock index developments in various types of enterprises in the period 1997–2002 (Figure 6), one notices that the value of chemical shares grew substantially slower than in other sectors. The strategy of growth in the chemical process industry at


FIGURE 6 History of the Standard & Poor stock index in the chemical sector and other selected sectors of the economy, 1997–2002. (From: www. bloomberg.com.)

the present time is based mainly on mergers, splits, takeovers, and modifying the structure of the product portfolio. It is basically a strategy of growth via trade, not via technological innovation. In most chemical companies nowadays, opportunities are sought in cost reductions via optimization of the primary business work processes (e.g., “operational excellence”) and via opening up bottle necks in the existing production facilities. Unfortunately, neither of these activities can make a company very attractive to shareholders. In the optimization of work processes, a critical limit in cost reductions will soon be reached, and competitors will also follow more or less the same path, so the company’s competitive advantage will only be temporary unless a shakeout takes place. Opening up bottlenecks, squeezing out yet another few percent from existing plants, is also not the way to convince investors that the company is capable of delivering an adequate growth in earnings. One of the most obvious solutions to that problem lies in innovations and technological breakthroughs, because only innovations and technological breakthroughs can ensure a sustainable technological advantage, cost leadership, and growth potential. Innovations and technological breakthroughs are what process intensification is all about.


5.

PROCESS INTENSIFICATION TOOLBOX

The toolbox for process intensification is schematically shown in Figure 7. It includes process-intensifying equipment (PI hardware) and process-intensifying methods (PI software). Obviously, in many cases overlap between these two domains can be observed as new methods may require novel types of equipment to be developed and, vice versa, novel apparatuses already developed sometimes make use of new, unconventional processing methods. In Figure 7, examples of both PI hardware and PI software are shown. Many of them will be discussed in detail in other chapters of this book. Here, we give only a brief overview of the more important PI items. 5.1.

Process-Intensifying Equipment

As already mentioned, one classic example of technological breakthroughs in process engineering was the invention and commercialization of static (motionless) mixers, examples of which are shown in Figure 8. Nowadays, static mixers not only offer a more size- and energy-efficient method for mixing or contacting fluids. In the SMR static mixer reactor by Sulzer, mixing elements are made of

FIGURE 7 Process intensification toolbox.


FIGURE 8 Various types of static mixers. (Courtesy: Koch-Glitsch.)

heat transfer tubes (Figure 9). Thanks to that, the SMR units can successfully be applied in processes in which simultaneous mixing and intensive heat removal/ supply are necessary, e.g., in nitration, neutralization, and polymerization reactions. For the cases when efficient mixing has to be coupled with a solid-catalyzed reaction a whole family of open-crossflow-structure catalysts has been developed. The best known of them are the so-called KATAPAK®s, commercialized by Sulzer. One of them, KATAPAK-M® is shown in Figure 10. It has good mixing properties and can simultaneously be used as the support for catalytic material.

FIGURE 9 Static mixer reactor developed by Sulzer. (Courtesy: Sulzer Chemtech.)


FIGURE 10 Sulzer’s KATAPAK-M®. (Courtesy: Sulzer Chemtech.)

KATAPAK®s are applied in catalytic distillation and in some gas-phase exothermic oxidation processes traditionally carried out in fixed beds. In these processes KATAPAK®s exhibit very good radial heat transfer characteristics (37). In nonreactive distillation processes structured packings are also widely used. One of the most recent and most promising types is the Super X-Pack developed by Nagaoka International Corporation, shown in Figure 11. This wire-based packing is claimed to be able to reduce the height of a distillation column by a factor of 5 compared to a conventional tray design, as shown in Figure 12 (38).


Also, the Super X-Pack can save up to 80% energy due to a substantially lower pressure drop. Heterogeneous catalytic processes can often be intensified by the use of monolithic catalysts (39). These are metallic or nonmetallic bodies forming a multitude of straight, narrow channels of defined uniform cross-sectional shapes (Figure 13). In order to ensure sufficient porosity and to enhance the catalytically active surface, the inner walls of the monolith channels are usually covered with a thin layer of washcoat, which acts as the support for the catalytically active species. The most important features of the monoliths are: Very low pressure drop in single- and two-phase flow, one to two orders of magnitude lower than in conventional packed-bed systems

FIGURE 11 Super X-Pack, developed by Nagaoka International Corp. (Courtesy: Nagaoka International Corp.)


FIGURE 12 Height reduction of a distillation column claimed by Super X-Pack. (Courtesy: Nagaoka International Corp.)


FIGURE 13 Monolithic catalysts.

High geometrical areas per reactor volume, typically 1.5– 4 times higher than in the reactors with particulate catalysts Very high catalytic efficiency, practically 100%, due to very short diffusion paths in thin washcoat layer Stankiewicz (40) gives a spectacular example of reactor size reduction by a factor of ca. 100, as a result of replacement of the conventional system with a monolithic reactor operated horizontally in a pipeline. For highly exothermic reactions the so-called HEX reactors present a very promising option. The basic common feature of all HEX reactors is much more favorable heat transfer conditions in comparison with conventional reactors (heat transfer coefficients typically 3500–7500 W/m2K, heat transfer areas up to 2200 m2/m3). A HEX reactor developed by BHR Group Ltd. (Figure 14) was able to decrease the by-product formation in one of ICI Acrylics’ processes by 75% (41) and to decrease the processing time in a Hickson & Welch fine chemical process from 18 hours to 15 minutes, saving 98.6% of batch time (42). Even higher values of heat transfer coefficients than those in the HEX reactors can be achieved in microreactors. Here, values up to 20,000 W/m2K are reported (43). Microreactors (Figure 15) are chemical reactors of extremely small dimensions that usually have a sandwich-like structure, consisting of a number of slices (layers) with micromachined channels (10–100 µm in diameter). The layers perform various functions, from mixing to catalytic reaction, heat exchange, or separation. Integration of these various functions within a single unit is one of the most important advantages of microreactors. The very high heat transfer rates achievable in microreactors allow for isothermal operation of highly exothermic processes (also important in carrying out kinetic studies). The very low reaction-volume-to-surfacearea ratios make microreactors potentially attractive for carrying out reactions involving poisonous or explosive reactants (think about partial oxidation reactions).


FIGURE 14 HEX reactor developed by the BHR Group. (Courtesy: BHR Group Ltd.)

Also, microchannel heat exchangers have channel sizes around or lower than 1 mm and are fabricated via silicon micromachining, deep X-ray lithography, or nonlithographic micromachining. The reported values of heat transfer coefficients in the microchannel heat exchangers range from ca. 10,000 to ca. 35,000 W/m2K (21,44). High heat transfer coefficients, though not as high as in the previous case, are also achievable in spinning disk reactors (Figure 16). This type of reactor has been developed by Ramshaw’s group at Newcastle University and is primarily applied to fast and very fast liquid–liquid reactions with large heat effect, such as nitrations, sulphonations, and polymerizations. At very short residence times (typically 0.1 s), heat is efficiently removed from the reacting liquid at heat transfer rates reaching 10,000 W/m2K. The spinning disk reactor investigated in one of SmithKline Beecham’s processes offered a 99.9% reduction in reaction time, 99% reduction in inventory, and 93% reduction in impurity level (45). Rotational movement and centrifugal forces are used not only in spinning disk reactors. The earlier-mentioned high-gravity (HiGee) technology, started at ICI’s New Science Group in the late 1970s as a spinoff of a NASA research project in deep space (microgravity environment), has developed into one of the most promising branches of process intensification. It consists of intensifying the mass


transfer operations by carrying them in rotating packed beds, in which high centrifugal forces (typically 1000 g) occur. This way, not only mass transfer but also heat and momentum transfer can be intensified. The rotating bed equipment, originally dedicated to separation processes (such as absorption, extraction, distillation), can also be applied to reacting systems (especially those mass transfer limited). It can potentially be applied not only to gas–liquid combinations but also to other phase combinations, including three-phase gas–liquid–solid systems. The HiGee technology has already been successfully applied on a commercial scale, for deaeration of flood water in Chinese oil fields (20), where conventional vacuum towers of ca. 30-m height have been replaced by the rotating machines of ca. 1-m diameter. The earlier-mentioned hypochlorous acid technology of Dow Chemical presents another example of a successful application of rotating packed beds (25). Also, successes have been achieved in the crystallization of nanoparticles. In the group of Chong Zheng, very uniform 15- to 30-nm crystals of CaCO3 are made in a rotating crystallizer at processing times 4–10 times shorter than in the conventional stirred-tank process (46). Another interesting example of process-intensifying equipment, also undergoing commercialization, is the centrifugal adsorber. This is a new continuous device for carrying out ion exchange and adsorption processes. By using a centrifugal field for establishing countercurrent flow between the liquid phase and the adsorbent, very small adsorbent particles (10–50 µm) can be used. This allows for extremely compact separation equipment (see Figure 17) with very short contact times and high capacities (typically 10 –50 m3/h, (17)).

FIGURE 15 An example of a microreactor. (Courtesy: Institut für Mikrotechnik Mainz.)


FIGURE 16 Spinning disk reactor.

Other examples of interesting PI hardware include the supersonic gas– liquid reactor developed at Praxair Inc. (47) and the jet impingement reactor of NORAM Engineering and Constructors (Hauptmann et al. (48)). The former is based on using a supersonic shock wave to disperse gas into very tiny bubbles in a supersonic inline mixing device, while the latter uses a system of specially configured jets and baffles in order to divide and remix liquid streams with high intensity. Also, rotor/stator mixers (49) are dedicated for processes requiring very fast mixing on the micro scale. They contain a high-speed rotor spinning close to a motionless stator. Fluid passes through the region where rotor and stator interact and experiences highly pulsating flow and shear. Inline rotor/stator mixers resemble centrifugal pumps and therefore may simultaneously contribute to pumping the liquids. 5.2.

Process-Intensifying Methods

As seen in Figure 7, three well-defined categories of PI software can be distinguished: Novel processing methods, such as integration of reaction and one or more unit operations in so-called multifunctional reactors and integration of two or more separation techniques in hybrid separations Use of alternative forms and sources of energy for chemical processing Novel methods of process/plant development and operation Multifunctional reactors can be described as reactors that, alongside chemical conversion (and for the sake of it), integrate at least one more function (usually unit operation) that conventionally would have to be performed in a separate


piece of equipment. A pretty widely known example of the integration of reaction and heat transfer in a multifunctional unit are reverse-flow reactors (50). Also, a number of interesting reactor concepts for combining endo- and exothermic reactions have been developed (51,52). Reactive separations present probably the most significant class of multifunctional reactors, of which reactive distillation is one of the better-known and commercially applied examples. The multifunctional reactor is in this case a packed distillation column, in which the packing material acts simultaneously as the catalyst carrier. Chemicals are converted on the catalyst and reaction products are continuously separated by fractionation (thus overcoming equilibrium limitations). Besides the continuous removal of the reaction products and higher yields due to the equilibrium shift, the advantages of catalytic distillation units consist mainly in lower energy requirements and lower capital investment (53). Currently, numerous studies are being carried out in the field of reactive distillation modeling, as reviewed recently by Taylor and Krishna (54). Also, research on new internals for catalytic distillation columns attracts a lot of attention. Reactive distillation originates from and finds most applications in the hydrocarbon processing. Recently, interesting papers on the application of reactive distillation in fine chemistry began to emerge (55). The reverse process to the reactive distillation, i.e., reactive condensation, has also been studied (56). Reactive extraction processes involve simultaneous reaction and liquid–liquid phase separation and can be effectively utilized to obtain significant improvements in yields of desired products and selectivities to desired products in multireaction systems, thereby reducing recycle flows and waste formation. The combination of

FIGURE 17 Progress in size reduction of adsorption equipment, up to the most recent centrifugal adsorption technology. (From Ref. 118.)


reaction with liquid–liquid extraction can also be used for separation of waste byproducts that are hard to separate using conventional techniques (57,58). Reactive extraction can also be used for selective separation of amino acids (59). Reactive crystallization, or precipitation, has been investigated by numerous research groups. Processes of industrial relevance include liquid-phase oxidation of para-xylene to terephthalic acid, the acidic hydrolysis of sodium salicylate to salicylic acid, and the absorption of ammonia in aqueous sulfuric acid to form ammonium sulfate (60). A very special type of reactive crystallization is diastereomeric crystallization, widely applied in the pharmaceutical industry for the resolution of enantiomers (61). Another fine example of reactive precipitation is the earlier-described production of nano-size particles of CaCO3 in high-gravity fields (46). Reactive absorption is probably the most widely applied type of a reactive separation process. It is used for production purposes in a number of classical bulk-chemical technologies, such as nitric or sulfuric acid. It is also often employed in gas purification processes, e.g., to remove carbon dioxide or hydrogen sulfide. Other interesting areas of application include olefin/paraffin separations, where reactive absorption with reversible chemical complexation appears to be a promising alternative to the cryogenic distillation (62). Numerous investigations are being carried out in reactive adsorption processes, for instance, in chromatographic reactors, which integrate continuous countercurrent chromatographic separation with chemical reaction (63,64), and in periodic separating reactors, which are a combination of a pressure swing adsorber with a periodic flow-forced packed-bed reactor (65). This allows achieving higher conversions and better yield by separating educts and products of an equilibrium reaction from each other. In the simulated moving bed reactor (SMBR), the movement of the bed with regard to the reactant inlets/outlets is usually realized in a rotating system. One of the more interesting developments here is the rotating cylindrical annulus chromatographic reactor, shown in Figure 18 (66). In this design the inlets of the mobile phase are uniformly distributed along the annular bed entrance, while the feed stream is stationary and confined to one sector. As a result of the rotation of the reactor, the selectively adsorbed species take different helical paths through the bed and can be continuously collected at fixed locations. Another interesting example of reactive adsorption is the so-called gas– solid–solid trickle flow reactor, in which adsorbent trickles through the fixed bed of catalyst, removing selectively in situ one or more of the products from the reaction zone. In the case of methanol synthesis this led to conversions significantly exceeding the equilibrium conversions under the given conditions (67). A huge research effort is devoted nowadays to membrane reactors. The membrane can play various functions in the reactor systems (68); it can, for instance, be used for selective in situ separation of the reaction products, as a result of which an advantageous equilibrium shift can be achieved. It can also be


FIGURE 18 Rotating cylindrical annulus chromatographic reactor.

applied for a controlled distributed feed of some of the reacting species, either to increase the overall yield/selectivity of the process or to facilitate the mass transfer (e.g., direct bubble-free oxygen supply/dissolution in the liquid phase via hollow-fiber membranes (69)). The membrane can also be used for the in situ separation of catalyst particles and even homogeneous catalysts from the reaction products (70). Finally, the membrane can incorporate catalytic material, thus becoming itself a highly selective reaction-separation system. Membranes are more and more frequently employed in the life sciences sector, in manufacturing of pharmaceuticals, in combination with a bioreactor in which enzymatic reaction takes place (71). Multifunctional reactors may also combine reaction and phase transition. A well-known example of such a combination is reactive extrusion. Reactive extruders have been used increasingly in polymer industries. They enable reactive processing of highly viscous materials without a need for using large amounts of solvents, as is the case in stirred-tank reactors. Most of the reactions carried out in extruders are single- or two-phase reactions. Recently, however, new types of extruders have been investigated with catalyst immobilized on the surface of the screws, which enables carrying out three-phase catalytic reactions (72).


Fuel cells present another widely investigated type of multifunctional reactors, in which chemical reaction is integrated with the generation of electric power (73). Simultaneous gas–solid reaction and comminution in a multifunctional reactor has also been investigated (74). Among hybrid separations, the integration of membranes with another separation technique presents the most important category. In membrane absorption and stripping the membrane is used as a permeable barrier between the gas and liquid phases. By using hollow-fiber membrane modules, large mass transfer areas can be created, which results in compact equipment. Besides, absorption membranes offer operation independent of gas- and liquid-flow rates, with no entrainment, flooding, channeling, or foaming (75,76). In membrane distillation, two liquids (usually two aqueous solutions) held at different temperatures are mechanically separated by a hydrophobic membrane. Vapors are transported via the membrane from the hot solution to the cold one. The most important (potential) applications of membrane distillation are in water desalination and water decontamination (77–79). Other possible fields of application include recovery of alcohols (e.g., ethanol, 2,3-butanediol) from fermentation broths (80), concentration of oil–water emulsions (81), and removal of water from azeotropic mixtures (82). Membrane (pervaporation) units can also be coupled with conventional distillation columns, for instance, in esterifications or in production of olefins, to split the azeotrope (83,84). Membrane chromatography systems include microporous or macroporous membranes that contain functional ligands attached to their inner pore structure, which act as adsorbents. In this sense, membrane chromatography is a hybrid combination of liquid chromatography and membrane filtration. Its most important potential applications include separations of biomolecules, such as proteins, polypeptides, and nucleic acids (85,86). Among hybrid separations not involving membranes, adsorptive distillation (87) offers interesting advantages over conventional methods. In this technique a selective adsorbent is added to a distillation mixture. This increases separation ability and may present an attractive option in the separation of azeotropes or close-boiling components. Adsorptive distillation can be used, for instance, for the removal of trace impurities in the manufacturing of fine chemicals (it may allow for switching some fine chemical processes from batchwise to continuous operation). Several unconventional processing techniques using alternative forms and sources of energy have been investigated thus far and are of importance for process intensification. The use of the centrifugal fields instead of gravitational ones was discussed earlier in this chapter. On the other hand, the formation of microbubbles (cavities) in the liquid reaction medium as a result of ultrasound waves has opened new possibilities for chemical syntheses. These cavities can be thought of as high-energy microreactors. By their collapse, “microimplosions”


take place with very high local energy release (temperature rises up to 5000 K, and negative pressures up to 10,000 bar are reported (88)). This may have various effects on the reacting species, from homolytic bond breakage with free radicals formation, to the fragmentation of the polymer chains by the shock wave in the liquid surrounding the collapsing bubble. In case of solid-catalyzed (slurry) systems, the collapsing cavities can additionally affect the catalyst surface—this can, for example, be used for in situ catalyst cleaning/rejuvenation (89). Sonochemistry has also been investigated in combination with other techniques, e.g., electrolysis, in the case of the oxidation of phenol in wastewater (90). The use of solar energy in chemical processing has also been investigated. Studies describe, for example, the cycloaddition reaction of a carbonyl compound to an olefin carried out in a solar furnace reactor (91) or oxidation of 4-chlorophenol in a solar-powered fiber-optic cable reactor (92). The concept of using solar light for the synthesis of -caprolactam was evaluated, and it was shown that the return on investment was better than for the conventional technology (93). Solar reactors can also be used advantageously in water treatment plants (94). The use of microwave dielectric heating offers significant advantages for chemical synthesis (95–97). Microwave heating was shown to enable some organic syntheses to proceed up to 1240 times faster than by conventional techniques (98). The employment of electric fields to augment process rates and to control droplet size is known for a range of processes, including paint spraying, crop spraying, and coating processes. In these processes the electrically charged droplets exhibit much better adhesion properties. Electric fields can also enhance processes involving liquid/liquid mixtures, in particular liquid–liquid extraction, where rate enhancements of 200–300% were reported (99). Bioseparations (e.g., DNA separation) present another area in which electric fields can be advantageously applied (100). Interesting results have been reported concerning the so-called gliding arc technology, i.e., the use of plasma generated by the formation of gliding electric discharges (101–103). In this technology, gliding electrical discharges are produced between electrodes placed in the fast gas flow. They offer a low-energy alternative for conventional high-energy-consuming high-temperature processes. The applications tested so far in the laboratory and on industrial scale include: methane transformation to acetylene and hydrogen, destruction of N2O, reforming of heavy petroleum residues, CO2 dissociation, activation of organic fibers, air purification from volatile compounds, natural gas conversion to syngas, and SO2 reduction to elementary sulfur. A number of other methods, not falling within any of the earlier-mentioned categories, may prove useful for process intensification. Some of them, such as supercritical fluids, are already known and have been applied in other industries (104,105). Because of their unique properties, especially the high diffusion coefficient, supercritical fluids are attractive media for mass transfer operations,


FIGURE 19 Process intensification via process synthesis: methyl acetate plant of Eastman Chemical.


e.g., extraction (106), and for chemical reactions (107,108). On the other hand, cryogenic techniques (distillation or distillation combined with adsorption (109)), nowadays almost exclusively used for production of industrial gases, may in the future prove attractive for some specific separations in manufacturing of bulk or fine chemicals. Among novel methods of process/plant operation, the use of unsteady-state (periodic) operation of chemical reactors has been studied for more than three decades (110). In many processes studied on the lab scale, the intentional pulsing of flows or concentrations led to a clear improvement of product yields/ selectivities (111). Purposeful pulsing of the feed in trickle-bed reactors has been shown to bring significant improvement in mass transfer rates, in catalyst wetting, and in the radial uniformity of liquid flow (112). The commercial-scale applications of the periodic operation are scarce and are practically limited to the reverseflow reactors discussed earlier. One of the main reasons is that a stationary process has the advantage of providing constant production and product purity, without the need for additional investments to synchronize nonstationary with stationary parts of the process. Further developments in the field of advanced process control may definitely change this picture, especially where the time constant of the pulsing mode is small—synchronizing will not be problematic. Finally, in order to get a more or less complete picture of the process intensification toolbox, the PI-oriented methods for process/plant development must be mentioned. Among them, process synthesis (PS) definitely plays the most important role (113–115). Process synthesis is in some sense a sister discipline of process intensification that aims at the development of a cost-optimal process concept based on the required functionalities. It includes diverse levels of activities, starting from basic conceptual plant design (often based on the “out-of-thebox” approach), through the selection of optimal pieces of equipment and optimal interconnections between them (plant integration), up to cost estimates. Process synthesis permits early assessment and evaluation of the manufacturability of products resulting from potential new chemistries. A textbook example of a commercial application of process synthesis is the methyl acetate plant of Eastman Chemical (23), in which a task-oriented integration of reaction and separation in a multifunctional reactor reduced the number of pieces of equipment from 28 to 3 (see Figure 19). In recent years, process synthesis has gone beyond its traditional field of applications and has entered new sectors, such as bioprocessing and pharmaceuticals manufacturing (116,117). REFERENCES 1. 2.

Agricola G. De Re Metallica Libri XII. Basel: J. Froben & N. Episopius, 1556. Taber RE, Hawkinson DR. Fluid Mixing Device. US Pat. 2,894,732, Shell Development Company, 1959.


3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

16.

17.

18.

19. 20.

21. 22. 23. 24.

25.

Sulzer Chemtech. Mixing and Reaction Technology. Technical Information Brochure. Winterthur, Switzerland: 1997. Chvatal V. Hutnicke Listy 1965; 19:466–474. Chervotkin VV, Bokovikov BA, Gogolev YF, Privalov SI. Sb Nauch Tr, Vses Nauch-Issled Inst Met Teplotek 1969; 16:108–118. Grekov EA. Sb Tr Tsent Nauch-Issled Inst Chern Met 1970; 75:77–84. Kolpakov SV, Frantsenyuk IV, Teder LI, Beremblyum GB. Stal’ 1971; 31:401–402. Leszczynski Z. Przem Chem 1973; 52:161–163. Romankov PG. Khim Ind 1977; 49:82–89. Kleemann G, Hartmann K. Wiss Z Tech Hochsch “Carl Schorlemmer” LeunaMerseburg 1978; 20:417–425. Ramshaw C. Chem Engr (London) 1983; 389:13–14. Heggs P. Chem Engr (London) 1983; 394:13. Ramshaw C. Chem Engr (London) 1985; 416:30–33. Proc Eng (London) 1988; 69 (Issue: Energy Efficiency Suppl.): S13–S16. Ramshaw C, ed. 1st International Conference on Process Intensification for the Chemical Industry, Antwerp, Belgium, Dec 6–8, 1995. BHR Group Conference Series, Publication No. 18. London: Mechanical Engineering Publications Limited, 1995. Stankiewicz A, Smits HA, Glasz WC, Moulijn JA. In: C Ramshaw, ed. Proceedings of the 1st International Conference on Process Intensification for the Chemical Industry. BHR Group Conference Series, No. 18. London: Mechanical Engineering Publications Limited, 1995:167–171. Bisschops MAT, van der Wielen LAM, Luyben KChAM. In: Semel J, ed. Proceedings of the 2nd International Conference on Process Intensification in Practice. BHR Group Conference Series, No. 28. London: Mechanical Engineering Publications Limited, 1997:299–307. Thonon B. In: Ramshaw C, ed. Proceedings of the 1st International Conference on Process Intensification for the Chemical Industry. BHR Group Conference Series, No. 18. London: Mechanical Engineering Publications Limited, 1995:37–47. Ehrfeld W, Hessel V, Löwe H. Microreactors. Weinheim:Wiley-VCH, 2000. Zheng C, Guo K, Song Y, Zhou X, Ai D. In: Semel J, ed. Proceedings of the 2nd International Conference on Process Intensification in Practice. BHR Group Conference Series, No. 28. London: Mechanical Engineering Publications Limited, 1997:273–287. Tonkovich ALY, Call ChJ, Jimenez DM, Wegeng RS, Drost MK. AIChE Symp Ser 1996; 92(310):119–125. Quiram DJ, Jensen KF, Schmidt MA, Mills PL, Ryley JF, Wetzel MD. AIChE Symp Ser 2000; 96(323):147–162. Siirola JJ. AIChE Symp Ser 1995; 91(304):222–233. Meili A. In: Semel J, ed. Proceedings of the 2nd International Conference on Process Intensification in Practice. BHR Group Conference Series, No. 28. London: Mechanical Engineering Publications Limited, 1997;309–318. Trent D, Tirtowidjojo D. In: Gough M, ed. Proceedings of the 4th International Conference on Process Intensification for the Chemical Industry. Cranfield, U.K.: BHR Group Ltd., 2001:11–19.


26. 27. 28. 29. 30.

31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41.

42. 43.

44.

45. 46.

47.

48.

49.

Chisti Y, Moo-Young M. Trans IChemE 1996; 74(Part A):575–583. Hamilton GE, Morton PhH, Young TW, Lyddiatt A. Biotechnol Bioeng 1999; 64: 310–321. Stewart GG. Eur Brew Conv 2000; 28:94–114. Laird T. Org Proc Res Dev 2001; 5:612. Stankiewicz A. Between the chip and the blast furnace. Process intensification in industry and in academia. UEF Conference “Refocusing Chemical Engineering,” Barga, Italy, May 27–June 1, 2001. Douglas JM. Conceptual Design of Chemical Processes. New York: McGraw-Hill, 1988. The future is now! DSM Magazine 1999; 148:4–10. Elverding P. DEFACTO 2001; 15(5):28–32. Hendershot DC. Chem Eng Prog 2000; 96:35–40. Ramshaw C. The desktop fine chemical process. TG Lustrum Congress 2000, “Honey, I Shrunk the Plant,” Delft, Nov 8–9, 2000. Goldhill J, Arthur D. Little Chemical Executive 2001; 1:1–2. Stringaro J-P, Collins P, Bailer O. In: Cybulski A, Moulijn JA, eds. Structured Catalysts and Reactors. New York: Marcel Dekker, 1998:393–416. Manteufel RPC. Surer 3X-Pack and Super 4X-Pack. A new generation of structured packings. ACHEMA, Frankfurt, May 25, 2000. Kapteijn F, Heiszwolf JJ, Nijhuis TA, Moulijn JA. CATTECH 1999; 3(1):24–41. Stankiewicz A. Chem Eng Sci 2001; 56:359–364. Phillips CH. In: Green A, ed. Proceedings of the 3rd International Conference on Process Intensification for the Chemical Industry. BHR Group Conference Series, No. 38. London: Professional Engineering Publishing Ltd., 1999:71–87. Calder R. BHR News. 2000; Summer:4. Jäckel K-P. In: Microtechnology: Application Opportunities in the Chemical Industry. DECHEMA Monographs Vol. 132. Frankfurt am Main: DECHEMA, 1995:29–50. Thonon B, Mercier P. In: Semel J, ed. Proceedings of the 2nd International Conference on Process Intensification in Practice. BHR Group Conference Series, No. 28. London: Mechanical Engineering Publications Limited, 1997:49–62. Oxley P, Brechtelsbauer C, Ricard F, Lewis N, Ramshaw C. Ind Eng Chem Res 2000; 39:2175–2182. Chen J, Wang Y, Jia Z, Zheng C. In: Semel J, ed. Proceedings of the 2nd International Conference on Process Intensification in Practice. BHR Group Conference Series, No. 28. London: Mechanical Engineering Publications Limited, 1997:157–161. Cheng ATY. In: Semel J, ed. Proceedings of the 2nd International Conference on Process Intensification in Practice. BHR Group Conference Series, No. 28. London: Mechanical Engineering Publications Limited, 1997:205–219. Hauptmann EG, Rae JM, Guenkel AA. In: Ramshaw C, ed. Proceedings of the 1st International Conference on Process Intensification for the Chemical Industry. BHR Group Conference Series, No. 18. London: Mechanical Engineering Publications Limited, 1995:181–184. Bourne JR, Studer M. Chem Eng Proc 1992; 31:285–296.


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. 79. 80. 81. 82. 83. 84.

Matros YS, Bunimovich GA. Catal Rev-Sci Eng 1996; 38:1–68. Frauhammer J, Eigenberger G, von Hippel L, Arntz D. Chem Eng Sci 1999; 54: 3661–3670. Kolios G, Frauhammer J, Eigenberger G. Chem Eng Sci 2000; 55:5945–5967. Stadig WP. Chem Processing 1987; 50:27–32. Taylor R, Krishna R. Chem Eng Sci 2000; 55:5183–5529. Clarkson JS, Walker AJ, Wood MA. Org Proc Res Dev 2001; 5:630–635. Ben Amor H, Halloin VL. Chem Eng Sci 1999; 54:1419–1423. Minotti M, Doherty M, Malone MF. Ind Eng Chem Res 1998; 37:4746–4755. Samant KD, Ng KM. Chem Eng Technol 1999; 22:877–880. Cascaval D, Oniscu C, Galction A-I. Biochem Eng J 2001; 7:171–176. Kelkar VV, Ng KM. AIChE J 1999; 45:69–81. Bayley CR, Vaidya NA. In: Collins AN, Sheldrake GN, Crosby J, eds. Chirality in Industry. New York: Wiley, 1992:69–77. Safarik DJ, Eldridge RB. Ind Eng Chem Res 1998; 37:2571–2581. Dünnebier G, Fricke J, Klatt K-U. Ind Eng Chem Res 2000; 39:2290–2304. Lode F, Houmard M, Migliorini C, Mazzotti M, Morbidelli M. Chem Eng Sci 2001; 56:269–291. Vaporciyan GG, Kadlec RH. AIChE J 1989; 35:831–844. Herbsthofer R, Bart H-J, Wolfgang J, Prior A. Chem Ing Tech 2001; 73:56–59. Westerterp KR, Bodewes TN, Vrijland MS, Kuczynski M. Hydrocarbon Processing 1988; 67:69–73. Sirkar KK, Shanbhag PV, Kovvali AS. Ind Eng Chem Res 1999; 38:3715–3737. Shanbhang PV, Guha AK, Sirkar KK. J Hazardous Materials 1995; 41:95–104. Huizenga P. The continuously filtering slurry reactor. Ph.D. dissertation, University of Twente, Enschede, Netherlands, 1998. Wöltinger J, Drauz K, Bommarius AS. Appl Cat A: General 2001; 221:171–185. Ebrahimi-Moshkabad M, Winterbottom JM. Catalysis Today 1999; 48:347–355. Carrette L, Friedrich KA, Stimming U. Chem Phys Chem 2000; 1:162–193. Uhde G, Sundmacher K, Hoffmann U. Proceedings of the 5th World Congress of Chemical Engineering, San Diego, 1996: Vol. I, 167–172. Poddar TK, Majumdar S, Sirkar KK. J Membrane Sci 1996; 120:221–237. Jansen AE, Klaassen R, Feron PHM. In: Ramshaw C, ed. Proceedings of the 1st International Conference on Process Intensification for the Chemical Industry. BHR Group Conference Series, No. 18. London: Mechanical Engineering Publications Limited, 1995:145–153. Banat FA, Simandl J. Sep Sci Technol 1998; 33:201–226. Couffin N, Cabssud C, Lahoussine-Turcaud V. Desalination 1998; 117:233–245. Banat FA, Simandl J. Chem Eng Sci 1996; 51:1257–1265. Qureshi N, Meagher MM, Hutkins RW. Sep Sci Technol 1994; 29:1733–1746. Gryta M, Karakulski K. Desalination 1999; 121:23–29. Udriot H, Araque A, von Stockar U. Chem Eng J 1994; 54:87–93. Parkinson G, Ondrey G. Chem Eng 2001; 108(11):31–37. Al-Rabiah AA, Timmerhaus KD, Noble RD. Proceedings of the 5th World Congress on Chemical Engineering, San Diego, California, 1996: Vol. IV, 335–340.


85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101.

102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113.

Roper DK, Lightfoot EN. J Chromatogr A 1995; 702:3–26. Zeng X, Ruckenstein E. Biotechnol Prog 1999; 15:1003–1019. Yu KT, Zhou M, Xu CJ. Proceedings of the 5th World Congress of Chemical Engineering, San Diego, 1996: Vol. I, 347–352. Mason TJ. Practical Sonochemistry. User’s Guide to Applications in Chemistry and Chemical Engineering. New York: Ellis Horwood, 1991. Mikkola J-P, Salmi T. Chem Eng Sci 1999; 54:1583–1588. Trabelsi F, Aït-Lyazidi H, Ratsimba B, Wilhelm AM, Delmas H, Fabre P-L, Berlan J. Chem Eng Sci 1996; 51:1857–1865. Pohlmann B, Scharf H-D, Jarolimek U, Mauermann P. Sol Energy 1997; 61:159–168. Peill NJ, Hoffmann MR. J Sol Energy Eng 1997; 119:229–236. Sattler C, Müller F-J, Riffelman K-J, Ortner J, Funken K-H. J Phys IV France 1999; 9:Pr3 723–727. Alfano OM, Bahnemann D, Cassano AE, Dillert R, Goslich R. Catalysis Today 2000; 58:199–230. Mingos DPM, Baghurst DR. Chem Soc Rev 1991; 20:1–47. Galema SA. Chem Soc Rev 1997; 26:233–238. Gabriel C, Gabriel S, Grant EH, Halstead BSJ, Mingos DMP. Chem Soc Rev 1998; 27:213–223. Gedye RN, Smith FE, Westaway KC. Can J Chem 1988; 66:17–26. Weatherley LR. Heat Recovery Systems CHP 1991; 13:515–537. Park YG. Biochem Eng J 2001; 7:213–221. Czernichowski A, Czech T. Proceedings HAKONE III, 3rd International Symposium on High-Temperature Low-Pressure Plasma Chemistry, Strasbourg, France, 1991: 147–152. Czernichowski A, Leuseur H. Proceedings Symposium “Plasma Applications to Waste Water Treatment,” Idaho Falls, ID, 1991:1–13. Czernichowski A, Polaczek J, Czech T. Proceedings ISPC-11, 11th International Symposium on Plasma Chemistry, Loughborough, 1993:674–679. McClain J. Chem Eng 2000; 107(2):72–79. Marr R, Gamse T. Chem Eng Proc 2000; 39:19–28. McHugh MA, Krukonis VJ. Supercritical Fluid Extraction. Boston: ButterworthHeinemann, 1994. Savage PE, Gopalan S, Mizan TI, Martino CJ, Brock EE. AIChE J 1995; 41: 1723–1178. Hyde JR, Licence P, Carter D, Poliakoff M. Appl Catal A: General 2001; 222:119–131. Jain R, Tseng JT. Production of high-purity gases by cryogenic adsorption. AIChE 1997 Annual Meeting, Paper 33b. Los Angeles, 1997. Zwijnenburg A, Stankiewicz A, Moulijn JA. Chem Eng Prog 1998; 94(11):39–47. Silveston PL. Composition Modulation in Catalytic Reactors. Amsterdam: Gordon & Breach, 1998. Boelhouwer JG, Piepers HW, Drinkenburg AAH. Chem Eng Sci 2001; 56(8): 2605–2614. Han C, Stephanopulos G, Liu YA. AIChE Symp Ser 2000; 96(323):148–159.


114. 115. 116. 117. 118.

Terril DL. AIChE Symp Ser 2000; 96(323):329–332. Johns WR. Chem Eng Prog 2001; 97(4):59–65. Steffens MA, Fraga ES, Bogle IDL. Comput Chem Eng 1999; 23:1455–1467. Johnson DB, Bogle IDL. Chem Papers 2000; 54:398–405. Bisschops MAT. Centrifugal adsorption technology. Ph.D. dissertation, Delft University of Technology, Delft, Netherlands, 1999.


2 Chemical Processing in High-Gravity Fields David L. Trent The Dow Chemical Company, Freeport, Texas, U.S.A.

1.

INTRODUCTION

The use of high-gravity, or centrifugal, fields for chemical processing has generated much interest in recent years. Fluid acceleration creates an environment in which mass transfer rates are two to three orders of magnitude higher than rates achieved in more conventional equipment, such as packed towers and stirred tanks. Heat transfer is also enhanced. Short contact time and fast transfer rates allow a reduction in equipment size and in-process inventory. Many chemical processes could benefit from these unique properties by reducing the cost of construction, reducing working capital, improving safety, producing less waste, etc. In addition, the use of high-gravity fields may provide solutions to processing problems more effectively and more economically than conventional equipment. Before exploring the wide range of applications, a brief review of the history of development, a discussion of the process fundamentals, and an introduction to mechanical design issues will help to set the boundaries for use of high-gravity fields in chemical processing. 2.

HISTORICAL DEVELOPMENT

The use of centrifugal fields in chemical processing dates to the beginning of the industry with such physical transport operations as pumping, compression, and solid/liquid separations. Extending this use into mass and heat transport operations


such as liquid/liquid extraction, gas/liquid interactions, reactions, crystallization, and heat transfer, however, is a much more recent development. Although the commercial application of high-gravity fields has been limited in these recent areas of interest, the potential to improve existing chemical processes and develop new processes remains high. The extension of the use of centrifugal fields into the commonplace of chemical processing had its start in 1945 with the commercial application of a centrifugal liquid extractor to the recovery of penicillin (1). The liquid extractor, based on earlier patents by Podbielniak (2), employed perforated concentric plates for contacting the two countercurrent liquid streams. Pilo and Dahlbeck (3) introduced an apparatus employing a variety of rotor internals and liquid distributors for “countercurrent contact of two fluids having different specific gravities” in their 1960 U.S. patent. Although the authors suggest use of the apparatus for gas scrubbing, distillation, heat exchange, and reactions, commercial application was not exploited. In 1966 Podbielniak (4) described a centrifugal device with concentric perforated plates for gas/liquid contact. Todd (5) extended the concepts of Pilo and Podbielniak to a multistage device in 1969. In 1981 Ramshaw and Mallinson (6) provided for filling the rotor with a high-surface-area material such as glass beads or wire gauze in order to effect improved mass transfer. They also described a rotor for distillation, complete with vapor/liquid contacting, condenser, and boiler. Following this patent, considerable academic interest developed in an attempt to exploit the high mass and heat transfer rates, high-throughput capability, and short contact times afforded by countercurrent fluid contact using centrifugal fields. Sustained commercial application, however, did not occur until 1997, when Zheng et al. (7) described the successful stripping of oxygen from water used in secondary oil field recovery. In 1999 Trent et al. (8,9) introduced the first commercial application involving simultaneous absorption, reaction, and stripping. Both of these involve gas/liquid contact using a woven wire screen for the rotor internals. 3.

PROCESS FUNDAMENTALS

In order to fully appreciate the application opportunities available from use of high-gravity fields, an understanding of what happens within the rotor would be helpful. Since many of the models used to describe chemical processes in more conventional equipment do not apply to high gravity, much of the information available is empirical. However, where possible the traditional models are being modified to match the observed behavior. The process fundamentals of interest here are mass transfer, heat transfer, fluid distribution and holdup, flooding, pressure drop, power requirements, and rotor internals. Although this last item could be considered with the mechanical design discussion, it is included here because the rotor internals can have a significant impact on the other process variables.


FIGURE 1 Schematic of gas continuous operation of a rotating packed-bed gas/ liquid contactor.

Before addressing these process fundamentals, however, a description of the basic process equipment would be helpful. Figure 1 shows a schematic of a simplified rotating packed-bed (RPB) contactor. This RPB illustrates countercurrent gas/liquid operation. Liquid enters at the eye of the rotor, being distributed on the rotor packing at the inside diameter. The centrifugal force of the spinning rotor accelerates the liquid radially outward. Gas enters the stationary housing and passes through the rotor from outside to inside. The gas exits at the eye of the rotor, while liquid drains from the housing. Seals on the drive shaft and on the rotor ensure that the gas moves through the rotor. Figure 1 shows a gas continuous configuration. A liquid continuous arrangement is also possible (10). For liquid/liquid extraction the preceding description applies if we consider the liquid to be the heavy phase and the gas to be the light phase (11). In both of these scenarios the light phase enters through the drive shaft and channels radially in the rotor end plate to the outer periphery of the rotor for distribution into the heavy phase (see Figure 2). 3.1.

Hydrodynamics

Understanding the flow of liquid and gas through a rotating packed bed is important to understand the performance results achieved. Liquid flow involves two


FIGURE 2 Schematic of liquid continuous operation of a rotating packed bed for gas/liquid and liquid/liquid contacting.

components, liquid introduction to the packing and flow through the packing. Flow through the packing occurs in the radial direction with very little tangential or axial spreading (see Figure 3). A slight curvature in the radial flow results from the direction of rotation. The degree of curvature and spreading is primarily a function of rotor speed and liquid viscosity and less a function of packing type and liquid flow rate (12,13). Gas flow does not impact liquid flow through the rotor (14). The flow pattern described results in incomplete wetting of the packing at the outer diameter of the rotor (13). Thus, not all of the packing surface area is utilized for mass transfer operations. For maximum use of the packing surface area, the ratio of the outside diameter to the inside diameter should be minimized. Scale-up from small-diameter to large-diameter rotors typically provides more efficient use of the packing. In spite of this incomplete wetting of the packing, very high mass transfer rates are achieved. Enhanced mass transfer performance results from the initial contact of the liquid feed with the rotor. Studies at the Higravitec Center of Beijing University of Chemical Technology (HCBUCT), using a video camera attached to the rotor, revealed a breakup of the liquid feed into smaller droplets that filled the void spaces


of the packing. This effectively increases the interfacial surface area of the liquid beyond that of the surface area of the packing. At a packing depth of about 7–10 mm, most of the liquid has been accelerated to the rotor speed and wets the packing. Liquid flow continues mostly along the packing surface, but some additional droplets fly across the void spaces in the packing. The most intense mixing and mass transfer occur in the inlet zone of the packing. The degree of mass transfer enhancement at the inlet is a function of the type of packing (porosity, shape of packing structure, etc.), rotor speed, method of liquid distribution, and liquid properties (15). Liquid distribution on the rotor affects the initial contact zone performance. Some of the variables to consider include angle of impingement, velocity of the liquid spray, and acceleration of the liquid via rotation. Optimum performance requires full axial wetting of the packing, whereas full tangential wetting is not necessary. The use of nozzles rotating in the same direction and at the same speed as the rotor gave poor results for a hypochlorous acid process due to little surface area or liquid-side mass transfer enhancement (8). Although nozzles rotating in the opposing direction of the rotor would be expected to provide the best mass transfer performance, the increased cost of manufacture of the equipment may not be justified. Gas distribution in the rotor has not been studied as thoroughly as that of the liquid. Gas entering the rotor at the outside diameter accelerates radially inward due to the reducing diameter. Gas tangential velocity relative to that of the rotor varies depending on the rotor packing. With parallel flat plates the low frictional drag (high slippage) makes the gas spiral inward with a path length much longer than the radial thickness of the packing (Figure 4a). The gas path length approaches the radial thickness of the packing as the packing surface area increases and the

FIGURE 3 General flow pattern of liquid in a gas continuous rotor.


FIGURE 4 Relative path of gas (a) in a rotor of low resistance (e.g. parallel flat plates), and (b) in a rotor of high resistance (low porosity, high surface area).

porosity decreases, due to the increased drag of the rotor on the gas (Figure 4b). This behavior significantly impacts gas-side mass transfer performance (16). The thickness of the liquid film on the rotor packing helps determine mass transfer rates. Film thickness can be shown to be inversely proportional to rotor speed to the 0.8 power (17). Visual measurements using a video camera attached to the rotor show a water film thickness of 20–80 microns on foam metal packing and 10 microns on wire gauze packing (15). Theoretical models estimate similar film thickness values (13,18,19). Film flow is expected to be laminar. In addition to rotor speed, liquid flow rate and fluid properties affect the film thickness (14).


The foregoing discussion focuses on gas continuous operations, in which the bulk of the rotor volume is filled with gas. Liquid continuous operation can be accomplished by collecting the liquid at the outside diameter of the rotor and channeling it back for discharge at a diameter slightly larger than the feed radial position, as illustrated in Figure 2 (20). Gas introduced into the liquid pool at the outer diameter moves countercurrent to the liquid inside the packing of the rotor. Because of the hydraulic pressure of the liquid from the centrifugal force, the gas bubbles start out small and steadily expand. Contact of the bubbles with the rotor packing results in breakup of the bubbles, to maintain a high interfacial surface area. References 21 and 22 illustrate this approach for water deaeration and a centrifugal field bioreactor, respectively. Liquid–liquid contact in an RPB involves introduction of the heavy liquid at the inside diameter of the rotor and the light phase at the outside diameter. The two liquid phases move countercurrent to one another within the rotor packing. Centrifugal force causes the heavy liquid to move radially outward. This displaces the light liquid, which moves radially inward. The design of the rotor packing influences the contact between the two liquid phases (23,24). 3.2.

Flooding

Rotating packed-bed devices handle high volumes of fluids in a small equipment volume, compared to packed towers, due to the acceleration of gravity. The Sherwood flooding correlation for packed towers (25) is expressed as a plot of UG2 a p  G  0.2   L gε 3  L 

versus

L  G    G  L 

0.5

Early RPB researchers discovered that this flooding correlation for packed towers applied equally well to RPBs when the gravity term (g) was replaced by centrifugal acceleration (r2). As acceleration increases, the gas flooding velocity (UG) increases in order to maintain the same value of the first term. Since the ratio of liquid (L) to gas (G) flow remains constant, liquid flow increases commensurately with gas flow. Most researchers observed higher gas velocities before the onset of flooding than predicted by the Sherwood correlation (17,26,27). Measurement of flooding by traditional means of observing a sharp pressure increase as gas rates increase is not effective with the RPB (28). Flooding can be determined experimentally by adjusting rotor speed and holding gas and liquid rates constant. Flooding will occur at the point of maximum pressure drop over a range of rotor speeds (28) or by observing the increased pressure drop change (inflection point) as a function of decreased rotor speed (27,29). In liquid–liquid contact, two types of flooding can occur. The design of the centrifugal contactor defines the throughput capability before flooding occurs. The second type of flooding relates to the principal interface moving into either


the light-phase or heavy-phase takeoff. The movement of the principal interface is controlled by the back pressure of the light-phase takeoff, which is a function of the rotor speed and phase density difference (24). 3.3.

Residence Time

Liquid residence time in the packed rotor varies as a function of packing depth, packing type, rotor speed, and liquid properties (26). Two basic approaches have been applied to the measurement of liquid in the rotor. The first measure is the average residence time of the liquid within the rotor, and the second is the liquid holdup on the packing. Due to the flow patterns described previously, not all of the rotor packing is wetted and not all of the liquid resides on the packing surface. As a result, average residence time and liquid holdup are distinct measures of liquid flow, contrary to the experience with packed towers. Tracer methods, both visual and electrical conductivity sensors, have been applied to measure the residence time of the liquid in the rotor (15,26). Measured liquid residence time ranges from about 0.2 seconds to about 1.8 seconds. Time decreases as the rotor speed increases, as liquid flow rate increases, and as the radial position increases. Gas flow rate and liquid viscosity (narrow test range) have little impact on residence time (15,19,26). Since liquid does not completely wet the packing and since film thickness varies with radial position, classical film-flow theory does not explain liquid flow behavior, nor does it predict liquid holdup (30). Electrical resistance measurements have been used for liquid holdup, assuming liquid flows as rivulets in the radial direction with little or no axial and transverse movement. These data can then be empirically fit to film-flow, pore-flow, or droplet-flow models (14,19). The real flow behavior is likely a complex combination of these different flow models, that is, a function of the packing used, the operating parameters, and fluid properties. Incorporating calculations for wetted surface area with the film-flow model allows prediction of liquid holdup within 20% of experimental values (18). Liquid holdup in liquid–liquid extractors must be defined for both the heavy and light phases. The light-phase outlet pressure is used to control the relative liquid holdup of the two phases. Higher light-phase outlet pressure increases the light-phase holdup. This pressure has been correlated with the phase density difference, rotor speed, and rotor dimensions (24). In addition, packing characteristics of volumetric surface area and porosity influence liquid holdup and throughout capability (23). 3.4.

Mass Transfer

Developing correlations to describe mass transfer in rotating packed beds has proven to be a challenge. Penetration theory (31), film-flow theory (32), and modified surface-renewal theory (12) are some examples of leveraging previous work


to the rotating packed bed. At issue is the need to include a multiplicity of variables as well as to understand the hydrodynamics of gas/liquid contact within the rotor. Each modeling approach compared experimental data to the models, with reasonable fit. However, these expressions have not yet been validated on multiple-RPB designs and operating conditions. A key to developing a generalized model for mass transfer performance is the proper understanding and treatment of the fluid flows within the rotor, as discussed previously in Section 3.1. The combination of high surface area, high velocities, thin films, and intense mixing in the packing provides an environment for intensive mass transfer, resulting in values for height of transfer unit (HTU) of 1.5–4 cm (26). Mass transfer has been described using HTU, number of transfer units (NTU), mass transfer coefficient (kL, kG), and volumetric mass transfer coefficient (kLa, kGa, k S a). To accommodate the variation in packing surface wetting with radial distance, an area transfer unit (ATU) has been proposed (33). Another proposed method of evaluation uses a volume transfer unit (VTU) to account for the entire volume of the packed rotor (34). Although the ATU and VTU methods may have merit in evaluating RPB performance, these methods make comparison with other transfer devices based on HTU more difficult. The possible physical design parameters affecting mass transfer include packing and packing supports. Atomization of the liquid as it impacts the spinning rotor packing creates high-surface-area liquid drops, in addition to the film wetting of the high-surface-area packing. This atomization results in significant mass transfer apart from the packing surface. As a result, low-surface-area packings produce equivalent, if not better, volumetric mass transfer coefficients than do high-surface-area packings (35,36). This implies that low-surface-area packing with high porosity can effectively replace high-surface-area packing, contrary to the experience with packed towers. The result is lower-cost packing, reduced pressure drop, and higher throughput. Packing supports at the inside diameter of the rotor generally provide a positive effect on mass transfer (36). Due to the atomization of liquid exiting the rotor, additional mass transfer occurs in the space between the rotor and the housing (37). Operational parameters of importance to mass transfer include rotor speed, liquid rates, and gas/liquid ratios. Mass transfer increases proportionately to rotor speed, decreases with increasing liquid flow, and increases with gas/liquid ratio (17,26,36). Although most references present rotor speed as revolutions per minute (rpm), expression as either tangential velocity (r) or multiples of gravity (r2) provides a better basis for comparison among the different rotor designs and for scale-up. Gas-side mass transfer in rotating packed beds does not show the same level of enhanced performance as liquid-side mass transfer. Average volumetric gas mass transfer values for a wire screen packing increased with gas flow rate but decreased with increased rotor speed. Compared to a packed tower, the RPB


mass transfer coefficient (1–8 s1) was similar when operated at similar superficial gas velocities (1 m/s). However, when gas velocities were increased (4–12 m/s) in studies with parallel flat plates as the rotor, the mass transfer coefficient also increased, to a high of 45 s1 (16). In a commercial-scale RPB, average volumetric gas-side mass transfer coefficients of 40–50 s1 were achieved using a wire screen packing and gas velocities of 4–5 m/s (9). The factors affecting gasside mass transfer are less understood than those of liquid-side mass transfer. Liquid–liquid mass transfer in centrifugal extraction contactors shows similar trends on performance as the gas–liquid contactors. Mass transfer improves at higher rotor speeds, higher solvent ratio, and higher phase density difference. Since the light-phase outlet pressure controls the liquid holdup of both phases, decreasing the light-phase outlet pressure decreases the light-phase holdup and increases the number of transfer units (24). Packing characteristics of pore size, porosity, and volumetric surface area also play a role in performance (23). Single centrifugal extractors have achieved up to 10 theoretical stages of extraction (38), but they could achieve up to 20 stages with suitable rotor design (11). Liquid–solid mass transfer has also been studied, on a limited basis. Application to systems with catalytic surfaces or electrodes would benefit from such studies. The theoretical equations have been proposed based on film-flow theory (32) and surface-renewal theory (39). Using an electrochemical cell with rotating screen disks, liquid–solid mass transfer was shown to increase with rotor speed and increased spacing between disks but to decrease with the addition of more disks (39). Water flow over naphthalene pellets provided 4–6 times higher volumetric mass transfer coefficients compared to gravity flow and similar superficial liquid velocities (17). 3.5.

Pressure Drop

Gas pressure drop through the RPB rotor is an important consideration when comparing the performance of the RPB with other mass transfer devices, such as a packed tower. Numerous studies on pressure drop in RPB rotors employing a variety of packings have yielded some surprising differences from conventional packed towers. For example, lower pressure drop for wetted packing compared to dry packing has been reported (26,40,41). Not all researchers observed this phenomena, because pressure drop was found to be a function of packing type, rotor design, gas rates, liquid rates, and rotor speed (41). Pressure drop has been reported for a number of rotor internals, including corrugated structured packing (28), foam metal (26,40,42), rectangular and elliptical cylinder plastic grains randomly packed (41), wire screen (43), and glass beads (17). In spite of the variation in porosity from 0.38 to 0.95 and in volumetric surface area from 500 to 4000 m2/m3, all of these studies showed similarities of increased pressure drop as rotor speed increased and gas rates increased.


In addition to the differences in packing type and characteristics, these studies also had a variety of rotor and housing designs for the gas and liquid inlet and outlet. All of them measured a gas pressure drop from the inlet gas piping or housing to the outlet gas pipe. In an effort to develop models of the pressure drop, most have attributed the pressure drop to centrifugal and frictional forces in the rotor, using a film-flow model (43). Pressure drop is proportional to the square of rotor speed (26,42). Compared to conventional packed towers, the pressure drop is lower per NTU (26) and about 15 times higher at flooding conditions (28). To account for the differences in machine configuration and to better explain the pressure-drop observations (e.g., lower pressure drop with onset of liquid flow), a model based on conservation of mass and momentum, in particular gas angular momentum, was developed (40). This model divided the pressure drop into four increments that included the gas inlet to the machine housing, the rotor, the eye of the rotor, and the gas exit nozzle from the machine. Although the models for pressure drop have a basis in theory, all are fit empirically to data generated from specific equipment. Application of these models may not be relevant to machines of different configurations and packing types (9). In liquid–liquid contactors, pressure drop is defined by the light phase. The heavy phase enters at near atmospheric pressure and is accelerated by the rotor to its discharge pressure. The pressure drop of the light phase is a function of phase density difference, rotor speed, rotor diameter, and location of the principal phase interface (24). 3.6.

Heat Transfer

Most of the experimental work on heat transfer in centrifugal fields has been done on spinning discs, which is the subject Chapter 3 in this book. A brief review of the enhanced heat transfer rates is relevant here. Studies on a smooth, flat spinning disc show heat transfer coefficients as high as about 20 kW/m2K. The coefficient is highest at the inlet to the disc, due to disturbances as the liquid is accelerated to the angular velocity of the disc. The heat transfer coefficient increases with increased rotor speed but decreases with increased radial position. Higher-viscosity fluids decrease the heat transfer rate. Heat transfer is generally higher than predicted from film or penetration theory (44). Modifying the surface of the disc allows enhancement of the heat transfer by introducing liquid film instabilities (waves). Best results come from the combination of thin films and large instabilities as revealed from the study of four surface geometries: smooth, sprayed metal, and two types of concentric grooves. In general, increased rotor speed gives higher heat transfer. However, with the grooved disc, high speeds cause liquid separation from the disc, resulting in lower heat transfer (45).


Heat transfer involving non-Newtonian fluids has not been studied in rotating devices. Models have been developed for gravity-driven heat transfer for power-law fluids (46). These models may be useful as a starting point to evaluate performance in higher-gravity fields. Another method of introducing heat to fluids in rotating devices involves the generation of eddy currents by rotation through a stationary magnetic field. This approach was successfully used in a polymer devolatilization process (47). 3.7.

Power

Estimating power consumption in rotating systems depends on several factors, including acceleration of the liquid, windage effect of gas drag on the rotor, friction in the bearings and seals, and gas pressure drop (33,48). As would be expected from power consumption in such devices as centrifugal pumps, the largest power component involves acceleration of the liquid to the angular velocity of the rotor at the outer diameter. Gas pressure drop actually decreases power consumption in the rotor. Frictional losses are defined by the design of the rotor, bearings, and seals. 3.8.

Rotor Internals

The rotor packing has an impact on all of the previously mentioned process fundamentals. Hydrodynamics, especially at the liquid inlet to the packing, is a function of packing porosity and volumetric surface area (12,13). These same packing properties influence pressure drop, residence time, and flooding velocity (26). Liquid-side mass transfer performance is best with wire gauze as compared to glass beads or parallel flat plates (8). Gas-side mass transfer is better with parallel flat plates than in wire gauze (16). Flat plates provide the best medium for heat transfer (34). 4.

MECHANICAL DESIGN

Since the development of high-gravity fields requires rotating equipment, the mechanical design is very important when considering operating performance, cost of design and fabrication, ease of maintenance, and overall reliability. Although most public reports on RPB studies describe the particular RPB design used in the reported studies, very little information has been published on the mechanical design principles. Original equipment manufacturers of rotating equipment provide an effective resource for proper design and fabrication of RPBs. The following discussion outlines some of the basic issues to be considered in the machine design. Overriding all of the following discussion is the need to design a stable rotor with minimal vibration under the desired operating conditions.


4.1.

Cantilevered Versus Centerhung Rotor

Once the rotor dimensions of inside diameter, outside diameter, and axial height have been determined from the operating performance requirements, the rotor orientation on the shaft must be determined. Two options are available, cantilevered, also called overhung, and centerhung. These relate to the position of the rotor relative to the shaft and the support bearings. The cantilever design places the rotor at the end of the shaft, while the centerhung design positions the rotor in the middle of the shaft, with bearings on either side of the rotor. Often the determining factor for selection is the ratio of axial height (AH) to outside rotor diameter (OD). The conservative approach limits cantilever selection to AH/OD 0.5, though designs with ratios up to about 0.85 are possible. Numerous examples of rotating equipment, such as pumps, compressors, and centrifuges, can be found for each design configuration. Figure 5 illustrates the vertical-shaft cantilever design; Figure 6 illustrates the horizontal-shaft cantilever design; and Figure 7 illustrates the horizontal-shaft centerhung design. In addition to the rotor dimensions, other considerations for selection of rotor shaft position include impact on operating performance, cost of manufacture, maintenance, and number and type of seals. The operating performance is not expected to deviate significantly based on rotor position. The possible considerations include rotor imbalance due to flooding of the housing and liquid distribution on the rotor. In general the centerhung design is considered more stable, but it has a slightly higher cost of manufacture due to the split case housing, is more difficult to maintain, and requires two shaft seals instead of one. Standard equations for fatigue and rigidity are used to determine shaft diameter for both orientations. 4.2.

Horizontal Versus Vertical Shaft Orientation

The centerhung design is restricted to a horizontal shaft orientation. A verticalshaft cantilever design is expected to have slightly lower maintenance costs than the horizontal-cantilever design. Both cantilever options should have similar design and fabrication costs. Flooding of the housing due to insufficient liquid drainage would be less of a problem with the vertical-shaft arrangement with respect to rotor imbalance. Liquid distribution on the rotor can be influenced by gravity more on the vertical shaft, but the effect should be minimal. Reference 7 illustrates both the centerhung and cantilever horizontal-shaft arrangements and discusses an application for use of the vertical-shaft cantilever design. 4.3.

Seals

Two types of seals are needed to prevent fluid leakage from the housing and to ensure that gas passes through the rotor countercurrent to the liquid. Seals on the shaft as it passes through the housing can be of a design appropriate for the fluids


FIGURE 5 Pilot-scale RPB illustrating the vertical-shaft cantilever design with direct motor drive. (Photo courtesy of The Dow Chemical Company.)


FIGURE 6 Pilot-scale RPB illustrating the horizontal-shaft cantilever design with direct motor drive. (Photo courtesy of Higravitec Center of Beijing University of Chemical Technology.)

being handled. Mechanical seals, lip seals, and packing glands are some suitable examples. As mentioned previously, a centerhung rotor requires two shaft seals, whereas the cantilever rotor requires only one. To seal the rotor to prevent gas bypassing, labyrinth seals and liquid ring seals are options. Figure 1 shows the position of seals for a vertical-shaft cantilever design.


FIGURE 7 Commercial water deaeration RPB using the horizontal-shaft centerhung design and direct motor drive. (Photo courtesy of Higravitec Center of Beijing University of Chemical Technology.)

4.4.

Power Train

Options for connecting the motor drive to the shaft depend on the shaft orientation. A vertical-shaft cantilever design would prefer a belt drive to reduce the cost of manufacture of the support structure and to facilitate maintenance. A horizontal shaft has the additional option of direct coupling. Variable speed can be accomplished through a gearbox or preferably through variable frequency control on the motor. In addition to the power requirements discussed previously, the startup power to overcome the torque of the rotor must be considered. This startup power is related to the time required to reach the desired rotor speed. 4.5.

Liquid Distribution

As discussed previously, proper liquid distribution on the rotor is critical to performance, but it is also important to prevent rotor imbalance. Rotor imbalance


from liquid maldistribution is especially a problem with high-viscosity fluids. Liquid distributor pipes extending the axial length of the rotor, depending on the pipe diameter and number, can be a source of additional gas pressure drop. Gas flow around the distributor pipes can also be a source of vibration. In the case of a vertical-shaft arrangement, the liquid head in the vertical distributor pipe must be considered to ensure equal liquid distribution on the axial length of the rotor. The rotor can be used to assist liquid distribution either by attaching the distributor pipes to the rotor or by introducing the liquid onto the rotor and allowing the centrifugal force to move the liquid to the packing. In the former option the liquid must enter the RPB through the shaft, requiring machining of a channel in the shaft and an additional seal. In the case of liquid–liquid extraction, the shaft must have at least one channel for introduction of one of the liquid phases (see Figure 2). 4.6.

Rotor Packing

The selection of the type of rotor packing depends largely on the performance requirements. However, there are some mechanical design considerations. Examples of packing include woven wire screen, pellets randomly packed, foam metal, and structured packing. The materials of construction must have physical properties sufficient to withstand the hydraulic forces created by the accelerating liquid. The packing must be dimensionally stable during operation to avoid rotor imbalance issues. Some packing materials may require supports to keep them in place. Proper design of the supports will consider porosity to prevent flooding, strength, impact on fluid distribution, and pressure drop. 4.7.

Multiple Rotors

Several designs involving multiple rotors have been proposed. To accommodate the need of additional transfer units in countercurrent gas/liquid contact, a vertical shaft with at least two rotors and appropriate internals to conduct the gas from the bottom to the top and liquid from top to bottom can be built (5). Another variation allows for heat transfer in addition to the mass transfer. By providing rotors for condensation and for boiling and multiple packed rotors for gas/liquid contacting, a self-contained distillation column on a single shaft is envisioned (6). Obviously, these multiple-rotor devices are more complex from a mechanical design and construction perspective. However, they offer some interesting possibilities for reducing plant size by combining multiple unit operations and additional stages of separation in one piece of equipment. 5.

APPLICATIONS

The operating and design principles given previously provide a basis for understanding the performance enhancement available to a wide variety of applications.


These include standard mass transfer operations, such as absorption and stripping, but also include reaction systems. The following section highlights some of the applications for which data are available and suggests other opportunities for exploitation of the intensified mass and heat transfer capabilities. 5.1.

Absorption

Absorption of a component of a gas stream into a liquid is a common practice in the chemical industry to affect cleanup of vent gases, conduct chemical reactions, purify products, or to recover products from process streams. The enhanced mass transfer capability of RPBs provides the opportunity to perform absorption processes in smaller equipment, to lower inventories, to shorten startup and shutdown times, and to lower pressure drop (48). Figure 8 provides a visual comparison of the size of a conventional absorber tower next to three RPBs that handle the equivalent gas and liquid flows (9). An example of industrial relevance is the removal of sulfur dioxide (SO2) from vent gases by absorption into water or a lime slurry (48). In the water absorption process, both gas-film and liquid-film resistance to mass transfer occurs. As a result the overall mass transfer rate is proportional to gas-flow rate and acceleration but inversely proportional to liquid-flow rate. Due to the fast reaction of SO2 with lime, this system is only gas-film diffusion limited. The overall mass transfer rate is largely unaffected by gas- or liquid-flow rate and is proportional to acceleration, but to a lesser extent than the water absorption process. In both cases the overall mass transfer rate is reportedly much higher than the corresponding conventional packed towers. In another study of gas-side mass transfer–limited absorption involving SO2 absorption into a sodium hydroxide solution using a wire screen packing, the overall mass transfer coefficient was found to be lower than reported data for packed towers (16). Replacing the wire screen packing with two parallel rotating plates significantly enhanced the mass transfer performance. Absorption of hypochlorous acid into water, a liquid-side mass transfer– limited process, showed HTU values as low as 4 cm, with a strong dependence on liquid-flow rate. Heat of absorption removal was identified as a potential issue with absorption in rotating beds (9). 5.2.

Stripping

Removal of volatile components from the liquid phase to a gas phase has been the object of much study in RPB devices. One of the early successful applications was oxygen removal from water for use in secondary oil field recovery and boiler water feed (7). The oil field application demonstrated oxygen removal from 6–14 ppm to less than 50 ppb in both 50-T/h and 300-T/h RPBs using natural gas for stripping. The packing had 92% porosity and 500-m2/m3 volumetric surface area


FIGURE 8 Commercial use of RPB technology in HOCl process. (Photo courtesy of The Dow Chemical Company.)


and was constructed of wire mesh. Comparison with conventional vacuum desorption in a packed tower combined with chemical reduction agents showed lower cost and equipment size for the RPB approach. The proposed application on oil platforms promises further advantages of weight reduction over conventional process equipment. The boiler feed water deaerator reduced oxygen content to less than 7 ppb by the use of exhaust steam (low pressure). This system operated at lower temperature (110C), used lower-pressure steam (0.05-MPa gauge), and achieved the oxygen specification without the use of chemical reducing agents as compared to conventional thermal desorption in a packed tower (7). The preceding approaches to water deaeration used a gas continuous process in the RPB. A liquid continuous RPB has been designed and tested for this application as well (21). The liquid continuous process allows design of the RPB for reduced power requirements, but it does require higher-pressure gas to overcome the hydraulic head of the liquid. The schematic in Figure 2 shows the liquid takeoff near the eye (inside diameter) of the rotor, thus recovering the power needed to accelerate the liquid. In the case of oil platform water deaeration using produced methane gas, boosting the pressure of the available gas would not be necessary. As with the gas continuous process, mass transfer is enhanced by increasing rotor speed and increasing gas-flow rate. Sampling at various radial positions in the polyurethane foam packing revealed the possibility of liquid back-mixing within the rotor that reduced the mass transfer efficiency, i.e., fewer transfer units than expected (21). Further work on the hydrodynamics of the gas and liquid interaction may be warranted in order to realize the full potential of this energy-saving approach for stripping (49). A novel example of stripping in rotating packed beds is the stripping of residual monomer and solvent from polymers (47). In polystyrene production, conventional vacuum desorption achieves residuals reduction to about 500 ppm. Steam-stripping technology is available to reduce residuals to about 200 ppm. Compared to steam stripping, the RPB technology is expected to reduce capital cost, energy costs, and equipment size and to eliminate the potential for side reactions of steam with the polymer. A pilot-scale devolatilizer, called an Accelerator, and a larger demonstration unit showed the viability of this approach. Data collected at 5–10 mm Hg pressure followed the equilibrium curve for residual styrene and solvent. This indicates that the mass transfer capability is more than adequate to achieve equilibrium conditions in the short residence time in the rotor. As a further effort to minimize process costs, the devolatilization rotating packed bed was combined with a centrifugal pelletizer. Heating of the rotor and polymer was accomplished by eddy currents generated by placing magnets on either side of the rotor. Since the high viscosity of the polymer melt requires higher g-force to achieve thin-film flow over the packing compared to earlier gas/liquid applications, the packing must have sufficient compressive strength to withstand


the generated forces. A reticulated foam metal packing of high porosity (90%) and high surface area (500 m2/m3) was used (47). Air stripping of volatile organic compounds from groundwater shows the possibility of using RPB technology for either continuous operation or intermittent remedial operation. Since a wide variety of processes (membranes, air stripping, biological activity, chemical oxidation, and carbon adsorption) are available to remove volatile organics from water, the selection of RPBs will depend on the performance requirements and the relative cost compared to the alternatives. Tests on air stripping of jet fuel components from groundwater show the viability of RPB use (33). Both a wire gauze packing and a reticulated foam metal packing proved effective in removing compounds such as benzene, o-xylene, toluene, 1,2,4-trimethylbenzene, and naphthalene. A demonstrated number of transfer units as high as 12 gave corresponding height-of-transfer-unit values of 2–3 cm. Another stripping application actually involves absorption and reaction as well. Chlorine gas absorbs into sodium hydroxide aqueous solution, reacts to produce hypochlorous acid (HOCl), and is then stripped using excess chlorine gas. The primary measure of performance of this operation is the recovery of stripped HOCl. This study showed the importance of liquid distribution (type of spray nozzle), gas/liquid ratio, and type of packing (wire gauze preferred over glass beads or flat plates). Above a minimal g-force, little performance improvement was seen. Low-surface-area wire gauze packing (660 m2/m3) was just as effective as high-surface-area (2800 m2/m3) packing (8). Scale-up to commercial operation of this process showed a doubling of the HTU for this gas-side mass transfer– limited stripping. The actual pressure drop in the commercial scale RPB was half the expected value. This same correlation, empirically based on centrifugal and frictional factors of film flow, effectively modeled the pilot RPB (9). 5.3.

Distillation

Distillation combines absorption and stripping in one device. Rotating packed beds perform distillation by use of external condensers and reboilers, as in conventional towers (29), or by use of internal heat exchangers as part of the rotor (6). Up to 20 theoretical plates were demonstrated in a rotor of 800-mm diameter (50). Distillation was demonstrated on a 3-tons/h pilot plant separating an ethanol/ propanol mixture at total reflux. The pilot plant consisted of two RPBs, one for stripping and one for rectification, along with external reboiler and condenser, respectively (51). Retrofit of existing distillation towers with an RPB has been proposed as a means to adding separation stages (52). Another pilot distillation study employed only one RPB along with external condenser and reboiler. The cyclohexane/n-heptane mixture was separated at rates up to 9 tons/h at total reflux. The system provided up to 6 transfer units (NTU) of separation in a 21-cm packing depth. The primary variable affecting


separation was rotor speed. Two operating pressures and a range of gas loading (reboiler duty) provided additional data for analysis (29,42). 5.4.

Heat Transfer

Heat input or removal in rotating systems is best accomplished using plates to separate the heat transfer fluids from the process fluids (45). Since spinning disc technology is discussed in Chapter 3 of this book, this section will cover only the application of heat transfer in conjunction with rotating packed beds and some of the issues related to further development needs. In the 1950s Hickman developed a centrifugal vapor compression evaporator for seawater desalination (53). This device consisted of multiple spinning discs. Seawater sprayed on one side of the disc evaporated, while the centrifugal force removed the residue from the plate surface. The vapor was compressed and returned to the opposite side of the plate, where condensation provided the heat for evaporation and the desired freshwater for recovery. Overall heat transfer coefficients of 18 kW/m2-K are about three times higher than those achieved in steam turbine condensers. A high-intensity heat pump, called Rotex, has been developed taking advantage of the enhanced heat and mass transfer performance of rotating discs (44). This single device carries out the processes of evaporation, condensation, absorption, and heat transfer to a working fluid. The higher heat transfer coefficients experienced by Hickman led to the concept of placing a peripheral reboiler and core condenser on either side of a rotating packed bed (50). This concept would be useful for distillation applications that need reflux and boilup. The internal exchangers as part of the rotor would decrease the required heat transfer surface area but would involve additional design and fabrication complexity. Although heat exchangers on either end of a packed rotor are an option for replacing external heat exchangers for distillation, the problem of heat transfer within a porous packed bed remains. Heat input can be achieved by use of eddy currents (47), microwaves, or sonic energy. Thus operations such as evaporation, stripping, and endothermic reactions can be envisioned. Heat removal, however, is more problematic. Exothermic reactions must be conducted adiabatically within the rotor, unless a suitable means of extracting the heat of reaction can be developed. One approach could be alternating packing and heat transfer plates. This raises the complexity of design and fabrication but could provide the needed cooling to approach isothermal operation. A simpler method of evaporative cooling is possible if the evaporation is compatible with the chemical process. 5.5.

Adsorption

Centrifugal adsorption technology (CAT) allows the use of very small adsorbent particles (microns) to increase the mass transfer efficiency. Application to ion


exchange, volatile organic removal from water, recovery of pharmaceutical proteins, and production of fine chemicals are examples of potential commercial interest. Process advantages include low inventory, low contact time, steady-state operation, and relatively small equipment (54). The CAT mode of operation involves introduction of the adsorbent near the axis of the rotor, allowing the centrifugal force to move the particles radially outward. Liquid introduced at the outer periphery of the rotor moves countercurrent to the adsorbent and is removed at the axis of the rotor. Adsorbent slurry collects at the periphery and is conducted to the rotor axis for discharge. Experiments using activated carbon to adsorb n-butanol from water revealed that the degree of back-mixing is the dominant factor in performance. Back-mixing is a function of rotor speed, density difference between the phases, and the particle diameter (54). The hydrodynamics of two-phase flow in CAT were compared to two-phase flow under gravity using a large-diameter (1.3-mm) particle in water with a small density difference and a small-diameter (81.8-micron) particle in water with a large density difference. The throughput capacity of the CAT was higher than predicted from the homogeneous-flow model, though the model works well for the gravity-flow column. Pressure drop estimates were used to predict void fractions in the range of 0.7–0.8. Higher rotor speeds resulted in higher void fraction (55). 5.6.

Liquid–Liquid Extraction

The use of centrifugal fields for liquid–liquid extraction was perhaps the first commercially successful application of rotating packed beds. Podbielniak modified a patented vapor–liquid contactor (2), using a perforated spiral passageway as the rotor packing, to solve problems with penicillin recovery in 1945 (1). Penicillin broth forms stable emulsions that require centrifugal force to break. Solvent extraction was effective only at low pH, which caused penicillin degradation. Multiple stages were needed to affect the necessary concentration. In addition, the fermentation liquor varied significantly from batch to batch and plant to plant. Conventional countercurrent solvent extraction, mixer-settlers, and mixercentrifuge combinations could not effectively solve these problems without product loss. The centrifugal solvent extractor achieved 98% product recovery by taking advantage of its low liquid holdup, short residence time, high centrifugal force, and multistage countercurrent contacting. Continuous glycerin washing of soap produced by saponification has been demonstrated in a countercurrent centrifugal extractor (38). The device achieves phase separation with as little as 0.02 specific gravity difference and accomplishes up to 10 theoretical stages of extraction. Some of the advantages over prior operations reportedly include flexibility in feed, low holdup, less waste due to more efficient separation, simple operation, rapid startup, and small space requirements. Rotors filled with ceramic foam instead of perforated cylinders have been tested for liquid extraction of trace contaminants from water (23). The test solution


used a C12 alkene to extract 1,2-dichloroethane from water. Flow patterns are similar to the previous applications, with the heavy phase introduced at the inner diameter of the rotor and the light phase at the outer diameter. Countercurrent flow is achieved as the heavy phase moves outward, displacing the light phase inward. The results indicate an optimum pore size for the ceramic foam, decreasing height of transfer unit with increasing rotor speed, and increasing holdup with increasing dispersed phase flow. The ability of the centrifugal extractor to solve difficult liquid–liquid separation problems, as illustrated in the previous examples, has allowed its use in a wide range of extraction applications. The long history of use has given it a general acceptance in chemical manufacturing—an acceptance not shared by the broader application of gas–liquid interactions. 5.7.

Crystallization

In the reactive precipitation process of reacting CO2 with Ca(OH)2 slurry to produce nanoparticles of CaCO3, the controlling steps of the process are absorption of CO2 and dissolution of solid Ca(OH)2. The degree of supersaturation depends on the reaction rate and controls the nucleation rate and, therefore, the particle size. The intense mass transfer and micromixing capability of the rotating packed bed provides the environment to produce CaCO3 particles of size 15–30 nm with a narrow size distribution. Reaction time reduces 4- to 10-fold, compared to stirred-tank reactors. Rotor speed, gas–liquid ratio, and initial calcium hydroxide concentration influence reaction rate. An increase in rotor speed reduces the average particle size. Addition of growth inhibitors also helps to control particle size and size distribution (56). High-gravity reactive precipitation (HGRP) has been extended to the production of aluminum hydroxide and strontium carbonate (57). Aluminum hydroxide fibrils precipitate from the reaction of sodium meta-aluminate (NaAlO2), water, and carbon dioxide and are formed in diameters of 1–10 nm and lengths of 50–300 nm. Rotor speed, gas- and liquid-flow rates, and initial reactant concentrations control particle size. Strontium carbonate particles of 40-nm mean diameter and narrow size distribution have been produced from the liquid–liquid reaction of strontium nitrate and sodium carbonate. Crystallization that occurs during evaporation can potentially be intensified by use of vapor recompression and spinning discs. In this scenario, the evaporated vapor is compressed and then condensed on the bottom of the discs to heat the crystallizing fluid (58). This approach may permit operation at higher temperatures, lower surface area, and less time. Recrystallization of an active pharmaceutical ingredient on a spinning disc, employing a solvent/antisolvent approach to induce rapid precipitation, results in the desired small particles (1–15 microns) and narrow particle size distribution (59).


5.8.

Reactions

A significant number of chemical reactions of commercial interest have reaction rates limited by heat or mass transfer rather than the intrinsic chemical kinetics. Many of these processes could be intensified by application of the RPB or spinning disc technology. Indeed, application of centrifugal fields to reacting systems may represent its greatest potential. Improved yield performance may also be possible (50). Spinning disc technology provides the advantage of heat transfer capability. Thus, temperature control of exothermic reactions such as styrene polymerization and intense mixing allow reduction in reaction time compared to conventional batch reactors, especially in the latter stages of reaction, where viscosity is higher (60,61). Polycondensation reactions that are equilibrium controlled, such as polyesters, could benefit from the thin films generated in the RPB or spinning disc. In these reactions, removal of the coproduct of polymerization, e.g., ethylene glycol, is necessary in order to advance the polymerization, a task that becomes increasingly difficult as the reaction proceeds. The thin films and short residence time of high-gravity devices aid the evaporation and may permit operation at higher temperatures than conventional reactors (50). As mentioned previously, RPB and spinning disc technology may provide benefits for reactions that are mass or heat transfer limited, i.e., for fast kinetic reactions. Unfortunately, the true chemical kinetics are often unknown. A smallscale RPB or spinning disc may prove to be a useful screening tool to determine the intrinsic chemical kinetics. In one such study, six different reactions of interest in the manufacture of pharmaceuticals were screened using a spinning disc reactor (59). Three of the six reactions were found to be limited by liquid–liquid mixing. These include a phase-transfer-catalyzed Darzen’s reaction, a crystallization, and a highly exothermic condensation reaction. In the Darzen’s reaction the reactant inventory was reduced 99% and the impurity level decreased 93% as compared to the conventional reactor. Crystallization achieved mean crystal size of 3 microns, with a narrow size distribution. The highly exothermic reaction had excellent temperature control. A lab-scale RPB has also been used to investigate reactions that are mass transfer constrained in conventional reactors. Testing a reaction that involves release of a volatile organic as part of molecular weight buildup revealed overall process reaction rates equivalent to conventional reactors in time frames more than two orders of magnitude lower. Here the high surface area and surface renewal capability of the RPB helped to overcome the transfer limitations across the liquid boundary. Similar results were seen on other reaction processes that were constrained by a liquid–surface interaction (62). Reactions can be combined with other unit operations, as in the example of reactive stripping in the production of hypochlorous acid (HOCl). An RPB was


used to absorb chlorine gas into an aqueous sodium hydroxide solution (caustic) where reaction to HOCl occurs. Since HOCl is unstable in the presence of the coproduct sodium chloride, stripping of the HOCl into the gas phase is necessary to recover a stable product. The reaction of chlorine with caustic is essentially instantaneous: therefore the process reaction rate is liquid-side mass transfer controlled. Stripping of HOCl, on the other hand, is believed to be gas-side mass transfer controlled. The intense mass transfer capability of the RPB allowed 10% higher yields while using less than half the stripping gas as compared to conventional spray tower operation. This study showed low-surface-area, high-porosity wire screen packing to perform better than glass beads or parallel flat plates. Packing support design and liquid distributor type influenced performance. Operating parameters of importance included rotor speed and the gas–liquid ratio (8). Scale-up issues and performance of the commercial HOCl operation (9) are discussed in Section 6. Fermentation reactions are often limited by oxygen transfer rates. The enhanced mass transfer achieved in centrifugal fields applied to bioreactions should be expected to increase productivity. A centrifugal field bioreactor (CFBR) demonstrated higher productivity in the overproduction of lipase with Staphylococcus carnosus as compared to conventional fermenters (22). Both batch and semibatch fermentation in the CFBR showed no influence on the biological activity of growth or exoprotein synthesis. Lipase productivity rates were proportional to oxygen transfer rates, which were 10 times higher than in shaken cultures. The CFBR process involved feed of air and liquid to the outer diameter of the rotor, with takeoff at the center. Air was dispersed in the liquid by either a sieve drum or a multilayer-sintered screen. The inward radial movement of the gas helped to suspend the bacteria in the culture against the centrifugal force. An external circulation loop for the liquid allowed heat exchange and product analysis. Since many fermentation reactions are characterized by foaming, the CFBR was equipped with a foam breaker—a stator with needles positioned at the inside diameter of the rotor. 5.9.

Other Applications

Centrifugal fields in an electrochemical cell facilitate the removal of gas bubbles from the electrodes, thus reducing the voltage requirement. A rotating chlorine cell showed a drop in voltage from 3.17 V to about 2.8 V at 3-kA/m2 current density when accelerated to 200 g (50). Demonstration of a rotating air cathode provided greater voltage drop at higher current density as compared to a stationary cell (50). Dedusting or demisting in rotating devices provides opportunity to remove small particles at very high throughput. A “mop-fan” built with flexible fibers in a conventional fan housing effectively removes 50% of two-micron particles of slaked lime dust in conjunction with water spray on the rotor. Inline rotary


demisters employing reticulated metal foam are used for oil separation from air on turbines (50). 6.

SCALE-UP AND COMMERCIAL USE

Although considerable studies on the use of centrifugal fields in chemical processing have been reported for lab- and pilot-scale operations, little information is public on scale-up criteria for either performance parameters or equipment design. Three examples of commercial use of centrifugal fields are available for review. These include liquid–liquid extraction, water deaeration, and reactive stripping for hypochlorous acid production. Commercial application of centrifugal fields encounters considerable resistance from both the technical and business communities due to both real and perceived risks. The real risks involve reliable mechanical design of rotating equipment, which includes seals, bearings, and rotor stability. Perceived risks on process performance may derive from a lack of understanding of the process fundamentals and how performance may change with the scale of operation. Overcoming the tendency to “use what we know and understand” represents a challenge that goes beyond the effort at technology development. Convincing the technical community to accept the risk of rotating equipment for chemical processing may be easier on applications with clear performance advantages over conventional process equipment. A good example is the liquid–liquid solvent extraction of penicillin (1), in which the low residence time and ability to handle emulsions and solids allowed 98% product recovery. The 10% higher yields and 50% reduction in stripping gas for the HOCl reactive stripping process provides another example of performance advantage (8). In addition to the lower operating costs associated with enhanced performance, the business community is interested in lower capital investment and assurance that the process will reliably perform as designed in terms of product capacity, on-stream time, and product quality. The smaller size of the centrifugal equipment may satisfy the capital investment question. This was the primary driver for implementation of water deaeration in China (7). Lower capital is also a driver for the polymer devolatilization application (47). The question of reliable performance is best addressed through convincing the technical community and leveraging the considerable industry experience with design and manufacture of rotating equipment, such as pumps, compressors, and centrifuges. 6.1.

Scale-Up Criteria

A number of parameters can be considered for scale-up of rotating packed beds, including rotor packing, liquid distribution, flooding, pressure drop, rotor speed, HTU, NTU, temperature, and pressure. Since the same packing material (same


porosity and volumetric surface area) is used on scale-up as in the pilot equipment, HTU is expected to scale directly. Results from water deaeration (7) and HOCl production (9) indicate this may be true for liquid-side mass transfer– controlled operations but not for gas-side mass transfer–controlled operations. Flooding of the rotor in gas continuous operations would be expected at the same gas velocities as in the pilot RPB. Flooding normally occurs at the inside diameter of the rotor due to the lower cross-sectional area and higher velocities. However, a check of the porosity of the outer packing support may be necessary to ensure that no flooding occurs. If scale-up of rotor speed is based on constant rotor tangential velocity instead of constant g-force, then the gas velocity at flooding will be lower with larger-diameter rotors. Throughput capability and backpressure control of the light-phase takeoff (24) control flooding in liquid–liquid or gas-dispersed systems. Using a pressure drop model based only on centrifugal acceleration and frictional drag, the HOCl scale-up overpredicted the pressure drop of the commercial RPB by a factor of 2 (9). A more rigorous approach to pressure drop calculation that takes into account the conservation of angular momentum and the inlet and outlet zones of the rotor and housing (40) should provide more predictable scale-up performance. Rotor speed has an impact on mass transfer performance, flooding, and pressure drop. Rotor speed on scale-up can be determined based on maintaining constant tangential velocity (r) or constant acceleration (r2). Rotor speed will be higher for constant-acceleration scale-up. Impact on both process performance and equipment design must be understood in making this determination. Scale-up based on constant acceleration is conservative for mass transfer and flooding performance, while constant tangential velocity is conservative for pressure drop. Liquid distribution may be an important parameter, as demonstrated in the HOCl process, where different liquid distributors provided significantly different results (8). The initial contact of the liquid with the rotor influences the mass transfer performance of the RPB in gas continuous operations (15). Although the use of a packing support at the inside diameter of the rotor would be expected to impact this initial liquid contact with the rotor, experiments did not show any reduced mass transfer performance (36). As mentioned earlier, the same rotor internals used in pilot tests should be used upon scale-up. The rotor dimensions of inner diameter and axial height are determined by maintaining a constant superficial gas velocity at the rotor eye. The radial packing depth, and thus the outer diameter, is based on the number of transfer units required. Adjustments in packing depth and packing type may be necessary to achieve the desired liquid holdup or residence time, e.g., for chemical reaction (26). As with any chemical operation, the physical properties of the fluids, such as density, viscosity, and heat capacity, must be known. If chemical reaction is


involved, then the intrinsic kinetics must be understood. Of particular relevance is the influence of mass transfer rates on overall reaction rates. 6.2.

Scale-Up Design

Once the rotor dimensions have been determined along with the operating conditions, attention shifts to the mechanical design of the RPB. The major concern with rotating equipment is maintaining stable operation, i.e., limiting vibration. Excessive vibration results in premature seal and bearing failure, poor process performance, metal fatigue, and increased maintenance costs and downtime. Proper mechanical design principles determine the option of cantilevered or centerhung rotor, the shaft diameter, the type and position of bearings, and seal design. The drive train, whether belt driven or direct coupled, is determined by the power requirements and the shaft orientation. The housing must be sufficient to contain the temperature and pressure of the operation and to provide adequate inlet and outlet nozzles for the process fluids. 6.3.

Commercial Examples

Two commercial examples of rotating packed-bed operation are water deaeration for the Chinese oil fields (7) and HOCl reactive stripping in the United States (9). These two cases illustrate nicely the range of process conditions and design features available for successful scale-up. Water deaeration (Figure 7) uses a direct-coupled drive on a horizontal-shaft centerhung rotor to process a low gasto-liquid operation (3 : 1 vol/vol). The HOCl process (Figure 9) employs a beltdriven, vertical-shaft cantilever rotor to contact a high gas-to-liquid ratio. The water deaeration process employed a staged scale-up program. From the lab operations, a 50 tons/hour (T/h) pilot RPB was built and tested in the oil field. Using natural gas for stripping at a gas/liquid ratio of over 2, the desired oxygen content in the exit liquid of less than 50 ppb was demonstrated. This successful demonstration led to the installation of a full-scale commercial RPB to process 300 T/h. This unit has rotor dimensions of 600-mm ID, 1000-mm OD, and 700-mm AH. The wire screen packing has high porosity (92%) and low surface area (500 m2/m3). The rotor spins at a modest 750 rpm. Performance matched that of the lab and pilot units, achieving a typical 30-ppb oxygen content. Two 250-T/h units have been designed for installation on an oil platform to process seawater, providing advantages in size and weight as compared to conventional technology (7). Figure 9 shows the commercial-scale RPB for the reactive stripping process for HOCl production. Figure 8 provides a visual impression of the process intensification that occurs using RPBs. The three RPBs shown in the lower left of the picture process the same volume of gas and liquid as the tall absorber tower to the right. The scale-up factor from the pilot unit of over 400 : 1 yielded a rotor of slightly less than 2 m in diameter. Performance of product yield from raw materials


FIGURE 9 Commercial RPB for HOCl production using the vertical-shaft cantilever design with belt drive. (Photo courtesy of The Dow Chemical Company.)

met or exceeded that of the pilot unit. Pressure drop half of that expected implies the need for better predictive correlations. Due to the higher gas-handling capability and the conservative scale-up design, much higher capacities are anticipated as compared to the design. The liquid-side mass transfer performance as measured by chlorate formation showed performance equivalent to or better than that of the pilot RPB. However, the gas-side mass transfer, as represented by HOCl stripping, showed a doubling of the HTU to about 8 cm. The mechanical reliability after two years of operation indicates no issues due to RPB operation. The RPB is very easy to start up and shut down (9). These two successful commercial applications of rotating packed beds prove that scale-up from pilot-scale equipment can achieve the desired process performance in commercial-scale operations. In addition, the mechanical reliability of the rotating equipment is in line with the experience with other rotating


devices. Thus, the risk concerns of performance and reliability can be managed acceptably. Both the technical and business communities can have confidence that future applications will meet expectations of process performance and onstream time. 7.

FUTURE

Extension of the use of centrifugal fields into chemical processing, beyond the physical movement of fluids, has shown limited niche application in the past, in spite of considerable research activity. Specialty application of centrifugal fields to liquid–liquid extraction has enjoyed success for more than 50 years. Advantages stem from operation at low density differences, breaking of emulsions, short contact times, and higher efficiencies as compared to other liquid–liquid extractors. The commercial use of rotating equipment for the broader field of gas– liquid operations has only a five-year history. Numerous examples of possibilities in the areas of absorption, adsorption, stripping, distillation, reactions, crystallization, and other operations have been referenced. The chief objections to the use of centrifugal fields have been associated with the risks of scale-up and the operation of rotating equipment. The two commercial applications of water deaeration and HOCl reactive stripping demonstrate the ability to reliably scale up processes involving a wide range of gas–liquid loadings. Process performance in both cases met or exceeded design criteria, with good operating reliability. Further application will likely require significant cost or performance advantages over more conventional process technology. Considerable commercial experience will be needed before centrifugal fields will enjoy common acceptance among both technical and business interests in the chemical industry. To gain that status, projects must be selected carefully to ensure that advantages are realized over alternative technologies. The most likely opportunities for exploitation will come from mass transfer– limited reactions and the combination of unit operations in one device. Examples of reactions mentioned earlier include polymerization, condensation reactions, crystallization, and heterogeneous catalysis. Combined unit operations are illustrated by reactive distillation, polymer devolatilization with pelletization, and the use of heat exchangers (reboilers and condensers) with distillation. In addition to research on process applications, research to define the fundamental performance characterizations is needed. A number of empirical correlations have been developed for pressure drop, residence time, power, flooding, etc. More generalized theoretical expressions for these parameters that accurately predict performance on a wide range of rotor designs and sizes would be very beneficial to confidently scale-up the technology.


REFERENCES 1. 2. 3. 4. 5. 6. 7.

8.

9.

10. 11. 12. 13. 14.

15.

16. 17.

18. 19.

20. 21.

Podbielniak WJ, Kaiser HR, Ziegenhorn GJ. Centrifugal solvent extraction. Chem Eng Prog Symp Ser 1970; 66(100):43–50. Podbielniak WJ. U.S. Patent 2,044,996, 1935; U.S. Patent 2,172,222, 1939; U.S. Patent 2,281,796, 1942. Pilo CW, Dahlbeck SW. U.S. Patent 2,941,872, June 21, 1960. Podbielniak WJ. U.S. Patent 3,233,880, Feb. 8, 1966. Todd DB. U.S. Patent 3,486,743, Dec. 30, 1969. Ramshaw C, Mallinson RH. U.S. Patent 4,283,255, Aug. 11, 1981. Zheng C, Guo K, Song Y, Zhou X, Ai D. Industrial practice of Higravitec in water deaeration. In: Semel J, ed. 2nd International Conference on Process Intensification in Practice. London: BHR Group, 1997:273–287. Trent D, Tirtowidjojo D, Quarderer G. Reactive stripping in a rotating packed bed for the production of hypochlorous acid. In: Green A, ed. 3rd International Conference on Process Intensification for the Chemical Industry. London: BHR Group, 1999:217–231. Trent D, Tirtowidjojo D. Commercial operation of a rotating packed bed (RPB) and other applications of RPB technology. In: Gough M, ed. 4th International Conference on Process Intensification in Practice. London: BHR Group, 2001:11–19. Peel J, Howarth CR, Ramshaw C. Process intensification: Higee seawater deaeration. Trans IChemE 1998; 76(Part A):585–592. Barson N, Beyer GH. Characteristics of a Podbielniak centrifugal extractor. Chem Eng Prog 1953; 49(5):243–252. Ding X, Hu X, Ding Y, Wu Y, Li D. A model for the mass transfer coefficient in rotating packed bed. Chem Eng Comm 2000; 178:249–256. Burns JR, Ramshaw C. Process intensification: visual study of liquid maldistribution in rotating packed beds. Chem Engr Sci 51 1996; (8):1347–1352. Basic A, Dudukovic MP. Hydrodynamics and mass transfer in rotating packed beds. In: Heat and Mass Transfer in Porous Media Conference Proceedings, 1992: 651–662. Guo K, Guo F, Feng Y, Chen J, Zheng C, Gardner NC. Synchronous visual and RTD study on liquid flow in rotating packed-bed contactor. Chem Engr Sci 2000; 55: 1699–1706. Sandilya P, Rao DP, Sharma A, Biswas G. Gas-phase mass transfer in a centrifugal contactor. Ind Eng Chem Res 2001; 40:384–392. Munjal S, Dudukovic MP, Ramachandran P. Mass transfer in rotating packed beds— II. Experimental results and comparison with theory and gravity flow. Chem Engr Sci 1989; 44(10):2257–2268. Lin CC, Chen YS, Liu HS. Prediction of liquid holdup in countercurrent-flow rotating packed bed. Trans IChemE 2000; 78(Part A):397–403. Burns JR, Jamil JN, Ramshaw C. Process intensification: operating characteristics of rotating packed beds—determination of liquid hold-up for a high-voidage structured packing. Chem Engr Sci 2000; 55:2401–2415. Ramshaw C. U.S. Patent 4,715,869. Dec. 29, 1987. Peel J, Howarth CR, Ramshaw C. Process intensification: Higee seawater deaeration. Trans IChemE 1998; 76(Part A):585–593.


22.

23.

24. 25. 26. 27.

28. 29. 30. 31. 32.

33.

34. 35.

36.

37.

38.

39.

Voit H, Gotz F, Mersmann AB. Overproduction of lipase with Staphylococcus carnosus (pLipPS1) under modified gravity in a centrifugal field bioreactor. Chem Eng Technol 1989; 12:364–373. Lee JGM, Howarth CR, Ramshaw CR. Trace effluent removal from large flowrate aqueous phase using a packed rotary extractor. In: Proceedings of Value Adding Solvent Extraction (Pap ISEC ‘96), Melbourne, 1996:1185–1190. Jacobsen FM, Beyer GH. Operating characteristics of a centrifugal extractor. AIChE J 1956; 2(3):283–289. Sherwood TK, Shipley GH, Holloway FAL. Flooding velocities in packed columns. Ind Eng Chem 1938; 30:768. Keyvani M, Gardner NC. Operating characteristics of rotating beds. Chem Engr Prog 1989; 85(9):48–52. Singh SP, Wilson JH, Counce RM, Villiers-Fisher JF, Jennings HL, Lucero AJ, Reed GD, Ashworth RA, Elliott MG. Removal of volatile organic compounds from groundwater using a rotary air stripper. Ind Eng Chem Res 1992; 31:574–580. Lockett MJ. Flooding of rotating structured packing and its application to conventional packed columns. Trans IChemE 1995; 73(Part A):379–384. Kelleher T, Fair JR. Distillation studies in a high-gravity contactor. Ind Eng Chem Res 1996; 35(12):4646–4655. Basic A, Dudukovic MP. Liquid holdup in rotating packed beds: examination of the film flow assumption. AIChE J 1995; 41(2):301–316. Tung H-H, Mah RSH. Modeling liquid mass transfer in Higee separation process. Chem Eng Commun 1985; 39:147–153. Munjal S, Dudukovic MP, Ramachandran P. Mass transfer in rotating packed beds— I. Development of gas–liquid and liquid–solid mass-transfer correlations. Chem Eng Sci 1989; 44(10):2245–2256. Singh SP. Air Stripping of Volatile Organic Compounds from Groundwater: An Evaluation of a Centrifugal Vapor–Liquid Contactor. Ph.D. dissertation, The University of Tennessee, Knoxville, 1989. Rumford F, Rae IJ. Performance characteristics of a centrifugal gas absorber. Trans Instn Chem Engrs 1956; 34:195–203. Zhu J, Feng Y, Zheng C, Gardner N. Modeling of mass transfer in rotating packed beds. Proceedings of First International Workshop on High-Gravity Engineering and Technology, Beijing University of Chemical Technology, Beijing, 1996. Guo F, Zhao Y, Cui J, Guo K, Chen J, Zheng C. Effect of inner packing support on liquid controlled mass transfer process in rotating packed beds. In: Gough M, ed. 4th International Conference on Process Intensification in Practice. London: BHR Group, 2001:107–113. Hassan-Beck HM, Ramshaw C. Process intensification: mass transfer for countercurrent rotating bed. Proceedings of First International Workshop on High Gravity Engineering and Technology, Beijing University of Chemical Technology, Beijing, Elsevier, 1996. Podbielniak WJ, Ziegenhorn GJ, Kaiser HR. Continuous soap washing and finishing, using multistage, countercurrent, centrifugal contactors. J Am Oil Chem Soc 1957; 34:103–106. Sedahmed GH, Al-Abd MZ, El-Taweel YA, Darwish MA. Liquid–solid mass transfer behavior of rotating screen discs. Chem Eng J 2000; 76:247–252.


40. 41. 42. 43. 44. 45. 46. 47.

48. 49.

50. 51. 52. 53. 54.

55. 56.

57.

58.

59.

Zheng C, Guo K, Feng Y, Yang C, Gardner NC. Pressure drop of centripetal gas flow through rotating beds. Ind Eng Chem Res 2000; 39:829–834. Liu H-S, Lin C-C, Wu S-C, Hsu H-W. Characteristics of a rotating packed bed. Ind Eng Chem Res 1996; 35:3590–3596. Kelleher T, Fair JR. Distillation studies in a high-gravity contactor. Separations Research Program, University of Texas, Austin, 1993. Kumar MP, Rao DP. Studies on a high-gravity gas–liquid contactor. Ind Eng Chem Res 1990; 29:920–924. Aoune A, Ramshaw C. Process intensification: heat and mass transfer characteristics of liquid films on rotating discs. Int J Heat Mass Transfer 1999; 42:2543–2556. Jachuck RJJ, Ramshaw C. Process intensification: heat transfer characteristics of tailored rotating surfaces. Heat Recovery Systems CHP 1994; 14(5):475–491. Shang D-Y, Anderson HI. Heat transfer in gravity-driven film flow of power-law fluids. Int J Heat Mass Transfer 1999; 42:2085–2099. Cummings CJ, Quarderer G, Tirtowidjojo D. Polymer devolatilization and pelletization in a rotating packed bed. In: Green A, ed. 3rd International Conference on Process Intensification for the Chemical Industry. London: BHR Group, 1999:147–158. Gardner N, Keyvani M, Coskundeniz A. Flue gas desulfurization by rotating beds. U.S. Department of Energy, DOE#DE-FG22-87PC 79924, 1993. Al-Shaban K, Balasundaram V, Howarth CR, Ramshaw C, Peel JRA. The hydrodynamic and mass transfer characteristics of a large centrifugal water deoxygenator. In: Proceedings of Energy Efficiency Process Technology, Process Intensification Conference, Athens, 1993:475–484. Ramshaw C. The opportunities for exploiting centrifugal fields. Heat Recovery Systems CHP 1993; 13(6):493–513. Ramshaw C. Higee distillation—an example of process intensification. Chem Eng 1983; 389:13–14. Fowler R. Higee—a status report. Chem Eng 1989; 456:35–37. Hickman K. U.S. Bureau of Saline Water R&D Progress Report No. 12, Nov. 1956. Bisschops MA, van der Wielen L, Luyben K. Centrifugal adsorption technology for the removal of volatile organic compounds from water. In: Semel J, ed. 2nd International Conference on Process Intensification in Practice. London: BHR Group, 1997:299–307. Bisschops MAT, Luyben K, van der Wielen L. Hydrodynamics of countercurrent two-phase flow in a centrifugal field. AIChE J 2001; 47(6):1263–1276. Chen J, Wang Y, Jia Z, Zheng C. Synthesis of nanoparticles of CaCO3 in a novel reactor. In: Semel J, ed. 2nd International Conference on Process Intensification in Practice. London: BHR Group, 1997:157–164. Chen J-F, Wang Y-H, Guo F, Wang X-M, Zheng C. Synthesis of nanoparticles with novel technology: high-gravity reactive precipitation. Ind Eng Chem Res 2000; 39: 948–954. Ramshaw C. Process intensification—incentives and opportunities. In: Doraiswamy LK, Mashelkar RA, eds. Frontiers in Chemical Reaction Engineering. Vol 1. New York: Wiley, 1984:685–697. Oxley P, Brechtelsbauer C, Ricard F, Lewis N, Ramshaw C. Evaluation of spinning disk reactor technology for the manufacture of pharmaceuticals. Ind Eng Chem Res 2000; 39:2175–2182.


60.

Boodhoo KVK, Jachuck RJ, Ramshaw C. Spinning disc reactor for the intensification of styrene polymerization. In: Semel J, ed. 2nd International Conference on Process Intensification in Practice. London: BHR Group, 1997:125–133. 61. Boodhoo KVK, Jachuck RJ, Ramshaw C. Process intensification: spinning disc polymerizer for the manufacture of polystyrene. In: Ramshaw C, ed. 1st International Conference on Process Intensification for the Chemical Industry. London: BHR Group, 1995:175–180. 62. Winnington TL, Drögemüller P. Speeding up slow processes. 4th International Conference on Process Intensification, Brugge, Belgium, 2001.


3 The Spinning Disc Reactor C. Ramshaw University of Newcastle upon Tyne, Newcastle upon Tyne, England

1.

INTRODUCTION

When considering the options for intensifying reactions that involve multiple fluids, it is helpful to identify the shortcomings of the conventional equipment that is currently in use. In this context, perhaps the most frequently used item is the stirred vessel fitted with a cooling jacket, shown in Figure 1. A turbine impeller generates a circulation comprising two toroidal vortices, and the turbine torque is normally prevented from driving a free vortex by the use of wall baffles, as shown. If a gas–liquid reaction is involved, then the gas is usually injected directly below the impeller via a suitable sparging arrangement. The popularity of the stirred vessel is due to its perceived simplicity and adaptability, coupled with the fact that it is superficially straightforward to scale-up from the laboratory beaker that was used when the process was being developed. Unfortunately, it suffers from several serious problems, as indicated later. In the normal case of a geometrically similar scale-up, it can be readily shown that the surface area per unit volume varies inversely with the vessel diameter. Thus larger vessels are more difficult to cool, since the heat generated by a reaction in a potential runaway situation is proportional to the vessel volume, whereas the surface area available to dissipate a given heat output is decreased. Vigorous reactions may require the reactor to be “detuned” by operating with more dilute feedstock in order to reduce the full-scale reaction intensity. This


FIGURE 1 Stirred vessel showing circulation pattern.

could influence the reaction temperature trajectory and compromise the yield and selectivity. A further unfortunate characteristic of the stirred vessel is that its mixing capability is also a strong function of its size. Scale-up usually proceeds on the basis of a constant impeller tip speed, and since the mean circulation speed in the vortices is broadly proportional to the tip speed chosen, the circulation time is proportional to the vessel diameter. Thus the turnover time of the vessel contents increases at the larger scale and the macro mixing performance deteriorates. These fundamental shortcomings of the stirred vessel have generated a considerable degree of uncertainty when fine chemical or pharmaceutical processes are being developed for full-scale operation. This has led the relevant regulating authorities, e.g., the U.S. Food and Drug Administration, to insist on a process validation at laboratory, pilot, and full scale. Since each validation entails significant administration and delay, the procedure can hold up the implementation of commercial production by several years. Because a new metabolically active molecule will be patented as soon as possible and certainly before clinical trials and process development, this delay significantly erodes the time available under patent cover to recoup a company’s R&D expenditure and make a profit from a potential “blockbuster” drug. 1.1.

The “Desktop” Continuous Process

The predominant culture that prevails for the production of fine chemicals/drugs, with an output of up to (say) 500 tons per year, is to operate batchwise. As already noted, this stems from the fact that the process is almost always developed from a batch-operated beaker or flask. However, it is worth observing that an output of 500 tons/year of active substance corresponds to a continuous process flow rate of around only 70 mL/second. This allows various items of intensified equipment


to be assembled and operated continuously literally on a “desktop” to meet the production demand. The decision to switch from batch to continuous processing immediately confers an intensification benefit because the peak batch process loads (e.g., heat output, liquid removal, etc.) are distributed in time, so equipment size can be reduced. Thus with a new process, the laboratory scale becomes the full scale when allowed to run continuously and the scale-up delays described earlier are largely avoided. This strategy is generating considerable industrial interest as the commercial pressure to bring new molecules to market rapidly continues to increase. A further factor that favors continuous “desktop” manufacture is its potential impact on the overall business process of making and marketing fine chemicals. Thus it goes without saying that with very short process residence times, the operation can be much more responsive so that grade changes can be effected in seconds rather than hours. This facilitates just-in-time manufacture, which can lead to dramatic reductions in the capital costs associated with the multiple grades of stock that may be needed rapidly to satisfy demanding customers. 1.2.

Exploitation of Centrifugal Fields

Approximately two-thirds of the unit operations performed in process engineering involve multiphase contact (e.g., distillation, gas/liquid reaction, boiling). In the absence of an imposed acceleration field, the system fluid dynamics are dominated by surface forces so that the interfacial area developed is relatively small, and, with no buoyancy force, there can be no countercurrent interfacial motion. When these conditions prevail, the intensity of the operation is very low, with little if any process performance (e.g., reaction, separation, heat transfer) being exhibited. This scenario leads naturally to the suggestion that a high-acceleration field would stimulate the generation of smaller bubbles, higher flooding velocities, and more intense shear stresses. This “Higee” strategy has been championed over many years because of its profound and beneficial impact upon many important multiphase operations: Absorption Distillation Boiling Condensation Liquid extraction Particle disengagement Heat pumps Etc. One particular embodiment of this approach is the rotating packed bed, which was originally conceived as the “Higee” equivalent of a packed column (1),


FIGURE 2 The “Higee” contactor (continuous gas phase).

shown schematically in Figure 2. The equivalent stage height within the toroidal packing is about 1.5 cm for a gas film—limited system, compared with about 60 cm in a conventional packed column. Equivalent flooding velocities may be estimated from the Sherwood plot and can be very high, even for packing with a specific surface area exceeding 1000 m1. Since the Higee duty was originally envisioned as being purely orientated to mass transfer, no specific heat transfer capability was provided. However, the spinning disc reactor (SDR) may be regarded as an alternative to the “Higee” rotor. It can act as a mass transfer/ contacting device (possibly with multiple discs) or as a particularly intense gas–liquid reactor (when fitted with heating/cooling provision). Its attraction lies in the high heat and mass transfer rates that can be stimulated between the disc and the thin liquid film generated on its surface, and between the film and the adjacent gas. The performance and applications of the SDR are considered in detail later. As might be expected, the enhanced acceleration field is established on a permanent basis within a rotor that receives and discharges the working fluid. The alternative approach, which relies upon a permanent vortex field, i.e., a cyclone,


will not be considered here, primarily because the fluid residence time within a vortex cannot be easily controlled independent of the desired acceleration. At this juncture it is helpful to consider the basic physics of motion in a rotating system. 1.3.

Free Motion of a Particle Around an Axis

Consider the free (frictionless) motion of a particle P of mass m rotating around a fixed axis O on a smooth surface, as shown in Figure 3. The particle is constrained to move in a circular trajectory by a light string that exerts an inward tension T and generates a corresponding acceleration. This can be estimated as follows: In time t, P moves along an arc that subtends an angle to the axis. The angular velocity of P is given by d兾dt and its speed is v r, where r is the length of the string. During the time t the change of velocity of P is v v sin () and as → 0, v is directed along the radius toward O. The acceleration is dv v sin ( d ) v as → 0 dt dt Hence, acceleration r2 and the string tension needed to maintain this is T m2r

FIGURE 3 The free motion of a particle around an axis.


FIGURE 4 Spiral trajectory as a particle moves toward the center of rotation.

If the string is broken, then the acceleration ceases and the particle leaves its circular trajectory and continues along a tangent at a velocity v. If the string is slowly shortened then: 1. The particle moves toward O in a spiral trajectory involving many turns. 2. Work must be done in order to overcome the string tension T. For the proposed frictionless system, this work input results in an increase in the particle’s kinetic energy. Conversely, if the string were lengthened, the particle’s velocity would decrease. Figure 4 shows the spiral trajectory of the particle and the corresponding velocity diagram. The radial and tangential velocity components are dr兾dt, v, respectively, giving a resultant VR that is a tangent to the spiral trajectory. Noting that dr兾dt << v, the angle between v and VR is , where tan (dr兾dt)兾v. As the particle moves inward towards O, the component of T along the spiral trajectory is responsible for increasing its speed. Hence, with an inward tension deemed to be negative we have: m

dv v 2 dr 1 T sin m ⋅ ⋅ dt r dt v

(as → 0)

or dv dr v ⋅ dt dt r

(1)

This confirms that v increases for an inward spiral trajectory (i.e., when dr兾dt is negative). From Eq. (1) we have dv dr v r and integration from v1r1 to v2r2 gives


ln

v2 r1 ln v1 r2

i.e., v2 r2 v1r1

(2)

This is an important result because it shows that in a frictionless or free situation, angular momentum is conserved. The same conclusion can be reached from energy considerations, noting that the work done by the string tension as r is reduced is equal to the gain in particle kinetic energy, i.e., T dr m

 v2  v2 dr d  m  mv dv r  2

Hence v dr dv r leading again to Eq. (2). The practical consequence of Eq. (2) for rotational fluid flow can be quite dramatic, as demonstrated by the high wind speeds that may be generated near the center of free vortex flows—e.g., tornadoes and typhoons. 1.4.

Flow Over a Rotating Surface

If we now consider the behavior of a liquid film on a rotating disc, the motion is no longer “free,” because the film is influenced by the disc via the shear force generated at the solid/liquid interface. Liquid supplied to the inner region of the disc is first brought up to the disc’s rotational velocity by the tangential shear force and then moves radially outward, to be discharged from the disc periphery. In a stationary frame of reference, the liquid trajectory is therefore a spiral with arms separated by a radial distance given by 2 dr ⋅ dt In a rotating frame of reference (i.e., that of an observer anchored to the disc), the trajectory is nearly radial. Since this reference frame is most relevant when we consider the disc/fluid interaction, it is helpful to evaluate the flow on this basis. We shall assume that frictionless flow occurs through a closed radial channel that is fixed to the rotating disc. Thus as the fluid moves outward, the only


interaction with the disc/channel is via a tangential force. In a time dt, the change per unit mass in the fluid angular momentum as it moves a distance dr is dM d (rr ) 2r dr

(3)

Note that since the fluid is forced to rotate at the disc speed, is constant. The change in angular momentum given by Eq. (3) is brought about by the torque generated by a tangential force F exerted on the fluid at the radius r by the radial channel and acting for a time dt. Since dM Fr dt, the tangential acceleration imposed by the channel on the fluid as it moves radially is F 1 dM dr 2 m r dt dt

(4)

This is the Coriolis acceleration, which is imposed on particles moving in a rotating reference frame—e.g., liquids on a rotating surface or winds in the earth’s atmosphere. The resultant acceleration experienced by a particle is a combination of the radial and tangential components, making an angle to the radius, where tan 2

dr 2 2 r dt v

 dr     dt 

In general dr << v dt and the unconstrained flow will be largely radial. This can be readily confirmed during the operation of a spinning disc reactor or a rotating packed bed, because any deposits from the fluid flow lie close to the radius vector. 2.

THE SPINNING DISC REACTOR

As pointed out earlier, a spinning disc, or more generally a rotating surface of revolution, is an alternative to the “Higee” rotating packed bed. It is particularly effective when high heat fluxes or viscous liquids are involved. The object is to generate a highly sheared liquid film when a liquid is supplied to the unit at or near its center. The film is initially accelerated tangentially by the shear stresses established at the disc/liquid surface. This causes the liquid to approach the disc’s angular velocity and then move outward as a thinning/diverging film under the prevailing centrifugal acceleration. The phenomenon was studied in detail by Woods (2), who photographed the behavior of a fully wetting dilute film of ink as


it traveled over a spinning glass disc. Care was taken to supply the liquid from a central axisymmetric distributor in a particularly uniform manner. After calibration, the local film thickness was inferred from the density of the photographic image at that point. Even when great care was taken to ensure that the liquid feed was introduced to the disc in an axisymmetric manner with the minimum disturbance, the smooth inner film always broke down into an array of spiral ripples, as shown in Figures 5 and 6. These spiral structures then broke down further until the wave pattern became utterly chaotic, provided that the disc was big enough. It is known that liquid film flow over a surface is intrinsically unstable, and the phenomenon has been studied by several workers (3–7). It appears to be qualitatively equivalent to the breakdown of a smoke plume rising from a lighted cigarette, where a chaotic zone is generated about 20 cm above the source. The behavior can also be observed when a liquid film flows over a stationary surface such as a windowpane or a dam spillway. Woods concluded that two types of wave existed: nearly two-dimensional (2D) and three-dimensional (3D). The amplitude of the two-dimensional spiral waves grew rapidly, and therefore a theory based on the assumption of small amplitude is not valid across the whole disc. A transition from 2D to 3D waves occurred once their amplitude reached about three to four times the local film mean thickness. Higher liquid flow rates stimulated a more rapid breakup of the wavelets. Only about 1% increase in liquid surface area was ascribed to the presence of waves. Thus any improvement in mass/heat transfer performance generated by the waves is due to the additional shear they induce. It will be appreciated that even in the absence of ripples, highly sheared thin liquid films, such as those that can be readily generated on a spinning surface, provide an ideal fluid dynamic environment for the rapid transmission of heat, matter, and momentum. This is due to the short diffusion path length involved for transfer between the adjacent gas phase to the liquid film and thence to the disc surface. These characteristics of a spinning disc (or more generally a rotating surface of revolution) make it ideal for performing any intrinsically rapid physical or chemical transformation in a liquid, even if it is viscous. Typical examples include polymerization, precipitations, and rapid exothermic organic reactions. Some of these are described in more detail later. 2.1.

The Nusselt-Flow Model

While the fluid dynamics of the actual film-flow process across the disc is dauntingly complex, a very approximate interim flow model may be based upon Nusselt’s treatment of the flow of a condensate film. This assumes that the flow is stable (i.e., ripple free), that there is no circumferential slip at the disc/liquid surface, and that there is no shear at the gas/liquid interface. The treatment is based


FIGURE 5 Liquid film behavior on a rotating disc, with Q 19 cm3/s and

(a) 100, (b) 200, (c) 300, (d) 400, (e) 500, (f) 600 rpm.


FIGURE 5 (cont.)


FIGURE 5 (cont.)


FIGURE 6 Local liquid film behavior on a rotating disc, with (a) Q 19 cm3/s,

100 rpm; (b) Q 19 cm3/s, 200 rpm; (c) Q 13 cm3/s, 400 rpm; (d) Q 19 cm3/s, 500 rpm; (e) Q 19 cm3/s, 600 rpm; (f) Q 19 cm3/s,

600 rpm.


FIGURE 6 (cont.)


FIGURE 6 (cont.)


FIGURE 7 (a) Sketch of a liquid film on a rotating disc. (b) Detail of a liquid film on a rotating disc.

on the schematic representation given in Figure 7. Part b represents the local film at a radius r. The shear stress on the annular plane at a distance y from the disc provides the radial acceleration for the fluid lying between y y and y s. Thus a force balance on the film lying between r r and r r dr, with zero shear stress at the gas–liquid interface, gives

2 r ( s y)

du dy

(5)

The boundary conditions are: 1. u 0 at y 0 since there is no fluid slip at the disc/liquid interface. 2. du兾dy 0 at y s since there is no shear stress at the gas/liquid interface.


Hence

2 r  y2   sy   2

u

(6)

The average film velocity is given by Uav

1 s

s

∫0

u dy

w 2 rs 2 3

The maximum film velocity (at y s) is Umax

w 2 rs 2 1.5 Uav 2

(7)

Referring to Figure 7a, the liquid is supplied to the disc at a radius ri and a mass flow rate M. It is deemed to instantaneously acquire and maintain the disc angular velocity as it moves over the disc to be discharged at its periphery. At a radius r the mass flow rate is given by M Uav s2r

(8)

Eliminating Uav from Eqs. (7) and (8) gives  3M  s    2 w 2 2 

1/ 3

r −2 / 3

(9)

Inserting Eq. (9) into Eq. (8) gives Uav

w 2 r  3M    3  2 r 2 w 2 2 

 M 2w2     12 2 

2/3

1/ 3

r −1/ 3

(10)

Hence the average time required for the liquid to travel from ri to ro is

∫

t

R0

Ri

dr 3  12 2    Uo 4  M 2 2 

1/ 3

(ro

4/3

ri4 / 3

)

(11)

If we consider a typical example of water flowing over a disc under the following conditions: M 3 102 kg/s

103 N-s/m 2


10 3 kg/m 3

ri 5 102 m

ro 0.25 m

100 s1 (955 rpm)

then from Eq. (11) the average liquid transit time on the disc is 0.25 s, and from Eq. (9) the film thickness at the disc edge (provided that the film does not break up into rivulets) is 28 microns. A more viscous liquid, such as a polymer (say, 10 N-s/m2), would have a thickness at the disc periphery of 600 microns and a transit time of about 5 seconds. As already noted, the foregoing calculations must be regarded as a guide only, since the films are intrinsically unstable, with waves being amplified as the liquid proceeds to the edge of the disc. It will be appreciated that this process proceeds more rapidly with relatively inviscid liquids. 2.1.1.

Mass Transfer

A conservative estimate of the disc’s mass transfer performance may be obtained from the Nusselt model, assuming that there is no film mixing as it proceeds to the edge of the disc. For unsteady diffusion into a finite stagnant slab, the plot shown in Figure 8 from (8) gives the relative concentration distribution within the slab at various times, with a zero initial concentration and a surface concentration C0 imposed at time t 0. The parameter on the curves is the Fourier number, Fo, where Fo

Dte s2

(12)

and D solute diffusivity within the film te exposure time of the film surface s film thickness As can be seen from Figure 8, if Fo 0.02, the concentration changes within the film are confined largely to the surface layer and the local mass transfer coefficient is given by the Higbie penetration theory (9) as  D kL    te 

1/ 2

(13)

For the previous example of a polymer flowing over the disc, a typical Fourier number may be calculated from: D ⬇ 109 m 2 /s te ⬇ 5 s s ⬇ 6 104 m


FIGURE 8 Concentration distribution at various times in a slab x for zero initial concentration and surface concentration C0.

Thus Fo 0.014. Equation (7) shows that the film surface velocity is given by  9 M 2 2  Umax 1.5 Uav  ⋅ 2   32 

1/ 3

r −1/ 3

Hence from Eq. (11),  32 2  te  2 2  9 M 

1/ 3

(

3 4/3 r r14 / 3 4


)

(14)

Inserting this into Eq. (13) gives D kL    

1/ 2

 2 M 2 2    2  3 

1/ 6

(r

1 4/3

r14 / 3

(15)

)

1/ 2

However, it must be noted that as the film flows over the disc, the film thickness progressively decreases, provided the liquid fully wets the disc. As this occurs, the concentration profiles normal to the disc plane are compressed, thereby causing a proportionate enhancement of the solute diffusion rate beyond that predicted by penetration theory. Thus the local value of kL can be corrected to account approximately for the steepened concentration gradients by multiplying by a factor s1兾s, where s1 is the film thickness at a radius r1 as given by Eq. (9). The corrected local value of kL is then D kL    

1/ 2

 2 M 2 2    2  3 

1/ 6

 r    r1 

2/3

(r

1 4/3

r14 / 3

)

1/ 2

(16)

At the point of film formation, where r r1, Eq. (16) shows that kL . However, the average value of kL over the disc surface is given by k Lav

1 2

r2 r12

(

)∫

r2

2 k L r dr

(17)

r1

This requires numerical integration. As pointed out at the outset, these estimates of the mass transfer performance are likely to be conservative as the disturbance of the film by ripples has been neglected. This will reduce the exposure time significantly, particularly with inviscid liquids. 2.1.2.

Heat Transfer

The Nusselt model was originally developed to correlate the performance of vapor condensers. In this case, the latent heat of condensation is discharged at the gas–liquid interface and subsequently conducted through the draining condensate film, the conduction path length being the local film thickness. When a liquid film is heated or cooled on a spinning disc, the conduction path length is less (about 50% of the thickness) because all of the sensible heat does not have to be conducted through the entire film. Since the thermal diffusivity of most liquids is typically of the order of 107 m2/s, compared with a mass diffusivity of around 109 m2/s, the Fourier numbers involved in the heat transfer version of Figure 8 are approximately 100 times their mass transfer equivalent. This implies that the heat transfer process involves the whole liquid film rather than merely


a thin layer near the disc surface. The Higbie model for heat transfer is therefore inappropriate. For the larger Fourier numbers involved in the heat transfer it is reasonable to represent the film temperature profile approximately by a quadratic expression: T A By Cy 2

(18)

A, B, and C are constants determined by these boundary conditions: 1. T Tw at y 0 2. T Ts at y s 3. dT兾dy 0 at y s It can be shown that y y2 T Tw 2(Tw Ts ) (Tw Ts ) 2 s s

(19)

dT 2(Tw Ts ) dy s

(20)

and  y − 1 s 

Since the film temperature gradient perpendicular to the disc will be much greater than that in the radial direction, the local heat flux (Q) into the film will be controlled by the value of dT兾dy at the disc surface. Hence  dT  2k Q k  (Tw Ts ) s  dy  y 0 Thus the effective film coefficient is h

Q 2k s Tw Ts

(21)

For our earlier example with water on a 0.5-m-diameter disc, Eq. (21) implies that the heat transfer film coefficient at the periphery is 43 kW/m2k, with the predicted film thickness of 28 microns. For this estimate to be realistic it is essential that the film wet the disc and not break up into rivulets. This depends upon a force balance at an incipient “dry-out point,” as indicated in Figure 9. At the film stagnation point the film momentum is potentially destroyed by the action of the component of the surface forces parallel to the disc. Thus for an average film velocity Uav we must satisfy the following condition for rivulet maintenance: 2 T (1 cos ) Uav s


FIGURE 9 Schematic showing liquid film “dry-out” on a rotating disc.

where T is the surface tension per unit length, is the contact angle, is the liquid density, and s is the local film thickness. Coherent films are less likely as they become thinner and their velocity decreases. An inspection of Eqs. (9) and (10) reveals that U 2av s is proportional to 2/3M5/3. Hence the tendency to form rivulets is less at higher disc speeds and liquid flow rates and increases with large T and small . 2.1.3.

Film-Flow Instability

The existence of the wave structure within the film is of major practical interest, as was highlighted by some elegant experimental work conducted by Brauner and Maron (7). They monitored the instantaneous local film thickness of a liquid flowing down a stationary inclined plane using a capacitance technique.


Simultaneously the local mass transfer coefficient was measured between the disc and the liquid, using the limiting electrolytic current method. Their plots are reproduced in Figure 10. It was clear that the passage of a ripple was associated with a significant enhancement of the mass transfer coefficient, as a consequence of the flow field associated with ripple propagation. An analogous phenomenon may be observed when sand particles are disturbed by wavelets in shallow seaside pools. While the phenomenon has considerable theoretical interest, its immediate practical implication is very important because it suggests that the disc heat and mass transfer performance could be enhanced still further by appropriately engineering the disc surface profile. Some experimental results are discussed next.

FIGURE 10 Simultaneous time traces of local instantaneous film thickness and transfer rate.


2.2.

Heat/Mass Transfer Performance

An early application of the SDR to heat transfer duty was developed by Hickman (10), who was interested in the desalination of brackish water using a vapour compression evaporator. A sketch of his initial arrangement is shown in Figure 11, where brackish water flowing on one side of a rotating disc assembly was evaporated by condensing steam on the other. The very high overall heat transfer coefficients that could be achieved (up to 45 kW/m2K) ensured that the pressure ratio demanded from the vapor compressor was minimal, thereby establishing very efficient operation. In a further development of the idea (11), a series of disc assemblies was mounted on one vertical shaft that was enclosed within a tower, as shown in Figure 12. It is significant that Hickman did not report any problems associated with the deposition of crystal scale, even though his experimental runs lasted for several hundreds of hours. Radial rather than spiral stains were, however, exhibited, which suggested that the influence of the Coriolis acceleration was minimal. Since the temperature difference between the condensate and the evaporating brackish water was only 1–2C, it is presumed that the equivalent supersaturation was insufficient to cause significant crystal nucleation. The ability of the spinning disc to operate with very small driving temperature and concentration differences can improve the thermodynamic efficiency of the overall process system. This is clearly the case with the Hickman vapor compression evaporator, and it is also exemplified in the applications described next. In general, the power needed to rotate the spinning disc assembly is a small fraction of that saved by virtue of the establishment of an intensified fluid dynamic environment. 2.2.1.

The Rotating Electrolytic Cell

A laboratory-scale rotating chlor-alkali membrane cell was constructed and tested some years ago in ICI. The electrodes comprised closely spaced catalyzed discs that were separated by a Nafion membrane. The anolyte and catholyte concentrations corresponded to those in the brine and sodium hydroxide solutions used in the standard (FM21) industrial version of the membrane cell. As can be seen from Figure 13 (taken from Ref. 12), while the industrial cell voltage at a current density of 3 kA/m2 was 3.17 V, that of the rotating unit was a function of the applied acceleration, falling to about 2.75 V at 100 “g.” At a higher current density, the benefits of enhanced acceleration were even more marked. It will be recognized that the enhanced buoyancy forces generated by the high acceleration can eliminate the polarization effects associated with bubbles that adhere to the electrodes/ membranes or remain in the electrolytes. In principle a compact rotor comprising a bipolar cell assembly of closely spaced discs is capable of an exceptionally high chlorine production rate while operating at exceedingly competitive voltage.


FIGURE 11 Schematic of the single-element Hickman still.


FIGURE 12 (a) Schematic of the multiple-disc Hickman still. (b) Full-scale Hickman still.

2.2.2.

The “Rotex” Absorption Heat Pump

The main factor that has been responsible for the slow adoption of absorption heat pumps for heating and air conditioning duties has been their high capital cost compared with that of vapor compression equivalents. This is due largely to the cycle complexity, as shown in Figure 14, which displays the four principal cycle elements, all of which involve vapor–liquid systems: 1. 2. 3. 4.

Condenser Evaporator Generator/boiler Absorber


FIGURE 12 (cont.)


FIGURE 13 Voltage characteristics of a rotating chlorine cell.

The only single-phase item involved is the solution heat exchanger, which is intensified by the use of laminar flow in a matrix of fine channels. A sketch of the single-effect Rotex (13) design is shown in Figure 15, where it can be seen that a hermetically sealed rotating disc assembly fulfills the four functions listed. The working fluid consists of a water solution of either mixed alkali metal hydroxides or lithium bromide. The evaporator receives low-grade heat from the circulating ambient air and vaporizes the refrigerant at low pressure, the vapor being promptly absorbed at the absorber disc immediately opposite. Working fluid from the absorber sump, now rich in refrigerant, is returned to the generator via a solution pump and a solution heat exchanger. The latter consists of a matrix of closely spaced metal foil, which, as discussed earlier, gives very efficient heat transfer in a small volume. The heat of condensation and absorption is removed from the condenser/ absorber disc assembly by circulating water that enters and leaves via a mechanical seal. The working fluid is pumped around the cycle by a pitot tube assembly, with the tubes dipping into a peripheral liquid trough. Since the Rotex machine operates with a horizontal axis, the pitot tube arm is counterweighted to resist the frictional torque exerted by the trough. Information recently released (14) shows that the double effect air conditioning version of Rotex has achieved a coefficient of performance of 1.0 at a temperature lift of 35C using lithium bromide solution. This unit is about to enter field trials. Its high performance is entirely due to the intensity of the heat and mass transfer environment generated on the liquid film flowing over the discs. It is also worth noting that Alfa Laval has developed a process evaporator for fruit juice and milk concentration using a nested stack of cones. Figure 16


shows the arrangement, in which steam is caused to condense outside the cones while the process fluid is concentrated on the film flowing on the inside surface. In this context it may be observed that the disc and the cone are specific examples of the general case of a rotating surface of revolution. The acceleration causing the outward movement of the liquid film is the resolved component of 2r along the surface in question. Koerfer (15) performed an interesting study with a series of perforated and smooth rotating discs 600 mm in diameter at speeds up to 600 rpm. The mass transfer performance was measured using the oxygen/water system, with the results shown in Figure 17. Very good performance was recorded with the perforated discs, and this was attributed partly to the short exposure time of the film as it negotiated each perforation and partly to the extra film area created. Interestingly, the film behavior was much more predictable when it flowed over the disc surface containing the raised lips arising from the punching operation. Film flow on the alternate side tended to “leak” through the disc, particularly at lower liquid flows, presumably due to the Coanda effect as liquid negotiated the rounded edge of the holes. As part of a general development to use spinning discs in an intensified absorption heat pump, Aoune and Ramshaw (16) measured both the local and average heat transfer performance on smooth rotating surfaces. The disc surface

FIGURE 14 Single-effect absorption heat pump cycle.


temperature was estimated by extrapolating the values given by the digitized output from thermocouples embedded at depths of 1 mm and 9 mm in a brass disc. The local film temperature was measured by a thermistor contained in a stylus that could be traversed over the disc surface. Using water, the heat transfer coefficient on the 50-cm-diameter disc regularly exhibited a minimum value at a radius of about 17 cm. On the other hand, with the use of a water/60% monopropylene glycol mixture, no minimum was observed and the absolute performance was much poorer than that obtained with water. This behavior is attributed to the tangential fluid slip generated as the feed liquid is brought up to the rotor’s angular velocity. This slip appears to be more marked with low-viscosity liquids, which seems intuitively reasonable.

FIGURE 15 (a) Rotex design concept. (b) Rotex prototype.


FIGURE 15 (cont.)

Mass transfer studies were then performed, based upon the aeration of liquid that had been previously stripped. The local liquid oxygen concentration was established by carefully abstracting a film sample via a quill that could be traversed over the disc. An Orbisphere oxygen analyzer was used. Rather poor agreement was obtained between the experimental results for water (kL 4 → 10 m/s 104) and the Higbie predictions (approx 1 m/s 104) based upon total exposure time of liquid on the disc. Clearly, liquid mixing within the film generates exposure times that are much shorter than the liquid residence time on the disc. Another study, by Jachuck and Ramshaw (17), explored the influence of surface profile upon the heat transfer performance of a spinning disc. Using a smooth disc as a benchmark it was shown that disc surfaces disrupted with metal powder or grooves gave a significantly improved performance—presumably due to the better film mixing. The best performance at modest disc speeds was obtained with “undercut” grooves (Figure 18), which were originally conceived as a technique for improving the circumferential distribution of any radial rivulets. At higher disc speeds, the film radial velocity was such that liquid was projected off the disc, thereby compromising the heat transfer process.


FIGURE 16 Alfa Laval concentrator.

2.3. 2.3.1.

Reactor Applications Strategic Considerations

At the most basic level, the SDR is an extremely effective gas–liquid contacting device. This makes it ideal for performing many intensified heat or mass transfer operations and, as will be discussed later, it may be deployed as an evaporator or an aerator/desorber. However, its principal application in the process industry is likely to be as a very high-performance reactor. Since the reactor is the heart of


FIGURE 17 Mass transfer performance of a rotating disc (O2/H2O system).


FIGURE 18

Types of disc grooves tested.

any process, the SDR can radically improve the economics and efficiency of many key processes, both in the commodity and in the fine chemical area. In order to exert full control over the progress of a chemical reaction or physical transformation, the fluid dynamic environment must be sufficiently intense so as to ensure that the mixing and heat transfer rates are faster than the intrinsic chemical kinetics. This concept is shown diagrammatically in Figure 19, which illustrates the progress of a reaction represented simply as A B → C, with the reactants A, B traveling in plug flow along a tubular reactor. When the interdiffusion of A, B is slow compared with the reaction rate, then C is produced near the original plane of A兾B separation. This represents a total loss of control on two counts: 1. The A兾B stoichiometric ratio varies wildly across the reactor diameter. Therefore the selectivity for the desired product C is likely to be compromised because a more realistic reaction scheme will usually include many side reactions.


2. Most of the reaction to C occurs in the immediate neighborhood of the plane of A–B separation. Thus only a small fraction of the available reactor volume is utilized and an opportunity for intensification is lost. On the other hand, when mixing is fast, the A兾B ratio is uniform and control over the product spectrum can be maintained. All the reactor space is used to maximum effect. Since the intrinsic kinetics are allowed free rein, the reactor is able to operate at the maximum intensity permitted by the specific chemical system. While it should be self-evident that a rational reactor design demands a knowledge of both the fluid dynamic environment and the detailed process kinetics, the latter are rarely available. In many instances this leads to the severe limitation of many important reactions by an inadequate fluid dynamic intensity. Some of these are known to be fast, e.g., liquid-phase nitrations, while others are (incorrectly) assumed to be slow, e.g., most polymerizations. In these circumstances the pragmatic approach is to use a high-intensity reactor for each system and then to assess the impact upon the space–time productivity. Obviously, an intrinsically slow system is resistant to further acceleration and this will rapidly become evident. One significant qualification of this contention involves the very

FIGURE 19 The influence of mixing and reaction rates on reactor behavior.


short residence time in the SDR compared with its conventional counterparts. In certain reactions the process temperature is restricted to one that avoids product breakdown in the time available. Since the residence time in the SDR when performing a polymerization reaction is up to 10 seconds rather than the several hours involved in a conventional stirred vessel, we must re-examine the process’s temperature trajectory. A higher operating temperature may well be acceptable because the undesired breakdown components may not have time to be generated. However, the higher temperature will reduce the liquid viscosity and accelerate the reaction. The lower viscosity will reduce the residence time still further. Therefore the SDR can exploit a process operating envelope that is much larger than what is accessible to conventional technology. With regard to the processing of viscous liquids, by far the most important application relates to the manufacture of polymers. The key processes are: 1. 2. 3.

Condensation reactions Radical reactions Devolatilization

The progress of a condensation reaction is controlled by an equilibrium with a volatile product, which, if continuously removed, drives the reaction forward. Unfortunately, as polymerization proceeds, the liquid viscosity increases, rendering the removal of the volatile component much more difficult. The batch stirred vessel, which is conventionally used for polymer manufacture, has a limited ability to remove a volatile component from the increasingly viscous polymer melt. On the other hand, the SDR can maintain effective mass transfer and, as will be shown in Section 2.3.2, can achieve in one pass (taking several seconds) the same increment in polymerization as would conventionally require tens of minutes. The SDR with one or more discs on the same shaft is therefore capable of performing polycondensation extremely rapidly. The short residence time also facilitates rapid changes of product grade with minimum wastage. The rate of a polymerization that proceeds via a series of radical reactions is controlled by the micromixing environment within the polymer melt. Once again the stirred vessel is a poor means to achieve the high desired intensity, whereas the SDR has an impressive capability in this respect. It is well known that UV radiation is a very effective means of radical generation, and this technique has been proposed in the past for stimulating certain radical polymerizations. However, the radiation extinction distance in a polymer melt is only a few millimeters, so a polymerization reactor comprising a stirred vessel having a diameter of several meters is not a rational option. On the other hand, the thin polymer films that can be created and maintained on the SDR allow all of the polymer to receive a continuous, uniformly high radiation dose and thereby maintain a very high reaction intensity, as described later for the manufacture of butyl acrylate.


FIGURE 20 Schematic of an SDR styrene polymerizer.

The industrial manufacture of polymers is rarely taken to completion, and this requires unreacted monomer to be removed from the product—when its viscosity is highest. This devolatilization procedure is notoriously difficult because it usually involves the vacuum stripping of a stirred vessel’s contents for many hours. Just as the SDR promotes the removal of the volatile component of a condensation reaction, it is also effective in dramatically accelerating the devolatilization process. 2.3.2.

Polymerization

Polystyrene. The manufacture of polystyrene from various grades of prepolymer has been performed (18,19) on a 36-cm brass SDR using the arrangement shown in Figure 20. A series of concentric grooves was machined in the disc surface in order to improve liquid mixing within the film. The reaction operates via free radicals, which were initiated in this case using benzoyl peroxide. In the first instance a series of batch runs was performed in a conventional laboratory-scale stirred vessel in order to produce a calibration curve (Figure 21) of conversion


FIGURE 21 Free-radical polymerization of styrene.

versus time. This vessel was then used to produce about 200 mL of prepolymer at a range of conversions that was supplied to the inner spinning disc surface over a period of about 30 seconds. The SDR was heated from below by a stationary radiant ring, and the polymer produced was collected in a cooled annular trough surrounding the disc. The styrene was diluted with about 16% w/w toluene in order to reduce the viscosity. Figure 21 also shows the increment in polymer conversion in one pass over the disc as a function of the initial conversion in the preliminary batch. It can be seen that the equivalent batch time that can be ascribed to one pass on the disc increases (up to 58 minutes) as the initial conversion increases to 63%. This implies that the benefits of the SDR become more marked as the polymer viscosity increases. It is envisaged that the process can be scaled up either by using a larger disc or by mounting several discs on one shaft. The latter approach (i.e., several discs in series) does, however, involve the problem of transferring polymer from the peripheral collection trough to the center of the next disc. An alternative may be to operate discs on one shaft in parallel. For the experiments just described, the feed rate was roughly 5–10 mL/s, which is equivalent to an output of up to 250 tons/year on a continuous basis, though at this early stage this should not be considered the ultimate limitation.


The fundamental reasons for the high performance of the SDR are still a matter for debate. The significance of micromixing and the consequent improved probability of radical interaction has already been mentioned. However, another factor is expected to be the divergent character of the flow on the disc. This may be expected to align the polymer molecules and thereby encourage the juxtaposition of the reactive groups. Polycondensation. The reaction between maleic anhydride and ethylene glycol has been studied as an example of polycondensation (19). Since the reaction proceeds on an equilibrium basis, in order to drive it to completion the water produced must be eliminated from the increasingly viscous polymer melt. The grooved brass 36-cm disc described earlier for the polystyrene experiments was used at a temperature of 200C and a disc speed of 1000 rpm. As before, the experimental procedure involved the establishment of a benchmark batch calibration against which the subsequent disc runs could be compared. A typical acid number plot versus batch time is presented in Figure 22. As the acid number decreases, the conversion to polymer increases. The water of reaction was removed from the polymer film by maintaining a large nitrogen purge to the vapor space. This technique, rather than the application of a vacuum, was the preferred method for reducing the water vapor partial pressure. It can be seen that the increment in polymerization following one pass in the SDR corresponds to many minutes of reaction in the small batch reactor used as a reference. This is particularly encouraging because the mass transfer intensity in the laboratory stirred reactor is likely to be much greater than its industrial-scale

FIGURE 22 Time savings in the SDR for polyestirification.


equivalent, and it therefore provides a demanding benchmark for the spinning disc performance. 2.3.3.

Fine Chemical Manufacture

The intense heat and mass transfer environment that can be established within the liquid film flowing over the disc allows high selectivities and conversions to be achieved when fast liquid-phase reactions are performed. Very encouraging results were achieved in an industrial study of a phase-transfer-catalyzed (p-t-c) Darzen’s reaction to produce a drug intermediate (20). In comparison with the currently used batch processes, the ptc reaction on the SDR had a 99.9% reduced reaction time, a 99% reduced inventory, and a 93% reduced impurity level. A more recent study has involved a 20-cm-diameter SDR with a catalytically activated surface to perform the rearrangement of -pinene oxide to campholenic aldehyde (21), which is an important intermediate used in the fragrance industry. The comparative performance of the batch reactor and the SDR is shown in Table 1. For equivalent conversion and selectivity, the unoptimized SDR gave a much higher throughput than the equivalent batch reactor and avoided the need to separate a catalyst slurry from the product. Figure 23 shows the variation of selectivity as a function of disc speed and feed flow rate; Figure 24 gives the conversion levels achieved. While conversion falls from 100% at the higher flows and speeds, presumably due to the reduced liquid residence time on the disc, the selectivity increases. Thus it might be expected that a larger disc (or a sequence of small discs) could combine high conversion and high selectivity. The batch reactor performance is summarized in Figure 25, where it can be seen that 100% conversion requires 5 minutes ( 1 second on the disc) and a maximum selectivity of 65% is reached. 2.3.4.

Precipitation/Crystallization

The operation of crystallizers and precipitators is critically dependent upon the supersaturation environment prevailing within the crystal magma because this influences both the nucleation of new particles and the growth of those that TABLE 1 Comparison of the Best SDR Runs with Batch Results for Conversion of -Pinene Oxide to Campholenic Aldehyde Batch process Process time (s) Processed feed Conversion (%) Selectivity (%) Note

300 1.2 kg/h 100 64 Catalyst separated from the product mixture


SDR (continuous) 1 209 kg/h 100 62 No loss of catalyst

FIGURE 23 Selectivity towards campholenic aldehyde at 85C at various feed rates.


FIGURE 24 Conversion of -pinene oxide at 85C and various disc speeds.


FIGURE 25 Batch reaction: conversion and selectivity towards campholenic aldehyde.


already exist. In general, the growth rate has a first-order dependence on supersaturation for diffusion-limited systems and 1.5- to second-order where the surface integration resistance is significant (22). On the other hand, the nucleation rate, whether primary or secondary, has a higher-order dependence on supersaturation, typically in the range 2–9 (23), with the lower values being obtained mainly with low-molecular-weight solutes. Hence it will be recognized that high supersaturation can be readily achieved on SDRs. This feature makes them attractive for producing very small or nanosized particles, which are currently of intense industrial interest. Supersaturation can be generated in several ways. Perhaps the simplest technique is merely to cool a solution saturated at a higher temperature. Alternatively, supersaturation can be created by removing the solvent or adding an antisolvent for systems where the solubility is only a weak function of temperature. Finally, supersaturation can be created by reaction—between either two liquids or a liquid and a gas. In all these cases the intense environment created within an initially crystal-free liquid film moving in plug flow over the disc can generate very high supersaturation and consequently small and fairly uniform crystals. This characteristic of the SDR may be attractive in several industries (e.g., pharmaceuticals and coatings), where the product quality is intimately related to the fineness of the crystals and the tightness of the size distribution. The concept has recently been tested (24) in a spinning cone precipitator (Figure 26), which shares most of the characteristics of a spinning disc except that the centrifugal acceleration vector is not aligned to the cone surface. Barium sulfate was generated by mixing equimolar solutions of BaCl2 and Na2SO4 in a central reservoir. A thin liquid/slurry film flowed to the cone rim, from where it was collected and subsequently analyzed in a Malvern Mastersizer. Equivalent batch experiments were performed, for the purpose of comparison, in a 50-mL agitated beaker. At a supersaturation of 500, defined as Molar concentration of Na 2 SO 4 or BaCl 2 Molar solubility of BaSO4 the cone produced crystals at 6000 rpm that had a Sauter mean diameter of 3.2 microns, compared to 6.85 microns from the batch runs. However, at a supersaturation of 5000, the batch yielded a Sauter mean diameter of 0.75 microns, compared with 0.18–0.32 microns from the disc. The particle size distributions reproduced in Figure 27 highlight the cone behavior at 8000 rpm more starkly, with a decided shift to 0.1–1 microns, compared with 1–10 microns in the batch. Thus it can be seen that for many systems, spinning precipitators hold out the prospect of generating the crystal size distributions that have considerable industrial interest. This view is further reinforced by some earlier work by SmithKline Beecham (20), which crystallized an unnamed product (API) on a


FIGURE 26 Spinning cone reactor layout.

15-cm spinning disc. Figure 28, reproduced from their work, compares the product size distribution from the SDR with that of the standard industrial material. Crystallization was induced by adding an antisolvent, and it can be seen that a very significant impact could be made on the normal product size distribution, which was smaller and narrower on the disc. The stainless steel disc was subject to crystal scaling after a few runs. However, a thin layer of PTFE suppressed this without significantly impairing the disc’s heat transfer performance. 2.4.

Comparative Spinning Disc Reactor Costs

It should be borne in mind that the SDR is most effectively exploited when it is run on a continuous basis. The industrial units constructed to date have had disc diameters up to 30 cm and have been capable of processing around 30 g/s of feedstock. This corresponds to a continuous annual output of (e.g., polymer) 1000 tons/year. For a typical pharmaceutical product, a 15-cm disc could process about 7 g/s, equivalent to an annual output of 200 tons. With conventional stirred-vessel technology, a roughly equivalent unit to the 30-cm SDR is a 2000-L batch reactor


FIGURE 27 BaSO4 size distribution for a supersaturation of 5000.

fabricated in 316 stainless steel, complete with computer control and a service unit delivering a disc temperature in the range 20C–150C. The cost of such a reactor is estimated to be £250,000 with a range of £200–£400K. Comparative SDR costs have been supplied by the most experienced manufacturers (Triton Chemical Systems Ltd), which has been collaborating closely with Newcastle University during the last 5 years of SDR development. 2.4.1.

Spinning Disc Reactor Layout/Specification

The SDR system is available as a “desktop” or a “floor-standing” unit. The latter is easier to use at pressure and is more versatile for use with ancillaries such as feed units. The desktop system is restricted to the core system shown within the circle in Figure 29, which is a schematic representation of the floor standing arrangement. Experience has shown that SDR design and manufacturing details are very important in ensuring that the unit operates satisfactorily over its full performance range. Thus care must be taken with respect to the feed arrangements to ensure that instantaneous stoichiometric ratios are held constant and that feed liquids are mixed exceedingly rapidly on the disc. A sophisticated feed system is provided to ensure that the delivery of two liquid feeds is accurate. If this level of complexity is not required, then there may be a significant cost saving. As shown in Figure 29, feed liquids are supplied by means of two ram injectors accurately driven by precision step motors. Product is drained from the disc casing into a


lower receiving vessel, which can be isolated during SDR operation. This allows product removal without interfering with the progress of the reaction. Specifications for all systems include: a. Constructed with 316L stainless steel contact parts for the reactor (other materials are available). b. Constructed with a vessel to take 3-atm pressure or vacuum (other pressures available). Vessel has twin-wall construction to allow heating or cooling of the jacket. c. Supplied complete with: i. Temperature measurements at the disc. ii. Electronics and interfaces for: 1. Up to 8 measurements of temperature. 2. Up to 8 measurements of pressure etc. iii. An advanced computer control and data-logging system. The main variables on a specific spinning disc reactor are: d. The material of construction. e. The range of ambient conditions under which the reaction take place (temperature and pressure). f. The range of disc speeds available. g. The degree of instrumentation required.

FIGURE 28 Recrystallization of an API: comparison of size distributions.


2.4.2.

Prices

These prices are for the spinning disc reactor as just defined in points a to c, including advanced control and data-logging software with computer and flat screen. The following ancillaries are usually necessary for the system, and these may exist, or Triton can recommend units or can supply them integrated into the system. a. Feed systems b. Heat transfer fluid system(s) for the disc/vessel/feeders c. Reaction atmosphere control systems, e.g., pressure/vacuum/gas feed Example prices are given in Table 2. As can be seen from the table, the cost of the SDR system is significantly less than that for a stirred vessel with a similar productive capacity. However, cost considerations are likely to be much less significant than the competitive edge that SDR technology is likely to bring in terms of improved selectivity and product quality.

FIGURE 29 Schematic of the SDR manufactured by Triton.


TABLE 2 Spinning Disc Reactor Costs M/C type Type K-01

Type P-01

Description

Small benchtop system, speeds from 100 to 2000 rpm, suitable for disc temperature to 100C 15-cm disc £40,000 30-cm disc £60,000 Free-standing system, speeds from 200 to 4000 rpm, suitable for disc temperatures to 100C 15-cm disc £75,000 30-cm disc £90,000 70-cm disc £120,000 Options

For either system, typical options are: a. Heat/cool system, to provide cooling at the walls to 20C and heating/cooling at the disc for 5C to 120C b. Injector system, 2000-mL capacity, with twin walls to allow heating cooling; feed rates from 0.07 to 7 mL/s; fast down, with index and control system c. Stirrers for the injectors d. Cooled deflector ring system e. Receivers, for operation up to 10 bar 1L 2L 4L f. Additional instrumentation Measurement of heat transfer fluid flow rate Measurement of temperature difference across disc Added temperature measurements (per channel) Added pressure measurement (per channel)

3.

Price

Price £17,500 £5,000

£1,000 £1,400 £980 £1,400 £1,960 £1,400 £400 £100 £400

CONCLUSION

It will be apparent from the examples cited that centrifugal fields in general, and rotating surfaces in particular, can exert a profound influence upon process engineering. Our experience at Newcastle University has repeatedly demonstrated that the greatest gains are achieved when the SDR is presented with the most severe process conditions. While our original target was simply to reduce equipment size and installed cost, it rapidly became apparent that reduced size alone would be insufficiently persuasive for the acceptance and adoption of SDR technology. However, it has now been demonstrated that in addition to yielding


a better product (i.e., tighter size/molecular weight distributions, improved purity, etc.), the SDR can transform the whole business process. As with any radical step, the main obstacle to its implementation is the normal human resistance to change. In order to overcome this it is vital that chemists who are developing new processes be as fully conversant with SDRs as with their beakers and flasks. There is therefore a latent market for miniature versions of this equipment throughout the world’s industrial and teaching laboratories. Once the cultural block is overcome and company decisionmakers fully appreciate the breadth of impact in prospect, then the process industry will be fully prepared to meet the challenges of the new century. REFERENCES 1. 2. 3.

4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

15.

Mallinson R, Ramshaw C. European Patent No. 2568B, 1969. Woods WP. The hydrodynamics of thin liquid films flowing over a rotating disc. Ph.D. dissertation, Newcastle University, Newcastle, U.K., 1995. Thomas S, Faghri A, Hankey W. Experimental analysis and flow visualization of a thin liquid on a stationary and rotating disc. Trans Asme—J Fluids Eng 1991; 113 (March):73–80. Wasden FK, Dukler AE. A numerical study of mass transfer in free-falling wavy films. AIChE 1990; 36(9):1379–1390. Oron A, Davies SH, Bankoff SG. Long-scale evolution of thin liquid films. Rev Modern Physics 1997; 69(3):931–980. Brauner N, Maron DM, Dukler AE. Modeling of wavy flow in inclined thin films in the presence of interfacial shear. Chem Eng Sci 1985; 40(6):923–937. Brauner N, Maron DM. Characteristics of inclined thin films, waviness and the associated mass transfer. Int J Heat Mass Trans 1982; 25(1):99–110. Carslaw HS, Jaeger JC. Conduction of Heat in Solids. 2d ed. Oxford University Press, Oxford, 1959:101. Higbie R. The rate of absorption of pure gas into a still liquid during short periods of exposure. Trans Am Inst Chem Eng 1935; 31:365. Saline Water Conversion Report for 1959. U.S. Dept. of Interior Office of Saline Water, 1959:40. Saline Water Conversion Report for 1957. U.S. Dept. of Interior Office of Saline Water, 1957:7. Ramshaw C. The opportunities for exploiting centrifugal fields. Heat Recovery Systems CHP 1993; 13(6):493–513. Ramshaw C, Winnington TL. An intensified absorption heat pump. Proc Inst Refrig 1988; 85:26–39. Gilchrist K, Lorton R, Green RJ. Process intensification applied to aqueous Li B rotating absorption chiller with dry heat rejection. 7th UK National Conference on Heat Transfer, Nottingham, U.K., Sept. 10–12, 2001. Koerfer M. Hydrodynamics and mass transfer of thin liquid films flowing on rotating perforated discs. Departmental Report, Chemical Engineering Department, Newcastle University, Newcastle, UK, 1986.


16. 17. 18.

19.

20. 21.

22. 23. 24.

Aoune A, Ramshaw C. Process intensification: heat and mass transfer characteristics of liquid films on rotating discs. Int J Heat Mass Trans 1999; 42:2543–2556. Jachuck RJJ, Ramshaw C. Process intensification: heat transfer characteristics of tailored rotating surfaces. Heat Recovery Systems CHP 1994; 14(5):475–491. Boodhoo KVK, Jachuck RJJ, Ramshaw C. Process intensification: spinning disc polymerizer for the manufacture of polystyrene. In: Ramshaw C, ed. 1st International Conference on Process Intensification in the Chemical Industry, Antwerp, Dec. 1995. Boodhoo KVK, Jachuck RJJ, Ramshaw C. Spinning disc reactor for the intensification of styrene polymerisation. In: Semel J, ed. 2nd International Conference on Process Intensification in Practice, Antwerp, Oct. 1997. Oxley P et al. Evaluation of spinning disc reactor technology for the manufacture of pharmaceuticals. IEC Res 2000; 39(7):2175–2182. Vicevic M, Jachuck RJ, Scott K. Process intensification for green chemistry: rearrangement of -pinene oxide using a catalyzed spinning disc reactor. 4th International Conference on Process Intensification for the Chemical Industry, Brugge, Belgium, Sept. 10, 2001. Mullins JW, ed. Crystallization. 2d ed. Butterworths, London, 1972:162. Mullins JW, ed. Crystallization. 2d ed. Butterworths, London, 1972:162. Hetherington P, Scalley MJ, Jachuck RJ. Process intensification: continuous production of barium sulphate using a spinning cone precipitator. 4th International Conference on Process Intensification for the Chemical Industry, Brugge, Belgium, Sept. 10, 2001.


4 Compact Multifunctional Heat Exchangers: A Pathway to Process Intensification B. Thonon and P. Tochon Greth, CEA–Grenoble, Grenoble, France

1.

INTRODUCTION

In the first part of this paper, the different technologies of compact heat exchangers are presented and their range of application is given. The second part presents the state of the art for heat transfer and fluid flow characteristics for single-phase, evaporation, condensation, and heat and mass transfer. The last part presents applications of compact multifunctional heat exchangers. As proposed by Shah and Mueller (1), heat exchangers may be characterized by the compactness factor, in m2/m3, and it is generally admitted that values greater than 700 m2/m3 characterize compact heat exchangers. Often, compact heat exchangers also refer to nontubular heat exchangers, even if shell-and-tube heat exchangers can have high compactness factors. For the heat exchanger considered, the hydraulic diameter ranges from less than 1 mm to 10 mm. There are mainly two types of compact heat exchangers: the plate type (primary surface heat exchanger) and the plate–fin type (secondary surface heat exchanger). These two types of heat exchangers are described and new technologies are presented. In the process industry, there are only four basic operations: reaction, separation, mixing, and heat transfer. The traditional unit operation is to perform each task in one or more pieces of equipment sequentially, for example, heat transfer in heat exchangers and reaction in reactors. Combining two or more tasks in one piece


of equipment is implemented only for control or enhancement purposes; one example is cooling in a jacketed stirred-tank reactor. But this is not an intensified process, because it requires a batch operation, which has poor efficiency, and the reaction cannot be controlled effectively. Examples of multifunctional heat exchangers are: Multistream heat exchanger (heat transfer between more than two fluids) Reactor heat exchanger (reaction and heat transfer) Reflux condenser (heat transfer and separation) This chapter presents the state of the art in compact heat exchanger technology and provides heat transfer and mass transfer characteristics of these devices and their use in the process industry. Heat transfer is commonly required in the process industry for heating, cooling, vaporizing, or condensing. In most cases, only two streams (one hot and one cold) are in thermal contact within the heat exchanger. The most commonly used heat exchanger is the shell-and-tube heat exchanger, which has poor heat transfer performance and requires a significant volume and ground area. Compact heat exchangers and enhancement technologies allow reducing the heat exchanger volume, to increase its effectiveness and to reduce capital and operating costs. Compact heat exchanger technologies are sufficiently advanced, but their use and acceptance in the process industry are not yet widespread. Reasons are the lack of awareness of their benefits and the absence of reliable design methods and investigations under actual operating conditions. Compact heat exchangers include plate heat exchangers as well as plate–fin heat exchangers, which are characterized by hydraulic diameters between 1 and 10 mm. But recent developments in manufacturing techniques, such as printed circuit heat exchangers and diffusion-bonded and superplastic-formed heat exchangers, allow reaching hydraulic diameters below 1 mm. These heat exchangers offer compactness greater than 1000 m2/m3 and are suitable for industrial processes. Rapid advances in range of design and operational reliability have made compact heat exchangers attractive for many applications in various industries. Their high performance has already made their use widespread in the automotive, aerospace, air conditioning, refrigeration, and electrical equipment industries for single-phase and phase-change duties. In the automotive industry, plate-type heat exchangers are used as heaters, evaporators, and condensers (2), and since the 1970s, the volume/heat capacity ratio has been divided by a factor of 2 (Figures 1 and 2). This improvement has been achieved because of a radical change in heat exchanger technology and by the adoption of mass production systems integrating innovative technologies. Compact heat exchangers produced individually are generally more expensive than a conventional shell-and-tube unit, and their payback time will be longer. But taking space, weight, and convenience into account compact heat exchangers can be used cost effectively in a wider range of applications than the niches currently being used in the process industry (3).


FIGURE 1 Progress in evaporator technology. (From Ref. 2.)

2. 2.1.

COMPACT HEAT EXCHANGER TECHNOLOGY Classification of Compact Heat Exchangers

Heat exchangers can be classified in many different ways, such as according to transfer processes, number of fluids, surface compactness, flow arrangements, heat transfer mechanisms, type of fluids (gas–gas, gas–liquid, liquid–liquid, gas– two-phase, liquid–two-phase, etc.), and industry. Heat exchangers can also be classified according to the construction type and process function (Figure 3). Refer

FIGURE 2 Progress in condenser technology. (From Ref. 2.)


FIGURE 3 Classification of heat exchangers.

to Shah and Mueller (1) for further details. In the follo1wing sections, only nontubular heat exchangers will be described. 2.2.

Plate Heat Exchangers

Plate heat exchangers (PHEs) were formerly used for milk pasteurization and gradually became the standard choice for heat treatment in the liquid-food industry. Actually, pasteurization must be considered a biological reaction, because the native composition of the liquid is denatured during the heat transfer process. In practice this denaturation leads to fouling. The facility of dismantling plate heat exchangers is one of the main reasons for their extensive use in the food industry. Furthermore, because the heat transfer coefficients are high, the fluid path length will be shorter and relatively well defined. Due to the lack of large dead areas in the channels, the corresponding residence time distribution is short and homogeneous. Eventually, with the development of larger plates, their use began to grow quickly in the chemical, petrochemical, district heating, and power industries, but essentially for single-phase duties. The concept of phase change in PHEs originated in the 1970s for ocean thermal-energy conversion (OTEC) applications; the working fluids were Freon R22 or ammonia (4). These first studies on evaporation and condensation have been used for the development of PHEs in the refrigeration


industry (5–10). Now PHEs have come to be used more often in the process industry (11–12), but their use is still not widespread. In terms of technology, PHEs are made from corrugated plates (Figure 4) that are pressed together. The plate size ranges from 0.02 m2 to over 3 m2 with conventional pressing technology (Figure 5), but can reach up to 15 m2 for explosionformed plates (Figure 6). The hydraulic diameter lies between 2 and 10 mm for most common plates, but free passages and wide gap plates exist for viscous fluid applications. Typically, the number of plates is between 10 and 100, which gives 5–50 channels per fluid. Furthermore, the use of high-quality metal and manufacturing techniques makes lead plate heat exchangers less prone to corrosion failure than shell-and-tube units (13). To ensure tightness, three technologies are available: gaskets, semiwelded or totally welded, and brazing. The gasketed PHE is the most common type, with the gasket material selected as a function of the application (temperature, fluid nature, etc.). Temperatures up to 200C and pressure up to 25 bars can be achieved by such heat exchangers. For applications where gaskets are undesirable (high pressure and temperature or very corrosive fluids), semiwelded or totally welded heat exchangers are available (Figure 7). The last variant is the brazedplate heat exchanger. The plate pattern is similar to conventional gasketed units, but tightness is obtained by brazing the pack of plates. For common applications copper brazing is used, but for ammonia units nickel brazing is possible. This technology leads to inexpensive units, but the plate size is generally limited to less than 0.1 m2. The drawback is that the heat exchanger cannot be opened, and fouling will limit the range of application.

FIGURE 4 Plate heat exchanger. (Courtesy of Alfa-Laval Vicarb.)


FIGURE 5 Corrugated plates. (Courtesy of Alfa-Laval Vicarb.)

2.3.

Spiral Heat Exchangers

The spiral heat exchanger consists of two metal sheets that are welded together and then rolled to obtain spiral passages. The passages can be either smooth or corrugated; in some cases, studs or spacers are introduced between the metal sheets. These devices have two functions: (1) to adjust the spacing and (2) to induce turbulence and increase heat transfer. The general flow configuration can be crossflow (single or multipass) or counterflow, depending on the configuration of the inlet and outlet distribution boxes. The heat transfer surface ranges from 0.05 m2 for refrigeration applications (Figure 8) to 500 m2 for industrial processes (Figure 9). Spiral heat exchangers are often used for phase-change applications, because the geometry of the hot and cold stream channels can be adapted to the process specifications. Recent developments in manufacturing technologies (laser welding) have allowed the manufacture of cost-effective recuperators based on the spiral concept (Figure 10) or the folded-plate recuperator (14–15).


2.4.

Plate-and-Shell Heat Exchangers

The basic principle of these heat exchangers is to insert a bundle of plates in a shell (Figure 11). On the plate side, the fluid flows inside corrugated or embossed channels (more often in two passes). On the shell side, the flow is similar to shelland-tube heat exchangers, and baffles can be inserted. This technology can be used for revamping an application, because the shell can be kept identical to that for a bundle of tubes. These heat exchangers are often used in the process industry as boilers (boiling on the shell side) because high pressures can be reached very easily on the shell side. Furthermore, a large gap on the shell side allows the use of dirty services, because cleaning is possible via removal of the bundle of plates. 2.5.

Plate–Fin Heat Exchangers

Aluminium plate–fin heat exchangers (PFHEs) were initially developed in the 1940s to provide the aerospace industry compact, light, and highly efficient heat exchangers for gas/gas applications. Because the mechanical characteristics of aluminum are increased at low temperatures, this technology has been used since 1950 for the liquefaction of natural gases. Nowadays, aluminum plate– fin heat exchangers are extensively used in applications such as air separation, hydrocarbon separation, and industrial and natural gas liquefaction (16). The plate–fin heat exchanger offers process integration possibilities (12 simultaneous

FIGURE 6 Explosion-formed plate. (Courtesy of Packinox.)


FIGURE 7 Various welded plate heat exchangers.


FIGURE 8 Spiral heat exchanger for the refrigeration industry. (Courtesy of Spirec.)

different streams and more in one single heat exchanger) and high efficiency under close temperature approach (1–2C) in a large variety of geometric configurations. The brazed plate–fin exchanger consists of stacked corrugated sheets (fins) separated by flat plates, forming passages that are closed by bars, with openings for the fluid inlet and outlet (Figures 12 and 13).

FIGURE 9 Spiral heat exchanger for the process industry. (Courtesy of Kapp.)


FIGURE 10 Laser welding of a spiral recuperator. (Courtesy of ACTE.)

In its simplest form, a heat exchanger may consist of two passages, with the cooling fluid in one passage and the warming fluid in the other. The flow direction of each of the fluids relative to one another may be countercurrent, cocurrent, or crossflow. The fins and the parting sheets are assembled by fusion of a brazing alloy to the surface of the parting sheets. The brazing operation happens in a vacuum furnace in which the brazing alloy is heated to its point of fusion. All parts in

FIGURE 11 Shell plate heat exchangers. (Courtesy of ACM and Barriquand.)


FIGURE 12 Components of a brazed aluminum plate–fin heat exchanger. (From Ref. 12.)

contact are bonded by capillary action (Figure 14). Once the brazing alloy has solidified, the assembly becomes one single block. All passages for flow distribution and heat transfer of the streams are contained in the internal geometry of the block. Inlet and outlet headers with nozzles for the streams are fitted, by welding, around the openings of the brazed passages. These nozzles are used for connecting the heat exchanger to existing plant pipework. Numerous fin corrugations have been developed, each with its own special characteristics (Figure 15). Straight fins and straight perforated fins act like parallel tubes with a rectangular cross section. Convective heat exchange occurs due to the friction of the fluid in contact with the surface of the fin. The channels of serrated fins are discontinuous, and the walls of the fins are offset. For air flows,


FIGURE 13 Plate–fin heat exchangers. (Courtesy of Nordon Cryogénie.)

louver fins are extensively used; for process applications (single- and two-phase), continuous or offset strip fins are used. For higher-temperature applications or when aluminum is not acceptable, stainless steel (temperatures up to 700C) or copper materials can be used. For very high temperatures (gas turbine heat recovery; T 1200C), a ceramic plate–fin heat exchanger has also been developed (17) (Figure 16). For high-pressure applications in the hydrocarbon and chemical processing industries, a titanium compact heat exchanger has been developed by RollsLaval. This heat exchanger consists of diffusion-bonded channels that are created by superplastic forming of titanium plates (18). This heat exchanger can handle high pressure and corrosive fluids and is suitable for marine applications. 2.6.

Flat Tube-and-Fin Heat Exchangers

The concept of flat tube and fins in heat exchangers has been developed in the automobile industry for engine cooling and air conditioning (19–21). In such applications one of the two fluids is air and the other is either water or a refrigerant. The nonequilibrium of the heat capacities of the two fluids leads to the adoption of different enhancement technologies for both fluids. Generally on the air side the surface is finned (plain or louver fins), and on the other side the fluid flows in small-diameter channels (Figures 17 and 18). The technology is based on


FIGURE 14 Stainless steel brazed plate–fin heat exchanger. (Courtesy of Nordon.)

assembling aluminum elements, by either mechanical expansion or brazing. For conventional applications, the pressure can be up to 20 bars. Recently, heat exchangers with operating pressures up 140 bar have been manufactured (22) for car air conditioning systems, using carbon dioxide as refrigerant. 2.7.

Microchannel Heat Exchangers

Microchannel heat exchangers are compact heat exchangers where the channel size is around or lower than 1 mm. Such heat exchangers have been developed for severe environments, such as offshore platforms (23). New applications are also arising for high-temperature nuclear reactors (24). To manufacture such small channels, several technologies are available (25): chemical etching, micromachining, electrodischarge machining, etc. The most common one is the printed circuit heat exchanger developed by the Heatric Company. The channels are manufactured by chemically etching into

FIGURE 15 Different fin geometries.


FIGURE 16 Ceramic recuperator. (Courtesy of Céramiques et Composites.)

a flat plate. The plates are stacked together and diffusion bonded. These heat exchangers can support pressures up to 500–1000 bar and temperatures up to 900C (not simultaneously with high pressure). The typical size of the channels is 1.0 2.0 mm, and the plate size can be up to 1.2 0.6 m (Figure 19). The processing technique is as flexible as for plate–fin heat exchangers, and crossflow

FIGURE 17 Condenser. (Courtesy of Livernois.)


FIGURE 18 Extruded aluminum minitube.

and counterflow configurations are employed. The main limitation of the microchannel heat exchanger is the pressure drop, which is roughly inversely proportional to the channel diameter. For high-pressure applications, the pressure drop is not a constraint; but for other fields of application it will be the main barrier to the use of such heat exchangers. More recently, Chart-Marston has developed the Marbon heat exchanger (26). This heat exchanger is made of stainless steel plates stacked and bonded together (Figure 20). Several configurations are possible: (1) shell and tube, and (2) plate–fin. The use of such a heat exchanger as a chemical reactor is under consideration, and the thermal and hydraulic characterization has been undertaken as Europeanfunded project (27). Very compact heat exchangers are also used for cooling electronic devices or microreactors (Figure 21). In these heat exchangers the channel size ranges from 50 m to 1 mm. Single phase and boiling are encountered in such applications (28). Applications in the chemical processing industries are also foreseen (29). These units can be very small in size and the heat duty per unit volume is very high, up to15 kW/cm3. 2.8.

Matrix Heat Exchangers

Perforated, or matrix, heat exchangers are highly compact and consist of a stack of perforated plates made of high-thermal-conductivity material, such as copper or aluminum, alternating with spacers of low thermal conductivity, such as plastic or stainless steel. The pack of alternate low- and high-thermal-conductivity


FIGURE 19 Printed circuit heat exchanger. (Courtesy of Heatric.)

plates are bonded together to form leakproof passageways between the streams (Figure 22). The main bonding technique adopted is diffusion bonding; more information can be found in Ref. 30. Such heat exchangers have been developed for cryogenic and lowtemperature applications (31) and for fuels cells (32). They are suitable for a large range of operating conditions, but there is very little information on their thermal and hydraulic behavior. Furthermore, as the heat is transferred by conduction in the plate, the temperature distribution is not homogeneous. 2.9.

Selection of Heat Exchanger Technology

The selection of the technology of compact heat exchangers depends on the operating conditions, such as pressure, flow rates, and temperature, as well as on other parameters, such as fouling, corrosion, compactness, weight, maintenance, and reliability. Table 1 summarizes the major limits for the different types of compact heat exchangers.


FIGURE 20 Marbond heat exchanger. (Courtesy of Marston.)

3. 3.1.

SINGLE-PHASE FLOW Flow Pattern

For corrugated heat exchangers, the flow is almost three-dimensional, and the velocity field is difficult to measure. Flow visualization (33–34), realized in a high-scale channel, clearly shows a recirculation area downstream of the corrugation edges (Figure 23). These areas are large at low Reynolds number (left of picture). But the transition to turbulent flow (Figure 24), which occurs at Re 200, reduces the size of these areas. Local information (35–38) on the heat

FIGURE 21 Microchannel heat exchangers.


FIGURE 22 Matrix heat exchanger.

transfer and velocity fields indicates that the heat transfer coefficients are linked to the flow pattern and to the mixing intensity in the channel; variations of 50% were measured. Furthermore, the size of the recirculation areas could be limited by a proper design or by choosing appropriate operating conditions. The work of Amblard (39) and Hugonnot (34) has shown that choosing the appropriate corrugation angle, and with the Reynolds number above a critical value, a quasi–plug flow can be obtained. This indicates that corrugated heat exchangers could be suitable for chemical reactions. The angle of corrugation has some influence on the global flow pattern. If we consider the entire channel as a two-dimensional medium, the flow behavior can be studied. At this scale the flow can be considered homogeneous if there is the same flow rate through the channel width. Thonon et al. (37) have shown that for low aspect ratio and low corrugation angle ( 30), there is flow inhomogeneity, up to 15% of the flow rate distribution. But at high corrugation angle or for higher aspect ratio (Ar 2), the flow is almost homogeneous. In plate–fin heat exchangers, the flow structure depends on the fin geometry. Continuous fins can be assimilated to rectangular channels and the flow is almost identical to pipe flows. For offset strip fins or louvered fins, there is a high degree of mixing (40), and the flow becomes turbulent even at low Reynolds number (Figure 25). 3.2.

Heat Transfer and Pressure Drop

For corrugated heat exchangers, extensive information is available in the literature (37–42); these studies have shown that the major geometric parameter is the


corrugation angle. The enhancement of heat transfer, compared to a smooth channel, is up to six times greater. But at the same time, the pressure drop can be increased by a factor of 100. Other geometric parameters are also influential, and manufacturers are continuously improving the plate’s design. For plate–fin heat exchangers in single-phase flow, the heat transfer coefficients are related to the developed heat transfer surface, and the area ratio must be taken into account. As related to the projected surface, the overall heat transfer coefficient is very high. Heat transfer and pressure drop can be estimated from correlations (43–44), but these correlations give only an estimate of the performance, because local modification of the fin geometry will affect heat transfer and pressure drop. For microchannel heat exchangers there is a large discrepancy between various experimental reports (45). Recent studies (46) have shown that down to TABLE 1 Operating Conditions of Compact Heat Exchangers

Technology Aluminum plate–fin heat exchanger Stainless steel plate–fin heat exchanger Ceramic plate–fin heat exchanger Diffusion-bonded heat exchanger Spiral heat exchangers Matrix heat exchangers Flat tube-and-fin heat exchanger Brazed-plate heat exchanger Welded-plate heat exchanger Plate-and-shell heat exchanger Gasketed-plate heat exchanger Graphite-plate heat exchanger Plastic-plate heat exchanger

Maximal pressure (bars)

Maximal temperature (C)

80–120

70–200

10

No

80

650

2

No

4

1300

2

No

500–1000

800–1000

2

No

30 1000 200

400 800 200

2 2 2

Yes No No

30

200

2

No

30–40

300–400

2

Yes/no

30–40

300–400

2

Yes/no

20–25

160–200

2

Yes

7

180

2

Yes

5

200–250


Number of streams Fouling

2

Yes/no

FIGURE 23 Flow structure in a corrugated channel: laminar flow (Re 150).

0.5 mm, there is no significant deviation, compared to large tube correlations, but the size of the channels has to be accurately measured. More work is still needed to fully understand heat transfer and fluid flow in submicronic channels. 3.3.

Fouling

The design of heat exchangers under foulant conditions results in oversizing, thus substantially raising the cost of plants. Fouling is also responsible for process inefficiencies, due to increased thermal resistance. In water-cooling applications, particulate and precipitation fouling are frequently responsible for the decrease in heat transfer performance. Hence, the thermal and hydraulic performances need to be well understood if the heat exchange capability of practical equipment needs to be accurately predicted.

FIGURE 24 Flow structure in a corrugated channel: turbulent flow (Re 5000).


FIGURE 25 Flow structure in plate–fin heat exchangers.

Plate heat exchangers are frequently used in industrial processes because they are more compact and have higher thermal performance than conventional shell-and-tube heat exchangers, and it is generally admitted that plate heat exchangers are less prone to fouling than conventional shell-and-tube heat exchangers due to the higher level of the shear stress. Measured fouling resistances (47–50) clearly indicate that fouling resistance values are about 10 times lower in corrugated channels than on a plain surface and that the geometry (Figures 26 and 27) and the fluid velocity (Figure 28) are the major influential parameters. For instance, it has been shown that the fouling resistance is almost inversely proportional to square of the fluid velocity. This implies that the fluid velocity has to be controlled properly while operating plate heat exchangers.

FIGURE 26 View of the deposit (30 corrugation angle at a velocity of 0.5 m/s).


FIGURE 27 View of the deposit (60 corrugation angle at a velocity of 0.5 m/s).

While sizing plate heat exchangers, the great sensitivity of fouling to the fluid velocity and channel geometry precludes the use of a single value for the fouling resistance. If the TEMA fouling resistance values are applied to plate heat exchangers, excess heat transfer surface will be required, which can lead to poor efficiency. It is often recommended that the fouling margin not exceed 25% of the

FIGURE 28 Void fraction in a corrugated heat exchanger.


extra heat transfer surface. Furthermore, the inverse velocity dependence of the fouling resistance needs to be taken into account at the design stage. If the extra surface required for fouling is provided by adding plates, a maximum heat duty is reached; but adding more plates will finally reduce the heat duty. In design procedures, this needs to be taken into account. In the case of severe particulate fouling conditions, 60 corrugation angle plates should be selected rather than lower corrugation angles, and a minimum fluid velocity of 0.3 m/s is suggested. 4. 4.1.

PHASE-CHANGE HEAT TRANSFER Two-Phase Flow Characteristics

During vaporization or condensation, thermal and hydraulic performances depend essentially on the two-phase flow structure. Furthermore, as very often in industrial processes, the heat exchanger operates in thermosyphon or under natural circulation; knowledge of the pressure drop and liquid holdup is of major importance. But there is very little information on two-phase flow characteristics in compact geometries. Carey (51) has studied pressure drop and void fraction in different types of compact heat exchangers and has outlined the differences with plain tube geometries. Kreissig and Muller-Steinhagen (52) and Margat et al. (53) have shown that the principles of the methods developed for plain tubes can be used but need to be adapted. The main results of these studies are that the liquid holdup is significantly affected by the mass flow rate (Figure 28); the liquid holdup is underestimated by conventional correlation; the two-phase flow multiplier can be estimated from a Chisholm-type correlation. Winkelmann et al. (54) have studied air–water flows in a corrugated heat exchanger. Flow visualization and two-phase pressure drop measurements have been performed. The flow visualizations have shown that the flow pattern is complex and that a wavy or a film flow occurs in most cases (Figure 29). The two-phase pressure drop depends on the total flow rate and vapor quality, and Chisholm-type correlation is proposed. More work is required to characterize the flow structure in compact heat exchangers and to develop predictive methods for the frictional pressure drop and the mean void fraction. 4.2.

Vaporization

Vaporization is the most common unit operation to be found in the process industry; the use of compact heat exchangers as evaporators began 40 years ago for the concentration of sugar or salt solutions. Nowadays, compact heat exchangers are used in several industrial processes, and this is particularly true for plate–fin heat exchangers, which are closely integrated in distillation and separation processes of natural and industrial gases. In most cases, evaporation takes place in an upward


FIGURE 29 Two-phase flow visualization in plate heat exchangers.

flow and the fluid enters subcooled. The heat exchanger operates either in natural or forced circulation. In the refrigeration industry, the heat exchangers are often located directly after the expansion valve, so a two-phase mixture is fed at the bottom of the heat exchanger. The main problem with such a configuration is to obtain a homogeneous phase distribution among the channels. Some manufacturers insert a distribution device in the inlet port, and it has been shown that a significant improvement in thermal performance can be achieved. For large compact heat exchangers operating under two-phase flow at the inlet, each phase of the mixtures is fed independently in the channel in order to ensure an effective phase distribution. The design of compact heat exchangers for vaporization duties requires knowledge of the heat transfer coefficients, and it is generally admitted that the basic mechanisms occurring during flow boiling are similar to those encountered in plain tubes (55–57), but no general predictive method is available. For plate heat exchangers, most of the data published have been obtained with pure refrigerants, and the operating conditions are rather different than those encountered in the process industry. The general trend of these studies is that the heat transfer coefficients are significantly higher than those obtained in conventional plain tubes, but the pressure drop is also increased. Concerning boiling of mixtures, compact heat exchangers provide high single-phase heat transfer coefficients; hence the vapor phase will be well mixed, and no major degradation of the heat transfer coefficient should be observed. Concerning plate–fin heat exchangers, the design of such units is much more complicated because up to 12 different fluids can flow in the heat exchangers.


For distillation and separation processes, the fluids encountered are mixtures, and specific pinch methods have been developed (58–59); a temperature approach of less than one degree can be obtained. To optimize the number of layers per fluid and the type of fins, the heat transfer coefficients need to be accurately predicted in the different boiling zones (subcooled, nucleate, convective, dryout, etc.). Therefore, tests were performed with cryogenic fluids, hydrocarbons (60–61), and refrigerants (62) in order to measure the heat transfer coefficients under actual flow and geometric conditions. Tests with cryogenic fluids are required because the behavior of such fluids is significantly different than that of organic fluids. Microchannel heat exchangers are used for boiling applications, but there is a lack of data for process fluids. Studies using water or refrigerant at low pressures have outlined differences in the flow pattern and in the heat transfer coefficients (63–64). This comes from the fact that the bubble diameter is limited by the channel size. For low and intermediate vapor qualities, the heat transfer coefficients are increased as compared to plain tubes of larger diameters. But for higher vapor qualities, partial dryout may occur and will reduce heat transfer. Because microchannel heat exchangers might operate under large temperature differences, estimation of the critical heat flux is important; but most of the correlations were obtained for flat plates and single tubes, so their extension to microchannels is doubtful (65). To select and apply a boiling method to compact heat exchangers, several facts must be taken into account. The method must be based on the fundamental mechanisms occurring, because purely empirical and curve-fit methods cannot be generalized. The basic assumptions for developing such a method is that both nucleate boiling and convective boiling occur and that the dominant mechanism depends on geometric and operating conditions (Figures 30 and 31). The effect of

FIGURE 30 Vaporization in a plate–fin heat exchanger.


FIGURE 31 Vaporization in plate heat exchanger.

geometry must be taken into account in the convective area, and that of the operating conditions in the nucleate regime. It is assumed that nucleate boiling correlations, established for plain tubes, are sufficiently accurate and general to be applied for compact heat exchangers. In the convective regime, the heat transfer coefficient is reported to the single-phase liquid heat transfer coefficient, and an enhancement factor is introduced to take into account the liquid–gas interactions. The main questions are to evaluate the single-phase heat transfer coefficient of the given geometry, especially at low Reynolds numbers, and to characterize the enhancement factor. The fundamental problem is to learn whether enhancement factors developed for plain tubes can be applied successfully to compact geometries. Work is being carried out in several R&D organizations, and knowledge on boiling in compact geometries should be improved in the near future. 4.3.

Condensation

Condensation occurs in many industrial processes, but rarely with pure fluids. The fluids encountered are mixtures, and noncondensable gases are often present; this makes the condensation process very complex. In compact geometries the heat transfer coefficient depends on the two-phase flow pattern (51–67). For low condensation rates, the heat transfer is gravity controlled, and the heat transfer coefficient depends on the liquid film thickness. For higher condensation rates, the heat transfer coefficient depends on the vapor shear effect, and for small passages the liquid–vapor interaction leads to high heat transfer coefficients.


In the case of mixtures or in the presence of a noncondensable gas, the condensing vapor must diffuse through the gas to the phase interface. For this to happen a partial pressure gradient toward the phase interface is necessary. The partial pressure of the vapor falls from a constant value at a rather large distance from the phase interface to a lower value at the interface. Correspondingly, the accompanying saturation temperature also falls toward the interface. Therefore, during condensation, the condensing vapor arrives by diffusion at the condensate surface, and it is the thermal resistance in the vapor that limits the process. Hence in order to improve the heat transfer, one must reduce the thermal resistance on the vapor side. Several factors can enhance the condensation process by reducing the vaporside resistance. During condensation of mixtures or of vapors that contain noncondensables, the heat transfer on the vapor side can be improved by raising the vapor velocity. It has been shown that the heat transfer coefficient can be improved by approximately 30% by increasing the vapor velocity. The use of a finely undulated surface can also achieve significant augmentations in heat transfer during condensation. It has been shown that corrugation can promote turbulent equilibrium between the phases and thus contribute to the increase in heat transfer. Compact heat exchangers are characterized by small hydraulic diameters (1–10 mm), and there is no reliable design method to estimate heat transfer coefficients during condensation in such small passages. In the available literature, condensation of mixtures and of vapor in the presence of noncondensables has been studied, but essentially for conventional geometries (plain tubes), and only few results have been published with fluids representative of actual process conditions (hydrocarbons). A 2002 study (68) has shown that in laminar regime there was a significant mixture effect (decrease in the heat transfer coefficient), but in turbulent regime the mixtures behave as pure fluids (Figure 32). This outlines the high degree of mixing encountered in such heat exchangers, which is not surprising, since the geometric pattern of the plates is similar to modern column packing. 5. 5.1.

HEAT TRANSFER AND MASS TRANSFER Macromixing

Macro- and micromixing are two major issues for compact multifunctional heat exchangers. Macromixing is closely linked to the flow pattern and has often been studied using flow visualization techniques. An effective macromixing will give high transfer rates and homogeneous flow distribution in the channel. The typical scale of macrostructures ranges from half of the channel height to one-tenth of the channel height. Micromixing is associated with much smaller scales and will affect the reaction rate. In this section, we will focus on two technologies able to produce both heat transfer and mixing: the corrugated heat exchanger and the plate–fin heat exchanger.


FIGURE 32 Mixture effect during condensation.

These heat exchangers have a flexible design and high heat transfer performance and might be suitable for combined heat transfer and mass transfer services. The thermal-hydraulic performances of the two technologies were presented in previous sections. Flow visualization has often been used for analyzing flow structures, but it only gives access to qualitative information or time-average measurements. Using advanced numerical methods allows simulating single-phase flow in complex geometries, such as those encountered in compact heat exchangers. The determination of the mixing ability for water at a Reynolds number equal to 2000 is managed under the aspects of the computational fluid dynamics (CFD) method. An extensive review of turbulence models useful for compact heat exchanger simulation is available in the literature (70). The most appropriate model for each selected geometry will be discussed next. For corrugated heat exchangers, the flow is almost three-dimensional. Analyzing the stream function inside the 60 heat exchanger (Figure 33), we can see that flow is mainly in the direction of the flow inlet, which is a characteristic of subchannel flow. However, a small part of the fluid is deviated by the channels: furrow flow. So for this angle, quasi–plug flow can be obtained, which is in


FIGURE 33 Stream function inside the 60-angle plate heat exchanger.

accord with the work of Amblard (39) and Hugonnot (34). The flow can be considered homogeneous since there is the same flow rate through the channel width. This indicates that high teta corrugated plate heat exchangers could be suitable for chemical reactions where macromixing is needed. Analyzing the stream function inside the 30-angle heat exchanger (Figure 34), we can see that the flow feature is equally in the direction of the flow inlet (subchannel flow) and inside the furrow (furrow flow). This result is in good agreement with Gaiser (35). For a two-dimensional wavy channel, a numerical analysis performed at GRETh (69) has shown that the mixing efficiency of corrugated channels is excellent because one corrugation is nearly sufficient to have perfect mixing (Figure 35). For finned passages, all the described phenomena (Figure 36) are in good accord with the regime predicted by the literature (70). Indeed, the flow hits the front edge of the rectangular obstacle and separates immediately. The shear layer reattaches to the wall and splits in two parts: one part flows upstream, creating a recirculating and high-shear area; the other is convected downward by the mean flow. With time, the shear layer becomes unstable near the reattachment point, and it oscillates; it generates a vortex production inside the bubble and a growth of the recirculating area. Due to impinging shear layer instabilities, the

FIGURE 34 Stream function inside the 30-angle plate heat exchanger.


FIGURE 35 Numerical analysis of the mixing efficiency in a corrugated channel.


FIGURE 36 Closeup of the flow pattern inside the offset strip fin geometry.

long bubble is broken, and an eddy is liberated and convected toward the trailing edge. At the same time, the bubble length decreases. This vortex shedding occurs on both faces of the obstacle, and a Von Karman street is formed in the wake of the fin. In plate–fin heat exchangers, the flow structure depends on the fin geometry. For offset strip fins, there is a high degree of mixing (25), and the flow becomes turbulent even at low Reynolds numbers (Figure 37). Indeed, after six rows of fins (e.g., about 20 mm), the flow is homogeneous, so the macromixing in the plate–fin heat exchanger is very efficient. 5.2.

Micromixing

Micromixing technologies have only recently been applied to the design of miniaturized devices for chemical applications, so called microreactors. The main components of such microreactors are mixers and heat exchangers.

FIGURE 37 Evolution of a passive scalar inside the offset strip fin geometry.


Two ranges of applications can be distinguished: The laminar regime for highly viscous fluids (such as glue), slow reactions, and long residence time The turbulent regime for classical gas- or liquid-phase reactions, fast reactions, and low residence time The present work will only focus on the turbulent regime, which is characterized by eddies with a wide range of length and time scales. The largest eddies are typically comparable in size to the characteristic length of the mean flow. The smallest scales are responsible for the dissipation of turbulent kinetic energy, which is related to the turbulent fluctuation of the velocity. The higher the fluctuation, the smaller the scales in the flow and the better the micromixing efficiency. So for highly turbulent flows (high velocity or large hydraulic diameter), the micromixing will be very efficient, without any inventive turbulent generators. However, for compact heat exchangers where the hydraulic diameter is not large (a few millimeters) and the velocity not very high (maximum 1 m/s), enhancement techniques are needed in order to lower the turbulent-transition limit. The mixing efficiency is determined by both the value of the pressure drops and the turbulent energy dissipation, (m2/s3):

where is given by

pu L

and L is the total length of the passage, is the voidage of the passage, p is the pressure drop through the passage, and is the density of the fluid. The higher the mixing efficiency, the better is the micromixing. For corrugated heat exchangers, using the numerical simulation for both 60 and 30, we can evaluate directly from the computation and from classical correlation (5–7) (Table 2). The lower the angle, the lower the mixing efficiency. Indeed, while lowering the angle, the flow pattern moves from subchannel flow (helicoidal motion) to furrow flow (duct flow), which reduces the mixing. The mixing efficiency is very low (below 10%), so corrugated heat exchangers, which are able to produce high heat flux, e.g., high macromixing, are not able to produce micromixing. For plate–fin heat exchangers, using the numerical simulation for the OSF geometry, we can evaluate directly from the computation and from classical correlation (8) (Table 2). The OSF geometry is able to produce both macro- and microturbulent scales. For the Reynolds number considered, the fully turbulent


TABLE 2 Micromixing Performance for Plate Heat Exchangers (PHE) and Offset Strip Fin (OSF) Heat Exchangers PHE 30 angle Turbulent intensity (%) (m2/s3) (W/kg) (%)

6.2 0.43 13.1 3.2

60 angle

OSF

11.0 2.01 51.0 5.7

10 17.1 24.1 71

regime is not achieved (one must increase the velocity or the plate width by a factor of 2), but this geometry is efficient for heat transfer and mixing. It could be a good concept in the area of temperature-controlled reactions. 5.3.

Two-Phase Flow Heat Transfer and Mass Transfer

Numerous industrial operations involve a heat transfer between a liquid phase and a gaseous phase, with an important mass transfer effect, either as desorptionevaporation or as absorption-condensation. Here are some examples: reconcentration, by evaporation, of solvents, toxic industrial effluents; production, by absorption, of industrial aqueous acid solutions; reversible or irreversible chemical reactions (oxidation, hydrogenation, sulfonation); purification of permanent gases (air, smoke) by scrubbing of soluble vapors; desorbers and absorbers for heat pumps, where these two operations occur simultaneously. In these multifunctional processes, heat transfer and mass transfer are two combined and simultaneous functions, and the objective is to substantially improve these functions in order to save energy, to increase the process efficiency, and to reduce the size and cost of industrial plants. Corrugated pads are often used in the dehumidification process or in chemical heat pumps, but a higher efficiency could be reached by using diabatic units, where the wall could exchange heat with the liquid film. 6. 6.1.

APPLICATIONS Feed/Effluent Heat Exchangers

Feed/effluent heat exchangers are used in many industrial processes to warm up the fluid before the reactor and to cool it down after treatment at high temperature. The conventional design of such heat exchangers is based on shell-and-tube units. But to increase the thermal effectiveness of the heat exchangers, the required heat length becomes very important, and high pressure drop will occur.


The alternative to conventional shell-and-tube units is to use plate-type heat exchangers (71). The advantages of PHEs are a high thermal effectiveness for a low pressure drop, compactness, and a low propensity to fouling (72). In the refining industry, large PHEs are used in hydroteater (HDT and HDS); the aim of these treatments is to reduce the naphtha contaminants (sulfur, hydrogen, arsenic, lead, etc.). With the use of such units, two main problems arise: two-phase flow distribution and fouling. At the inlet of the effluent, the flow is a two-phase mixture (gas oil hydrogen) with vapor qualities up to 20%. To ensure a homogeneous flow distribution, the two phases are fed independently at the inlet of each channel (up to 200 channels in parallel). These devices allow reaching high thermal effectiveness and a low-temperature approach. Concerning fouling, because these heat exchangers are totally welded and cannot easily be cleaned, tests have been performed under laboratory conditions and on site to measure the fouling resistance. Full-scale tests, realized at the Belgium Refining Corp in Antwerp (Figure 38), have shown that after two years the thermal performance remained constant. A more precise study was also performed using the Alcor fouling apparatus. A PHE and shell-and-tube unit were connected in parallel; under similar operating conditions, the PHE exhibited no fouling while the shell-and-tube heat exchanger reached fouling resistance values up to 3 104 m2 K/W. Work is still required to optimize the design of such PHEs because the heat transfer mechanisms are complex. On the feed side, a complex mixture is vaporized; on the effluent side, the mixture is condensed with the presence of noncondensable gases. 6.2.

Process Evaporators

Concentration of liquid by evaporation is widely used in industry, and a large variety of techniques have been adopted. Two cases are considered: (1) the effluent is desolved in water (salt solution, for example); (2) the effluent and the water act as a mixture (water–acid solutions, for example). In the first case, the concentration process is very efficient and high concentration can be obtained. In specific cases, crystallization of the effluent can be achieved. Most of the technological developments have been obtained on the projects dealing with desalination of seawater or the sugar industry. For mixtures, the liquid effluent to be concentrated is partially evaporated in a heat exchanger; at the outlet, the vapor phase is richer with more volatile compounds and the liquid phase is richer with less volatile compounds. The evaporator acts as a first stage of distillation. The effectiveness of the evaporation process depends essentially on the mixture-phase equilibrium. In most of the cases, the heating fluid is steam. As a consequence, the channel geometry must me adapted on each side of the heat exchanger to achieve optimal performance. Furthermore, as in most cases, the evaporating fluid


FIGURE 38 Process flow sheet. (Courtesy of Packinox.)

has high fouling capacities, so the evaporating side must be cleanable and the dead zone should be avoided. Different types of evaporators are used in concentration processes: (1) flash evaporation through a discharge valve, (2) horizontal tubular or plate reboilers (submerged or falling film), (3) vertical tubular or plate evaporators (climbing or falling film), (4) specific evaporators (direct contact, scraped surface, etc.). Compact heat exchangers are mostly used as vertical evaporators with either falling film (Morgenroth et al. (73)) or climbing film (Brotherton (12), Patel and Thomson (11)). A special high-capacity compact falling-film evaporator has been developed for sugar plants. The use of falling-film evaporators allows one to reach higher heat transfer coefficients than with climbing-film evaporators, especially for low temperature differences. Compared to conventional systems, the overall heat transfer performance can be up to two to five times higher, as a function of the fluid


viscosity. Plate heat exchangers are also used as evaporator in sugar plants, with a similar level of enhancement. All the tests realized (lab scale, pilot and industrial units) have shown that plate heat exchangers, even with small hydraulic diameters, are not more sensitive to fouling than plain tubes. Other applications in concentration processes of such compact heat exchangers should be developed, especially in the area of mixture concentration. 6.3.

Integrated Heat Exchangers in Separation Processes

Distillation and separation processes for purifying products requires up to 40% of the overall energy demand in chemical processes (Trambouze (74)), and great efforts have been made by engineers to reduce this energy consumption. The integration of the reboiler and the condenser in distillation columns has already been studied, but the entire potential of integrated heat exchangers has not been achieved (Lyon et al. (75)). As an example, plate heat exchangers have been installed in an isopropanol dehydration plant as reboilers and condensers, and they have been found to be very effective from the viewpoint of heat transfer. Condensers and evaporators used in distillation processes are generally based on the horizontal shell-and-tube design, and this conventional design leads to large units, which have generally low energy efficiency. The integration is not optimal, and the heat exchanger is still a part of the unit. Introducing compact heat exchangers in place of shell-and-tube units will permit greater compactness and lower energy consumption. Because the heat exchanger operates purely in countercurrent, heat is available at a higher temperature than with horizontal shells. For long-term developments, diabatic distillation and separation units must be studied. Reflux condensers are often used in the top of a distillation column, as a first stage to separating the lighter and heavier components. Because the liquid flow and gas flow are countercurrent, a critical gas flow rate exists above which flooding occurs. Although considerable work in the literature has been devoted to free-falling films (mostly in pipes), the issue of flooding in countercurrent gas/ liquid flow has not been settled yet. Furthermore, the literature concerning flooding in narrow passages is extremely poor. Thus, for the case of compact condensers, there is no reliable tool for engineering-type predictions of flooding. From the viewpoint of heat transfer, reflux flows are more complex to study than downward flows. Heat and mass transfer occur simultaneously between the phases, and heat transfer is present between the liquid phase and the walls. Literature provides neither experimental data nor a reliable prediction method for these special system conditions. Studies about condensation inside channels have been carried out only for concurrent downward flow of pure vapors in tubes with large diameters. At present the design parameters for reflux condensation can only be estimated by the classical annular-flow model presented in heat transfer handbooks. However, this model is not applicable under the given conditions, and


it is known from practical applications that the real condensation rate in compact heat exchangers deviates drastically from predictions according to the classical model. Modeling reflux condensation (dephlegmation) in a compact plate–fin heat exchanger has been carried out (Urban et al. (76)), and the critical aspect raised is the distribution of the liquid film on the fins and the importance of the flowdistribution device. An integrated heat exchanger distillation column has been developed by Nakaiwa et al. (77), and a plate technology has been adopted. The analysis outlines the energy savings that can be obtained as well as the higher compactness. Aluminum plate–fin heat exchangers are often used as condensers in distillation and separation processes, but they require nonfouling and noncorrosive fluids. In the chemical industry, stainless steel or welded-plate heat exchangers have been used as top condensers of distillation columns, because they can be either directly installed inside the column or closely integrated outside (Figure 39),

FIGURE 39 Overhead reflux condensers.


thereby avoiding extra piping and a pump for the condensate. In these compact heat exchangers, condensation can take place in downward or reflux flows. For reflux condensation, flooding may occur at high gas-flow rates, and the design of nontubular heat exchangers is essentially based on correlations and methods developed for plain tubes or larger hydraulic diameters. Studies on reflux flows in noncircular and small-hydraulic-diameter channels are required for a better design of such apparatus. Looking more closely at the basic structure of compact heat exchangers, it can be noticed that the geometry is quite similar to the packing for distillation columns. Because the wetting area is significantly higher than for plain tubes and higher mixing occurs in liquid films, an additional rectification effect may occur during reflux condensation. This phenomenon needs to be evaluated because it may reduce the height of the distillation column. 6.4.

Reactor Heat Exchangers

The development of integrated chemical reactor heat exchangers requires sizing tools for aiding the design and operation of the process. The thermal performance of these heat exchangers is of prime importance for a global analysis of energy efficiency. Furthermore, a local analysis of flow and heat transfer conditions is also required for a better characterization in terms of chemical reactors (mixing intensity and residence time distribution). At bHr Group (Phillips et al. (78)), tests performed on fast exothermic reactions have shown that energy savings up to 40% could be achieved and that the amount of by-product was significantly reduced. Extension of this work to commercial compact heat exchangers is currently being considered; the first results indicate that chemical heat exchangers (CHEs) could be suitable as continuous chemical reactors. Catalytic plate reactors already exist, but their range of application is extremely limited (Jachuck and Ramshaw (79)). The basic idea is to bring into contact the heat source (catalytic reaction) and the heat sink (cooling medium). The catalyst is coated as a thin layer on one side of the plate, and on the other side flows the coolant fluid. For dehydrogenation applications, a plate-type catalytic reactor has been developed (Arakawa et al. (80)); the benefits are its higher flexibility, via the control of the process temperature, and its higher productivity, via the reduction in by-product formation. An alternative solution is to pack small balls of catalyst between two plates. Plate catalytic reactors can operate under high heat fluxes and probably allow innovative reaction schemes (reaction with pure oxygen or under pressure). At ECN, an example of coupling reactions has been studied for reforming of methanol. Combustion and reforming of methanol are done in two catalytic


compartments of the reactors. The catalyst has been coated on the internal surfaces of a stainless steel plate–fin heat exchanger. A monolith structure is used for each reacting channel, and a model has been developed to calculate the heat transferred between the channels. This heat exchanger reactor has been developed for integration in a fuel cell plant. The design of a heat exchanger reactor has to match several objectives: The residence time must be sufficient to ensure a complete reaction within the heat exchanger. The fluid temperature must be controlled, which implies high heat transfer coefficients. If the feed and the reactant are not premixed and well dispersed, a sufficient turbulent intensity must be generated by the channel geometry. The pressure drop must be acceptable. The chemical process gives the enthalpy of reaction, the flow rate, the reaction time, and the required reaction temperature. The first step in the sizing procedure is to calculate the required number of channels for the heat exchanger. Then the pass arrangement is selected in order to achieve the highest possible Reynolds number within an acceptable pressure drop. For example, if the total number of channels is fixed by the residence time: channels in series will induce high velocities and high pressure drop; channels in parallel will induce low velocities and low pressure drop. The second step is to estimate the heat transfer coefficient and to check that the heat flux can effectively be controlled by the secondary fluid (the lower heat transfer coefficient should be on the reaction side). The use of compact heat exchangers, where the channel characteristic dimension is between 1 and 10 mm, allow high heat transfer and mass transfer coefficients, even for low Reynolds numbers. The limitation will come from the mixing intensity, which may not be sufficient to ensure droplet breakdown and to avoid droplet coalescence, which will directly affect the reaction. Preliminary studies have shown that corrugated heat exchangers and OSF plate–fin heat exchangers are in turbulent flow even at low Reynolds numbers (Re 300) and that they should be suitable for chemical reactions. Investigation of the fluid flow is of prime interest for such applications, and advanced numerical methods provide local and instantaneous values that can be used to characterize the chemical reaction. These advanced CFD methods may also be used to develop specific compact heat exchangers for chemical reactions as well as mixing devices. Microchannels are also envisioned as a structure for heat exchanger reactors. If the channel dimensions range between 100 and 500 m, the area per unit volume is very high and allows catalytic reactions within the heat exchanger. An example of such a heat exchanger reactor is given by Rebrov et al. (81). Several studies are ongoing for applications in fine chemicals, reaction screening, and micro hydrogen reformers (29).


7.

CONCLUSIONS

In this chapter, the different types of compact heat exchangers have been reviewed and their applications described. Single-phase flow applications are now common in the process industry; due to their high thermal effectiveness, compact heat exchangers are real alternatives to conventional shell-and-tube units. For multiphase applications, the use of compact heat exchangers is still not widespread, but in some industries these heat exchangers are widely used. In particularly, in the refrigeration industry, plate heat exchangers are often used as boilers and condensers. A great opportunity to transfer knowledge and technology can be applied to compact heat exchangers. On the one hand, within the new environmental requirements, mixtures will replace the conventional refrigerants (pure fluids), and the refrigeration industry is not used to dealing with mixtures. Transfer of knowledge from the process industry, which is used to dealing with mixtures, could help the refrigeration industry. On the other hand, because the refrigeration industry already uses compact heat exchangers, the transfer of technology to the process industry will be fruitful. New applications for compact heat exchanger should also arise in environmental systems (Shah et al. [82]). Heat exchangers can also be considered an active component in the process and not only a utility. For instance, heat exchanger reactors (Phillips et al. [27]) or diabatic distillation columns (Nakaiwa et al. [77]) may be designed using compact heat exchanger technology. Process intensification (PI) is described as a key for future development in process plants (Green [83]), and because the cost of energy is now decreasing in Europe, the search for compactness in equipment is the goal to be achieved. Therefore, adopting compact heat exchangers is probably the most effective way to intensify a process. To support and develop intensive technologies, there is a need for basic studies on heat transfer and mass transfer in compact geometries (R&D projects) and also targeted actions on specific applications (demonstration projects). As outlined at the Compact Heat Exchangers for the Process Industry conference (Shah [84–86]), the development of new products must be realized in collaboration with the process industry, and the reliability of compact heat equipment is the first goal to achieve. Furthermore, manufacturers must propose manufacturing standards and design methodologies. Finally, the high potential of compact heat exchangers clearly matches the objectives of process intensification, and a much wider use should emerge within 5–10 years. REFERENCES 1. 2.

Shah RK, Mueller AC. Heat exchange. In: Ullmann’s Encyclopedia of Industrial Chemistry, Unit Operations II. Vol. B3. Weinheim, Germany: VCH, 1988:Chapter 2. Ohara T. The latest heat exchanger technology for vehicle’s. In: Shah RK et al., eds. Compact Heat Exchangers and Enhancement Technology for the Process Industry. Begell House, 1999:9–16.


3. 4. 5. 6.

7. 8.

9.

10.

11.

12. 13. 14.

15. 16. 17.

18.

19.

20.

Rom L. Heat exchangers: is it time for a change? Chem Eng 1996; (Feb):70–73. Panchal CB, Rabas TJ. Thermal performance of advanced heat exchangers for ammonia refrigeration systems. Heat Transfer Eng 1993; 14(4):42–57. Kumar H. The design of plate heat exchangers for refrigerants. Proceedings of the Institute of Refrigeration, 1991–92, 1992:5.1–5.5. Syed A. The use of plate heat exchangers as evaporators and condensers in process refrigeration. In: Foumeny EA, Heggs, PJ, eds. Heat Exchange Engineering. Ellis Horwood Series in Chemical Engineering. Vol. 1: Design of Heat Exchangers, London, 1992: Chap. 10. Navarro JM, Bailly A. Compact brazed plate heat exchanger. In: Recent Developments in Heat Exchanger Technology. Paris: Elsevier, 1994, pp 276–283. Sterner D, Sunden B. Performance of some plate and frame heat exchangers as evaporators in refrigeration systems. Proceedings of the 5th UK Heat Transfer Conference, IMechE, 1997. Pelletier O, Palm B. Condensation and boiling of hydrocarbons in small plate heat exchangers. Nordiske kôle-ogvarmepumpedager, Reykjavik Iceland, June 19–22, 1997. Palm B, Thonon B. Thermal and hydraulic performances of compact heat exchangers for refrigeration systems. In: Compact Heat Exchangers and Enhancement Technology for the Process Industry. Begell House, 1999:455–462. Patel N, Thompson. Plate heat exchangers for process evaporation and condensation. In: Foumeny EA, Heggs PJ, eds. Heat Exchange Engineering. Ellis Horwood Series in Chemical Engineering. Vol. 2. Compact Heat Exchangers: Technique of Size Reduction, 1992, Chap. 20. Brotherton F. Evaporation in plate heat exchangers. Heat Recovery Systems CHP 1994; 14(5):555–561. Turisini et al. Prevent corrosion failures in plate heat exchangers. Chem Eng Prog 1997; (Sept):44–50. Bacquet N. The spiral heat exchanger: concept and manufacturing technique. In: Shah et al., eds. Compact Heat Exchangers and Enhancement Technology for the Process Industry. Begell House, 2001:265–267. McDonald C. Low-cost compact primary surface recuperator concept for microturbines. Appl Thermal Eng 2000; 20:471–497. Alpema. The standards of the brazed aluminium plate–fin heat exchanger. Houston: Association Alpema, 1994. Ferrato M, Thonon B. A compact ceramic plate–fin heat exchanger for gas turbine heat recovery. In: Shah RK, ed. Compact Heat Exchangers for the Process Industry. Begell House, 1997. Adderley CI, Fowler JO. High-performance titanium plate–fin exchanger using a novel manufacturing process. In: Design and Operation of Heat Exchangers. Springer Verlag, 1992. Cowell T, Achaichia N. Compact heat exchangers in the automobile industry. In: Shah RK, ed. Compact Heat Exchangers for the Process Industry. Begell House, 1997: 11–28. Trauger P, Hughes GG. Construction and Performance Characteristics of the PFE Evaporator. SAE paper No. 930155, 1993.


21.

22. 23. 24.

25. 26. 27.

28.

29. 30. 31. 32. 33. 34.

35.

36.

37. 38. 39.

Webb RL. Advances in air-cooled heat exchanger technology. In: Proceedings of the Heat Exchangers for Sustainable Development Conference, paper IL.6, Lisbon, June 1998:677–692. Pettersen J, Hafner A, Skaugen G, Rekstad H. Development of compact heat exchangers for CO2 air conditioning systems. Int J Refrigeration 1998; 21(3):180–193. Johnston N. Development and applications of printed circuit heat exchangers. Compact Heat Exchangers for the Process Industry Conference, Snowbird, UT, June 1997. Takeda T, Kunitomi K, Horie T, Iwata K. Feasibility of the applicability of a diffusionwelded compact intermediate heat exchanger to next-generation high-temperature gas-cooled reactors. Nucl Eng Design, 1997; 168:11–21. Tonkovich AL et al. Microchannel heat exchanger for chemical reactors. AICHE Symposium Series 1996; 92(310):119–125. Watton B, Symonds K, Symonds S. Heat exchanger. International Patent, WO 97/21064, June 1997. BHR PI 3 Phillips C. Development of a novel compact chemical reactor heat exchanger. In: Proceedings of the 3rd International Conference on Process Intensification, BHR-Group publication, No. 38, 1999, pp 71–88. Mehendale S, Jacoby A, Shah R. Heat exchangers at micro- and meso-scale. In: Shah RK et al., eds. Compact Heat Exchangers and Enhancement Technologies for the Process Industry. Begell House, 1999:55–74. Schubert K et al. Microstructure devices for application in thermal and chemical process engineering. Microscale Thermophys Eng 2001; 5:17–39. Krishnan J et al. Manufacture of a matrix heat exchanger by diffusion bonding. J Material Processing Tech 1997; 66:85–97. Navarro JM, Bailly A. Compact brazed-plate heat exchanger. In: Recent Developments in Heat Exchanger Technology., Paris: Elsevier, 1994. Ahuja V, Green R. Application of matrix heat exchangers to thermomechanical energy recovery from liquid hydrogen. Cryogenics 1998; 38(9):857–867. Focke WW, Knibbe PG. Flow visualization in parallel ducts with corrugated walls. J Fluid Mech 1986; 165:73–77. Hugonnot P. Etude locale et performances thermohydrauliques à faibles nombres de Reynolds d’un canal plan corrugué: application aux échangeurs à plaques. Thesis, University of Nancy I, June 1989. Gaiser G, Kottke EV. Effect of corrugation parameters on local and integral heat transfer in plate heat exchangers and regenerators. In: Proceedings of the 9th Heat Mass Transfer Conference, Jerusalem, Vol. 5. New York: Hemisphere Publishing, 1990:85–90. Béreiziat D, Devienne R, Lebouché M. Local flow structure for non-Newtonian fluids in a periodically corrugated wall channel. J Enhanced Heat Transfer 1995; 2(1–2): 71–77. Thonon B, Vidil R, Marvillet C. Recent research and developments in plate heat exchangers. J Enhanced Heat Transfer 1995; 2(1–2):149–155. Stasiek J, Collins MW, Ciofalo M. Investigation of flow and heat transfer in corrugated passages. I. Experimental results. Int J Heat Mass Transfer 1996; 39(1):149–164. Amblard A. Comportement hydraulique et thermique d’un canal plan corrugé: application aux echangeurs de chaleur à; plaques. Thesis, INP Grenoble, March 1986.


40.

41. 42. 43. 44.

45. 46.

47.

48.

49.

50. 51. 52. 53.

54.

55.

56. 57.

Mercier P, Tochon P. Analysis of turbulent flow and heat transfer in compact heat exchangers by a pseudo-direct numerical simulation. In: Shah RK, ed. Compact Heat Exchanger for the Process Industry. Begell House, 1997:223–230. Raju KSN, Bansal JC. Plate heat exchanger design. In: Kakaç et al., eds. LowReynolds-Number-Flow Heat Exchanger. Hemisphere, 1983:913–932. Shah RK, Focke WW. Plate heat exchangers and their design theory. In: Shah et al., eds. Heat Transfer Equipment Design. Hemisphere, 1988:227–254. Shah RK, London AL. Laminar Flow Forced Convection in Ducts. New York: Academic Press, 1978. Manglik RM, Bergles AE. Heat transfer and pressure drop correlations for the rectangular offset strip fin compact heat exchanger. Experiments in Thermal Fluid Sci 1995; (10):171–180. Sobhan C, Garimella S. A comparative analysis of studies on heat transfer and fluid flow in microchannels. Microscale Thermophys Eng 2001; 5:293–311. Agostini B, Watel B, Bontemps A, Thonon B. Experimental study of single-phase flow friction factor and heat transfer coefficient in mini-channels. CHE Symposium, Grenoble: Edizioni ETS, August 2002, 85–89. Müller-Steinhagen HM. Fouling in plate and frame heat exchangers. In: Panchal CB, ed. Fouling Mitigation of Industrial Heat Exchange Equipment. Begell House, 1997: 101–112. Clarke RH, Pritchard AM. Design of compact heat exchangers with practically no resistance. In: Proceedings of Advances in Industrial Heat Exchangers, HEE96, Birmingham, 17–19 April, 1996. Thonon B, Grillot JM. Fouling mitigation in plate heat exchanger by a proper design. In: eds. Understanding Heat Exchanger Fouling and Its Mitigation, Lucca, Italy, 1997. Thonon B, Grandgeorge S, Jallut C. Effect of geometry and flow conditions on particulate fouling in plate heat exchangers. Heat Transfer Eng 1999; 20(3):12–24. Carey VP. Two-phase flow in small-scale ribbed and finned passages for compact evaporators and condensers. Nucl Eng Design 1993; 141:249–268. Kreissig G, Müller-Steinhagen HM. Frictional pressure drop for gas/liquid twophase flow in plate heat exchangers. Heat Transfer Eng 1992; 13(4):42–52. Margat L, Thonon B, Tadrist L. Heat transfer and two-phase flow characteristics during convective boiling in a corrugated channel; In: Shah R.K., ed. Compact Heat Exchangers for the Process Industry. Begell House, 1997:323–329. Winkelmann D, Thonon B, Auracher H, Bontemps A. Two-phase flow characteristics of a corrugated channel. Proceedings of the 20th International Congress of Refrigeration, IIR/IIF, Sydney, 1999. Haseler LE, Butterworth D. Boiling in compact heat exchangeers: industrial practice and problems. In: Chen JC, ed. Convective Flow Boiling. Begell House, 1996: 57–60. Thonon B, Feldman A, Margat L, Marvillet C. Transition from nucleate boiling to convective boiling in compact heat exchangers. Int J Refrigeration 1997; 20(8):592–597. Wambsganss MW, et al. Vaporization in compact heat exchangers. Proceedings of the Experimental Heat Transfer, Fluid Mechanics and Thermodynamics Conference, Brussels, 1997.


58. 59.

60. 61. 62. 63.

64. 65.

66. 67. 68. 69.

70. 71.

72.

73.

74. 75.

Lunsford K. Understand the use of brazed heat exchangers. Chem Eng Prog 1996; (Nov):44–53. Prasad B. The thermal design and rating of multistream plate–fin heat exchangers. In: Shah RK, ed. Compact Heat Exchangers for the Process Industry. Begell House, 1997:79–100. Wadekar VV. Flow boiling of heptane in plate–fin heat exchanger passage. Compact Heat Exchanger for Power and Process Industries. ASME, HTD Vol. 201, 1992:1–6. Watel B, Thonon B. Flow boiling of propane in a plate–fin heat exchanger. In: European Heat Transfer Conference. Heidelberg: Edizioni ETS, 2000:1171–1176. Feldman A, Marvillet C, Lebouché M. An experimental study of boiling in plate–fin heat exchangers. 2d European Thermal Sciences Conference, Rome, 1996:445–450. Kew P, Cornwell K. Confined bubble flow and boiling in narrow spaces. Proceedings of the 10th International Heat Transfer Conference, paper 18-FB-12, New York: Taylor and Francis, 1994:473–478. Tran TN, Wambsganss MW, France DM. Small circular- and rectangular-channel boiling with two refrigerants. International J Multiphase Flow 1996; 22(3):489–498. Agostini B, Watel B, Bontemps A, Thonon B. Experimental study of ascendant boiling flow in mini-channels. Zero Leakage—Minimum Charge, IIR/IIF, Stockholm, August 2002. Thonon B, Chopard F. Condensation in plate heat exchangers: assessment of a general design method. Eurotherm 47, Heat Transfer in Condensation. Elsevier, 1996:10–18. Srinavasan V, Shah RK. Condensation in compact heat exchangers. J Enhanced Heat Transfer 1997; 4:237–256. Thonon B, Bontemps A. Condensation of pure hydrocarbons and their mixtures in cross corrugated heat exchangers. Heat Transfer Eng 2002; 23(6):3–17. Thonon B, Mercier P. Flow structure, thermal and hydraulic performances of compact geometries used as integrated heat exchanger reactor. Process Intensification Conference, Antwerp, October 1999. Shah RK, Heikal M, Thonon B, Tochon P. Performances of compact heat exchanger surfaces with emphasis on numerical analysis. Adv Heat Transfer 2000; 34:363–444. Fauconnier JC, Gorenflo F, Thomas R. A shell-and-plate heat exchanger for the catalytic reforming unit at O.M. refinery Karlsruhe. Eurotherm 33, Recent Developments in Heat Exchanger Technology, Paris: E.E.T.I., 1994:313–320. Shibuya H, Morohashi M. Fouling tests using a pilot-scale Packinox heat exchanger with untreated straight-run gas oils. In: Panchal CB, ed. Fouling Mitigation of Industrial Heat Exchange Equipment. Begell House, 1997:525–536. Morgenroth B, Austmeyer KE, Mauch W. Experiences with the falling-film plate evaporator and concept for energy-efficient process schemes in the cane sugar industry. Zuckerindustrie 1995; 120. Trambouze P. Energy saving in the petrochemical industry. In: Pilavachi PA, ed. Energy Efficiency in Process Technology. Elsevier Applied Science, 1991:61–72. Lyon H, Jachuck R, Ramshaw C. Compact heat exchangers—comes Out on top: results of a survey. Heat Exchange Engineering, European Research Meeting, Birmingham, AL, April 1996.


76. Urban et al. Dephlegamator for ethylene plant—modeling of dephlegmation. In: Shah RK, ed. Compact Heat Exchangers for the Process Industry. Begell House, 1997: 525–532. 77. Nakaiwa M et al. Potential energy saving in an ideal heat-integrated distillation column. Appl Thermal Eng 1998; 18:1077–1087. 78. Phillips CH, Lauschke G, Peerhossaini H. Intensification of batch processes using integrated chemical reactors–heat exchangers. Appl Thermal Eng 1997; 17(8.10): 809–824. 79. Jachuck RJ, Ramshaw C. Developments in compact heat exchangers. Heat Exchange Engineering. European Research Meeting, Birmingham, AL, April 1996. 80. Arakawa ST, Mulvaney RC, Felch DE, Petri JA, Vandenbussche K, Dandekar HW. Increase productivity with novel reactor design. Hydrocarbon Processing 1993; (March):93–100. 81. Rebrov EV, Croon MHJM, Schouten JV. Design of a microstructured reactor with integrated heat-exchanger for optimum performance of a highly exothermic reaction. Catalysis Today 2001; 69:183–192. 82. Shah RK, Thonon B, Benforado DM. Opportunities for heat exchanger applications in environmental systems. Appl Thermal Eng 2000; 20:631–650. 83. Green A. Process intensification: the key to the survival in global markets Chem Industry1998; (March):168–172. 84. Shah RK, ed. Compact Heat Exchangers for the Process Industry. Begell House, 1997. 85. Shah RK et al., eds. Compact Heat Exchangers and Enhancement Technologies for the Process Industry. Begell House, 1999. 86. Shah RK et al., eds. Compact Heat Exchangers and Enhancement Technologies for the Process Industry. Begell House, 2001.


5 Process Intensification Through Microreaction Technology Wolfgang Ehrfeld Ehrfeld Mikrotechnik AG, Wendelsheim, Germany

1. MICROTECHNOLOGY AS A KEY FOR THE ADVANCED DESIGN OF CHEMICAL PLANTS Since the middle of the 20th century, general technological progress has been dominated essentially by a unique strategy of success, which constantly aims at comprehensive miniaturization and integration of functional elements in technical systems. The most outstanding development took place in microelectronics, where integrated circuits with hundreds of millions or even billions of transistors have become products of our daily lives. More recently, micromechanical, microoptical, microfluidic, and many other microdevices have become the basis for a multibillion dollar business, the market for microtechnology (1–3). The products of microtechnology have achieved a key position in information, communication, entertainment, automotive, and medical technologies. In the chemical and pharmaceutical industries, biochemists were the first or, at least, the fastest to make an interdisciplinary move into the promising field of microtechnology. Terms like lab on a chip, microarrays, microfluidics, and micro total analysis systems have become familiar to all working in the life sciences and at the front line of genomics, proteomics, glycomics, metabolomics, and all the other “omics” in this area. They regard biomolecules to some extent like a source of data and, consequently, have no problems applying information-based technologies to


innovations in their own field. In the same way, large information technology (IT) companies have started to enter the area of life sciences, utilizing their traditional strengths in combination with the concepts of microreaction technology. Researchers in chemical engineering are also intensively analyzing the possibilities potentially offered by the general strategy of miniaturization and integration to realize a radical change in design philosophy for modern chemical plants. This research and development work started at ICI in the late 1970s, and the term process intensification was used to characterize the novel concept. The main intention was to achieve much lower investment, operating, and maintenance costs for chemical plants, without decreasing their production capacity, by means of a dramatic reduction in plant size; they aimed at a reduction factor of 100 or even 1000 (4,5). This intention may look more like a dream than a serious concept. However, the technological progress even in standard plant items has proven, beyond doubt, that this concept has a realistic basis. One may just consider, on the one hand, a standard stirred-tank reactor with a cooling jacket having a volume of about 10 m3 and, on the other hand, a potentially equivalent reactor for the same production capacity consisting of a static mixer and a compact heat exchanger having a volume of about 0.1 m3 (6). This simple comparison demonstrates the superiority of continuous operation over batch processing with regard to specific plant volume and its importance in process intensification. Many other potential examples exist, such as spinning disk reactors, vortex scrubbers, reactor-mixing systems, and, of course, multifunctional reactors, which integrate reactions and unit operations. There is no doubt that the ultimate development of process intensification leads to the novel field of microreaction technology (Figure 1) (7–9). Because of the small characteristic dimensions of microreaction devices, mass and heat transfer processes can be strongly enhanced, and, consequently, initial and boundary conditions as well as residence times can be precisely adjusted for optimizing yield and selectivity. Microreaction devices are evidently superior, due to their short response time, which simplifies the control of operation. In connection with the extremely small material holdup, nearly inherently safe plant concepts can be realized. Moreover, microreaction technology offers access to advanced approaches in plant design, like the concept of numbering-up instead of scale-up and, in particular, the possibility to utilize novel process routes not accessible with macroscopic devices. As a matter of fact, microfabrication methods have to be introduced into chemical engineering in order to profit from the potential advantages of microreaction technology. Although this is a difficult hurdle, a few chemical companies have successfully started to utilize microreaction technology for commercial syntheses of fine and special chemicals. Nevertheless, much effort must still be spent to transfer further promising research results into commercial application and to


FIGURE 1 Evolutionary development through process intensification.

get away from traditional strategies in chemical engineering. In the following, a comprehensive analysis covering these aspects will be given. 2. EFFECT OF MINIATURIZATION ON UNIT OPERATIONS AND REACTIONS 2.1. Enhancement of Heat Transfer and Mass Transfer Processes Diffusion, thermal conductivity, and viscosity are physically similar phenomena that involve the transport of a physical quantity through a gas or liquid. The driving forces for the corresponding transport fluxes of mass, energy, and momentum are the gradients in concentration, temperature, and velocity, respectively, where in all three cases the fluxes are in the same direction as the gradients. For given differences in these properties, a decrease in the characteristic dimensions results in an increase in these gradients and, correspondingly, in higher mass and heat transfer rates as well as in higher viscous losses. Accordingly, mixing and heat exchange systems with extremely high transfer rates per unit volume can be realized by miniaturization; on the other hand, however, the effect of viscous losses has to be taken into account. Besides the effect of decreasing linear dimensions on the corresponding gradients, the effective surface area for exchange processes has to be considered. With decreasing characteristic dimensions, the surface-area-to-volume ratio of


the system increases. This results in a corresponding enlargement of the specific interface area, i.e., of the area per unit mass or unit volume, for transfer processes, so, in connection with the enhancement of the gradients, i.e., the driving forces for heat and mass transfer, extremely efficient mixers and heat exchangers can be realized by miniaturization. Furthermore, the amount of material in a system is reduced with the reciprocal third power of the characteristic dimensions, and, consequently, the response time of the microdevice is extremely reduced so that in most cases large differences concerning temperatures and concentrations are diminished immediately. It was demonstrated in many cases that highly exothermal reactions can be performed under isothermal conditions using the channels of micro heat exchangers as reaction volumes (Figure 2) (10). Pioneering work on this subject started in the late 1980s, when micro heat exchangers with extremely high transfer rates per unit volume were produced by means of advanced mechanical micromachining methods (Figure 3) (11). Meanwhile, specific heat transfer rates of more than 20 kW per cm3 have been achieved, and a broad spectrum of materials has been successfully applied. A wide variety of micromixers are also available that allow mixing times in the submillisecond range (8,12). They utilize mainly the concept of multilamination, where two streams of fluids are split into a large number of small substreams and fed alternately into an interdigital flow system, where they merge into a joint stream (Figure 4). Other concepts are based on the principle of splitting, side-to-side arrangement, and further splitting to generate an increasing number of substreams with different compositions, as known from large-scale static mixers. Vortex-type micromixers have also abeen applied. A wide range of applications for micromixers exists in the fields of gas–liquid suspensions and liquid–liquid emulsions, with extremely small bubble and droplet sizes, respectively. A high uniformity concerning size distribution is achievable; in particular, the specific power consumption for generating suspensions and emulsions is much lower than in the stirring devices or high-pressure jets usually applied in the macroscopic range (8,13). Accordingly, micromixers are promising tools to improve the performance of phase transfer and other exchange processes. 2.2. Inherent Process Restrictions in Miniaturized Devices and Their Potential Solutions As a matter of fact, miniaturization inevitably results in a number of process restrictions, and completely new problems arise, too. There are, above all, the problems of blockage of microstructures by solid particles and fouling effects. Moreover, corrosion might be much more dangerous for microscopic than for macroscopic devices.


FIGURE 2 Microreactor for parallel screening of catalysts for partial oxidation of methane. (Source: D. Hönicke, TU Chemnitz.)


FIGURE 3 Micro heat exchanger produced by means of mechanical micromachining. (a) Platelet with grooves of 30-m depth and 70-m width. (b) Assembly of crossflow heat exchanger. (c) Final devices. (Source: Forschungszentrum Karlsruhe.)

Nevertheless, if the solid particles are small enough, they will have no negative effect on the operation of a microreactor. On the contrary, microreactors can even produce pigments of higher quality, i.e., smaller size and better uniformity, than macroscopic devices. This positive result was obtained experimentally at Clariant Company; consequently, a microreactor pilot plant for pigment production is under construction (14). By means of highly efficient micromixers, Siemens Axiva Company succeeded in improving the synthesis of acrylate resins. They could avoid a detrimental portion of high-molecular-weight resin and, consequently, fouling of the main continuously operating reactor. Evidently, there are at least concrete chances to get around some of the problems resulting from small characteristic dimensions. There is, of course, no possibility of avoiding all problems inherently connected with small dimensions. For instance, gravitational forces cannot be efficiently utilized to transport fluids at small characteristic dimensions, since the effects of


surface forces might far exceed those of mass or bulk forces. This problem is immediately evident when regarding the reflux in a miniaturized distillation column or the settler in a micromixer/microsettler system. To some extent, rotating devices can be applied and centrifugal forces can be utilized for material transport. This approach has been demonstrated successfully in microfluidic systems, but it is not a general solution. Consequently, other methods for phase separation are required for miniaturized process devices, such as microfiltration to break emulsions and the utilization of hydrophobic and hydrophilic surfaces or capillary effects. Finally, surface effects will become more and more dominant in chemical reactions when the characteristic dimensions are reduced, which may produce advantages or disadvantages, depending on the respective type of reaction. 2.3. Consequences for the Selection of Reaction Routes and Plant Design The extreme enhancement in mass and heat transfer rates through miniaturization of process devices results in fundamentally novel design possibilities with respect to selecting alternative reaction routes and plant design. In contrast to macro devices like large stirring tanks, the starting conditions for a chemical reaction can be set precisely with respect to time and concentration because of the much faster mixing of educts in a micromixer. The reaction starts at precisely defined time and position with a spatially uniform composition. Thus, unfavorable reaction conditions

FIGURE 4 Micromixers. (a) Interdigital structure of a multilamination micromixer. (b) Principle of split-and-recombine static micromixers. (Source: IMM.)


due to incomplete mixing are minimized that eventually result in undesired side and secondary reactions and, consequently, losses in yield and selectivity. The high heat transfer rates achievable in micro heat exchangers and reactors avoid unfavorable reaction conditions resulting from hot spots or thermal runaway effects. An optimum temperature or temperature profile for the reaction can be chosen with respect to spatial distribution and time. Thus, a fast-flowing fluid element can be cooled down or heated up very rapidly, in fractions of a millisecond. Because of the small thermal mass of microdevices, a periodic change of temperature of the reactor can be realized, with a typical time constant of some seconds. All these examples offer possibilities to improve yield and selectivity. Since micro reactors—except for high-throughput screening in combinatorial materials research—are usually operated under continuous conditions, it seems simple to adjust the optimum residence time by means of a suitable delay loop or channel that is also favorable with respect to yield and selectivity (Figure 5). However, the flow conditions in microdevices are generally characterized by a low Reynolds number; consequently, a parabolic Hagen–Poiseuille profile will exist in long channels and ducts. This flow profile results in an unfavorable broadening of the distribution of residence time. Special channel configurations allow one to reduce this effect. 2.4.

Process Control and Safety

The inherent advantage of precise adjustment of the starting and boundary conditions for chemical reactions and unit operations in microdevices provides a novel basis for process control. Taking into account, in addition, the small holdup,

FIGURE 5 Process intensification by setting the optimum residence time.


FIGURE 6 Explosion-proof continuous synthesis in the explosive range. The reaction system consists completely of flame-retarding microchannels.

it is evident that an extremely short response time is a further inherent advantage of microreaction devices with respect to process control. As a result, there is a unique chance to utilize alternative reaction routes for chemical synthesis that so far have not been applied commercially, for reasons of safety or difficulties in process control or because it is fundamentally impossible to realize such reaction routes using macroscopic devices. This is the case, in particular, for controlled reactions in the explosive regime (Figure 6) (15). This is accessible by means of microreaction devices, since, due to their small characteristic dimensions, they act like flame retention baffles. Moreover, the small dimensions allow reactions to be performed at extremely high pressure, which is of importance for chemical processes using supercritical solvents. 2.5. Sustainable Development by Numbering-Up and Distributed Production The safety problems connected with the storage of large quantities of educts and products remain, of course, unchanged when a conventional plant is replaced by a microreaction plant with the same production capacity. Nevertheless, this problem may be reduced by replacing a large plant by several small plants for distributed production. In contrast to conventional plants with macroscopic process devices, where scale-up usually results in a considerable reduction of specific investment costs, microreaction plants may instead profit from the mass production of microdevices in reducing specific investment costs. Scale-up for achieving the desired production capacity can be done only at one site, while a plant comprising a large


number of identical chemical microdevices according to the numbering-up concept can be split up for production at several sites. As a result, microreaction technology may contribute to the strategy of sustainable development by saving resources due to a higher yield and, in particular, by flexible production on site and on demand. There are a number of further advantages to the numbering-up concept. Research results can be transferred into production faster, plants can be constructed in a shorter time, and the production capacity can be adjusted more flexibly to variations in demand. Since mass production of microdevices may result in relatively a low cost per piece, novel cost-saving maintenance and repair concepts based on disposable elements might be introduced. 3. FROM BASIC PROPERTIES TO TECHNICAL DESIGN RULES In contrast to microelectronics, where extremely powerful software tools and detailed design rules exist for the development of ultralarge-scale integrated circuits, there are no corresponding comprehensive means in microreaction technology available to date. Such design tools should comprise mathematical modeling of flow and chemical reactions in miniaturized systems as well as specifications for suitable materials and simulation of manufacturing processes applicable to the respective microreaction devices. Since it will take several years to realize such an integral software toolbox, individual approaches with separate steps have to be applied to meet gradually the requirements of microreactor design. Standard software for computational fluid dynamics is directly applicable in this context, and there are also powerful software tools for the simulation of special steps in microfabrication processes. However, there has been rather little experience with materials for microreactors, optimization of microreactor design, and, in particular, the treatment of interdependent effects. Consequently, a profound knowledge of the basic properties and phenomena of microreaction technology just described is absolutely essential for the successful design of microreaction devices. For instance, proper design rules must take into account that mixing and heat exchange systems with extremely high transfer rates per unit volume can be realized via miniaturization but that an increase in viscous losses may counterbalance the positive effects. Accordingly, suitable figures of merit must be defined for micromixers and micro heat exchangers that consider the ratio of mass or heat fluxes to pressure losses. However, the value of such a figure of merit should be always considered in context with further boundary conditions of the process and the interdependence of several process properties. Decreasing the characteristic dimensions of a system results, as already explained, in a reduction in the material holdup and a simultaneous enlargement of the surface-area-to-volume ratio of the system. These aspects also determine the speed of mixing and heat transfer and,


consequently, the degree of miniaturization required in a specific case. Such aspects have to be considered for a favorable design of a microreaction system; in some cases, extremely small dimensions are not necessary to avoid unfavorable reaction conditions resulting from hot spots or thermal runaway effects. If large amounts of materials have to be transported, a favorable design should instead consider large pumps rather than arrangements with many micropumps, which, in most cases, are commercially unattractive for cost reasons and technically less suitable because of their comparatively low efficiency. 4. MICROFABRICATION OF REACTION AND UNIT OPERATION DEVICES 4.1.

General Requirements

Since the production of chemicals in a continuous process is inevitably connected to a transport of material, three-dimensional microfabrication processes are required in order to realize sufficiently large cross sections for channels and ducts as well as reaction volumes. Meanwhile, a wide variety of such processes as well as design and test methods exist that all essentially originated from either semiconductor technology or precision engineering. Thin-film methods, applied to a large extent in semiconductor technology, are less suitable for the generation of threedimensional microreaction devices but are widely used for surface processing and protection as well as for manufacturing sensor elements. Because of the extremely wide variety of reactions, educts, products, and process conditions, a sufficiently broad spectrum of materials is required to realize suitable microdevices for chemical processes. Metals and metal alloys, plastics, glass, ceramic materials, semiconductor materials like silicon, and various auxiliary materials for sealing, surface treatment, etc. have been successfully applied for realizing microreaction devices. Besides such basic aspects concerning the shape of and materials for microreaction devices, costs play a major role in the selection of a microfabrication process. In this respect, the number of pieces and the precision that is really required, as well as aspects like availability and manufacturing experience, must be taken into account. In contrast to the situation some years ago, the prerequisites for cost-effective mass fabrication as well as small-scale production or rapid prototyping have essentially changed. Modern commercial equipment for the production of microdevices is available that allows unreliable and uneconomic laboratory-scale manufacturing devices to be replaced. Mathematical modeling of the device function may also help to cut costs, since it allows more realistic specifications to be worked out with regard to functional requirements. In addition, mathematical modeling of the process sequence for microfabrication and assembly will be useful for cost saving. Such hard and


soft aspects will be considered in more detail in the following analysis of microfabrication methods for reaction devices. 4.2.

LIGA Technology

The LIGA technology allows the production of ultraprecise micro structures with an extreme aspect ratio from a wide variety of materials (16,17). It is based on a combination of deep lithography, electroforming, and molding processes. In the first step of the manufacturing sequence, a pattern from a mask or by means of a serial beam-writing process is transferred into a thick resist layer on an electrically conductive substrate. Ultraprecise microstructures with an extreme aspect ratio can be generated by deep X-ray lithography. Using special epoxy resists like SU 8, which utilizes intrinsic optical waveguide properties of irradiated crosslinked regions, favorable results are also achievable by means of UV lithography. In the second step, the three-dimensional relief-like structure of the resist polymer generated by deep lithography is transferred into a complementary metallic structure by means of electroforming, starting from the electrically conductive substrate. Usually a nickel sulfamate electrolyte is applied, but there are also proven electrolytes available for deposition of other metals and metal alloys. The metal structure generated by means of electrodeposition may be the final product in some special cases. In general, however, it is used in a third step as a master tool for a replication process, such as injection molding, casting, or embossing, for mass fabrication of microstructures. A wide variety of mold materials can be applied for micromolding, e.g., organic polymers, preceramic polymers, and ceramic and metallic powders with organic binders for subsequent sintering, so that most material requirements for chemical microdevices can be favorably met. It should be emphasized that the development of the LIGA technology originated from a special requirement in nuclear process engineering. Curved micronozzles with characteristic dimensions in the micrometer range were needed as mass products for aerodynamic separation of the uranium isotopes in the framework of a large technological development work at the Karlsruhe Nuclear Research Center (Figure 7). Today there are a number of LIGA products that evidently have promising markets in the fields of micro-optics and integrated optics, molecular biotechnology, and microactuators. More recently, LIGA components and systems have been successfully applied to chemical engineering and microreaction technology, respectively. A number of chemical companies and, of course, research institutes utilize devices such as micromixers, micro heat exchangers, and micro bubble columns as well as modular systems with integrated functional elements for reaction, heat transfer, mixing, separation, and fluid distribution for process development. LIGA devices are also seriously being considered by the chemical industry for the production of fine chemicals.


FIGURE 7 Double-deflecting micronozzle for aerodynamic separation of uranium isotopes manufactured by LIGA technology from nickel. The smallest characteristic dimensions achieved in such devices are below 10 m. (Source: Institute of Nuclear Process Engineering at the former Karlsruhe Nuclear Research Center, now Forschungszentrum Karlsruhe, Siemens.)

4.3.

Wet and Dry Etching Processes

Wet etching processes are widely used to produce microstructures by means of transferring resist patterns into various materials. However, for most materials only isotropic etching processes exist, so, because of lateral underetching of the resist pattern, only shallow microchannels or other shallow structures can be generated at the surface of a bulk material. Three-dimensional structures can be manufactured when the pattern is etched completely through thin foils, which then have to be stacked in order to realize deep microchannels with a high aspect ratio (Figure 8). Isotropic etching has been applied several times for manufacturing microreaction devices. The technological expenditure is relatively low, but there are some restrictions concerning accuracy, surface roughness, and geometrical design. The product spectrum comprises various types of heat exchangers, micromixers, separators, reaction units, and even integrated devices with several functional elements. Wet chemical anisotropic etching of monocrystalline silicon has been widely applied in microtechnology (18,20). This method is based on the dependence of etching velocity on crystal orientation, so only a few basic geometries can be


FIGURE 8 Microetched foil of stainless steel for manufacturing micro heat exchangers by stacking and diffusion welding. (Source: Ehrfeld Mikrotechnik, Ätztechnik Herz.)

realized. Besides silicon, there has been very little manufacturing experience with other monocrystalline, inevitably very expensive, materials. Consequently, wet chemical anisotropic etching is in general not very attractive for manufacturing chemical microdevices because of strong restrictions with respect to shape and material. Nevertheless, the technological expenditure is low, and material problems can also be solved via the deposition of protection layers. A number of microfluidic devices have been manufactured by means of this method, such as micropumps, microvalves, and flow-distribution systems. Besides anisotropic etching of monocrystalline materials, another wet chemical etching process exists that uses a special type of photosensitive glass (19). A wafer consisting of such glass is irradiated through a mask with UV light and subsequently heated to a temperature between 800 and 900 K. This results in a crystallization of the irradiated regions that can be dissolved much faster in hydrofluoric acid than the nonirradiated parts. This method has been successfully applied to produce microreaction devices such as mixers, heat exchangers, and micro titer plates from glass. Precise microstructures with nearly any cross-sectional shape can be generated by means of anisotropic plasma-etching methods, where again silicon is the most important and proven material (18,20). Usually, a mask pattern is transferred into a thin layer consisting of a material resistant to plasma etching on a silicon


wafer. Subsequently, silicon is etched by means of a fluorine-containing lowpressure plasma that generates gaseous silicon compounds. In order to generate microstructures with an extremely high aspect ratio, the directed etching process is connected with a subsequent deposition process from the plasma where the walls oriented in parallel to the etching direction are covered with a plasma polymer resistant to the reactive plasma (21). By means of multiple repetition of directed etching and side wall passivation, channels and other structures with nearly vertical walls can be realized; accordingly, extremely high aspect ratios are achievable for nearly any cross-sectional shape (Figure 9).

FIGURE 9 Channel structure of a phase separator generated by ASE deep etching of silicon. (Source: IMM.)


This so-called advanced silicon etching (ASE) process achieves etching velocities on the order of 0.2 mm per hour. The ASE dry etching method has, of course, limitations concerning, e.g., material selection and surface smoothness or the brittleness of silicon, which makes it nearly impossible to use it directly as a mold insert in micromolding processes. Nevertheless, it is possible to transfer ASE silicon structures into complementary metal structures by electroforming. For a number of applications, ASE is evidently a favorable alternative to LIGA in manufacturing devices for microreaction technology. 4.4.

Mechanical Micromachining

In the past few years, impressive progress has been made in so-called mechanical micromachining, utilizing technologies based on so-called ultraprecision machining. Complex three-dimensional microstructures have been generated with shape accuracies in the submicrometer range by means of milling, turning, and grinding (11,22). Three- and five-axis ultraprecision micromilling machines are available as commercial products. Using diamond tools, an extremely low surface roughness of a few nanometers is achievable for nonferrous materials. Progress has also been made in machining stainless steel by using ultrafine-grain hard metal tools and novel technologies like vibration cutting. In addition, mechanical micromachining has been successfully applied with brittle materials. Micromixers, micro heat exchangers, and reaction systems have been successfully produced by means of this technology (Figure 3). It is evident that there are hardly any limitations concerning the generation of microstructures for chemical microdevices with complex geometries, extremely high aspect ratio, and high precision from a wide variety of materials by means of mechanical micromachining. Rather, restrictions may exist when manufacturing closely packed channels or other structures, because of the finite size of the tools. Also, manufacturing costs may become a problem in mass fabrication; but in such a case, mechanical micromachining may be helpful for manufacturing mold inserts for mass fabrication by means of micromolding. Moreover, there are other mechanical methods for high-volume production, like punching and embossing, that have been successfully applied in fabricating, e.g., micro heat exchangers. 4.5.

Microelectrodischarge Machining

An interesting alternative to standard mechanical micromilling, turning, drilling, and grinding methods is microelectrodischarge machining (EDM), which is virtually unlimited with respect to the geometrical shape of the work piece (23). Material is removed in a discharge between the electrically conductive work piece and an electrode by small sparks in a dielectric fluid such as oil or deionized water. An important advantage in micromachining is that the forces acting on the work piece


in EDM are extremely low. The disadvantages of micro EDM are a relatively large surface roughness, limitations in miniaturization because of the finite size of the electrodes and the spark gap in the electrical discharge, and very long machining times, so this method is essentially used to manufacture mold inserts or prototypes. The methods of mechanical micromachining and micro EDM have been extensively applied to the fabrication of components such as micro heat exchangers, mixers, and reaction channels as well as chemical microsystems with integrated heat exchange, reaction, mixing, and distribution elements (Figure 10). 4.6.

Micromachining by Means of Laser Radiation

Microfabrication by means of laser radiation covers a wide range of different methods (24,25). On the one hand, these are processes where material is removed in an intense electromagnetic field by melting, evaporation, decomposition, photoablation, or a combination of these phenomena. On the other hand, generating processes exist where structures are built up from liquid resins, laminated layers, or powders using, e.g., photochemically induced crosslinking of organic compounds

FIGURE 10 Micromixing element generated by microelectrodischarge machining. (Source: Ehrfeld Mikrotechnik, Zumtobel.)


FIGURE 11 Part of a static micromixer manufactured by laser ablation from aluminum oxide. (Source: Ehrfeld Mikrotechnik, Heidelberg Instruments Microtechnologies.)

in stereolithography or powder solidification by laser sintering. In addition, welding by means of laser radiation is of major importance for the connection and assembly of microdevices. There are no restrictions worth mentioning concerning materials in micromachining by laser radiation, which is a real advantage for chemical microdevices (Figure 11). However, limitations exist to achieving critical dimensions below 10 m and low surface roughness. Removal of material is also often connected with the generation of debris, which reduces accuracy. Since laser-based microfabrication processes, except lithography, are essentially serial rather than parallel machining methods, their productivity is comparatively low. Nevertheless, they offer a huge potential in rapid prototyping. Laser-based micromachining processes have been applied to date only on a relatively small scale for manufacturing chemical microdevices (27). This will probably change relatively soon, since rapid prototyping will become more and more important for developing novel microreaction devices. 5.

IMPLEMENTATION OF MICROREACTION TECHNOLOGY

Microreaction technology has created a novel basis for: Accelerating screening in combinatorial material development Realizing extremely powerful tools for the evaluation of new reaction pathways Implementing comprehensively the concept of process intensification for the production of fine and special chemicals (Figure 12)


Following the development route of miniaturization in the life sciences, the implementation of microreaction technology in combinatorial material development has been very successful. Companies like Symyx, located in Silicon Valley, whose business is based on the highly effective synthesizing and screening of a huge number of chemical compounds, have demonstrated that faster development and cost savings are achievable by means of microreaction devices. Not only can the amount of reactants, auxiliary substances, waste, energy, and space be minimized, but all the other advantages of microreaction devices mentioned earlier can also be favorably utilized (see, e.g., Ref. 27). The research work of such companies is focused on more efficient catalysts, new polymers, high-performance phosphors for illumination, and, of course, drug development and many other substances. Promising work in this direction is also being done at universities and government research centers (7,8,26,28). Researchers at BASF have shown that microreactors can be utilized that give access to operating conditions that cannot be realized by means of macroscopic equipment. They succeeded in improving yield and selectivity in a highly exothermal two-phase reaction in connection with the synthesis of a vitamin precursor. At Degussa company, a microreactor test facility for proprietary reactions is under construction. The major focus in this context is the implementation of microreaction devices as powerful tools for process development and, in particular, for the evaluation of new reaction pathways. Companies like Clariant and Merck use microreactors for production, and they are obviously convinced that the ultimate development of process intensification leads to microreaction technology. In contrast to other companies, Clariant

FIGURE 12 Microreaction technology aims at production of (a) information, (b) tools for process development, and (c) chemicals.


FIGURE 13 Modular microreaction system consisting of functional elements for reactions and unit operations arranged on a base plate. The cube-shaped modules of stainless steel with built-in microstructures have a side length of 25 mm and can be operated at pressures up to 100 bar. (Source: Ehrfeld Mikrotechnik.)

has reported about its work and its promising progress (14). Researchers at Clariant assume that about 15% of future production facilities will be based on microreaction technology. However, microfabrication methods that are usually unfamiliar to chemical engineers have to be introduced to profit comprehensively from microreaction technology. This transition from standard manufacturing methods of plant components to the development and production of microdevices is also inevitably connected with the application of special materials that are not yet proven in chemical engineering. In addition, novel design rules that have not existed until now should be implemented for the long term to speed up the development of novel devices. Essential progress is to be expected from the introduction of so-called modular microreaction systems. The system developed by Ehrfeld Mikrotechnik comprises single functional elements for reactions, unit operations, transport,


measurement, and control. The modules can be arranged and connected in a wide variety of configurations and serve as a toolbox for realizing development platforms similar to microplants (Figure 13). By means of such platforms, the optimum operation conditions of chemical processes as well as favorable plant configurations can be determined and novel reaction routes tested. There is also a wide range of applications in combinatorial chemistry. Since the microplants are usually set up for continuous operation, they have a comparatively high productivity and can be utilized directly for small-scale production of special chemicals. 6.

CONCLUSIONS

Future progress in chemical engineering will be strongly determined by process intensification through microreaction technology. It offers fundamentally novel opportunities to save direct costs in the areas of development, investment, operation, and maintenance as well as to reduce indirect follow-up expenditures in connection with storage, transport, and changes in demand or market trends. A roadmap of microreaction technology for novel process routes and efficient production is shown in Figure 14. Nearly all major chemical, chemical engineering, and pharmaceutical companies are interested in or even active in analyzing the potential of microreaction technology. Moreover, there are a number of powerful three-dimensional microfabrication technologies that should meet nearly all requirements concerning geometries as well as materials of microreaction devices in prototyping and mass fabrication. However, the implementation of a novel technology needs time. It is necessary to prove carefully the potential advantages, to develop a sufficiently broad scientific basis, to implement reliable and cost-effective fabrication of chemical microdevices on an industrial basis, to gain experience in the design, construction,

FIGURE 14 Roadmap of microreaction technology for novel process routes and efficient production.


and operation of microreaction plants, and finally to demonstrate real commercial success. Meanwhile, a lot of effort has gone in this direction, but there is still a multitude of tasks to solve until decisionmakers will be convinced enough of the commercial prospects of microreaction technology to accept the inevitable financial risks of technological progress. REFERENCES A comprehensive overview of the international development work on microreaction technology can be found in the Proceedings of the International Conferences on Microreaction Technology, which are listed in the following. Ehrfeld W, ed. Proceedings of the 1st International Conference on Microreaction Technology. Berlin: Springer, 1998. Ehrfeld W, Rinard I, Wegeng R, eds. Process Miniaturization: 2nd International Conference on Microreaction Technology, IMRET 2; Topical Conference Preprints. AIChe, New Orleans, 1998. Ehrfeld W, ed. Proceedings of the 3rd International Conference on Microreaction Technology. Berlin: Springer, 2000. Rinard I, ed. 4th International Conference on Microreaction Technology. Topical Conference Proceedings. AIChE Spring National Meeting, Atlanta, GA, March 5–9, 2000. Matlosz M, Ehrfeld W, Baselt JP, eds. Proceedings of the 5th International Conference on Microreaction Technology. Berlin: Springer, 2001. Rinard I, ed. 6th International Conference on Microreaction Technology, Conference Proceedings. AIChe Spring Meeting, New Orleans, March 10–14, 2002.

The literature cited in this contribution is listed here. 1. 2. 3. 4. 5. 6. 7.

8. 9.

Ehrfeld W, Ehrfeld U, Kiesewalter S. Progress and profit through microtechnologies. Proceedings VDE World Microtechnologies Congress, MICRO.tec, Vol. 1, 2000: 9–17. Market Analysis for Micro Systems II, 2000–2005. A NEXUS Task Force Report, 2002. Bundesministerium für Bildung und Forschung. Förderkonzept Mikrosystemtechnik 2000, Bonn, Germany, Jan 2000. Stankiewicz AI, Moulijn JA. Process intensification: transforming chemical engineering. Chem Eng Prog 2000; (Jan):22–33. Green A, Johnson B, John A. Process intensification magnifies profits. Chem Eng 1999; (Dec):66–73. Wood M, Green A. A methodological approach to process intensification. IchemE Symposium Series No. 144, 1998:405–416. Jensen KF, Hsing I-M, Srinivasan R, Schmidt MA, Harold MP, Lerou JJ, Ryley JF. Reaction engineering for microreactor systems. Proceedings of the 1st International Conference on Microreaction Technology. Berlin: Springer, 1998:2–9. Ehrfeld W, Hessel V, Haverkamp V. Microreactors. In: Ullmann’s Encyclopedia of Industrial Chemistry. 6th ed. Weinheim: Wiley-VCH, 1999. Jäckel K-P. Microreaction Technology—Vision and Reality. Plenary Lecture, ACHEMA 2000, Frankfurt.


10.

11. 12.

13.

14. 15.

16. 17.

18. 19.

20.

21. 22.

23.

24.

25.

Mayer J, Fichtner M, Wolf D, Schubert K. A microstructured reactor for the catalytic partial oxidation of methane to syngas. Proceedings of the 3rd International Conference on Microreaction Technology. Berlin: Springer, 2000:187–196. Schubert K, Bier W, Linder G, Seidel D. Herstellung und Test von kompakten Mikrowärmeübertragern. Chem Ing Tech 1969; 61:172–173. Löwe H, Ehrfeld W, Hessel V, Richter T, Schiewe J. Micromixing technology. Proceedings of the 4th International Conference on Microreaction Technology. AIChE Spring National Meeting, Atlanta, GA, March 2000. Bayer T, Heinichen H, Natelberg T. Emulsification of silicon oil in water—comparison between a micromixer and a conventional stirred tank. Proceedings of the 4th International Conference on Microreaction Technology. Atlanta, GA, AIChE Spring National Meeting, March 2000:167–173. Wochner M. Mikroreaktoren—kleine Ergänzung für grosse Kessel, Clartext No. 3/ 2002. Hagendorf U, Jänicke M, Schüth F, Schubert K, Fichtner M. A Pt/Al2O3 coated microstructured reactor/heat exchanger for the controlled H2 /O2 reaction in the explosion regime. Proceedings of the 2nd International Conference on Microreaction Technology, AIChE Spring Meeting, New Orleans, LA, March 1998, 81–87. Ehrfeld W, Münchmeyer D. Three-dimensional microfabrication using synchrotron radiation. Nucl Inst Meth Phys Res 1991; A303:523–531. Ehrfeld W, Ehrfeld U. Microfabrication for process intensification. In: Matlosz M, Ehrfeld W, Baselt JP, eds. Proceedings of the 5th International Conference on Microreaction Technology. Berlin: Springer-Verlag, 2001:3–12. Koehler M, Ätztechniken. In: Ehrfeld W, ed. Handbuch Mikrotechnik. München: Carl Hanser Verlag, 2001:279–322. Freitag A, Dietrich TR, Scholz R. Glass as a material for microreaction technology. Proceedings of the 4th International Conference on Microreaction Technology. AIChE Spring National Meeting, Atlanta, GA, March 2000:48–54. Rangelow IW, Kassing R. Silicon microreactors made by reactive ion etching. Proceedings of the 1st International Conference on Microreaction Technology. Berlin: Springer, 1998:169–174. Laermer F, Schilp A (Robert Bosch GmbH). Method of Anisotropically Etching Silicon. U.S. Patent No. 5501893, 1996. Weck M. Ultraprecision machining of microcomponents. In: Weck M, ed. Proceedings of the International Seminar on Precision Engineering and Microtechnology, Aachen: European Society for Precision Engineering and Nanotechnology, July 2000. Michel F, Ehrfeld W, Koch O, Gruber H-P. EDM for microfabrication—technology and applications. In: Weck M, ed. Proceedings of the International Seminar on Precision Engineering and Microtechnology, Aachen, July 2000. Bremus E, Gillner A, Hellrung D, Höcker H, Legewie F, Poprawe R, Wehner M, Wild M. Laser processing for manufacturing microfluidic devices. In: Proceedings of the 3rd International Conference on Microreaction Technology. Berlin: Springer, 2000:187–196. Gillner A, Klotzbücher T. Lasermikrobearbeitung. In: Ehrfeld W, ed. Handbuch Mikrotechnik. München: Carl Hanser Verlag, 2001:105–143.


26.

Wiessmeier G, Schubert K, Hönicke D. Monolithic microreactors possessing regular mesopore systems for the successful performance of heterogeneously catalysed reactions. In: Ehrfeld W, ed. Proceedings of the 1st International Conference on Microreaction Technology., Berlin: Springer, 1998:20–26. 27. Jandeleit B, Schaefer DJ, Powers TS, Turner HW, Weinberg WH. Combinatorial materials science and catalysis. Angew Chem Int Ed 1999; 38:2494–2532. 28. Claus P, Hönicke D, Zech T. Miniaturization of screening devices for the combinatorial development of heterogeneous catalysts. Catal Today 2001; 67:319–339.


6 Structured Catalysts and Reactors: A Contribution to Process Intensification Jacob A. Moulijn, Freek Kapteijn, and Andrzej Stankiewicz Delft University of Technology, Delft, The Netherlands

1.

INTRODUCTION

In this book the chemical plant is focused upon. Therefore, the present chapter emphasizes chemical reactors for the chemical process industry. But it should be made clear that structured packings and catalysts also have a large potential in consumer products. Chemical reactors form the heart of a (petro-)chemicals production plant. Given the large variety of plants it is no surprise that a wide variety of chemical reactors are used. Catalytic reactors can be roughly divided into random and structured reactors. It is useful to start with a summary of the major basic concerns (apart from high activity, selectivity, etc.) for catalytic reactors: Catalyst quality on a microscopic length scale (quality, number of active sites) Catalyst quality on a mesoscopic length scale (diffusion length, loading, profiles) Ease of catalyst separation and handling Heat supply and removal Hydrodynamics (regimes, controllability, predictability)


Transport resistance (rate and selectivity) Safety and environmental aspects (runaways, hazardous materials, selectivity) Costs On each of these, random and structured reactors behave quite differently. In terms of costs and catalyst loading, random packed-bed reactors usually are most favorable. So why would one use structured reactors? As will become clear, in many of the concerns listed, structured reactors are to be preferred. Precision in catalytic processes is the basis for process improvement. It does not make sense to develop the best possible catalyst and to use it in an unsatisfactory reactor. Both the catalyst and the reactor should be close to perfect. Random packed beds do not fulfill this requirement. They are not homogeneous, because maldistributions always occur; at the reactor wall these are unavoidable, originating form the looser packing there. These maldistributions lead to nonuniform flow and concentration profiles, and even hot spots can arise (1). A similar analysis holds for slurry reactors. For instance, in a mechanically stirred tank reactor the mixing intensity is highly non-uniform and conditions exist where only a relatively small annulus around the tip of the stirrer is an effective reaction space. Catalytic conversion and separation are conventionally carried out in separate pieces of equipment. A combination of functions in single units is an elegant form of process intensification. When one of the functions is a chemical reaction, it is referred to as a multifunctional reactor. A good example is catalytic distillation technology from the CDTech Company. They have introduced elegant technology for desulfurization of oil (2). Structured reactors will play a key role in the design of novel processes based on multifunctional reactors (3). A monolith is a good example. Monolithic catalysts are shown in Figure 1.

FIGURE 1 Monolithic structures of various shapes. Square-channel cordierite structures (1, 3, 5, 6), internally finned channels (2), washcoated steel monolith (4).


FIGURE 2 Low-pressure-drop reactors used for tail-gas treating.

Depending on one’s point of view, a monolith can be considered a reactor or a catalyst; the borders between catalyst and reactor vanish (4,5). Other structured reactors also deserve attention, but for the “message” it suffices to limit the discussion to monolithic reactors. 2.

OVERVIEW OF STRUCTURED REACTORS

Structured reactors and catalysts are encountered in a large variety (3,6). Structured catalytic reactors can be divided into two categories. The first involves a structured catalyst, whereas the second one involves “normal” catalyst particles arranged in a nonrandom way. In the first category, the catalyst and the reactor are essentially identical entities. Because of their low pressure drop, structured reactors in practice dominate the field for treating tail gases. Figure 2 presents the major types of reactor. The monolithic reactor represents the class of “real” structured catalytic reactors, whereas the parallel-passage reactor and the lateral-flow reactor are based on a structured arrangement of packings with “normal” catalyst particles. Structuring is possible at all length scales. In structured reactors the level is considered above that of a single particle. Structuring can be done based on dedicated structured catalyst shapes in such a way that the catalyst is an integrated part


TABLE 1 Subdivision and Typical Examples of Structured Reactors Structuring on the level of the catalyst and the reactor Monoliths Extruded parallel-channel systems (honeycombs), usually ceramic Emission reduction for cars Ozone decomposition in airplanes Selective catalytic reduction of NOx Arrays of corrugated plates Arrays of fibers Gauzes Ag Methanol → formaldehyde Pt/Rh NO production from ammonia HCN production from methane Foams Catalytic membranes

Structuring exclusively on the level of the reactor Three-levels-of-porosity (TLP) reactors Bead-string reactors Membrane-enclosed catalytic reactors

of the reactor shape. An alternative is to arrange catalyst particles in such a way that a structured reactor is the result. Table 1 gives the subdivision together with typical examples. By inspection of Figure 2, most of them can be recognized. 2.1.

Monolithic Catalysts and Reactors

Monoliths usually are made from ceramics, but metals are also used. They can be produced by extrusion of support material (often cordierite is used, but various types of clays or typical catalyst carrier materials, such as alumina and titania, are also used), a paste containing catalyst particles (e.g., zeolites, V-based catalysts) or a precursor for the final product (e.g., polymers for carbon monoliths). Alternatively, catalysts, supports, or their precursors can be coated onto a monolithic support structure (“washcoating”). Zeolites have been coated by growing them directly on the support during the synthesis (7). The coating literature and patents represent a large field, and, in principle, a variety of preparation procedures are available. All major catalyst support materials, ceramic and polymeric, have been extruded as monolith (4,8). Metallic support structures are used for automotive applications (9). The choice for a certain catalyst type will strongly depend on the balance between maximizing the catalyst inventory and catalyst effectiveness. For slow reactions, a high catalyst loading is desired and the pure


catalyst-type monolith is desired; for fast reactions or if diffusion is slow, a thin coating with a maximum geometric area is preferred. Monoliths are the dominant catalyst structures for three-way catalysts in cars (10–13), selective catalytic reduction catalysts in power stations (14–17) and for ozone destruction in airplanes. What causes this popularity? The catalyst consists of one piece, so no attrition due to moving particles in a vibrating case occurs. The large open frontal area and straight channels result in an extremely low pressure drop, essential for end-of-pipe solutions like exhaust pipes and stack gases. The straight channels prevent the accumulation of dust. In all these applications the reaction system is relatively simple; a single fluid phase (gas) has to be treated at reasonable conditions. More demanding applications of monoliths are now being investigated, fast reactions at high temperatures such as steam reforming, partial oxidation of hydrocarbons to syngas, and oxidative dehydrogenation (18–20). These examples are limited to single-phase applications. As will be discussed later, monoliths for multiphase applications have already proven their value. Monoliths are industrially produced in large quantities by extrusion. This leads to the attractive situation that, although they are sophisticated structures, they are commercially available at reasonable cost. Of course, monolithic catalysts have disadvantages. They share with packed-bed catalysts the requirement of sufficient stability or in any case good regenerability. With respect to mass transfer and heat transfer characteristics, the major limitations are the laminar flow through the channels, no interconnectivity between the channels, and a poor radial heat conductivity. The latter two properties are much better for the foamtype monoliths, but with a trade-off in a higher pressure drop and/or lower catalyst loading (sites/m3). In principle, a laminar flow velocity profile is associated with low mass transfer rates and a wide residence time distribution. Fortunately, for gases, due to the small channel size and high diffusivity, this radial transport in the channels is sufficiently fast. Typical time scales for diffusion are given in Table 2. In liquid phases the diffusivity is three orders of magnitude smaller, which is one of the reasons that monoliths do not enjoy a high popularity in liquid-phase operations. It will be shown that this is based on a misconception. TABLE 2 Diffusion Time Scales in Catalytic Reactors (lD2/2D)

Gas Liquid Liquid in cat pore Liquid in zeolite pore

D (m2/s)

lD for 1 mm

lD for 0.1 mm

lD for 1 m

105 109 1010 1011

50 ms 500 s 5000 s 50,000 s

0.5 ms 5s 50 s 500 s

50 ns 500 ns 5 ms 50 ms


The honeycomb-type monoliths are characterized by a very high geometric surface area. Dependent on the cell density, this can exceed 3000 m2/m3reactor! Figure 3 shows the values for three different cell densities, i.e., for 200, 400, and 600 cpsi (cells per square inch). These examples are quite realistic. At present the normal monolith for cars is a 400-cpsi monolith. The values for the geometrical surface amount to 3440 m2/m3reactor. In packed beds this value is much lower in order to avoid unrealistic pressure drops. It is to be expected that future monoliths will exhibit even larger geometric surface areas. That alone makes them highly useful for process intensification programs. Metal monoliths can be shaped rather freely. A good example is given in Figure 4 (9), where it can be seen that in these parallel-channel systems the structure of the channels is such that the turbulence increases. The reasoning behind that is the wish to counteract the low mass transfer rates associated with laminar flow in the thin channels of the monolith. 2.2. Gauzes The appearance of gauzes is illustrated by Figures 5 and 6 (9). The use of noble metal gauzes goes back to the beginning of the 20th century for the oxidation of ammonia into NO. This work followed up work of Ostwald, who applied platinized asbestos and later a roll of corrugated strip of Pt. Probably, this was the first application of a structured reactor.

FIGURE 3 Geometric surface areas for three different cell densities.


FIGURE 4 Shaped channels in metal monoliths in order to increase mass transfer in gas-phase applications.

2.3. Structured Packings It is often suggested to coat structured packings of the type given in Figure 7 with catalysts, in the same way as monoliths. However, in normal applications this does not lead to a satisfactory reactor. The geometric surface area is orders of magnitude smaller than packed beds and monoliths. Of course, this problem can be solved by packing the channels with catalyst particles. Also, they can be applied as mixing device to be used for a good inlet distribution in the case of


FIGURE 5 Pt/Rh gauze for the production of NO from ammonia.

multiphase reactors. Later it will be seen that combinations of static mixers and monolithic catalysts have a high potential in process intensification. 2.4.

Foams

Foams are to some extent the negative images of packed beds. They can be used when turbulent mixing is important. Figure 8 gives an example of a foam that is used as a carrier for a molten salt catalyst in diesel soot trapping and combustion. The openness and the mixing characteristics of foams have stimulated research in the potential application in soot trapping. An advantage is the robustness of the


FIGURE 6 Details of a Pt/Rh gauze before (A) and after (B) use in the oxidation of ammonia into NO.


FIGURE 7 Structured packing (Sulzer Katapak-K).

system: Plugging does not occur. A disadvantage is the relatively low trapping efficiency. In combustion foams result in stable combustion behavior. 2.5.

Arranged Catalysts—Three-Levels-of-Porosity (TLP) Reactors

Three-levels-of-porosity (TLP) reactors (21–23) are alternatives for monolith reactors in certain applications. Conventional catalyst particles can be arranged in any geometric configuration. In such arrays, three levels of porosity can be distinguished: the pore space within the particles, the intraparticle space, and the space between the arrays. An example of such a TLP reactor is the parallelpassage reactor (PPR); see Figure 9 (24). The catalyst particles are confined between wire gauze screens that divide the reactor into a large number of catalyst layers with empty passages in between. The gas flows along the catalysts layers instead of through the bed as in a traditional fixed-bed reactor. Because the gas flows through straight channels


(ca. 10 mm wide), the pressure drop over the PPR is much lower than over a fixed-bed reactor. Reactants are transferred from the gas to the catalyst inside the gauzes, mainly by diffusion. The PPR is very suitable for treating dust-containing gases, e.g., flue gases from power plants, because dust will not be collected on the catalyst particles as a result of the straightness of the gas passages. Bead-string reactors represent the limit of parallel-passage reactors: They contain single-catalyst-particle subunits. Figure 10 gives a schematic representation (25). Bead-string reactors have the advantages of lateral-flow reactors but not the disadvantage of the low mass transfer rates in the units of the lateral-flow reactors.

FIGURE 8 Foam as a catalyst support (alumina impregnated with Cs2SO4.V2O5).


FIGURE 9 Example of the TLP reactor: parallel-passage reactor (PPR). (Adapted from Ref. 24.)

The disadvantage, of course, is the high cost, due to the difficult, labor-intensive production. In a recent patent of ABB, an arrangement is claimed that combines the advantages of the bead-string reactor with an easily produced arranged catalyst configuration: Monolith channels are packed with catalyst particles, resulting in strings of particles. This was described as a structured packed bed (26). The reactor internals, consisting of structured packings packed with catalyst particles, are also examples of arranged catalysts, e.g., the Sulzers Katapak-S type. Multifunctional reactors often are also structured reactors. A good example is the membrane reactor (27,28). Two types can be distinguished, those based on a catalytic membrane and those in which the membrane only provides a selective separation function without being catalytically active itself. The former is an example of a structured catalyst, while the latter belongs to the category of arranged catalysts. The reactor containing a nonactive membrane is referred to as a membrane-enclosed catalytic reactor (MECR). In the following, an example of a MECR is described. 2.6.

Membrane-Enclosed Catalytic Reactor (MECR)

Catalytic membrane reactors are not yet commercial. In fact, this is not surprising. When catalysis is coupled with separation in one vessel, compared to separate pieces of equipment, degrees of freedom are lost. The MECR is in that respect more promising for the short term. Examples are the dehydrogenation of alkanes in order to shift the equilibrium and the methane steam reforming for hydrogen production (29,30). An enzyme-based example is the hydrolysis of fats described in the following.


2.6.1. Production of Fatty Acids by Fat Hydrolysis in a Membrane Reactor A recent development at laboratory scale is the application of an enzyme (lipase) to catalyze the hydrolysis: Water and fat are mixed at low temperature (300 K) in a continuous stirred-tank reactor (CSTR). The water phase contains the enzyme. A much purer glycerol solution is obtained than in the conventional process. The disadvantage is that the equilibrium is not favorable. An elegant solution has been proposed based on a membrane reactor consisting of a module with hollow cellulose fibers [see Fig. 11 (31)]. The enzyme is placed at the inner side of the fibers, to which the fat is fed. Water passes at the

FIGURE 10 Schematic (a) and reactor configuration (b) of the bead-string reactor.


FIGURE 11 Membrane reactor for the production of fatty acids. (Adapted from Ref. 31.)

outside and diffuses through the membrane to react at the fat/lipase interface. The fatty acid formed stays in the oil phase, whereas the glycerol formed is transported through the membrane into the water phase. Laboratory studies show nearly complete conversions. 3.

GAS-PHASE REACTIONS

It is fair to state that by and large the most important application of structured reactors is in environmental catalysis. The major applications are in automotive emission reduction. For diesel exhaust gases a complication is that it is overall oxidizing and contains soot. The three-way catalyst does not work under the conditions of the diesel exhaust gas. The cleaning of exhaust gas from stationary sources is also done in structured catalytic reactors. Important areas are reduction of NOx from power plants and the oxidation of volatile organic compounds (VOCs). Structured reactors also suggest themselves in synthesis gas production, for instance, in catalytic partial oxidation (CPO) of methane. 3.1.

Environmental Catalysis

Converters for cars are usually ceramic monoliths and occasionally metal based. Without much exaggeration, they can be claimed to be one of the major successes of recent decades in the area of chemical engineering and catalysis. In the beginning, the catalytic converter was placed underbody, where sufficient space was available and where the temperature was expected to be mild. There was no need


for process intensification! Later, performance became critical, mainly because of the observation that under steady-state conditions the converter works well but cold start emission is relatively large. Options that are offered include a catalyst close to the engine, electrically heated catalysts, and combinations of catalyst and adsorbent. Initially, packed beds were also used. They, however, were no success, and at present monoliths are applied exclusively. This should not be misunderstood. Monolith means literally “a single stone.” However, metal-based analogues are also included in the definition of monolith. In fact, for catalytic converters in cars, in addition to ceramics, metal-based monoliths have been and still are used. A major advantage of metal was the thin wall thickness that could be achieved. Later, industry succeeded in manufacturing ceramic structures of comparable wall thickness. In view of their higher resistance against corrosion, ceramic monoliths are now more generally applied than metal ones. Structured catalysts are also essential in diesel exhaust gas purification. State-of-the-art solutions are marketed by PSA and by Johnson Matthey. The truck market is dominated by diesel engines. In that application, space requirement is a major issue, and intensification is badly needed. Space velocities exceeding 100,000 h1 are demanded. Reactive structured filters are the way to go. In the wake of the spectacular application of monoliths in the treatment of automobile exhaust gas, the potential of monoliths in other applications was studied. Gas-phase reactions were the major area. Catalytic oxidation has received a lot of attention. Low-NOx burners based on monoliths were designed, catalytic oxidation of VOCs also benefits from structured catalysts, basically because of the low pressure drop and the resistance against dust. Originally, packed-bed reactors were applied in selective catalytic reduction (SCR). They could be used only in low-dust applications (15). They were successfully replaced by several types of structured catalysts, viz., honeycombs, plate-type catalysts, and parallel-flow systems. Also, this technology is without doubt successful. Volatile organic compounds are destroyed by combustion in structured catalysts usually containing Pt or Pd. Compared to automotive applications, the size of the reactors is large. Figure 12 explains the engineering for the destruction of VOCs of the large gas flows in industry (9). 3.2.

Production of Syngas

In the production of syngas, the following reactions are usually undesired. The desired reaction is the production of CO/H2 mixtures according to 2CH4 O2 → 2CO 4H2 whereas sequential oxidation giving CO2 and H2O is not desired. This calls for short residence times (ms), short diffusion length (as small as possible a diameter


FIGURE 12 Industrial unit for SRC containing metal monolith units.


of the catalyst particle), and the absence of nonuniformity with respect to temperature and concentrations. Moreover, the pressure drop in this type of application should be minimal. It is obvious that a structured reactor should be used. This leaves three candidates: Monolith Gauze Foam Foams have the highest turbulence and the highest pressure drop. Probably because of the first phenomenon they are the most suited, although gauzes also might be good. Monoliths have the advantage of being well defined, but the absence of radial heat transport will lead to scale-up problems: When a catalyst in a channel would “die,” the temperature will drop, and so will the viscosity, leading to a “leak” in the reactor. It is clear that radial heat transfer is a key issue (as it is in packed beds). Advanced designs have been described in the literature. By adapting the geometry, turbulence can be enhanced. Figure 4 illustrates this. Gauzes are the state of the art for many millisecond-reactions performed in industry. The best-known examples are the oxidation of ammonia to NO for the production of nitric acid and the Andrussov process, in which HCN is produced from methane and ammonia (32): NH3 CH4 1.5O2 → HCN 3H2O The temperature in this process is quite high, 1100–1200C. It is not surprising that under these severe conditions extended reorganization of the alloys takes place. This is shown in Figure 6 for the oxidation of ammonia into NO. Many more options are imaginable. A good example is the crosscurrent monolith (Figure 13). In theory such a system allows ideal heat exchange between adjacent channels. Such an elaborate structure might look improbable. However, Corning recently filed a patent claiming the direct extrusion of crosscurrent structures. So these advanced reactor types might be applied in practice in the future. Naphtha cracking, as a large-scale endothermic reaction, might be a good case for such a reactor. 3.3.

Scale-Up

Scale-up of structured reactors is usually easier than for packed-bed reactors. The major point is that the hydrodynamics are independent of the scale of the reactor (assuming a good inlet device). When the radial temperature profile is also independent of the scale, scale-up is straightforward. This is the case for millisecond reactors. In these reactors, rates are very high; as a consequence, in exothermic reactions they operate adiabatically. So they scale easily.


FIGURE 13 Crosscurrent monolithic structure.

4.

MULTIPHASE REACTIONS

Various types of reactors are being used commercially for multiphase applications, the major ones being the slurry reactor, the bubble-column reactor, and the trickle-bed reactor (5). Figure 14 gives a schematic of these three types of reactor. Each reactor has its own advantages and disadvantages. Slurry catalysts are small (typically 50 m), while trickle-bed particles are larger (millimeter scale), in view of the allowable pressure drop over the bed. The particle size is a crucial parameter. In general it can be stated that larger particles are less efficient and, even more important, are less selective in those reactions where the desired product is subject to the following undesired reaction (A → B → C, with B as the


desired product). In that often-encountered case, the slurry reactor is more selective than the trickle bed reactor. In terms of process intensification, a mechanically stirred-tank reactor often is not a good choice. In practice it is no exception that gas–liquid mass transfer is rate determining. This implies that only the part of the space close to the tip of the stirrer(s) is well used. A large part of the reactor does not contribute much to the productivity and, depending on the kinetics, will lead to low selectivity. Moreover, the major disadvantages of the slurry reactor are the separation of product and catalyst and catalyst attrition. The trickle-bed reactor is much more convenient, but large particle sizes are unavoidable. An important limitation of trickle-bed reactors is that, in practice, they are nearly always operated cocurrently, to avoid liquid entrainment by the gas (“flooding”). Some important commercial applications, however, would benefit from a countercurrent operation, especially for equilibrium-limited reactions and in the case of strong product inhibition (33). Examples are hydrotreating processes like hydrodesulfurization (HDS), hydrodenitrogenation (HDN), and hydrocracking. Only for large particles or low flow rates could this operational mode be achieved in a packed bed [Synsat process (34)]. Deep desulfiding is a good example of a reaction where the concentration profile in countercurrent operation is more optimal from a reaction kinetics point of view (2,35). Also, more active catalysts (e.g., noble metals) can be used in the last part of the reactor (“catalyst profiling”) that are more susceptible to H2S poisoning and, as a consequence, are not suitable for cocurrent operation. Overall, countercurrent operation leads to deeper desulfurization with smaller catalyst units or to larger throughputs (21). In all

FIGURE 14 Schematic representations of three basic gas–liquid–solid reactor systems.


three reactor types, in principle a runaway is possible because when hot spots would be formed, large amounts of reactants can reach the hot spot, leading to a classical runaway. It will be shown that structured reactors based on monoliths are imaginable that do not possess the unfavorable properties mentioned. In current practice, one application is known. AKZO-NOBEL (previously EKA-NOBEL) operates five plants based on the anthraquinone process, in which the reduction step is carried out in monolith reactors (36,37). Many multiphase reactions have been carried out at laboratory scale, and in industry interest is also increasing, as is apparent from patents (26,38–46). Recently, monolith structures have been tried in photocatalysis. This might well be an important application in the future. 4.1. 4.1.1.

Hydrodynamics and Mass Transfer in Monoliths Cocurrent Operation

For cocurrent gas–liquid flow, several flow regimes can occur. The preferred one is usually the so-called Taylor, or slug, flow (47– 49). This type consists of gas bubbles and liquid slugs flowing consecutively through the small monolith channels. The gas bubble fills up the whole space of the channel, and only a thin liquid film separates the gas from the catalyst (Figure 15). For two reasons, the rate of mass transfer is large. First, the liquid layer between bubble and catalyst coating is thin, increasing mass transfer. Second, the liquid slugs show an internal recirculation during their travel through a channel. Because of this, radial transfer of mass is increased. Moreover, the gas bubbles push the liquid slugs forward as a piston, and a type of plug flow is created. Compare this with single-phase liquid flow through the channels. Because of the low channel diameter, the flow will be laminar and, as a consequence, the radial transport will be extremely slow, leading to very poor reactor performance: Rates are slow and the reactor exhibits strong nonplug-flow behavior. For multiphase operation under slug-flow conditions, the mass transfer increase is an order of magnitude larger than for singlephase liquid flow, whereas the increase in friction—that is, pressure drop—is much less [Figure 16 (50)]. A fortunate finding is that Taylor flow conditions are easily realised under practical conditions. Ideally, in contrast to packed beds, scale-up of monolithic reactors is very simple. When we know the behavior of one channel, we should be able to predict the whole reactor. Is this really true? Compared to a packed bed, a monolithic reactor differs in radial transport. When the initial distribution of liquid in the radial direction is nonideal, going down through the reactor, this unfavorable distribution does not change. In a packed-bed reactor this happens to a certain degree. Therefore in scale-up, the reactor inlet system has to be designed well so that the distribution of the liquid in the top of the reactor is ideal. We found that if a bubble emulsion on top of the monolith is present, a satisfactory distribution seems to be guaranteed, as found for trickle-bed reactor operation. We carried out a large experimental program and defined the conditions where this happens to be the case. It appeared that


FIGURE 15 Taylor flow through a single tube. Left: picture of air–water flow; middle: schematic representation of the gas and liquid slugs; right: CFD velocity pattern in a liquid slug showing the liquid recirculation.

the flow rate has to be above a specified minimal value. Stacking of monolith pieces on top of each other or with some spacing in between, to allow some radial mixing, does not seem to have a negative impact on the flow characteristics. So the flow rates have to be sufficiently high (linear velocities > 0.1 m/s) in order to guarantee a good distribution of liquid over the cross section of the reactor. One might wonder if upflow of gas and liquid is not to be preferred, because lower flow rates might be applied. This appeared not to be the case. Again, high flow rates are needed to establish a good gas–liquid flow distribution. It might be worthwhile to investigate whether systems can be developed or conditions established that allow low flow rates. Combinations of monolithic catalyst packages with the Sulzer type of contactors are being conceptually investigated in our group. They might increase the window of operability toward lower flow rates.


FIGURE 16 Relative increase of friction and mass transfer due to gas–liquid Taylor flow, compared to developed laminar flow in small tubes. represents the dimensionless length of a liquid slug, Re the Reynolds number based on the liquid.

Moreover, they might lead to flexibility, allowing more compact reactor systems. The first results are promising. 4.1.2.

Mass Transfer

Mass transfer was studied experimentally in various ways. Nonreactive studies involved the uptake or release of oxygen by the liquid for the measurement of gas–liquid transfer (51–53), while in reactive studies the overall gas–solid or liquid–solid transfer could be determined. As an example of the performance, a monolith in the hydrogenation of -methylstyrene was compared with a tricklebed reactor under identical reaction conditions in cocurrent mode. Per unit reactor volume, the washcoated 400-cpsi monolith yielded a hydrogenation rate more than four times higher. For a reaction that is mass transfer controlled, this stresses the better mass transfer in the monolith. Overall, the Ni was used 40 times more efficiently in the monolith than in the trickle-bed reactor, even in spite of the use of an eggshell catalyst in both cases. In spite of the high rates observed, it was felt that not all the Ni in the washcoat layer was optimally used (5,54). In subsequent work a more eggshell type of coating was realized and the rates observed were an order of magnitude higher. Mass transfer is usually expressed as the factor kla, the mass transfer coefficient times the exposed surface area per unit volume a. Values of kla depend strongly on the gas and liquid properties, but for many systems values of 0.5 s1 are found, and values even much larger than 1 s1 possibly apply.


TABLE 3 Comparison of Gas-to-Liquid Mass Transfer in Three Common Three-Phase Reactor Types Reactor type

kla (s1)

Trickle bed Slurry Monolith

0.05–0.2 0.1–0.3 1

This is one order of magnitude higher than in conventional reactor types (1), which underlines the process intensification potential of monolithic reactors. In Table 3 the three common reactor types are compared. Obviously, the monolithic reactor in the Taylor-flow regime leads to a high degree of process intensification. When these numbers are recalculated into production rates, values of 40 mol/m3reactor-s were found. Figure 17 illustrates the high value in relation to the “Weisz window of reality.” This demonstrates the attractiveness of using monoliths in fast catalyzed gas–liquid–solid reactions. 4.1.3. Countercurrent Operation in Monoliths and Arranged Packings Under practical conditions, countercurrent operation in a packed bed reactor is not feasible, because flooding occurs (55,56). The reason is that in the small interstitial space, extended momentum transfer takes place between the liquid flowing down and the gas flowing upward. At velocities used in industry this would imply

FIGURE 17 High productivity of multiphase monolith reactors.


that the particle size has to be increased by an order of magnitude. This leads to unacceptable internal diffusion limitations. Clearly, momentum transfer has to be decreased while maintaining high rates. This can be done by structuring the catalyst or by clever arranging of the catalyst particles in the reactor. Various arranged catalyst structures are used or can be envisaged. Figure 18 gives an overview of the most important ones. The principle of these structures is that relatively large channels are present, leaving space for countercurrent flow without extended momentum transfer. In catalytic distillation a lot of experience has already been gained in packings based on particles arranged in bales (2). From an extensive study, it appeared that in structured reactors as well, countercurrent operation is possible at industrially relevant conditions. The breakthrough was the design of optimal monolithic structures and dedicated inlet and outlet systems. For example, good results were obtained by cutting the monolith under an angle of 70 as the optimal value (57) or by a special outlet construction, guiding the liquid away from the exit (Figure 19). Finned tubes exhibit outlet flooding mainly only, whereas the unfinned tube also exhibits inlet flooding. The unfinned tube has a larger hydraulic diameter due to the absence of the fins and hence a wider flooding-free region. Injecting the gas via a capillary and guiding away the liquid through quartz wool plugs even enlarges this region for the finned channel. This graph illustrates that current operating region for trickle bed reactors (HDS) is well covered by the finned monoliths (58).

FIGURE 18 Various arranged particle packings.


FIGURE 19 Flow map for countercurrent gas–liquid flow (n-decane/air) through finned channels with different outlet geometries. Indicated are the flooding limits for single tubes. Finned tubes exhibit mainly only outlet flooding, whereas the unfinned tube also exhibits inlet flooding. The unfinned tube has a larger hydraulic diameter due to the absence of the fins and hence a wider flooding-free region. Injecting the gas via a capillary and guiding away the liquid through quartz wool plugs even enlarges this region for the finned channel. This graph illustrates that current operating region for trickle-bed reactors (HDS) is well covered by the finned monoliths (From Ref. 58.)

4.1.4.

Monolith Reactors

The catalyst to be used in a reactor operation can be coated as a thin layer on the channel walls, and, hence, the reactor can be described as a “frozen slurry reactor.” The diffusion length is small and well controllable. The catalyst loading often is relatively small, but using thicker coatings or using a monolith extruded from the catalyst support, e.g., an all-alumina monolith, can increase it. The high cell density of the monoliths creates a high geometric surface area. Using a packed bed, unrealistically small particles would be needed to achieve this. Catalyst separation and handling are as convenient as in a common packed bed. Scale-up is in principle straightforward. Larger channel geometries (e.g., in the internally finned monolith channels) allow countercurrent operation of gas and liquid. Monolith reactors are intrinsically safer. The monolith channels have no radial communication in terms of mass transport, and the development of runaway by local hot spots in a trickle-bed reactor cannot occur. Moreover, when the feed of liquid or gas is stopped, the channels are quickly emptied. From the foregoing it should be evident that monolithic reactors (and other structured reactors) in many respects are superior to classical reactors. Indeed, for


several reactions, monolithic catalysts have been reported, although, except for one case, only at the bench or pilot scale. The interesting points are to demonstrate that the “theoretically” outlined advantages are indeed present. Compared to classical reactors they in fact boil down to a larger reactor productivity, a better selectivity control, and a higher efficiency. The first point also implies a better catalyst utilization. Obvious is the fact the catalyst is fixed in a reactor and pressure drops are low. Highly exothermal reactions can be applied by external heat exchange (1,39). If a CSTR-type reactor is not desired, the horizontal reactor with interstage cooling is an attractive alternative. The hydrogenation step in the anthraquinone process of AKZO-Nobel is an industrial realization of a monolithic reactor and includes a lot of pioneering work from the Anderson group (59–63). More examples of the use of monoliths can be found in Refs. 5 and 64. In our own group, in cooperation with a chemical industry, we have studied the selective hydrogenation of pyrolysis gasoline, a by-product of the naphtha cracking that can be upgraded to gasoline by selectively removing gum-forming dienes and styrene-like molecules, leaving intact the internal alkenes. This study (65) demonstrated the plug-flow behavior needed for such a selective conversion and the efficient use of the active phase, which was at least a factor of 3– 4 better than in a trickle-bed operation. The hydrogenation of -methylstyrene, mentioned earlier, is an even more appealing example of better active-phase utilization and confirms the good mass transfer properties. An attractive property of monolithic reactors is their flexibility of application in multiphase reactions. These can be classified according to operation in (semi)batch or continuous mode and as plug-flow or stirred-tank reactor or, according to the contacting mode, as co-, counter-, and crosscurrent. In view of the relatively high flow rates and fast responses in the monolith, transient operations also are among the possibilities. The cocurrent monolith reactor, with its plug-flow characteristics, can in principle be used in downflow, upflow, and horizontal-flow modes, provided a good gas–liquid distribution is secured (66). The last mode might solve a major problem in practical applications of monoliths: Because, for hydrodynamic reasons, high flow rates are needed, the reactor length tends to be very large. The process intensification potential of horizontal configuration, the so-called in-line monolithic reactor (ILMR), has recently been demonstrated by Stankiewicz (67) for one of the large-scale hydrogenation processes of DSM. It has been shown that the conventional reactor system, consisting of a stirred-tank reactor and a packed-bed reactor, could be replaced by an ILMR ca. 30–100 times smaller (depending on the type and thickness of the washcoat) (Figure 20). Research with respect to this type of reactor is in progress. An important outcome of the research might be that coupling of monolithic elements, mixing units and heat exchangers,


FIGURE 20 Process intensification in the in-line monolithic reactor, ILMR.


leads to flexible cascade reactor setups, enabling multistep synthesis in one pass. Further extension of the in-line monolithic reactor concept on other unit operations could possibly lead in the future to much more compact, safer, and environmentally friendly chemical plants, in which pipelines would not only serve for sending gases and liquids, but be made functional and used for reactions or separations (68). The best studied mode is cocurrent downflow. It can be envisaged in two ways, with either a controlled flow of gas or a free recirculation due to entrainment by the liquid at the entrance of the monolith (Figure 21). This reactor is an alternative to the bubble-column reactor often used in biotechnological applications. Since high reactor types are being used and large gas-flow rates are required, the energy input to introduce and compress the gas for injection at the bottom is relatively high. In the downflow monolith reactor, this gas injection is automatically achieved. The cocurrent reactor type can easily be used as a stirred reactor type by a large recirculation flow without extremely high energy input due to the low pressure drop. An external heat exchanger can be scaled independent of the reactor to deliver the required heat duty (1,5). Of course, monoliths have disadvantages. They are at this moment more expensive than particle catalysts. In fixed-bed operation, they will have to exhibit

FIGURE 21 Configuration of a cocurrent downflow monolith reactor with free gas recirculation. Only liquid is recirculated, and an external heat exchanger can be scaled independent of the reactor to deliver the required heat duty.


FIGURE 22 Cocurrent/countercurrent Syn Technology process scheme.

a sufficiently long lifetime. In quickly (irreversibly) deactivating reactions, they will not be used. Of extreme importance is that the inlet distribution should be secured. In cocurrent flow, both gas and liquid have to be in contact evenly with the catalyst at the monolith walls. Countercurrent operation is appealing in many respects and is already executed in practice. An example is the desulfurization process of Syn Tech, where co- and countercurrent operation are combined (see Figure 22). Apart from the GLS-type, LLS- and GLLS-type catalytic reactions are also possible using monoliths. The attractive property is to bring reactants efficiently in contact with the solid catalyst. But there is more. They can be applied in stripping, extraction, evaporation, drying, and distillation, in coas well as in countercurrent modes. Monoliths are then used as low-pressure-drop and low-energy-consuming contacting devices. The combination with catalysis is then obvious to arrive at a multifunctional reactor system in which reaction and controlled reactant addition or product removal is achieved. These applications are not restricted to gaseous or liquid phases, but also work in solid phases. The straight channels are ideal for fixed- or moving-bed applications (Figure 23), the former to combine an optimized catalyst inventory and liquid holdup while still having a relatively low flow resistance of a single pellet-string reactor. Moreover, existing catalysts can be applied. The use of finned channels gives even more freedom. This could be


FIGURE 23 Monolith structure packed with catalyst particles (“catalyst bale”) for structured fixed-bed or moving-bed applications.

considered a “structured trickle-bed reactor,” where longer residence times can be achieved than in a straight-channel monolith. Blocks of monoliths filled with particles may find application in catalytic distillation or three-levels-of-porosity reactors (26), replacing the catalyst “bales” (5). Channels filled with a single particle string have much better solid flow characteristics than a packed bed, so application of monoliths as the moving-bed reactor internal is seducing. This opens a wide range of applications, covering moving-bed adsorption processes, moving-bed applications for deactivating catalysts (reforming, hydrodemetallization, dehydrogenation), solid trickle-flow reactors, and regenerative processes where a moving catalyst is alternatingly subjected to different atmospheres and transport reaction intermediates and/or heat (FCC, butane to maleic anhydride oxidation). The channel structure also works as a flow straightener, providing better plug-flow characteristics in large-diameter entrainedflow reactors, which suffer from back-mixing of catalyst at the reactor wall. Evaluating the properties of catalytic reactors, there are three important aspects that strongly determine the overall performance: the amount of catalyst and intrinsic kinetics, the transport phenomena (diffusion inside and outside the catalyst), and the hydrodynamics in the reactor. In classical reactors these are strongly interrelated and cannot be defined and designed independently. As an example, for fast


reactions small catalyst particles are desired from the point of view of catalyst effectiveness, but a packed bed with small particles will result in an unacceptable pressure drop. Therefore an optimum has to be sought for the particle size. The elegance of structured reactors is that these three aspects can be designed and optimized fairly independently, resulting in an optimized reactor performance. Figure 24 shows the situation for a monolith channel in a gas–liquid reaction. The zeolite catalyst should be very small to take advantage of its high activity. It is embedded in a washcoat layer on the wall of the monolith channel of a thickness that yields the required catalyst effectiveness and selectivity. The channel diameter determines the type of flow, in this case Taylor flow, which optimizes the mass transfer from the gas and liquid phases to the solid catalyst. The straight monolith channels already ensure a low pressure drop across the structure. This is a structured system covering about 10 orders of magnitude, from nanometers to several meters. If the aim of the catalytic process is to optimize yield and selectivity, one can distinguish two extremes: fast reactions and slow reactions (Figure 25). In slow reactions, the intrinsic reaction kinetics control the process, so the catalyst inventory should be as high as possible. Increasing the wall thickness of a monolith can have the desired effect. In fact the degree of variation in this way is virtually from 10–90 volume %, whereas a packed bed will always yield an inventory of around 60% or lower if hollow catalyst particles are used. In fast reactions, mass transfer or intraparticle diffusion becomes controlling. Thinner catalyst coatings, Taylor flow, etc. can be applied to optimize these

FIGURE 24 Schematic representation of the operation of a monolith channel, washcoated with a zeolite catalyst, under Taylor-flow conditions.


FIGURE 25 Aspects controlling the performance of a three-phase catalytic reactor, indicating the flexibility of the use of monolithic catalysts.

requirements. If mass transfer is controlling, the productivity is proportional to the geometric surface area of the monolithic structure. Increasing cell densities are recommended, without yielding unacceptable pressure drops. These examples exemplify the potential power of the application of monolithic structures in catalytic reactors. 5.

CONCLUSIONS

Monolithic and other structured catalysts exhibit favorable properties with respect to practical convenience, high rates, high selectivity, and low energy consumption. From an engineering point of view, the easy scale-up and the potential of high safety are also appealing. This is not limited to single-phase processes, but they are also well placed for multiphase processing. Monoliths exhibit a large flexibility in operation. They are well suited for optimal semibatch, batch, continuous, and transient processing. Catalytic conversion can be combined with in situ separation, catalytic reactions can be combined, heat integration is possible, and all lead to process intensification. In the short term, catalytic monoliths will be applied to replace trickle-bed reactor and slurry-phase


operations in view of the better overall conversion and selectivity performance. Monoliths allow efficient use of small catalyst particles, e.g., zeolites, and have a substantial flexibility with respect to catalyst inventory in a reactor. Multifunctional reactor operations like reactive stripping and distillation are challenging applications that are not too far away. Several options exist for applications in the oil refinery and the chemical process industry. The essence of the use of a structured reactor is that it allows the decoupling of intrinsic reaction kinetics, transport phenomena, and hydrodynamics. In this way those phenomena that control the overall behavior of a catalytic reactor can be optimized independently, giving rise to excellent reactor performance. REFERENCES 1. Heiszwolf JJ, Engelvaart LB, Gvd Eijnden M, Kreutzer MT, Kapteijn F, Moulijn JA. Hydrodynamic aspects of the monolithic loop reactor. Chem Eng Sci 2001; 56: 805–812. 2. Dautzenberg FM. Novel reactor concepts in hydrotreating. Cattech 1999; 3:54–63. 3. In: Cybulski A, Moulijn JA, eds. Structured catalysts and reactors. New York: Marcel Dekker, 1998:670. 4. Cybulski A, Moulijn JA. Monoliths in heterogeneous catalysis. Catal Rev Sci Eng 1994; 36(2):179–270. 5. Kapteijn F, Heiszwolf JJ, Nijhuis TA, Moulijn JA. Monoliths in multiphase catalytic processes—aspects and prospects. Cattech 1999; 3:24–41. 6. Cybulski A, Moulijn JA. Monoliths in heterogeneous catalysis. Catal Rev Sci Eng 1994; 36:179–270. 7. Jansen JC, Koegler JH, Bekkum Hv, Calis HP, Bleek CM, Kapteijn F, Moulijn JA, Geus ER, Puil Nvd. Zeolitic coatings and their potential use in catalysis. Microporous Mesoporous Materials 1998; 21:213–226. 8. Gulati ST. Ceramic catalyst supports for gasoline fuel. In: Cybulski, A Moulijn JA, eds. Structured Catalysts and Reactors. New York: Marcel Dekker, 1998:15–58. 9. Twigg MV, Webster DE. Metal and coated-metal catalysts. In: Cybulski A, Moulijn JA, eds. Structured Catalysts and Reactors. New York: Marcel Dekker, 1998:59–90. 10. Twigg MV, Wilkins AJJ. Autocatalysts—past, present and future. In: Cybulski A, Moulijn JA, eds. Structured Catalysts and Reactors. New York: Marcel Dekker, 1998: 91–120. 11. Heck RM, Farrauto RJ. The automobile catalyst. Cattech 1997; 1:117–124. 12. Marin GB, Hoebink JHBJ. Kinetic modeling of automotive exhaust catalysis. Cattech 1997; 2:137–148. 13. Misono M. Catalytic reduction of nitrogen oxides by bifunctional catalysts. Cattech 1998; 2:53–69. 14. Beretta A, Orsenigo C, Tronconi E, Forzatti P, Berti F. Analysis of plate-type monolith SCR-DeNO(x) catalysts. Kinet Catal 1998; 39:646–648. 15. Beretta A, Tronconi E, Groppi G, Forzatti P. Monolithic catalysts for the selective reduction of NOx with NH3 from stationary sources. In: Cybulski A, Moulijn JA, eds. Structured Catalysts and Reactors. New York: Marcel Dekker, 1998:121–148.


16. 17. 18. 19. 20. 21.

22. 23. 24. 25.

26. 27.

28. 29. 30. 31. 32. 33.

34. 35. 36.

Tronconni E, Forzatti P. Adequacy of lumped parameter models for SCR reactor with monolith structure. AIChE J 1992; 38:201–210. Topsøe N-Y. Catalysis for NOx abatement. Cattech 1997; 1:125–136. Iordanoglou DI, Bodke AS, Schmidt LD. Oxygenates and olefins from alkanes in a single-gauze reactor at short contact times. J Catal 1999; 187:400–409. O’Connor RP, Schmidt LD. Catalytic partial oxidation of cyclohexane in a singlegauze reactor. J Catal 2000; 191:245–256. Beretta A, Ranzi E, Forzatti P. Production of olefins via oxidative dehydrogenation of light paraffins at short contact times. Catal Today 2001; 64:103–111. Hasselt BWv, Lebens PJM, Calis HP, Kapteijn F, Sie ST, Moulijn JA, Bleek CMvd. A numerical comparison of alternative three-phase reactors with a conventional trickle-bed reactor. The advantages of countercurrent flow for hydrodesulfurization. Chem Eng Sci 1999; 54:4791–4799. Hasselt BWv, Calis HP, Sie ST, Bleek CMvd. Pressure drop characteristics of the three-levels-of-porosity reactor. Chem Eng Sci 1999; 54:3701–3708. Hasselt BWv, Calis HP, Sie ST, Bleek CMvd. Liquid holdup in the three-levels-ofporosity reactor. Chem Eng Sci 1999; 54:1405–1411. Sie ST, Calis HP. Parallel passage and lateral-flow reactors. In: Cybulski A, Moulijn JA, eds. Structured Catalysts and Reactors. New York: Marcel Dekker, 1998:323–354. Calis HP, Takács K, Gerritsen AW, Bleek CMvd. Bead-string reactor. In: Cybulski A, Moulijn JA, eds. Structured Catalysts and Reactors. New York: Marcel Dekker, 1998: 355–392. Strangio VA, Dautzenberg FM, Calis HP, Gupta A. Fixed-Bed Catalytic Reactor. Patent WO 99/48604, World, 1999. Falconer JL, Noble RD, Sperry DP. Catalytic membrane reactors. In: Noble RD, Stern SA, eds. Membrane Separations Technology. Principles and Applications. Amsterdam: Elsevier, 1995:669–712. Hsieh HP. Inorganic Membranes for Separation and Reaction. Amsterdam: Elsevier, 1996:1–591. Thomas S, Schafer R, Caro J, Seidel-Morgenstern A. Investigation of mass transfer through inorganic membranes with several layers. Catal Today 2001; 67:205–216. Kikuchi E. Membrane reactor application to hydrogen production. Catal Today 2000; 56:97–101. Moulijn JA, Makkee M, Diepen AEv. Chemical Process Technology. West Sussex, UK: Wiley, 2001:453. Bodke AS, Olschki DA, Schmidt LD. Hydrogen addition to the Andrussow process for HCN synthesis. Appl Catal A General 2000; 201:13–22. Sie ST, Lebens PJM. Monolithic reactors for countercurrent gas-liquid operation. In: Cybulski A, Moulijn JA, eds. Structured Catalysts and Reactors. New York: Marcel Dekker, 1998:305–321. Maxwell IE. Innovation in applied catalysis. Cattech 1997; 1:5–13. Schanke D, Bergene E, Holmen A. Fischer–Tropsch synthesis. Patent WO 98/38147, U.S., 1998. Bengtsson E. Process in the production of hydrogen peroxide. Patent EP 0 384 905, European, 1990.


37. 38. 39.

40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51.

52.

53.

54. 55.

56.

Berglin T, Herrmann W. A method in the production of hydrogen peroxide. Patent EP 0 102 934, European, 1983. Schanke D, Bergene E, Holmen A. Fischer–Tropsch synthesis. Patent WO 98/38147, World, 1998. Machado RM, Parillo DJ, Boehme RP, Broekhuis RR. Use of a monolith catalyst for the hydrogenation of dinitrotoluene to toluendiamine. Patent US6,005,143, US, 1999. Schanke D, Bergene E, Holmen A. Fischer–Tropsch synthesis. Patent US6,211,255, US, 2001. Reesink BH, Vaarkamp M. Chemical process in reactor with structured catalyst. Patent EP 1 121 976 A1, Europe, 2000. Cornelison RC, Alcorn WR, Baillie IC. Process for hydrogenation of organic compounds. Patent EP 0 233 642 A2, Europe, 1987. Edvinsson RK, Moulijn JA. Monolietreactor. Patent application NL 1004961, The Netherlands, 1999. Makkee M, Kapteijn F, Moulijn JA. Reactorvat. Patent NL93/00231, The Netherlands, 1999. Sie ST, Moulijn JA, Cybulski A. Internally finned channel reactor. Patent PCT94901066.4, International (Europe), 1999. Sie ST, Cybulski A, Moulijn JA. Process for catalytically reacting a gas and a liquid. Patent EP 0 667 807 B1, Europe, 1998. Irandoust S, Andersson B. Simulation of flow and mass transfer in Taylor flow through a capillary. Computers Chem Eng 1989; 13:519–526. Thulasidas TC, Abraham MA, Cerro RL. Axial dispersion of bubble-train flow in capillaries. Chem Eng Sci 1996. Thulasidas TC, Abraham MA, Cerro RL. Flow patterns in liquid slugs during bubble-train flow inside capillaries. Chem Eng Sci 1997; 52:2947–2962. Kapteijn F, Nijhuis TA, Heiszwolf JJ, Moulijn JA. New nontraditional multiphase catalytic reactors based on monolithic structures. Catal Today 2001; 66:133–144. Heibel AK, Heiszwolf JJ, Kapteijn F, Moulijn JA. Influence of channel geometry on hydrodynamics and mass transfer in the monolith film-flow reactor. Catal Today 2001; 69:153–163. Lebens PJM, Stork MM, Kapteijn F, Moulijn JA. Hydrodynamics and mass transfer issues in a countercurrent gas–liquid internally finned monolith reactor. Chem Eng Sci 1999; 54:2381–2389. Lebens PJM, Heiszwolf JJ, Kapteijn F, Moulijn JA. Gas-liquid mass transfer in an internally finned monolith operated countercurrently in the film-flow regime. Chem Eng Sci 1999; 54:5119–5125. Nijhuis TA, Kreutzer MT, Romijn ACJ, Kapteijn F, Moulijn JA. Monolithic catalysts as more efficient three-phase reactors. Catal Today 2001; 66:157–165. Lebens PJM, Meijden Rvd, Edvinsson RK, Kapteijn F, Sie ST, Moulijn JA. Hydrodynamics of gas–liquid countercurrent flow in internally finned monolithic structures. Chem Eng Sci 1997; 52:3893–3899. Lebens PJM, Kapteijn F, Sie ST, Moulijn JA. Potentials of internally finned monoliths as a packing for multifunctional reactors. Chem Eng Sci 1999; 54: 1359–1365.


57.

58.

59.

60.

61.

62. 63. 64.

65.

66. 67. 68.

Lebens PJM, Edvinsson RK, Sie ST, Moulijn JA. Effect of entrance and exit geometry on pressure drop and flooding limits in a single channel of an internally finned monolith. Ind Eng Chem Res 1998; 37:3722–3730. Lebens PJM, Edvinsson RK, Sie ST, Moulijn JA. Effect of entrance and exit geometry on pressure drop and flooding limits in a single channel of an internally finned monolith. Ind Eng Chem Res 1998; 37:3722–3730. Andersson B, Irandoust S, Cybulski A. Modeling of monolith reactors in three-phase processes. In: Cybulski A, Moulijn JA, eds. Structured Catalysts and Reactors. Chemical Industries, Vol. 71. New York: Marcel Dekker, 1998:267–304. Hatziantoniou V, Andersson B, Schöön N-H. Mass transfer and selectivity in liquidphase hydrogenation of nitro compounds in a monolithic catalyst reactor with segmented gas–liquid flow. Ind Eng Chem Process Des Dev 1986; 25:964–970. Hatziantoniou V, Andersson B. The segmented two-phase flow monolithic catalyst reactor. An alternative for liquid-phase hydrogenations. Ind Eng Chem Fundam 1984; 23:82–88. Irandoust S, Andersson B. Mass transfer and liquid-phase reactions in a segmented two-phase flow monolithic catalyst reactor. Chem Eng Sci 1988; 43:1983–1988. Irandoust S, Andersson B. Monolithic catalysts for nonautomobile applications. Catal Rev Sci Eng 1988; 30:341–392. Cybulski A, Moulijn JA. The use of monolithic catalysts for three-phase reactions. In: Moulijn JA, Cybulski A, eds. Structured Catalysts and Reactors. New York: Marcel Dekker, 1998:239–266. Smits HA, Stankiewicz A, Glasz WC, Fogl THA, Moulijn JA. Selective three-phase hydrogenation of unsaturated hydrocarbons in a monolithic reactor. Chem Eng Sci 1996; 51:3019–3025. Stankiewicz A, Moulijn JA. Process intensification: transforming chemical engineering. Chem Eng Prog 2000; 96:22–34. Stankiewicz A. Process intensification in in-line monolithic reactor. Chem Eng Sci 2001; 56:359–364. Stankiewicz A, Moulijn JA. Proces intensification. Ind Eng Chem Res 2002; 41: 1920–1924.


7 Inline and High-Intensity Mixers Andrew Green BHR Group Limited, Cranfield, England

1.

INTRODUCTION

High-intensity inline devices are often used to mix fluids in the process industries. Such devices include simple pipes, baffled pipes, tees, motionless mixers, dynamic mixers, centrifugal pumps, ejectors, and rotor/stator mixers. In addition to their traditional application in physical processes such as mixing and dispersion, such devices can provide very effective environments for mass transfer and chemical reaction to take place. Furthermore, combining effective inline mixing with heat transfer is the basis of combined heat exchanger reactors (HEX reactors). The chapter provides insight on the importance of mixing and how it relates to process intensification using inline mixers. Design information for inline devices such as motionless mixers, T mixers, ejectors, and HEX reactors is provided. This should assist the reader to: (a) understand the advantages and disadvantages of these devices as process tools for single-phase, gas–liquid, and liquid–liquid applications, (b) evaluate manufacturers bids, and (c) identify opportunities for intensifying processes, as either a retrofit for existing plant or as a new process. 1.1.

Why Is Mixing Important?

Consider a simple chemical reaction, where two reactants A and B come together and produce a product R: AB → R


The reaction will have an intrinsic kinetic rate, usually dependent on the local concentrations of A and B. Often it will produce heat (exothermic reaction) or require heat input (endothermic reaction). If this is not removed (or supplied) fast enough, the temperature will rise (or fall), possibly by tens or even hundreds of degrees. Clearly this could have disastrous consequences, particularly because the rate of reaction will increase with temperature, potentially leading to a runaway reaction. For the reaction to take place, A and B need to be brought together; the reactor must be mixed. This is usually not a problem in a chemist’s beaker, where mixing can be very rapid. However, if it scaled up to a batch stirred vessel, mixing inevitably becomes slower and may take several minutes in a typical production-scale vessel. If this mixing time is slower than the reaction time, the reaction will be artificially slowed down. It becomes mixing, rather than kinetic, limited. In other words, process inefficiency is built in. For highly exothermic reactions, matters become even worse. As a vessel is scaled up, the ratio of heat transfer area to volume reduces, so its ability to remove heat reduces. To cope, a process design chemist will alter conditions to slow the reaction down. This might involve running at lower concentrations (i.e., more solvent) or operating semibatch—feeding B in slowly over many minutes or hours so that the system can cope with the heat release. If a chemical reaction that would naturally take place in a few seconds is slowed down to take 12 hours or more, it is clearly inefficient. Reactions are rarely as simple as this. There will often be other reactions competing with the desired reaction; for example, AB → R RB → S In other words, the desired product R reacts with reactant B to form by-product S. If the second reaction is much slower than the first, there should not be too much S formed. However if mixing is slow, the first reaction can be artificially slowed down, which will then tend to favor production of S—and yield will reduce. The flow pattern in the reactor will also influence the production of S. For “backmixed” flow, as occurs in a stirred vessel, the product stream from the reaction zone will be continually recirculated back into contact with the reactant stream, exposing R to fresh B. In a “plug flow” reactor, reactants are brought together in the reaction zone and then removed, reducing the likelihood of the formation of S. To summarize, production of R will be optimized by ensuring that mixing is faster than the desired reaction step and that the reactor operates in plug flow. The reactor is the nucleus of the process. Getting the fluid dynamics right in the reactor means improved safety, productivity, and selectivity, which in turn influences upstream (reduced raw material costs) and downstream (reduced separation and waste treatment costs); see Figure 1.


FIGURE 1 Importance of mixing on reactor design.

1.2.

Process Intensification

Process intensification (PI) has various definitions, but from the point of view of this chapter it is considered to be a design philosophy in which the fluid dynamics of the plant are designed to meet the chemical and physical requirements of process so that it can proceed at its optimal rate. As such, it integrates chemistry and chemical engineering approaches. This can be illustrated by the generalized “S curve” shown in Figure 2. If “plant performance” is poor (e.g., the mixing rate is much lower than the natural speed of the desired reaction), then so is “process performance” (e.g., selectivity). As plant performance improves (e.g., the mixing rate is increased), so does process performance, up to the point where it becomes chemistry limited. An optimum PI design will be one where the chemistry is

FIGURE 2 S curve of plant and process performance.


designed to give (in the absence of plant restrictions) the desired process performance; the plant is then designed to operate at the point at the top of the S curve as it “flattens out.” Moving further to the right means overdesign and increased capital and/or running costs. In summary, PI aims to match: Mixing rate to reaction rate Heat transfer performance to heat generation Residence time to reaction time Flow pattern to reaction scheme 1.3.

Motionless (Static) Mixers

A wide range of motionless (static) mixers is available on the market (Figure 3). They are pipe inserts that generate radial mixing (i.e., across the pipe) and (for multiphase systems) interfacial surface area (e.g., to produce fine bubbles or droplets). The energy for mixing is extracted from the mean flow; as such, an extra pumping

FIGURE 3 Motionless mixers: Chemineer HEV; Sulzer/Koch SMV and SMX; Chemineer Kenics.


duty is incurred. Originally designed for laminar-flow applications, they find wide use in all flow regimes. The number of elements required for any application is dependent on the difficulty of the mixing duty, more elements being necessary for difficult tasks. Motionless mixers (and inline mixers in general) present an alternative to the more traditional agitated vessel. These devices are particularly useful for the continuous processing of chemicals but are also incorporated as part of a batch system in pump-around loops. 1.3.1.

Attributes and Benefits

Mixing in a motionless mixer is rapid and is achieved by the action of splitting and twisting of the flow by the mixer elements. Energy dissipation rates are high, with typical values between 10 and 1000 W/ kg, compared to an upper limit of around 5 W/kg in conventional equipment, such as stirred tanks. These large dissipation rates give rise to much higher mixing rates for intensified mixers when compared to stirred tanks. When two phases are mixed together (gas–liquid, immiscible liquid–liquid), a fine dispersion of bubbles or drops and a high specific interfacial area are produced because of the intensive turbulence and shear. For this reason, resistance to interphase mass transfer is considerably smaller than in conventional equipment. In addition, a wide range of gas–liquid flow ratios can be handled, whereas in stirred tanks the gas-flow rate is often limited by the onset of flooding. Mass transfer coefficients (kLa) can be 10–100 times higher than in a stirred tank. The flow pattern in a motionless mixer is approximately plug flow; i.e., different elements of fluid spend similar time periods in the mixer. Residence time is usually short. The combination of rapid mixing and uniform, short, residence times is specifically favorable for carrying out reactions with fast kinetics. Motionless mixers are compact, thus requiring a small site and a lower capital expenditure (CAPEX). Inherent safety is improved due to a smaller reacting inventory. In addition, since there are no moving parts, sealing problems are reduced and maintenance is minimized. 1.3.2.

Limitations

High-intensity mixers are not suited to slow reactions (i.e., reaction times greater than a few minutes) where long residence times are required. However, it should always be questioned whether the reaction is intrinsically slow or whether it has been artificially slowed to operate safely in a stirred tank. Because these reactors are almost by definition designed to meet the needs of a specific reaction, there can be a lack of flexibility if a multiproduct plant is required. This has, however, been addressed by BHR Group’s FlexReactor (Figure 4), which combines motionless mixers in a highly reconfigurable package.


FIGURE 4 FlexReactor (from BHR Group).

2. 2.1.

MIXING CONCEPTS Reynolds Number

The Reynolds number is the ratio of inertial to viscous forces in a flow. For a pipe: Re p

u p d p

(1)

The value of Re indicates the flow regime for a specific system. A particular regime is a property of the flow field, not the fluid, which is why the Reynolds number is useful. Re is an important parameter for mixing considerations because the flow regime determines the mixing mechanisms of the flow field. At high Re, inertial forces dominate. Energy input is required to sustain turbulent eddies, which are active at different length scales; a degree of “self-mixing” exists. At low Re, viscous forces dominate. External energy input is required to stretch, chop, and fold fluid and accelerate molecular diffusion. Ultimately, all energy input is dissipated to heat.


2.2.

Hydraulic Diameter

The empty-pipe Reynolds number is based on the inner pipe diameter and superficial fluid velocity [Eq. (1)]. If the pipe contains a motionless mixer, Eq. (1) needs to be modified to take into account the “metal” in the mixer, which reduces the effective diameter but increases the fluid velocity (because it blocks part of the cross section). The theoretically sound characteristic dimension for a motionless mixer is the hydraulic diameter, given by dH 4

Area open to flow Wetted perimeter

(2)

The mixer velocity is the superficial velocity divided by the mixer voidage (), giving Re H

u p d H

(3)

For a motionless mixer, ReH Rep, and values of dH /dp vary significantly from mixer to mixer, as shown in Table 1. 2.3.

Pressure Drop

In motionless mixers the energy input for mixing is provided by the pressure loss from the mean flow. All manufacturers can provide pressure drop data. These are usually given as a friction factor or as a multiplier for the empty-pipe pressure drop. Values range from 30 to 1000 times the empty-pipe friction factor. 2.3.1.

Friction Factors

Care must be taken when comparing the pressure drop in motionless mixers, because three definitions exist. In this chapter, Moody’s friction factor is adopted, TABLE 1 Ratios of Hydraulic Diameter to Pipe Diameter for Motionless Mixers Manufacturer

Mixer type

dH /dp (%)

Sulzer/Koch

SMV SMX SMXL Kenics KMS HEV

7–25 33 48 48 86

Chemineer


T1

which is the ratio of the pressure loss in one diameter’s length of pipe to the mean velocity pressure: fM

2p( d p / Lm )

(4)

u 2p

For fully turbulent flow (generally ReH 10,000), fM is roughly constant. For laminar flow (ReH 1000), the product fM Re is a constant. Between these values (transitional flow), fM is a function of Re. Approximate fully turbulent friction factors for motionless mixers are given in Table 2. (Note: These figures are approximate and for comparison purposes only; they should not be used for design. The true friction factors vary slightly with Re and scale.) The other friction factors in common use are the Newton number and Fanning’s friction factor. The relationship between the three is: fM 2 Ne Moody’s Newton number

4 fF Fanning’s

(5)

2.3.2. Pressure Drop and Energy Dissipation: Turbulent Flow For motionless mixers, energy is extracted from mean flow. Data can be correlated using an analogy with a rough pipe: Re p

u p d p

(6)

This can then be used to determine the total energy dissipation rate in the mixer (in W/kg):

Qp f u3 M Vm 2d

(7)

TABLE 2 Approximate Friction Factors for Motionless Mixers Manufacturer

Mixer type

fM (approx.)

Sulzer/Koch

SMV SMXL Kenics (KMS) HEV Empty pipe

6 2.5 2 0.4 0.001–0.03

Chemineer


Equation (7) illustrates how critical the pipe diameter is to the energy dissipation rate: At fixed throughput, ∝

I d7

(8)

The energy dissipation rate is a critical parameter in motionless mixers, because it affects the rate of mixing. However, not all of the energy dissipated is useful for mixing; in particular, laminar dissipation due to shear at the pipe wall or the mixer does not contribute to mixing. Total energy dissipation can be split into dissipative losses (ED) and turbulent energy dissipation ( ): ED 2.4.

(9)

Turbulent-Mixing Length Scales

Turbulent mixing is a complex phenomenon that takes place at a number of scales. Three scales of mixing can be defined (macromixing, mesomixing, and micromixing). Macromixing, or blending, is the spreading of an additive by convective flow patterns and turbulent dispersion. It occurs at scales of typically 102 –103 m. Following this dispersion, the largest turbulent eddies are broken down into the smallest turbulent eddies; this is the process of mesomixing, which occurs at scales of typically 103 –104 m. Below the size of the smallest eddy, viscous forces dominate; this is the scale of micromixing. Various processes occur at this scale, starting with folding and wrapping (“engulfment,” at scales of 104 –105 m), followed by stretching of small eddies with diffusion (106 –107 m). 3.

MIXING AND REACTION

For a reactive process, the reactants must be brought into contact by mixing before a reaction can occur. In a motionless mixer in turbulent flow, the pressure drop defines the turbulent energy dissipation rate, which then determines the macro-, meso-, and micromixing rates. 3.1.

Slow Reactions

For a “slow” reaction, the mixing rates are all much faster than the inherent kinetics; in this case the mixing and reaction processes are decoupled. For a motionlessmixer system, there must be sufficient residence time downstream of the mixer for the reaction to go to completion. For long reaction times, a stirred tank can be used to give the required residence time. The process then becomes: Component mixing

t 1/2m ⇒

reaction


t 1/2r ⇒

products

The term slow needs to be used with care in PI applications. With mixing times typically well below 1 second, reactions of only a few seconds’ duration can be considered slow for a motionless mixer. 3.2.

Fast Reactions

For a “fast” reaction, the time scales for mixing and reaction are of the same order (i.e., t 1/2m ≈ t 1/2r ), so mixing and reaction are no longer consecutive processes but simultaneous: Component mixing and reaction

t 1/2m , t 1/2r ⇒

products

For fast reactions, the mixing rate can limit the product rate of formation and, as described in the next section, product quality/yield. 3.3.

Multiple Fast Reactions

Chemical processes often involve multiple, competing reactions. A common situation is that of a competitive-consecutive reaction, such as that described in Section 1.1, where reactant A and the desired product R are competing for reactant B. The selectively for waste product S can be defined as XS

2cS c R 2c S

(10)

With very fast mixing (t1/ 2 m << t1/ 2), the distribution of products is determined by the relative kinetics of the two reactions: If the desired reaction is much faster than the undesired reaction, Xs will tend toward zero. However, slow mixing compared to the fastest reaction (i.e., t1/ 2 m >> t1/ 2 ) slows down the desired reaction, leading to high waste selectivities, Xs → 1. It should be noted that in this analysis, t1/ 2 m is the maximum of the three mixing time scales ( macro, meso, micro). Dependent on the mixing conditions, geometry, and chemistry, any one mixing time scale can be rate determining. 4. 4.1.

MIXING PERFORMANCE OF INLINE MIXERS Macromixing (or Blending) Performance

Measurements of macromixing by, for example, a motionless mixer are based on the coefficient of variation (CoV), which is a statistical measure of radial homogeneity at the macroscale. It is defined as the standard deviation of concentration measurements made at the exit of a mixer divided by the mean concentration:


CoV

  n  (ci c ) 2 ] / (n 1)   i 1 c c

∑

(11)

n is the number of measurements (e.g., conductivity probes or sampling positions) used over the pipe cross section (usually 9). ci is the time-averaged concentration of the ith probe. CoV characterizes the degree of blending achieved between an additive and the bulk stream. The lower the CoV, the better the streams are mixed. A CoV of 5%, or 0.05, is often used as the benchmark. The physical meaning of this value is that there will be a 95% probability that all samples taken will be 2CoV (i.e., 10%) of the mean mixed concentration. Correlations for CoV are usually expressed in terms of the ratio between the coefficient of variation downstream to the coefficient of variation at the inlet to the mixer [i.e., CoV/(CoV)0]. 4.2.

Blending Correlations

4.2.1.

Empty Pipe: Turbulent Flow

Blending performance for an empty pipe is critically dependent on where and how the fluid is injected [Eqs. (12) and (13) and Ref. 1]. Centerline injection: CoV/(CoV)0 2 exp(0.75f D1/2 L /d ) Wall injection (low velocity): CoV/(CoV)0 2 exp(0.25 f

(12) 1/ 2 D

L /d ) (13)

It can be shown that (CoV)0 (Q/q)1/2

(14)

For example, for q/Q 104, Rep 105, fD 0.02, and CoV 0.05, L/d for centerline injection is 78, whereas for wall injection it is 234. It should be noted that when additive flow has significant momentum, much more rapid blending is possible (so-called T mixer). An optimum value of momentum ratio between main flow and additive can be found (see Ref. 2 for details). 4.2.2. Turbulent and Transitional Flow Mixing in Motionless Mixers Motionless-mixer manufacturers usually have experimentally based correlations to predict macromixing performance in turbulent flow. These often use slightly different bases, so care has to be taken when comparing performance. For the Koch/Sulzer SMV in turbulent flow (Rep 2000), significant mixing continues to be achieved for several diameters downstream of the mixer.


Usually mixing is measured/defined two diameters downstream of the mixer, and performance is given by (3) CoV/(CoV)0 2 exp(1.5 L /d p )

(15)

where (mixing length) L Lm 2dp Lm nLe (mixer length) Equation 15 applies for viscosity ratios µB/µA 100 and shows that the mixing length to achieve a given CoV is 13 times shorter than for an empty pipe. To take advantage of the mixing downstream of the mixer, SMV elements are often spaced out in pairs in turbulent flow. The length of one element depends on the mixer diameter: For mixer diameters 100 mm, the length of an element (Le) is equal to one pipe diameter, but for mixer diameters 100 mm, Le 0.5dp. For the Chemineer Kenics and HEV mixers, a correlation has been developed that covers both mixers (4). For fully developed turbulent flow (ReH 8700): log10 [CoV/(CoV) 0 ] 1.65Re 0.043 (0.0879n 0.7363) H

(16)

For 1000 ReH 8700: log10 [CoV/(CoV) 0 ] 0.27Re 0.24 H ( 0.0879n 0.7363)

(17)

Equations (16) and (17) are for measurements three pipe diameters downstream of the mixer and are valid for viscosity ratios µB/µA 100. The macromixing length (say, to give CoV 0.05) is insensitive to Re under fully turbulent conditions. So for a higher velocity, though the mixer length remains constant, the time for mixing will be shorter. Put another way, macromixing time is inversely proportional to pipe velocity. For transitional flow, precise correlations are not available; but for both SMV and Chemineer Kenics mixers, extra elements are required to achieve a certain degree of mixing. The SMV does not achieve significant mixing downstream of the mixer as in turbulent flow, so elements are not spaced out. The HEV is not recommended for transitional flow. 4.2.3. Mixer Rankings for Turbulent-Flow Blending Applications The ranking of mixers for a blending application will depend on what the user is trying to achieve. If blending efficiency is most critical (i.e., achieving the required mixing for minimum pressure drop/energy use), the most efficient mixer is, in fact, an empty pipe. After that the ranking is: Empty pipe HEV Kenics or SMV SMX


However, if the rate of mixing is important (i.e., the need to achieve the most mixing in the shortest length), the ranking becomes: SMV HEV Kenics SMX Empty pipe The performance of the HEV should be noted: It was designed to be hydrodynamically efficient and to use flow vortices, rather than “metal,” to achieve mixing. Despite a low pressure drop, the HEV can achieve mixing more efficiently and in a shorter length than the Kenics [as shown in Eqs. (16) and (17), it has the same design correlation with respect to number of elements, but element spacing is shorter]. However, care needs to be taken in its installation, because performance can be significantly degraded if there is an uneven flow distribution at its inlet (e.g., if it directly follows a bend). The poor performance of the SMX is to be expected because it has been designed specifically for laminarflow applications and is not recommended for turbulent-flow applications by its manufacturers. 4.2.4.

Axial Dispersion

The coefficient of variation is a measure of mixing across the pipe cross section (“radial dispersion”). In pipe-flow and motionless mixers, mixing along the length of the mixer (axial dispersion) occurs. This can be described in terms of the residence time distribution (or RTD), which is a measure of age distributions for fluid elements passing through the mixer. An empty pipe has relatively high axial dispersion, primarily caused by the flow profile that is established in a pipe (i.e., the fluid in the center of the pipe flows faster than that near the walls). Motionless mixers tend to have a much tighter RTD. 4.2.5.

Blending Correlations: Laminar Flow

Despite the wide use of the “striation thickness” concept in the early commercial literature, the CoV is now the most widely used mixing index. The following correlations are valid for viscosity ratios 0.01 µB/µA 100, feeding into the center of the pipe and with CoV measured 2dp downstream for Sulzer mixers, 3dp downstream for Kenics mixer. SMX (Rep 200, Le /dp 1.0): log10 (CoV/(CoV) 0 ) ⬇ 0.19n ⬇ 0.19 Lm /d p

(18)

SMXL (Rep 200, Le /dp 3.3): log10 (CoV/(CoV) 0 ) ⬇ 26n ⬇ 0.078 Lm /d p

(19)

Kenics (Rem 200, Le /dp 1.5): log10 (CoV/(CoV) 0 ) ⬇ 0.098n ⬇ 0.067 Lm /d p


(20)

where n number of motionless-mixer elements Le length of one mixer element These data are consistent between the manufacturers’ and independent investigators. 4.2.6.

Mixer Rankings for Laminar-Flow Applications

As with turbulent flow, ranking depends on the requirements of the application. Based on various literature data, where at least two datasets exist, the following rankings can be made. Energy efficiency (most efficient first): SMXL → Kenics → SMX → Hi-mixer → Komax → Lightnin → Ross ISG Mixing rate (most rapid first): SMX → Ross ISG → Hi-mixer → SMXL → Kenics → Komax → Lightnin 4.2.7.

High Viscosity Ratios

Laminar-flow blending duties involving high viscosity ratios (greater than 1000 : 1) are classified as difficult. The SMX mixer appears to have the best track record in achieving satisfactory results. 4.3.

Mixing with Reaction in Inline Mixers

If a mixer is to be used for reactive processes, it should be designed such that the longest mixing time scale (whether micro-, meso-, or macromixing) is significantly shorter than the characteristic time scale of the desired chemical reaction. As mentioned in Section 3, any of the time scales can be rate determining. 4.3.1.

Micromixing Limited

If micro meso macro , then the process is micromixing controlled. Micromixing is a complex phenomenon (Section 2.4), but for most liquids engulfment is the longest step. In this case, micromixing time is the inverse of engulfment rate (E) and can be estimated by

micro

1 1     E 0.058  ε 

1/ 2


(21)

Detailed models have been developed that can predict product distribution for competitive reactions with known kinetics under micromixing- (engulfment-) controlled conditions (5). 4.3.2.

Mesomixing Limited

In motionless mixers, mesomixing time can be estimated from  Q

meso 2.17  B   nf uε 

1/ 3

(22)

Numerical models for mesomixing control with reactions are being developed, although they are more complex than micromixing models and require the input of empirically determined length scales. Mesomixing limitations give rise to worse process performance than if micromixing alone were limiting, so if possible mesomixing time should be reduced (e.g., by increasing the number of additive feeds or reducing the additive flow rate) to the point that micromixing controls. However, in practice this is often not possible. 4.4. Scale-Up/Scale-Down of Motionless Mixers (Single Phase) For systems involving fast reactions where reactor performance has been established at one scale and equal performance is required at different scales, the criteria for scale-up/scale-down are: The mixing rate of the limiting step (characteristic time scale) should be kept constant. Residence time in the mixer should be constant. The limiting mixing mechanism should not change. The process conditions should remain the same, e.g., reactant concentrations, flow rate ratio, mixer type, relative feed position. If the friction factor, mixer voidage, and turbulence-generating efficiency do not vary significantly with scale, then the following scale-up rules can be applied (where k Qnew /Qold, the mixer diameter should be rescaled from dold to dnew). 4.4.1.

Macromixing or Mesomixing Limitation

For a macro- or mesomixing limitation, dnew k 1/ 3 dold


(23)

The resulting design will have the same number of elements at the new scale (resident time constant), so the length of the mixer will be d  lnew  new  lold  dold  4.4.2.

(24)

Micromixing Limitation

For a micromixing limitation, dnew k 3/ 7 dold

(25)

If residence time is kept constant, fewer elements will be required at the larger scale, and 1/ 3

d  lnew  new  lold  dold  5.

(26)

GAS–LIQUID MIXING

5.1.

Introduction

Gas–liquid reactions form an integral part of the production of many bulk and specialty chemicals, such as the dissolution of gases for oxidations, chlorinations, sulfonations, nitrations, and hydrogenations. When the gaseous reactant must be transferred to the liquid phase, mass transfer can become the rate-limiting step. In this case, the use of high-intensity mixers (motionless mixers or ejectors) can increase the reaction rate. Conversely, for slow reactions a coarse dispersion of gas, as produced by a bubble column, will suffice. Because a large variety of equipment is available (bubble columns, sieve trays, stirred tanks, motionless mixers, ejectors, loop reactors, etc.), a criterion for equipment selection can be established and is dictated by the required rate of mass transfer between the phases. 5.2.

Mixer Types

5.2.1.

Motionless Mixers

When a gas stream is introduced into a turbulent liquid flow in a motionless mixer, the gas is broken up into bubbles. The breakup is due mainly to the turbulent shear force of the liquid but also partly to the collision between gas and the leading edge of an element. There are two basic operating modes: Continuous, as a stand-alone mixer Loop operation (either semibatch or continuous).


A stand-alone mixer requires the mass transfer/reaction to be completed within the mixer. If the gas flow rate matches the stoichiometry of the liquid phase, all the gas should be dissolved and reacted at the end of the mixer. This generally involves very high volumetric ratios between gas and liquid. If there is excess gas, there will be some gas at the mixer outlet, which needs to be separated. 5.2.2. Gas–Liquid Ejectors Ejectors consist of four main sections (Figure 5): Spinner—orients and stabilizes the flow. Nozzle—provides a high-velocity jet of fluid. Gas chamber—the high-velocity jet creates suction in the gas chamber, entraining gas into the ejector.

FIGURE 5 Gas–liquid ejector.


Mixing tube—on leaving the gas chamber, the liquid jet attaches itself to the mixing tube wall, resulting in a rapid dissipation of kinetic energy, creating an intensive mixing zone known as the mixing shock region. High turbulence in this region breaks up the gas, producing a fine dispersion of bubbles and consequently a large interfacial area for mass transfer. By and large, ejectors and motionless mixers have similar mass transfer performance at a given gas-to-liquid flow ratio and energy input. However, ejectors have a number of benefits and drawbacks compared to a motionless mixer. On the positive side, the ejector suction means that a pressurized gas supply is not required. The unrestricted mixing tube means that solid formation due to reaction is not problematic. Against this, the operation is sensitive to changes in the gas–liquid flow ratio and diameter/length ratio. Gas-to-liquid flow ratios are also more limited in ejectors. 5.3.

Loop Reactors

Motionless mixers and ejectors are useful for applications requiring short residence times (on the order of seconds or less). If long residence times are required, e.g., if the reaction is relatively slow, the use of a motionless mixer alone would lead to a very long mixer, which may not be practical. One way to overcome this problem is to use a loop reactor, which combines a high-intensity mixer, such as a motionless mixer or ejector, with a separation tank. 5.4.

Guide to Equipment Selection

Tables 3 and 4 summarize where different mixers/configurations are most appropriate. TABLE 3 Application of Mixer Types and Configurations Motionless-mixer stand-alone Solids present Slow reaction Fast reaction Energy efficiency important Low gas Pressure


Motionlessmixer loop

Ejector stand-alone

Ejector loop

TABLE 4 Application of Chemineer Kenics and Koch/Sulzer SMV Mixers Mixer type Situation p constraint Space constraint Solids present Energy efficiency important Need for heat removal

5.5.

Kenics

SMV

Mixing Concepts

5.5.1.

Rate of Mass Transfer

The rate of mass transfer for motionless mixers and ejectors can be described by NA KLa CLM

(27)

where NA is the amount of transferred species per unit time per unit dispersion volume, KL is the overall mass transfer coefficient, a is the specific surface area for mass transfer, and CLM is the log mean concentration driving force. 1/KL, the overall mass transfer resistance, is usually dominated by the resistance in the liquid phase, 1/kL. Consequently, the gas-phase resistance can be neglected and KL kL. However, it is imperative that this assumption be checked, because it does not always hold for very soluble gases or when kLa is enhanced by reaction (6). The large levels of turbulent energy dissipation produced in high-intensity mixers act to reduce the bubble size, typically from 0.5 to 2.0 mm in high-intensity mixers, compared to 1.0 to 5.0 mm in stirred tanks and bubble columns. In addition, much higher gas-to-liquid ratios can be achieved, and turbulence enhances kL, leading to overall mass transfer coefficients (kLa) 10–100 times greater than for a stirred tank. 5.5.2.

Reaction Regime

The relative speed of kinetics to mixing is described by the Hatta number, t  Ha  MT   tR 

0.5


(28)

where tMT and tR are the time constants for mass transfer and reaction, respectively. The lower the value of Ha, the faster the mixing relative to the intrinsic reaction rate. Ha can be calculated from D 2  ( n A 1) nB Ha  2A k AB CAL CBL  n A1  kL

0.5

(29)

Reactions are often classified into four categories: Ha 0.02 0.02 Ha 2 Ha 2

slow reaction moderately fast reaction very fast or instantaneous reaction

For Ha 0.02, there is a considerable scope for process intensification. If a reaction is intrinsically fast (a large reaction rate constant) the design aim is to provide sufficiently intense mixing to move it into the slow reaction regime (Ha 0.02) such that the reaction is limited by the intrinsic reaction rate rather than the mass transfer rate. In order to establish the reaction regime and to design equipment, the following need to be known: Flow pattern Mass transfer coefficient Bubble sizes These can be determined from gas–liquid flow rates, the energy dissipation rate (driven by the pressure drop), and the physical properties of the fluids. 5.6.

Design Guidelines and Correlations

5.6.1. Flow Patterns Gas–liquid flows are much more complicated than single-phase flows, due to the existence of the gas–liquid interface. The phases can be present in a range of possible flow regimes (flow patterns), which are dependent upon the physical properties of both phases, the flow rates, and the equipment size and orientation. The most commonly noted flow patterns are (7): Annular flow—a liquid film on the walls and a continuous gas phase, containing a mist of liquid droplets, in the core Intermediate slug flow—large gas voids containing liquid droplets Bubble flow—continuous liquid flow with a dispersion of gas bubbles Figure 6 shows the various flow patterns in horizontal flow, and similar patterns can be seen in vertical upflow or downflow. In general, bubble flow develops under high liquid-flow rates and low gas-flow rates; annular flow develops under


FIGURE 6 Gas–liquid flow patterns in horizontal flow.

low liquid-flow rates and high gas-flow rates; stratified flow develops in low gasand liquid-flow rates. Bubble flow is generally more desirable for liquid-film controlled mass transfer processes because of the high turbulence level in the liquid phase, while an annular flow is more desirable for gas-film controlled processes, where the turbulence level in the gas phase is high. However, in reactive systems, stoichiometry will often define the gas- and liquid-flow rates, leaving no choice for the flow pattern. Having said this, motionless mixers and ejectors can maintain the bubble flow regime even at high gas-to-liquid flow ratios, where flooding of the impeller would occur in a stirred tank or annular flow develop in empty pipes (Figure 7 (8)). 5.6.2.

Pressure drop

Pressure drop is a critical parameter, in that it determines pumping requirements and enables the power input to the mixer to be calculated. Motionless Mixers. Major mixer manufacturers agree that the Lockhart and Martinelli parameters for two-phase flow in pipes (9) can also be applied to motionless mixers. To estimate the pressure drop, the single-phase liquid and gas pressure drops are first calculated. The Lockhart and Martinelli para-meter X is found from X

pL pG


(30)

FIGURE 7 Flow pattern maps for cocurrent air–water upflow through motionless mixers. (From Ref. 8.)


and the two-phase pressure drop can be calculated from p L2 pL G2 pG

(31)

where L and G are functions of X for the liquid and gas phases, respectively, obtained from the Lockhart and Martinelli charts (Figure 8, Ref. 9) or from the following empirical equations: L ( 4.6 X 1.78 12.5 X 0.68 0.65) 0.5

(32)

G

(33)

X2

L

Ejector. Gas chamber pressure needs to be known in order to calculate ejector power input. A semiempirical equation was developed by Henzler (10) that related the entrainment ratio to other system variables:  QG D B1 0.38 m QL Dn 

 L     G

0.09

 p2   1  ps 

1/ 6

2( p2 ps )   L u 2j 

(34)

Gas chamber pressure, ps, can be calculated from this equation through iteration. Factor B depends on the mixing tube/nozzle diameter ratio for a given ejector type and needs to be determined experimentally.

FIGURE 8 Lockhart and Martinelli parameters for pressure drop in multiphase flow.


The nozzle pressure, which determines the selection of the liquid pump, is given by   D 4 1 p1 ps L u 2j 1  n   2   D1   5.6.3.

(35)

Power Input

The power input, required for calculation of the mass transfer coefficient, is calculated from Eqs. (36) and (37). For a motionless mixer, the power comes from the gas and liquid phases; for the ejector, power comes from the liquid only. In a motionless mixer: p  1 P QL p QG pav ln 1  QL u L21 u L2 2 2  p2 

(

)

(36)

)

(37)

In an ejector: p  1 P QL p QG pav ln G  QL u L21 u L2 2 2 p  2

(

Generally, kLa values at the same power input are similar between the two devices (11). 5.6.4.

Mass Transfer Coefficients

The amount of gas transferred is proportional to the product of the mass transfer coefficient (kL ) and the specific area (a). Because most measurement techniques measure this product, many correlations for kLa appear in the literature. However, caution is advised, because they can give different predictions for the same operating conditions. Equations (38) and (39) are two examples from independent investigators for motionless mixers, from Refs. 12 and 8, respectively: P k L a 1.74 104   V P k L a 0.64    m

0.8

(38)

0.75

G

(39)

The majority of reported correlations for ejectors are for loop-type configurations, e.g., (13): P k L a 0.044   V

0.76

with very little reported on the stand-alone configuration.


(40)

Comprehensive correlations for motionless mixers and ejectors have been developed as part of BHR Group’s HILINE Consortium, but these are available only to members. 5.6.5.

Bubble-Size Calculations

When a gas stream is introduced into a turbulent liquid flow in a motionless mixer, the gas is broken up into bubbles. The breakup is due mainly due to the turbulent shear force of the liquid but, for motionless mixers, also partly to the collision between the gas and the leading edge of an element. Gas dispersion is a physical process and involves bubble breakup and coalescence, which can both take place in the same mixer/reactor. Bubble breakup and coalescence are both complex processes. In a turbulentflow field, bubbles are broken up mainly due to the turbulent shear force, and the eventual bubble size is a balance between this force and the surface tension force. For a given gas–liquid system and flow field, a maximum bubble size exists. Any bubbles larger than this size will be broken up. According to theory (14), this maximum bubble size relates to gas–liquid physical properties and flow characteristics:  We ′crit  dmax    2 

0.6

(

0.6 c2 d

)

0.2

ε 0.4

(41)

Wecrit is the modified critical Weber number, which is close to 1. Coalescence occurs when two bubbles approach each other, collide, and become one bigger bubble. Two important factors are: Frequency of collision Efficiency of coalescence The frequency of collision relates to the flow pattern and gas volume fraction: The more random the flow pattern or the higher the gas volume fraction, the higher the frequency. The efficiency of coalescence relates to physical properties of the gas–liquid system. Some systems, such as air–water, have a high efficiency of coalescence and are often called “coalescing systems.” Other systems, such as gas–alcohol or gas–salt solution, have a low efficiency of coalescence and are called “noncoalescing systems.” 5.6.6.

Characterization of Bubble-Size Distribution

It is useful to define an appropriate average to characterize bubble-size distribution. For heat and mass transfer, the Sauter mean diameter (d32) is generally used: d32

ni di3 ni di2


(42)

Hesketh et al. (2) proposed that the general equation of Hinze (14) was valid for the turbulent dispersion of a gas using any motionless mixer or an empty pipe:  We′c  d32 Cn    2 

0.6

 0.6  2   L G

(

)

 0.4 2 ε  

(43)

where Cn ⬇ 0.6 and Wec ⬇ 0.6–1.6. The Sauter mean diameter (d32) is related to the interfacial area per unit volume (a) and dispersed phase volume fraction by a6

d32

(44)

where QG/(QG Q1). If kLa and a are known, kL can be estimated and, provided the reaction kinetics are known, Ha can be calculated from Eq. (29) and the reaction regime deduced. 6. 6.1.

LIQUID–LIQUID DISPERSIONS Introduction

Motionless mixers are highly effective for producing dispersions of immiscible liquids. Applications can be physical (e.g., for liquid–liquid extraction) or chemical (e.g., many nitration reactions). As with gas–liquid mixing, the most relevant parameter to measure for such applications is the Sauter mean diameter. 6.2.

Turbulent-Flow Correlations

A range of correlations is available from the literature, usually relating the Sauter mean diameter to the Weber number, which is the ratio of shear forces to surface tension forces: We c u 2p d p /

(45)

Most correlations show that d32 is proportional to the Weber number raised to the power of 0.6, which is consistent with the theory of drop breakup by turbulent shear forces. Strictly, these correlations should be applied only where the drop size is in the inertial subrange of turbulence, i.e., k d32 dp /4


(46)

where k is the Kolmogorov length scale k (v3/ )1/4 6.2.1.

(47)

Kenics Mixer

Two very similar correlations have been developed for Kenics mixers (15,16) of the form d32 CWe0.6 dp

(48)

where C 0.45 for the former and 0.49 for the latter. These are valid for fully turbulent flow in a pipe (Rep 12,000), inviscid drops (d ⬇ c) and were measured with water as the continuous phase. The dispersed-phase fraction has little effect up to a value of 0.25. Equilibrium drop size is achieved with only eight mixer elements. For viscous drops (d up to 200 mPa) in turbulent flow, Berkman and Calabrese (16) developed the correlation further to give 1/ 3   d32   d32 0.6 0.49 We 1 1.38Vi    dp  d p   

0.6

(49)

where  d u p   c  Vi      d 

1/ 2

(50)

For higher viscosity ratios, more elements were needed (24 in these experiments). 6.2.2.

Sulzer (Koch) Mixers

Sulzer published correlations in the open literature for drop size. An early correlation for the SMV is (3) d32 0.21We ′0.5 Re 0H.15 dH

(51)

where We is the Weber number based on hydraulic diameter, i.e., We ′

c u 2p d H

(52)

The correlation was developed for five 50-mm SMV elements (dH 8 mm) and covered Reynolds numbers (ReH) in the range 200–20,000 and dispersed-phase volume fraction up to 0.25.


More recent work (17) has covered viscous drops in turbulent flow, over a wide range of mixer types (SMV, SMX, SMXL, SMR, SMF), diameters (up to 80 mm), lengths, and dispersed-phase viscosities (up to 70 mPa): (1 BVi)We c  d32 0.65(1 kd )   2

0.6

    c 

0.6

0.1

 c  0.4   ε  d 

(53)

where Wec 1.8 and Vi is the viscosity number [Eq. (50)]. 6.3. Comparison Between Motionless Mixers and Stirred Tanks Comparison between drop sizes for a motionless mixer and a (well-) stirred tank under typical operating conditions yields an interesting result that similar drop sizes can be obtained in both despite very different average energy dissipation rates. This is down to the extremely inhomogeneous energy dissipation in a stirred tank, where the peak rate (usually close to the impeller tip) can be similar in magnitude to that in the more homogeneous motionless mixer. However, a stirred tank may take hours of operation to achieve an equilibrium drop size, whereas a motionless mixer will achieve it with a few elements and in less than a second. This has implications for two-phase applications when it is important to rapidly achieve mass transfer to complete fast chemical/physical reactions or to minimize byproduct formation from complex fast reactions. 6.4.

Scale-Up/Scale-Down

When scaling up or scaling down a liquid–liquid process in turbulent flow, the energy dissipation rate needs to be kept constant, giving dnew (Qnew /Qold )3/ 7 dold

(54)

However, checks should be made that flow is fully turbulent at both scales and that the drop size remains within the inertial subrange of turbulence [Eq. (46)]. As a minimum, residence time should be maintained, i.e., 1/ 3

d  lnew  new  lold  dold 

(55)

This suggests that fewer elements will be required at a larger scale; however, for a conservative design on scale-up, the same number of elements should be maintained. On scaling up or scaling down, the same mixer type and feed arrangement should be maintained.


7. 7.1.

COMBINED HEAT EXCHANGER-REACTORS (HEX REACTORS) HEX Reactor: Concepts

The mixers discussed in Sections 4–6 are particularly suitable for reactions where the required heat input (endothermic reaction) or heat production (exothermic reaction) is modest (i.e., temperature changes on reaction would be only a few degrees in the absence of any heat transfer). HEX reactors can be used for rapid, highly exothermic (or endothermic) reactions; not only are the mixing rate and residence time of a reactor matched to the kinetic rate and reaction time, but heat transfer performance is also matched to heat production (Figure 9). 7.2.

HEX Reactor Types

HEX reactors generally fall into three basic types (18). a. Jacketed motionless-mixer reactors (Figure 10): Motionless mixers provide a highly effective and efficient mixing environment for rapid reactions. Heat transfer capacity can be provided by utilizing either single mixers in jackets or multiple mixers in a shell-and-tube geometry. The “FlexReactor” (Figure 4) has been designed to package motionless mixers in a highly reconfigurable unit with effective heat transfer. b. Compact heat exchangers: There is a wide variety of commercially available compact heat exchangers available (e.g., enhanced shell and tube, plate and frame, plate–fin, and diffusion-bonded). Such devices provide extremely effective heat transfer but have not been optimized as reactors, compromising their efficiency.

FIGURE 9 HEX reactors: principle of operation.


FIGURE 10 Jacketed motionless mixer.

TABLE 5 HEX Reactor Considerations Attribute

Bespoke HEX

Motionless mixer

FlexReactor

Typical heat transfer coefficient HT density Material of construction Scale-up procedures

U 1500 W/m2-K

U 800 W/m2-K

U 600 W/m2-K

5000 m2/m3 Limited

900 m2/m3 Wide

900 m2/m3 Wide

Use more units in parallel

Cost Mixing Multipoint feed

Increase tube length/ diameter Medium High Yes—variable

Availability Residence time

High Low Yes—but fixed once constructed Low Low

Increase tube length/ diameter Low High Yes—but fixed once constructed High High

Flexibility

Low

Low


Medium High and variable High

c. Bespoke HEX reactors: A good example is the MarbondTM reactor, supplied by Chart Industries, which has been designed specifically as a heat exchanger-reactor, combining high heat transfer with effective mixing (although its high surface area results in relatively low micromixing efficiencies). Passageways are typically of the order of a few millimeters. Its diffusion-bonded construction is very robust, and it can be constructed to provide an optimized feed arrangement. The choice of reactor will be very dependent on the requirements of the chemical reaction scheme, the relative importance of mixing and heat transfer, and practical considerations (e.g., the effect of solids in the process; materials of construction; flexibility). A comparison of the typical performance of different designs is given in Table 5. HEX Reactors are discussed in more depth in Chapter 4. NOMENCLATURE Symbol

Explanation

Units

a ci c CoV Cn C*AL CBL D DAL Dm Dta, Dax Dtr d dB dH dmax dn dp d32 E EB Ep fM g KL k kL

Interfacial area Concentration of species i Mean concentration Coefficient of variation Constant ⬃ 0.6 Equilibrium concentration of A Bulk liquid concentration of B Stirred-tank impeller or rotor diameter Diffusion coefficient of A in liquid Molecular diffusivity Axial turbulent dispersion coefficient Radial turbulent dispersion coefficient Droplet diameter Feed pipe diameter Hydraulic diameter of motionless mixer Maximum stable drop size nozzle diameter Internal pipe diameter Sauter mean diameter Engulfment rate coefficient Ratio of viscous to interfacial forces Rate of direct energy dissipation Moody’s friction factor Acceleration due to gravity Overall mass transfer coefficient Second-order reaction rate constant Liquid-side mass transfer coefficient

m1 mol-m3 mol-m3 — — mol-m3 mol-m3 m m2-s1 m2-s1 m2-s1 m2-s1 m m m m m m m s1 — W-kg1 — ms2 ms1 m3-mol1-s1 ms1


Lc Le Lm LR m N NA Ni n ne nf P P1 P2 Ps p Q T t up v V X Xi z

Concentration integral length scale Length of one mixer element Length of motionless mixer Length of reaction zone mass Impeller rotational speed Mass transfer rate Total number of moles of species i Number of sampling positions used in CoV measurement Number of motionless mixer elements Number of additive feed tubes Power Inlet pressure Exit pressure Gas chamber or suction pressure Pressure drop over motionless mixer Volumetric flow rate Dimensionless time Stirred-tank diameter Time Superficial pipe velocity Velocity Liquid volume Lockhart and Martinelli parameter Selectivity for product i Ratio of mixer friction factor to pipe friction factor

m m m m kg rps mol/s mol — — — W Pa Pa Pa Pa m3-s1 — m s m-s1 m s1 m3 — — —

Greek Symbols

k

Flow rate ratio (QA,C /QB) Shear rate Turbulent energy dissipation rate Efficiency of turbulence generation Voidage of a motionless mixer Turbulence-generating length scale Kolmogorov microscale Dynamic viscosity Kinematic viscosity Fluid density Standard deviation Interfacial tension Characteristic time scale Total energy dissipation rate Dispersed phase volume fraction


— s1 W-kg1 — — m m mPa-s m2-s1 kg-m3 — N-m1 s W-kg1 —

Dimensionless Groups Da Re Sc We Wecrit Ha

Damkohler number (k2-o cB0 /E) Reynolds number (updp/v) Schmidt number (v/Dm) Motionless-mixer Weber number (cup2dp /) Stirred-tank Weber number (c N 2D3/) Critical Weber number (0.6–1.6) Hatta number

Subscripts 1 2 av c d E G H j L MT m macro meso micro mix p Q R R T 1/2r

At inlet At exit Average Continuous phase Dispersed phase Engulfment Gas Based on hydraulic diameter Jet Liquid Mass transfer Motionless mixer Macromixing Mesomixing Micromixing Slowest mixing step Empty pipe Competitive-parallel reactions Radial Reaction Tangential Reaction half-life

REFERENCES 1. 2. 3. 4.

Henzler HJ, Chem Ing Tech, 1980; 52:659–661. Hesketh RP, Russel T, Etchells AW. R&D notes. AIChE J 1987; 33(4):663–667. Streiff FA. Sulzer Tech Rev 1977; 3. Knight CS. Experimental investigation of the effects of a recycle loop/static mixer/ agitated vessel system on fast, competitive-parallel reactions. PhD dissertation, University of Arkansas, 1994. 5. Baldyga J, Bourne JR. Principles of micromixing. J Fluid Mechanics 1986; 1:147. 6. Middleton JC. In: Harnby N, Edwards MF, Nienow AW, eds. Mixing in the Process Industries. 2d ed. London: Butterworth Heineman, 1992:Chapter 15.


7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

Drahos J, Cermak. Chem Eng Processes 1989; 26:147. Roes AWM, Zeeman AJ, Bukkems FHJ. I Chem E Symposium Series, 1984; (87): 231–238. Lockhart RW, Martinelli RC. Chem Eng Prog 1949; 45:39. Henzler HJ. Chem Eng Tech 1980; 52:659–661. Zhu M. Proc. 1st International Conference on Process Intensification, Antwerp, Belgium, December 6–8, 1995, organized by BHR Group, Cranfield, UK, 51–59. Middleton JC. AIChE 71st Annual Meeting, Miami Beach, Paper 74E, 1978. Dutta NN, Raghavan KV. Chem Eng J 1987; 36:111–121. Hinze JO. AIChE J 1955; 1:289–295. Middleman S. Ind Eng Proc Des Dev 1974; 13:78. Berkmann PD, Calabrese RV. AIChE J 1988; 34(4). Streiff FA, Mathys P, Fischer TU. New fundamentals for liquid–liquid dispersion using static mixers. Récents Progrès en Génie des Procédés 1997; 11(51):307–314. Green A, Johnson B, Westall S, Bunegar M, Symonds K. Combined chemical reactor/ heat exchangers: validation and application in industrial processes. 4th International Conference in Process Intensification, Brugge, Belgium, September 2001.


8 Reactive and Hybrid Separations: Incentives, Applications, Barriers Andrzej Stankiewicz DSM Research, Geleen, The Netherlands

1.

INTRODUCTION

Integration of various steps/operations presents one of the most promising ways for intensifying (bio)chemical processes. It can be achieved either by combining reaction and separation in a single reactive separation step or by combining two (or more) separation techniques in a hybrid separation unit. Such an integration may bring a number of advantages to the process under consideration, not just a decrease in the size of equipment. This chapter provides a general overview of the reactive and hybrid separations and discusses their place in the intensification of (bio)chemical processes. Written from an industrial point of view, it focuses on the application aspects of those integrative technologies. Potential application fields are reported, along with already existing commercial-scale operations. Special attention is given to the barriers that hamper a broader introduction of the reactive and hybrid separations into industrial practice and the ways to overcome those barriers. The modeling and design aspects of three reactive separation methods (reactive distillation, reactive absorption, and reactive extraction) are discussed in more detail in Chapter 9.


2.

REACTIVE SEPARATIONS—WHY INTEGRATE?

In the simplest case, integration of reaction and separation may take place on the purely equipment level, without introducing any new functional interrelations between the operations involved—the reaction does not influence the separation, nor has the separation process any effect upon the reaction. The aimed result of such combination can be: Lower investment costs (compact plant layout, integral design) Smaller inventory (safety aspects) Improved heat management/energy utilization. The Urea 2000plus™ technology, developed by Stamicarbon B.V. (1) and described further in Chapter 12, presents a typical example of such a “noninterrelating” integration. The integration resulted here in a considerably smaller and cheaper plant, with much less high-pressure equipment/piping needed and less energy consumption. Yet the interrelations between the reaction and other operations remained basically the same as in the conventional technology. In most cases, however, the reaction and separation are integrated in order to benefit from the interaction effect between those two, for instance, To improve yield/selectivity (e.g., via equilibrium shift) To facilitate separation (e.g., azeotrope problems) For other reasons, e.g., to extend the catalyst lifetime One speaks in those cases about reactive separations or separative reactors. The industrially important reactive separations include: Reactive distillation Membrane-based reactive separations Reactive adsorption Reactive absorption Reactive extraction Reactive crystallization 2.1.

Reactive Distillation

In most industrial applications the reactive distillation is used to improve the yield/selectivity of the required product. Figure 1 shows three examples of industrial processes, in which combination of reaction and distillation shifts the equilibrium of the reaction A B ↔ C D in the required direction (2). The length of the reacting, distillation, and stripping zones as well as the positioning of the reactant inlets vary in each particular case, depending on the process requirements. On the other hand, in selective hydrogenations of dienes and aromatics (3), reactive distillation is used to remove the single-hydrogenated product from the reaction


FIGURE 1 Examples of industrial processes employing reactive distillation: (a) methyl tert-butyl ether (MTBE) from isobutene and methanol; (b) cumene via alkylation of benzene with propylene; (c) ethylene glycol via hydration of ethylene oxide.

zone, thus preventing its further hydrogenation and increasing its yield. In the methyl acetate technology of Eastman Chemical, integration of reactive distillation with extractive distillation in a single unit totally solved the azeotrope problem (4). Reactive distillation can also be used as a powerful separation method in case of mixtures containing reactive and inert components with close boiling points. The method is schematically depicted in Figure 2. Here, a reactive entrainer is introduced to the first reactive distillation column, to form an intermediate product having a boiling point much more distant from the boiling point of the inert components. In the first column, inert components are therefore easily separated, while the intermediate product is fed to the second reactive distillation step, where the reversed reaction takes place and the original reactive component is recovered and separated from the entrainer. Stein et al. (5) investigated the application of this principle to the separation of close-boiling i-butene and n-butene, using methanol as a reactive entrainer. Obviously, reactive distillation may lead to significant savings on energy. Hydrolysis of methyl acetate presents an industrial example of such energy savings.


The heat requirements of the reactive distillation-based process are ca. 50% lower than in the conventional technology. In the alkylation of benzene to cumene reactive distillation effectively eliminated the hot spots and reduced the oligomerization of propylene (6). Table 1 gives an overview of the possible applications of reactive distillation reported in the literature. Very few of them have been realized so far on the commercial scale. One of the common factors that hinders a broader application of reactive distillation is a small feasible operation window. The overlap region in the pressure–temperature domain, in which chemical reaction and separation and apparatus design are feasible, is usually quite narrow (see Figure 2 in Chapter 9). A possible remedy for this limitation is sought in the development of new types of catalysts that would allow one to significantly broaden the feasible operation window for chemical reaction. 2.2.

Membrane-Based Reactive Separations

Sirkar et al. (64) give an interesting overview of various functions that a membrane may play in a chemical reactor. Those functions are schematically shown in Figure 3 and summarized in Table 2.

FIGURE 2 Separation of reactive and inert components with close boiling points, facilitated by reactive distillation. (From Ref. 5.)


TABLE 1 Reported Existing and Potential Applications of Reactive Distillation Product/process Butyl acrylate from butanol and acrylic acid n-Butyl acetate from n-butanol and acetic acid Ethyl acetate from ethanol and acetic acid Methyl acetate from methanol and acetic acid Hydrolysis of methyl acetate 2-Methylpropylacetate from 2-methylpropanol and acetic acid Amyl acetate from amyl alcohol and acetic acid Ethyl pentenoate from ethanol and pentenoic acid Esterification of fatty acids Methylal from formaldehyde and methanol TAME (tertiary amyl ether) MTBE (methyl tert-butyl ether) ETBE (ethyl tert-butyl ether) from ethanol and isobutene ETBE (ethyl tert-butyl ether) from bioethanol and tert-butylalcohol Diisopropyl ether from propene TAA (tert-amyl alcohol) via hydration of isoamylene Isopropanol via hydration of propene Cyclohexanol via hydration of cyclohexene Phenol from cumene Ethylene glycol via hydration of ethylene oxide Isobutene via dehydration of tert-butanol Isoamylenes via dehydration of 2-methyl-1-butanol Isophorone from acetone MIBK (Methyl iso-butyl ketone) from acetone Diacetone alcohol (DAA) and mesityl oxide (MO) via aldol condensation of acetone Acetone via dehydrogenation of propanol Tetrahydrofuran from butanediol Xylenes via toluene disproportionation Hydrogenation of unsaturated hydrocarbons Isomerization of C5-C6 paraffins Isobutene via hydroisomerization of C4 alkenes Cumene via alkylation of benzene with propylene Ethylbenzene via alkylation of benzene with ethylene Cyclopentane and/or cyclopentene from dicyclopentadiene Purification of hydrofluorocarbons


Selected refs. 7 8 9,10 11,12 13 14 15 16 17 18,19 20–22 23,24 25,26 27 28 29 30 31,32 33 34,35 36–38 39 40 41,42 43,44 45 46 47 48,49 50 51 52 53,54 55 56

TABLE 1 (cont.) Product/process Naphtha desulfurization Dihydroxy polyether polyol via alkanolysis of corresponding diesters Glycine from glycinonitryle DEC (diethylcarbonate) via carbonylation of ethanol with dimethylcarbonate Polyamides (e.g., Nylon 6) via hydrolytic polymerization of amino nitriles Nylon 66 via polycondensation Propylene oxide from propylene chlorohydrin and calcium hydroxide

Selected refs. 57 58 59 60 61 62 63

Among all the membrane functions listed in Table 2, catalytic membranes probably attract the most attention. The scientific literature on catalytic membrane reactors is exceptionally rich and includes many interesting ideas, such as heat- and mass-integrated combination of hydrogenation and dehydrogenation processes in a single membrane unit. Yet practically no large-scale industrial applications of catalytic membrane reactors have been reported so far, perhaps with the exception of the Russian vitamin K technology (65). The primary reason for this is the relatively high price of membrane units, although other factors, such as low permeability, sealing problems, as well as mechanical and thermal fragileness of the membranes, also play an important role. Further developments in the field of material engineering will surely change this picture. Possible application areas of catalytic membrane reactors include: Dehydrogenations, e.g., ethane to ethene, ethylbenzene to styrene, methanol to formaldehyde Methane steam reforming Water–gas shift reaction Selective oxidations, e.g., propane to acroleine, butane to maleic anhydride, ethylene to ethylene oxide Oxidative dehydrogenations of hydrocarbons Oxidative coupling of methane Methane oxidation to syngas An excellent review of all potential applications of catalytic membrane reactors studied so far can be found in the 2002 book by Sanchez Marcano and Tsotsis (66). On the other hand, membranes are frequently employed in combination with a bioreactor, for instance, in enzymatic pharmaceutical processes. An example


of such an application of an ultrafiltration membrane-reactor system for the production of s-ibuprofen is discussed further in Chapter 12. 2.3.

Reactive Adsorption

The vast majority of possible applications of reactive adsorption aim at the improvement of the product yield by shifting the equilibrium in the required direction. In contrast to the nonreactive adsorption techniques, such as simulated moving beds and pressure-swing adsorption, and despite its great potential [for example, a 12-fold higher conversion per pass in oxidative methane coupling (67)], the

FIGURE 3 Membrane functions in chemical reactor. (Reproduced with permission from Sirkar et al. (64), copyright (1999) American Chemical Society.)


TABLE 2 Membrane Functions in Chemical Reactor Systems Function Separation of products from the reaction mixture

Separation of a reactant from a mixed stream for introduction into the reactor

Controlled addition of one reactant or two reactants

Nondispersive phase contacting, with reaction at the phase interface or in the bulk phases

Segregation of a catalyst (and cofactor) in a reactor Immobilization of a catalyst in (or on) a membrane

Membrane is the catalyst

Examples In situ product removal from enzymatic reactor via a nanofiltration or ultrafiltration membrane Removal of selected enantiomer via a liquid membrane Removal of water in esterification reactions via a pervaporation membrane Hydrogen removal in catalytic dehydrogenation reactions Separation of oxygen from air for oxidizing methane to syngas Separation of hydrogen from dehydrogenation reaction to oxidize it with oxygen on permeate side Separation of organic priority pollutants from wastewater for biological purification Controlled oxygen addition in partial oxidation reactions (to increase selectivity) Controlled air introduction in oxidative dehydrogenations Emulsion-free enzymatic splitting of fats Bubble-free oxygen/ozone supply in wastewater treatment via hollowfiber membranes Segregation of enzymes with respect to molecular weight on ultrafiltration membranes Immobilization of enzymes or cells on polymeric membranes Immobilization of metals (Pd, Pt) on ceramic membranes Cation exchange membranes for esterification reactions Palladium membranes for hydrogenation/dehydrogenation reactions


TABLE 2 (cont.) Function Membrane is the reactor

Solid electrolyte membrane supports the electrodes, conducts ions, and achieves the reactions on its surface Transfer of heat

Immobilizing the liquid reaction medium

Examples Reactions in flow-through membrane systems (“pore flow-through reactors”) Solid electrolyte membranes such as H and O2 conductors in fuel cells

Membranes coupling endo- and exothermic reaction zones (e.g., hydrogenation–dehydrogenation) Supported liquid membranes (SLM) for homogeneous catalytic processes

Source: Ref. 64.

industrial-scale applications of adsorptive reactors remain to be seen. Challenges involve materials development of catalysts/adsorbents and matching of process conditions (same temperature) for both reaction and adsorption so that high yields/ selectivities can be achieved. Reactive adsorption processes investigated in the bench or pilot scale are numerous, as shown in Table 3. One of the more promising types of adsorptive reactors is the so-called gas– solid–solid trickle-flow reactor (GSSTFR), in which fine adsorbent trickles through the fixed bed of catalyst (Figure 4), removing selectively in situ one or more of the products from the reaction zone. In the case of methanol synthesis this led to conversions significantly exceeding the equilibrium conversions under the given conditions (98). The economics of the methanol process based on the gas–solid– solid trickle-flow reactor was evaluated and compared with the conventional lowpressure Lurgi process (99). For the production scale of 1000 tons per day, the new technology offered considerable reductions in cooling water consumption (50%), recirculation energy (70%), raw materials (12%), and catalyst amount (70%). Further improvement of the GSSTFR concept could be seen in applying a moving bed of adsorbent through straight, parallel channels of a monolithic catalyst, similar to the one shown in Figure 23 of Chapter 6. 2.4.

Reactive Extraction

Similar to reactive adsorption, the reactive extraction can be applied primarily in multireaction systems, for improvement in yields and selectivities to desired products. The combination of reaction with liquid–liquid extraction can also be used


TABLE 3 Processes Investigated in Reactive Adsorption Systems Process

Selected refs.

Reactor type

Esterification of glycerin with acetic acid

68

Methyl tert-butyl ether (MTBE) synthesis Hydrolysis of methyl formate Oxidative coupling of methane Enzymatic production of L-amino acids Oxidation of phenols Ethyl acetate from ethanol and acetic acid Enzymatic inversion of sucrose

69,70 71 67,72 73 74 75 76 77

Dehydroisomerization of n-butane to isobutene Mesitylene hydrogenation Hydrogenation of 1,3,5-trimethylbenzene Biosynthesis of dextran polymer from sucrose Dissociation of dicyclopentadiene Dehydrogenation of cyclohexane

78

Simulated moving bed chromatographic reactor (SMBCR) SMBCR Discontinuous chromatographic reactor SMBCR Centrifugal partition chromatographic reactor Chromatographic reactor Chromatographic reactor SMBCR Rotating cylindrical annulus chromatographic reactor (RCACR) Chromatographic pulse reactor

72 79,80 81

SMBCR SMBCR Chromatographic pulse reactor

82,83 84 85 86 87

SMBCR Chromatographic pulse reactor RCACR SMBCR SMBCR

88

Batch and fixed-bed adsorptive reactors

Ascorbic acid synthesis Regioselective enzymatic diol esterification


TABLE 3 (cont.) Process Oxidation of lean VOC mixtures Hydrolysis of methyl formate Enzymatic production of lactosucrose from sucrose and lactose Diethylacetal from ethanol and acetaldehyde Steam methane reforming Propene metathesis to ethene and 2-butene 1-Butene dehydrogenation to 1,3-butadiene Sulfur from H2S (Claus process) HCN from carbon monoxide and ammonia 6-Aminopenicillanic acid from penicillin G Methanol synthesis


Selected refs.

Reactor type

89 90 91

SMBCR RCACR SMBCR

92

Fixed-bed adsorptive reactor

93 94

Pressure-swing adsorptive reactor (PSAR) PSAR

95

Rapid PSAR

96 96

Reverse-flow adsorptive reactor Reverse-flow adsorptive reactor

97

Trickle-flow fluidized-bed reactor

98,99

Gas–solid–solid trickle-flow reactor (GSSTFR)

FIGURE 4 Gas–solid–solid trickle-flow reactor. (From Ref. 98.)

for the separation of waste by-products that are hard to separate using conventional techniques (100,101). An overview of processes studied in reactive extraction systems is shown in Table 4. In 2002, an interesting concept was proposed for coupling a CO2-based supercritical extraction with air oxidation in order to remove and decompose pollutants from gases or liquids (134). An exemplary process scheme according to this preliminary concept is shown in Figure 5. Possible (future) environmental applications of such an integrated supercritical extraction-reaction system include treatment of liquid effluents, regeneration of catalysts and adsorption materials, and soil decontamination.


2.5.

Reactive Crystallization/Precipitation

Reactive crystallization/precipitation plays a role in a number of industrially relevant processes, such as liquid-phase oxidation of para-xylene to produce technical-grade terephthalic acid, the acidic hydrolysis of sodium salicylate to salicylic acid, and the absorption of ammonia in aqueous sulfuric acid to form ammonium sulfate (135). Reactive crystallization/precipitation is also widely applied in the pharmaceutical industry, to facilitate the resolution of the enantiomers (diastereomeric crystallization). Here, the racemate is reacted with a specific optically active material (resolving agent) to produce two diastereomeric derivatives (usually salts) that are easily separated by crystallization: ( DL )-A ( L )-〉 → ( D)-A ⋅ ( L )-〉 ( L )-A ⋅ ( L )-〉 racemate

resolving agent

n-salt

p-salt

TABLE 4 Some Processes Studied in Reactive Extraction Systems Product/process Penicillin G recovery Downstream separation of 1,3-propanediol Separation of lactic acid Separation of organic acids from the products of partial oxidation of paraffins Separation of salicylic acid Separation of D,L-phenylalanine Separation of citric acid Separation of aspartic acid Cephalosporin C recovery Separation of metals (e.g., zinc) Phenolic wastewater treatment Separation of dicarboxylic acids (e.g., oxalic, malonic, succinic, adipic acid) Recovery of gallium from coal fly ash Recovery of palladium, platinum, rodium from leaching solutions Fractionation of amino acids Recovery of 7-ACA (7-aminocephalosporanic acid) Recovery of erythromycin Removal of toxic heavy metals from wastewater streams Production of dioxolane from aldehyde Recovery of aldehydes and ketones from hydrocarbon mixtures Production of cyclic ester oligomers from linear polyesters


Selected refs. 102–106 107 108–111 112 113,114 113,115 116 117 118 119–121 122 123 124 125 126 127 128 129,130 131 132 133

FIGURE 5 Countercurrent supercritical extraction coupled with air oxidation to remove and decompose pollutants from gases and liquids. (From Ref. 134.)

Diastereomeric crystallization is commonly used in the production of a number of pharmaceuticals, such as ampicillin, ethambutol, chloramphenicol, diltiazem, fosfomycin, and naproxen (136). Somewhat similar are the so-called adductive crystallization processes, often (wrongly) called extractive crystallization, where reactions of complex/ adduct formation are used to separate compounds that are otherwise difficult to separate. Examples of adductive crystallization include separation of p- and mcresols (137), separation of o- and p-nitrochlorobenzenes (138), separation of quinaldine and isoquinoline (139), separation of nonaromatic compounds from naphtha-cracking raffinate (140), and separation of p-cresol from 2,6-xylenol (141). Other examples of reactive crystallization/precipitation reported in the literature are listed in Table 5. Reactive crystallization/precipitation can also be conducted in high-gravity (Higee) fields using rotating equipment. In China this technique has been used successfully for the production of nano-size particles of CaCO3. Ultrafine particles


with the mean size of 15–40 nm and a very narrow size distribution were produced by carbonation of a lime suspension in a rotating packed-bed reactor (RPBR) (158). The reaction times in RPBR were 4- to 10-fold shorter than the corresponding reaction times in a conventional stirred-tank unit. A similar technique was used for the production of nanofibrils of aluminum hydroxide with a diameter of 1–10 nm and 50–300 nm long as well as nanoparticles of SrCO3 with a mean size of 40 nm (159). 2.6.

Reactive Absorption/Stripping

Reactive absorption is very old as a processing technique and has been used for production purposes in a number of classical bulk-chemical technologies, such as nitric or sulfuric acid. The Raschig process for the production of hydroxylamine, an important intermediate in classical caprolactam technologies (Stamicarbon, Inventa), is also an example of a multistep reactive absorption process. Here, water, ammonia, and carbon dioxide react together in an absorption column to give a solution of ammonium carbonate, which subsequently forms an alkaline

TABLE 5 Examples of Reactive Crystallization/Precipitation Product

Selected refs.

Calcium carbonate

142,143

Methyl -methoximino acetoacetate Magnesium hydroxide Calcium phosphate Magnesium ammonium phosphate

144

Lead sulfate Magnesium carbonate Nickel hydroxide Ziprasidone–HClH2O

150 151 152 153

Barium carbonate

154

Boric acid

155

Procaine benzylpenicillin Sulfamic acid

156 157

145,146 147,148 149


Remarks Liquid–liquid and gas–liquid reaction systems

Removal of ammonium and phosphate ions from wastewater

Conducted in impinging fluid jet stream system To remove CO2 from waste gas Reaction of borax solution with solid oxalic acid From urea and fuming H2SO4

solution of ammonium nitrite by reactive absorption of nitrous oxide at low temperature. In a further step, the ammonium nitrite is converted to hydroxylamine disulfonate with sulfur dioxide. The hydroxylamine disulfonate solution is drawn off and the salt is hydrolyzed and neutralized to give hydroxylamine sulfate and ammonium sulfate as coproduct. Carbon dioxide removal by reactive absorption in amine solutions is also applied on the commercial scale, for instance, in the treatment of flue gas (see later in this chapter). Another possible application field of the technique is gas desulfurization, in which H2S is removed and converted to sulfur by means of reactive absorption. Aqueous solutions of ferric chelates (160–162) as well as tetramethylene sulfone, pyridine, quinoline, and polyglycol ether solutions of SO2 (163,164) have been proposed as solvents. Reactive absorption can also be used for NOx reduction and removal from flue or exhaust gases (165,166). The separation of light olefins and paraffins by means of a reversible chemical complexation of olefins with Ag(I) or Cu(I) compounds in aqueous and nonaqueous solutions is another very interesting example of reactive absorption, one that could possibly replace the conventional cryogenic distillation technology (167). 3.

HYBRID SEPARATIONS

Generally speaking, hybrid separations can be described as processing methods that integrate two or more different separation techniques in a single operation, making use of the synergy between them. The industrially most important (or promising) hybrid separations include: Extractive distillation Adsorptive distillation Membrane distillation Membrane absorption/stripping Adsorptive membranes (membrane chromatography) Membrane extraction 3.1.

Extractive Distillation

Extractive distillation is probably the oldest and most widely applied type of hybrid separation, particularly useful in close-boiling-point problems or in systems in which components form azeotropes. In the method, an extra component (solvent) is added to the system, which does not form azeotropes with feed components. The solvent alters the relative volatility of original feed components, allowing one to distill overhead. The solvent leaves the column with the bottom products and is separated in a binary column. Energy savings represent the most important advantage of extractive distillation over the conventional (nonhybrid) separation methods (168,169).


FIGURE 6 Scheme of aromatics separation via extractive distillation in the BTX process of the GTC Technology Corp.

Originally, extractive distillation was limited to two-component problems. However, recent developments in solvent technology enabled applications of this hybrid separation in multicomponent systems as well. An example of such application is the BTX process of the GTC Technology Corp., shown in Figure 6, in which extractive distillation replaced the conventional liquid–liquid extraction to separate aromatics from catalytic reformate or pyrolysis gasoline. This led to a ca. 25% lower capital cost and a ca. 15% decrease in energy consumption (170). Some other examples of existing and potential applications of the extractive distillations are listed in Table 6. Solvents used for extractive distillation vary considerably, but in almost all cases solvent selection presents a trade-off between its selectivity and solvency (194). The effectivity of the solvent can sometimes be improved by the addition of a salt (195). 3.2.

Adsorptive Distillation

Although considered by some authors a “novel process,” adsorptive distillation is a relatively old hybrid separation, originating in the early 1950s (196). It is a


TABLE 6 Reported (Potential) Applications of Extractive Distillation Separation process High-purity cyclohexane from petroleum Benzene and toluene from nonaromatics Isopropyl ether from acetone Methyl acetate from methanol Anhydrous ethanol from fermentation broth Ethyl acetate from ethanol/water Ternary acetate–alcohol–water systems (propyl, butyl, amyl, hexyl) m-Xylene from o-xylene MTBE from impurities Binary mixtures of lower-boiling alcohols Binary mixtures of phenolic compounds (chlorophenol, phenol, cresol, xylenol) Acetone from water Ethanol dehydration C2 alcohols from water Cyclohexane-cyclohexenebenzene MTBE from ethanol Methylcyclohexane from toluene Anhydrous ethanol recovery from wastewater streams Propylene from propane 1-Butene from 1,3-butadiene

Selected refs.

Remarks

171

Close-boiling-point problem

172,173


174 175

Azeotrope problem Azeotrope problem

176

Azeotrope problem

177

Azeotrope problem

178

Azeotrope problem

179 180 181

Close-boiling-point problem Close-boiling-point problem Close-boiling-point problem

182,183


184 185 186 187

Azeotrope problem Azeotrope problem Azeotrope problem Close-boiling-point problem

188 189

Azeotrope problem Close-boiling-point problem

190

Azeotrope problem

191 192,193

Close-boiling-point problem Close-boiling-point problem

three-phase mass transfer operation in which distillation is carried in presence of a solid selective adsorbent. The adsorbent usually consists of a fine powder (particle size in the 10-m range), fluidized and circulated by an inert carrier. The process is typically carried in two columns: an adsorptive distillation column for increasing separation ability, and a distillative desorption column for enhancing


the regeneration of the adsorbent. As in the case of extractive distillation, adsorptive distillation also can be used for the separation of mixtures containing closeboiling components or to bypass the azeotrope. Another interesting potential application field is the removal of trace impurities in the production of fine chemicals. The simplest scheme of an adsorptive distillation system for separating a binary mixture of azeotrope-forming components is shown in Figure 7. Here, adsorbent S carried by an inert fluid carrier enters the adsorptive distillation column, selectively adsorbs component B from the feed, and flows to the desorption (stripping) column, in which separation and enrichment of B takes place. Figure 8 shows another variant of adsorptive distillation, as proposed in a patent by Shell (197), for improved separation of closely-boiling hydrocarbon mixtures. Here an extra stripping medium (e.g., pentane) is used to remove the adsorbate in the stripping column. Despite an almost 50-year history, no large-scale commercial processes using adsorptive distillation have been reported so far. Some potential application fields for this hybrid separation are listed in Table 7.

FIGURE 7 Scheme of an adsorptive distillation system for the separation of azeotrope-forming components A and B (S–adsorbent).


FIGURE 8 Simplified scheme of an adsorptive distillation–based separation of closely boiling hydrocarbon mixtures. (From Ref. 197.)

3.3.

Membrane Distillation

Membrane distillation is considered a promising separation method applicable primarily in environmental technologies. In membrane distillation a microporous and hydrophobic membrane separates aqueous solutions at different temperatures and compositions, as shown in Figure 9. The temperature difference existing across the membrane results in a vapor pressure difference. The molecules are transported through the pores of the membrane from the high-vapor-pressure side to the lowvapor-pressure side. At least one side of the membrane remains in contact with the liquid phase. Benefits offered by membrane distillation include (202): 100% (theoretical) rejection of ions, macromolecules, colloids, cells, and other nonvolatiles Lower operating temperatures than conventional distillation Lower operating pressures than conventional pressure-driven membrane separation Reduced chemical interaction between membrane and process solutions Less demanding membrane mechanical property requirements Reduced vapor spaces compared to conventional distillation


Membrane distillation systems may be classified into four different categories (203): Direct contact membrane distillation (DCMD), in which the membrane is in direct contact with the liquid phase on both sides Air-gap membrane distillation (AGMD), in which an air gap is interposed between the membrane and the condensation surface Vacuum membrane distillation (VMD), in which the vapor phase is evacuated from the liquid through the membrane and the condensation takes place in a separate apparatus Sweeping-gas membrane distillation (SGMD), in which a stripping gas, instead of vacuum, is used as a carrier Currently, the most important application area for membrane distillation is water desalination technology. Figure 10 shows one of the water desalination processes developed by a Japanese organization, the Water Re-Use Promotion Center, in cooperation with Takenaka Corporation and Organo Corporation (204). The process uses solar energy and can therefore be installed at locations without an electricity supply. Other application areas for membrane distillation reported in the literature are summarized in Table 8. In 2002, the TNO Environment, Energy and Process Innovation institute in the Netherlands developed a membrane-based distillation concept that radically improves the economy and ecology of existing desalination technology for TABLE 7 Potential Application Fields of Adsorptive Distillation Reported in the Literature System Toluene–methylcyclohexane mixtures and other closely boiling hydrocarbons Naphtha reformate and other close-boiling hydrocarbons o-Xylene–m-xylene mixture Ethanol–water and ethyl acetate–water–n-butanol mixtures p-Xylene–m-xylene mixture

Selected refs. 196

197

198 199,200

201


Remarks Suitable adsorbents: silica gel, activated carbon, coconut charcoal, bauxite, activated alumina Silica gel as adsorbent

Modeling study Zeolite (4A molecular sieve) as adsorbent, glycol as carrier Zeolite (NaY molecular sieve) as adsorbent, n-decane as carrier

FIGURE 9 Scheme of a membrane distillation process.

seawater and brackish water. This so-called Memstill® technology (Figure 11a) combines multistage flash and multieffect distillation modes into one membrane module. Since the Memstill® module houses a continuum of evaporation stages in an almost ideal countercurrent flow configuration, a very high recovery of the evaporation heat is possible. The economic advantage of the Memstill® technology, compared to the “classical” desalination techniques, is shown in Figure 11b. An academic-industrial consortium is currently developing and improving the Memstill® process concept and module design (226). The same TNO Institute has also developed a concept of another membrane-based distillation technology for fractionation of non-water-based systems (227). The technology, called MEMFRAC, offers high energy efficiency in compact equipment. The study carried out for fractionation of benzene from toluene showed that with MEMFRAC technology a HETP between 5 and 10 cm could be obtained. Additional advantages of the MEMFRAC technology include: lack of entrainment, flooding, foaming, or channeling (due to indirect gas–liquid contact), independent gas/liquid control, and the possibility for modular plant design. Such a modular MEMFRAC distillation unit is schematically presented in Figure 12. On the other hand, a pervaporation membrane can be coupled with a conventional distillation column, resulting in a hybrid membrane/distillation process (228,229). Some of the investigated applications of such hybrid pervaporation membrane/distillation systems are shown in Table 9. In hybrid pervaporation/ distillation systems, the membrane units can be installed on the overhead vapor of the distillation column, as shown in Figure 13a for the case of propylene/ propane splitting (234), or they can be installed on the feed to the distillation column,


FIGURE 10 Scheme of the demonstration test plant for water desalination using solar energy and membrane distillation. (Courtesy: CADDET, Center for Renewable Energy, Harwell, UK).


TABLE 8 Application Areas of Membrane Distillation System/process

Selected refs.

Concentration of H2SO4, H3PO4, NaOH, HNO3, and HCl solutions Concentration of 2,3-butanediol from fermentation broths Wastewater treatment in the textile industry

205–207

Radioactive wastewater treatment

210

Removal of benzene traces from water Concentration of protein solutions Removal of halogenated VOCs Concentration of oil–water emulsions Concentration of sugar/sucrose solutions Separation of water and glycols Ethanol–water separation Acetone and ethanol removal from aqueous solutions Propanone removal from aqueous streams Acetone–butanol–ethanol (ABE) solvent recovery Fermentative ethanol production

211

Concentration and purification of fluosilicic acid Removal of trichloroethylene

Remarks

208

VMD process

209

Integrated reverse osmosis/membrane distillation process Integrated reverse osmosis/membrane distillation process VMD process

212 213 214 215,216

VMD process

217 218 219

DCMD and AGMD processes DCMD process AGMD process AGMD process

220

AGMD process

221

AGMD process

222,223

Integration of MD in fermentation resulted in ca. 2 increase of production rate

224 225


VMD process

FIGURE 11 Memstill® technology of seawater desalination developed at the TNO institute: (a) principle of the process; (b) cost comparison with other desalination techniques. (From Ref. 226.)


FIGURE 12 Scheme of a modular MEMFRAC distillation unit for fractionation of non-water-based systems, developed at the TNO Institute. (From Ref. 227.)

as shown in Figure 13b for the case of aromatic/aliphatic hydrocarbon separation (235). Shortcut design methods for hybrid pervaporation/distillation processes can be found in Ref. 236. 3.4.

Membrane Absorption/Stripping

Membrane absorption is one of the processes that Mother Nature had invented long before engineers did. Human lungs and intestines present perfect examples of membrane absorption systems. In the simplest case a gaseous component is selectively transported via a membrane and dissolved in the absorbing liquid, as shown in Figure 14. It is also possible to carry a membrane-based absorption-desorption process, with two liquids on both sides of the membrane (237), or a membrane stripping process, in which selected components are removed from the liquid phase through the membrane by a stripping gas (238). An important characteristic feature


of membrane absorption is that it proceeds without creating a real gas–liquid interface in the form of bubbles. Such a bubbleless gas–liquid mass transfer can be of advantage in certain processes, for instance, in shear-sensitive biological systems. One of the most important application areas of the membrane absorption is the capture of CO2 from flue gas. Kværner has recently developed a membrane absorption–based technology for the removal of CO2 from turbine exhaust gases in offshore applications (239). The process, based on membrane-facilitated CO2 absorption in amine, followed by membrane-facilitated stripping with steam, is schematically shown in Figure 15. The expected cost reduction, in comparison with a conventional amine separation process, ranges between 30 and 40%, for both investment and operating costs. The new membrane-based process also offers a very significant reduction in the weight and size of equipment (70–75% and 65%, respectively; see Figure 16), a great advantage in the case of offshore technology. Some other possible applications of membrane absorption/stripping are shown in Table 10. 3.5.

Membrane Chromatography (Adsorptive Membranes)

Membrane chromatography is a separation technique used almost exclusively in the downstream processing of proteins. Traditionally, most chromatographic purification steps in the downstream processing of proteins take place in columns packed with bead-shaped particles. Membrane chromatography presents a hybrid combination of liquid chromatography and membrane filtration based on microporous or macroporous membranes that contain functional ligands attached to their inner pore structure, which act as adsorbents. The main feature and advantage of this technique, compared to the conventional ones, is the absence of pore TABLE 9 Possible Applications of Hybrid Pervaporation/Distillation Systems System/process investigated

Selected refs.

Benzene–cyclohexane separation

230

Ethanol dehydration

231

Propylene/propane splitting Propylene/propane splitting

232 233


Remarks/effects Combination of extractive distillation and one-stage pervaporation; high-purity (99.2–99.5%) products; estimated cost savings of 20% Simulation study; 50% cost reduction in comparison with conventional azeotropic distillation Pilot-plant studies; 20–50% savings on operating costs 26–30% savings on capital investment

FIGURE 13 Combined distillation/pervaporation systems for (a) propylene/ propane splitting and (b) aromatic/aliphatic hydrocarbon separation. (Part a from Ref. 234; part b from Ref. 235.)


FIGURE 14 Membrane absorption.

diffusion, which is the main transport resistance in traditional chromatography. Dissolved molecules are carried directly to the adsorptive sites in the membranes by convective flow (Figure 17), which increases the throughput of the process. Membrane chromatography presents a process-intensive option for the protein A, G, or L affinity chromatography (247–253), as well as for metal affinity, ionexchange, hydrophobic interaction or reversed-phase chromatography (254–258). In recent years some new potential application fields for membrane chromatography have been demonstrated. Those are listed in Table 11.

FIGURE 15 Membrane absorption–based technology for removal of CO2 from turbine exhaust gases, developed by Kværner Process Systems. (Courtesy: Kværner.)


FIGURE 16 Plant-size reduction in Kværner’s technology for CO2 removal from exhaust gases: (a) conventional process; (b) membrane absorption process. (Courtesy: Kværner.)

3.6.

Membrane Extraction

In membrane extraction, the treated solution and the extractant/solvent are separated from each other by means of a solid or liquid membrane. The technique is applied primarily in three areas: wastewater treatment (e.g., removal of pollutants or recovery of trace components), biotechnology (e.g., removal of products from fermentation broths or separation of enantiomers), and analytical chemistry (e.g., online monitoring of pollutant concentrations in wastewater). Figure 18a shows schematically an industrial hollow fiber–based pertraction unit for water treatment, according to the TNO technology (263). The unit can be integrated with a film evaporator to enable the release of pollutants in pure form (Figure 18b).


Other more promising possible applications of membrane extraction, reported in the literature, are listed in Table 12. 3.7.

Other Hybrid Separations

In extractive crystallization (nonadductive), the driving force for the separation process is created by altering the solid–liquid phase relationships via the addition of a third component (usually liquid solvent) to the system. The solvent is chosen in such a way that it binds strongly at the crystallization temperature but separates easily at another temperature, where it is usually regenerated via distillation. Examples of such defined extractive distillations include separation of m- and pcresols using acetic acid as the solvent (297), separation of o- and p-nitrochlorobenzenes using p-dichlorobenzene (298), separation of lithium sulfate and lithium formate using n-butanol or 2-propanol (299), and separation of p-xylene from m-xylene using pentene (300). TABLE 10 Reported Possible Applications of Membrane Absorption/ Stripping System/process investigated Ammonia absorption/ desorption from ammonia water CO2 and/or SO2 removal

Selected refs. 240

241

Cyanide recovery from wastewater

242

H2S removal from gas streams

243

CO2 production for the horticultural industry (greenhouses) H2S and SO2 removal

244

VOC removal from wastewater

245

246


Remarks/effects Pilot-plant study in a polypropylene hollow-fiber column; ammonia is absorbed in diluted sulfuric acid Absorption in NaOH, K2CO3, alkanolamines, and Na2SO3 using hydrophobic microporous hollowfiber modules Recovery via a gas-filled membrane (GFM) placed between the wastewater and a chemical stripping solution Asymmetric hollow-fiber modules coupled with concentrated alkaline solution Possible energy saving of more than 30% reported Polyvinylidene fluoride (PVDF) hollow fibers and concentrated NaOH solution used Air stripping process via a polypropylene hollow-fiber module

FIGURE 17 Comparison of transport mechanisms in (a) conventional chromatography and in (b) membrane chromatography.

Smith, Bryson, and Valsaray (301,302) investigated solvent sublation, an adsorptive bubble process, combining the transport mechanisms of liquid/liquid extraction and gas stripping. The technique exploits the surface-active nature of organic compounds in their removal from water systems. The mechanism of solvent sublation is shown in Figure 19a. Gas bubbles are used to transport adsorbed solute from the bulk solution to the solvent layer. Solvent sublation is particularly promising in the removal of (hydrophobic) organic compounds from wastewater streams. An exemplary process scheme is shown in Figure 19b. An important advantage of the method is that the intimate contact between the extracting solvent and the wastewater is prevented (no problem of residual solvent in the treated water). Zeitsch (303) conducted a preliminary research on the removal of acetic acid from the vapor stream of furfural reactors by means of extractive condensation. It is a hybrid vapor-phase extraction process, in which solvent (triethylamine, TEA) forms a high-boiling complex with acetic acid. As a result, a “fog”


fraction is formed that can be separated via a coalescence filter or an electrostatic separator. The technique is reported to be highly selective. In 2002, Drioli and coworkers (304) investigated a process for obtaining protein crystals by means of membrane crystallization, which actually combines membrane distillation and crystallization techniques. The solvent evaporates at the membrane interface, migrates through the pores of the membrane, and condenses on the opposite side of the membrane. The reported preliminary results indicate interesting potentialities of this new method with respect to macromolecular crystallization. 4.

BARRIERS AND FUTURE PROSPECTS

Despite many ongoing research activities in the field and a number of successful commercializations, there still exist numerous technical and nontechnical barriers that hinder a wider introduction of reactive and hybrid separations into industrial practice. Two workshops held in 1998 by the Center for Waste Reduction Technologies of AIChE (305) identified some of the barriers for reactive separations and divided them into three categories: a. Technical gaps, such as lack of simulation and scale-up capability, lack of validated thermodynamic and kinetic data, lack of materials (e.g., TABLE 11 Reported New Potential Applications of Membrane Chromatography System/process investigated

Selected refs.

Separation of polynucleotides

259

Separation of oligonucleotides and peptides Separation of small hydrophobic molecules

260 260

Enantiomeric separation

261

Separation of trace metals

262


Remarks/effects Supercoiled plasmid DNA investigated as model

Benzene, toluene, homologues of 4-hydroxybenzoate investigated Racemic mixtures of tryptophan and thiophenal investigated in microfluidicbased membrane chromatography La-Ce-Pr-Nd-Sm separation and Zr-Hf separation investigated

FIGURE 18 Pertraction technology for wastewater treatment from the TNO Institute: (a) scheme of the hollow-fiber pertraction unit; (b) integration of pertraction with film evaporation. (Courtesy: TNO.)


System/process investigated

Selected refs.

Separation of acetic acid from aqueous solution

264

Separation of (S)-naproxen from racemic naproxen thioesters Separation of D,L-alanine and D,L-phenylalanine racemic mixtures Removal of sulfanilic acid from wastewater Lactic acid purification and concentration Enrichment of bisphenol A Phenol recovery from aqueous solutions Zinc(II) recovery from HCl solution Hydrogen separation from methane steam conversion products Separation of liquid olefin/paraffin mixtures Removal of 2-chlorophenol Ethanol removal from aqueous solutions Separation of cephalosporin C from fermentation broth Separation of penicillin G from aqueous streams Enrichment of amino acids Separation of cephalexin from a mixture of 7-ADCA Separation of butyric acid from fermentation broth

265

R

TABLE 12 Reported Potential Applications of Membrane Extraction Remarks

266,267

Microporous polypropylene membrane, MIBK as solvent Reactive extraction via hollow-fiber membrane Hollow-fiber zeolite membrane

268 269 270 271,272 273 274

Hollow-fiber modules Emulsion liquid membranes (ELMs) Liquid membrane Various liquid and solid membranes Bulk liquid and hollow-fiber membranes Palladium alloy membranes

275 276 277 278

Nonporous polymeric membranes Liquid membrane from aqueous solutions Microporous polypropylene membrane Bulk and emulsion liquid membranes

279

Supported liquid membrane (Amberlite LA-2) Supported liquid membrane (Aliquat 336) Supported liquid membrane (Aliquat 336) Liquid membrane

280 281 282

(continued)


TABLE 12 (cont.) System/process investigated

Selected refs.

Remarks

Separation of propionic and acetic acid from fermentation broth Separation of citric acid from fermentation broth Separation of lactic acid from fermentation broth Production of acetone, butanol, and ethanol (ABE) from potato wastes Separation of long-chain unsaturated fatty acids

283

Polymeric membranes

284 285 286

Liquid membrane Emulsion liquid membrane Polypropylene membrane

287

(Heavy) metals recovery from wastewater Removal of organic contaminants from wastewater

288–292 293–296

Microporous membrane, MeCN and n-heptane as solvents Various membranes Various membranes


FIGURE 19 Solvent sublation: (a) process mechanism and (b) an exemplary process scheme. (From Refs. 301 and 302.)


integrated catalysts/sorbents, membrane materials), and lack of highlevel process synthesis methodology b. Technology transfer barriers, such as lack of multidisciplinary team approaches to process integration, lack of commonality of problems (technology is application-specific) and lack of demonstrations/ prototypes on a reasonable scale (reactive and hybrid separations are still regarded more as a science than a technology) c. General barriers, such as higher standards, to which new technologies must be held, compared to conventional technologies, lack of information on process economics (early economic and process evaluation), and fear of risk in using new technologies. Most of these barriers also hold for hybrid separations. Two more factors that clearly play a hindering role in the commercial application of many reactive and hybrid separations are: the already-mentioned small feasible operation windows and the reduction of the degree of freedom caused by the integration of reaction and separation or by the coupling of two separations in one processing unit. Figure 20 shows an example of how the integration of reaction and membrane separation reduces the degree of freedom in a membrane reactor, resulting in decreased operational flexibility (306).

FIGURE 20 Degree of integration versus degree of freedom, in an example of a membrane reactor. (From Ref. 306.)


Despite the existing barriers, the coming years are expected to bring a significant increase in the number of industrially applied reactive and hybrid separation technologies. In particular, progress can be expected in the application of reactive distillation, reactive adsorption, and membrane-based operations. In hybrid separations, expansion of research activities on new product/process areas has already been seen. Reactive and hybrid separations have enormous potential for process intensification. Making full use of that potential will lead to substantially smaller, cleaner, and more energy-efficient chemical and biochemical plants.

REFERENCES 1. A low-cost design for urea. Nitrogen 1996; (222):29–31. 2. Taylor R, Krishna R. Chem Eng Sci 2000; 55:5183–5529. 3. Gildert GR, Rock K, McGuirk T. In: Proceedings of the International Symposium on Large Chemical Plants 10. Advances in process technology through catalytic distillation. Antwerp, 1998:103–113. 4. Siirola JJ. AIChE Symp Ser 1995; 91(304):222–233. 5. Stein E, Kienle A, Sundmacher K. Chem Eng 2000; 107(13):68–72. 6. Podrebarac GG, Ng FTT, Rempel GL. CHEMTECH 1997; 27(5):37–45. 7. Schwarzer S, Hoffmann U. Chem Eng Technol 2002; 25:975–980. 8. Steinigeweg S, Gmehling J. Ind Eng Chem Res 2002; 41:5483–5490. 9. Georgiadis MC, Schenk M, Gani R, Pistikopoulos RN. Computer-Aided Chem Eng 2001; 9:997–1002. 10. Kenig EY, Bader H, Górak A, Bessling B, Adrian T, Schoenmakers H. Chem Eng Sci 2001; 56:6185–6193. 11. Siirola JJ. AIChE Symp Ser 1995; 91(304):222–233. 12. Moritz P, Blagov S, Hasse H. Chem Ing Tech 2002; 74:1207–1218. 13. von Scala C, Gotze L, Moritz P. Sulzer Tech Rev 2001; 83(3):12–15. 14. Smejkal Q, Hanika J, Kolena J. Chem Eng Sci 2001; 56:365–370. 15. Chiang S-F, Kuo C-L, Yu C-C, Wong DSH. Ind Eng Chem Res 2002; 41:3233–3246. 16. Bollyn M, Wright AR. Chem Eng World 1999; 34(6):103–108. 17. Omota F, Dimian AC, Bliek A. Computer-Aided Chem Eng 2001; 9:9463–468. 18. Masamo J, Matsuzaki K. J Chem Eng Japan 1994; 27:1–5. 19. Kolah AK, Mahajani SM, Sharma MM. Ind Eng Chem Res 1996; 35:3707–3720. 20. Ward DJ. Etherification of isoamylenes by catalytic distillation. U.S. Patent 5196612, UOP Inc., 1993. 21. Baur R, Taylor R, Krishna R. Computer-Aided Chem Eng 2001; 9:93–98. 22. Baur R, Krishna R. Chem Eng Proc 2002; 41:445–462. 23. Yeh A-I, Berg L. Chem Eng Commun 1992; 113:147–153. 24. Preston KL. Use of reactive distillation in the manufacture of methyl tertiary-butyl ether. U.S. Patent 5741953, Huntsman Specialty Chemicals Corp., 1998. 25. Tian Y-C, Zhao F, Bisowarno BH, Tade MO. J Proc Control 2002; 13:57–67. 26. Al-Arfaj MA, Luyben WL. Ind Eng Chem Res 2002; 41:3784–3796.


27. Quitain AT, Itoh H, Goto S. In: Abraham MA, Hesketh RP, eds. Reaction Engineering for Pollution Prevention. Amsterdam: Elsevier Science, 2000:237–246. 28. Marker TL, Funk GA, Barger PT, Hammershaimb HU. Two-stage process for producing diisopropyl ether using catalytic distillation. U.S. Patent 5,744,645, UOP Inc., 1998. 29. Castor Gonzales J, Subawalla H, Fair JR. Ind Eng Chem Res 1997; 36:3845–3853. 30. Xu Y, Chuang KT, Sanger AR. Chem Eng Res Des 80:686–694, 2002. 31. Watanabe R, Nakanishi A. Production of cyclohexanol from cyclohexene by reactive distillation. JP 10287600, Mitsubishi Chemical Industries Ltd., 1998. 32. Steyer F, Qi Z, Sundmacher K. Chem Eng Sci 2002; 57:1511–1520. 33. Levin D, Santiesteban JG. Production of phenol using reactive distillation. WO 0146102, Mobil Oil Corp., 2001. 34. Kumar A, Daoutidis P. AIChE J 1999; 45:51–68. 35. Monroy-Loperena R, Perez-Cisneros E, Alvarez-Ramirez J. Chem Eng Sci 2000; 55:4925–4937. 36. Knifton JF, Sanderson JR, Stockton ME. The use of reactive distillation in the catalytic dehydration of teriary-butyl alcohol to produce anhydrous isobutylene. U.S. Patent 5,811,620, Huntsman Specialty Chemicals Corp., 1998. 37. Knifton JF, Sanderson JR, Stockton ME. Catal Lett 2001; 73:55–57. 38. Abella LC, Gaspillo P-AD, Itoh H, Goto S. J Chem Eng Japan 1999; 32:742–746. 39. Aiouache F, Goto S. J Chem Eng Japan 2002; 35:443–449. 40. Yoshida Y. Preparation of isophorone from acetone by using a reactive distillation column. JP 08245486, Daicel Chem, 1995. 41. Lawson KH, Nkosi B. Production of MIBK from acetone and hydrogen using catalytic distillation technology. U.S. Patent 6,008,416, Catalytic Distillation Technologies, 1999. 42. Saayman N, Lund GJ, Kindermans S. Process for the production of MIBK using catalytic distillation technology. WO 02022542, Catalytic Distillation Technologies, 2002. 43. Huang C, Yang L, Ng FTT, Rempel GL. Chem Eng Sci 1998; 53:3489–3499. 44. Podrebarac GG, Ng FTT, Rempel GL. Chem Eng Sci 1998; 53:1067–1075. 45. Gaspilla P-AD, Abella LC, Goto S. J Chem Eng Japan 1998; 31:440–444. 46. Liu Q, Zhang F, Gao H. Huaxue Gongcheng (Xi’an China) 2002; 30(2):75–78. 47. Stitt EH. Chem Eng Sci 2002; 57:1537–1543. 48. Maraschino MJ. Apparatus and process for the catalytic distillation of unsaturated, nonaromatic C3–5 hydrocarbons. U.S. Patent 5,847,249, Catalytic Distillation Technologies, 1998. 49. Stanley SJ, Sumner C. Catalytic distillation and hydrogenation of heavy unsaturates in an olefins plant in the manufacture of ethylene and propylene. WO 9909118, ABB Lummus Global Inc., 1999. 50. Lebas E, Jullian S, Travers C, Capron P, Joly J-F, Thery M. Process for the isomerization of paraffins by reactive distillation. EP 787786, Institut Français du Pétrole, 1997. 51. Dorbon M, Chodorge JA, Cosyns JA, Viltard JC, Didillon B. Method for producing high-purity isobutene through hydroisomerization reactive distillation and skeletal isomerization of c4 alkenes. FR 2757506, Institut Français du Pétrole, 1998. 52. Shoemaker JD, Jones EM Jr. Hydrocarbon Processing 1987; 66(6):57–58. 53. Smith LA Jr, Arganbright RP, Hearn D. Preparation of ethylbenzene in a catalytic distillation column reactor. U.S. Patent 5,476,978, Chemical Research and Licensing Company, 1995.


54. Zhang J, Li D, Fu J, Cao G. Process and apparatus for preparation of ethylbenzene by alkylation of benzene with dilute ethylene contained in dry gas by catalytic distillation. U.S. Patent Appl. 2001018545, 2001. 55. Silvergerg SE, Lattner JR, Sanchez LE. Use of hydrogenative catalytic distillation to produce cyclopentane and/or cyclopentene from dicyclopentadiene. WO 0029358, Exxon Chemical Patents Inc., 2000. 56. Cassel WR, Wuttke KG. Purification of hydrofluorocarbons by reactive distillation. WO 9640606, E.I. Du Pont de Nemours and Company, 1996. 57. Podrebarac GG, Gildert GR. Catalytic distillation for desulfurization of full-range naphtha by thioetherification and hydrodesulfurization. WO 0266580, Catalytic Distillation Technologies, 2002. 58. Dorai S. Alkanolysis of polyether polyol esters by reactive distillation. WO 9723559, E.I. Du Pont de Nemours and Company, 1997. 59. Aoki T, Otsubo K. Reactive distillation for manufacturing of glycine from glycinonitrile. JP 2001258586, Asahi Chemical Industry Co., Ltd., 2001. 60. Luo H-P, Xiao W-D. Chem Eng Sci 2001; 56:403–410. 61. Leemann M, Hildebrandt V, Thiele H, Espig S. Production of polyamides by reactive distillation. WO 9943732, BASF AG, 1999. 62. Grosser JH, Doherty MF, Malone MF. Ind Eng Chem Res 1987; 26:983–989. 63. Bezzo F, Bertucco A, Forlin A, Barolo M. Sep Pur Technol 1999; 16:251–260. 64. Sirkar KK, Shanbhag PV, Kovvali AS. Ind Eng Chem Res 1999; 38:3715–3737. 65. Gryaznov VM, Smirnov VS. Kinet Katal 1977; 18:579–580. 66. Sanchez Marcano JG, Tsotsis TT. Catalytic Membranes and Membrane Reactors. Weinheim: Wiley-VCH, 2002. 67. Bjorklund MC, Carr RW. Catal Today 1995; 25:159–168. 68. Carvoli G, Gelosa D, Morbidelli M, Ranghino G, Ruggieri R, Tricella A. Process for the esterification of alcohols in a chromatographic reactor. EP 1213278, La Chemial Spa, 2002. 69. Zhang Z, Hidajat K, Ray AK. Ind Eng Chem Res 2002; 41:3213–3232. 70. Zhang Z, Hidajat K, Ray AK. Ind Eng Chem Res 2001; 40:5305–5316. 71. Falk T, Seidel-Morgenstern A. Chem Eng Sci 2002; 57:1599–1606. 72. Bjorklund MC, Kruglov AV, Carr RW. Ind Eng Chem Res 2001; 40:2236–2242. 73. den Hollander JL, Wong YW, Luyben KC, van der Wielen LAM. Chem Eng Sci 1999; 54:3207–3215. 74. Deng Z, Dieckmann GR, Langer SH. J Chem Soc Perkin Trans 2: Phys Org Chem 1998; (5):1123–1128. 75. Mazzotti M, Neri B, Gelosa D, Morbidelli M. Ind Eng Chem Res 1997; 36:3163–3172. 76. Meurer M, Altenhoener U, Strube J, Untiedt A, Schmidt-Traub H. Starch/Staerke 1996; 48:452–458. 77. Herbstfhofer R, Bart H-J, Prior A, Wolfgang J. Recents Prog Genie Procedes 1999; 13(68):195–203. 78. Sad MR, Querini CA, Cornelli RA, Figoli NS, Parera JM. Appl Catal A General 1996; 146:131–143. 79. Ray AK, Carr RW, Aris R. Chem Eng Sci 1994; 49:469–480. 80. Ray AK, Carr RW. Chem Eng Sci 1995; 50:2195–2202. 81. Zafar I, Barker PE. Chem Eng Sci 1988; 43:2369–2375.


82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112.

113. 114. 115. 116.

Langer SH, Patton JE. J Phys Chem 1972; 76:2159–2170. Griffith TD, Chu AH, Langer SH. Chem Eng J 1987; 36:73–91. Schweich D, Villermaux J. Ind Eng Chem Fundam 1982; 21:47–51. Wardewell AW, Carr RW, Aris R. In: Wei J, Georgakis C, eds. Chemical Reaction Engineering. Boston: ACS Symposium Series Vol. 196, 1982:297–306. Arumugam BK, Perri ST, MacKenzie EB, Blair LW, Zoeller JR. Process for the preparation of ascorbic acid. U.S. Patent 6,476,239, Eastman Chemical Co., 2002. Meissner JP, Carta G. Ind Eng Chem Res 2002; 41:4722–4732. Migliorini C, Meissner JP, Mazzotti M, Carta G. Biotechnol Prog 2000; 16:600–609. Fissor D, Barresi AA. Chem Eng Technol 2002; 25:421–426. Cho BK, Carr RW, Aris R. Sep Sci Technol 1980; 15:679–696. Kawase M, Pilgrim A, Araki T, Hashimoto K. Chem Eng Sci 2001; 56:453–458. Silva VMTM, Rodrigues AE. AIChE J 2002; 48:625–634. Xiu G-H, Li P, Rodrigues AE. Chem Eng Sci 2002; 57:3893–3908. Gomes VG, Yee KWK. Chem Eng Sci 2002; 57:3839–3850. Sheikh J, Kershenbaum LS, Alpay E. Hung J Ind Chem 1999; 1:52–55. Elsner MP, Dittrich C, Agar DW. Chem Eng Sci 2002; 57:1607–1619. van der Wielen LAM, Diepen PJ, Houwers J, Luyben KCAM. Chem Eng Sci 1996; 51:2315–2325. Kuczynski M. The synthesis of methanol in a gas–solid–solid trickle-flow reactor. Ph.D. dissertation, University of Twente, Enschede, Netherlands, 1987. Westerterp KR, Bodewes TN, Vrijland MS, Kuczynski M. Hydrocarbon Processing 1988; 67:69–73. Minotti M, Doherty M, Malone MF. Ind Eng Chem Res 1998; 37:4746–4755. Samant KD, Ng KM. Chem Eng Technol 1999; 22:877–880. Patnaik PR. J Biotechnol 1992; 23:95–101. Yang C, Cussler EL. Biotechnol Bioeng 2000; 69:66–73. Likidis Z, Schügerl K. Biotechnol Lett 1987; 9:229–232. Müller B, Schlichting E, Bischoff L, Schügerl K. Appl Microbiol Biotechnol 1987; 26:206–210. Reschke M, Schügerl K. Chem Eng J 1984; 28:B25–B29. Malinowski JJ. Biotechnol Prog 2000; 16:76–79. Jarvinen M, Myllykoski L, Keiski R, Sohlo J. Bioseparation 2000; 9:163–166. Wasewar KL, Heesink ABM, Versteeg GF, Pangarkar VG. J Biotechnol 2002; 97: 59–68. Nurcan T, Bayraktar E, Mahmetoglu Ü. J Chem Technol Biotechnol 2001; 76:764–768. Pai RA, Doherty MF, Malone MF. AIChE J 2002; 48:514–526. Crandall JW, Grimm RC. Separation of organic acids from the products of partial oxidation of paraffins by reactive extraction with amines. U.S. Patent 3,541,121, Union Carbide Corp., 1970. Schlichting E, Halwachs W, Schügerl K. Chem Eng Proc 1985; 19:317–328. Haensel R, Halwachs W, Schügerl K. Chem Eng Sci 1986; 41:135–141. Haensel R, Halwachs W, Schügerl K. Chem Eng Sci 1986; 41:1811–1815. Bauer U, Marr R, Rueckl W, Siebenhofer M. Berichte der Bunsen-Gesellschaft 1989; 93:980–984.


117. 118. 119. 120. 121. 122.

123. 124. 125. 126. 127. 128. 129. 130. 131.

132.

133. 134. 135. 136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148. 149.

Aliwarga L, Schügerl K. DECHEMA Biotechnol Conf 1988; 1:483–488. Hano T, Matsumoto M, Ohtake T, Hori F. J Chem Eng Japan 1992; 25:293–297. Bart HJ, Maier S, Marr R, Weiss S. Chem Tech (Leipzig) 1993; 45:107–115. Bart HJ. Reactive Extraction. Berlin: Springer-Verlag, 2001. Bart HJ. Chem Eng Sci 2002; 57:1633–1637. Verhoeven W, Sluyts D, Denecker G, van Osselaer T, Hinz J, Vaes J, van Herck W, de Vos S. Treatment of phenolic wastewaters from manufacture of polycarbonates, bisphenols or diphenyl carbonate by reactive extraction. DE 19510063, Bayer Antwerpen N.V., 1996. Cascaval D, Tudose R, Oniscu C. Hung J Ind Chem 1997; 25:245–248. Gutierrez B, Pazos C, Coca J. Waste Manag Res 1997; 15:371–382. Nowottny C, Halwachs W, Schügerl K. Sep Purif Technol 1997; 12:135–144. Cascaval D, Oniscu C, Galaction A-I. Biochem Eng J 2001; 7:171–176. Bora M, Dutta NN, Bhattacharya KG. Chem Eng Commun 2000; 179:15–34. Oniscu C, Cascaval D, Galaction A-I, Dumitru IF. Roumanian Biotechnol Lett 2000; 5:439–447. Kueke F. Procedure for the reactive extraction of chromium-containing substances from aqueous feed solutions. DE 19943232, 2001. Elshani S, Smart NG, Lin Y, Wai CM. Sep Sci Technol 2001; 36:1197–1210. Grinbaum B, Adda M, Eidelman C. In: Cox M, Hidalgo M, Valiente M, eds. Solvent Extraction for the 21st Century, Proc ISEC ’99, Barcelona. London: Society of Chemical Industry, 2001:549–552. Kuzmanovic B, Jallo R, Kuipers NJM, de Haan AB, Kwant G. In: International Solvent Extraction Conference, Cape Town, South Africa, March 2002. Marshalltown: South African Institute of Mining and Metallurgy, 2002:1271–1276. Brugel EG. Production of cyclic ester oligomers from linear polyesters by reactive extraction. WO 0268496, E.I. Du Pont de Nemours and Co., 2002. Kruse A, Dahmen N, Dinjus E. Chem Eng Technol 2002; 25:882–885. Kelkar VV, Ng KM. AIChE J 1999; 45:69–81. Bayley CR, Vaidya NA. In: Collins AN, Sheldrake GN, Crosby J, eds. Chirality in Industry. New York: Wiley, 1992:69–77. Chivate MR, Parikh NC. Indian Chem Eng 1966; 8:111–115. Tare JP, Chivate MR. AIChE Symp Ser 1976; 72(153):95–99. Tare JP. Inst Chem Eng Symp Ser 1978; 54:253–256a. Lee CH, Kim KJ, Ryu SK. J Chem Eng Japan 1992; 25:585–591. Jadhav VK, Chivate MR, Tavare NS. J Chem Eng Data 1991; 36:249–251. Tsuge H, Kotaki Y, Hibino S. J Chem Eng Japan 1987; 20:374–379. Kotaki Y, Tsuge H. Can J Chem Eng 1990; 68:435–442. Momonaga M, Yazawa H, Kagara K. J Chem Eng Japan 1992; 25:237–242. Tsuge H, Matsuo H. ACS Symp Ser 1990; 438:344–354. Tsuge H, Okada K, Yano T, Fukushi N, Akita H. ACS Symp Ser 1997; 667: 254–256. Tsuge H, Yoshiza S, Tsuzuki M. Chem Eng Res Des 1996; 74:797–802. Tsuge H, Tanaka Y, Yoshizawa S, Kuraishi T. Chem Eng Res Des 2002; 80: 105–110. Hirasawa I, Nakagawa H, Yosikawa O, Itoh M. ACS Symp Ser 1997; 667:267–276.


150. Hirasawa I, Katayama A, Kitayama M, Toyokura K. In: Proceedings of the 14th International Symposium on Industrial Crystal, Cambridge, UK, 1999. Rugby: Institution of Chemical Engineers, 1999:916–924. 151. Tsuge H, Takahashi Y, Fujiwara E. In: Proceedings of the 8th World Salt Symposium, The Hague, 2000. Amsterdam: Elsevier Science, 2000:733–738. 152. Ito T, Tsuge H, Fujita A, Fukatani M. In: Proceedings of the 14th International Symposium on Industrial Crystal, Cambridge, UK, 1999. Rugby: Institution of Chemical Engineers, 1999:1510–1517. 153. Am Ende DJ, Crawford TC, Weston NP. Reactive crystallization method to improve particle size. EP 1157726, Pfizer Products Inc., 2001. 154. Chen P-C, Kou KL, Tai HK, Jin SL, Lye CL, Lin CY. J Crystal Growth 2002; 237–239 (Pt 3):2166–2171. 155. ZareNezhad B, Garside J, Tavare NS. In: Proceedings of the 14th International Symposium on Industrial Crystal, Cambridge, UK, 1999. Rugby: Institution of Chemical Engineers, 1999:1316–1330. 156. Lu J, Wang J. Chinese J Chem Eng 2000; 8:68–73. 157. Toyokura K, Tawa K, Ueno J. J Chem Eng Japan 1979; 12:24–28. 158. Chen J, Wang Y, Zheng C. In: Semel J, ed. 2nd International Conference on Process Intensification in Practice. BHR Group Conf. Series, No. 28. London: Mechanical Engineering Publications, 1997:157–164. 159. Chen J-F, Wang Y-H, Guo F, Wang X-M, Zheng C. Ind Eng Chem Res 2000; 39:948–954. 160. Diaz Z. H2S removal from gas streams. U.S. Patent 4,518,576, Shell Oil Company, 1983. 161. Demmink JF, Beenackers AACM. Ind Eng Chem Res 1998; 37:1444–1453. 162. Demmink JF, Mehra A, Beenackers AACM. Chem Eng Sci 2002; 57:1723–1734. 163. Hasebe N, Hasebe N. Desulfurization of a gas by absorption. U.S. Patent 5,389,351, 1995. 164. Hix RM, Lynn S. Ind Eng Chem Res 1991; 30:930–939. 165. Gestrich W. Treatment of gaseous stream containing nitrogen oxide. DE 3640119, 1987. 166. Zapfel W, Marr R, Siebenhofer M. Sep Sci Technol 1997; 32:617–640. 167. Safarik DJ, Eldridge RB. Ind Eng Chem Res 1998; 37:2571–2581. 168. Lucas K, Roth H, Radgen P. AES (ASME) 1993; 30:291–297. 169. Sucksmith I. Chem Eng 1982; 89(13):91–95. 170. Hairston D. Chem Eng 2001; 108(10):31–39. 171. Tripathi RP, Sagar JM, Gulati IB. J Appl Chem Biotechnol 1973; 23:581–588. 172. Berg L. AIChE J 1983; 29:961–966. 173. Ruiz C, Coca J, Vega A, Díez FV. Ind Eng Chem Res 1997; 36:4934–4939. 174. Berg L, Yeh A-I. AIChE J 1984; 30:871–874. 175. Berg L, Yeh A-I. Chem Eng Commun 1984; 30:113–117. 176. Lee F-M, Pahl RH. Ind Eng Chem Process Des Dev 1985; 24:168–172. 177. Berg L, Yeh A-I, Ratanapupech P. Chem Eng Commun 1985; 39:193–199. 178. Berg L, Yeh A-I. Chem Eng Commun 1986; 48:93–101. 179. Berg L, Yeh A-I. Chem Eng Commun 1987; 54:149–159. 180. Berg L. Chem Eng Commun 1987; 52:105–108.


181. Berg L, Vosburgh MG, Christensen RW, Shanahan MJ. Chem Eng Commun 1988; 66:1–21. 182. Mahapatara A, Gaikar VG, Sharma MM. Sep Sci Technol 1988; 23:429–436. 183. Gaikar VG, Sau GK. Sep Sci Technol 1996; 31:1979–1988. 184. Yeh A-I, Berg L, Warren KJ. Chem Eng Commun 1988; 68:69–79. 185. Yeh A-I, Berg L. Chem Eng Commun 1992; 113:147–153. 186. Liu F, Zhang C, Huang F, Zhang C. Fuel Sci Technol Int 1993; 11:1537–1550. 187. Vega A, Díez F, Esteban R, Coca J. Ind Eng Chem Res 1997; 36:803–807. 188. Arce A, Rodil E, Soto A. Can J Chem Eng 1999; 77:1135–1140. 189. Tiverios PG, van Brunt V. Ind Eng Chem Res 2000; 39:1614–1623. 190. Schlovsky G, Loftus B. Chem Eng 2000; 107(2):96–98. 191. Liao B, Lei Z, Xu Z, Zhou R, Duan Z. Chem Eng J 2001; 84:581–586. 192. Lei Z, Zhou R, Duan Z. J Chem Eng Japan 2002; 35:211–216. 193. Lei Z, Zhou R, Duan Z. Chem Eng J 2002; 85:379–386. 194. Lee F-M. Chem Eng 1998; 105(11):112–121. 195. Lei Z, Wang H, Zhou R, Duan Z. Chem Eng J 2002; 87:149–156. 196. Herbst WA. Adsorptive distillation. U.S. Patent 2,665,315, Standard Oil Development Company, 1954. 197. Marple S, Jr, Foecking NJ. Adsorptive distillation process. U.S. Patent 2,768,942, Shell Development Company, 1956. 198. Rice RG. Chem Eng Commun 1981; 10:111–126. 199. Ming Z, Chunjian X, Guocong YU. Prog Nat Sci 1995; 5(3):291–298. 200. Yu KT, Zhou M, Yu CJ. A novel separation process. Distillation accompanied by adsorption. In: Proceedings of the 5th World Congress on Chemical Engineering, San Diego, Vol. 1, New York: AIChE, 1996:347–352. 201. Zhou M, Xu CJ, Zhang GH, Yu KT. Study on the separation of p- and m-xylene by adsorptive distillation. In: Proceedings of the 5th World Congress on Chemical Engineering, San Diego, Vol. 1, New York: AIChE, 1996:962–967. 202. Lawson KW, Lloyd DR. J Membrane Sci 1997; 124:1–25. 203. Velazquez A, Mengual JI. Ind Eng Chem Res 1995; 34:585–590. 204. CADDET Center for Renewable Energy. A Solar Desalination System Using the Membrane Distillation Process. Technical Brochure No. 46, 2001. 205. Kimura S, Nakao S-I. J Membrane Sci 1987; 33:285–298. 206. Udriot H, Araque A, von Stockar U. Chem Eng J 1994; 54:87–93. 207. Tomaszewska M, Gryta M, Morawski AW. J Membrane Sci 1995; 102:113–122. 208. Qureshi N, Meagher MM, Hutkins RW. Sep Sci Technol 1994; 29:1733–1748. 209. Calabro V, Drioli E, Matera F. Desalination 1991; 83:209–224. 210. Chmielewski AG, Harasimowicz M, Zakrzewska-Trznadel G. In: Howell JA, Noworyta A, eds. Towards Hybrid Membrane and Biotechnology Solutions for Polish Environ/ / mental Problems. Wroclaw, Poland: Wroclaw Technical University Press, 1995. 211. Banat FA, Simandl J. Chem Eng Sci 1996; 51:1257–1265. 212. Ortiz de Zárate JM, Rincón C, Mengual JI. Sep Sci Technol 1998; 33:283–296. 213. Couffin N, Cabassud C, Lahoussine-Turcaud V. Desalination 1998; 117:233–245. 214. Gryta M, Karakulski K. Desalination 1999; 121:23–29. 215. Izquierdo-Gil MA, García-Payo MC, Fernández-Pineda C. Sep Sci Technol. 1999; 34:1773–1801.


216. Izquierdo-Gil MA, García-Payo MC, Fernández-Pineda C. J Membrane Sci 1999; 155:291–307. 217. Rijncón C, Ortiz de Zárate JM, Mengual JI. J Membrane Sci 1999; 158:155–165. 218. Banat FA, Simandl J. J Membrane Sci 1999; 163:333–348. 219. Banat FA, Al-Rub FA, Shannag M. Heat Mass Transfer 1999; 35:423–431. 220. Banat FA, Simandl J. J Chem Technol Biotechnol 2000; 75:168–178. 221. Banat FA, Al-Shannag M. Bioproc Eng 2000; 23:643–649. 222. Gryta M, Morawski AW, Tomaszewska M. Catal Today 2000; 56:159–165. 223. Gryta M. Sep Pur Technol 2001; 24:283–296. 224. Tomaszewska M. Ind Eng Chem Res 2000; 39:3038–3041. 225. Duan S, Ito A, Ohkawa A. J Chem Eng Japan 2001; 34:1069–1073. 226. TNO Environment, Energy and Process Innovation. Memstill®: Membrane Distillation. Technical information brochure, 2002. 227. TNO Environment, Energy and Process Innovation. MEMFRAC distillation: ModularEnergy Efficient-Compact. Technical information brochure, 2002. 228. Moganti S, Noble RD, Koval CA. J Membrane Sci 1994; 93:31–44. 229. Stephan W, Noble RD, Koval CA. J Membrane Sci 1995; 99:259–272. 230. Rautenbach R, Albrecht R. J Membrane Sci 1985; 25:25–54. 231. Goldblatt ME, Gooding CH. Ind Membrane Process 1988; 82:51–59. 232. Davis JC, Valus RJ, Eshragi R, Velikoff AE. Sep Sci Technol 1993; 28:463–476. 233. Al-Rabiah AA, Timmerhaus KD, Noble RD. Utilization of a hybrid membrane system in olefin production. In: Proceedings of the 5th World Congress of Chemical Engineering, San Diego, Vol. IV, New York: AIChE, 1996:335–340. 234. Office of Industrial Technologies, U.S. Department of Energy. Olefin Recovery from Chemical Industry Waste Streams. Chemicals Project Fact Sheet, 2000. 235. Office of Industrial Technologies, U.S. Department of Energy. Energy-Saving Separation Technology for the Petroleum Industry. Petroleum Project Fact Sheet, 2000. 236. Bausa J, Marquardt W. Ind Eng Chem Res 2000; 39:1658–1672. 237. Shichang W, Shichang X, Yingjie Q. Chinese J Chem Eng 1993; 1:160–170. 238. Mahmud H, Kumar A, Narbaitz RM, Matsuura T. J Mem Sci 2002; 209:207–219. 239. Herzog H, Falk-Pedersen O. The Kvaerner membrane contractor. Lessons from a case study in how to reduce capture costs. 5th International Conference on Greenhouse Gas Control Technologies. Cairns, Australia, Aug 13–16, 2000. 240. Shichang W, Shichang X, Yingjie Q. Chinese J Chem Eng 1993; 1:160–170. 241. Nii S, Takeuchi H. Gas Sep Purif 1994; 8:107–114. 242. Shen Z, Huang J, Qian G. Water Environ Res 1997; 69:363–367. 243. Li K, Wang D, Koe CC, Teo WK. Chem Eng Sci 1998; 53:1111–1119. 244. Klassen R, Feron PHM, Jansen AE. Flue gas treatment with membrane gas absorption. In: Proceedings of the 1998 International Joint Power Geneneration Conference, PWR-Vol. 33, ASME, Volume 2, 1998:309–316. 245. Li K. Chem Eng Technol 2002; 25:203–206. 246. Mahmud H, Kumar A, Narbaitz RM, Matsuura T. J Mem Sci 2002; 209:207–219. 247. Thömmes J, Kula M-R. Biotechnol Prog 1995; 11:357–367. 248. Bamford CH, Al-Lamee KG, McLaughlin PJ, Purbrick MD, Wear TJ. In: Kahovec J, ed. Macromolecules. IUPAC 34th International Symposium on Macromolecules, Utrecht, the Netherlands: VSP, 1994:431–440.


249. Malakian A. In: Street HG, ed. Highly Selective Separations in Biotechnology. Glasgow: Blacki, 1994:34–54. 250. Birkenmeier G, Dietze H. J Chromatography B 1997; 704:63–68. 251. Berruex LG, Freitag R. In: Bernard A, et al., eds. Animal Cell Technology: Products from Cells, Cells as Products. Dordrecht, Netherlands: Kluwer Academic, 1999: 223–225. 252. Zou H, Luo Q, Zhou D. J Biochem Biophys Methods 2001; 49:199–240. 253. Borcherding H, Hicke H-G, Jorcke D, Ulbricht M. Desalination 2002; 149:297–302. 254. Reif O-W, Nier V, Freitag R. In: Alberghinia L et al., eds. ECB6: Proceedings of the 6th European Congress on Biotechnology. Amsterdam: Elsevier Science, 1994:521–524. 255. Ghosh R. J Chromatography A 2001; 923:59–64. 256. Lütkemeyer D, Bretschneider M, Büntemeyer H, Lehmann J. J Chromatography 1993; 639:57–66. 257. Ghosh R. J Membrane Sci 2001; 192:243–247. 258. Tennikova TB, Svec F. J Chromatography 1993; 646:279–288. 259. Giovannini R, Freitag R, Tennikova TB. Anal Chem 1998; 70:3348–3354. 260. Podgornik A, Barut M, Jancˇar J, Sˇtrancar A. Anal Chem 1999; 71:2986–2991. 261. Wang P-C, Gao J, Lee CS. J Chromatography A 2002; 942:115–122. 262. Yang X-J, Pin C, Fane AG. Anal Chim Acta 1998; 369:17–20. 263. TNO Environment, Energy and Process Innovation. Pertraction for Water Treatment. Technical information brochure, 2002. 264. Yeh H-M, Cheng H-H, Hsieh M-J. Chem Eng Sci 2002; 57:2457–2466. 265. Lu C-H, Cheng Y-C, Tsai S-W. Biotechnol Bioeng 2002; 79:200–210. 266. Hadik P, Nagy E, Szabo L-P. Bulgarian Chem Commun 2001; 33:389–394. 267. Abe K, Goto M, Nakashio F. Sep Sci Technol 1997; 32:1921–1935. 268. Wang Y, Luo G, Cai W, Wang Y, Dai Y. Sep Sci Technol 2002; 37:1163–1177. 269. Berrama T, Pareau D, Stambouli M, Durand G. Purification and concentration of lactic acid by emulsion liquid membrane extraction. In: Cox, Hidalgo M, Valiente M, eds. Solvent Extraction for the 21st Century, Proceedings of ISEC ’99, Barcelona. London: Society of Chemical Industry, 2001:983–987. 270. Liu J, Chao J, Wen M, Jiang G. J Sep Sci 2001; 24:874–878. 271. Cichy W, Schlosser S, Szymanowski J. Solv Extr Ion Exchange 2001; 19:905–923. 272. Li Y, Luo G, Dai Y. Chinese J Chem Eng 1999; 7:104–109. 273. Kirschlig P, Nowak K, Miesiac I, Nitsch W, Szymanowski J. Solv Extr Res Dev, Japan 2001;8:135–143. 274. Murav’ev LL, Vandyshev AB, Makarov VM. Chem Petrol Eng 2000; 36:480–484. 275. Bessarabov DG, Theron JP, Sanderson RD, Schwarz H-H, Schossig-Tiedemann M, Paul D. Sep Purif Technol 1999; 16:167–174. 276. Lin SH, Pan CL, Leu HG. J Hazard Mater 1999; 65:298–304. 277. Burfeind J, Schügerl K. Chem Eng Technol 1999; 22:197–199. 278. Sahoo GC, Dutta NN, Dass NN. J Membrane Sci 1999; 157:251–261. 279. Juang R-S, Lee S-H, Shiau R-C. J Membrane Sci 1998; 146:95–104. 280. Dzygiel P, Wieczorek P, Mathiasson L, Jonsson JA. Anal Lett 1998; 31:1261–1274. 281. Sahoo GC, Dutta NN. J Membrane Sci 1998; 145:15–26. 282. Zigova J, Sturdik E. J Ind Microb Biotechnol 2000; 24:153–160. 283. Wodzki R, Nowaczyk J, Kujawski M. Sep Pur Technol 2000; 21:39–54. 284. Boey SC, Garcia del Cerro MC, Pyle DL. Chem Eng Res Des 1987; 65:218–223.


285. Scholler C, Chaudhuri JB, Pyle DL. Biotechnol Bioeng 1993; 42:50–58. 286. Grobben NG, Eggink G, Cuperus FP, Huizing HJ. Appl Microb Biotechnol 1993; 39:494–498. 287. Matsuba Y, Kitamura Y, Takahashi T. Proc Metallurgy 1992; 7B:1637–1642. 288. Degener W. Metall (Isernhagen, Germany) 1988; 42:817–820. 289. Hu S-Y, Wiencek JM. AIChE J 1998; 44:570–581. 290. Kim BM. AIChE Symp Ser 1985; 81(243):126–132. 291. Boyadzhyev L, Lazarova Z. Chimica Oggi 1993; 11(11–12):29–38. 292. Janssen AE, Klaassen R, Maanen HCHJV, Akkerhuis JJ. Rec Prog Genie Procedes 1992; 6:389–394. 293. Wang Y, Zhu S, Dai Y. Removal of VOCs from wastewater using pertraction. In: Cox M, Hidalgo M, Valiente M, eds. Solvent Extraction for the 21st Century. Proceedings of ISEC ’99, Barcelona. London: Society of Chemical Industry, 2001:177–182. 294. Klaassen R, Janssen AE, Akkerhuis JJ, Bult BA, Oesterhold FIHM, Schenider J. Rec Prog Genie Procedes 1992; 6:183–188. 295. Klaassen R. Chemie Technik (Sonderaus., Chemie Umwelt Technik) 1999; 27:24–28. 296. Livingston A, Ferreira F, Han S, Boam A, Zhang S. In: Preprints 8. Aachener Membran Kolloquium, 27–29 März 2001, Aachen. Mainz: VDI, 2001:1205–1214. 297. Chivate MR, Shah SM. Chem Eng Sci 1956; 5:232–241. 298. Dikshit RC, Chivate MR. Chem Eng Sci 1970; 25:311–317. 299. Carton A, Bolado S, Marcos MM. Informacion Tecnologica 2000; 11:73–82. 300. Rajagopal S, Ng KM, Douglas JM. AIChE J 1991; 37:437–447. 301. Smith JS, Valsaraj KT. Chem Eng Prog 1998; 94(5):69–77. 302. Bryson BG, Valsaraj KT. J Hazard Mater 2000; 2601:1–11. 303. Zeitsch KJ. Ind Eng Chem Res 1999; 38:4123–4124. 304. Curcio E, Di Profio G, Drioli E. Desalination 2002; 145:173–177. 305. Adler S, Beaver E, Bryan P, Rogers JEL, Robinson S, Russomanno C. Vision 2020: 1998 Separations Roadmap. New York: AIChE, Center for Waste Reduction Technologies, 1998. 306. Tlatlik S, Schembecker G. Process synthesis for reactive separations. In: Proceedings of ARS-1, Advances in Reactive Separations 1, University of Dortmund, Germany, October 12, 2000:1–10.


9 Reactive Separations in Fluid Systems E.Y. Kenig and A. Górak University of Dortmund, Dortmund, Germany

H.-J. Bart University of Kaiserslautern, Kaiserslautern, Germany

1. INTRODUCTION: AN OVERVIEW OF REACTIVE SEPARATIONS Chemical manufacturing companies produce materials based on chemical reactions between selected feed stocks. In many cases the completion of the chemical reactions is limited by the equilibrium between feed and product. The process must then include the separation of this equilibrium mixture and recycling of the reactants. The fundamental process steps of bringing material together, causing them to react, and then separating products from reactants are common to many processes. Conventionally, each unit operation—whether mixing or absorption, distillation, evaporation, crystallization, in fact, any of the heat-, mass-, and momentumtransfer operations so familiar to chemical engineers—is typically performed in individual items of equipment, which, when arranged together in sequence, make up the complete process plant. As reaction and separation stages are carried out in discrete equipment units, equipment and energy costs are added up from these major steps. However, this historical view of plant design is now being challenged by the combination of two or more unit operations into one plant unit. The potential for capital cost savings is obvious, but there are often many other process advantages that accrue from such combinations (1).


In recent decades, a combination of separation and reaction inside a single unit has become more and more popular. This combination has been recognized by chemical process industries as having favorable economics for carrying out reaction simultaneously with separation for certain classes of reacting systems, and many new processes (called reactive separations) have been invented based on this technology (2–9). Reactive separation units may also be treated as a kind of multifunctional reactor in which the functionalities of several processes are combined to generate the new reactor concept (Figure 1). The most important examples of reactive separation processes (RSPs) are reactive distillation (RD), reactive absorption (RA), and reactive extraction (RE). In RD, reaction and distillation take place within the same zone of a distillation column. Reactants are converted to products, with simultaneous separation of the products and recycling of unused reactants. The RD process can be efficient in both size and cost of capital equipment and in energy used to achieve a complete conversion of reactants. Since reactor costs are often less than 10% of the capital investment, the combination of a relatively cheap reactor with a distillation column offers great potential for overall savings. Among suitable RD processes are etherifications, nitrations, esterifications, transesterifications, condensations, and alcylations (2).

FIGURE 1 Reactive separation units as multifunctional reactors. (Inspired by Ref. 4.)


Similarly, in RA, reactions occur simultaneously with the component transport and absorptive separation, in the same column zone. These processes are used predominantly for the production of basic chemicals, e.g., sulphuric or nitric acids, and for the removal of components from gas and liquid streams. This can be either the cleanup of process gas streams or the removal of toxic or harmful substances in flue gases. Absorbers or scrubbers where RA is performed are often considered gas–liquid reactors (10). If more attention is paid to the mass transport, these apparatuses are instead treated as absorption units. Reactive extraction uses liquid ion exchangers that promote a selective reaction or separation. The solutes are very often ionic species (metal ions or organic/inorganic acids) or intermediates (furfural phenols, etc.), and the extraction chemistry is discussed elsewhere (11–13). Reactive extraction can be used for separation/ purification or enrichment or conversion of salts (14). A 2001 review on reactive phase equilibria, kinetics, and mass transfer and apparative techniques is given in Ref. 8. Reactive extraction equipment is discussed in detail in Ref. 15, and recent advances are given in Ref. 16. Reactive absorption, distillation, and extraction have much in common. First of all, they involve at least one liquid phase, and therefore the properties of the liquid state become significant. Second, they occur in moving systems; thus the process hydrodynamics plays an important part. Third, these processes are based on the contact of at least two phases, and therefore the interfacial transport phenomena have to be considered. Further common features are multicomponent interactions of mixture components, a tricky interplay of mass transport and chemical reactions, and complex process chemistry and thermodynamics. On the other hand, RD, RA, and RE have a number of specific features that should be considered with care and described by different approaches. Before going into detail, it is worthy to note that the operating window of reactive separations may be somewhat limited, since these operations are feasible only if they allow for both separation and reaction within the same range of temperature and pressure and, on the other hand, for the safe operation from the constructional point of view (Figure 2). 1.1.

Reactive Absorption

The main purposes of absorption processes are the removal of one or more components from the gas phase, production of particular substances in the liquid phase, and gas mixture separation (3). Industrial absorption operations are usually realized by combining absorption and desorption units. The example given in Figure 3 illustrates this combination of two processes. In an absorber, one or several gas components are absorbed by a lean solvent, either physically or chemically. A rich solvent, after preheating in heat exchangers H1 and H3, is transported to the top of a desorption unit, which usually operates


FIGURE 2 Feasibility of reactive separation, depending on mechanical design, chemical reaction, and separation performance.

under a pressure lower than that in the absorber. Part of the gas absorbed by the rich solvent is desorbed due to flashing and heating. The other part has to be desorbed in the stripper via countercurrent contact of liquid with the inert gas or steam. The lean solvent then flows through heat exchanger H1 to recover heat necessary for heating the reach solvent, passes through heat exchanger H2 to cool down to a desired temperature, and finally enters the absorber (3). Usually a small amount of fresh solvent should be added to the column in order to equalize the solvent loss due to evaporation in the desorber or to irreversible chemical reactions occurring in the whole system (3). Reactive absorption represents a process in which a selective solution of gaseous species by a liquid solvent phase is combined with chemical reactions. As compared to purely physical absorption, RA does not necessarily require elevated pressure and high solubility of absorbed components; because of the chemical reaction, the equilibrium state can be shifted favorably, resulting in enhanced solution capacity (17). Most RA processes involve reactions in the liquid phase only; in some of them, both liquid and gas reactions occur (18,19). Usually the effect of chemical reactions in RA processes is advantageous only in the region of low gas-phase concentrations, due to limitations stemming from the reaction stoichiometry or equilibrium (20). Further difficulties of RA applications may be caused by the reaction heat through exothermic reactions and by relatively difficult solvent regeneration (21,22). Most RA processes are


steady-state operations, either homogeneously catalyzed or auto-catalyzed. Some important industrial applications of RA are given in Table 1. Reactive absorption can be realized in a variety of equipment types, e.g., in film absorbers, plate columns, packed units, or bubble columns. This process is characterized by independent flow of both phases, which is different from distillation and permits both cocurrent (downflow and upflow) and countercurrent regimes. Reactive absorption is essentially an old process, known since the foundation of modern industry. This is a very important process, too, being the basic operation in many technological chains. More recently, the role of RA as a key environmental protection process has grown up significantly. Despite the clear importance of RA, its behavior is still not properly understood. This can be attributed to a very complex combination of process thermodynamics and kinetics, with intricate reaction schemes including ionic species, reaction rates varying over a wide range, and complex mass transfer and reaction coupling. As compared to distillation, RA is a fully rate-controlled process, and it definitely occurs far from the equilibrium state. Therefore, practitioners and theoreticians are highly interested in establishing a proper rate-based description of this process. 1.2.

Reactive Distillation

Reactive distillation is a combination of chemical reactions and distillation (Figure 4b). This operation provides promising process alternatives to traditional

FIGURE 3 Scheme of an absorber–desorber link. (Adapted from Ref. 3.)


TABLE 1 Applications of RA Processes Aim of the process

Example

Application area

Refs.

Removal of harmful substances

Coke oven gas purification, amine washing Solvent regeneration

Gas purification

23–25

Gas separation

26

Manufacture of sulphuric acid, formaldehyde preparation, manufacture of soda ash Manufacture of nitrogenous fertilizers Water removal from natural gas, air drying Synthesis gas conditioning

Chemical synthesis

27–29

Fertilizer industry

30

Gas drying

31,32

Gas separation/ gas purification

26

Retrieval/regeneration of valuable substances or nonreacted reactants Production/preparation of particular products

Water removal

Conditioning of gas streams

sequential operations, shown in Figure 4a. Among potential advantages of RD are: New, less expensive products Higher efficiency because of overcoming thermodynamic and kinetic limitations Better selectivity due to suppressing of undesired reactions Higher raw material conversion Avoiding of hot spots Savings due to smaller equipment Less environmental pollution One can distinguish between homogeneously and heterogeneously catalyzed RD; the latter is often called catalytic distillation (CD). The applicability of the RD process is highly dependent on the properties of the chemical system at hand. A classical example for which RD is recommended may be the reaction in which the products are generated by a reversible reaction, e.g., in the production of methyl acetate. This system is very complex because of the occurrence of several azeotropes between reactants and products.


FIGURE 4 Reactive distillation (b) as alternative to the sequential operation (a).

A usual solution in this case is a sequence of a reactor and several separation units (Figure 5). Another way—an integrated RD process such as shown in Figure 6— allows for simultaneous formation of methyl acetate in the reaction zone, extractive distillation and product enrichment in the upper part of the column, and methanol separation in the stripping zone. The production of esters such as methyl acetate, ethyl acetate, and butyl acetate has for years been an interesting RD application. The most important application of RD today seems to be the production of ethers such as methyl tertiary butyl ether (MTBE), ethyl tertiary butyl ether (ETBE), and tertiary amyl methyl ether (TAME), which are widely used as modern gasoline components. Figure 7, upper part, shows a traditional process for MTBE production, which is a strongly exothermic reaction. The disadvantages of that process can be avoided if the reaction and separation take place within the same zone of the reactor (Figure 7, lower part). Table 2 gives a short overview of possible RD applications. The design of RD is currently based on expensive and time-consuming sequences of laboratory and pilot-plant experiments, since there is no commercially available software adequately describing all relevant features of reactions (catalyst, kinetics, holdup) and distillation (VLE, thermodynamics, plate and packing behavior) as well as their combination in RD. There is also a need to improve catalysts and column internals for RD applications (1,51). Figures 8 and 9 show some examples of catalytic internals, applied for reactive distillation. 1.3.

Reactive Extraction

Liquid–liquid extraction is based on partial miscibility of liquids. In the simplest extraction system, two compounds have to be separated. This can be done by extracting with a carefully selected solvent, in which one compound (solute) easily dissolves whereas the other (nonsolute) does not. The solvent has to be recovered


FIGURE 5 Methyl acetate synthesis: conventional scheme. (From Ref. 33.)


FIGURE 6 Methyl acetate synthesis: reactive distillation scheme. (From Ref. 33.)

FIGURE 7 MTBE synthesis: conventional scheme (above) and reactive distillation scheme (below).


from the extract for recycling. In countercurrent extraction processes, there is a light phase and a heavy phase, with one phase dispersed in the other. Which phase has to be dispersed is an important question in the design of the process. In reactive extraction, the use of liquid ion exchangers is recommended in order to extract ionic solutes. These exchangers can be applied to manifold extraction processes in the chemical industry (e.g., extraction of furfural, organic, and inorganic acids), biochemical and pharmaceutical productions (e.g., penicillin, amino acids), hydrometallurgy (e.g., mining of metals), and all related environmental applications. These last are especially attractive, since liquid ion exchangers react very selectively and have an advantageous performance at very low feed concentrations. For practical purposes, an ion exchanger is usually diluted in a nonaromatic, high-boiling diluent (boiling point about 500 K) that is immiscible with water. This prevents solvent losses and toxic problems and gives the organic phase the required physical properties (high interfacial tension, low viscosity, low density), since most liquid ion exchangers are highly viscous or even solid. In some cases a modifier, usually a long-chain alcohol, is added to help in the solubilization of the solute–ion exchanger complex. At very high solute loadings, a split of the organic phase in a solvent-rich and a solvent-poor fraction may occur, especially when using aliphatic diluents. The organic phase in RE is thus not a single substance, as in physical extraction systems in which such three-phase liquid systems are not encountered. Re-extraction is usually performed with chemicals, for instance, with strong mineral acids. All liquid ion exchangers can be mixed together in order to generate synergistic effects. As a special case, an equimolar mixture of cation and anion exchangers gives a “mixed” extraction system, which can extract salts or acids. In this case the re-extraction occurs by shift of either temperature, aqueous ionic strength, or acidity/basicidity. Equilibrium and selectivity constitute important aspects of reactive and nonreactive extraction processes. Another important factor is the reaction kinetics, which has to be reasonably fast. Most RE processes are close to equilibrium in less than five minutes. Many ion exchangers need reaction times of less than one minute, and thus diffusion of the solute complex in the organic phase is the rate-determining step. The cation and anion exchangers are amphiphilic substances that are adsorbed at the interface. The latter is then rigid and independent of the droplet diameter, since friction forces are shielded. This is similar to physical extraction systems, in which an analogous behavior is caused by surfactants in the aqueous feed accumulated at the interface. The problems concerning reaction equilibrium and kinetics description based on chemical potentials rather than on concentrations are extensively discussed in Refs. 54 and 55, using the zinc system. The latter is recommended as a reactive liquid–liquid reference extraction test system by the European Federation of


TABLE 2 Applications of RD Processes Reaction type Esterification

Transesterification

Hydrolysis

Etherification

Alcylation Condensation

Dismutation Hydration

Nitration

a

Synthesis Methyl acetate from methanol and acetic acid Methyl acetate from methanol and acetic acid Ethyl acetate from ethanol and acetic acid Butyl acetate from butanol and acetic acid Ethyl acetate from ethanol and butyl acetate Diethyl carbonate from ethanol and dimethyl carbonate Acetic acid and methanol from methyl acetate and water MTBE from isobutene and methanol ETBE from isobutene and ethanol TAME from isoamylene and methanol Cumene from propylene and benzene Diacetone alcohol from acetone Bisphenol-A from phenol and acetone Monosilane from trichlorsilane Mono ethylene glycol from ethylene oxide and water 4-Nitrochlorobenzene from chlorobenzene and nitric acid

Catalysta hom.

33

het.

34,35

no data

36

hom.

37

hom.

38

het.

39

het.

40

het.

41,42

het.

43

het.

44

het.

45

het.

46

no data

47

het.

48

hom.

49

hom.

50

hom.: homogeneously catalyzed, het.: heterogeneously catalyzed.


Refs.

FIGURE 8 (a) Schematic of an RD column filled with catalytic internals CD TECH (1—catalytic balls, 2—feed, 3—distillate, 4—bottom product, 5—sieve tray) and (b) catalytic structured packing Sulzer Katapak-S. (Part a from Ref. 52.)

Chemical Engineering (EFCE) and is thus well documented (http://www.icheme.uk/ learning/ or http://www.dechema.de/extraktion). The selection of the right solvent is the key to successful separation by nonreactive and reactive liquid–liquid extraction. In this respect, different criteria should be taken into account, e.g., Selectivity Capacity Recoverability of solvent Density Viscosity and melting point Insolubility of solvent Interfacial tension Toxicity and flammability Corrosivity Thermal and chemical stability Availability and costs Environmental impact.


Some of these criteria are crucial, while others are desirable properties improving separation and/or making it more economical. Solvent selectivity, recoverability, and a large density difference in respect to the raffinate are essential. Some of the requirements on the solvent can be in conflict, and thus a compromise may be necessary. Because aromatic diluents are more expensive and more toxic than aliphatic ones, the latter are preferably used in industrial practice (see earlier). Aromatic diluents, with equivalent molecular weights comparable to those of aliphatic ones, are more polar and thus more water soluble. The degradation of the diluent is usually negligible in comparison with that of the ion exchanger. The latter one can be chemically and thermally degrading and also can be poisoned by an irreversibly extracted compound. Reactive extraction is closely related to the droplet phenomena, and thus most theoretical models are based on droplet consideration. Their experimental evaluation can be done using either a rising (falling) droplet apparatus (Figure 10a) for short residence times or a Venturi tube for long contact times (Figure 10b) (56).

FIGURE 9 (a) Catalytic structured packing Montz Multipak and (b) an example of reactive trays. (Part b from Ref. 53.)


FIGURE 10 (a) Rising-droplet apparatus and (b) Venturi tube for droplet mass transfer experiments. (1—feed storage, 2—metering pump, 3—double-flow valve, 4, 5—pumps, 6—heat exchanger, 7—collecting funnel, 8—stream to analysis, 9—valve).

Monodispersed droplets can be produced and in the latter case captured by the counterflowing continuous phase in the conus of the Venturi tube (see Figure 10b). The RE process proceeds in three major types of equipment: mixer-settler systems, column extractors, and centrifugal extractors. Countercurrent column extractors can be further subdivided into nonagitated nonproprietary columns and agitated proprietary extractors. Agitating the liquid–liquid system breaks up droplets and increases the interfacial area to improve the mass transfer and column efficiency. Various forms of energy input are used, e.g., rotation of propellers, impellers, and discs; pulsation, vibration, and ultrasonic devices; and centrifugal devices. Some examples of mechanically agitated contactors are the rotating-disk contactor (RDC), Karr, Oldshue–Rushton, Scheibel, and Kühni columns shown in Figure 11. There are three types of nonproprietary nonagitated types of extraction columns (see Figure 12). The spray columns are the simplest type of extractors, containing only distributors for the feed (often through perforated pipes). This


FIGURE 11 Agitated extractors. Left to right: RDC, Karr, Oldshue–Rushton, Scheibel, Kühni extraction columns.

makes them cheap; however, they are limited in use due to significant axial mixing in the column and the fact that the phases are not coalesced and redistributed. This often results in low efficiencies, which are comparable to one or two theoretical equilibrium stages. Packed columns are much more efficient since the packing reduces back-mixing and enhances drop reformation. The packing types that can be used are the same as those for normal distillation operations (e.g., rings, saddles, or slightly modified structured packings of corrugated metal sheets). Compared to packed beds, structured packings need a reduced crosssectional area for liquid flow, resulting in smaller column diameters. Sieve-tray columns resemble the distillation column design, except that there is no weir. In

FIGURE 12 Nonagitated extractors. Left to right: spray, packed, sieve-tray (light), sieve-tray (heavy) column.


an extraction process, either phase, the light or the heavy one, can be dispersed. This means that there are also two sieve-tray designs in respect to downcomer and top- or bottom-settler/distributor design. Generally, the selection of a specific RE contactor is complicated due to the large number of types available and the number of design parameters. The practical handling and design of a reactive solvent extraction processes can be found elsewhere (see, e.g., Refs. 12 and 13). 2. 2.1.

FUNDAMENTALS OF PROCESS MODELING General

As already mentioned, all three considered RSPs reveal significant similarity, and hence their modeling methods are based largely on the same framework. Because of their multicomponent nature, RSPs are affected by a complex thermodynamic and diffusional coupling, which, in turn, is accompanied by simultaneous chemical reactions (57–59). To describe such phenomena adequately, specially developed mathematical models capable of taking into consideration column hydrodynamics, mass transfer resistances, and reaction kinetics are required. Homogeneously catalyzed RD, with a liquid catalyst acting as a mixture component, and auto-catalyzed RD present essentially a combination of transport phenomena and reactions taking place in a two-phase system with an interface. In this respect they are very similar to RA and RE, and, generally, reaction has to be considered both in the bulk and in the film region. For slow reactions, a reaction account exclusively in the bulk phase is usually sufficient. For heterogeneous systems (CD), it is generally necessary to consider additionally the phenomena in the solid catalyst phase. In this case, very detailed models using intrinsic kinetics and covering mass transport inside the porous catalyst arise (see, e.g., Refs. 60–62). However, it is often assumed that all internal (inside the porous medium) and external mass transfer resistances can be lumped together (35,63,64). In this case each catalytically active site is in contact with the liquid bulk, i.e., the catalyst surface is totally exposed to the liquid bulk phase and can be completely described by the bulk variables (9,64). This results in the socalled pseudo-homogeneous models. If the reaction (either homogeneous or heterogeneous) is very fast, it does not depend on the reaction kinetics and thus can be described using the data on chemical equilibrium only. Modeling of hydrodynamics in gas/vapor/liquid–liquid contactors includes an appropriate description of axial dispersion, liquid holdup, and pressure drop. The correlations giving such a description have been published in numerous papers and are collected in several reviews and textbooks (e.g., Refs. 65 and 66). Nevertheless, there is still a need for a better description of the hydrodynamics in catalytic column internals; this is being reflected by research activities in progress (67).


The description of thermodynamics and chemical properties of the RSP is very process specific, and hence its general detailed discussion would constitute a separate issue. Therefore, we will give only a brief discussion of these topics in the context of the following case studies (Section 3). Further related details can be found in Refs. 68–74. In order to model large industrial reactive separation units, a proper subdivision of a column apparatus into smaller elements is usually necessary. These elements (the so-called stages) are identified with real trays or segments of a packed column. They can be described using different theoretical concepts, with a wide range of physicochemical assumptions and accuracy. 2.2.

Equilibrium-Stage Model

In recent decades, the modeling and design of RSPs has usually been based on the equilibrium-stage model. Since 1893, when the first equilibrium-stage model was published by Sorel (75), numerous publications discussing various aspects of model development, application, and solution have appeared in the literature (76). The equilibrium-stage model assumes that each gas/vapor/liquid stream leaving a tray is in thermodynamic equilibrium with the corresponding liquid stream leaving the same tray. For the packed columns, the idea of the height equivalent to the theoretical stage (HETS) is used. In case of RSPs, the chemical reaction has to be additionally taken into account, either via reaction equilibrium equations or via rate expressions integrated into the mass and energy balances. In this respect, much depends on the relation between the mass transfer and reaction rates in a particular RSP. The definition of the Hatta number representing the reaction rate in reference to that of the mass transfer helps to discriminate between very fast, fast, average, and slow chemical reactions (68,77). If a fast reaction system is considered, the RSP can be satisfactory described assuming a reaction equilibrium. Here, a proper modeling approach is based on the nonreactive equilibrium-stage model, extended by simultaneously using the chemical equilibrium relationship. Such descriptions can be appropriate enough for instantaneous reactions and those close to them. In contrast, if the chemical reaction is slow, the reaction rate dominates the whole process, and therefore, a reaction kinetics expression has to be integrated into the mass and energy balances. This concept has been used in a number of studies, for RA (e.g., Refs. 78 and 79), RD (e.g., Refs. 80 and 81), and RE (e.g., Refs. 8 and 12) process simulations. In practice, RSPs rarely operate at thermodynamic equilibrium. Therefore, some correlation parameters, such as tray efficiencies or HETS values, have been introduced to adjust the equilibrium-based theoretical description to reality. For multicomponent mixtures, however, this concept often fails, since diffusion interactions of several components result in unusual phenomena such as osmotic or reverse


diffusion and mass transfer barrier (82,83). These effects cause a strange behavior of the efficiency factors, which are different for each component, vary along the column height, and show a strong dependency on the component concentration (57,83,84). The acceleration of mass transfer due to chemical reactions in the interfacial region is often accounted for via the so-called enhancement factors (27,68,69). They are either obtained by fitting experimental results or derived theoretically on the grounds of simplified model assumptions. It is not possible to derive the enhancement factors properly from binary experiments, and significant problems arise if reversible, parallel, or consecutive reactions take place. The equilibrium-stage model seems to be suitable for esterification reaction in CD processes (see Refs. 35 and 74). However, it cannot be recommended for all reaction types, especially those with higher reaction rates. 2.3.

Rate-Based Approach

A more physically consistent way to describe a column stage is known as the ratebased approach (57,85,86). This approach implies that actual rates of multicomponent mass and heat transfer and chemical reactions are taken into account directly. Considering homogeneous RSPs, mass transfer at the gas/vapor/liquid– liquid interface can be described using different theoretical concepts (57,59). Most often the two-film model (87) or the penetration/surface renewal model (27,88) is used, in which the model parameters are estimated via experimental correlations. In this respect the two-film model is advantageous since there is a broad spectrum of correlations available in the literature, for all types of internals and systems. For the penetration/surface renewal model, such a choice is limited. In the two-film model (Figure 13), it is assumed that all of the resistance to mass transfer is concentrated in thin stagnant films adjacent to the phase interface and that transfer occurs within these films by steady-state molecular diffusion alone. Outside the films, in the bulk fluid phases, the level of mixing is so high that there is no composition gradient at all. This means that in the film region, only one-dimensional diffusion transport normal to the interface takes place. Multicomponent diffusion in the films is described by the Maxwell–Stefan equations, which can be derived from the kinetic theory of gases (89). The Maxwell–Stefan equations connect diffusion fluxes of the components with the gradients of their chemical potential. With some modification these equations take a generalized form in which they can be used for the description of real gases and liquids (57): n

di

∑ j1

xi N Lj x j N Li c Lt Dij

i 1, . . . , n


(1)

FIGURE 13 Two-film model.

where di is the generalized driving force: di

xi di ℜT dz

i 1, . . . , n

(2)

Similar equations can be also written for the gas/vapor phase. Thus the gas/vapor/liquid–liquid mass transfer is modeled as a combination of the two-film model and the Maxwell–Stefan diffusion description. In this stage model, the equilibrium state exists only at the interface. The film thickness represents a model parameter that can be estimated using mass transfer coefficient correlations. These correlations reflect the mass transport dependence on physical properties and process hydrodynamics and are available from the literature (see, e.g., Refs. 57, 68 and 90). The two-film model representation can serve as a basis for more complicated models used to describe heterogeneously catalyzed RSPs or systems containing suspended solids. In these processes a third solid phase is present, and thus the two-film model is combined with the description of this third phase. This can be done using different levels of model complexity, from quasi-homogeneous description up to the four-film presentations that provide a very detailed description of both vapor/gas/liquid–liquid and solid/liquid interfaces (see, e.g., Refs. 62, 68 and 91). A comparative study of the modeling complexity is given in Ref. 64 for fuel ether synthesis of MTBE and TAME by CD. 2.4.

Computational Fluid Dynamics

Every separation unit operation is governed by the continuum conservation laws, and thus, in principle, everything meaningful to know in the continuum for any process can be determined with computational fluid dynamics (CFD) (92). In recent years there have been significant academic and industrial efforts to enable


the use of CFD for the design, scale-up, and optimal operation of a variety of chemical process equipment. Special attention has been given to the CFD modeling of two-phase flows. The most frequently encountered computational techniques for calculating multiphase flows are Euler–Lagrange and Euler–Euler methods. Euler–Lagrange models are applicable to dispersed flows (93). In these models the flow of the so-called “carrier phase” is simulated by solving continuum-flow equations. The motion of individual particles (or group of particles) of the dispersed phase is tracked through the flow domain using the calculated carrier-phase flow field as input; afterwards, mass, momentum, and energy transfer between the two phases are computed and applied to the carrier-phase flow field prediction. This procedure requires several iterations (94). Euler–Euler models assume interpenetrating continua to derive averaged continuum equations for both phases. The probability that a phase exists at a certain position at a certain time is given by a phase indicator function, which, for steady-state processes, is equivalent to the volume of fraction of the correspondent phase (volume-of-fluid technique). The phase-averaging process introduces further unknowns into the basic conservation equations; their description requires empirical and problem-dependent input (94). In principal, Euler–Euler models are applicable to all multiphase flows. Advantages and disadvantages of both methods are compared, e.g., in Refs. 95 and 96. The volume-of-fluid technique can be used for a priori determination of the morphology and rise characteristics of single bubbles rising in a liquid (97,98). Considerable progress has been made in CFD modeling of bubbling gas–solid fluidized beds by adoption of the Eulerian framework for both the dilute (bubbles) and dense (emulsion) phases (99–102). The use of CFD models for gas–liquid bubble columns has also aroused significant interest in recent years, and both Euler–Euler and Euler–Lagrange methods have been employed for the description of the gas and liquid phases (94–96,103–113). There are also some attempts available in the literature to model tray hydrodynamics using CFD (114–119). Despite considerable success in some fields of application, the CFD simulations are still not fully mastered, especially where the considered processes reveal clearly nonhomogeneous, segregated fluid flow patterns. The latter are usually the basic phenomenon in packed or filmlike units used in reactive and nonreactive separations. One of the important issues with RSPs is the development of efficient column internals. Such internals have to enhance both separation and reaction and maintain a sound balance between them. This is valid for both homogeneously and heterogeneously catalyzed processes, being especially important for CD. An understanding of the complex, multiphase flow on the internals interrelated with the mass transport and chemical reaction constitutes a very important challenge for the future. Some first steps in this respect have been done concerning the performance


of SULZER packings KATAPAK-S® and SULZER BX and OPTIFLOW (67,120–122) as well as trays on which chemical reaction occurs (119). Recently, a substantial effort has been made to optimize column internals for reactive separations and to reduce the number of expensive hydrodynamics experiments via the CFD simulations (67,119,122). Such simulations can be regarded as virtual experiments carried out in order to predict the performance of the internals by varying geometrical and structural parameters, thus reducing the optimization time. Necessary new hydrodynamic models have to be formulated and tested for three-dimensional description of two-phase flow through the internals. Since accurate resolution of the trickle-flow scale is not feasible at the moment, such flow details have to be simplified and are subject to the subgrid modeling supported by experimental investigations of small-scale phenomena. The CFD simulations should be linked with the rate-based process simulator, providing important information on the process hydrodynamics in the form of correlations for mass transfer coefficients, specific contact area, liquid holdup, residence time distribution, and pressure drop. An ability to obtain these correlation via the purely theoretical way rather than by the traditional experimental one should be considered a significant advantage, because this brings a principal opportunity to virtually prototyping of new optimized internals for reactive separations. The local aspects of liquid–liquid two-phase flow in RE has been the focus of CFD analysis by different research groups (123–126). In principle, all aspects concerning single-phase flow phenomena (residence time distribution, impeller discharge flow rate, etc.) can be tackled, even with complex geometries. However, the two-phase CFD is still a challenge, and the droplet interactions (breakup and coalescence) and mass transfer are not implemented in commercially available codes. Thus these issues constitute an open area for further research and development (127). 3.

CASE STUDIES

3.1. 3.1.1.

Absorption of NOx Chemical System

The reactive system considered is a basic one in the production of nitric acid as well as in some other industrial processes (19). It consists of 10 components, including air (N2, O2), water (H2O), oxyacids of nitrogen (HNO2, HNO3), and nitrogen oxides (NO, NO2, N2O, N2O3, N2O4). The components are involved in simultaneous, parallel, and consecutive reactions occurring in both phases. The reactions are of high orders and most of them are exothermic. Reaction kinetics is described by the scheme suggested in Ref. 128 and modified in Ref. 129. This scheme involves eight reactions and can be regarded


as the most extensive reaction system so far. The gas-phase reactions are governed by the following equations: 2NO O 2 → 2 NO 2

H R0 114 kJ/mol

(R1)

NO NO 2 ↔ N 2 O 3

H R0 39.9 kJ/mol

(R2)

2NO 2 ↔ N 2 O 4

H R0 57.2 kJ/mol

(R3)

3NO 2 H 2 O ↔ 2HNO 3 NO

H R0 35.4 kJ/mol

(R4)

whereas the corresponding equations for the liquid phase are 2 NO 2 H 2 O → HNO 2 HNO 3

H R0 10.72 kJ/mol

(R5)

N 2 O 3 H 2 O → 2HNO 2

H R0 3.99 kJ/mol

(R6)

N 2 O 4 H 2 O → HNO 2 HNO 3

H R0 5.03 kJ/mol

(R7)

3HNO 2 ↔ HNO 3 H 2 O 2NO

H R0 7.17 kJ/mol

(R8)

The liquid-phase reactions are valid for nitric acid concentrations below 34 wt %. In the case of higher nitric acid concentrations, Reactions (R5) to (R7) become reversible. The oxidation of NO (Reaction (R1)) is the slowest reaction in this system. Therefore, the total gas-phase holdup in absorbers can be determined using the kinetic data for this reaction (130). The other gas-phase reactions are instantaneous equilibrium reactions. 3.1.2.

Process Setup

Measurements of an industrial NOx absorption process, schematically shown in Figure 14, were described in Ref. 131. The absorption plant constitutes a sequence of four units used for the removal of nitrogen oxides from the waste gas of an adipin acid factory. Each unit is separated by a metal plate into two sections. In fact there are eight columns joined together as a countercurrent absorption plant. This plant is operated at atmospheric pressure. Columns 1–7 have a pump around for cooling of the liquid. The diameter of each column is 2.2 m; the height is 7 m. The packing height is 3.2 m. The packing consists of 35 mm INTALOX ceramic saddles.


FIGURE 14 Absorption plant consisting of four units (eight columns).

The liquid feeds entering columns 7 and 8 are low-concentration nitric acids. The liquid product has a HNO3 concentration of about 35 wt %. The gas feed has a concentration of NOx of about 60,000 vppm. A quarter of NOx is NO; the rest is NO2. 3.1.3.

Results and Discussion

The sensitivity analysis performed in Ref. 129 shows that the suggested model provides concentration profiles that are qualitatively correct. For the simulation of the industrial absorption process shown in Figure 14, the following correlations ensuring the most reliable results are selected: The rate constant of Reaction (R1) (the slowest and hence the most important reaction in the system) according to Ref. 132 The liquid-side mass transfer coefficient according to Ref. 133 The gas-side mass transfer coefficient according to Wehmeier (see Ref. 134) Figures 15 and 16 give an illustration of the model quality. Figure 15 shows a comparison of the simulated and measured gas-phase concentrations of NO and NO2 throughout the whole absorption plant, whereas in Figure 16, experimental and simulated liquid-phase concentrations of HNO3 and HNO2 are demonstrated.


FIGURE 15 Experimental and simulated gas-phase concentrations of NO and NO2 throughout the absorption plant.

The zigzag form of the simulated concentration profiles results from switching different sections of each single column (see Ref. 135). Good agreement between experimental and simulation results can be readily observed, except for the first two columns. Here the larger deviations between experiments and simulated results can be attributed to the fact that at high concentration of HNO3 Reactions (R5) to (R7), assumed to be irreversible reactions, convert to reversible ones; the data on their rate constants are lacking. 3.2.

Coke Gas Purification

3.2.1.

Chemical System

Coke oven gas consists mainly of a mixture of carbon monoxide, hydrogen, methane, and carbon dioxide. It is contaminated with a variety of organic and inorganic compounds that have to be separated in absorption columns before its further use as a synthesis gas. The selective absorption of coke plant gas contamination results from a complex system of parallel liquid-phase reactions. Instantaneous reversible reactions: NH 3 H 2 O ↔ NH4 OH

( R9)

H 2 S H 2 O ↔ HS H 3O

( R10)


HCN H 2 O ↔ CN H 3O

( R11)

HCO3 H 2 O ↔ CO 32 H 3O

( R12)

H 3O OH ↔ 2 H 2 O

( R13)

Finite-rate reversible reactions: CO 2 OH ↔ HCO3

( R14)

CO 2 2H 2 O ↔ HCO3 H 3O

( R15)

CO 2 NH 3 H 2 O ↔ H 2 NCOO H 3O

( R16)

The reactions including CO2 obey first- and second-order kinetics, whereas the other reversible reactions are based on simple proton transfers and are therefore regarded as instantaneous by the corresponding mass action law equations. The formation of bicarbonate ions (HCO3) takes place via two different

FIGURE 16 Experimental and simulated liquid-phase concentrations of HNO3 and HNO2 throughout the absorption plant.


mechanisms. The rate of the direct reaction between carbon dioxide and hydroxyl ions (the most important step) is taken from Ref. 28. Usually the reaction between CO2 and water is very slow and hardly contributes to the total rate of reaction of carbon dioxide. Nevertheless, in this work it was considered of the first order with respect to CO2, since the reaction kinetics depends on the carbonation ratio (136). The absorption rate of carbon dioxide increases in the presence of amines or ammonia. Therefore, the reaction kinetics of NH3 and CO2 has been considered in the model equations, too. The rate constant as a function of the temperature has been determined according to Ref. 136. The coefficients for the calculation of the chemical equilibrium constants in this system of volatile weak electrolytes are taken from Ref. 137. The CO2 absorption is hindered by a slow chemical reaction by which the dissolved carbon dioxide molecules are converted into the more reactive ionic species. Therefore, when gases containing H2S, NH3, and CO2 contact water, the H2S and ammonia are absorbed much more rapidly than CO2, and this selectivity can be accentuated by optimizing the operating conditions (23). Nevertheless, all chemical reactions are coupled by hydronium ions, and additional CO2 absorption leads to the desorption of hydrogen sulfide and decreases the scrubber efficiency. 3.2.2.

Process Setup

Today’s coke plant gas purification processes are mostly carried out under atmospheric pressure, employing a circulated ammonia-based absorbent. The consumption of the external solvent is reduced via the use of ammonia available in the coke gas (138). An example of innovative purification processes is the ammonia hydrogen sulfide circulation scrubbing (ASCS) (Figure 17), in which the ammonia contained in the raw gas dissolves in the NH3 absorber and then the absorbent saturated with the ammonia passes through the H2S absorber to selectively absorb the H2S and HCN components from the coke gas. The next step is the thermal regeneration of the absorbent with the steam in a two-step desorption plant, whereas a part of the deacidified water is fed back into the H2S absorber (25). Pilot-plant experiments have been carried out at real process conditions in the coke plant “August Thyssen” (Duisburg, Germany). The DN 100 pilot column (Figure 17) was made of stainless steel and equipped with about 4 m of structured packing (Sulzer MELLAPAK® 350Y), three liquid distributors, and a digital control system. Several steady-state experiments have been compared with the simulation results and supported the design optimization of the coke gas purification process (25). 3.2.3.


A number of steady-state simulations have been performed with the aim of analyzing the influence of numerical and physicochemical parameters, beginning


FIGURE 17 Ammonia hydrogen sulfide circulation scrubbing process for the coke oven gas purification (right) and H2S absorber (left).


with a single stage and ending with a simulation of a column. Different film and packing section discretizations, several mass transfer and hydrodynamic correlations, and different driving forces and diffusion models have been thoroughly tested (Figure 18). The most sensitive components appeared to be those involved in finite-rate reactions, especially CO2. Furthermore, the impact of electrical forces enhances the absorption of the strong electrolytes H2S and HCN by 3–5%, while the CO2 absorption rate is dominated by the reaction in the film (139,140). Significant changes in the concentration profiles and the component absorption rates due to the film reaction have been established (141,142). Single-stage simulations reveal that intermolecular friction forces do not lead to reverse diffusion effects, and thus the molar fluxes calculated with the effective diffusion approach differ only slightly from those obtained via the Maxwell–Stefan equations without the consideration of generalized driving forces. This result is as expected for dilute solutions and allows one to reduce model complexity for the process studied (143). As a further model simplification, a linearization of the film concentration profiles has been studied. This causes no significant changes in the simulation results and at the same time reduces the total number of equations by half and stabilizes the numerical solution (142). The assumption of chemical equilibrium in

FIGURE 18 Absorption rates calculated with different model assumptions concerning reaction consideration.


FIGURE 19 Liquid-phase axial concentration profiles for the H2S scrubber: comparison between experimental and simulation results based on different model approaches.

the liquid bulk phase does not change the absorption rates significantly, which indicates fast conversion. Therefore, neglecting the film reaction unrealistically reduces the absorption rates. On the other hand, neglecting the reaction kinetics within the film results in completely different orders of magnitude for the calculated absorption degree. As a consequence, the reactions of carbon dioxide should not be regarded as instantaneous, although the corresponding Hatta number of about 7 characterizes the reaction as very fast (3). The model optimized with respect to the numerical parameters and physicochemical properties has been validated against experimental data, whereas the axial concentration and temperature profiles for both phases demonstrated good agreement (Figure 19). It has also been found that the simulations of the scrubber based on the equilibrium-stage model extended by the chemical reaction kinetics yield results completely inconsistent with the experimental studies; namely, the selectivity toward H2S and HCN absorption cannot be reflected (Figure 19). In this case, the film reaction represents an essential element of the rate-based approach that has to be considered in the model. As a result, the only feasible simplification is represented by a linearization of the film concentration profiles, including the implementation of the average reaction kinetics in the liquid film region (143).


3.2.4.

Dynamic Modeling

Steady-state modeling is not sufficient if one faces various disturbances in RA operations (e.g., feed variation) or tries to optimize the startup and shutdown phases of the process. In this case, a knowledge of dynamic process behavior is necessary. Further areas where the dynamic information is crucial are the process control as well as safety issues and training. Dynamic modeling can also be considered as the next step toward the deep process analysis that follows the steady-state modeling and is based on its results. The dynamic formulation of the model equations requires a careful analysis of the whole system in order to prevent high-index problems during the numerical solution (144). As a consequence, a consistent set of initial conditions for the dynamic simulations and suitable descriptions of the hydrodynamics have to be introduced. For instance, pressure drop and liquid holdup must be correlated with the gas and liquid flows. The model optimized based on steady-state analysis allows for a dynamic real-time simulation of the entire absorption process. Because dynamic behavior is determined mainly by process hydraulics, it is necessary to consider those elements of the column periphery that lead to larger time constants than the column itself. Therefore, major elements of the column periphery, such as distributors, stirred tanks, and pipelines, have been additionally implemented into the dynamic model. The dynamic behavior of the coke gas purification process has been investigated systematically (139,140,145). For instance, local perturbations of the gas load and its composition have been analyzed. A significant dynamic parameter is represented by the liquid holdup. Figure 20 demonstrates the changes of the solvent composition after a decrease of the gas-flow rate from 67 m3/h to 36.4 m3/h and a simultaneous small increase in the liquid-flow rate. The liquid holdup of the packing section decreases, which leads to a lower conversion of the kinetically controlled reactions of CO2 and a reduction in the CO2 absorption rate. As a consequence, the solvent mole fractions of HCO3 and carbamate decreases whereas the relative fraction of HS increases. The selectivity of the absorption process toward the H2S and HCN reduction is enhanced by minimizing the liquid holdup of the column. At the same time, a larger interfacial area improves the performance of the plant. Therefore, modern industrial sour gas scrubbers should be equipped with structured packings. Figure 21 illustrates the system response after a sudden increase in the gas flow by 20% and its H2S load by 100%. As expected, the H2S load increases everywhere along the column height in the gas phase. The change is more significant in the lower part of the absorber than at the top because some additional hydrogen sulfide is absorbed. The new steady state is already achieved after 30 minutes, which justifies the implementation of dynamic models for the column periphery.


FIGURE 20 Dynamic change of solvent composition after a sudden significant decrease in the gas-flow rate and a simultaneous small increase in the liquid-flow rate.


FIGURE 21 Dynamic change of the H2S gas-phase concentration along the column after a sudden increase in the gas flow and its H2S load.

3.3.

Methyl Acetate Synthesis, Batch Distillation

3.3.1.

Chemical System

The synthesis of methyl acetate from methanol and acetic acid is a slightly exothermic equilibrium-limited liquid-phase reaction: CH 3OH (CH 3 )COOH ↔ (CH 3 )COO(CH 3 ) H 2 O H R0 4.2 kJ/mol

(R17)

The low equilibrium constant and the strongly nonideal behavior that causes the forming of the binary azeotropes methyl acetate/methanol and methyl acetate/ water make this reaction system interesting as a possible RD application (33). Therefore, methyl acetate synthesis has been chosen as a test system and investigated in a semibatch RD column. Since the process is carried out under atmospheric pressure, no side reactions in the liquid phase occur (146). 3.3.2.

Process Setup and Operation

The catalytic packing MULTIPAK® (147) applied in this case study consists of corrugated wire gauze sheets and catalyst bags of the same material assembled in alternate sequence. Sufficient mass transfer between gas and liquid phase is


guaranteed by segmentation of the catalyst bags and numerous contact spots with the wire gauze sheets. The packing was equipped with the acid ion exchange resin known as an effective catalyst for esterification processes (34,148). A batch distillation column with a diameter of 100 mm and a reactive packing height of 2 m (MULTIPAK I®) in the bottom section and an additional meter of conventional packing (ROMBOPAK 6M®) in the top section was used. The flow sheet of the column is shown in Figure 22. At first, the distillation still was charged with methanol—the low-boiling reactant—and heated under total reflux until steady-state conditions were achieved. At that moment, acetic acid—the high-boiling reactant—was fed above the reaction zone to the second distributor. After 30 min the reflux ratio was changed from infinity to 2 and the product withdrawal at the top of the column began. During the

FIGURE 22 Reactive distillation column, batch operation.


column operations, the liquid-phase concentration profiles along the column and the temperature profiles were measured. For the determination of the liquid-phase composition, two methods were applied simultaneously. On the one hand, samples were taken and analyzed by gas chromatography. On the other hand, an online NIR spectrometer was used to determine the concentration without taking any samples (149). 3.3.3.


Figures 23 and 24 show the liquid-phase compositions for, respectively, the reboiler and condenser as functions of time. After column startup, the concentration of methanol decreases continuously whereas the distillate mole fraction of methyl acetate reaches about 90%. A comparison of the rate-based simulation (with the Maxwell–Stefan diffusion equations) and experimental results for the liquid-phase composition at the column top and in the column reboiler demonstrates their satisfactory agreement (Figures 23 and 24). Figure 25 shows the simulation

FIGURE 23 Liquid mole fractions in the column reboiler: lines, simulations; dots, experiments.


FIGURE 24 Liquid mole fractions in the column condenser: lines, simulations; dots, experiments.

results obtained with different modeling approaches, after an operation time of 10,000 s. The reference model employs the rate-based approach and the Maxwell– Stefan diffusion equations. Another rate-based model assumes effective diffusion coefficients instead of the Maxwell–Stefan equations. The third model used is an equilibrium-based one. Both the reference model and effective-diffusion model show similar results. The equilibrium-stage model is only able to describe the process behavior qualitatively. This is in contrast to the reactive absorption processes (see Sections 3.1 and 3.2) and can be explained by the low reaction rate, which dominates the whole process kinetics. 3.4.

Methyl Acetate Synthesis, Steady-State Distillation

3.4.1.

Chemical System

The synthesis of methyl acetate from methanol and acetic acid analyzed in this case study is the same as described by Reaction (R17): CH 3OH (CH 3 )COOH ↔ (CH 3 )COO(CH 3 ) H 2 O


.

FIGURE 25 Axial concentration profiles for the semibatch column (t 10,000 s ) .

3.4.2.

Process Setup

The column (35) has an inner diameter of 50 mm and a total packing height of 4 m composed of a reactive section of 2 m and two nonreactive sections of 1 m each, below and above the reactive part of the column (Figure 26). The acetic acid feed is located above the catalytic packing, while methanol is fed to the column below the reactive section. A similar column design was presented in Ref. 150. 3.4.3.


A series of experiments have been performed with a stoichiometric feed ratio of acetic acid and methanol. The reflux ratio was kept constant at a value of 2.0, the feed flow rate at a value of 3.0 kg/h, while the heat duty to the reboiler was varied over a wide range. A comparison of experimental results and model prediction for the liquid-phase composition profiles along the column is given in Figure 27 for different reboiler duties (151). The theoretical values are displayed with continuous lines and empty symbols, whereas the experimental data measured along the


column are shown by the relevant filled symbols. It can be seen that the process behavior is reflected by the simulations with high accuracy. A maximum concentration of methanol and acetic acid can be observed at the respective feed locations, while methyl acetate is enriched toward the top and water toward the bottom of the column. 3.5.

Synthesis of Methyl Tertiary Butyl Ether

3.5.1.

Chemical System

The synthesis of methyl tertiary butyl ether (MTBE) is one of the most important applications of RD. MTBE is produced via an acid-catalyzed reaction between methanol and isobutylene: CH 3OH C(CH 3 ) 2 CH 2 ↔ C(CH 3 )3 OCH 3 H R0 37.7 kJ/mol

(R18)

This reaction has been extensively studied by several authors, e.g., Refs. 152–154. 3.5.2.

Process Setup

MTBE synthesis was investigated both theoretically and experimentally. Here, some results for a pilot-scale RD column at Neste Oy Engineering, Finland, are

FIGURE 26 Reactive distillation column, steady-state operation.


FIGURE 27 Liquid-phase composition profiles along the 50-mm-diameter catalytic distillation column for methyl acetate synthesis at reflux ratio of 2.0 and different reboiler duties: (a) 295W (b) 873W (c) 1161W.

presented (155). The column (used as an example here) has a catalytic section in the middle part. This catalytic section may consist either of a packed bed of catalytically active rings (91) or of structured catalytic packing (147). The rectifying and stripping sections are filled with Intalox Metal Tower Packing. The methanol feed is introduced just above the catalyst section of the column and the hydrocarbon feed just below. 3.5.3.


Figure 28 demonstrates the simulated and measured concentration profiles for the pilot test made in the column, with the reactive section filled with catalytically


active rings. In the simulations four components—methanol, isobutene, MTBE, and 1-butene—were chosen to describe the system under consideration. Here, segment 1 corresponds to the reboiler. A satisfactory agreement between calculated and measured values can be clearly observed. In Figure 29, the simulation results for the column with different reactive internals, catalytic packing MULTIPAK®, are shown. Here, 16 components were considered. Again, the liquid bulk composition profiles demonstrated in Figure 29 agree well with the experimental data. 3.6. 3.6.1.

Reactive Extraction of Zinc Chemical System

The extraction of zinc with the cation exchanger di(2-ethylhexyl)phosphoric acid, RH, is recommended by the EFCE as a test system for RE. Physical properties, handling, equilibrium data, etc. are documented on the internet (http:// www. dechema.de/Extraction, http://www.icheme.org/learning). In brief, the ion exchanger is dimeric, R 2 H 2 , in aliphatic diluents (156), and the overall reaction is then aZn 2 b R 2 H 2 ↔ Zn a R 2 a (RH)2 b2 a 2 H

(R19)

At low concentrations, polynuclear complexes do not exist (157); thus a 1: Zn 2 b R 2 H 2 ↔ ZnR 2 (RH)2 b2 2 H

(R20)

FIGURE 28 Calculated and experimental liquid compositions for experiments with catalytically active rings.


FIGURE 29 Calculated and experimental liquid compositions for experiments with catalytic structured packing.


The mass action law on the basis of concentrations then yields K eq

[ ZnR 2 ( RH) 2 b2 ][H ]2 [H ]2 D Zn [ Zn 2 ][ R 2 H 2 ]b [ R 2 H 2 ]b

(3)

where DZn denotes the partition coefficient of zinc between the organic and the aqueous phase (bar indicates organic species). In logarithmic form, the stoichiometry of the complex, b, is determined by slope analysis: log DZn b log([R 2 H 2 ]) 2 pH log K eq

(4)

As can be seen, the partition increases with increasing ion exchanger concentraeq tion, R 2 H 2 , and with pH. The determination of the equilibrium constant, K , is discussed, e.g., in Ref. 158. 3.6.2.

Process Setup

Apart from the nuclear industry, the most frequently installed pieces of process equipment are mixer-settler cascades. The advantage is the easy control of each stage regarding the pH value, selection of the phase to be dispersed, etc. (12,15). The disadvantages are the high investment costs and large solvent inventory. Nowadays, modern designed extractants (with fast kinetics) allow a process design with highly efficient extraction columns (159). Applications are found in the chemical industry, e.g., with sulfonic acid extraction (160). The first application on a big scale in hydrometallurgy was reported in 1997 at WMC Olympic Dam, Australia, where 10 pulsed Batman columns (0.5 to 3 m in diameter and 35 m tall) were used for uranium recovery. An increased recovery (from 90% to 97%) was found after replacing the formerly used mixer-settler units. The RE of zinc is reported in detail when using an RDC, including the discussion of the stripping process to regenerate the ion exchanger for cyclic reuse (161–163). 3.6.3.


A comparison of predicted and experimental mass transfer coefficients is given in Figure 30. The simulated overall mass transfer coefficient originates from a model that is a combination of the microkinetic reaction according to Eq. (B9) and eddy diffusion according to Eq. (B11) (see Appendix B). Figure 30 shows that the mass transfer coefficient at higher concentrations is generally underestimated, thus including some safety value for the process design. As discussed in detail in Ref. 56, combinations of microkinetics [Eq. (B9)] with other eddy diffusion correlations instead of Eq. (B11) (e.g., Refs. 164 and 165) are also appropriate to describe the system. In contrast to this, the combination of microkinetics with molecular diffusion concepts fails, and the same is true when the equilibrium


FIGURE 30 Comparison of experimental and simulated overall mass transfer coefficients at different initial zinc concentrations and droplet diameters (1, 2, or 3 mm ) .

approach is used neglecting the kinetics rate law. The results of the column simulations are discussed in Refs. 162 and 163, and a special discussion on contamination effects is given in Ref. 166. 4.

SUMMARY AND OUTLOOK

This chapter concerns the most important reactive separation processes: reactive absorption, reactive distillation, and reactive extraction. These operations combining the separation and reaction steps inside a single column are advantageous as compared to traditional unit operations. The three considered processes are similar and at the same time very different. Therefore, their common modeling basis is discussed and their peculiarities are illustrated with a number of industrially relevant case studies. The theoretical description is supported by the results of laboratory-, pilot-, and industrial-scale experimental investigations. Both steady-state and dynamic issues are treated; in addition, the design of column internals is addressed. Reactive absorption, reactive distillation, and reactive extraction occur in multicomponent multiphase fluid systems, and thus a single modeling framework for these processes is desirable. In this respect, different possible ways to build such a framework are discussed, and it is advocated that the rate-based approach provides the most rigorous and appropriate way. By this approach, direct consideration


of the diffusional and reaction kinetics is realized. Special attention is paid to the application of CFD, which could become a powerful theoretical tool to predict the flow behavior for different column units and internals geometries for engineering applications. In particular, CFD can play an outstanding role in the development of the column internals for reactive separations. Fundamental advances in the understanding of the underlying physicochemical phenomena when coupled with CFD would go a long way toward support of reactive separation technology. The modeling of RA is illustrated by the absorption of NOx and by the coke gas purification process. The first case is modeled by using an analytical treatment of the film phenomena, whereas the second one is solved by a purely numerical technique. The simulation results are compared with the experimental data obtained at an industrial absorption plant consisting of eight units with pump around (NOx) and at a pilot column for the ammonia hydrogen sulfide circulation scrubbing process (coke gas purification). For the latter case, both steady-state and dynamic conditions are considered. The comparison results, on the one hand, demonstrate a good agreement between the rate-based simulations and experimental data, and, on the other hand, warn of using the equilibrium approach, which appears completely inappropriate in the case of complex finite-rate reactions. The modeling of RD processes is illustrated with the heterogenously catalyzed synthesis of methyl acetate and MTBE. The complex character of reactive distillation processes requires a detailed mathematical description of the interaction of mass transfer and chemical reaction and the dynamic column behavior. The most detailed model is based on a rigorous dynamic rate-based approach that takes into account diffusional interactions via the Maxwell–Stefan equations and overall reaction kinetics for the determination of the total conversion. All major influences of the column internals and the periphery can be considered by this approach. As an application example, the dynamic model was used for the simulation of the steady-state and semibatch production of methyl acetate, performed in a packed column with a catalytic packing. For the model validation, several experiments were carried out in a pilot-plant column. For the investigated operation range, the simulation results are in good agreement with the experimental data. The use of this model for model-based process control calls for suitable model reductions without a significant decrease in the predictivity. For the methyl acetate process, a simplified description of the mass transfer using effective diffusion coefficients and neglecting diffusional interactions seems to be sufficient. On the other hand, a detailed description of the reaction, including the specific phenomena of the heterogeneous catalysis by an adequate consideration of the solid phase, is required for the predictive simulation of even more complex systems, including side and consecutive reactions. Optimal functioning of reactive distillation depends on careful process design, with appropriately selected column internals, feed locations, and catalyst placement. Greater understanding of the general and particular features of the process behavior is equally essential.


The modeling of RE processes is strongly related to the knowledge of the reaction equilibrium and kinetics and mass transfer regime. The latter is decisive for column simulations, whereas eddy diffusivity concepts often have to be used. The parameters can easily be obtained in small-scale laboratory devices, with a minimum of substance involved. A real challenge is the correct hydrodynamic description of the two-phase flow. The assumption of plug flow gives acceptable results with column diameters smaller than 0.1 m. The commonly applied axial dispersion model or back-mixing model uses one parameter to account for nonideal flow in each phase. Here the dispersed phase is considered to be pseudo-continuous and monodisperse. Droplet population models taking into account the dynamic processes of coalescence and breakup of droplets should give a more realistic picture and thus a more firm design of a process. The use of CFD calculations in liquid– liquid dispersed-phase flow is limited to single-droplet flow or low column holdup. The simulation of large industrial columns especially is not feasible nowadays. Some important general aspects of rate-based modeling as well as further peculiarities of the specific process applications and the different solution strategies are given in Appendices A and B. The key reactive separation topics to be addressed in the near future are a proper hydrodynamic modeling for catalytic internals, including residence time distribution account and scale-up methodology. Further studies on the hydrodynamics of catalytic internals are essential for a better understanding of RSP behavior and the availability of optimally designed catalytic column internals for them. In this regard, the methods of computational fluid dynamics appear very helpful. The development of new methodologies enabling the creation of intelligent, tailor-made column internals and consequent RSP optimization constitutes one of the burning present-day challenges. Such a development is already in progress in some European research projects. Despite the recent rapid development of computer technology and numerical methods, the rate-based approach in its current realization still often requires a significant computational effort, with related numerical difficulties. This is one of the reasons the application of rate-based models to industrial tasks is rather limited. Therefore, further work is required in order to bridge this gap and provide chemical engineers with reliable, consistent, robust, and comfortable simulation tools for reactive separation processes. ACKNOWLEDGMENTS We would like to thank our colleagues at the Chair of Fluid Separation Processes, Dortmund University, and all other project partners who have been involved in the research activities. We are also grateful to the German Research Foundation (DFG, Grants No. Schm 808/5-1, Ba 1569/2-1 2-2, Ba 1569/6-1), the Volkswagen Foundation (Project No. I/70 875, 876, 877), the European


Commission (BRITE-EURAM program, CEC Project No. BE95-1335), the German Federal Ministry of Education and Research (BMBF, Project No. 03C0306), the Foundation “Rheinland-Pfalz für Innovation” (836-386261/193), as well as BASF AG, Bayer AG, Axiva GmbH, Degussa. NOMENCLATURE aI As B c CIP dC di dp D Dax Deff DZn E E F FC G h H R0 ky Ki Keq [K] l L n Ni Q R R Re Sc Sh t T uL U

specific gas–liquid interfacial area column cross section liquid load molar concentration adjustable parameter, Eq. (B10) column diameter generalized driving force for component i droplet diameter Maxwell–Stefan diffusion coefficient axial dispersion coefficient effective diffusion coefficient partition coefficient of zinc length-specific energy holdup dimensionless residence time distribution Faraday’s constant gas capacity factor gas molar flow rate molar enthalpy reaction enthalpy overall mass transfer coefficients distribution coefficient equilibrium constant reaction matrix [Eq. (B1)] axial coordinate liquid molar flow rate number of components of mixture molar flux of component i heat flux total component reaction rate column vector with elements Ri gas constant Reynolds number Schmidt number Sherwood number time temperature liquid velocity length-specific molar holdup


m2/m3 m2 m3/(m2s) mol/m3 m 1/m m m2/s m2/s m2/s J/m 9.65 104 C/mol Pa0.5 mol/s J/mol J/mol m/s

1/s m mol/s mol/(m2s) W/m2 mol/m3s mol/m3s 8.3144 J /(mol K)

s K m/s mol/m

w xi x yi z zi

terminal velocity first fluid-phase (liquid) mole fraction of component i column vector with elements xi second fluid-phase (gas, vapor, or liquid) mole fraction of component i film coordinate ionic charge of component i

m/s mol/mol mol/mol mol/mol m

Greek Letters r c d

film thickness dimensionless film coordinate forward-reaction constant backward-reaction constant thermal conductivity chemical potential dynamic viscosity of continuous phase dynamic viscosity of dispersed phase volumetric holdup electrical potential

Subscripts G i, j L t

gas or second fluid phase component/reaction indices liquid phase mixture property

Superscripts B I

bulk phase phase interface

Abbreviations ADM CD PDE RA RD RE RH RSP

axial dispersion model catalytic distillation piston flow model with axial dispersion and mass exchange reactive absorption reactive distillation reactive extraction di(2-ethylhexyl) phosphoric acid reactive separation process


m m3/2/(mol1/2s) s1 W/(m K) J/mol Pa s Pa s m3/m3 V

REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.

20. 21. 22. 23. 24. 25.

Noble T. Multifunctional equipment—all for one, and one for all? Chem Eng Prog 2001; 97(11):10–13. Doherty MF, Buzad G. Reactive distillation by design. Trans IChemE 1992; 70: 448–458. Zarzycki R, Chacuk A. Absorption: Fundamentals and Applications. Oxford: Pergamon Press, 1993. Agar DW. Multifunctional reactors: old preconceptions and new dimensions. Chem Eng Sci 1999; 54:1299–1305. Kelkar VV, Ng KM. Design of reactive crystallization systems incorporating kinetics and mass-transfer effects. AIChE J 1999; 45:69–81. Stankiewicz AI, Moulijn JA. Process intensification: transforming chemical engineering. Chem Eng Prog 2000; 96(1):22–34. Fricke J, Schmidt-Traub H, Kawase M. Chromatographic reactors. In: Ullmann’s Encyclopedia of Industrial Chemistry. Weinheim, Germany: Wiley-VCH, 2001. Bart HJ. Reactive Extraction. Berlin: Springer, 2001. Noeres C, Kenig EY, Górak A. Modelling of reactive separation processes: reactive absorption and reactive distillation. Chem Eng Process 2003; 42:157–178. Hatcher WJ. Reaction and mass transport in two-phase reactors. In: Cheremisinoff NP, ed. Handbook of Heat and Mass Transfer. Vol. 2. Houston: Gulf, 1986:837–868. Ritcey GM, Ashbrook AW. Solvent Extraction Principles and Applications to Process Metallurgy. Vols. I, II. New York: Elsevier, 1979. Lo TC, Baird MHI, Hanson C, eds. Handbook of Solvent Extraction. New York: Wiley, 1983. Thornton JD, ed. Science and Practice of Liquid–Liquid Extraction. Vols. I, II. Oxford: Oxford University Press, 1992. Marr R, Bart HJ. Metallsalzextraktion. Chem Ing Tech 1982; 54:119–129. Godfrey JC, Slater MJ, eds. Liquid–Liquid Extraction Equipment. Chichester, UK: Wiley, 1994. Marcus Y, Sengupta AK, eds. Ion Exchange and Solvent Extraction. Vol. 15. New York: Marcel Dekker, 2002. Sherwood TK, Pigford RL, Wilke CR. Mass Transfer. New York: McGraw-Hill, 1975. Sherwood TK, Pigford RL. Absorption and Extraction. New York: McGraw-Hill, 1952. Counce RM, Perona JJ. Designing packed-tower wet scrubbers: emphasis on nitrogen oxides. In: Cheremisinoff NP, ed. Handbook of Heat and Mass Transfer. Vol. 2. Houston: Gulf, 1986:953–966. Sattler K. Thermische Trennverfahren. Weinheim, Germany: VCH, 1988. Strelzoff S. Choosing the optimum CO2-removal system. Chem Eng 1975; 82: 115–120. Falbe J. Chemierohstoffe aus Kohle. Stuttgart: Georg-Thieme-Verlag, 1977. Kohl AL, Riesenfeld FC. Gas Purification. Houston: Gulf, 1985. Yu WC, Astarita G. Selective absorption of hydrogen sulphide in tertiary amine solutions. Chem Eng Sci 1987; 42:419–424. Thielert H. Simulation und Optimierung der Kokereigaswäsche. Ph.D. dissertation, Technical University Berlin, Berlin, 1997.


26. 27. 28. 29.

30. 31.

32.

33. 34. 35. 36. 37. 38.

39.

40. 41. 42. 43.

44.

45.

Kobus A. Ein heuristisch-numerischer Ansatz zum systematischen Entwurf und Design von Absorptionsverfahren. Düsseldorf: VDI, 1999. Danckwerts PV. Gas–Liquid Reactions. New York: McGraw-Hill, 1970. Astarita G, Savage DW, Bisio A. Gas Treating with Chemical Solvents. New York: Wiley, 1983. Pal SK, Sharma MM. Solvay-type process: absorption of CO2 in ammoniated aqueous and nonaqueous solutions of NaNO3, Ca(NO3)2, and KCl. Ind Eng Chem Proc Des Dev 1983; 22:76–79. Sauchel V. Chemistry and Technology of Fertilizers. New York: Reinhold, 1960. Gandhidasan P. Heat and mass transfer in solar regenerators. In: Cheremisinoff NP, ed. Handbook of Heat and Mass Transfer. Vol. 2. Houston: Gulf, 1986: 1475–1499. Kenig EY, Kholpanov LP, Katysheva LI, Markish IH, Malyusov VA. Calculation of two-phase nonisothermal absorption in a liquid film in downward cocurrent flow. Theor Found Chem Eng 1985; 19:97–102. Agreda VH, Partin LR, Heise WH. High-purity methyl acetate via reactive distillation. Chem Eng Prog 1990; 86(2):40–46. Kreul LU, Górak A, Dittrich C, Barton PI. Dynamic catalytic distillation: advanced simulation and experimental validation. Computers Chem Eng 1998; 22:371–378. Górak A, Hoffmann A. Catalytic distillation in structured packings: methyl acetate synthesis. AIChE J 2001; 47:1067–1076. Komatsu H, Holland CD. A new method of convergence for solving reacting distillation problems. J Chem Eng Japan 1977; 4:292–297. Hartig H, Regner H. Verfahrenstechnische Auslegung einer Veresterungskolonne. Chem Ing Tech 1971; 43:1001–1007. Davies B, Jeffreys GV. The continuous trans-esterification of ethyl alcohol and butyl acetate in a sieve-plate column. Part III: trans-esterification in a six-plate sieve-tray column. Trans IChemE 1973; 51:275–280. Luo HP, Xiao WD. A reactive distillation process for a cascade and azeotropic reacting system: carbonylation of ethanol with dimethyl carbonate. Chem Eng Sci 2001; 56:403–410. Fuchigami Y. Hydrolysis of methyl acetate in distillation column packed with reactive packing of ion exchange resin. J Chem Eng Japan 1990; 23:354–358. DeGarmo JL, Parulekar VN, Pinjala V. Consider reactive distillation. Chem Eng Prog 1992; 88(3):42–50. Sundmacher K, Hoffmann U. Activity evaluation of a catalytic distillation packing for MTBE production. Chem Eng Technol 1993; 16:279–289. Thiel C. Modellbildung, Simulation, Design und experimentelle Validierung von heterogen katalysierten Reaktivdestillationsprozessen zur Synthese der Kraftstoffether MTBE, ETBE und TAME. Ph.D. dissertation, Technical University of Clausthal, Clausthal, Germany, 1997. Oost C. Ein Beitrag zum Umweltschutz: Reaktionstechnische Untersuchungen zur Flüssigphasesynthese des Antiklopfmittels TAME. Ph.D. dissertation, Technical University of Clausthal, Clausthal, Germany, 1995. Shoemaker JD, Jones EM. Cumene by catalytic distillation. Hydroc Proc 1987; 66:57–58.


46.

47. 48.

49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60.

61. 62.

63.

64.

65. 66.

Podrebarac GG, Ng FTT, Rempel GL. The production of diacetone alcohol with catalytic distillation. Part I: catalytic distillation experiments. Chem Eng Sci 1998; 53:1067–1075. Ung S, Doherty MF. Vapor–liquid equilibrium in systems with multiple chemical reactions. Chem Eng Sci 1995; 50:23–48. Müller D, Schäfer JP, Leimkühler HJ. The economic potential of reactive distillation processes exemplified by silane production. Proceedings of VDI-GVC, DECHEMA and EFCE Meeting, Section “Thermal Separations,” Bamberg, Germany, 2001. Ciric AR, Miao P. Steady-state multiplicities in an ethylene glycol reactive distillation column. Ind Eng Chem Res 1994; 33:2738–2748. Belson DJ. A distillation method of aromatic nitration using azeotropic nitric acid. Ind Eng Chem Res 1990; 29:1562–1565. Fair JR. Design aspects for reactive distillation. Chem Eng 1998; 105(11):158–162. Rock K, Gildert GR, McGuirk T. Catalytic distillation extends its reach. Chem Eng 1997; 104(7):78–80. Marion MC, Viltard JC, Travers P, Harter I, Forestiere A. U.S. Patent 5,776,320, Institute Français du Petrol, 1998. Klocker H, Bart HJ, Marr R, Müller H. Mass transfer based on chemical potential theory: ZnSO4/H2SO4/D2EHPA. AIChE J 1997; 43:2479–2487. Bart HJ, Rousselle HP. Microkinetics and reaction equilibria in the system ZnSO4/ D2EHPA/isododecane. Hydrometallurgy 1999; 51:285–298. Mörters M, Bart HJ. Mass transfer into droplets in reactive extraction. Chem Eng Process, 2003, in press. Taylor R, Krishna R. Multicomponent Mass Transfer. New York: Wiley, 1993. Kenig EY, Górak A. A film model based approach for simulation of multicomponent reactive separation. Chem Eng Process 1995; 34:97–103. Kenig EY. Modeling of Multicomponent Mass Transfer in Separation of Fluid Mixtures. Düsseldorf: VDI, 2000. Sundmacher K, Hoffmann U. Multicomponent mass and energy transport on different length scales in a packed reactive distillation column for heterogeneously catalyzed fuel ether production. Chem Eng Sci 1994; 49:4443–4464. Krishna R, Wesselingh JA. The Maxwell–Stefan approach to mass transfer. Chem Eng Sci 1997; 52:861–911. Higler AP, Krishna R, Taylor R. Nonequilibrium modeling of reactive distillation: a dusty fluid model for heterogeneously catalyzed processes. Ind Eng Chem Res 2000; 39:1596–1607. Yuxiang Z, Xien X. Study on catalytic distillation processes. Part II. Simulation of catalytic distillation processes—quasi-homogenous and rate-based model. Trans IChemE 1992; 70:465–470. Sundmacher K, Uhde G, Hoffmann U. Multiple reactions in catalytic distillation processes for the production of fuel oxygenates MTBE and TAME: analysis by rigorous model and experimental validation. Chem Eng Sci 1999; 54:2839–2847. Mackowiak J. Fluiddynamik in Kolonnen mit modernen Füllkörperpackungen und Packungen für Gas/Flüssigkeitssysteme. Frankfurt am Main: Sauerländer, 2001. Billet R. Packed Towers. Weinheim, Germany: VCH, 1995.


67.

68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89.

Kenig EY, Klöker M, Egorov Y, Menter F, Górak A. Towards improvement of reactive separation performance using computational fluid dynamics. Proceedings of International Symposium on Multifunctional Reactors (ISMR-2), Nuremberg, 2001. Doraiswamy LL, Sharma MM. Heterogeneous Reactions: Analysis, Examples and Reactor Design. New York: Wiley, 1984. Westerterp KR, van Swaaij WPM, Beenackers AACM. Chemical Reactor Design and Operation. Chichester, UK: Wiley, 1984. Chen CC, Evans LB. A local composition model for the excess Gibbs energy of aqueous electrolyte systems. AIChE J 1986; 32:444–459. Mock B, Evans LB, Chen CC. Thermodynamic representation of phase equilibria of mixed-solvent electrolyte systems. AIChE J 1986; 32:1655–1664. Reid R, Prausnitz JM, Poling BE. The Properties of Gases and Liquids. New York: McGraw-Hill, 1987. Froment GF, Bischoff KB. Chemical Reactor Analysis and Design. New York: Wiley, 1990. Taylor R, Krishna R. Modeling reactive distillation. Chem Eng Sci 2000; 55: 5183–5229. Sorel E. La rectification de l’alcool. Paris: Gauthiers—Villais et Fils, 1893. Henley EJ, Seader JD. Equilibrium-Stage Separation Operations in Chemical Engineering. New York: Wiley, 1981. van Krevelen DW, Hoftijzer PJ. Kinetics of gas–liquid reactions—Part I: General theory. Recl Trav Chim Pays-Bas 1948; 67:563–586. Holma H, Sohlo J. A mathematical model of an absorption tower of nitrogen oxides in nitric aid production. Computers Chem Eng 1979; 3:135–141. Carta G. Scrubbing of nitrogen oxides with nitric acid solutions. Chem Eng Commun 1986; 42:157–170. Suzuki I, Yagi H, Komatsu H, Hirata M. Calculation of multicomponent distillation accompanied by a chemical reaction. J Chem Eng Japan 1971; 4:26–33. Chang YA, Seader JD. Simulation of continuous reactive distillation by homotopy continuation method. Computers Chem Eng 1988; 12:1243–1255. Stewart WE, Prober R. Matrix calculation of multicomponent mass transfer in isothermal systems. Ind Eng Chem Fundam 1964; 3:224–235. Toor HL. Solution of the linearized equations of multicomponent mass transfer. AIChE J 1964; 10:448–455, 460–465. Górak A. Simulation thermischer Trennverfahren fluider Vielkomponentengemische. In: Schuler G, ed. Prozesssimulation. Weinheim, Germany: VCH, 1995:349–408. Seader JD. The rate-based approach for modeling staged separations. Chem Eng Prog 1989; 85(10):41–49. Katti SS. Gas–liquid–solid systems: an industrial perspective. Trans IChemE 1995; 73(part A):595–607. Lewis WK, Whitman WG. Principles of gas absorption. Ind Eng Chem 1924; 16:1215–1220. Higbie R. The rate of absorption of a pure gas into a still liquid during short periods of exposure. Trans Am Inst Chem Engrs 1935; 31:365–383. Hirschfelder JO, Curtiss CF, Bird RB. Molecular Theory of Gases and Liquids. New York: Wiley, 1964.


90.

91. 92.

93.

94.

95. 96.

97. 98.

99. 100. 101. 102.

103. 104. 105.

106.

107.

Pelkonen S, Kaesemann R, Górak A. Distillation lines for multicomponent separation in packed columns—theory and comparison with experiment. Ind Eng Chem Res 1997; 36:5392–5398. Sundmacher K, Hoffmann U. Development of a new catalytic distillation process for fuel ethers via a detailed nonequilibrium model. Chem Eng Sci 1996; 51:2359–2368. Davidson DL. The enterprise-wide application of computational fluid dynamics in the chemicals industry. Proceedings of the 6th World Congress of Chemical Engineering, Melbourne, 2001. Sommerfeld M. Modellierung und numerische Berechnung von partikelbeladenen turbulenten Strömungen mit Hilfe des Euler/Lagrange-Verfahrens. Aachen, Germany: Shaker Verlag, 1996. Scheuerer G. Solution algorithms and implementations strategies for Eulerian– Eulerian multiphase-flow models. Proceedings of ACFD 2000 International Conference on Applied Computational Fluid Dynamics, Beijing, 2000. Kuipers JAM, van Swaaij WPM. Application of computational fluid dynamics to chemical reactor engineering. Rev Chem Eng 1997; 13:1–118. Sokolichin A, Eigenberger G, Lapin A, Lübbert A. Direct numerical simulation of gas–liquid two-phase flows. Euler/Euler versus Euler/Lagrange. Chem Eng Sci 1997; 52:611–626. Krishna R, van Baten JM. Simulating the motion of gas bubbles in a liquid. Nature 1999; 398:208. Krishna R, Urseanu MI, van Baten JM, Ellenberger J. Wall effects on the rise of single gas bubbles in liquids. Int Commun Heat Mass Transfer 1999; 26: 781–790. Ding J, Gidaspow D. A bubbling fluidization model using the kinetic theory of granular flow. AIChE J 1990; 36:523–538. Fan LS, Zhu C. Principles of Gas–Solid Flows. Cambridge: Cambridge University Press, 1998. Marschall KJ, Mleczko L. CFD modeling of an internally circulating fluidized-bed reactor. Chem Eng Sci 1999; 54:2085–2093. van Wachem BGM, Schouten JC, Krishna R, van den Bleek CM. Validation of the Eulerian simulated dynamic behavior of gas–solid fluidized beds. Chem Eng Sci 1999; 54:2141–2149. Grienberger J, Hofmann H. Investigations and modelling of bubble columns. Chem Eng Sci 1992; 47:2215–2220. Lapin A, Lübbert A. Numerical simulation of the dynamics of two-phase gas– liquid flows in bubble columns. Chem Eng Sci 1994; 49:3661–3674. Sokolichin A, Eigenberger G. Gas–liquid flow in bubble columns and loop reactors. Part I. Detailed modelling and numerical simulation. Chem Eng Sci 1994; 49: 5735–5746. Grevskott S, Sannaes BH, Dudukovic MP, Hjarbo KW, Svendsen HF. Liquid circulation, bubble-size distributions, and solids movement in two- and three-phase columns. Chem Eng Sci 1996; 51:1703–1713. Lin TJ, Reese J, Hong T, Fan LS. Quantitative analysis and computation of twodimensional bubble columns. AIChE J 1996; 42:301–318.


108.

109. 110. 111. 112.

113. 114. 115. 116. 117. 118. 119.

120. 121.

122.

123. 124.

125.

126.

Delnoij E, Lammers FS, Kuipers JAM, van Swaaij WPM. Dynamic simulation of dispersed gas–liquid two-phase flow using a discrete bubble model. Chem Eng Sci 1997; 52:1429–1458. Jakobsen HA, Sannaes BH, Grevskott S, Svendsen HF. Modeling of bubble-driven vertical flows. Ind Eng Chem Res 1997; 36:4052–4074. Sanyal J, Vasquez S, Roy S, Dudukovic MP. Numerical simulation of gas–liquid dynamics in cylindrical bubble-column reactors. Chem Eng Sci 1999; 54: 5071–5083. Thakre SS, Joshi JB. CFD simulation of bubble-column reactors: importance of drag-force formulation. Chem Eng Sci 1999; 54:5055–5060. Krishna R, Urseanu MI, van Baten JM, Ellenberger J. Influence of scale on the hydrodynamics of bubble columns operating in the churn-turbulent regime: experiments vs. Eulerian simulations. Chem Eng Sci 1999; 54:4903–4911. Pfleger D, Becker S. Modelling and simulation of the dynamic flow behavior in a bubble column. Chem Eng Sci 2001; 56:1737–1747. Fischer CH, Quarini GL. Three-dimensional heterogeneous modeling of distillation tray hydraulics. Proceedings of AIChE Annual Meeting, Miami Beach, 1998. Metha B, Chuang KT, Nandakumar K. Model for liquid-phase flow on sieve trays. Trans IChemE 1998; 76:843–848. Krishna R, van Baten JM, Ellenberger J, Higler AP, Taylor R. CFD simulations of sieve-tray hydrodynamics. Trans IChemE 1999; 77:639–646. Lju CJ, Yuan XG, Yu KT, Zhu XJ. A fluid-dynamic model for flow pattern on a distillation tray. Chem Eng Sci 55:2287–2294, 2000. van Baten JM, Krishna R. Modelling sieve-tray hydraulics using computational fluid dynamics. Chem Eng J 2000; 77:143–151. van Baten JM, Ellenberger J, Krishna R. Hydrodynamics of reactive distillation tray column with catalyst-containing envelopes: experiments vs. CFD simulations. Catal Today 2001; 66:233–240. van Gulijk C. Using computational fluid dynamics to calculate transversal dispersion in a structured packed bed. Computers Chem Eng 1998; 22:767–770. Higler AP, Krishna R, Ellenberger J, Taylor R. Countercurrent operation of a structured catalytically packed-bed reactor: liquid-phase mixing and mass transfer. Chem Eng Sci 1999; 54:5145–5152. van Baten JM, Ellenberger J, Krishna R. Radial and axial dispersion of the liquid phase within KATAPAK-S structure: experiments vs. CFD simulations. Chem Eng Sci 2001; 56:813–821. Rieger R, Weiss C, Wigley G, Bart HJ, Marr R. Investigating the process of liquid– liquid extraction by means of CFD. Computers Chem Eng 1996; 20:1467–1475. Zahiarieva A, Masberat O, Aoun Nabli M, Giraud P, Gourdon C. Experimental and numerical 2D turbulent field in a pulsed extraction column. In: Shallcross DC et al., eds. Value Adding Through Solvent Extraction. Vol. 2. Melbourne: University of Melbourne, 1996:1233–1238. Fei WY, Wang YD, Wan YK. Physical modelling and numerical simulation of velocity fields in rotating-disc contactor via CFD simulation and LDV measurement. Chem Eng J 2000; 78:131–139. Modes G, Bart HJ. CFD simulations of nonideal dispersed-phase flow in stirred extraction columns. Chem Eng Technol 2002; 24:1242–1245.


127. 128. 129. 130.

131. 132.

133. 134. 135. 136.

137. 138. 139. 140. 141. 142.

143. 144. 145.

146.

Bart HJ. From single droplet to column design. Keynote lecture at 6th World Congress on Chemical Engineering, Melbourne, 2001. Suchak NJ, Jethani KR, Joshi JB. Modeling and simulation of NOx absorption in pilot-scale packed columns. AIChE J 1991; 37:323–339. Wiesner U, Wittig M, Górak A. Design and optimization of a nitric acid recovery plant from nitrous waste gases. Computers Chem Eng 1996; 20:S1425–S1430. Ulrichs K, Laue W, Renovanz HD. Salpetersäure und salpetrige Säure, Stickstoffoxide, Nitrate, Nitrite. In: Ullmanns Encyclopädie der technischen Chemie. Vol. 20. Weinheim, Germany: VCH, 1981:305–362. Wiesner U, Górak A, Kenig EY, Steude H, Börger GG. Modelling of absorption of nitrous waste gases. IChemE Symp Ser 1997; 142(1):323–333. Kafarov VV, Shestopalov EA, Novikov EA, Belkov VP. Mathematical model of absorption of NOx in production of weak nitric acid. Soviet Chem Ind 1975; 7: 1224–1226. Kolev N. Wirkungsweise von Füllkörperschüttungen. Chem Ing Tech 1976; 48: 1105–1112. Brauer H. Stoffaustausch einschliesslich chemischer Reaktionen. Aarau: Sauerländer, 1971. Kenig EY, Wiesner U, Górak A. Modeling of reactive absorption using the Maxwell–Stefan equations. Ind Eng Chem Res 1997; 36:4425–4434. Danckwerts PV, Sharma MM. The absorption of carbon dioxide into solutions of alkalis and amines (with some notes on hydrogen sulfide and carbonyl sulfide). The Chemical Engineer 1966; (July–Aug):244–280. Maurer G. On the solubility of volatile weak electrolytes in aqueous solutions. ACS Symp Ser 1980; 133:139–172. Tippmer K. Progress in the desulphurization of coke oven gas and utilization of H2S-Still Otto’s SOLOX process. Coke Making Int 1994; 6:32–43. Schneider R, Kenig EY, Górak A. Dynamic modelling of reactive absorption with the Maxwell–Stefan approach. Trans IchemE 1999; 77(part A):633–638. Kenig EY, Schneider R, Górak A. Rigorous dynamic modelling of complex reactive absorption processes. Chem Eng Sci 1999; 54:5195–5203. Schneider R, Górak A. Model optimization for the dynamic simulation of reactive absorption processes. Chem Eng Technol 2001; 24:979–989. Schneider R, Kenig EY, Górak A. Complex reactive absorption processes: model optimization and dynamic column simulation. Proceedings of ESCAPE 11 European Symposium on Computer-Aided Process Engineering, Kolding, Denmark, 2001. Kenig EY, Schneider R, Górak A. Reactive absorption: optimal process design via optimal modelling. Chem Eng Sci 2001; 56:343–350. Pantelides CC. The consistent initialization of differential-algebraic systems. SIAM J Sci Stat Comp 1988; 9:213–231. Schneider R, Kenig EY, Górak A. Dynamische Simulation reaktiver Absorptionsprozesse am Beispiel einer Sauergaswäsche: Modellentwicklung, -analyse und -optimierung. Chem Ing Tech 2000; 72:1224–1229. Song W, Venimadhavan G, Manning JM, Malone MF, Doherty MF. Measurement of residue curve maps and heterogeneous kinetics in methyl acetate synthesis. Ind Eng Chem Res 1998; 37:1917–1928.


147. 148. 149. 150. 151. 152.

153.

154. 155.

156. 157.

158. 159. 160. 161. 162. 163. 164. 165. 166.

Górak A, Kreul LU, Skowronski M. Strukturierte Mehrzweckpackung. German patent 19701045 A1, 1998. Schneider R, Noeres C, Kreul LU, Górak A. Dynamic modeling and simulation of reactive batch distillation. Computers Chem Eng 1999; 23:S423–S426. Schneider R, Noeres C, Kreul LU, Górak A. Dynamic modeling and simulation of reactive batch distillation. Computers Chem Eng 2001; 25:169–176. Krafczyk J, Gmehling J. Einsatz von Katalysatorpackungen für die Herstellung von Methylacetat durch reaktive Rektifikation. Chem Ing Tech 1994; 66:1372–1375. Hoffmann A, Górak A. Scale-up of reactive distillation columns with catalytic packing. Chem Eng Process (submitted). Gicquel A, Torck B. Synthesis of methyl tertiary butyl ether catalyzed by ionexchange resin. Influence of methanol concentration and temperature. J Catal 1983; 83:9–18. Rehfinger A, Hoffmann U. Kinetics of methyl tertiary butyl ether liquid-phase synthesis catalyzed by ion exchange resin. I. Intrinsic rate expression in liquid-phase activities. Chem Eng Sci 1990; 45:1605–1617. Parra D, Tejero J, Cunill F, Iborra M, Izquierdo JF. Kinetic study of MTBE liquidphase synthesis using C4 olefin cut. Chem Eng Sci 1994; 49:4563–4578. Kenig EY, Jakobsson K, Banik P, Aittamaa J, Górak A, Koskinen M, Wettmann P. An integrated tool for synthesis and design of reactive distillation. Chem Eng Sci 1999; 54:1347–1352. Kolarik Z. Critical evaluation of some equilibrium constants involving acidic organophosphorous extractants. Pure Appl Chem 1982; 54:2593–2674. Kolarik Z, Grimm R. Acidic organophosphorous extractants XXIV: the polymerization behavior of Cu(II) Cd(II), Zn(II) and Co(II) complexes of D2EHPA in fully loaded organic phases. J Inorg Nucl Chem 1976; 38:1721–1727. Mörters M, Bart HJ. Extraction equilibria of zinc with bis(2-ethyl-hexyl) phosphoric acid. J Chem Eng Data 2000; 45:82–85. Bart HJ. Reactive extraction of acids and metals—the state of the art in column design. Chem Eng Sci 2002; 57:1633–1637. Schäfer JP, Sluyts D. Aufarbeitung eines Prozessabwassers mittels Reaktivextraktion. In: VDI-Jahrbuch. Düsseldorf: VDI, 1996:285–295. Veglio F, Slater MJ. Design of liquid–liquid extraction columns for the possible test system Zn/D2EHPA in n-dodecane. Hydrometallurgy 1996; 42:177–195. Mansur MB, Slater MJ, Biscaia EC. Kinetic analysis of the reactive liquid–liquid test system ZnSO4/D2EHPA/n-heptane. Hydrometallurgy 2002; 63:107–116. Mansur MB, Slater MJ, Biscaia EC. Equilibrium analysis of the reactive liquid– liquid test system ZnSO4/D2EHPA/n-heptane. Hydrometallurgy 2002; 63:117–126. Steiner L. Mass-transfer rates from single drops and drop swarms. Chem Eng Sci 1986; 41:1979–1986. Slater M. A combined model of mass transfer coefficient for contaminated drop liquid– liquid systems. Can J Chem Eng 1995; 73:462–469. Brodkorb M, Slater MJ. Multicomponent and contamination effects on mass transfer in a liquid–liquid extraction rotating disk contactor. Trans IChemE 2001; 79(part A):335–346.


167.

168.

169. 170. 171. 172.

173. 174.

175. 176. 177. 178. 179.

180. 181. 182. 183. 184. 185. 186.

Kenig EY, Butzmann F, Kucka L, Górak A. Comparison of numerical and analytical solutions of a multicomponent reaction–mass transfer problem in terms of the film model. Chem Eng Sci 2000; 55:1483–1496. Noeres C, Benvenuti C, Hoffmann A, Górak A. Reactive distillation: nonideal-flow behavior of the liquid phase in structured catalytic packings. Proceedings of International Symposium on Multifunctional Reactors (ISMR-2), Nuremberg, 2001. Higler AP, Krishna R, Taylor R. Nonequilibrium-cell model for multicomponent (reactive) separation processes. AIChE J 1999; 45:2357–2370. Danckwerts PV. Continuous flow systems—distribution of residence times. Chem Eng Sci 1953; 2:1–13. van Swaaij WPM, Harpentier JC, Villermaux J. Residence time distribution in the liquid phase of trickle flow in packed columns. Chem Eng Sci 24:1083–1095, 1969. Noeres C, Górak A. Zur dynamischen Modellierung der Reaktivrektifikation unter Berücksichtigung des Verweilzeitverhaltens katalytischer Packungen. Proceedings of GVC—Fachausschuss “Thermische Zerlegung,” Bingen, Germany, 2002. Rod V. Calculating mass transfer with longitudinal mixing. Br Chem Eng 1966; 6:483–488. Casamatta G. Compartment de la population des gouttes dans une colonne d’extraction: transport, rupture, coalescence, transfer de matiere. Ph.D. dissertation, Institute National Polytechnique de Toulouse, Toulouse, France, 1981. Kronberger T, Ortner A, Zulehner W, Bart HJ. Numerical simulation of extraction columns using a drop population model. Computers Chem Eng 1995; 19:639–644. Ortner A, Kronberger T, Zulehner W, Bart HJ. Tropfen-Populanz-Bilanzmodell am Beispiel einer gerührten Extraktionskolonne. Chem Ing Tech 1995; 67:984–988. Gerstlauer A. Herleitung und Reduktion populationsdynamischer Modelle am Beispiel der Flüssig–Flüssig Extraktion. Düsseldorf: VDI, 1999. Modes G. Grundsätzliche Studie zur Populationsdynamik einer Extraktionskolonne auf Basis von Einzeltropfenuntersuchungen. Aachen, Germany: Shaker, 2000. Attarakih M, Bart HJ, Faqir NM. An approximate optimal moving-grid technique for the solution of discretized population balances in batch systems. Proceedings of ESCAPE 12 European Symposium on Computer-Aided Process Engineering, The Hague, 2002. Simon M, Bart HJ. Experimental studies of coalescence in liquid/liquid systems. Chem Eng Technol 2002; 25:481–484. Wei J, Prater CD. The structure and analysis of complex reaction systems. Adv Catal 1962; 13:203–392. Hikita H, Asai S. Gas absorption with (m, n)-th order irreversible chemical reactions. Int Chem Eng 1964; 4:332–340. Toor HL. Dual diffusion-reaction coupling in first order multicomponent systems. Chem Eng Sci 1965; 20:941–951. DeLancey GB. Multicomponent film-penetration theory with linearized kinetics— I. Linearization theory and flux expression. Chem Eng Sci 1974; 29:2315–2323. Astarita G, Sandler SI. Kinetic and Thermodynamic Lumping of Multicomponent Mixtures. Amsterdam: Elsevier, 1991. Kenig EY, Kholpanov LP. Analysis of formulation and solution of multicomponent reaction-diffusion problems. Theor Found Chem Eng 1992; 26:510–521.


187. 188. 189. 190. 191.

192. 193. 194. 195.

196. 197. 198.

199.

200. 201. 202. 203. 204.

205. 206.

Onda K, Takeuchi H, Okumoto Y. Mass transfer coefficients between gas and liquid phases in packed columns. J Chem Eng Japan 1968; 1:56–62. Fuller EN, Schettler PD, Giddings JC. A new method for prediction of binary gasphase diffusion coefficients. Ind Eng Chem 1966; 58:19–27. Siddiqi MA, Lucas K. Correlations for prediction of diffusion in liquids. Can J Chem Eng 1986; 64:839–843. Mika V. Model of packed absorption column. I. Physical absorption. Coll Czech Chem Commun 1967; 32:2933–2943. Schwartz SE, White WH. Solubility equilibria of the nitrogen oxides and oxyacids in dilute aqueous solutions. In: Pfafflin JR, Zeigler EN, eds. Advances in Environmental Science and Engineering. Vol. 4. New York: Gordon and Breach, 1981:1–45. Engels H. Anwendung des Modells der lokalen Zusammensetzung auf Elektrolytlösungen. Ph.D. dissertation, RWTH Aachen, Aachen, Germany, 1985. Horvath AL. Handbook of Aqueous Electrolyte Solutions. Chichester, UK: Ellis Horwood, 1985. Rocha JA, Bravo JL, Fair JR. Distillation columns containing structured packings— 2. Mass transfer model. Ind Eng Chem Res 1996; 35:1660–1667. Ranzi E, Rovaglio M, Faravelli T, Biardi G. Role of energy balances in dynamic simulation of multicomponent distillation columns. Computers Chem Eng 1988; 12:783–786. Ronge G. Überprüfung unterschiedlicher Modelle für den Stoffaustausch bei der Rektifikation in Packungskolonnen. Düsseldorf: VDI, 1995. Xu ZP, Chuang KT. Kinetics of acetic acid esterification over ion exchange catalysts. Can J Chem Eng 1996; 74:493–500. Powers MF, Vickery DJ, Arehole A, Taylor R. A nonequilibrium-stage model of multicomponent separation processes—V. Computational methods for solving the model equations. Computers Chem Eng 1988; 12:1229–1241. Ellenberger J, Krishna R. Countercurrent operation of structured catalytically packed distillation columns: pressure drop, holdup, and mixing. Chem Eng Sci 1999; 54:1339–1345. Moritz P, Hasse H. Fluid dynamics in reactive distillation packing Katapak-S. Chem Eng Sci 1999; 54:1367–1374. Stichlmair JG, Fair JR. Distillation: Principles and Practice. New York: WileyVCH, 1998. Sundmacher K, Hoffmann U. Macrokinetic analysis of MTBE synthesis in chemical potentials. Chem Eng Sci 1994; 49:3077–3089. Mörters M, Bart HJ. Fluorescence-indicated mass transfer in reactive extraction. Chem Eng Technol 2000; 23:1–7. Mack C. Untersuchungen zum Stofftransport und Stoffaustauschverhalten bei der Extraktion von Chrom (III) und Zink mittels D2EHPA. Ph.D. dissertation, Technical University of Darmstadt, Darmstadt, Germany, 2000. Marangoni CGM. Über die Ausbreitung der Tropfen einer Flüssigkeit auf der Oberfläche einer anderen. Ann Phys Chem (Poggendorf) 1871; 143:337–354. Thomson W (Lord Kelvin). The influence of wind on waves in water supposed frictionless. Phil Mag 1871; 10:330–333.


207.

208. 209. 210.

Pitzer KS. Theory: ion interaction approach. In: Pytkowicz RM, ed. Activity Coefficients in Electrolyte Solutions. Vol. 1. Boca Raton, FL: CRC Press, 1979: 157–208. Hildebrand JM, Scott RL. The Solubility of Nonelectrolytes. New York: Reinhold, 1950. Henschke M, Pfennig A. Mass transfer enhancement in single-drop extraction experiments. AIChE J 1999; 45:2079–2086. Handlos AE, Baron T. Mass and heat transfer from drops in liquid–liquid extraction. AIChE J 1957; 3:127–136.

APPENDIX A. A DETAILED DESCRIPTION OF RATE-BASED MODELING A.1.

Balance Equations

The mass balance equations of the traditional multicomponent rate-based model (see, e.g., Refs. 57 and 58) are written separately for each phase. In order to give a common description to all three considered RSPs (where it is possible, of course) we will use the notion of two contacting fluid phases. The first one is always the liquid phase, whereas the second fluid phase represents the gas phase for RA, the vapor phase for RD and the liquid phase for RE. Considering homogeneous chemical reactions taking place in the fluid phases, the steady-state balance equations should include the reaction source terms: 0 0

d ( LxiB ) ( N LiB a I RLiB L ) As dl

d (GyiB ) ( NGiB a I RGiB G ) As dl

i 1, . . . , n i 1, . . . , n

(A1)

(A2)

If chemical reactions take place in the (first) liquid phase only (this is valid for most of RD processes), the phase balances for the second fluid phase simplify to 0

d (GyiB ) NGiB a I As dl

i 1, . . . , n

(A3)

The bulk-phase balances are completed by the summation equation for the liquid and second fluid bulk mole fractions: n

∑x

B i

1

i1


(A4)

n

∑ y 1 B i

(A5)

i1

The volumetric liquid holdup, L, depends on the gas/vapor and liquid flows and is calculated via empirical correlations (e.g., Ref. 65). For the determination of axial temperature profiles, differential energy balances are formulated, including the product of the liquid molar holdup and the specific enthalpy as energy capacity. The energy balances written for continuous systems are as follows: 0 0

d 0 ( LhLB ) (QLB a I RLB L H RL ) As dl

(A6)

d 0 (GhGB ) (QGB a I RGBG H RG ) As dl

(A7)

In the dynamic rate-based stage model, molar holdup terms have to be considered in the mass balance equations, whereas the changes in both the specific molar component holdup and the total molar holdup are taken into account. For the liquid phase, these equations are as follows: ∂ ∂ U Li ( LxiB ) ( N LiB a I RLiB L ) As ∂t ∂l U Li xiBU Lt xiB ( L c Lt As )

i 1, . . . , n

i 1, . . . , n

(A8)

(A9)

The gas/vapor holdup can often be neglected due to the low gas-phase density, and the component balance equation reduces to Eq. (A2) (see also Ref. 139). A.2.

Mass Transfer and Reaction Coupling in Fluid Films

The component fluxes NiB entering into Eqs. (A1)–(A3) are determined based on the mass transport in the film region. Because the key assumptions of the film model result in the one-dimensional mass transport normal to the interface, the differential component balance equations including simultaneous mass transfer and reaction in the film are as follows: dNLi RLi 0 dz

i 1, . . . , n

(A10)

Equations (A10), which are generally valid for both liquid and second fluid phases, represent nothing but differential mass balances for the film region, with the


account of the source term due to the reaction. To link these balances to process variables like component concentrations, some additional relationships, often called constitutive relations (see Ref. 57), are necessary. For the component fluxes Ni, these constitutive relations result from the multicomponent diffusion description [Eqs. (1) and (2)] and, for the source terms, from the reaction kinetics description. The latter strongly depends on the specific reaction mechanism, the stoichiometry, and the presence or absence of parallel reaction schemes (69). The rate expressions for Ri usually represent nonlinear dependences on the mixture composition and temperature. Specifically for the coupled reaction–mass transfer problems, such as Eqs. (A10), it is always essential as to whether or not the reaction rate is comparable to that of diffusion (68,77). Equations (A10) should be completed by the boundary conditions relevant to the film model. These conditions specify the values of the mixture composition at both film boundaries. For example, for the liquid phase: xi ( z 0) xiI ,

xi ( z L ) xiB

i 1, . . . , n

(A11)

Combining Eqs. (A10) with the boundary conditions (A11) written in vector form and using constitutive relations such as Eqs. (1) and (2), we obtain a vector-type boundary-value problem, which permits the component concentration profiles to be obtained as functions of the film coordinate. These concentration profiles, in turn, allow one to determine the component fluxes. Thus the boundaryvalue problem describing the film phenomena has to be solved in conjunction with all other model equations. The composition boundary values entering into Eqs. (A11) represent external values for Eqs. (A10). With some further assumptions concerning the diffusion and reaction terms, this allows an analytical solution of the boundary-value problem [Eqs. (A10) and (A11)] in a closed matrix form (see Refs. 58 and 135). On the other hand, the boundary values need to be determined from the total system of equations describing the process. The bulk values in both phases are found from the balance relations, Eqs. (A1) and (A2). The interfacial liquid-phase concentrations xiI are related to the relevant concentrations of the second fluid phase, yiI , by the thermodynamic equilibrium relationships and by the continuity condition for the molar fluxes at the interface (57,135). Due to the chemical conversion in the liquid film, the molar fluxes at the interface and at the boundary between the film and the bulk of the phase differ. The system of equations is completed by the conservation equations for the mass and energy fluxes at the phase interface and the necessary linking conditions between the bulk and film phases (see Refs. 57, 59, and 84). Generally all these considerations are also valid for the second fluid film phase, provided that reactions occur there (135). Both analytical and numerical solutions of the coupled diffusion-reaction film problem are analyzed at full length in Ref. 167; their particular applications are considered in Section 3.


A.3.

Nonideal Flow Behavior in Catalytic Column Internals

The mass balances [Eqs. (A1) and (A2)] assume plug-flow behavior for both the gas/vapor and liquid phases. However, real flow behavior is much more complex and constitutes a fundamental issue in multiphase reactor design. It has a strong influence on the reactor performance, for example, due to back-mixing of both phases, which is responsible for significant effects on the reaction rates and product selectivity. Possible development of stagnant zones results in secondary undesired reactions. To ensure an optimum model development for CD processes, experimental studies on the nonideal flow behavior in the catalytic packing MULTIPAK® are performed (168). The experimental results confirm that the fluid flow in MULTIPAK® deviates from plug-flow behavior (Figure 31). Calculated axial dispersion coefficients are about 104–102 m2/s, which are several orders of magnitude larger than that for molecular diffusion (Figure 32). Therefore, in the investigated operating range, nonideal mixing effects are caused by hydrodynamic rather than molecular diffusion effects. Calculated Bodenstein numbers are one order of magnitude

FIGURE 31 Comparison between the experimental RTD curve for the catalytic packing MULTIPAK® (dC 0.1 m), the ADM model, and the PDE model.


FIGURE 32 Axial dispersion coefficients for the catalytic packing MULTIPAK® (dC 0.1 m ) , calculated based on the ADM model.

smaller than those for fixed-bed reactors, which may be caused by two effects: the occurrence of stagnant zones in the catalyst layer, and liquid bypassing due to the hybrid structure of the catalytic packing (168). The rate-based models suggested up to now do not take liquid back-mixing into consideration. The only exception is the nonequilibrium-cell model for multicomponent reactive distillation in tray columns presented in Ref. 169. In this work a single distillation tray is treated by a series of cells along the vapor and liquid flow paths, whereas each cell is described by the two-film model (see Section 2.3). Using different numbers of cells in both flow paths allows one to describe various flow patterns. However, a consistent experimental determination of necessary model parameters (e.g., cell film thickness) appears difficult, whereas the complex iterative character of the calculation procedure in the dynamic case limits the applicability of the nonequilibrium cell model. A far more promising approach is represented by the so-called differential models, such as the axial dispersion model (ADM) (170) as well as the pistonflow model with axial dispersion and mass exchange (PDE) (171). Experimental studies (168) show that the ADM gives an appropriate description of the nonideal flow behavior of the liquid phase in catalytic packings (see Figure 31). Considering


the nonideal flow behavior via the ADM, the dynamic mass balances [cf. Eqs. (A8)] take the following form: D ∂2 ∂ ∂ ULi ax 2 ( LxiB ) ( LxiB ) ( N LiB a I RLiB L ) As u L ∂l ∂t ∂l

i 1, . . . , n (A12)

A thorough investigation of the influence of the flow nonideality in catalytic packings on the dynamic process behavior of specific CD processes is an objective of some current studies (172). Equation (A12) is widely used in RE, but it does not account for the specific interactions of the dispersed phase. In this respect current research is focused on drop population balance models, which account for the different rising velocities of the different-size droplets and their interactions, such as droplet breakup and coalescence (173–180). APPENDIX B. MODELING PECULIARITIES AND MODEL PARAMETERS FOR THE CASE STUDIES B.1.

Absorption of NOx

In terms of the concentration vector, Eq. (A10) is a nonlinear differential equation of the second order. The boundary-value problem [Eqs. (A10) and (A11)] is usually solved numerically. However, it is also possible to linearize the reaction term using the method suggested in Ref. 181: R ≅ [ K ] x

(B1)

Equation (B1) provides a satisfactory representation for many processes over the entire reaction range and is a good linear approximation for most systems in a sufficiently small range (see, e.g., Refs. 68 and 182–184). Equation (B1) has gained widespread acceptance in various chemical and reactor engineering areas (185) and is recommended for use in the modeling of reactive separation operations (59,184). The approximation of Eq. (B1) allows one to reduce Eqs. (A10) and (A11) to a linearized boundary-value problem (183,184,186). The latter can then be solved analytically and yields a compact matrix-form solution for the concentration profiles in the film region [58]. Such a solution gives simple analytical expressions for the component fluxes with regard to the homogeneous reaction in the fluid films (see Ref. 135), which can be of particular value when large industrial reactive separation units are considered and designed. The methods of determination of the reaction matrix [K] are considered in Refs. 167, 181, 183, 184 and 186. Another important matrix parameter entering into the linearized film mass transport equation is the multicomponent diffusion matrix [D]. The latter results from the transformation of the Maxwell–Stefan Eqs. (1) to the form of the generalized Fick’s law (83). Matrix [D] is generally a function of


the mixture composition and it is assumed constant along the diffusion path (83). The direct expressions for the elements of the diffusion matrix [D] can be found, for example, in Ref. 57. The linearization of the initial film mass transport equation and its analytical solution were applied to simulate the industrial NOx absorption process considered. In order to calculate the multicomponent diffusion matrices [D], the binary diffusivities in both phases should be known. The film thickness representing an important model parameter is estimated via the mass transfer coefficients (57,83). The binary diffusivities and mass transfer coefficients were calculated from the correlations summarized in Table 3. The correlations of Billet (66) and Onda et al. (187) are valid for various mixtures and packings and cover both absorption and distillation processes. The correlation of Kolev (133) is obtained for RA and certain random packings. In general, the mass transfer coefficient correlations need to be compared to one another and validated using experimental data. This shows, in particular, the way the mass transfer correlations influence the concentration profiles of the components and other relevant process characteristics. Nitric acid is a strong electrolyte. Therefore, the solubilities of nitrogen oxides in water given in Ref. 191 and based on Henry’s law are utilized and further corrected by using the method of van Krevelen and Hoftijzer (77) for electrolyte solutions. The chemical equilibrium is calculated in terms of liquid-phase activities. The local composition model of Engels (192), based on the UNIQUAC model, is used for the calculation of vapor pressures and activity coefficients of water and nitric acid. Multicomponent diffusion coefficients in the liquid phase are corrected for the nonideality, as suggested in Ref. 57.

TABLE 3 Binary Diffusion Coefficients and Mass Transfer Coefficients

Phase

Binary diffusion coefficient

Gas

Ref. 188

Liquid

Ref. 189


Mass transfer coefficient correlation Ref. 187 Ref. 66 Wehmeier (see Ref. 134) Ref. 187 Ref. 66 Ref. 133 Ref. 190

B.2.

Coke Gas Purification

In Ref. 139 a purely numerical approach to the solution of the considered complex RA problem was suggested. The liquid film is treated as an additional balance region, in which reaction and mass transfer occur simultaneously. Therefore, the reactions are considered both in the liquid-bulk-phase mass balances, Eq. (A1), and in the differential balances for the liquid film, Eq. (A10). To be able to describe the presence of electrolytes in the system, the electrical driving force also needs to be taken into account (57). Therefore, the gradient of the electrical potential is introduced into the generalized driving force di [cf. Eq. (2)]: di

xi 1 ∂i F 1 d x i zi ℜT L ∂ ℜT L d

i 1, . . . , n

(B2)

In dilute electrolyte systems, the diffusional interactions can usually be neglected, and the generalized Maxwell–Stefan equations are reduced to the Nernst–Planck equations (B3): N Li

c Lt DLi, eff  dxi F d  x i zi   xi N Ln L ℜT d   d

i 1, . . . , n 1

(B3)

where n is the solvent index. The consideration of the electrical potential requires an additional condition, the electroneutrality, which has to be met in each point of the liquid phase: n

∑x z 0 i i

(B4)

i1

Thermodynamic nonidealities are considered both in the transport equations (A10) and in the equilibrium relationships at the phase interface. Because electrolytes are present in the system, the liquid-phase diffusion coefficients should be corrected to account for the specific transport properties of electrolyte solutions. The thermodynamic equilibrium at the gas–liquid interface is described as follows: yiI Ki xiI

i 1, . . . , n

(B5)

where the distribution coefficient Ki comprises fugacities in both phases and activity coefficients in the liquid phase. For the system considered, the values of Ki, Eq. (B5), are determined from the electrolyte NRTL method (70,71). The liquid-phase diffusion coefficients are found with the Nernst–Hartley equation (193), which describes the transport properties in weak electrolyte systems. The gas-phase diffusion coefficients are estimated according to the


Chapman–Enskog–Wilke–Lee model (72). The correlations for the mass transfer coefficients are taken from Ref. 194. B.3.

Methyl Acetate System, Batch Distillation

The rate-based models usually use the two-film theory and comprise the material and energy balances of a differential element of the two-phase volume in the packing (148). The classical two-film model shown in Figure 13 is extended here to consider the catalyst phase (Figure 33). A pseudo-homogeneous approach is chosen for the catalyzed reaction (see also Section 2.1), and the corresponding overall reaction kinetics is determined by fixed-bed experiments (34). This macroscopic kinetics includes the influence of the liquid distribution and mass transfer resistances at the liquid–solid interface as well as diffusional transport phenomena inside the porous catalyst. For the determination of conversion corresponding to the average residence time, the reaction kinetics is integrated into the mass balances, and the liquid

FIGURE 33 Film model for a differential packing segment with heterogeneous catalyst.


holdup, as the accumulation term, is accounted for simultaneously, as in Eqs. (A8) and (A9). Because of low vapor-phase density, the vapor holdup is neglected, and the vapor-phase-component balance equation reduces to Eq. (A2). Ranzi et al. [195] found that the full energy balances, including the accumulation term, have to be considered in order to predict correct dynamic process behavior. Therefore, the differential dynamic energy balance for the liquid phase is applied as follows: ∂ ∂ 0 ) As EL ( LhLB ) (QLB a I RLB L H RL ∂t ∂l Q L (TLB T I ) L B L

2

∑N h

Li Li

(B6)

i 1, . . . , n

i1

where EL hLB ( L c Lt As )

(B7)

Similar to the mass balance equation, the vapor-phase energy balance simplifies to Eq. (A7). Experimental studies were carried out to derive correlations for mass transfer coefficients, reaction kinetics, liquid holdup, and pressure drop for the packing MULTIPAK® (35). Suitable correlations for ROMBOPAK 6M® are taken from Refs. 90 and 196. The nonideal thermodynamic behavior of the investigated multicomponent system was described by the NRTL model for activity coefficients concerning nonidealities caused by the dimerisation (see Ref. 72). Binary diffusion coefficients for the vapor phase and for the liquid phase were estimated via the method proposed by Fuller et al. and Tyn and Calus, respectively (see Ref. 72). Physical properties such as densities, viscosities, and thermal conductivities were calculated from the methods given in Ref. 72. Heat losses through the column wall were measured at pilot scale. B.4.

Methyl Acetate System, Steady-State Distillation

The model is based on the film theory and comprises the material and energy balances of a differential element of the two-phase volume in the packing. Each element consists of an ideally mixed vapor and liquid bulk phase and a vapor film region adjacent to the interface, as shown in Figure 33. A first guess of the bulk phase compositions and temperatures was provided by the solution of an equilibriumstage model without reactions, as suggested in Ref. 198. The catalyzed reaction is described by the quasi-homogeneous approach of Ref. 197, since the concentration of acid sites has been determined as aCat 4.7 molH/gCat for dry Lewatit K2621, which is close to the data of Ref. 197 given for Amberlyst 15.


In order to determine the model parameters, several experiments were performed at laboratory scale. Pressure drop experiments were carried out in glass columns, with a total packing height of 1 m at ambient pressure. Air/water was used as a test system, with a circulating liquid phase set at a constant temperature of 20C. The experimental data cover a wide range of possible column loads. The gas load for the column with 100-mm diameter was restricted to 1.7 Pa0.5. Therefore, the liquid load was increased to higher values to reach the flooding region of the catalytic packing. Two different flow regimes similar to those of conventional structured packings can be observed. Flooding of the packing can be observed at a pressure drop above 103 Pa/m. The possible column loads for MULTIPAK® are very similar to those reported in Refs. 199 and 200 for KATAPAK-S. The number of theoretical stages per meter of the catalytic packing was determined as a function of the gas capacity factor. For the whole range of column loads, the separation efficiency is at least four theoretical stages per meter. Moritz and Hasse (200) determined an NTSM value of 3 for the laboratory-scale KATAPAK-S. The separation efficiency remains constant for a wide loading range of the packing. For lower column loads, the NTSM value increases to 6, a phenomenon already reported in Ref. 90 for the conventional structured packing Montzpak A3-500. A simple transfer-unit concept assuming all mass transfer resistance in the vapor phase was used to determine the vapor-side mass transfer coefficients (201). The mass transfer correlation Sh G 0.009 Re G0.92 Sc1/3

(B8)

represents all experimental data with an accuracy of 13%. A comparison with experimental data is shown in Figure 34. B.5.

Synthesis of Methyl Tertiary Butyl Ether

The mathematical description considered in Section 2.3 and Appendix A was used as a modeling basis for the specially developed completely rate-based simulator DESIGNER (155). This tool consists of several blocks, including model libraries for physical properties, mass and heat transfer, reaction kinetics, and equilibrium, as well as a specific hybrid solver and thermodynamic package. DESIGNER also contains different hydrodynamic models (e.g., completely mixed liquid–completely mixed vapor, completely mixed liquid–vapor plug flow, mixed pool model, eddy diffusion model) and a model library of hydrodynamic correlations for the mass transfer coefficients, interfacial area, pressure drop, holdup, weeping, and entrainment that cover a number of different column internals and flow conditions. In DESIGNER, different ways of taking account of heterogeneous reaction kinetics are available, depending on the reaction rate and character. One further possibility is to use a detailed model for the heterogeneous catalyst mass transfer


efficiency based on the approach of Ref. 91. When applying this type of kinetic model, the intrinsic kinetics data are needed. Another way is the pseudo-homogeneous approach, with effective kinetic expressions, by which the kinetics description is introduced as source terms into the balance equations [cf. Eqs. (A1) and (A2)]. For the system considered here, the reaction is slow as compared to the mass transfer rate. For this reason the pseudo-homogeneous approach is used, the reaction being accounted for in the liquid bulk only. Basically, DESIGNER can use different physical property packages that are easy to interchange with commercial flowsheet simulators. For the case considered, the vapor–liquid equilibrium description is based on the UNIQUAC model. The liquid-phase binary diffusivities are determined using the method of Tyn and Calus (see Ref. 72) for the diluted mixtures, corrected by the Vignes equation (57), to account for finite concentrations. The vapor-phase diffusion coefficients are assumed constant. The reaction kinetics parameters taken from Ref. 202 are implemented directly in the DESIGNER code. B.6.

Reactive Extraction of Zinc

In conventional RE processes, the diffusive resistance is concentrated mainly inside the droplet, whereas the aqueous-side resistance can be neglected. This has been proven in Ref. 203 using the laser-induced-fluorescence (LIF) technique. Usually the organic phase is more viscous and the diffusion coefficients of the organic complexes are larger than those at the aqueous side, which supports this finding. The mass transfer within a rigid droplet is determined by the Maxwell– Stefan diffusion. The appropriate diffusion coefficients experimentally determined

FIGURE 34 Sherwood number correlation for MULTIPAK®.


TABLE 4 Ternary Fick Diffusion Coefficients for the System ZnR2 ( RH ) (1), RH (2), and Diluent (3) at 298.15 K . Isododecane (3) D11 D12 D21 D22

1.01 0.11 4.11 2.79

0.30 1010 0.10 1010 0.80 1010 0.30 1010

Toluene (3) m2/s m2/s m2/s m2/s

3.60 1.10 1010 m2/s 0.35 0.42 1010 m2/s 0.16 3.50 1010 m2/s 7.68 0.21 1010 m2/s

for this zinc extraction system in Ref. 204 are presented in Table 4. With nonrigid droplets, a mass transfer enhancement by internal convection has to be considered. However, with industrial feed solutions there are always impurities present that may dampen the mass transfer (8). In contrast, there also might be a mass transfer increase due to Marangoni effects (205,206). Therefore, for a final design of a column, mass transfer measurements are recommended. The macrokinetics of zinc extraction is discussed in detail in Ref. 8. It is a combination of a reaction kinetics term (55) with the Maxwell–Stefan (54) or eddy diffusion (56). The rate law is as follows:

d[ Zn 2 ] dt

2

[ R2 H2 ]  (B9) [ R2 H2 ]1.5 [ Zn 2 ] r [ H ]2 [ ZnR2 ( RH )]  v   C [ R H ]  [ R2 H2 ]1.5 C1 [ H ]2  2 2 2 

where C1, C2, , and r, are the estimated kinetics parameter (see EFCE test systems discussed earlier). The rate constant for the backward reaction, r, can be replaced by the thermodynamic equilibrium constant: K eq

v r

(B10)

The species concentrations are formulated in activities using the Pitzer model (207) for the aqueous phase and the Hildebrand–Scott solubility parameter (208) for the organic phase. The effective diffusion coefficient is calculated according to the model of Ref. 209, which accounts for interfacial instabilities. This model includes a Handlos–Baron-like correlation (210) and one adjustable parameter, CIP : Deff

w dp   2048 CIP 1 d  c  


(B11)

10 Multifunctional Reactors: Integration of Reaction and Heat Transfer David W. Agar University of Dortmund, Dortmund, Germany

1.

INTRODUCTION

Reaction engineers devote a lot of time and ingenuity to enhancing reactor performance by attempting to follow an optimal trajectory for the reaction system (1) and by overcoming the limitations imposed by the accompanying heat and mass transfer processes. These objectives are often interrelated: Achieving the concentration and temperature profiles required to maximize conversion rates and minimize by-product formation, for example, dictates the absence of gradients that might lead to local deviations from these values. Process intensification can be considered to be the use of measures to increase the volume-specific rates of reaction, heat transfer, and mass transfer and thus to enable the chemical system or catalyst to realize its full potential (2). Catalysis itself is an example of process intensification in its broadest sense. The use of special reaction media, such as ionic liquids or supercritical fluids, high-density energy sources, such as microwaves or ultrasonics, the exploitation of centrifugal fields, the use of microstructured reactors with very high specific surface areas, and the periodic reactor operation all fall under this definition of process intensification, and the list given is by no means exhaustive. Reactor performance is dictated by the inputs, by the contacting pattern, i.e., how and when individual elements pass through the reactor and contact one


FIGURE 1 Basic strategies for manipulating temperature and concentration profiles in chemical reactors.

another and how long they retain their identity, and by the cumulative kinetics and thermodynamics to which elements are exposed along their reaction trajectory (3). Identifying the most suitable reactor configuration, e.g., an ideal plug flow, is a well-established procedure in chemical reaction engineering. Less appreciated are the full range of possibilities available in manipulating the local rates of reaction by imposing favorable temperature, concentration, and activity profiles along a catalytic reactor. While the first option has received extensive attention, the last two have been somewhat neglected, although they often offer a more selective intervention in the progress of the reaction and complement the more common tailoring of the temperature profile. In general, temperature and concentration profiles may be externally influenced by convective or recuperative and, less commonly, regenerative or reactive strategies (Figure 1). The “convective” addition or withdrawal of side streams along the reactor represents a simple technique for temperature control or for improving selectivity by restricting availability of one reactant. In a recuperative process, examples of which are provided by the cooled tubular reactor and the membrane reactor, heat or material is exchanged, in the latter case usually in a


selective manner, via diffusive transport processes, such as heat conduction, with spatially distinct external sources and sinks. Regeneration exploits the storage of heat and mass, by adsorption, for example, on reactor internals—usually a fixed bed—to yield beneficial temperature and concentration profiles that could not arise under steady-state operation. Regenerative processes are inherently unsteady state in nature and entail a chronologically separate recharging of the storage capacity drawn down during the reaction phase. The use of a supplementary reaction to supply or consume heat and/or reactants or products, as encountered in oxydehydrogenations, is a method requiring considerable finesse and almost perfect compatibility between the individual reactions. 2.

CONVECTIVE HEAT TRANSFER

Each of these approaches has its pros and cons, as can be illustrated for convective cooling in a cold-shot reactor, employed in ammonia synthesis, for example. Plotting conversion against temperature for a reversible exothermic reaction (Figure 2a) shows that intermittent cooling by the discrete introduction of cold feed along the reactor enables one to circumvent the equilibrium limitation imposed on adiabatic operation but also that the cooling effect desired is accompanied by a less welcome loss in conversion. Furthermore, as the slope of the cooling line approaches that of the adiabatic reaction path, the efficacy of cooling and its benefits diminish. A possible solution to this problem of limited cooling capacity is to employ an inert side stream in place of feedstock (Figure 2b) as coolant (3). However, this would result in dilution of the reactor product stream and complicate the downstream processing steps. A process intensification technique to overcome the difficulties indicated is, for gaseous systems, to inject inert liquid between adiabatic reactor stages. Exploiting the heat of evaporation means that much lower quantities of inert are needed than with a gaseous coolant. 3.

RECUPERATIVE HEAT TRANSFER

The best-known recuperative reactor—the multitubular reactor used in the partial oxidation of hydrocarbons, for instance—is a ubiquitous piece of equipment in the chemical processing industry (4). This should not blind one to the fact that it exhibits several serious shortcomings: The reactors are costly, and, despite the use of up to 25,000 tubes in a single reactor to provide a suitably extensive area for heat transfer (⬃100 m2/m3), one often observes large temperature excursions from the desired temperature level—so-called hot spots—in both the axial and radial directions (Figure 3). These hot spots arise due to a bottleneck in the heat removal process, arising from a combination of the locally accelerated reaction rates and poor heat transport through the catalyst bed (⬃100 W/m2K)—an order of magnitude smaller than what the coolant side is capable of providing.


FIGURE 2 Interstage “convective” cooling of an exothermic equilibrium reaction through introduction of (a) cold-shot feed by-pass and (b) cold-shot inert side-stream between adiabatic stages.


FIGURE 3 Schematic of temperature profiles and hot spot formation in a multitubular reactor.

Hot spots usually dictate the attainable reactor performance in terms of conversion, selectivity, safety issues, operating lifetimes, and materials of construction, since most of the critical processes are confined to the immediate vicinity of the temperature maximum. 3.1.

Catalyst Dilution

The elimination of reactor hot spots has also attracted considerable interest over the years. Because a panacea remains elusive, a variety of countermeasures have been adopted reflecting different compromises between the demands of the reaction, heat removal, and pressure drop (Figure 4). Perhaps the simplest procedure for avoiding the formation of pronounced hot spots is to dilute the catalyst at the endangered locations (5). In this manner, the reaction is spread more evenly over the length of the reactor, and a better harmonization between the heat production by the reaction and the heat removal via the reactor wall is realized (Figure 5). The resultant reactors are, of course, larger, and the inclusion of inert packing leads to increased pressure drops. But the technique is reliable and involves no additional developments. Interestingly, the precise activity profile employed is seldom decisive for the improvement in performance; i.e., simple arrangements suffice.


FIGURE 4 Various measures for the harmonization of reaction, heat transport, and pressure drop in chemical reactors benchmarked against a multitubular reactor.

3.2.

Linde™ Isothermal Reactor

An alternative solution is to pack still more heat exchange surface into the catalyst bed and to try to augment the heat transfer coefficients modestly by inducing a greater degree of turbulence for the gas flow over the cooling surface. The so-called Linde™ isothermal reactor (6) inverts the situation in the conventional multitubular reactor by operating with the coolant—normally pressurized boiling water—within the tubes and the catalyst outside (Figure 6). The tubes are also no longer parallel but assume a convoluted spiral geometry that raises the specific surface area and increases heat transfer coefficients by an order of 50%. Use of the reactor is confined to temperatures below 550 K, since the pressure required for evaporative cooling above this value becomes exorbitant, and molten salt cooling is not an option due to the high pressure drop in the cooling circuit. In addition, the removal of the catalyst from such reactors, e.g., for regeneration or replacement, may present problems owing to “arch” formation between cooling tubes. In economic terms, the greater compactness of the Linde™ isothermal reactor must be set against its increased construction complexity. 3.3.

Fluidized Beds

An extremely effective means of enhancing heat removal from a reactor is to make use of fluidized-bed technology (3). Heat transfer coefficients for gaseous systems are increased to values of around 600 W/m2K or more by virtue of the very efficient convective-regenerative particle transport mechanism of heat transfer. Further


advantages include lower resistances to mass transfer (particle size ⬃ 100 m) and facile catalyst exchange for regeneration purposes if necessary. Less favorable aspects of fluidized beds include the high degree of back-mixing that occurs, the limited range of hydrodynamic loading, and the uncertainty involved in scale-up. Moreover, the catalyst must be very mechanically resilient, and adapting catalysts for fluidized-bed operation can be a time-consuming and frustrating exercise. 3.4.

Catalytic Microreactors

A measure that has been the subject of extensive publication is that of microreactors with catalytically coated walls (7,8). A microreactor has been defined as: “a miniaturized reaction vessel with characteristic dimensions in the range 10–300 m which has been fabricated using state-of-the-art high-precision engineering” (7). Such reactors exhibit well-defined laminar-flow patterns and permit facile scale-up by simple “numbering up” of the number of channels and flexible

FIGURE 5 Hot spot reduction using spatially structured catalyst dilution. Selectivity profiles for base case with constant coolant temperature, cocurrent coolant strategy and axially profiled catalyst activity strategy. The base case chosen in the calculations is the one in which the coolant temperature is constant and the activity profile along the length of the reactor is at the level unity. (From Ref. 5.)


FIGURE 6 Linde™ isothermal reactor for intensified cooling of strongly exothermic reactions. (From Ref. 6.)

arrangements of individual modules. Furthermore, they have extremely low holdups and, as a consequence, comparatively short dynamic response times. Most importantly, they enable one to operate under isothermal conditions with even the most exothermic reactions. The list of applications for which microreactors are suitable includes the reaction engineering for fuel cell hydrogen production, the synthesis of hazardous chemicals, high-throughput screening for chemicals and catalysts, as instruments for obtaining insights into chemical reaction mechanisms, and as components of “intelligent” chemical sensors. The excellent heat transfer characteristics of microreactors result not only from their considerably enhanced specific heat exchange surface areas of 30,000 m2/m3—a value roughly 300 times higher than that in a conventional multitubular reactor—but also from the rapid lateral heat transfer across the channels.


For the small microchannel dimensions involved (Figure 7), conductive heat transfer plays an important role, augmenting the heat transfer coefficients by a factor of 5 or more to values in the vicinity of 700 W/m2K. This phenomenon is also responsible for good transverse mixing, counteracting the negative effects of laminar-flow profiles. Less advantageous are the high unit production costs for microreactors and the complexities involved in their manufacture. The basic material of construction is usually metal or plastic, which has restricted heterogeneous catalytic applications to metallic catalysts or often unsatisfactory metallic-catalytic composites.

FIGURE 7 Structure and typical dimensions of a microreactor. (From Ref. 7.)


Despite the high specific surface areas, the amount of accessible catalyst remains low due to the limited thickness of the porous catalytic layer dictated by considerations such as the adhesion to the substrate. The susceptibility of the fine channels to blockage with solid impurities or deposits formed in the reaction, together with the problems of integrating connections with the external macroenvironments and ensuring uniform gas distribution between the individual channels, a prerequisite for numbering up, represent further questions that have to be resolved for the industrial application of microreactors to become practicable. 3.4.1.

Catalytic Millireactors

The drawbacks associated with this particular technique of process intensification may be ameliorated by analyzing the characteristic length scales for the component processes. The chemical activity of a heterogeneous catalyst, for example, is expressed in its nanostructure, i.e., the pore diameter of 10–100 nm reflected in the typical specific surface area of a catalyst, say, 100 m2/g. This scale provides an adequate number of active sites for the reaction to proceed without imposing unnecessary limitations on the underlying chemical kinetics. Catalyst systems lacking this nanostructure will seldom provide sufficient activity and do not permit the catalyst chemistry to realize its full potential. Even in a microreactor, geometric surface areas of catalytic layers fall short of the internal specific pore surface area by a factor of almost 10,000. The diffusive transport through the pore system to and from the active sites dictates the maximal thickness of porous catalytic layers. For the diffusion coefficients in gas-phase systems, the conventional analysis based on the Thiele or Weisz modulus suggests that diffusive mass transfer is sufficient to maintain the reactant flow to the active catalytic sites over distances of about 1 mm. The corresponding value for liquid systems is two orders of magnitude lower. The validity of this result is confirmed by observing the typical dimensions of industrial catalysts employed for gas- and liquid-phase reactions, respectively. Considering a typical intrinsic rate of catalytic reaction in an intermediate temperature range of 1 mmol/kg cats and a strongly exothermic reaction enthalpy indicates that the rates of heat generation one needs to master can be of the order of 500 kW/m3 of catalyst. For tolerable temperature gradients (2 K) and characteristic thermal conductivity values for porous catalysts (1 W/mK) one obtains the result that specific heat exchange surface areas of 1000 m2/m3 should be sufficient to remove the heat of reaction. This analysis thus implies that while the traditional multitubular reactor with its tube diameters of several centimeters may well be overtaxed, the microreactor offers an unnecessarily excessive specific surface area, way beyond the demands actually being imposed by the chemistry of the catalytic system. The resulting insight is that, in terms of the heat transfer demands, a reduction of the characteristic dimensions down to the millimeter scale yields an entirely adequate performance (Figure 8) (9).


FIGURE 8 Analysis of the appropriate characteristic dimensions for specific heat transfer surface requirements in a chemical reactor exhibiting typical reaction rates. Da Damkoehler number, NTU number of thermal transfer units, Nu Nusselt number, a thermal diffusivity. (From Ref. 9.)

3.4.2.

Ceramic Catalytic Millireactors

Effective process intensification requires that all superfluous bottlenecks be removed, but only to the extent at which a controlling step—usually the intrinsic catalytic kinetics—can no longer be accelerated further. The considerations presented earlier suggest an optimal reactor structure to meet this criterion. Since the intrinsic catalytic activity may be expressed only when a suitable nanostructure provides access to the maximum number of active sites and because the mass of the catalyst rather than its external surface area determines the total activity, all microreactor concepts based on bulk metal catalysts or on thin catalytic films produced by CVD, sol-gel coating, or anodic oxidation on a metallic support are basically unsuitable because they provide insufficient catalytic material. In addition, the conflicting demands inherent in producing a catalytic layer with the nanostructure thickness and adhesion to the substrate desired often lead to unsatisfactory compromises, and new composite structures must be developed for each specific catalyst system. Whereas the traditional approach to catalytic microreactors has been to coat a metallic micro-heat exchanger structure with catalyst, a superior technique might be to produce a heat exchanger out of the proven catalytic material itself or something similar, i.e., to utilize a nanoporous ceramic substrate. For the small dimensions involved, the use of slightly more poorly conducting ceramic leads to no deterioration in performance in comparison to metallic walls and can even be advantageous for thermal efficiency by maintaining “axial” temperature gradients without conductive short-circuiting.


FIGURE 9 Monolithic catalyst modified to serve as a heat exchanger.

The use of proven catalyst recipes would greatly curtail development times, and the absence of extraneous material avoids unwanted catalytic effects and enhances thermal stability. The fixation of catalyst on ceramic substrates such as washcoats is a well-known, reliable, and relatively straightforward procedure. The fabrication of complex small-scale ceramic structures is, however, more awkward than for metals or plastics, and they exhibit relatively poor mechanical strength. Furthermore, the porous ceramic nanostructure must be sealed to prevent contact between the reaction medium and coolant. In the past, the principles described have been implicitly recognized in several attempts to convert monolithic catalysts into catalytic heat exchangers. While the use of millimeter dimensions and nanoporous ceramic supports meets the primary criteria already mentioned, the parallel channel structure of monoliths is not ideally tailored for heat exchanger applications, and complex header structures are required to uniformly distribute and collect reaction medium and coolant to and from the individual channels (Figure 9). The unsatisfactory interface between the “milli-” and “macroscale” has been a major weakness of such concepts.


3.4.3.

Ceramic Catalytic-Plate Heat Exchanger

An existing item of chemical processing equipment—the plate heat exchanger— suggests a possible solution to this drawback. The integration of manifolds into the plate stack arrangement means that one can dispense with special headers for fluid distribution. Furthermore, internal structures on the plates can be used to provide intermittent mixing and induce secondary flows, thus overcoming the principle shortcomings of simple laminar-flow conditions. A variety of techniques, including an efficient countercurrent CVD method, have been proposed for sealing porous ceramic structures to prevent the unwanted diffusive slip between reaction medium and coolant (10). Although the fragility of ceramic structures probably precludes the application of evaporative cooling or molten salt coolants, the use of oil up to 300°C and gas for higher temperatures is feasible, since the poor performance of gas coolant can, to some extent anyway, be compensated for using by a favorable arrangement of the coolant-side geometry to cut the pressure drop. The coolant pressure in both cases can be adapted to approximate that of the reaction medium, thus reducing to a minimum the mechanical stress on the ceramic plates due to pressure differentials. One can also select the coolant gas to have good thermal properties (heat capacity, conductivity) and to be inert, so slight leakages into the reaction medium can, if necessary, be tolerated. Heat could be withdrawn effectively from the gas coolant circuit using an external inert fluidized-bed or other proven techniques. The resulting piece of equipment can be referred to as a “catalytic plate heat exchanger,” with structure and scale similar to that of a conventional plate heat exchanger. The plates comprise porous ceramic of 1- to 2-mm thickness and separate alternating reaction and cooling chambers (Figure 10). The detailed channel profile desired may be introduced on to the plain support by means of an intermediate laser-engraved separating plate. The individuals plates are stacked and cemented with countercurrent, cocurrent, or crossflow cooling configurations, as specified. More complex temperature profiles can also be attained by combining individual stack modules as required. Special attention must be paid to the uniform distribution of reaction medium and coolant both within and between the plates, to ensure an approximation to plug-flow behavior despite the usually unavoidable laminar-flow conditions. Fortunately, such flows are amenable to precise modeling using computational fluid dynamics, enabling one to develop suitable geometries. Nevertheless, the delusion that numbering-up is a trivial procedure quickly dissipates when one is confronted with the complexity of guaranteeing almost identical flow conditions in each of several ten thousand channels! In addition to the harmonization of the underlying physical and chemical processes, the catalytic-plate heat exchanger offers a cost-effective alternative to both conventional multitubular reactors and catalytic microreactors for industrial


FIGURE 10 Schematic layout of a catalytic plate heat exchanger.

applications. Because it is based on proven catalyst compositions and state-of-the-art fabrication methods, development costs and times should be low, a feature enhanced by the nonspecific, generally applicable nature of the concept. The successful design of such reactors can be ensured by the use of reliable modeling tools and the flexibility available in the reactor architecture and operation. Finally, the facile regeneration, recycling, or disposal of deactivated catalyst is an additional advantage over metallic catalyst composites. 4.

REACTIVE-RECUPERATIVE HEAT TRANSFER

A second method of process intensification for recuperative reactors is to enhance performance by using a reactive coolant or heating medium, since the heat effects associated with reactions are usually much larger than those available with phase changes or simpler heating and cooling procedures. The coupling of an exothermic auxiliary heat source reaction with the desired endothermic reaction, or vice


versa, through a double catalytically coated wall has also been attempted in a modified monolithic catalyst (Figure 11) for the steam reforming reaction, with the methane combustion reaction being used to supply the heat required (11). The work done demonstrated that it is extremely difficult to localize the reactions and regulate the temperatures in the interests of performance—the system tends to assume a state dictated by the intrinsic kinetics and heat effects of the two reactions. With more “passive” coolants or heating media, the operator is in a better position to influence the reactor behavior and can manipulate temperature profiles to a greater extent. 5.

REGENERATIVE HEAT TRANSFER

Regenerative reactors, that is to say, those exploiting heat storage on fixed beds, remain a somewhat neglected option in reaction engineering. Although the principle of heat regeneration had been utilized previously in the chemical industry,

FIGURE 11 Monolithic catalyst adapted for the thermal coupling of endo- and exothermic reactions. (From Ref. 11.)


for example, in a process for high-temperature thermal for nitrogen fixation (12), and is a common feature in related process industries, such as power generation and steel manufacture, its most recent renaissance was brought about the development (13) and refinement (14) of the reverse-flow reactor concept (Figure 12), which has now established itself firmly in the niche of oxidative waste-gas treatment. Additionally, regenerative heat transfer is an important feature of conventional reactor operation, for instance, in the occurrence of the transitional “wrong-way” behavior with temperature levels in excess of the adiabatic value that can arise when a fixed-bed reactor is shut down (15). Regenerative heat exchange in chemical reactors offers clear benefits, such as simplicity, robustness, low costs, and high efficiencies, against which must be set its inherently unsteady-state operation, the limited potential for an exact regulation of temperature profiles, and the restriction of its use to gaseous reaction media. 5.1. Comparison of Regenerative with Convective, Recuperative, and Reactive Heat Transfer In the evaluation of the regenerative heat exchange option, it is instructive to consider the heat exchange techniques presently employed in the following chemical processes: styrene synthesis, steam reforming, and hydrogen cyanide production (Table 1).

FIGURE 12 Reverse flow reactor concept.


TABLE 1 Heat Exchange Employed for High-Temperature Endothermic Reactions Ethylbenzene Dehydrogenation

Steam Reforming

Hydrogen Cyanide Manufacture

C8H10 ↔ C8H8 H2

CH4 H2O ↔ CO 3H2

CH4 NH3 ↔ HCN 3H2

600C

900C

1200C

T-Profiling technique Convection

Badger/Mobil “adiabatic” process

Recuperation

BASF “isothermal” process

conventional primary steam reforming

Regeneration Reaction


Degussa BMA process ?

autothermal reforming (fuel cell applications)

Andrussov ammonoxidation process

The moderately endothermic dehydrogenation of ethylbenzene to styrene is carried out catalytically at temperatures of around 600°C. The dominant technology for meeting the heat requirements of this reaction involves the intermediate introduction of superheated steam as a heating medium between adiabatic reactor stages, i.e., a convective process. Several additional features of the reaction favor the convective approach: Steam can easily be removed from the products by a phase separation following condensation, steam acts as an inert dilutant enhancing the equilibrium conversion, and steam also helps to maintain catalytic activity. An alternative recuperative technology in which combustion gases are used as a spatially segregated heat source in a multitubular suffers from the low heat transfer coefficients in gas–gas recuperative heat exchangers and entails a more costly reactor construction. The steam reforming of methane to synthesis gas, a strongly endothermic reaction carried out catalytically at around 900°C, primarily utilizes recuperative heat transfer. The higher reaction temperature makes the convective supply of heat problematic, particularly at the preferred operating pressures of about 25 bar. The heat required can also be supplied reactively by simultaneously carrying out the exothermic partial oxidation of methane to carbon monoxide and hydrogen. The main impediment for this so-called “autothermal” reforming was previously the rapid deactivation of the catalysts used. In connection with research on the “on-board” generation of hydrogen for mobile fuel cell applications, novel noble metal catalysts have been developed that maintain an adequate activity. Such catalysts are, however, almost certainly too costly for regular industrial purposes. The prevalent manufacturing process for hydrogen cyanide—the Andrussov process—represents the successful industrial application of “reactive” heat exchange. In the catalytic ammonoxidation of methane at 1100°C, the actual endothermic synthesis reaction between methane and ammonia is thermal supported by the oxidation of the hydrogen formed, so no additional heat need be supplied (Figure 13). Reactive heat transfer of this sort is, of course, very efficient, in that it entails virtually no temperature gradients, due to the almost molecular scale at which it occurs. On the other hand, the inclusion of an extra reactant—oxygen—gives rise to unwanted side reactions, such as the formation of carbon oxides, and encourages undesirable ammonia decomposition. Furthermore, because it is uneconomic to use pure oxygen in place of air, the nitrogen introduced into the process results in the need for much larger reactors and downstream units to treat the diluted gas streams. The alternative recuperative BMA process offers much higher yields, due to the absence of side reactions, and higher product concentrations with lower flows. Recuperative heat exchange at such high temperatures necessitates the use of catalytically coated ceramic tubular reactors, which are costly and lack operational robustness. In addition, only about half of the heat supplied in the form of combustion gases can actually be utilized for the reaction.


FIGURE 13 Possible thermal coupling mechanisms between the endothermic synthesis reaction and the exothermic hydrogen oxidation in hydrogen cyanide manufacture.

5.2. Regenerative Heat Transfer Process for Hydrogen Cyanide Manufacture Although regenerative heat exchange solutions have been proposed for the first two examples presented (16), hydrogen cyanide synthesis is, in fact, a more promising candidate by virtue of the higher and broader temperature range and the lower operating pressures. A cursory analysis reveals that a regenerative heat exchange process could in fact combine the most important advantages of the reactive and recuperative processes while surmounting their shortcomings (Table 2) (17). More detailed modeling exposes some weaknesses, for example, the need to use coarse monolithic catalyst structure to achieve reasonable reheat periods without excessive pressure drops. The cycle time of approximately 4 minutes is somewhat short for practical purposes. The “cold spot” formation in the reaction phase (Figure 14) and the resultant inability of the reaction to distribute itself over the catalyst to utilize the stored heat optimally is probably a modeling artefact caused by too low literature values for the activation energy, possibly reflecting an incorrectly interpreted film transport limitation.


TABLE 2 Advantages and Disadvantages of Reactive, Recuperative, and Regenerative Heat Exchange in Hydrogen Cyanide Manufacture

5.3. Desorptive Cooling for Enhanced Regenerative Heat Transfer Process intensification measures to extend regenerative cycle times could enable the reaction engineer to utilize the strengths of this form of heat exchange more fully. The use of the larger enthalpies associated with phase changes rather than simple specific heat effects to store thermal energy is a technique already exploited with evaporative coolants in recuperative processes and in the liquid injection mentioned earlier in the context of convective processes. The regenerative analogy to this principle can be illustrated using a technique that can be referred to as desorptive cooling (18). If an inert material is initially adsorbed on the fixed bed comprising an appropriate adsorbent and a catalyst, the heat of adsorption—having the same order of magnitude as the latent heat of evaporation—will be released (Figure 15). Since no reaction takes place in this phase, moderate temperature excursions are acceptable, and recycle flows over external heat exchangers or injection of liquid adsorptives may serve as heat sinks. In the subsequent reaction phase, the heat liberated by an exothermic reaction on the catalyst is taken up by the desorption of the inert from the previously loaded neighboring adsorbent particles. As long as this desorption occurs, the heat of reaction will not lead to major temperature increases. Sooner or later, of course, the adsorbent will be depleted and the temperatures will drift upward, at which point the adsorption phase must be repeated. Such a system yields intensive cooling without the need for an extensive heat exchange surface within the reactor. The operation of the reactor in the reaction


phase is adiabatic. One can even customize the heat removal process by distributing the adsorbent and catalyst in different ratios along the length of the fixed bed to maximize cycle duration. Furthermore, if the desorption process exhibits greater sensitivity toward temperature changes than the reaction, the process will become self-regulating to a certain extent. On the negative side, desorptive cooling is, as are all regenerative processes, inherently unsteady state, space time yields are diminished by the presence of adsorbent in the fixed bed, and the adsorbent-adsorptive system must be compatible, i.e. inert, with respect to the catalytic reaction being conducted. To demonstrate the potential available, simulations were carried out for the oxidation of carbon monoxide on a palladium shell catalyst with water desorption from 3A zeolite as a heat sink, based on experimentally validated model parameters for the individual steps (Figure 16). The calculations indicated that the reaction cycle time could be lengthened by a factor of 10, to a total 20 minutes, in comparison to a simple regenerative process with a similar amount of inert material instead of adsorbent in the fixed bed and for the same threshold for temperature deviation from the initial value.

FIGURE 14 Temperature profiles of gas and catalysts in a regenerative process for hydrogen cyanide manufacture at the start and finish of the reaction cycle. Cycle duration is 4 minutes and the monolithic catalyst used has the following dimensions: wall thickness 4 mm, channel width 20 mm, length 2000 mm.


FIGURE 15 Operating principle of “desorptive” cooling.

Experimental measurements yielded only a fivefold extension of the reaction cycle time, a difference largely caused by heat storage effects in the smallscale equipment used, which disproportionately enhanced the cooling effect observed in the inert bed control experiment. Despite this less satisfactory result, the desorptive cooling concept would still seem to offer potential for dramatic improvement in performance for regenerative heat exchange processes. The desorptive cooling principle is effectively equivalent to the distributed injection of a cooling liquid along the length of a fixed-bed reactor into a gaseous reaction medium undergoing exothermic reaction. An extension of the idea, the adsorption of a reaction product to both enhance the equilibrium position and provide some of the heat required for the endothermic reaction, has also been proposed (19). More mundanely, the latent heat effects of wax solidification have been exploited in temperature-regulating fabrics incorporating microencapsulated wax particles! A certain analogy can also be drawn with the previously mentioned use of catalyst bed dilution with inert material to better harmonize recuperative


heat removal with the rate of heat formation due to reaction, even though this measure must be considered a example of process deintensification in the strictest sense. 5.4.

Regenerative Heat Transfer in Adsorptive Reactors

The benefits obtained by integrating adsorption into regenerative heat exchange demonstrate the synergies available between these two related processes. Similar advantages accrue when heat regeneration is incorporated into adsorptive reactors, in which concentration profiles are manipulated to improve reactor performance through selective ad- and desorption of components in the reaction medium. An example of the potential in this second direction is provided by an innovative regenerative reactor for carrying out the Deacon reaction (20), in which hydrogen chloride is catalytically oxidized to chlorine—an important step in chlorine recycling for the chemical industry. By resolving the gas-phase reaction into two sequential gas–solid subreactions corresponding to adsorption-desorption steps (Figure 17), one can overcome the equilibrium thermodynamics that otherwise

FIGURE 16 Simulated temperature profiles along a reactor with and without “desorptive” cooling at various times for the oxidation of CO on a Pt catalyst with water vapor desorption from 3A zeolite in a fixed bed comprising equal proportions of catalyst and adsorbent. The solid curves give the simple regenerative behavior and the dotted curves describe the desorptively cooled case. Initial reactor temperature is 125°C, initial adsorbent loading 0.12 kg/kg, inlet CO-concentration 0.2 mol/l, gas loading 6000 h1.


FIGURE 17 Operating concept of an unsteady-state Deacon process for chlorine recycling.

limit conversion to around 85% and complicate downstream processing of the resultant partially converted reactor output. In the exothermic chlorination step, hydrogen chloride reacts with the oxidized form of the catalyst to yield copper chloride and water vapor. During the following endothermic oxidation stage, this copper chloride is converted back to its oxidized form, liberating chlorine in the process. In this manner, one can achieve total conversion and considerably simplify the subsequent processing necessary. In view of the extremely corrosive nature of the reaction (and catalyst!) system a regenerative transfer of heat between the two phases is to be preferred over any recuperative alternative. The use of the fixed bed to store both heat and chlorine between the two steps of the reaction cycle leads to contradictions and bottlenecks in catalyst and reactor design—a common drawback of multifunctionality as a technique for process intensification. For example, the removal of the excess heat of reaction from the reactor favors cocurrent operation of the fixed bed in its function as a heat regenerator. For optimal utilization of the concentration profiles and maximal chlorine capacity of the fixed-bed countercurrent, operation of the two cycle phases is to be preferred. Despite these difficulties, it still proved possible to modify the catalyst and reactor design sufficiently to fulfill both the “adsorptive” and the “regenerative” function. 6.

ELECTROMAGNETIC HEAT TRANSFER

The use of electromagnetic techniques represents the final method for process intensification of heat transfer that will be dealt with here. Developments in the fuel cell sector hold out the promise of cheap electric power based on surplus hydrogen in chemical plants. The liberalization of the energy sector has also cut electricity prices and enhanced the attractiveness of using electricity in connection with chemical reactors. The precise regulation possible with electrical processes and their clean, environmentally friendly nature are further inducements for their application.


Once again it must be said that the use of electricity in chemical reactors is hardly new. Apart from numerous electrochemical processes, electrothermal processes are employed for the manufacture of phosphorus, hydrogen cyanide, and calcium carbide. Furthermore, syntheses based on electric arcs have been used in the past for the large-scale production of nitric oxide and acetylene and are still employed today for ozone manufacture or chemical vapor depostion coating procedures. A major advantage of the electrical heating employed in electrothermal processes that offsets the higher operating costs is the absence of any limit on the temperature level at which it can be applied, making it particularly suitable for processes above 1000°C, where conventional heat transfer becomes difficult to realize. 6.1.

Ohmic Heating

Ohmic heating of catalyst is often used as a simple method of igniting the chemical reaction during reactor startup, for instance, in the oxidation of ammonia on platinum-rhodium gauze catalysts. Another application is the prevention of “coldstart” emissions from automotive catalysts responsible for much of the residual pollution still produced from this source (21). The startup times needed for the catalyst to attain its operating temperature can be cut by a factor of 5 or more by installing an electrically heated catalyst element with a metallic support upstream of the main catalyst unit. Direct electrical catalyst heating permits facile temperature control but requires a well-defined catalyst structure to function effectively. 6.2.

Dielectric Heating

Much attention has recently also been devoted to dielectric heating of reactors using microwaves (! 1 cm to 1 m, 30 GHz to 300 MHz) (22). As in domestic applications, the primary attraction lies in the absence of heat exchange surface, and thus of fouling, and local overheating. Dielectric heating is especially suitable for temperature-sensitive materials for which even slight nonuniformities or temperature gradients might prove damaging, which explains its use in the manufacture of ceramics and catalysts and in the sterilization of complex fermentation media. It has also been proposed for the local production of limited quantities of hazardous chemicals, such as hydrogen cyanide (Figure 18) (23). The rapid startup and exact temperature regulation possible, even at varying throughputs, together with the low inventories entailed by such production reactors compensate for the higher costs of the microwave heating. The speed of microwave heating also makes it a suitable technology for the oxidative regeneration of diesel particulate filters. Although the efficiency of microwave heating is high, the generation of microwaves is by no means free of losses. Specific chemical effects due to microwaves


FIGURE 18 Yields of hydrogen cyanide in catalytic and noncatalytic microwave-heated reactors. (From Ref. 23.)

remain controversial and are probably of little significance for industrial purposes. Questions of scale-up, capacity limits, and materials of construction still need to be addressed before industrial applications of microwaves in combination with chemical reactors become more widespread. An interesting application, derived from the use of microwaves for selective desorption processes, is the modification of catalyst performance by the imposition of a temperature profile on a catalyst pellet, which is usually dictated by the interaction between the heat of reaction and the thermal conductivity of the pellet. Microwave heating together with the use of carrier materials of various permittivities and conductivities would permit one to regulate the temperature conditions within the catalyst pellet independently. An extension of this principle would be the selective thermal activation of one sort of catalyst in a mixed fixedbed system. The external “switching” of catalytic activity in this manner could be employed expediently to realize multistep syntheses in a single reactor. To prevent thermal short-circuiting it would be necessary to isolate the individual catalyst particles from one another in an insulating matrix. 6.3.

Electric Arc Processes

Electric arc processes have been given a new lease on life in the guise of plasma reactors, especially those involving “cold,” or nonthermal plasmas, with electron “temperatures” of 104–105 K and gas temperatures of 102–103 K. Plasmas of this kind can be used to activate and functionalize inert molecules, but usually with only poor selectivities and low energy yields (⬃ 0.01 mol/kWh!). The use of catalytic additives may offer some potential for improvement, but reactive plasma processes will probably remain restricted to a few specific applications.


6.4.

Peltier Cooling

A major drawback of the electrical techniques described is that they are, without exception, heating processes, whereas the reaction engineer is often more interested in reactor cooling. The active cooling of a catalytic surface using the Peltier effect enables one to achieve almost perfect temperature control, even in the face of strongly exothermic reaction behavior, but the low efficiency of the phenomenon means that a large additional amount of heat also has to be dissipated. As a Peltier element only “pumps” heat from one location to another (Figure 19), the question of the ultimate heat sink must still be resolved. For these reasons such techniques will probably be restricted to laboratory use in the form of special microreactors for the foreseeable future. 7. CATALYST MODIFICATION FOR ENHANCED HEAT TRANSFER All of the recuperative, regenerative, and electrical methods described for intensifying heat transfer in chemical reactors illustrate the importance of developing appropriate catalyst systems tailored to the behavior being sought. Improving catalyst performance in this way by using “active” multifunctional supports while leaving the catalytic chemistry unchanged can be referred to as “commensal” catalysis, in analogy to the natural relationship between two species that benefits one partner (the catalyst) while being neither advantageous nor disadvantageous for the other (the carrier). Bifunctional catalysts, with spillover diffusion effects to reduce coke formation, and zeolitic catalysts with their selective access to active sites represent two conventional examples of commensal catalysis. However, the support properties can also be modified to enhance interfacial areas, heat and mass transport, heat and mass storage, mechanical, thermal, or chemical

FIGURE 19 Active electrical reactor cooling using Peltier elements.


TABLE 3 Performance Enhancement Through the Modification of the Physical Properties of Catalyst Supports • Bifunctional Catalysts e.g. propene oxidation, Pt-doped zeolites, spillover-oxygen • Supported Catalysts e.g. impregnated catalysts, SLPC, SAPC • Mass Transport e.g. zeolites, membrane encapsulation, Aerogel • Heat Transport e.g. graphite carriers, coated-wall reactors, full metal catalysts • Mass Storage e.g. active C-carriers, hydride-containing catalysts • Heat Storage e.g. metallic monoliths • Mechanical Resistance e.g. protective coat, “washcoat”, gauzes • Thermal Resistance e.g. non-oxide ceramics, doped Ba-hexaaluminate • Chemical Resistance e.g. non-oxide ceramic, silicon dioxide, Al-phosphate • Electromagnetic Properties e.g. dielectrically heated catalysts, magnetic fluidized beds

resistance, and electromagnetic behavior (Table 3). The use of such microstructured hybrid catalysts can make an important contribution to process intensification measures in other areas as well. 8.

SUMMARY

The reaction engineer has a variety of tools at his disposal when attempting to intensify heat transfer in chemical reactors, ranging from well-established methods to innovative technologies. For recuperative heat transfer, the most dramatic improvements can be achieved by using catalytic or catalytically coated heat exchange surfaces and working at the millimeter scale to harmonize the physical and chemical processes taking place and render the catalytic chemistry the performance-limiting step. These two measures overcome the most serious bottlenecks in the traditional multitubular reactor. The operation of regenerative and reactive processes for the manipulation of temperature profiles in chemical reactors is usually more complex than convective and recuperative techniques, due to the inherent dynamics and high sensitivities involved. Regenerative and reactive processes can, however, permit


a more active (self-) regulation of the temperature conditions within the reactor, which can, in some cases, simplify their implementation. The full potential of hybrid operation, employing more than one of the fundamental processes for manipulating temperature and concentration profiles, remains to be realized. Special catalysts and reactors must be developed to accommodate the conflicting demands that often arise in the design process. REFERENCES 1.

2. 3. 4. 5. 6.

7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17.

Nicol W, Hernier M, Hildebrandt D, Glasser D. The attainable region and process synthesis: reaction systems with external cooling and heating. Chem Eng Sci 2001; 56:173–191. DECHEMA. Trendbericht No. 5: Prozessintensivierung. Frankfurt: DECHEMA Gesellschaft für Chemische Technik und Biotechnolgie eV, Press release, 2000. Levenspiel O. The Chemical Reactor Omnibook; Corvallis, Or: OSU Book Stores, 1996. Eigenberger G. Fixed-bed reactors. In: Ullmann’s Encyclopedia of Industrial Chemistry. Weinheim, Germany: VCH, 1992, Vol. B4:199–238. Krishna R, Sie ST. Strategies for multiphase reactor selection. Chem. Eng. Sci. 1994; 49:4029–4065. Lahne U, Lohmüller R. Schüttschichtreaktoren mit gewickelten Kühlrohren, eine konstruktive Neuentwicklung zur Durchführung exothermer katalytischer Prozesse. Chem Ing Tech 1986; 58:212–215. Ehrfeld W, Hessel V, Löwe H. Microreactors. Weinheim, Germany: Wiley VCH, 2000. Gavrilidis A, Angeli P, Cao E, Yeong KK, Yan YSS. Technology and applications of microengineered reactors. Chem. Eng. Res. Des. 2002; 80:3–30. Gerhardt W. BASF AG. Personal communication, 2000. Dummann G, Pahlke T, Agar D. Versiegelung microporöser Strukturen mit Gegendiffusion-CVD. Chem Ing Tech 2002; 74:824–827. Frauhammer J, Eigenberger G, Hippel LV, Arntz D. A new reactor concept for endothermic high-temperature reactions. Chem. Eng. Sci. 1999; 54:2661–3670. Hendrikson WG, Daniels F. Fixation of atmospheric nitrogen in a gas-heated furnace. Ind. Eng. Chem 1953; 45:26113–2615. Matros YS. Catalytic processes under unsteady-state conditions. In: Studies in Surface Science and Catalysis. Amsterdam: Elsevier, 1989, Vol. 43. Nieken U. Abluftreinigung in Katalytischen Festbettreaktoren bei Periodischer Strömungsumkehr. VDI-Fortschrittberichte No. 328; Düsseldorf: VDI-Verlag, 1993. Pinjala V, Chen YC, Luss D. Wrong-way behavior of packed-bed reactors. II. Impact of thermal dispersion. AIChE J 1988; 34:1663–1672. Kolios G, Eigenberger G. Styrene synthesis in a reverse-flow reactor. Chem. Eng. Sci 1999; 54:2637–2646. Agar D. Multifunctional reactors: old preconceptions and new dimensions. Chem. Eng. Sci. 1999; 54:1299–1305.


18.

19.

20.

21.

22. 23.

Franke M. Bewertung desorptiver Kühlung von Festbettreaktoren. Master’s thesis, Chemical Engineering Department, University of Dortmund, Dortmund, Germany, 2000. Yongsunthon I, Alpay E. Design of periodic adsorptive reactors for the optimal integration of reaction, separation and heat exchange. Chem. Eng. Sci 1999; 54: 2647–2657. Agar D, Watzenberger O, Hagemeyer A. A novel unsteady-state process for HCl oxidation—multifunctional reactor operation with regenerative Cl storage in a fixed bed. In: Proceedings of R ’97 Meeting, Recovery, Recycling, Re-integration. Geneva, Switzerland, 1997: Vol. 4, IV.45–IV.50. Kirchner T, Eigenberger G. Optimization of the cold-start behavior of automotive catalysts using an electrically heated precatalyst. Chem. Eng. Sci. 1996; 51: 2409–2418. Bathen D, Schmidt-Traub H. Alternative für Nischen—Anwendungspotentiale der Mikrowellentechnologie in der Verfahrenstechnik. ChemieTechnik 1998; 27:80–83. Lerou J, Ng K. Chemical reaction engineering: a multiscale approach to a multiobjective task. Chem. Eng. Sci. 1996; 51:1595–1614.


11 Process Synthesis/Integration Patrick Linke and Antonis Kokossis University of Surrey, Surrey, England

Henk van den Berg University of Twente, Enschede, The Netherlands, and Ghent University, Ghent, Belgium

1.

INTRODUCTION

Process synthesis, also referred to as process integration, deals with the systematic development of process flowsheets. The process synthesis activity has been described as “the automatic generation of design alternatives and the selection of the better ones based on incomplete information” (1). Design technology is required to help the engineer find novel, improved solutions to process design problems in the context of the incomplete information available. The ultimate aim of chemical process design is to synthesize a process that enables the production of desired chemicals in the most cost-effective and environmentally benign manner possible and is flexible as well as easily operated. Ideally, process synthesis tools should allow one, out of the set of all feasible alternative structural and operational process design options, to systematically determine the most promising process designs, those that approach the performance limits of the system closely and meet the constraints. An enormous variety of decisions need to be made in order to solve process design problems. These range from the selection of the most promising process chemistry (reaction paths and catalysts) to the optimal exploitation of reaction,


mass, and heat transfer interactions at the process level. In order to arrive at a truly optimal design, all decisions have to be made with respect to the overall objectives in order to achieve the best balance of process trade-offs. Due to the complexity of the overall problem, applicabilities of systematic decision-making technologies are generally confined to closely defined design subproblems. The “onion” model (see e.g., Ref. 2) captures a shared view that the process design problem is best decomposed into a reactor design (the heart of the process) subproblem, a separation system design subproblem, an energy system design subproblem, and a utility system design subproblem. These subproblems appear challenging for a variety of reasons. Challenges in reactor optimization arise mainly from the high nonlinearities that need to be addressed. In energy systems such as heat exchanger networks, on the other hand, the challenge is to search the vast number of combinatorial design options. Separation system design presents examples of intermediate complexity in terms of model nonlinearities. Despite the challenges, the design problem decomposition itself leaves plenty of unanswered questions to be addressed by future research. Clearly, there is significant scope for improvement by looking at a bigger picture and systematically exploiting interactions between solutions to the classical design subproblems. Recent developments in process synthesis technology aim at giving a more global perspective by solving less confined but more conceptual problems in a variety of areas. Process synthesis and integration tools are generally developed in academia. Successful technology development requires close collaboration between the academic and the industrialist to ensure applicability to real-life problems. The most successful example of process synthesis and integration technology development is pinch analysis (3). After many years of close academia–industry collaboration, pinch analysis has led to significant energy savings in the chemical industries and has become a standard tool employed in most energy system synthesis projects worldwide. In contrast to the simpler techniques for energy systems synthesis that have relatively quickly been taking up applications in industry, the methods for separation and reaction system synthesis have been slow in being widely accepted by industrialists. However, these tools have the potential to significantly help design engineers in their quest to develop new plants faster and with fewer resources. The next section will reflect on common conceptual process design practices for overall flowsheet development. The remainder of this chapter reviews recent developments in process synthesis methods for reaction and separation systems that systematically guide the design decision-making process toward novel and improved designs. 2.

CONVENTIONAL CONCEPTUAL DESIGN PRACTICES

In this section a number of issues of process synthesis and conceptual design will be reviewed. Based on available textbook knowledge (2,4–6), a synergistic


approach is discussed that uses project organization, decomposition, process building, selection of alternatives, and design rounds. Process functions are taken into account at a level above the unit operations. Due to the importance of project organization (6), management techniques also need to be applied to process engineering. This section reflects on experiences with industrial applications as well as graduate courses on process design. New aspects on the combination and arrangement of activities will also be discussed (7). 2.1.

Elements of Conceptual Process Design

This section deals with a number of basic points of process design strategy, cornerstones that have to be applied in a synergistic procedure. The application realizes a synergy of process functional analysis, process synthesis, and project organization to generate structure and trace alternatives and decisions. This synergy in conceptual process design uses the elements of heuristics, mathematics, and creativity. Components of conventional design approaches include: Existing flowsheet and technology analysis—collection of know-how. Flowsheet decomposition into functional sections where tasks prevail over equipment and unit operations. A black box approach as the endpoint of the analysis; raw materials and products, overall process-relevant data (e.g., yields, reactor conditions), and boundary conditions are given, leading to a first evaluation of the overall process; economics are based on raw materials and products. Formulation of goals for the new design; the black box is the start of the rebuilding of the process from outside to inside. Systematic process rebuilding, while including known and creative technologies and breakthroughs and using heuristics, expert systems, and process simulation. Application of management tools in the form of tree diagrams (interrelationship of goals and means) and work diagrams (to structure the project step by step), document alternatives and choices. The approach has similarities with a bow-tie model: One starts broad and collects all available process information. A structure is used, e.g., a tree diagram, to create a format for the information needed and collected. One should concentrate on the essentials and collect all relevant factors that influence the process, from the outside to the inside. Setting the boundaries of the process early reduces rework later on; however, this clearly limits the degree of design novelty that can be achieved. Phasing of process design is a normal issue in the process industries. A process design or substantial modification is done stepwise. Sequential steps in process design are carried out after management approvals of the process proposals. This means that project organization is of vital importance to direct activities and to


avoid rework. This aspect is hardly covered in the textbooks for process synthesis. Biegler et al. (5) use several levels or loops in their book, but they are not linked to the phases of process design. The concept of process flowsheet decomposition is used by Douglas (4) and Biegler et al. (5) but needs more emphasis from the moment a project is started. Flowsheets have to be translated into functional activities so that a critical and systematic analysis can be started early. The results can be used during the buildup phase after the black (grey) box model development. Conventional technology is easily used in the buildup phase because it is mature and has less or no risk, but new technology has to be taken into consideration to enable step changes in performance improvements and to avoid copy engineering. All design textbooks mention that important decisions are made in the early stages of design and that the know-how to make these decisions is developed gradually during the design process. The consequences of choices made in early stages of design cannot be foreseen at the moment of the decision. It is clear that minor decisions about the type of a centrifugal pump to be used for the reflux of a distillation column are less important in the early stages of design than in the performance of the reactor. Douglas (4) advocates the application of rejection of less attractive alternatives and proposes to continue the development of all alternatives that cannot be rejected. The conventional way is to continue with a single alternative; however, this tends to cause problems later in the project. Biegler et al. (5) and Seider et al. (6) show how to document alternatives and choices. Criteria for rejection are, e.g., process yield and selectivity, costs, safety, ecology, and reliability of equipment. These criteria and the decision process have to be documented. It must be possible to return to the design path in case the results have to be rejected and early decisions need to be reconsidered, as shown in Figure 1. A systematic process design cannot be done in one step. It is common practice to develop a draft flowsheet first. This will be a feasible, unoptimized flowsheet,

FIGURE 1 Alternatives and choices.


where the consequences of the choices made are not yet clear. Biegler et al. (5) refer to this as the base-case design and propose four sequential rounds to complete the design of the process. The process know-how is extended from round to round, and the consequences of choices from among alternatives become more and more clear. In the fourth step Biegler et al. (5) use optimization techniques to search for design alternatives. Ideally, the project phasing applied in the process industry and the growing ability of the process engineers to collect and judge alternatives during the project need to be matched. The next section describes such an approach. 2.2.

Step-by-Step Conceptual Process Design

This section gives a short description of a step-by-step design strategy that applies topics treated in the previous section. Assume that the task of a process is to convert raw materials A and B into product C and to minimize by-product or waste D. The states of materials A, B, C, and D are defined by: mass flow, composition, phase (vapor, liquid, solid), form (e.g., particle size), temperature, and pressure (6). The process design has to consider alternatives for raw materials and process routes that lead to a given product. This is done in step 0 of the design activities. An early set design basis is often the production capacity (ton/yr). Each process operation can be viewed as having a role in eliminating one or more of the property differences between the raw materials and the products. The first step is to eliminate differences in molecular type by chemical reaction. The function of the reactor is usually considered the heart of a chemical process. Raw materials are seldom converted into the desired products at the required purities, so a separation function needs to be defined in Step 2. The need for and task of separation is often strengthened by the reactor conditions, e.g., excess of a component may be required inside the reactor. Generally, the reactor outlet needs the separation function to match the product specification. This information is collected in the decomposition stage of process design. Other functions in the process area that one needs to take into account are change of phase, phase separation, change of temperature and pressure, as well as mixing and splitting streams. Each operation can be viewed as having a role in eliminating one or more of the property difference between raw materials and products. The transition from raw materials to products is schematically represented in Figure 2, based on the information given in Seider et al. (6) and Biegler et al. (5). Alternative arrangements of reactors and separators are considered as a next design action. Both Douglas (4) and Seider et al. (6) consider the choice between batchwise operation and continuous operation a first choice in process design. Our experience is that a black box evaluation—a mass balance and a first economic evaluation—based on raw material and product flows, yields, and prices can be made without making a choice between batch and continuous operation.


FIGURE 2 Moving from synthesis steps to process operations.

Connections between process units are created in a third step while considering in detail the following aspects: From raw materials to reactor: purification and conditioning of feeds (purification, heating, pressure) From separator to product: final purification (finishing) Reactor–separator connections: stream reconditioning (e.g., reduction of temperature and/or pressure) Connections of separator outlets: reactor performance, excess components, recovery Recycles from separators: connections to, e.g., reactor system Step 4 consists of defining and matching the mass flows, including, e.g., the excess of one component required in the reactor. The activity starts with building preliminary mass balances before other conservation laws, i.e., heat and impulse, are considered. Other relevant factors that have to be evaluated early in process design are controllability and safety. The outcome of Step 4 is an overview of process functions needed to realize the transition of raw materials into products, including alternative flowsheets and unit operations that could be applied. Step 5 considers the combination of unit operations or task integration to develop the opportunities for process intensification, viz., leading to smaller, more efficient, and cheaper processes. This step includes screening for options related to heat integration and reactive separations, among others. Step 6 consists of the setup of the mass and heat balances, and alternatives for the reaction and separation functions are considered. The alternatives are documented in a structure such as the one shown in Figure 1 or the synthesis tree proposed by Seider et al. (6). Proceeding with the steps, many process alternatives are generated form which the most promising ones need to be selected taking


into account profit, process yield, controllability, safety, and ecological aspects. Criteria for assessing preliminary designs are presented in Ref. 5. Between the steps, interactions will take place and decisions can create the need for loops to previous steps, e.g., to reconsider the reactions (Step 1) and separation (Step 2) in more detail. Process design is generally not done in a first and single round; a next round is required to reconsider choices made, reduce the number of alternatives, and select the most promising option—while taking into account best available technology, heuristics, and creative new solutions. A second round, as proposed by Biegler et al. (5), is carried out to create a more in-depth conceptual design and to extend the process design activities, e.g., toward process simulation, heat integration, and equipment design. This is done in line with the project phases mentioned. Various industrial applications of the step-by-step design approach have been reported (see, e.g., Ref. 8). A number of these and other confidential projects have shown the value of the coordinated approach to conceptual process design. From the foregoing discussion it is clear that the process alternatives are commonly generated based on intuition and case-based reasoning. This leaves a strong chance that promising design candidates are not arrived at and that novelty is not automatically accounted for in the design process. Ideally, a systematic approach should be able to capture all possible design alternatives and screen for the design that delivers the best possible performance for the specified performance measure. The following sections describe such technology developments for reaction system synthesis, separation system synthesis, and integrated reaction– separation systems. 3.

REACTION ENGINEERING

The reactor is undoubtedly the most important ingredient of a chemical process flowsheet, for it is the part where the product value is generated. From a decisionmaking perspective, the reactor design is more of a difficult process design task than the separation and energy systems design tasks. This is because the reaction, heat transfer, and mass transfer phenomena tend to occur simultaneously in this unit. Most reaction models of commercially relevant systems are highly complex, and the development of both graphical and computational design tools that enable quick decisionmaking so as to obtain high-performance reactor designs is challenging. Such tools are required to provide performance targets and design suggestions for mixing and operational policies to give maximum decision support to the designer. There is an increasing awareness that the commonly employed textbook knowledge and heuristics (9) is insufficient for the systematic development of high-performance reactor designs. The result is a lack of innovation, quality, and efficiency in many industrial designs. Researchers from various perspectives are


making efforts to develop systematic optimization tools to improve the performance of chemical reactors. At a conceptual level, the area of reactor network synthesis has resulted in new methods that focus on a systematic and thorough consideration of the available options. In the remainder of this section, we describe recent developments in reactor network synthesis that provide systematic decision support for various reactor design aspects. Reactor network synthesis aims at identifying high-performance reactor design candidates that exploit mixing, feeding, bypassing, recycling, and temperature effects such that the systems performance is maximized with respect to the objective functions employed. 3.1.

Single-Phase Systems

Virtually all efforts in reactor network synthesis have addressed single-phase systems. Design approaches can be broadly divided into superstructure optimizationbased and graphical synthesis. Achenie and Biegler (10–12) were the first to synthesize comprehensive reactor superstructures using optimization technology. They developed superstructures using axial dispersion models, recycle-PFR representations, and environmental reactor models and applied optimization techniques in the form of nonlinear programming (NLP) methods to identify the most promising design candidates hidden in the them. Kokossis and Floudas (13–15) first introduced the idea of a reactor network superstructure modeled and optimized as a mixed-integer nonlinear programming (MINLP) formulation. Though general and inclusive, their representation did not follow previous developments but made an effort to facilitate the functionalities of the MINLP technology with the synthesis objectives. With the primary purpose of scoping, optimizing, and analyzing the reaction process, Kokossis and Floudas replaced detailed models with simple though generic structures, enough to screen for design options and estimate the limiting performance of the reaction system. In the same vein, dynamic components have been replaced with the use of CSTR cascades. A superstructure of generic elements (ideal CSTRs and PFRs) was postulated to account for all possible interconnections amongst the units (Figure 3). The representation was modeled and optimized as an MINLP model. Schweiger and Floudas (16) later revisited the approach and optimized superstructures with the PFRs, with side streams being replaced by rigorous DSR representations that avoid the inaccuracies introduced by the use of CSTR cascades. Around the same time, Glasser et al. (17) retrieved and extended the insightful methods of Horn (18) and presented graphical procedures known as the attainable region (AR) method. Their approach requires the graphical construction of the convex hull of the problem and helps to exemplify the need for a systematic and general methodology. In principle, the reactor network with maximum performance in terms of yield, selectivity, or conversion can be located on the boundary of the AR in the form of DSR and CSTR cascades with


FIGURE 3 Reactor superstructure. (From Ref. 31.)

bypasses. Though useful in two dimensions, in higher dimensions their developments face both graphical and implementation problems. Furthermore, the AR methods frequently result in complex designs with multiple DSR and CSTR units and complex feeding and bypassing strategies. Though fundamental limitations appear evident, persistent efforts to extend the graphical methods have been published (19–24). A more promising direction has been pursued by Biegler and coworkers. The motivation has been to instill better guarantees in the optimization efforts by exploiting ideas and rules established in the construction of the AR. Applications presented in this area include the work by Balakrishna and Biegler (25,26) and Lakshmanan and Biegler (27–29) and involved mathematical programming applications in the form of NLP and MINLP formulations. Hildebrandt and Biegler (30) presented a review of the attainable region approaches and suggested areas for future development of the concept. In 1999, Marcoulaki and Kokossis (31) presented a different interpretation of the synthesis problem. From a practical viewpoint, the nature of a useful approach for reactor network synthesis should primarily account for solid performance limits for the reaction system (targeting) and the systematic development of design candidates that approach this performance (screening and scoping). Targeting is particularly useful for the design decision making because it allows design evaluation in light of the ultimate performance possible for a given system, provided there is enough confidence in the optimization results. The targets can be used for synthesis and retrofit problems because they can provide the incentives to modify the operation and ideas for developing the reactor design. Because the design equations of chemical reactors feature a significant number of nonconvex terms, the importance of confidence assumes a significant place, and robust optimization technology is called for. On the other hand, the


borders of the attainable region and the strictly optimal solutions assume a relative meaning in reactor design. The reaction kinetics typically involve significant uncertainties and approximate models that give little justification for a strict emphasis on the optimum. In this context, the development of a CSTR-PFRCSTR-PFR-CSTR-PFR layout as an “optimal” configuration represents an academic exercise of limited interest. A single, slightly inferior CSTR can feature operational advantages and prove an even better choice in the form of an industrial back-mixed reactor. From a targeting viewpoint, layouts near the targets are considered equally important. Therefore, solutions in the interior of the attainable region make valid options as long as they remain close to the targets (31). Screening and scoping aims at identifying the range of design candidates that perform reasonably close to the targets so as to provide the design engineer with options on which the decision making can be based. The nonlinear, discontinuous, and discrete nature of the synthesis problem formulation considered here is not suitable to be addressed by mathematical programming techniques in the form of the commonly employed (mixed-integer) nonlinear programs ((MI)NLPs) because the synthesis aim is to establish performance targets in the different synthesis stages. Mathematical programming searches for local improvements and terminates at the nearest locally optimal point. For a general case, there is no reason to be confident that the obtained solution cannot be substantially improved. The type of information required from the targeting stage naturally relates to the results one can obtain with the application of a stochastic optimization approach to the reactor network superstructure synthesis. The application of stochastic optimization gives one confidence in the optimization results, can yield particularly nonlinear reactor models, and is not restricted by the dimensionality or the size of the problem. Marcoulaki and Kokossis have applied stochastic optimization in the form of the simulated annealing meta-heuristic to the single-phase reactor network synthesis problem (31). They optimize the rich and inclusive superstructures formulated by Kokossis and Floudas (13) to identify performance targets and to extract numerous design candidates that approach the targets. The implementation of the stochastic search over the superstructure schemes requires venues for the development and evolution of states, the assessment and acceptance criteria, and the cooling schedule. The synthesis perspective and the optimization methodology is discussed in detail in Marcoulaki and Kokossis (31). They found the methodology to systematically converge to the globally optimal domain, i.e., the performance targets of the system, and to produce numerous design alternatives with performances close to these targets for the numerous complex systems studied. To illustrate the methodology, consider the Denbigh reaction system. The reaction scheme involves five components and is described by: 2A  1→ B,

B 2→ C,

A 3→ D,


2B  4→ E

with the following kinetics: R [ R1 , R 2 , R 3 , R 4 ] [k1 x A2 , k2 x B , k3 x A , k4 x B2 ] The data are from Ref. 31. The objective for optimization is the maximization of the effluent concentration of component B. The performance limit of the system is identified with each stochastic run requiring an average of only 120 CPU sec on an HP 9000-C100 workstation. Numerous designs are obtained from the stochastic search that perform close to the performance target, mostly variations of series arrangements of PFRs and CSTRs. A detailed discussion of this and other studies is given in Ref. 31. 3.2.

Multiphase Systems

After fixed-bed reactors, multiphase reactors are the most widely used reaction systems in the chemical process Industries. The common industrial practice employs conventional designs based on empiricism, past experience, and qualitative reasoning on the basis of analogies with similar systems and case studies. The presence of multiple fluid phases in the system represents additional degrees of freedom that need to be exploited in process synthesis. As compared to the singlephase reactor network design problem, the multiphase case poses additional challenges in the form of a significantly larger number of possible network configurations as well as additional modeling complexities arising from the additional need to model multiple phases and to address mass transfer and hydrodynamic effects. Mehta and Kokossis (32) introduced a systematic methodology for the synthesis of multiphase chemical reaction networks that is based on a compact representation of design options. The approach accounts not only for conventional industrial reactors, such as bubble columns, cocurrent and countercurrent beds, and agitated reactors, but also for all possible combinations of compartments that can improve or enhance the performance of a multiphase reaction process. The representation is described in the form of a superstructure of generic synthesis units featuring shadow reactor compartments, and the synthesis scheme provides functionalities that are subjected to optimization. The implementation of the single-phase reactor network synthesis methodology of Marcoulaki and Kokossis (31,33) enables the development of targets and screening procedures that can help the engineer to assess system performance and review promising design options. The building block of the superstructure representation is the generic reactor unit, which follows the shadow reactor concept (32). This generic unit is illustrated in Figure 4. Each generic unit consists of reactor compartments in each phase of the system, and each processes the reaction. The shadow reactor compartment assumes a state from the set of homogeneous reactors. The default units in the set include CSTRs and PFRs with side streams. The interface between a given pair of


FIGURE 4 Generic multiphase reactor unit. (From Ref. 34.)

compartments in contacting phases establishes mass transfer links through which mass transfer with its shadows in other phases is accounted for. Mixing is accommodated for through links with other compartments of the same phase. Each generic reactor unit can represent all possible combinations of mixing and contacting patterns that can be associated with the ideal representations of the conventional reactor designs. Figure 5 shows the five possible conventional reactor designs and their counterparts in generic reactor unit instances for a system comprising two phases. The shadow reactor concept is generic and can be applied to systems with any number of phases. The modeling equations associated with the generic units are given in Mehta (34). The shadow reactor superstructure is generated for a specified number of generic reactor units by linking the compartments of a particular phase with a stream network realizing complete connectivity. By a selective combination of these streams, options related to feed distribution, product removal, bypasses, and recycles among the generic units in the network can be evaluated for each phase. The shadow reactor superstructure is illustrated in Figure 6 for a system with two phases. The flow directions in the compartments, along with the options available in the stream network, give rise to conventional as well as novel arrangements. These options can be explored through optimal search of the solution space defined by the superstructure. Nonisothermal systems are accounted for by the introduction of temperaturecontrol units into the generic reactor unit representation. These units consist of elements associated with the manipulation of temperature changes and constitute temperature profiles (profile-based approach) and heaters/coolers (unit-based approach). The assumption of thermal equilibrium between the contacting phases reduces the need for a single temperature per shadow reactor compartment. The profile-based system (PBS) finds the optimum profiles without considering the details of heat transfer mechanisms. Because the profiles are imposed rather than


developed, the approach adds no further computational difficulties as compared to the isothermal reactor network synthesis. By biasing the profile complexity of the final solutions, the profiles can be controlled effectively. The solutions obtained are easy to interpret, and thus the approach helps in understanding of dominant trade-offs in the problems. Results from the unit-based system (UBS) provides the target that can be obtained from a network of adiabatic reactors with consideration of direct and indirect intermediate heat transfer options. The synthesis of nonisothermal homogeneous and multiphase reactor networks is discussed in detail in Mehta and Kokossis (35). Following a similar reasoning that had led to the choice of an appropriate optimization technology for the single-phase reactor network synthesis as described in Section 3.1, stochastic optimization in the form of simulated annealing (SA) is adapted for the synthesis of the multiphase networks as well. The advantages are the identification of stochastic optima with confidence levels and the provision of a variety of solutions around the targets that can be reviewed as alternative designs. The synthesis framework consists of three stages. The targeting stage calculates performance targets and confidence levels using stochastic optimization in the form of SA. Simulated annealing is based on a randomized evolution of states developed through stepwise modification. The states are developed using the shadow reactor superstructure. In the screening and design stage, the results of the targeting stage are used to develop designs and layouts that feature performances that fall within a desired distance from the targets. These may consist of designs with the same or a different number of compartments, layouts with different networks of streams,

FIGURE 5 Conventional reactor designs and generic unit counterparts. (From Ref. 34.)


FIGURE 6 Shadow reactor superstructure. (From Ref. 34.)

recycles, and bypasses, or variations on the same structure. The stochastic search produces a multitude of solutions with similar performance, and its function is exploited as a major advantage. These designs can be improved further with the application of deterministic methods and more accurate models. The solutions developed from the screening stage are functional models based on the shadow reactor superstructure model. The analysis and validation stage requires the translation of these layouts into practical schemes. In general there are several ways to develop practical schemes from the functional models because one can opt for either physically distinct units or multicompartment reactors. Consider the reactor design for the production chlorination of butanoic acid as an example to illustrate the technology developments. A full study is given in Ref. 34, and only a brief summary of the results is presented here. The chlorination of butanoic acid (BA) involves two reactions in the liquid phase: 1.

BA Cl2

→ MBA HCl

2.

BA 2Cl2

→ DBA 2HCl

where MBA and DBA are abbreviations for monochlorobutanoic acid and dichlorobutanoic acid, respectively. The system involves two phases, a liquid phase where the reactions occur, and a gas phase consisting of chlorine feed and hydrogen chloride product. Solubilities are calculated using Henry’s law, and mass transfer rates are modeled according to film theory. The reaction kinetics and all model parameters and other problem data used in the study are given in Ref. 34. The objective of the study was to find those reactor networks with the


highest possible yield. Initially the study was performed assuming fixed phase holdups in the reactor units; i.e., hydrodynamic effects are neglected. Conventional reactor designs have been optimized to create a basis for comparison. The mechanically agitated vessel achieves the highest yield of 74.4%, followed by the bubble column reactor with a yield of 72.9%, and the coas well as the countercurrent reactors both achieving a yield of 69.5%. Reactor network optimization of a superstructure comprising three generic units produces a performance target of 99.6% yield of MBA. The selected designs with performances close to the target value range from simple designs employing only one reactor unit to designs featuring three units. The simplest solution consists of a single reactor with completely back-mixed gas and liquid phases and a bypass of fresh chlorine feed. Other, unconventional designs are also found that consist of two or three generic units, all approaching the target performance closely. Two of the simple designs are illustrated in Figure 7. For the unconventional designs, see Ref. 34. Mehta and Kokossis (36) also demonstrate how existing knowledge in the form of hydrodynamic correlations can be incorporated into the framework while maintaining the possibility of achieving design novelty. They also show how to deal with the ranges of application of the different available correlations so as to find meaningful results. For the example just described, they found that the performance target of 99.6% remains unchanged if the known hydrodynamic effects are considered for the reactor units via common correlations available in the literature. However, changes in the reactor designs from a mechanically agitated

FIGURE 7 Reactor design candidates. (From Ref. 34.)


vessel to a bubble column reactor are observed as a result of the higher interfacial areas available in the bubble column, according to the correlations. The methodology for the design of nonisothermal multiphase reactor networks is presented in detail in Refs. 35 and 36. 4.

COMPLEX DISTILLATION

Despite advances in other separation technologies, distillation is still the most widely used operation in chemical processes. Effective screening of separation systems constitutes a critical stage, for engineers need to review and understand trade-offs ahead of detailed modeling and simulation. In the separation, it is often desired to explore the use of complex rather than simple columns because the complex units reduce mixing losses, use available vapor and liquid more effectively, and improve the separation efficiency (37). Despite their recognized potential in energy savings, complex distillation applications are limited due to their difficult and demanding design-and-synthesis assignment. Synthesis challenges and operability issues that arise from a more complex dynamic behavior have discouraged wider acceptance in industry. A prohibitive number of configurations emerge from different allocations of side-rectifiers, side-strippers, prefractionators, and side-draw columns. Such options are difficult to enumerate and assess. The design alternatives increase rapidly, and the trade-offs are impossible to assess with an exhaustive (implicit or explicit) enumeration of the options. Previous efforts have focused on the development of shortcut methods that had a purpose of evaluating fixed configurations and initializing simulation models. Stupin and Lockhart (38) developed the equivalent arrangements of simple columns to represent complex configurations. Several other researchers (37, 39–47) extended knowledge from shortcut models as available for simple columns to evaluate the performance of complex configurations. Tedder and Rudd (48) performed a parametric analysis for complex designs and identified optimality regions as functions of the feed composition and the relative volatilities. Glinos and Malone (49) identified dimensionless parameters and proposed guidelines for the selection of complex distillation schemes. With less emphasis on fixed layouts, thermodynamic methods (50–53) produced procedures to assess energy efficiency in the integrated separation. Mathematical programming approaches promote process novelty with the use of superstructure development. Sargent and Gaminibandara (54) pioneered a progressive distillation train that Agrawal (55) later extended with additional connections of vapor–liquid streams to include satellite columns. Christiansen et al. (56) added more connections between component states to include structures with triangular walls in a single shell. Several researchers (57–62) have proposed different superstructures and developed mixed-integer nonlinear programming (MINLP) models for the synthesis of distillation systems. These superstructures


can be generalized as the combinations of two extreme representations, the state– task network (STN) of Sargent (63) and the state–equipment network (SEN) of Smith and Pantelides (59). These developments use superstructures of complicated interconnections with a large number of variables to account for the design options. More often than not, however, the optimization technology required to process these models is not addressed as a challenge. Researchers simply assume the technology is readily available and that the models can be solved and optimized robustly. Consequently, reported designs contain conventional rather than novel and unconventional designs. The development of new mathematical formulations for synthesis problems should not be detached from the optimization technology required to handle these models. Yeomans and Grossmann (64,65) have made some progress on a solution strategy using generalized disjunctive programming models, but their method is still unable to address major difficulties of solving industrial-scale distillation synthesis problems. Papalexandri and Pistikopoulos (66) propose a representation based on a superstructure of multipurpose heat and mass transfer modules. The authors refer to tasks as a more general form of a synthesis unit. They postulate an even more general (and difficult) problem and propose a detailed modeling framework that relies on the use of MINLP solvers. In contrast to the work by Papalexandri and Pistikopoulos, Shah and Kokossis (67,68) use the concept of a task (69) to reduce rather than enlarge the mathematical model and to simplify rather than complicate the optimization. In all cases the mathematical models are postulated as small mixed-integer linear programming (MILP) problems one can solve to global optimality. The remainder of this section gives a description of these developments. Shah and Kokossis (67,68) present a simultaneously systematic and rigorous approach for the automated development of optimal designs. The approach is computationally inexpensive, reliable, and very efficient to implement. It provides the engineer with a selected set of optimal designs on which further attention to dynamics and operability would yield the final design. The work embraces conceptual and engineering knowledge. In the past, conceptual knowledge has been restricted to heuristics and evolutionary approaches, both inappropriate for a systematic search. They typically employ a two-step approach where the best simple sequence is first identified. The sequence is next evolved into complex layouts. Although novelties are not excluded, they are left to coincidence and deprived from a systematic framework. The challenge is to develop systematic procedures that overcome the difficulties encountered with superstructure approaches. In contrast, the new synthesis approach is based on a supertask representation. The approach conceptualizes alternatives using simple tasks and hybrid tasks. The options include all complex column configurations related to side rectifiers, side strippers, prefractionators, Petlyuk columns, and side-draw columns. Furthermore, the representation accounts for sequences with sloppy separations and enables one to bracket optimum pressure limits. The approach assumes a number of components to be


separated into a predefined set of products. Given is the basic information about the components with respect to its physical and critical properties. The objective is to determine the appropriate and cost-effective separation schemes ahead of detailed simulations. In order to address complex systems effectively, Shah and Kokossis (67,68) replace superstructures with supertasks. The building blocks of the latter are not unit based, but task-based elements. Apart from simple columns, the synthesis representation is required to embed elementary complex distillation arrangements involving side columns, side-draw columns, and prefractionators. These complex arrangements are shown in Figure 8. The figure explains the two different types of side columns: the side-rectifier arrangement type (Figure 8a) and the sidestripper arrangement (Figure 8b); two types of side-draw columns: the vapor sidedraw column (Figure 8c) and the liquid side-draw column (Figure 8d); and the two prefractionator arrangements: with and without thermal coupling (Figures 8e and 8f). The arrangement shown in Figure 8f is known as a Petlyuk column.

FIGURE 8 Commonly used complex column arrangements. (From Ref. 82.)


Unlike simple columns, these complex configurations produce more than two products and feature more than a single light and heavy key component. A task-based representation of these schemes is accomplished with the ideas of hybrids and sloppy splits. Additional tasks are made up of different simple tasks. They are subsequently termed hybrid tasks and are defined as an ordered combination of simple distillation tasks, as illustrated in Figure 9. The sloppy separation improves the separation efficiency by distributing components between the lightest and the heaviest ones. Among the different options to distribute intermediate components (ICs), the maximum efficiency is accomplished by a general sequence of sloppy splits to separate the lightest and heaviest product at each stage. For an n-product system, a hybrid of order (n 1) accommodates a sequence of sloppy splits in the supertask representation. However, it processes each product (n 1) times (for an n-product system) at the expense of energy and column shells. These factors have an adverse effect on the capital and operating costs of the separation system. To capitalize on the advantages of the sloppy split, hybrids of a second order (over n-product systems) are included in arrangements, where sloppy splits are followed by sharp separation. These schemes combine the higher separation efficiency of the sloppy splits while the recurrent processing of products in the downstream sequence is minimized by the sharp separations. The sloppy split arrangements are introduced as an additional transformation on each hybrid. Instances of a task are replicas of the task operating under different conditions. The concept is used to optimize the operating conditions, such as the column pressure, and assumes the development of an operating range and a discretization scheme. Feasible ranges of pressure are identified by the physical properties (e.g., critical pressure) of the key components (upper limit) and the available utility levels (lower limit). The discretization scheme may be either uniform or based on the available utilities. The modeler can use a small or large number of discrete levels to capture associated trade-offs. The synthesis elements of the supertask representation are translated into modeling terms that combine available physical properties with design aspects. Shah and Kokossis (67,68) exploit modeling accomplishments from a variety of research groups (43,37,70) who have separately addressed the modeling and simulation challenges of complex distillation configurations. The calculations are used to set up the MILP optimization problem. Each task is assessed and evaluated independently and separately from the performance of the upstream or downstream units, because the composition and flow rate of the feed and the products streams are known a priori and only depend on the separation function of the particular task. Design and performance calculations can use any shortcut, semirigorous or rigorous method and any property package without influencing the constraints and solution strategy of optimization efforts.


FIGURE 9 Illustration of tasks and hybrid tasks. (From Ref. 82.)


The proposed formulation results in a mixed-integer linear programming problem that is solved using OSL/GAMS (71). The optimization simultaneously yields the optimal sequence, the optimal hybrids, and the optimal transformations. The formulation guarantees the global optimality in all cases because it involves only linear expressions of the continuous variables. An industrial application of the methodology is given in Section 6. 5. REACTION–SEPARATION AND REACTIVE–SEPARATION SYSTEM SYNTHESIS Particularly strong and complex interactions prevail among reaction and separation systems that are generally not at all or not fully exploited as a result of the application of the available synthesis methods for reactor networks and separation systems in isolation. The lack of generality in the synthesis methods is a tribute to the nonlinear process models required to capture the reaction and separation phenomena as well as to the vast number of feasible process design candidates. These complexities even make it difficult to synthesize the decomposed subsystems, which are typically reactor networks, separation systems, reactor–separator–recycle systems, and reactive separation systems. The development of reliable synthesis tools for these sub-systems is still an active research area. Potentially beneficial interactions between reaction and separation are manifold and can result from either separation within the reaction zones of the process (reactive separation systems), appropriate combinations of separate reaction and separation equipment (reactor–separator systems), or a combination of these. The realization of reactive separation options requires the introduction of mass separating options into the reaction equipment and results in additional phases or states that need to be considered in the process synthesis. Common examples are the introduction of solvents (reactive extraction, reactive absorption), stripping agents (reactive distillation), solids (reactive crystallization), and diffusion barriers (membrane reactors). In general reactor–separator systems, a number of separation tasks exist that enable the generation of intermediates or products of a performance-enhancing nature in the context of a general reacting system. The optimal separation tasks cannot generally be performed by the same unit operation, and a combination of unit operations is likely to be present in optimal-flow schemes. It is obvious that the simultaneous inclusion of all possible reactor–reactive separation–separator design options into an automated design framework quickly leads to combinatorial explosion. In combination with the nonlinear models used to describe the reaction, mass, and heat transfer phenomena that occur in the processing units, the resulting synthesis problem is beyond the scope of existing optimization technology, even for relatively small problems. This has led to the


development of synthesis tools, which are constrained to limited reaction and/or separation options; i.e., the general synthesis problem needs to be reduced in size significantly before any systematic synthesis methods can be applied (e.g., Refs. 14, 66, and 72–76). This problem-size reduction poses a major challenge: identification of the relevant modeling information. Currently, the design engineer needs to anticipate the important phenomena that dictate the performance of the system without the help of systematic screening methods. This demands a thorough understanding of the effects of component concentration and temperature manipulations on the system performance on the basis of process knowledge available at the earliest design stage. A typical example is the effect of component addition and removal to and from the reaction zones and the corresponding introduction of reactive separation options. Although an experienced designer can spot the relevant trade-offs for simple reacting systems and link them with the physical property differences to be exploited for separation, this task becomes nontrivial for systems with complex reactions where many components are present. However, the inclusion of all relevant design options is vital for the synthesis of optimal process configurations, because any opportunity excluded at this stage will inevitably lead to an inferior design. In view of the problem under consideration, Linke and Kokossis (77,78) propose a design framework that accounts for two synthesis stages: a screening stage that allows for identification of the dominant design trends in combined reaction and separation task systems, followed by a design stage that incorporates the insights gained in the screening stage and allows for the generation of optimal reactor–reactive separation–separation configurations. The screening stage aims at replacing the early problem decomposition based on intuition with the introduction of simple separation models that enable the investigation of major trade-offs in the context of reactor design. This is facilitated by structural optimization of relaxed reactor/separation task network superstructure models. Allowing any separation tasks to be present in the networks, the most important composition manipulations in the process network can be identified, regardless of separation feasibility. By introducing constraints on specific separation tasks, sensitivity studies allow one to investigate their impact on the performance targets and process layouts. Rather than final process layouts, the screening stage generates insights into optimal component separation, removal and recycle policies, along with reactor mixing and contacting patterns. The obtained process layouts feature reactor and separation task combinations, and only those important tasks are subsequently assessed for separation feasibility. The generated process design information is processed in a design stage by introducing the additional modeling information required for candidate reactive separation and mass exchange operations into a refined superstructure network model and by excluding infeasible separation tasks. In other words, more modeling rigor is added, to give a more detailed


account of the relevant reactive separation and separation options, while the combinatorial size of the problem is reduced by the irrelevant design options. The difference in targets between the screening and the design stages reflect potential savings of separation options termed infeasible upon analysis of the physical properties and give incentives for the investigation of novel separation techniques for the system under consideration, such as the development of novel solvents, membranes, or hybrid separation systems. Both synthesis stages allow for iterations to incorporate the insight gained during the synthesis process. Reasons for iterations are the identification of additional problem constraints, the removal of process units without positive impact on the system performance, and the inclusion of refined process models. Such refinements might result from the limited validity of the available reaction models in optimal design regions. In order to gain confidence about the designs, a wide range of operating conditions and structural design alternatives needs to be searched in the process synthesis stage. In many cases, the optimized operating conditions do not coincide with the regions of model validity, and the model trade-offs might not reflect the real system trade-offs. This creates the need for new process models for the operation regions identified as of interest, which may introduce additional model nonlinearities or an additional set of reactions and components. The proposed synthesis representation, along with the robust optimization framework, is not limited by the complexity of such models. The refinement of reaction models demands additional experiments to be carried out. The synthesis methodology is efficient in suggesting the concentration and temperature regions of interest for process design, which helps to minimize the experimental efforts. In both synthesis stages, superstructures of generic units are formulated, and the performance targets as well as a set of design candidates are obtained subsequently via robust stochastic optimization techniques. 5.1.

Unit Representation

The basic elements of the synthesis representation are the generic reactor/mass exchanger (RMX) units and the separation task units (STUs). The underlying phenomena exploited in chemical process design are reaction, heat exchange, and mass exchange. They are generally exploited simultaneously (nonisothermal multiphase reactors and reactive separators), in combinations (nonisothermal homogeneous reactors and mass exchangers), or in isolation (isothermal reactors and mass exchangers, heat exchangers), depending on the particular system under investigation and the location of the operation within the flowsheet. This work employs a generic reactor/mass exchanger unit for a flexible and compact synthesis representation of all possibilities of phenomena exploitation. The RMX unit follows the shadow compartment concept developed for nonisothermal multiphase reactor network synthesis (Section 3.2). The unit consists of compartments


in each phase or state present in the system under investigation, and the streams processed in the different compartments of the generic unit can, by definition, exchange mass across a physical boundary, which can be either a phase boundary or a diffusion barrier. Each compartment features a superset of mutually exclusive mixing patterns through which a compact representation of all possible contacting and mixing pattern combinations between streams of different phases can be realized within a single generic unit. Along with decisions on the existence of mass transfer links between compartments of the same generic unit and decisions on the consideration of reaction phenomena in the different compartments, a single RMX unit enables the representation of a reactor, a mass exchanger, a reactive mass exchanger, or a combination of these. Each compartment receives inlet streams within its state and corresponds with the compartments in the other states via diffusional mass exchange links across the state boundaries. The effluents from the compartments either leave the unit or are recycled within the RMX unit. Recycles can be present within a given state or across the state boundaries if technically possible, e.g., by reboiling, condensing, throttling, or compression. Throughout this work, all states that can receive streams from a reference state are termed adaptable states to the reference state. Mutually exclusive mixing options considered for the compartments include well-mixed and segregated flow. In order to avoid a mathematical model with both differential and algebraic equations, segregated flow behavior is approximated with a serial arrangement of well-mixed units of equal volume with the resulting cell model consisting of only algebraic equations. In each compartment, all inlet streams are connected to all mixers prior to the subunits. Each subunit effluent stream is split and connected to the subsequent subunit as well as to the final product mixer of the compartment. The recycle streams from the final product splitter distribute among all well-mixed units employed in the compartments of the adaptable states. Temperature effects are accounted for in the compartment models in accordance with the profile-based and the unit-based synthesis approaches (Section 3.2). In contrast to the rigorous representation of reaction and mass transfer phenomena by RMX units, the separation task units (STUs) represent venues for composition manipulations of streams without the need for detailed physical models. In accordance with the purpose of any separation system, the separation task units generate a number of outlet streams of different compositions by distributing the components present in the inlet stream among the outlet streams. In the screening stage, the aim is to identify the separation tasks, which have a positive impact on the performance of the reaction–separation system. Only sharp separations are considered between component lumps; i.e., each lump can be present in only a single outlet stream of the separation task unit. Sharp separations do not impose a limitation in terms of composition attainability, because any nonsharp separation can be obtained by sharp separation followed by stream mixing. Moreover, potential benefits that might arise from nonsharp separations as compared


to sharp separation followed by mixing cannot be quantified, because a particular separation process is not specified at this synthesis stage. In the design stage, the separation task unit is adoptable for the representation of separation tasks that can be performed using particular separation processes. In this case, the possible distribution policies of component lumps to the outlet streams are constrained by the separation orders of the separation process. Nonsharp separations arising from operational constraints on the separation tasks can easily be accomplished. The separation task unit performs a set of feasible separation tasks according to the separation order to define a set of outlet streams. Depending on the order in which the tasks are performed, a variety of processing alternatives exist for a single unit. 5.2.

Process Representation

The generic synthesis units presented earlier allow for the representation of all sections of general processes involving reaction and separation, i.e., reaction sections, reactive separation sections, mass exchange sections, and sections performing separation tasks. The aim of the superstructure network generations is to provide for a venue that enables the simultaneous exploration of all functionalities of the different synthesis units along with all possible interactions among them. The superstructure formulations evolve in the different synthesis stages according to the insights into the process obtained at the previous stages. Novelty is accounted for in the superstructures because the representations are not constrained to conventional process configurations but instead include all possible novel combinations of the synthesis units. The superstructures feature a number of generic RMX and separation task units as well as raw material sources and product sinks, the interconnections among which are realized by two types of stream networks: Intraphase streams establish connections between the synthesis units, products, and raw materials of the same state, whereas interphase streams establish those connections across the state boundaries; i.e., the source and sink of such a stream belong to adaptable states. The compositions of the streams remain unchanged on state transition. The transfer of material across the state boundary potentially requires the addition or removal of energy to or from the interphase streams. In this case, an energy transfer unit such as a total reboiler or a total condenser, a throttle or a compressor is associated with those mixers receiving a stream from a different state. Superstructures formulated in the design stage facilitate RMX units for representation of the relevant reaction, mass exchange, and reactive separation options obtained from the screening stage and separation task units associated with the different unit operations that allow one to perform the desired separation tasks. This incurs superstructures with additional states resulting from the inclusion of separating agents and diffusion barriers. Figure 10 illustrates two process


configurations that can be obtained by eliminating interphase and intraphase streams, as well as generic units from a superstructure network featuring two RMX units and two separation task units in two states. A flow scheme featuring a catalytic reactive distillation arrangement is shown in Figure 10a. The design shown in Figure 10b does not feature any interphase connections, and reaction

FIGURE 10 Design instances from the superstructure.


occurs in both RMX units. A separation task unit facilitates the intermediate separation of components in between the RMX units; a plug-flow and a well-mixed RMX compartment are arranged in series, followed by a separation task unit representing a separation sequence to separate the raw material from the product and the by-products. 5.3.

Optimization

In order to establish a basis for optimization, the reaction and separation superstructure is formulated as a mathematical model that involves the component material balances for the RMX units, the mixers prior to the separation task units, and the product mixers, as well as the energy balances for the RMX units (for the unit-based nonisothermal representation). The formulation incorporates models for the reaction kinetics, physical property and mass transfer models, shortcut models and regression expressions for equipment sizing and costing, and the objective function. General relationship-modeling terms will introduce nonlinearities into the superstructure network model. As for the reactor network synthesis problems, the reaction and separation superstructures are optimized using stochastic search techniques. Linke and Kokossis (77,78) have studied the performance of SA and tabu search (TS) for this type of problem. They found robust performances for both algorithms and TS to be the more efficient search metaheuristic. A detailed description of the implementation of the stochastic search algorithms is given in Ref. 79. For illustration purposes, consider the Williams–Otto flowsheeting problem (80). In the conventional design, raw materials A and B are fed to a reactor, where the following reactions occur: (1)

AB

→

C

(2)

BC

→

PE

(3)

PC

→

G

The reactor effluents are cooled and fed to a decanter for removal of heavy waste G, which requires further treatment. From the remaining mixture, the desired product P is removed via distillation. The unreacted raw materials as well as unwanted byproduct E and a fraction of P are partly recycled back to the reactor. As components E and P form an azeotrope, an amount of the desired product equivalent to at least 10% weight fraction of E is lost through the purge, which is used on site as a fuel. The volatilities i of components i in the system have the following descending order: P > E > C > B > A > G. The reaction rates of components A, B, C, P, E, and G respectively are functions of the weight fractions X and given by the vector


R(X) [k1XAXB; k1XAXB k2XBXC; 2k1XAXB 2k2XBXC k3XPXC; k2XBXC 0.5k3XPXC; 2k2XBXC; 1.5k3XPXC] The complete data are given in Ref. 79. The objective of the synthesis exercise is to find the designs that maximize the annual profit of the process for a minimum production rate of 400 kg/h of component P. 5.3.1.

Screening Stage

The stochastic search of superstructures featuring three RMX and three unconstrained separation task units yields an absolute profit target of around $618 k/yr. A variety of design alternatives exist that can achieve the targets, featuring one, two, or three reactors and component distributors. The process design requirements as identified in the screening stage can be stated as follows: Separation and removal of by-products E and G Separation and removal of desired product P Separation of components A, B, and C from the reacting mixture and distribution among the reactor units Excess of component B in the reaction zones minimizes by-product formation (low concentration of product P) Reaction zones exhibit plug-flow behavior 5.3.2.

Design Stage

Based on the insights gained in the first design instance, appropriate separation venues can be identified and included in the search. Distillation enables separation of mixtures according to the order of volatilities and hence allows separation in support of the raw material and intermediate recovery. To avoid fouling, G needs to be decanted prior to the operation, which can be achieved at a low cost in a decanter. However, component P forms an azeotropic mixture with component E, resulting in a loss of desired product to the low-value fuel. A solvent is available that allows selective extraction of desired product P from the mixture. The equilibrium relationship and maximum solvent loading are given in Ref. 79. The design stage therefore considers a superstructure of RMX units and STUs to capture the reaction–extraction–distillation–decanting system. Stochastic optimization yields a target performance of around $433 k/yr for the system. Designs with performances close to the target can be grouped into two main categories according to their use of the solvent: reactive extraction designs and reactor–hybrid separation designs, each achieving performances close to the target. Two sample designs are illustrated in Figure 11. Designs utilizing the distillation–extraction hybrid achieve higher selectivities in the conversion of the raw materials to product P than do the reactive extraction designs,


in which valuable components A, B, and C are lost to the fuel stream. Both designs achieve a virtually complete recovery of product P. The extraction of the desired product from the concentrated distillation overheads allows for higher solvent loadings and consequently a significantly reduced solvent flow as compared to the reactive extraction case. However, the benefits in terms of higher selectivities and lower solvent flows are offset by the expenditure required for the distillation operation.

FIGURE 11 Process design candidates from the design stage.


The methodology is illustrated with a further application to the design of activated sludge process systems in Section 6 and has seen further applications in multiphase reaction–separation, reactive distillation, and membrane network synthesis (79). 6.

INDUSTRIAL STUDIES AND REAL-LIFE APPLICATIONS

The methods described in Sections 3 through 5 have been tested on a number of case studies. This section presents three applications to problems of complexity typically encountered in industrial practice that have been performed using the synthesis schemes, partly in collaboration with industrial partners. 6.1. Reactor Design: Ammoxidation of Propylene to Acrylonitrile Mehta (34) has carried out a reactor network optimization study to find improved designs for the production of acrylonitrile in a collaboration between UMIST and one of its industrial partners. Most industrial installations employ fluidized-bed reactors (BP/Sohio process) with a well-mixed reaction zone. Previous process improvements have mainly resulted from better catalysts, which have produced an increase in yield from 58% to around 80%. The reaction model employed in the optimization study is taken from Ref. 81 and considers seven reactions and eight components. Air, pure oxygen, and propylene are available as raw material streams. The optimization study assumes negligible pressure drop along the reaction sections, isothermal and isobaric operation, and negligible mass gas–solid transfer effects. Acrylonitrile yield has been employed as the objective function. Initial targeting has shown that the network yield increases with the catalyst volume employed and asymptotically approaches a yield of around 90% toward infinitely large catalyst volumes. For very large volumes, the well-mixed reactor achieves the performance targets. Because only smaller catalyst volumes are relevant in industrial practice, a targeting study is carried out to develop the performance targets as functions of the catalyst volume utilized. For comparison, the targets for the conventional well-mixed reactor are developed alongside the targets that can be realized with reactor designs featuring novel unconventional mixing and feeddistribution policies. The targeting curve is shown in Figure 12. The targeting curve shows that the performance of the CSTR reactor significantly degrades toward smaller catalyst volumes, i.e., toward the industrially more relevant designs. Only a slight degradation of reactor yields with catalyst volume is observed for the novel designs. This suggests that the acrylonitrile yield can be kept high even with significantly reduced catalyst volumes, provided an appropriate reactor network is selected. When compared to the conventional reactor, the novel designs showed improvements in yield of up to 17% for practical values


FIGURE 12 Targeting curve for ammoxidation of propylene to acrylonitrile.

of catalyst volume. Mehta (34) observed a change in reactor design pattern toward smaller catalyst volumes, gradually moving single CSTRs to combinations of PFRs and CSTRs with distributed feed streams. The study was carried out over a short period of time and resulted in an impressive enhancement of insights into how an optimal reactor should be designed for maximum performance at low catalyst volumes. A commercial reactor design has subsequently been developed on the basis of the insights gained in this study. 6.2.

Complex Distillation

This case study (82,67,68) included a feed containing nine components and that had to be separated into four products: A (C4 fraction separated as lights), B (isopentane and some n-pentane), C (C6 fraction), and D (heavier components, C7+). Minimum energy cost is used as the synthesis objective, and the optimization results are summarized in Table 1. Designs I and II represent the favorable schemes. Design I separates the lightest product and uses a prefractionator for the downstream separation in hybrid B/C/D. Design II employs a prefractionator in hybrid AB/C/D to separate plentiful products, C and D, early in the sequence. Design III, the best simple sequence, separates the plentiful product, D, first and favors difficult separation, B/C, in the end. When simulated rigorously, the ranking of the design candidates remains identical to that from the conceptual design level. The initial energy costs match very well with the costs from rigorous simulation. An interesting point is that in this example the best simple sequence is close to an indirect sequence (Design III) but the best complex scheme corresponds to


TABLE 1 Promising Designs for the Industrial Complex Distillation Case Study Design I

II

III

Energy cost

Design

6.57

6.60

7.60

Column

Pressure (bar)

Trays

1

4.9

30

2

1.8

4

3

1.8

82

1

3.5

4

2

3.5

90

3

4.9

29

1

1.6

23

2

4.9

29

3

1.8

54

a direct sequence (Design I). The lowest pressures are selected for all the tasks in the designs. This results in the most favorable separation factors. The complex column configurations require high pressures, and the benefits of thermal coupling are counterbalanced by the requirement of higher reflux. Pressure effects were found to dominate the results and to minimize the thermal coupling. 6.3. Combined Reaction and Separation Systems: The Activated Sludge Process The reaction and separation synthesis approach of Section 5 has been adopted to the problem of activated sludge process design (83). The conventional designs as well as all novel schemes for combined oxidation/denitrification of wastewater are explored. The process is optimized using a novel methodology for optimal reaction/separation network synthesis, supplied with a comprehensive kinetic model (84). The activated sludge process is synthesized using the systematic


reaction/separation network synthesis framework of Section 4 to determine the optimal biochemical reactor network design along with the sludge separation and recycle policies. The representation provides for all possible mixing, contacting, and reaction/separation features, and the superstructures are optimized through application of the simulated annealing algorithm. Oxygen mass transfer is modeled using film theory. A weighted objective function is formulated so as to minimize both the effluent COD and total nitrogen content. An initial optimization study without considering any bounds on the reactor volumes yields a reference target (lowest value of objective function) of 100%, corresponding to a 96.8% reduction in COD and 84.8% in N. However, the reaction volumes observed in the designs that achieve the targets are unpractically large. For this reason, volume bounds are introduced in a subsequent study to achieve the mean hydraulic residence time of the conventional processes. Surprisingly, the target did not deteriorate significantly, the new target for the objective value being 101.4%. COD was reduced by 97.4% and N by 84.9%, a performance much better than those attained by any conventional process, especially as far as denitrification is concerned. An inspection of the structures revealed a design pattern: Many structures did not include an anoxic reactor and yet yielded very good denitrification. Moreover, the system appeared to seek ways to hinder or control the dissolution of oxygen, in contrast to common practice, which aims at dissolving the maximum amount possible into the oxic reactors. To find out how such good denitrification could be accomplished without the inclusion of an anoxic reactor, the detailed oxygen profiles within the aerated reactors (mostly PFRs) were examined. The concentration of oxygen is controlled within extremely low levels (0.1–0.3 ppm), an order of magnitude less than those currently used in industrial practice (2–9 ppm). This leads to the conclusion that both organic matter stabilization and denitrification processes occur simultaneously in these designs, due to the very low oxygen concentrations—a policy that seems to achieve maximum efficiency when low volume is required. A schematic design of the system is illustrated in Figure 13. Details on the synthesis study can be found in Ref. 83.

FIGURE 13 Conceptual activated sludge process design.


7.

CONCLUSIONS

The drive to improve processes and to create new process routes continuously force process engineers and management to develop new technology. Process synthesis or integration has proven its value to improve processes with reduced costs, less pollution, and lower energy consumption. Traditionally, most industrial applications of process integration techniques have been found in developing energy-efficient systems. After a brief discussion on common conceptual process design practices, this chapter illustrated recently developed process synthesis technology for process design problems involving reaction and separation. A major advantage of the methods is their ability to determine robust performance targets and to identify a variety of design options with similar close-to-target performances. The design tools offer decision support to the design engineer. This will allow the inspection of similarities and differences among high-performance candidates in order to select the most practical novel designs. However, for successful technology development and transfer, close collaboration between academia and industry is essential. REFERENCES 1. Westerberg AW. Synthesis in engineering design. Comp Chem Eng 1989; 13:365. 2. Smith R. Chemical Process Design. New York: McGraw-Hill, 1995. 3. Linnhoff B, Flower JR. Synthesis of heat exchanger networks. AIChE J 1978; 24:633. 4. Douglas JM. Conceptual Design of Chemical Processess. New York: McGraw-Hill, 1988. 5. Biegler LT, Grossmann IE, Westerberg AW. Systematic Methods of Chemical Process Design; Upper Saddle River, NJ: Prentice Hall, 1997. 6. Seider WD, Seader JD, Lewin DR. Process Design Principles, Synthesis, Analysis and Evaluation; New York: Wiley, 1999. 7. van den Berg H. Methods for process intensification projects. In: Proceedings of the 4th International Process Intensification Conference. Bruges, Belgium: BHR, 2001. 8. Green A. In Proceedings of the 4th International Conference on Process Intensification. Bruges, Belgium, 2001. 9. Levenspiel O. Chemical Reactor Engineering: An Introduction to the Design of Chemical Reactors. New York: Wiley, 1962. 10. Achenie LEK, Biegler LT. Algorithmic synthesis of chemical reactor networks using mathematical programming. Ind Eng Chem Fundam 1986; 25:621. 11. Achenie LEK, Biegler LT. Developing targets for the performance index of a reactor network. Ind Eng Chem Res 1988; 27:1811. 12. Achenie LEK, Biegler LT. A superstructure-based approach to chemical reactor network synthesis. Comput Chem Eng 1990; 14:23. 13. Kokossis AC, Floudas CA. Optimization of complex reactor networks: I. Isothermal operation. Chem Eng Sci 1990; 45:595.


14. 15. 16. 17.

18.

19. 20. 21. 22. 23. 24.

25. 26. 27.

28. 29. 30. 31. 32. 33. 34.

Kokossis AC, Floudas CA. Synthesis of isothermal reactor-separator-recycle systems. Chem Eng Sci 1991; 46:1361. Kokossis AC, Floudas CA. Optimization of complex reactor networks: II. Nonisothermal operation. Chem Eng Sci 1994; 49:1037. Schweiger CA, Floudas CA. Optimization framework for the synthesis of chemical reactor networks. Ind Eng Chem Res 1999; 38:744. Glasser D, Hildebrandt D, Crowe CM. A geometric approach to steady-flow reactors: the attainable region and optimization in the concentration space. Ind Eng Chem Res 1987; 26:1803. Horn F. Attainable and non-attainable regions in chemical reaction technique. In: Proceedings of the 3rd European Symposium on Chemical Reaction Engineering. London: Pergamon Press, 1964. Glasser D, Hildebrandt D. Reactor and process synthesis. Comput Chem Eng 1997; 21:775. Feinberg M, Hildebrandt D. Optimal reactor design from a geometric viewpoint. I. Universal properties of the attainable region. Chem Eng Sci 1997; 52:1637. Glasser D, Hildebrandt D, Godorr S. The attainable region for segregated maximum mixed and other reactor models. Ind Eng Chem Res 1994; 33:1136. Glasser B, Hildebrandt D, Glasser D. Optimal mixing for exothermic reversible reactions. Ind Eng Chem Res 1992; 31:1541. Hildebrandt D, Glasser D. The attainable region and optimal reactor structures. Chem Eng Sci 1990; 45:2161. Hildebrandt D, Glasser D, Crowe CM. Geometry of the attainable region generated by reaction and mixing: with and without constraints. Ind Eng Chem Res 1990; 29:49–58. Balakrishna S, Biegler LT. A constructive targeting approach to the synthesis of isothermal reactor networks. Ind Eng Chem Res 1992; 31:300. Balakrishna S, Biegler LT. Targeting strategies for the synthesis and energy integration of nonisothermal reactor networks. Ind Eng Chem Res 1992; 31:2152. Lakshmanan A, Biegler LT. Reactor network targeting for waste minimization. In: El-Halwagi M, Petrides D, eds. Pollution Prevention via Process and Product Modifications AIChE Symposium Series 90. New York: AIChE 128, 1994. Lakshmanan L, Biegler LT. Synthesis of optimal chemical reactor networks. Ind Eng Chem Res 1996; 35(5):1344. Lakshmanan L, Biegler LT. A case study for reactor network synthesis: the vinyl chloride process. Comput Chem Eng 1997; 21:785. Hildebrandt D, Biegler LT. Synthesis of chemical reactor networks. Snowmass, CO: FOCAPD, 1994. Marcoulaki EC, Kokossis AC. Screening and scoping complex reaction networks using stochastic optimization. AIChE J 1999; 45(9):1977–1991. Mehta VL, Kokossis AC. Development of novel multiphase reactors using a systematic design framework. Comp Chem Eng 1997; 21:325. Marcoulaki EC, Kokossis AC. Stochastic optimization of complex reaction systems. Comp Chem Eng 1996; 20:231. Mehta VL. PhD dissertation, UMIST, U.K., 1998.


35. 36.

37. 38. 39. 40. 41. 42. 43. 44. 45. 46.

47.

48. 49. 50. 51. 52. 53. 54. 55.

Mehta VL, Kokossis AC. Nonisothermal synthesis of homogeneous and multiphase reactor networks. AIChE J 2000; 46:2256. Mehta, VL, Kokossis AC. New generation tools for multiphase reaction systems: a validated systematic methodology for novelty and design automation. Comp Chem Eng 1998; 22:119. Triantafyllou C, Smith R. The design and optimization of fully thermally coupled distillation columns. Trans IChemE 1992; 70:118. Stupin WJ, Lockhart FJ. Thermally coupled distillation—a case history. Chem Eng Prog 1972; 68:71–72. Fidkowski Z, Krolikowski L. Thermally coupled system of distillation columns: optimization procedure. AIChE J 1986; 32:537. Fidkowski Z, Krolikowski L. Minimum energy requirements of thermally coupled distillation systems. AIChE J 1987; 33:643. Fidkowski Z, Krolikowski L. Energy requirements of nonconventional distillation systems. AIChE J 1990; 36:1275. Glinos K, Malone MF. Minimum reflux, product distribution, and lumping rules for multicomponent distillation. Ind Eng Chem Process Des Dev 1984; 23:764. Glinos K, Malone MF. Design of side-stream distillation columns. Ind Eng Chem Process Des Dev 1985; 24:822. Nikolaides IP, Malone MF. Approximate design of multiple-feed/side-stream distillation systems. Ind Eng Chem Res 1987; 26:1839. Nikolaides IP, Malone MF. Approximate design and optimization of a thermally coupled distillation with prefractionation. Ind Eng Chem Res 1988; 27:811. Carlberg NA, Westerberg AW. Temperature-heat diagrams for complex columns. 2. Underwood’s method for side strippers and enrichers. Ind Eng Chem Res 1989; 28:1379. Carlberg NA, Westerberg AW. Temperature-heat diagrams for complex columns. 3. Underwood’s method for the Petlyuk configuration. Ind Eng Chem Res 1989; 28:1386. Tedder DW, Rudd DF. Parametric studies in industrial distillation. AIChE J 1978; 24:303. Glinos K, Malone MF. Optimality regions for complex column alternatives in distillation systems. Chem Eng Res Des 1988; 66:229. Petlyuk FB, Platonov VM, Slavinskii DM. Thermodynamically optimal method for separating multicomponent mixtures. Int Chem Eng 1965; 5:555. Kaibel G. Distillation columns with vertical partitions. Chem Eng Technol 1987; 10:92. Kaibel G. Distillation column arrangements with low energy consumption. IChemE Symp Ser 1988; 109:43. Kaibel G, Blass E, Kohler J. Thermodynamics guideline for the development of distillation column arrangements. Gas Sep Purif B 1989; 109. Sargent RWH, Gaminibandara K. Optimum design of plate distillation columns. In: Dixon LCW, ed. Optimization in Action. London: Academic Press, 1976:267–314. Agrawal R. Synthesis of distillation column configurations for a multicomponent separation. Ind Eng Chem Res 1996; 35:1059.


56. 57. 58. 59. 60. 61. 62. 63. 64. 65.

66. 67. 68. 69. 70. 71. 72. 73.

74. 75. 76.

Christiansen AC, Skogestad S, Lien K. Complex distillation arrangements: extending their ideas. Comp Chem Eng 1997; 21:237. Floudas CA, Paules IV GE. A MINLP formulation for the synthesis of heat-integrated distillation sequences. Comp Chem Eng 1988; 12:531. Agarwal A, Floudas CA. Synthesis of heat integrated nonsharp distillation sequences. Comp Chem Eng 1992; 16:89. Smith EMB, Pantelides CC. Design of reaction/separation networks using detailed models. Comp Chem Eng 1995; 19:83. Bauer MJ, Stichlmair J. Superstructurers for the mixed integer optimization of nonideal and azeotropic distillation processes. Comp Chem Eng 1997; 20:25. Caballero JA, Grossmann IE. Aggregated models for integrated distillation systems. Ind Eng Chem Res 1999; 38:2330. Dunnebier G, Pantelides CC. Optimal design of thermally coupled distillation columns. Ind Eng Chem Res 1999; 38:162–176. Sargent RWH. Functional approach to process synthesis and its application to distillation systems. Comp Chem Eng 1998; 22:31. Yeomans H, Grossmann IE. Disjunctive programming models for the optimal design of distillation columns and separation sequences. Ind Eng Chem Res 2000; 39:1637. Yeomans H, Grossmann IE. Optimal design of complex distillation columns using rigorous tray-by-tray disjunctive programming models. Ind Eng Chem Res 2000; 39:4326. Papalexandri KP, Pistikopoulos EN. Generalized modular representation framework for process synthesis. AIChE J 1996; 42(4):1010. Shah PB, Kokossis AC. Knowledge-based models for the analysis of complex separation processes. Comp Chem Eng 2001; 25:867. Shah PB, Kokossis AC. New synthesis framework for the optimization of complex distillation systems. AIChE J 2002; 48:527. Hendry JE, Hughes RR. Generating separation process flowsheets. Chem Eng Prog 1972; 68:71. Koehler JP, Aguirre P, Blass E. Evolutionary thermodynamic synthesis of zeotropic distillation sequences. Gas Sep Purif 1992; 6:153. Brooke KE, Kendrick D, Meeraus A. GAMS: A User’s Guide. San Francisco, CA: Scientific Press, 1988. Barbosa D, Doherty MF. The simple distillation of homogeneous reactive mixtures. Chem Eng Sci 1988; 43(3):541. Ismail RS, Pistikopoulos EN, Papalexandri KP. Synthesis of combined reactive and reactor/separation systems utilizing a mass/heat exchange transfer module. Chem Eng Sci 1999; 54:2721–2729. Ng KM, Berry DA. Synthesis of reactive crystallization processes. AIChE J 1997; 43(7):1737. Ng KM, Samant KD. Synthesis of extractive reaction processes. AIChE J 1998; 44(6):1363. Stein E, Kienle A, Esparta ARJ, Mohl KD, Gilles ED. Optimization of a reactor network for ethylene glycol synthesis—an algorithmic approach. Comp Chem Eng 1999; 23:903–906.


77. 78.

79. 80. 81. 82. 83. 84.

Linke P, Kokossis AC. Attainable designs for reaction and separation processes from a superstructure-based approach. Submitted to AIChE J, 2002. Linke P, Kokossis AC. Systematic decision-making technology for optimal multiphase reaction and reaction/reactive separation system design. Comput-Aided Chem Eng 2002; 10:247. Linke P. PhD dissertation, UMIST, U.K., 2001. Ray H, Szekely J. Process Optimization; New York: Wiley, 1973. Chen BH, Dai QL, Lu DW. Development and modeling of a loop fluidized-bed reactor with baffle for propylene ammoxidation. Chem Eng Sci 1996; 51: 2983. Shah PB. PhD dissertation, UMIST, U.K., 1999. Rigopoulos S, Linke P. Systematic development of optimal activated sludge process designs. Comp Chem Eng 2002; 26:585. Henze M, Grady CPL Jr, Gujer W, Marais GVR, Matsuo T. A general model for singlesludge wastewater treatment systems. Water Res 1987; 21(5):505.


12 Process Intensification in Industrial Practice: Methodology and Application Remko A. Bakker DSM Fine Chemicals Austria, Linz, Austria

1.

INTRODUCTION

In this chapter a method is explained for how process intensification can be introduced in a commercial company. When one would like to introduce process intensification, certain steps can be followed that increase the chances of success. One very important aspect is the awareness of the drivers for introducing process intensification for certain processes and companies: the why of process intensification. This is explained in Section 3 in some detail, for without the knowledge of these drivers, one risks introducing process intensification ineffectively and inefficiently. After that, a short overview is given of the technologies available for process intensification, serving as a basis for the real process intensification study. The method of introducing process intensification is then explained in Section 5, which is the main section: What steps should be incorporated in a process intensification study, what a good group of participants is, and in what phase of a chemical process the study should be introduced. The chapter concludes with two concrete examples of process intensification, one in a bulk chemical process and one in a fine chemical process. Process intensification can bring many benefits regarding business, legislative, and environmental aspects and is therefore very worthwhile to be implemented.


2. SHORT OVERVIEW OF DEVELOPMENTS IN PROCESS INTENSIFICATION As discussed in Chapter 1, process intensification as a way of looking at chemical processes originated in the early 1970s. It was adopted by the ICI New Science Group in the United Kingdom, where mainly rotational fields were used for accelerating chemical process steps. After the first conferences on process intensification, organized by the British Hydromechanics Research (BHR) Group (1), the field became a separately identifiable field in chemical engineering and spread around the world. The general ideas behind process intensification—improving a chemical process via stepby-step changes—could already be found in various activities of development groups in chemical engineering (in industry and at universities), though at that time not all of these activities were called process intensification. By grouping the activities in this field under an umbrella name and by organizing meetings and conferences dedicated to the topic, many new ideas, developments, and applications were discovered, and they are still being discovered. This is one of the reasons the field of process intensification developed rapidly at the end of the 20th century. 3. WHY WORK ON THE INTENSIFICATION OF CHEMICAL PROCESSES? As with all developing technologies, there is always more risk in introducing them in commercial plants than in using the well-known, traditional technologies. Some conservatism in commercial industry is always present. A very good illustration of that can be seen in Figure 1 of Chapter 1, where a 16th century chemical process is compared with contemporary chemical processing. Why then would one work on these kinds of developments? By introducing step-by-step improvements in a process, process intensification offers a strong possibility to fulfill the current business, legislative, and environmental requirements. These aspects are becoming more and more important in contemporary chemical industry. 1. Business aspects are changing due to the expansion of the globalization of the world economy: There are more potential competitors in this enlarged market. The best low-cost position is necessary to obtain and maintain market share. This is true for the fine chemical and pharmaceutical industries, because transportation costs are only a small amount of total production costs and because the total global market may in theory be supplied by producers from any place in the world. On the other hand, for the bulk chemical industry as well, small cost


advantages per kilogram of product can make the difference between profit and loss. 2. Legislative and environmental aspects are usually strongly linked. They are getting a lot of attention in discussions on the license to operate (LTO) and sustainable technology (see Chapter 14). These two aspects are also reflected in the visions developed on this issue in the United States in the so-called Vision 2020 program (see Appendix 1). One can see that all aspects of introducing step-by-step changes are mentioned. And in the European Union many current activities are based on the topic of process intensification, reflected, for example, in the European-Supported Research Projects (see Appendix 2). On the industrial side, a lot of effort is being given to the topic of process intensification: the European Chemical Industry Council (CEFIC) has process intensification as one of its technology focus points in the so-called SUSTECH program (see Appendix 3). The intensification of chemical processes can help reduce the risks of chemical processing and lower the energy demand and waste production. These positive possibilities, which are evidently also linked to the business aspects, have contributed to the growing use of process intensification. This is also expressed in the paper of Elverding (2) at the Defacto conference on “Profit, Planet, and People” in combination with process intensification, reflecting the benefits for business and environment. Many different aspects relating to business, legislation, and environment can be defined. These are given in Table 1. In this important table one can find many of the benefits of process intensification summed up. All of these aspects in the end have to do with business, legislation, or the environment and show how process intensification can have a strong positive influence on these three main drivers and answers the question of why one would one choose to intensify a process. The overview in Table 1 holds, in principle, for any type of chemical industry. Additionally, different types of chemical industries may have different focuses, depending on their starting position and on their business requirements. Knowledge of the perspectives helps in determining the main cost drivers from the list given in Table 1 and helps therefore in finding the starting points for process intensification studies. Table 2 presents an overview of some of the main types of chemical industries and the specific demands they posed on process intensification when applied in these industries. A distinction can thus be made between: Large-scale chemical industries, producing—mostly in a continuous way— hundreds of kilotons of product Fine chemical industries, which can be split into: Multiclient chemicals, producing up to hundreds of tons of chemicals for various customers, mostly in a batch way


TABLE 1

Positive Results of Process Intensification on Various Aspects

Aspect

Business (costs)

Legislative requirements (environment and safety)

Environment

Energy

X

X

X

Space

X

X

X

Number of process steps Cost

X

X

X

X

X

X

X

X

X

X

Emissions

Waste

X

Less holdup


Examples

Less heat loss via integration of units (e.g., reactive distillation) Efficient heat transfer possible (high heat transfer rates) Reduction in floor space needed Less piping Reduced maintenance Faster grade/product changes Ease of dismantling and/or shipping Fewer units Less danger Fewer possibilities for yield losses Many aspects together may result in a lower cost Theoretical goal: no limitation, other than kinetics, gives optimal yield and least waste Fewer possibilities for yield losses Fewer process steps may result in lowered emissions Fewer possibilities for yield losses means less waste Theoretical goal: no limitation, other than kinetics, gives optimal yield and least waste Intrinsically safe design: prevent large volumes, little intermediate storage

Process X interruptions and downtime

Fewer units, hence less possibility for malfunctioning of parts

Flexible feedstock specifications Recyclable

X

X

X

X

—

X

X

X


Fewer stoppages due to lowered maintenance Faster grade/product changes Easier to adapt quickly to new process conditions due to smaller equipment, fewer intermediate material or grade changes Easier to dismantle and/or ship Easier to adapt quickly to new processes due to smaller equipment Many other aspects can be thought of

TABLE 2 Process Intensification for Different Types of Industries Fine chemicals Large-scale chemical industries

Multiclient chemicals

Custom manufacturing

Pharmaceutical industries

Characteristics Up to hundreds of kilotons Up to hundreds of tons Continuous Mostly batch, some parts continuous Total profits are 5–30% Total profits are 20–40% turnover turnover Material and energy costs Material and energy costs are 70–90%, labor costs are 70–90%, labor costs 10–30% of total costs 10–30% of total costs

Up to tens of tons Mostly batch, some parts continuous Total profits are 30–50% turnover Material and energy costs are 50–80%, labor costs 20–50% of total costs

Up to tens of tons Mostly batch Total profits are 30–70% of turnover Material and energy costs are 10–30%,labor costs 70–90% of total costs

Why PI? Material and energy cost reduction Higher efficiency, higher production capacity Space and waste reduction

Higher efficiency, labor cost reduction Higher production capacity, material cost reduction Energy cost reduction, optimal price

Higher efficiency, labor Labor cost reduction (also cost reduction in cleaning procedures) Higher production capacity, Higher production material cost reduction capacity Optimal price and time Optimal price and time of of delivery delivery

How? Focus on materials and energy (large impact on costs)


Focus on labor and Focus on labor and Focus on number of efficiency (efficiency so efficiency (efficiency so steps (labor and that all steps in the that all steps in the cleaning intensive) process are well balanced) process are well balanced)

Custom manufacturing, producing up to tens of tons of a chemical for one specific customer, mostly in a batch way Pharmaceutical industries, producing up to tons of a active pharmaceutical ingredient (API), mostly in a batch way Table 2 shows that different chemical industries have different cost focuses. It is important to know the main cost contributors of a certain process, because this is a good way to find step-by-step changes in the overall costs of a process. For different types of chemical industries one can find an idea of the corresponding main drivers in Table 2 (various examples of these main drivers can also be found in the goals defined in Appendix 1). Of course, apart from these general features every process is unique and requires an individual investigation and definition of the cost drivers before one can start a process intensification study. This will be discussed in more detail in the next section. Process Intensification is all about step-by-step changes. At what stage an optimization in a process is called a step change depends not only on the type of business but also on the maturity of a product in the lifecycle and the goals defined by the three main aspects of process intensification (business, legislative requirements, and environment). For example, the step change that one would like to achieve in efficiency, hence in optimizing a process without changing the hardware drastically, depends on the lifecycle of a product. This is shown in Figure 1. This last aspect should be taken into account when defining the reasons for performing a process intensification study. 4. KEY FEATURES OF PROCESS INTENSIFICATION TECHNOLOGIES As stated earlier and in other chapters of this book (e.g., Chapter 1), process intensification is making step-change improvements in processes. Many general possibilities are available for reaching that goal. A large number of technologies that have the characteristics of process intensification involve one or more of the following three imperatives. 1. Make it small. 2. Combine. 3. Intensify the driving force. Examples of these can be found in all chapters throughout the book. For example, belonging to imperative 1 are microreactors (Chapter 5), minireactors (Chapter 6), and compact heat exchangers (Chapter 4). Imperative 2 represents a large field comprising the combining of reaction and extraction, reaction and separation (Chapters 8, 9, and 10), reactive extrusion, reactive heat exchangers


FIGURE 1 Step changes in efficiency as a function of the lifecycle of a product (fictitious curve to show the differences).

(Chapter 4), and many more. The last imperative covers technologies that use centrifugal fields for contacting (Chapters 2 and 3), separations, and crystallization or that involve extremes of pressure and temperature, ultrasound waves, microwaves (for e.g., drying) and electric fields (for e.g., separation or dispersion). Basically, using these technologies one would like to move forward to the theoretical optimum of a chemical process, which is that there are no other limitations than chemical kinetics. Normally a chemical process is influenced by more than just kinetics: hydrodynamics (mixing), heat transfer, and mass transfer determine the quality of the process. Process intensification focuses on removing these three limitations to reaching the goal of kinetically limited processes. This is schematically depicted in Figure 2. These three basic imperatives can be used to begin thinking about when to try to intensify a process in a chemical plant. They are also part of another area of chemical engineering: safety aspects of plant designs. In the 1980s, effort was put into developing strategies for designing so-called “intrinsically safe designs,” which means that the process designs are such that dangers are intrinsically minimized so that the safety of a process does not depend on safety devices.


Examples of strategies for safe designs are: prevent large volumes, intensify contact, reduce the need for aids (like solvents), use little intermediate storage. One can immediately see the similarities between intrinsically safe design and process intensification. This also means that applying process intensification techniques is beneficial for process safety. 5. INTRODUCING PROCESS INTENSIFICATION IN A CHEMICAL PROCESS This section describes how one may intensify a process, or, said differently, how process intensification can be introduced for a process. When one would like to introduce process intensification for a process it is important to consider the development phase a process is in. Roughly said, the following phases can be discerned for a general process: New ideas for a process → Determining process chemistry → Pilot plants studies → Plant design → Plant startup → Debottlenecking or troubleshooting. Every phase requires a somewhat different approach, because the boundary conditions and possibilities for intensifying a process differ for each phase. What follows is a short description of the phases and the approaches.

FIGURE 2 Process Intensification: Approaching the kinetic limit by using the three imperatives: (1) Make it small; (2) combine; (3) intensify the driving force.


1. New ideas for a process: A product and a chemical route are defined. This is the phase in which the core of the process is determined. This means that it is already very important to think about the consequences that selecting a certain pathway has on the future final plant and on the plant costs. Choosing a nonoptimal pathway may result in a much more expensive final plant design than if another pathway (or, e.g., another solvent or catalyst) were to be chosen. This is the phase in which it is still relatively easy and cheap to make changes that produce the most intensified process. To be able to know the effect of a chemical route on the final plant requires a multidisciplinary team to be involved at this phase, not just chemists, because more aspects play a role than chemistry. This multidisciplinary team should consist of at least chemists and technologists. 2. Determining process chemistry: The specific chemical parameters of the chosen process route are determined. In this phase, the route itself is more or less frozen. The important parameters of the process are determined, such as the required temperatures, pressures, and concentrations of substances in the process steps. Making changes and optimizations in this phase is still relatively inexpensive. The effects of this phase on the efficiency of the final process can be large. Applying intensification possibilities in this phase will reduce greatly the costs of the final process. In this phase the kinetic limit, as given in Figure 2, is determined, hence this is the last phase where the intrinsically optimal route may be found—the route with the least by-products and the least waste generation (from auxiliaries such as solvents and from by-products). It is also the last phase in which, apart from the optimal kinetics, the optimal process conditions can be found that cause the fewest problems in technical implementation. Therefore, a multidisciplinary team is required to determine the optimal process route, not just chemists. Again, this multidisciplinary team should consist of at least chemists and technologists. 3. Pilot plants studies: The process is scaled up to pilot-plant size to observe scale-up effects and to determine parameters for the large-scale plant (though one should take into account that this phase is skipped more and more in process designs, to save time and money). In this phase, the basic technologies of the process are checked, and the scalability of the process route developed in the previous phases is checked. Changes in this phase are still relatively easy to make but more costly than in the previous phases, because pilot-plant studies themselves cost money, and larger changes often mean a delay in the development process because the previous phase will also have to be partly redone as well. In this phase one can look for the limitations that technical equipment pose for the process route. Heat and mass transfer limitations usually result in a poorer process performance than the optimal lab conditions gave. In the pilot phase one can try to optimize the equipment to overcome the limitations. For that purpose one can make use of intensified process technologies, as described throughout this book. These new technologies can best be tested during the pilot scale, to collect


experience with using these technologies and for the further scale-up to plant size. Because process technology, chemistry, and future plant equipment play a role, a multidisciplinary team is required consisting of at least chemists, technologists, and process engineers. 4. Plant design: A commercial-scale process is developed and designed in detail. Experience gathered in the previous phases is used for developing the large-scale plant. In this phase, it is difficult to make large changes in the process, because then the previous phases would have to be partly redone and the costs and time spent in the previous phases are lost. It is important that intensified techniques are used as much as possible, since normally more limitations in the process will arise at plant scale than in the lab or at pilot-plant scale. One wants to approach the kinetic limit as much as possible. 5. Plant startup: The commercial plant is started. In this phase, no actions for process intensification studies can be defined. 6. Debottlenecking or troubleshooting: While running the process, problems that may occur are solved (troubleshooting) or larger optimization projects started (debottlenecking). In this phase, the plant is built and the process is running. One gains experience from the large-scale process. Small changes in the process can be made, usually as troubleshooting projects or in creep projects. These small changes, though important, are not the changes meant in process intensification. When one would extend the plant capability with larger steps, this is usually called debottlenecking, which mostly involves investment in the adaptations for the bottlenecks in the process. In this phase one also tries to find possibilities to really make step changes in the cost efficiency of the process, which may be called process intensification. It again requires a multidisciplinary team, because all aspects play a role, from chemistry to process engineering. It can be seen that a multidisciplinary team is very important in almost every phase when the goal is to develop an intensified process. Of course, many of the phases just described in practice run partly in parallel, to speed up the development process. This means that it becomes even more important to develop the most intensified process using multidisciplinary teams as early as possible. In all phases of the process development just described one has opportunities to introduce intensification technologies to intensify the process. In phases 1 and 2, one should not forget that the choices made are crucial for the process afterwards. Also, in these phases one should already be considering the consequences of the choices on the commercial plant and the possible scale-up difficulties using specific chemical routes. The steps these teams can take to arrive at new, intensified solutions are given schematically in Table 3. Step 1. The first step is to define the goal of the study. What is the time frame of possible implementation? When the study is a strategic study, the time frame can be many years. The goal then is to


TABLE 3 Steps Needed to Develop New, Intensified Solution for Processes, from Pilot Phase to Debottlenecking Phase 1. Goal definition Determine the scope

The goal of the study Unit and section of the plant or the entire plant All the people involved (also; the internal customers)

2. Scouting of ideas Identify the thermodynamics, kinetics, and balances (mass, energy) plus the cost factors

Identify the breakthroughs

For existing process: from known data For new process from literature and thermo databanks Determine the theoretically optimal process (no limitations) In multidisciplinary teams Make cost estimates of different options

3. Selection Select the most feasible option

All the people involved (also, the internal customers)

4. Detailed design Identify the detailed thermodynamics and kinetics Final design of the unit or plant

Databanks and new measurements Concluding the intensification project

have new concepts ready for the future. In that case one can also use technologies that are still in development at the present date. When the goal is to achieve direct implementation, the time frame is short and one should use the available data and available technologies to optimize a process. The budget necessary for the latter study is then also different from that for the former type. What part of a process should be looked at? This can be one unit, one chemical conversion step, a total process, or even a combination of processes that are interlinked through streams of intermediate products or energy streams. Step 2. The second step is the actual study: the scouting of the ideas. This phase consists of two subparts. The first part is data collection: What are the kinetics and material properties in the process? What are the mass and energy balances?


What are the current process steps and their limitations? What is the theoretically optimal process (no limitations)? What are the main cost factors? Data should also be available to all team members of the new technologies on the market and in development. This overview should be available before the start of the actual idea generation: The available technologies and technologies under development should be known. This information should be available before starting process intensification studies. It is a one-time effort to collect these data and consequently a smaller continuous action to keep the data up to date. Overviews can of course have various formats and sources. An example of one technology from an extensive list used within DSM is given in Figure 3. Of course, this book is also a good reference for available technologies. The second part of Step 2 is the identification phase. In this phase ideas are collected to intensify the process within the boundaries as defined in Step 1. This phase requires a well-balanced team of participants from various disciplines, e.g., engineers, R&D, process technologists, chemical experts. Apart from representative of various disciplines, one should ideally also have participants that know the process well and people completely unfamiliar with the process, to get a wide variety of ideas and suggestions. Lastly, it is beneficial to mix older, experienced members with younger members who may be less biased to company-standard process solutions. As a start, the existing process should be described using all the known data gathered in the first part. It is very important that the process also be explained in functional units, instead of equipment. This means that the function of a step is described instead of just the details on how a piece of existing equipment works in the actual situation. This helps in finding the differences between the goal of a step and the way the equipment actually works and in finding other Intensification methods to make it work better. Next, free discussion and brainstorming begin. One can use many different starting points for discussions and ideas. An overview of available technologies also is important (e.g., Figure 3). Also helpful would be the list given in Table 4. Combine and select ideas. After brainstorming one should of course focus on the main ideas. Often various ideas can be combined to create clusters of ideas around main themes. Design a draft of the new process (alternatives). After having found the main ideas, a flowsheet of possible intensified processes should be drawn. Such flowcharts help in further evaluations and in gaining clear pictures of the ideas.


FIGURE 3 Example of one technology out of an extensive of technologies used within DSM.


Based on the selected flowcharts (process flow diagrams) one can make a first rough estimate of the costs of the new processes. Various standard methods for this are available, depending also on the company’s systems for cost estimation. This part of Step 2 should not be forgotten, because it can clearly show the benefits of a new process in terms of costs, which is one of the main drivers for changes in processes. It also helps in ranking ideas. Step 3. This step involves the selection of the most feasible route, which should be done in a large group, including the internal customers. Step 4. The last step is the engineering or further development of the intensified process route. This step follows the standard engineering and development processes used for developing conventional processes, although some new technologies may be involved. Pilot plant testing of these new technologies is very likely to be necessary. This method has been applied within DSM for 13 of its existing processes from its three main business clusters (polymers and industrial chemicals, life science products, and performance materials). See Figure 4 for the scheme used within DSM that follows this route. The results showed possibilities for reducing the cost of the present processes to 60–90% of the current costs. (This was shown earlier, in Figure 5 of Chapter 1.) This shows that the way of performing process intensification studies described in this chapter can result in very economical results for various types of businesses (from life sciences to bulk chemicals). Two concrete examples are presented in the next section to give a feeling of process intensification. TABLE 4 Starting Questions That Might Be Used in the Brainstorming Phase of an Intensification Study Keywords Intensify Segment Use a different aid Change conditions Change in an early stage Combine Fix Add or remove Periodic action Phase transition

Example Make the process smaller Divide into independent parts A different solvent, temporary shielding agent Change pressure or temperature Prevent the root-cause upstream from where the problem occurs Combine unit operations Fix one phase to prevent separation problems Quickly remove or add a product or reactants Change a continuous system into periodic actions Use phase transition in the process


FIGURE 4 Scheme to arrive at intensified processes as used within DSM.

6. 6.1.

EXAMPLES OF INTRODUCTIONS S-Ibuprofen

The first example of process intensification at DSM is the pilot-scale test of the enzymatic production of S-ibuprofen, a nonsteroidal, anti-inflammatory drug. The molecular scheme is given in Figure 5. More details can be found in Refs. 3 and 4. The conventional production consisted of many process steps, typical for a fine chemical process. These process steps are given in Figure 6. The selectivity of the enzyme is rather high. The downstream processing is rather laborious,


FIGURE 5 Production of S-ibuprofen with a racemization step and an enzymatic conversion using the carboxyl esterase enzyme.

because it requires that the product be completely enzyme free. Furthermore, the product, S-ibuprofen, causes enzyme deactivation. Altogether, many unit operations are required to obtain the product; due to the deactivation of the enzyme, a maximum conversion of 21% can be reached. Removing the product as fast as possible from the reaction mixture could possibly prevent the enzyme from deactivating. This was experimentally tried on pilot scale using an ultrafiltration unit, parallel to the reactor. In this filtration unit the product is separated and the unreacted components and the enzyme are returned to the reactor. An even better option, using a true membrane reactor in which the catalyst (enzyme) would remain on one side and the product would remain on the other side, was not tested. Both options are given schematically in Figure 7.

FIGURE 6 Processing steps for the production of S-ibuprofen.


FIGURE 7 S-Ibuprofen production options: ultrafiltration (left) and a membrane reactor (right).

FIGURE 8 Processing steps in the production of S-ibuprofen with the ultrafiltration unit.

FIGURE 9 Conversion of the S-ibuprofen process with and without the ultrafiltration unit.


Performing the S-ibuprofen production process with an ultrafiltration unit, 6 of the 11 unit production steps shown in Figure 6 could be skipped! The entire process scheme was reduced to the one given in Figure 8. Based on kinetic modeling of the reaction, including the enzyme deactivation, it was found that there would be an additional improvement, apart from skipping six unit operations. By removing the product quickly out of the reacting medium, the enzyme deactivation proceeds more slowly and the conversion increases. This has been experimentally validated in the pilot-scale system. The results are given in Figure 9. It can be seen that the conversion is increased from 21% to 50%! This example shows that process intensification can bring a substantial cost benefit to the process. 6.2.

Urea

The second example of process intensification at DSM is the urea process (5). The history of the urea process at DSM is rather long, as shown in Table 5. Urea is produced in a two-step process. The first step is the formation of carbamate from NH3 and CO2. This reaction is exothermic. The second step is the decomposition of carbamate into urea and water. This second reaction is slightly endothermic. Both reactions are equilibrium reactions. The conversion to urea in equilibrium is about 60%. This means that substantial recycle flow is necessary to obtain sufficient overall conversion. In the reaction section the main unit operations are: The stripper, in which remaining CO2 and NH3 are being stripped with CO2 from the product flow The scrubber, where the reactants in the reactor offgas are stripped

TABLE 5 Development of the Urea Process at DSM 1945 1956 1966 1974 1994 1995 1996 1997 1998

Start of R&D program by DSM First commercial plant (75 mt/d) CO2 stripping process World-scale capacity 1750 mt/d Implementation of first pool condenser 200th unit contracted Urea 2000plusTM technology Installation of first pool reactor, largest single-line urea plant in operation Startup of first Urea 2000plusTM pool reactor plant


The condensor, in which NH3 and CO2 are condensed into carbamate The reactor, in which the carbamate decomposes into urea and water In the old process, the Stamicarbon stripping process, these four units were positioned above each other. The overall height of the plant was 76 m! After various small and larger improvements, a new concept was developed in which the reactor and the stripper are combined in one unit, the horizontal pool reactor. A schematic diagram and a photo of the reactor are given in Figure 10. The layout of the new plant concept with the integrated and intensified pool reactor now has a height of 18 m, a height reduction by a factor of 3. Also, the number of units has been decreased. This is illustrated in Figure 11, in which the development of the urea process is depicted. In this figure it can be seen that even an old “proven” bulk chemical process can be intensified, resulting in a much more compact and economical plant.

FIGURE 10 Urea pool reactor used in the Urea 2000plusTM process.


FIGURE 11 Development of the urea process.

7.

CONCLUSIONS

As shown in this chapter, the methodology for applying process intensification in commercial industries requires a broad interest and cooperation within companies, including a vision of the company management and individuals. Some boundary conditions need to be met, such as a vision in place that offers the initial time and money to set up the necessary infrastructure for applying these new technologies: a basic knowledge of the possibilities and available technologies and of the methodology. Process intensification can then bring substantial benefit in terms of business, legislation, and environment, as is demonstrated in this chapter as well as in the entire book, therefore making PI very worthwhile to be implemented. 8. 8.1.

APPENDICES Appendix 1: Excerpt from U.S. Vision 2020

This excerpt is taken from Ref. 6. U.S. chemical companies must innovate and change to keep competitive in the global environment. Chemical Industry Vision 2020 Technology


Partnership (Vision 2020) is an initiative to leverage research and development (R&D) resources through the collaborative efforts of industry, government, and academe and focus them on priority targets identified by the chemical industry. New Process Chemistry Roadmap Section: Office of Industrial Technologies (OIT) This roadmap focuses on innovative new process chemistry to support the Process Science and Engineering Technology and Chemical Synthesis technical areas within the New Chemical Science and Engineering Technology section of Vision 2020: The U.S. Chemical Industry. However, the roadmap is integrally connected to other sections of the Vision and other supporting roadmaps. Process chemistry begins in the laboratory, but is further developed through reactor design and process engineering. Feedstock characteristics, catalytic mechanisms, and downstream processing (e.g., separations) all come into play in developing new process chemistries. Goals Reduce feedstock losses to waste or low-value by-products by 90%. Reduce industry-wide energy intensity (energy per unit product) by 30%. Reduce total emissions and effluents from chemical manufacturing by 30%. Increase usage of C1 and renewable resources to 33% of industry-wide carbon usage. Reduce cost of production by 25%. Accelerate introduction of new products by 15%. Reduce lead times and time to market for new products and technologies by 30%. 8.2.

Appendix 2: European Union Research Program

This excerpt is taken from Ref. 7. The 5th framework of the EU research program runs from 1998 to 2000. One of the four subprograms is called the GROWTH program. In this GROWTH program, key action 1 is called innovative products, processes, and organization. The description is given here: KEY ACTION 1: INNOVATIVE PRODUCTS, PROCESSES, AND ORGANIZATION (Budget 731 MEuro) a.

Contributing to modernization of industry and adaptation to change, through the combined effects of improved industrial capability and innovation capacity, while introducing more flexibility and capability to respond in real time to customer needs.


Research should stimulate cross-sectoral exchanges and participation of SMEs, taking into account their specific needs and roles in the supply chain as well as approaches able to create and hold in Europe sufficient jobs to arrest the decline of industrial employment while improving the overall quality of work. b. Substantially* improving overall quality within the value chain (quality is intimately linked to value for and timely satisfaction of customer needs at the lowest costs) and consequently reducing “inefficiencies” and overall lifecycle product costs by the same order of magnitude. c. Minimizing resource consumption (e.g., materials, energy, water) to reduce substantially the overall “lifecycle” impact of “product-service” provision and use. These goals should be dealt with in a synergistic way. They should not be regarded as absolute targets for individual projects but rather as broad indications of the direction toward which the European industrial system, supported by improved regulations, should evolve. 8.3. Appendix 3. Excerpt from the CEFIC Technology Program SUSTECH This excerpt is taken from Ref. 8. The SUSTECH program was launched by a consortium of the major chemical companies in Europe in April 1994. In doing so, they created a framework which would enable them to collaborate effectively in areas of research and development, which would be critical for the long term future of the process industries in Europe. The focus for collaboration would be those areas that while attracting public concern were nevertheless not areas in which the companies would normally compete. Such areas as sustainable development, protection of the environment, and technologies for making more efficient use of the earth’s resources were ready targets for collaborative action. Process Intensification. The objective of the Process Intensification (PI) Cluster is to promote PI and its benefits to a wide industrial audience by: Exposing technologies critically to industrial needs Identifying priority industrial applications * The term substantially means over 20–30% in the shorter term or over 10% per year in the longer term.


Facilitating partnerships between technology providers, equipment suppliers, and end users. Two sessions were held at SUSTECH 7. In the morning session, process intensification and the future direction of the cluster were discussed in general. In the afternoon, a mini-workshop was held on gas–liquid reactors. Total attendance was in excess of 60, including 23 representatives from the chemical industry. An industrial Steering Group has been established to determine the priority areas for the cluster and the mechanisms required to achieve its objectives. It was proposed to carry out analyses on representative processes to identify the potential business benefits arising from application of PI technologies. REFERENCES 1. BHR Group. Process Intensification Conferences. http://www.bhrgroup.co.uk/confsite/ pi01home.html. 2. Elverding P. DSM’s Triple Bottom Line. DEFACTO 2001; 15(5):28–32. 3. Bakker RA, Stankiewicz AI, Schyns VJAJ. Process intensification within DSM, general methodology and concrete examples. In: Proceedings of the 4th International Conference on Process Intensification for the Chemical Industry. Cranfield, UK, BHR Group, 2001. 4. Cauwenberg V, Vergossen P, Stankiewicz A, Kierkels H. Integration of reaction and separation in manufacturing of pharmaceuticals: membrane-mediated production of S-ibuprofen. Chem Eng Sci 1999; 54:1473–1477. 5. Technical information brochures on the Urea 2000plus™ process can be obtained from Stamicarbon B.V., the licensing company of DSM. 6. U.S. Vision 2020, section of the Office of Industrial Technologies (OIT). http://www. oit.doe.gov/chemicals/visions_new_chemistry.shtml. 7. European Union 5th Framework Research Program 1998–2002. Thematic program GROWTH: http://europa.eu.int/comm/research/growth/index.html and http:// europa.eu.int/comm/research/growth/pdf/growth-workprog2000_en.pdf. 8. SUSTECH Technology program, part of the CEFIC organization. http://www.cefic.be.


13 Process Intensification for Safety Dennis C. Hendershot Rohm and Haas Company, Bristol, Pennsylvania, USA

1.

INTRODUCTION

Process intensification is an important strategy in the development of inherently safer chemical processes and plants. By reducing the inventory of hazardous material or energy in the process, the potential consequence of failure to control that hazardous material or energy is reduced. Rather than relying on add-on safety features such as interlocks, procedures, and consequence mitigation systems, the safety of the plant is based on reducing the magnitude of the possible damage. While safety devices can be designed to be highly reliable, no safety device is perfect, and all will have a finite failure probability. If a chemical plant contains a large amount of hazardous material or energy, the consequences of the failure of the add-on safety devices may be large. A smaller plant is safer because we have reduced its inherent capability to cause damage, rather than because we have controlled that capability through additional safety devices. 2.

INHERENT SAFETY

Inherent is defined as “existing in something as a permanent and inseparable element, quality, or attribute” (1). A chemical process can be described as inherently safer if it reduces or eliminates a hazard when compared to another process alternative. To understand this definition, it is necessary to understand what is meant


by the term hazard. A hazard is a characteristic of a material or process that results in potential harm to people, the environment, or property. A hazard is an inherent property of a material or its conditions of use. Some examples of hazards include: Sulfuric acid is corrosive. Chlorine is toxic by inhalation. Heptane is flammable. Nitroglycerine is unstable and can detonate. Heat transfer oil at 300C contains a large amount of energy. A vessel full of compressed air at 40 bar contains a large amount of potential energy. The hazard cannot be changed, but it is possible to change the material or operating conditions. The magnitude of a potential incident that could result from the failure to control the hazard can also be reduced. An important way of accomplishing this objective is to reduce the quantity of material in the process through process intensification. It is generally more appropriate to describe processes as inherently safer when compared to alternatives rather than as inherently safe. All processes have multiple hazards, and it is not possible to eliminate all hazards. For example, it may be possible to replace a toxic or flammable solvent for an extraction process with supercritical carbon dioxide. While the carbon dioxide process is inherently safer with respect to flammability and toxicity hazards, it operates at high pressure, which introduces new hazards. Depending on the specific details of the alternative processes—what the flammability properties of the solvent are, how toxic the solvent is, what the size of the equipment is, what the operating pressure is—the supercritical carbon dioxide process may be inherently safer overall. However, it is important to consider all hazards in judging the overall inherent safety of process alternatives and to avoid focusing on a single hazard and forgetting about others. 3.

HISTORY OF INHERENTLY SAFER DESIGN

It can be argued that engineers have always attempted to eliminate hazards in their designs. For example, in 1828 the pioneering railway engineer Robert Stevenson argued for simplification and minimization—two of the principles of inherently safer design—of the newly developed steam locomotive when he discussed “an alteration which I think will considerably reduce the quantity of machinery as well as the liability to mismanagement. . . . [I]n their present complicated state they cannot be managed by ‘fools’; therefore they must undergo some alteration or amendment” (2). The concept of inherent safety as a specific set of design strategies to eliminate or reduce hazards in the chemical industry was first articulated by Trevor Kletz in the 1970s. Following the Flixborough explosion in England in 1974,


there was increased concern about safety in chemical plants, from the industry itself, government, and the general public. This resulted in an increased focus on controlling hazards in chemical plants. A specific worry was that the magnitude of potential accidents was now larger because of the large size of a new generation of world-scale petrochemical facilities. In 1977, while working as a safety advisor for the ICI Petrochemicals Division, Kletz suggested an alternative approach in the annual Jubilee Lecture to the U.K. Society of the Chemical Industry (3). He proposed that the industry should direct its attention to eliminating or reducing the hazards of its plants and processes rather than accepting those hazards and working to control them. Kletz called this approach inherently safer design. Furthermore, this hazard elimination or reduction would be accomplished by means that were inherent in the process and, thus, permanent and inseparable from it. Since 1977, concepts and approaches to inherently safer design in the chemical industry have been developed and promoted (4–6), including a brief discussion in Perry’s Chemical Engineers’ Handbook (7). 4. CONCEPT OF LAYER OF PROTECTION FOR PROCESS SAFETY The safety of a chemical process relies on multiple layers of protection to protect people, the environment, and property from the hazards associated with the process. A process designer recognizes that equipment can fail and that people will make mistakes. While we can design more reliable equipment and train and motivate people to reduce mistakes, we can never completely eliminate these failures and mistakes. Therefore, it is important to provide multiple layers of protection, a defense in depth, to reduce risk (Figure 1). Even so, there will always be some small probability that all of the layers of protection will fail simultaneously and an accident will occur. Also, the effectiveness of the layers of protection depends on the ongoing maintenance of equipment, the training and performance of people, and the management systems. If these systems deteriorate, the reliability of the protection layers will be reduced and the risk will increase (Figure 2). The Center for Chemical Process Safety has published a book that describes the layerof-protection concept in detail and extends it to a quantitative risk-management technique, layer of protection analysis (LOPA) (8). If the magnitude of the potential accident is very large, some people may never feel completely comfortable with the residual risk that it may occur, even if this risk is very small and management systems to maintain the protection systems continue to be good and effective. By reducing the magnitude of the potential accident, inherently safer design acknowledges the inevitability of the failure of equipment, people, and management systems, and it bases process safety on reducing the inherent hazard of the process. An inherently safer design reduces the need for layers of protection for a process, and, if the magnitude of the potential


FIGURE 1 Layers of protection for a chemical process.

consequence of an accident can be reduced sufficiently, it may eliminate the need for protection layers entirely (Figure 3). In general, the layers of protection applied to a chemical process can be placed into four categories: Inherent—Eliminate or reduce the hazard by using materials and conditions that are less hazardous or nonhazardous. Passive—Minimize or control the hazard using design features that reduce either the frequency or the consequence of incidents without the active functioning of any device.


Active—Control or mitigate incidents using controls, safety interlocks, or emergency shutdown systems to detect hazardous conditions and take appropriate action to place the plant in a safe condition. Procedural—Use operating procedures, administrative checks, emergency response, and other management systems to prevent incidents, to detect incidents in time for operators to place the plant in a safe condition, or to reduce the magnitude of the damage resulting from an incident. These categories are not rigidly defined, and a design may exhibit characteristics of more than one category. Safety strategies in the inherent and passive

FIGURE 2 Higher risk resulting from degraded layers of protection.


FIGURE 3 An inherently safer process meets risk requirements with fewer or no layers of protection.

categories are generally considered to be more robust and reliable. They depend on the physical and chemical properties of the system rather than on the successful operation of instruments, safety devices, and procedures. Inherent and passive strategies are often confused, but they are different. A truly inherent solution to a safety issue will either completely eliminate the hazard or reduce the potential magnitude of an incident associated with the hazard sufficiently that it cannot cause significant damage. On the other hand, passive strategies do not eliminate


the hazard, but instead prevent injury and damage by eliminating or reducing the exposure of people or property to the hazardous condition without the active functioning of any device. A containment dike around a storage tank is an example of a passive layer of protection. It performs its function of limiting the damage in case of a leak from the storage tank simply by being present. There is no need to detect a leak or for any device or person to perform any function for the containment dike to do its job. 5.

INHERENT SAFETY STRATEGIES

Approaches to inherently safer process design have been categorized in a number of ways. The Center for Chemical Process Safety (5) describes four strategies for inherent safety, derived from Kletz’s initial proposals (3,6): 11. Minimize—Use smaller quantities of hazardous substances, including process intensification approaches. Example: Use a small, continuous reaction system for the production of nitroglycerine in place of a large batch reactor. 12. Substitute—Replace a material with a less hazardous substance. Example: Aqueous latex paints reduce both flammability and toxicity hazards when compared to solvent-based paints, both during manufacture and for the final consumer. 13. Moderate—Use less hazardous conditions or a less hazardous form of a substance. Example: Plastic resins can be produced in pellet or granular form, reducing dust explosion hazards when compared to a powder form of the same material. 14. Simplify—Design processes and facilities that eliminate unnecessary complexity and that are tolerant of human error. Example: Design piping to permit gravity flow of hazardous materials in a plant, eliminating the need for pumps, which can leak. Additional discussion and more examples of these strategies can be found in books by Kletz (6) and CCPS (4,5). The remainder of this discussion will focus on minimization (process intensification) as an inherent safety strategy. 6. PROCESS INTENSIFICATION AS AN INHERENT SAFETY STRATEGY 6.1.

Smaller Is Safer

Reducing the size of chemical processing equipment enhances safety in two ways. The quantity of hazardous material that can be released in case of equipment leakage or rupture is obviously smaller if the equipment is smaller. In addition,


the potential energy contained in the plant equipment is also smaller if the equipment is smaller. This potential energy can be in many forms, such as high temperature, high pressure, or heat of reaction from a reactive chemical mixture in the equipment. If this potential energy is released in an uncontrolled manner, an incident such as a fire, explosion, or uncontrolled leak of material from the equipment may occur. Clearly the potential damage from the uncontrolled release of material or energy is reduced if equipment can be made smaller. There is also another benefit to smaller equipment—it may be more feasible to provide equipment to mitigate or control the consequences of an incident. For example, it may be feasible to totally enclose a small reactor in a blast-proof structure. Doing this for a large reactor may not be feasible because the blast-proof enclosure would have to be much larger. The enclosure would also have to be much stronger because it would have to withstand a larger potential explosion from the larger reactor. 6.2.

Traditional Approaches to Minimizing Inventory

There are many opportunities to minimize inventory of hazardous material in a chemical plant without fundamental changes in process technology. The accident in Bhopal, India, in 1984 released methyl isocyanate, causing approximately 2000 fatalities and injuring tens of thousands of people (9). This is by far the worst accident in the history of the chemical industry. Following the Bhopal accident, most chemical companies reviewed their operations to identify opportunities to reduce the inventory of toxic and flammable materials. Many significant reductions were reported as a result of this effort, and these reductions were accomplished relatively quickly. Obviously, these companies did not rebuild plants using different technology or make dramatic changes to the process equipment in the existing plants in such a short time. So how did chemical plants around the world reduce hazardous material inventories? They carefully evaluated existing equipment and operations and identified changes in operations that would allow the existing plants to operate with a reduced inventory of hazardous materials. The Bhopal tragedy focused the attention of creative engineers on the problem of reducing hazardous material inventory, and they quickly identified ways of accomplishing this objective, even for existing plants and technologies. 6.2.1.

Storage

The single biggest reduction in hazardous material inventory of plants was in storage facilities. This includes raw materials, in-process intermediates, and final products. The terrible consequences of Bhopal caused engineers and managers to question the need for storage of large quantities of raw materials and intermediates. In many cases, a large inventory of raw materials makes it easier to operate a plant. The company may have more flexibility in ordering raw materials, to take advantage of favorable market conditions to order large amounts of material when


prices are low. The plant is less likely to have to shut down because of transportation delays. However, if the raw material is very hazardous, and particularly if it is very toxic and a large leak has the potential to impact population in a large area surrounding the plant, the risks associated with storage of a large amount of the material may be significant. A better strategy would be to devote the company’s resources to ensuring a reliable and economic supply of material, effective transportation systems, and good inventory control systems. Strategic alliances with raw material suppliers and transportation companies, modern inventory management systems, and improved communication with supplier plants will allow a plant to reduce the quantity of hazardous material that must be stored at the manufacturing site. There is also an economic benefit—working capital is reduced because the inventory of raw material is reduced. Often, a plant design includes large storage tanks for hazardous in-process intermediates. This intermediate storage decouples sections of a plant from one another. Parts of a plant can continue to operate, either filling or emptying inprocess storage tanks, while another unit in the plant is shut down for maintenance or because of operating problems. In fact, reliability engineers interested in increasing the on-stream efficiency of a plant will encourage a plant designer to include intermediate storage buffers in a design to improve the overall plant reliability. This can be desirable in improving the inherent safety of a plant—a continuous plant is generally safer in normal, steady-state operation than while starting up or shutting down (10). The buffers will reduce the frequency of starting up and shutting down sections of the plant because of unavailability of feedstock from an upstream unit or a shutdown of a downstream unit resulting in no place to put the product. However, if the intermediate to be stored in the buffer tank is extremely hazardous (flammable, toxic, reactive), this benefit must be balanced against the risks inherent in the storage of a large quantity of hazardous material. There may be other approaches to enhancing the reliability of the units within a large plant that will eliminate or greatly reduce the need to store large quantities of hazardous intermediates. Perhaps the intermediate buffer can be provided at another point in the process, allowing storage of a less hazardous material or a less hazardous form of the same material. For example, a reactive material is produced in a continuous gas-phase reaction, then absorbed in water to form an aqueous solution, isolated as a pure material by distillation, and stored for feeding to downstream processing units. A buffer between this unit and the downstream units could be provided with intermediate storage of the distilled, purified material. However, it would be inherently safer to provide this storage buffer for the aqueous solution of the material, which is less hazardous. There are other ways to reduce the need for in-process storage of hazardous intermediates. Fully understanding the causes of unreliability in production units is the key to reducing the need for in-process buffer tanks. Causes of shutdown can be eliminated by using more robust and reliable equipment, improving the process to


make it less sensitive to variations in operating parameters, or providing redundancy for critical equipment so that the entire plant does not have to be shut down when the critical equipment requires repair. If the reliability of individual units in a plant can be increased, it will not be necessary to store large amounts of hazardous in-process intermediates to maintain the desired production capacity. The design engineer should question the need for all in-process storage. He should ask if the plant can operate just as efficiently without the intermediate storage tanks if the causes of plant shutdowns could be understood and eliminated. In many cases, the answer is yes, and in-process storage has been greatly reduced or eliminated. One company reported a reduction in total inventory of hazardous materials, including chlorine and hydrogen cyanide, of over 1 million pounds at a single plant site (11). One measure of the inherent safety of a process with respect to fire and explosion risk is the Dow Fire and Explosion Index (12). Table 1 shows examples of the impact of inventory reduction on the Fire and Explosion Index. The benefits of reduced inventory can also be quantified by estimating the potential consequence from a potential incident. For example, one of the hazards of storage of a liquified flammable gas such as propylene is a boiling-liquid expanding vapor explosion (BLEVE) and the associated fireball. Figure 4 shows the heat radiation intensity as a function of distance from a storage tank for a potential BLEVE for three sizes of propylene storage vessel, ranging from 500 to 50,000 kg. The magnitude of the consequence reduction from reducing propylene inventory is clear. 6.2.2.

Piping

When designing piping for hazardous materials, the designer should attempt to minimize inventory, by minimizing both pipe diameter and pipe length. Hazardous material piping should be large enough to transport the quantity of material required and no larger. Reducing pipe size from 50-mm diameter to 25-mm diameter reduces the inventory of hazardous material in the pipe by a factor of 4. This has a TABLE 1 Impact of Inventory on the Fire and Explosion Index Fire and explosion index Ethyl acrylate storage inventory 907,000 kg (2 million pounds) 91,000 kg (200,000 pounds) 23,000 kg (50,000 pounds) Agricultural product with in-process intermediate storage 23 cubic m (6000 gallons) 0 Source: Ref. 13.


151 130 120

185 140

FIGURE 4 Radiation intensity from a BLEVE and resulting fireball as a function of distance for propylene storage tanks in three different sizes.

major impact on the distance over which hazardous concentrations of material can occur in the atmosphere if the pipe is broken. Figure 5 shows the footprint of a toxic phosgene cloud for a specific set of conditions, where the only difference is the size of the phosgene pipe. The smaller pipe results in a much smaller toxic vapor cloud. Inventory of hazardous material in pipes can also be minimized by using the hazardous material as a gas rather than as a liquid. The Dow Chemical Exposure Index (14) is a tool that can be used to measure inherent safety with regard to potential toxic exposure risk. Table 2 shows the reduction in the Chemical Exposure Index that can be realized by handling a number of hazardous materials as a gas rather than as a liquid, assuming that the same-size pipe can deliver the required flow rate. Figure 6 shows the decrease in the hazard zone (toxic cloud footprint) that resulted from relocating a chlorine vaporizer from a production building to


482

Hendershot

FIGURE 5 Cloud footprint to atmospheric concentration of 1 ppm resulting from the rupture of three sizes of phosgene pipe. Release conditions: Complete rupture of pipe while connected to a large storage tank without shutoff, pipe elevation is 5 m above grade, wind speed is 5 m/sec, atmospheric stability class D, 1 ppm is the Emergency Response Planning Guideline-3 (ERPG-3) concentration for phosgene, the concentration at which life-threatening effects might result from exposure for 1 hour.

a storage area. Following the modification, the long chlorine transfer line contained gaseous chlorine instead of liquid chlorine. 6.3.

Process Intensification

Process intensification refers to a chemical process using significantly smaller equipment. Examples include novel reactors, intense mixing devices, heat and mass transfer devices that provide high surface area per unit of volume, devices


that combine one or more unit operations in a single piece of equipment, and devices that use alternate ways of delivering energy to processing equipment— for example, ultrasound, microwaves, laser beams or light, and radiation. Other chapters in this book discuss these and other process intensification techniques in more depth. These technologies can greatly increase the rate of physical and chemical processes, allowing a very high productivity from a small volume of inprocess inventory. Clearly, this is desirable from an economic perspective— a small, highly efficient plant can be expected to be cheaper and more cost effective. If the material contained in the plant is hazardous, because of either its physical and toxicological properties or its conditions of use, such as high temperature or pressure, then the risk associated with a small amount of the material will be less than for a large quantity. If the plant is small enough, the maximum possible accident may not pose a significant hazard to people, the environment, or property. This may result in an additional reduction in the equipment needed for the plant—it may not require as much (or any) safety equipment, emergency alarms and interlocks, or other layers of protection to manage risk. Even if the small plant still requires safety equipment, this equipment will be smaller and cheaper. Installation and ongoing operation of safety equipment is often a major expense; if it can be eliminated or reduced in size and complexity, there will be cost savings. Safety need not cost money—safer can also be cheaper if a small, efficient, inherently safer process can be invented. Some specific examples of process intensification resulting in safer as well as more economical processes follow. Many more examples, described in more detail, can be found throughout this book.

TABLE 2 Chemical Exposure Index for Failure of a 2-Inch Pipe Material

State

Chemical exposure index

Phosgene

Liquid Gas Liquid Gas Liquid Gas Liquid Gas

1000 850 1000 490 980 310 360 110

Chlorine Hydrogen sulfide Sulfur trioxide Source: Ref. 14.


FIGURE 6 Cloud footprint to an atmospheric concentration of 20 ppm resulting from the rupture of a 50-mm-diameter chlorine pipe containing either chlorine liquid or chlorine vapor. Release conditions: Complete rupture of pipe without shutoff, pipe elevation is 5 m above grade, wind speed is 5 m/sec, atmospheric stability class D, 20 ppm is the Emergency Response Planning Guideline-3 (ERPG-3) concentration for chlorine, the concentration at which life-threatening effects might result from exposure for 1 hour.

6.3.1.

Nitration

Nitration reactions are highly exothermic, can generate high pressure in a closed system from noncondensible by-products from undesired side reactions, and often produce unstable reaction products, such as explosives. Many years ago, products such as nitroglycerine were manufactured in large batch reactors. As engineers began to understand the physical and chemical processes involved in nitration chemistry, they recognized that the chemical reaction actually occurs very rapidly once the reactants come into contact with each other. The large reactor


size and long reaction times were actually the result of poor mixing, poor removal of the heat of reaction, and poor mass transfer in two-phase reaction mixtures. This knowledge of the physical and chemical processes involved in an industrial-scale nitration process was used to design new types of reactors to efficiently contact the reactants and liquid phases and remove the heat of reaction. Modern nitration plants use very small continuous stirred-tank reactors with intense mixing and large heat transfer area or jet reactors of various designs to provide intense mixing and rapid contacting of reactants (6). 6.3.2.

Polymerization

A number of innovative polymerization reactors using loop reactors, plug-flow and static mixer reactors, and continuous stirred-tank reactors have been reported. For example, Wilkinson and Geddes (15) describe a 50-liter reactor that has the same capacity as a 5000-gallon batch reactor. Extruders, thin-film evaporators, and other devices designed to provide intense mixing for viscous or unstable materials have also been used as reactors. 6.3.3.

Tubular or Jet Reactors

Continuous reactors, including simple plug-flow pipe reactors, tubular reactors containing static or other mixing devices, and jet reactors of various types, have been used to efficiently produce toxic materials for immediate consumption in downstream processing operations with little or no inventory. Some examples follow. Methyl isocyanate (MIC), the material that was released at Bhopal, can be produced and immediately converted to final product in a process that contains a total inventory of less than 10 kg of MIC (6). Caro’s acid, an equilibrium mixture of sulfuric acid, water, and peroxymonosulfuric acid, is used in the metal-extraction industry. It is manufactured by reacting concentrated sulfuric acid with hydrogen peroxide. Caro’s acid is a powerful oxidizing agent and decomposes readily. A process was developed to manufacture 1000 kg/day of Caro’s acid in a tubular reactor with a volume of 20 ml and a residence time of less than one second, with the product immediately mixed with the solution to be treated (16). A continuous tubular reactor was developed to manufacture phosgene for immediate consumption by a group of batch-processing buildings (17). One plant using the new design contains 70 kg of phosgene gas, compared to a total inventory of 25,000 kg of liquid phosgene in the old plant. Because the new plant is small, it is also feasible to provide a secondary containment building for the equipment containing phosgene, providing an additional barrier between a highly toxic material and people and the surrounding environment (18).


6.3.4.

Heat Exchangers

Heat exchange equipment varies widely in efficiency in terms of available heat transfer area per volume of inventory contained in the exchanger. Innovative heat exchanger design can reduce the volume of hazardous material in a heat exchanger by a factor of 10 or more when compared to a standard shell-and-tube exchanger. Some more efficient alternatives include plate, shell-and-finned tube, plate-fin, and rotary exchangers. With some of these devices, it is important to consider the potential for additional leak scenarios, such as gasket leaks in plate exchangers or hazards associated with seals and moving parts in rotary exchangers. The designer must balance the benefits of reduced inventory against the possibility that leaks may be more likely. Kletz (6) discusses alternative heat exchanger designs. 6.3.5.

Distillation

Innovative design of distillation devices can greatly reduce inventory and can also reduce the residence time in the distillation system, which may be important for safely processing thermally sensitive materials. Wiped film evaporators have been in use for many years, particularly for distillation of reactive materials, such as monomers, and for other heat-sensitive compounds. In more traditional distillation equipment, columns and trays can be designed to minimize the inventory of liquid in the column. Often the largest inventory of hazardous liquid is contained in the bottom of the column and in the reboiler. This is particularly undesirable for a reactive or thermally sensitive material because this is part of the distillation column that is at the highest temperature. It is possible to design the bottom of a column to minimize this inventory. For example, use a conical column bottom or a small-diameter column bottom that still provides sufficient liquid head for bottoms or reboiler recirculation pumps but reduces the total quantity of hot liquid. Centrifugal distillation equipment such as Higee (19) can be much smaller than conventional distillation equipment. 6.3.6.

Extraction

Extraction columns are often very large devices, and they may contain a large quantity of flammable or toxic solvent. The inventory of hazardous material in an extraction system can be greatly reduced by using a centrifugal extractor. In this case, the inherent safety benefits of reduced inventory must be balanced against the new hazards introduced by the centrifugal extractor, including operation at high speed and potential for leakage from seals, to determine the best choice for a particular application. 6.3.7.

Combined Unit Operations

Equipment that combines more than one unit operation in a single piece of equipment is another approach to process intensification. Reactive distillation is a good


example. In one case, a traditional process for the manufacture of methyl acetate uses a stirred-tank reactor, eight distillation columns, and an extraction column. A new, reactive distillation process uses one reactive distillation column and two additional columns (5). The total inventory of combustible material in the process is reduced. In this example, the reactive distillation system also results in a simpler plant—one with fewer major vessels. The process is also simpler because there is much less supporting equipment—condensers, reboilers, pipes, valves, pumps, instruments. Safety is enhanced because every valve seal, flange, pump seal, and instrument connection that is eliminated from the plant is one less place that the process can leak. This reduces fugitive emissions from this equipment, which contribute to plant emissions and pollution, and also the likelihood of bigger leaks that can cause a fire, personnel exposure or injury, or a significant environmental incident. 6.3.8.

Innovative Energy Sources

Energy sources such as laser light, ultraviolet light, microwaves, and ultrasonic energy can be used to apply energy in a controlled fashion to a chemical reaction or physical unit operation to increase efficiency. For example, technology for the distributed, small-scale manufacture of hydrogen cyanide using microwave energy to direct heat to the reactor catalyst is under development (20). This would allow hydrogen cyanide to be manufactured in small quantities where it is needed, rather than manufacturing it in a large central facility, storing it, and transporting it to the sites where it is needed. 7.

METRICS FOR INHERENT SAFETY

How do you measure inherent safety? This is an important question that must be answered to effectively promote the development of inherently safer process technologies, including process intensification. Development of new technology requires resources, and these resources can be more easily obtained if the costs and potential benefits of the new technology, including economics, safety, and environment, can be measured. Also, chemical processes always involve multiple hazards. A process that is inherently safer with respect to one hazard may be less safe with respect to a different hazard, or it may introduce new hazards. For example, our earlier discussion about extraction points out the benefits of reduced inventory that can be obtained with a centrifugal extractor but acknowledges the new hazards introduced by the use of high-speed rotating equipment and the potential for leakage from seals. Some method of quantifying these risks is essential to making the best technology decisions. The chemical industry is just beginning to develop tools for measuring inherent safety. Some of these tools have been used for risk management and loss prevention for some time, but we are just beginning to recognize their value in understanding the inherent safety of processes.


7.1.

Accident Consequence Analysis

The analysis of the potential consequences of an accident is a useful way of understanding the relative inherent safety of process alternatives. These consequences might consider, for example, the distance to a benchmark level of damage resulting from a fire, explosion, or toxic material release. Accident consequence analysis is of particular value in understanding the benefits of minimization, moderation, and limitation of effects. This discussion includes several examples of the use of potential accident consequence analysis as a way of measuring inherent safety, such as the BLEVE and toxic gas plume model results shown in Figures 4, 5, and 6. There are many commercial and public domain tools available for accident consequence modeling. Guidelines for Use of Vapor Cloud Dispersion Models (21) reviews many of the available models for flammable and toxic gas clouds in the atmosphere. Guidelines for Consequence Analysis of Chemical Releases (22) describes modeling techniques for fires and explosions, some of the simpler atmospheric dispersion models, and models for estimating the impact of different types of incidents on people and property. This book also includes a set of spreadsheets for the models discussed. The TNO “Yellow Book” (23) describes a wide range of models for many types of potential accidents in a chemical plant. The TNO “Green Book” (24) reviews models for estimating the potential impact of accidents on people and property. 7.2.

Risk Indices

A number of risk indices have been developed over the years as chemical process loss prevention and risk management tools. Many of these are based on the inherent characteristics of the processes, and they can be used as measures of process inherent safety. In general, these indices measure a single aspect of inherent safety, and it is necessary to use several indices to obtain a full understanding of the overall process characteristics. The Dow Fire and Explosion Index (FEI) (12) and the Dow Chemical Exposure Index (CEI) (14) are two commonly used tools that measure inherent safety characteristics. Gowland (25) reports on the use of the FEI and CEI in the development of safety improvements for a urethane plant. Tables 1 and 2 illustrate the application of the FEI and CEI in measuring inherent safety characteristics of process design options. These indices measure the inherent safety characteristics of processes in only two specific areas—fire and explosion hazards and acute chemical inhalation toxicity hazards. Other indices would be required to evaluate other types of hazards. Because the material inventory is considered in these indices, process intensification will result in a lower value (greater safety) for either index. However, the indices do also consider other factors, such as temperature and pressure, that


might be higher for an alternative process. The index is therefore valuable in helping to understand the relative value of a reduced inventory of hazardous material compared to other hazard factors that might increase. 7.3.

Overall Inherent Safety Index

Some initial work on the development of an overall inherent safety index has been done at Loughborough University in the United Kingdom (26). Other, similar work has been done at VTT in Finland (27). Both indices are considered prototypes by the developers, and more work is needed. These proposed inherent safety indices evaluate a number of factors related to inherent safety, including: Inventory Flammability Explosiveness Toxicity Temperature Pressure A single-number overall index characterizing the inherent safety of the overall process is generated by both proposed inherent safety indices. Process intensification will lower the value of the index (indicating an inherently safer process) because it will reduce the penalty for “inventory.” If the alternative process results in an increase in the inherent hazard due to other factors, the index will be useful in understanding the inherent safety characteristics of the different alternatives. The relative contributions of the various components of the index to the total value may also be useful in understanding process safety characteristics. Table 3 summarizes the application of this proposed inherent safety index to a number of alternative routes for the manufacture of methyl methacrylate.

TABLE 3 Evaluation of Alternate Methyl Methacrylate Processes Using the Edwards et al. Proposed Inherent Safety Index Process route Acetone cyanohydrin Ethylene based/propionaldehyde Propylene based Ethylene based/methyl propionate Isobutylene based Tertiary butyl alcohol based


Inherent safety index ⬃120 ⬃75 ⬃70 ⬃50 ⬃50 ⬃50

8. PROCESS INTENSIFICATION BENEFITS FOR PASSIVE AND ACTIVE LAYERS OF PROTECTION While a smaller process may not totally eliminate a hazard, it often can have the benefit of making effective passive layers of protection for the process more feasible and cost effective. Passive protection devices such as containment dikes, blast-resistant enclosures, and containment buildings to prevent the escape of toxic gas can be smaller. Active safety devices such as rupture disks, flares, and scrubbers will be reduced in size. Smaller processing equipment will also respond more quickly to other common safety interlock actions, such as shutting off reactant feeds, increasing cooling to a vessel, adding reaction shortstop agents, and emptying a reactor to a reaction quench tank. Some examples of the potential benefits of process intensification in the application of other safety features follow. Unstable materials such as explosives are sometimes manufactured in remotely controlled facilities protected by blast-resistant enclosures or bunkers. In this situation, if an explosion occurs the process equipment may be severely damaged or destroyed, but nobody will be injured and the environment and other property will be protected. This would be considered to be a passive safety feature— the blast-resistant enclosure does not require the action of any equipment or people to perform its function. While this type of blast-resistant enclosure may be practical for small equipment, it may become prohibitively expensive for large equipment. Clearly a larger structure would be required for larger equipment. However, the enclosure would also have to be much stronger because the potential explosion in the large vessel would be of greater magnitude. Process vessels are often protected from overpressurization by active devices such as relief valves and rupture disks. These devices open at a set pressure and allow gas and liquid to escape from the protected vessel to limit the pressure. Many years ago, the rupture disk or relief valve would simply discharge to the atmosphere. However, this is often unacceptable today because we cannot allow hazardous material to escape to the environment, even in an emergency situation. Emergency relief devices often must discharge to a collection-and-treatment system to ensure that no hazardous material is released to the atmosphere. This treatment system might include equipment such as vapor liquid separation devices, scrubbers, absorbers, flares, or thermal oxidizers. If the process vessels can be made smaller, the size of the required treatment system for the emergency relief discharge is correspondingly reduced. Total containment of the emergency relief device effluent in a large pressure vessel may even be possible, eliminating all discharge to the environment during the event (but the contents of the containment vessel will have to be treated and disposed of later on). For extremely exothermic reactions, emergency quench systems are sometimes used to protect against a runaway reaction. If the reactor temperature exceeds a predetermined maximum safe temperature, the reactor contents are rapidly


discharged to another vessel containing a material that will stop the reaction. These systems are more feasible for a small reactor. A reactor of a couple of cubic meters volume can be emptied to a quench tank in a few seconds through a largediameter dump valve, but it might take many minutes to discharge the contents of a large reactor. It may not be possible to empty the large reactor quickly enough to prevent the runaway reaction. 9.

SUMMARY

Safety considerations are an inseparable part of the development of a chemical process and the design and operation of a chemical plant. While risk management and safety features can be added on to a plant design or to an operating plant, safety is most reliably and robustly ensured by developing inherently safer processes. Safety strategies can be categorized as inherent, passive, active, and procedural. Inherent and passive strategies generally relate to the basic process technology and plant design and are nearly always implemented early in the design life cycle. They focus on elimination of hazards or minimizing the degree of hazard rather than on management of hazards. Process intensification is an important approach to the development of inherently safer chemical processes because it reduces the quantity of hazardous material in the process, thereby reducing the inherent risk. Active and procedural strategies are usually also a part of a chemical process risk management program—it is not often possible to eliminate all hazards. Process intensification can also make active and procedural safety features more effective and economical. Safety equipment can be made smaller and less costly. It may be feasible to use safety devices to protect against the hazards from small processing equipment that are impractical for use in a large plant. The faster response time of small equipment may allow effective automatic or manual intervention to detect an incipient problem and take action to prevent it from developing into a serious accident. Chemical process safety cannot be viewed in isolation from other process and plant design criteria. The chemical plant must meet many requirements for workers (safety, long-term health, employment and wages), owners (operating costs, capital investment, profitability), customers (product quality, reliability of supply, cost), neighbors (safety, health, environmental impact, economic impact), and government (compliance with laws and regulations). All of these are important, and they may be in conflict. The chemical process designer must work to select the optimum design that considers all stakeholders. Process intensification is an important approach to minimizing the hazards associated with chemical handling and manufacture and will be an important factor in the future for designing safe, environmentally friendly, and economically competitive chemical plants.


REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

12. 13. 14. 15. 16. 17.

18. 19. 20. 21.

American College Dictionary. New York: Random House, 1967. Rolt LTC. The Railway Revolution: George and Robert Stevenson. New York: St. Martin’s Press, 1960:147. Kletz TA. What you don’t have can’t leak. Chem Ind 1978; May 6, 287–292. Center for Chemical Process Safety (CCPS). Guidelines for Engineering Design for Process Safety. New York: American Institute of Chemical Engineers, 1993:5–52. Center for Chemical Process Safety (CCPS). Inherently Safer Chemical Processes: A Life Cycle Approach. New York: American Institute of Chemical Engineers, 1996. Kletz TA. Process Plants: A Handbook for Inherently Safer Design. Bristol, PA: Taylor and Francis, 1998. Perry RH, Green DW, eds. Perry’s Chemical Engineers’ Handbook, 7th ed. New York: McGraw-Hill, 1997: 26–5, 26–6. Center for Chemical Process Safety (CCPS). Layer of Protection Analysis. New York: American Institute of Chemical Engineers, 2001. Lees FP. Loss Prevention in the Process Industries, 2nd ed. Oxford, UK: ButterworthHeinemann, 1996:A5/1–A5/11. Post RL, Hendershot DC, Kers P. Synergistic design approach to safety and reliability yields great benefits. Chem Eng Prog 2002; 98(1):60–66. Wade DE. Reduction of risks by reduction of toxic material inventory. In: Woodward JL, ed. Proceedings of the International Symposium on Preventing Major Chemical Accidents. February 3–5, 1987, Washington, DC; New York: American Institute of Chemical Engineers, 1987:2.1–2.8. Dow Chemical Company. Dow’s Fire and Explosion Index Hazard Classification Guide. 7th ed. New York: American Institute of Chemical Engineers, 1994. Hendershot DC. Process minimization: making plants safer. Chem Eng Prog 2000; 96(1):35–40. Dow Chemical Company. Dow’s Chemical Exposure Index Guide. 1st ed. New York: American Institute of Chemical Engineers, 1994. Wilkinson M, Geddes K. An award-winning process. Chem Br 1993; (December): 1050–1052. Whiting MJL. The benefits of process intensification for Caro’s acid production. Trans IChemE 1992; 70(March):195–196. Osterwalder U. Continuous process to fit batch operation: safe phosgene production on demand. Symp. Pap.—Inst. Chem. Eng., North West. Branch. Rugby, U.K.: IChemE, 1996:6.1–6.6. Delseth R. Production industrielle avec le phosgene. Chemia 1998; 52(12):698–701. Kelleher T, Fair JB. Distillation studies in a high-gravity contactor. Ind Eng Chem Res 1996; 35:4646–4655. Koch TA, Krause KR, Mehdizadeh M. Improved safety through distributed manufacturing of hazardous chemicals. Process Safety Prog 1997; 16(1):23–24. Center for Chemical Process Safety (CCPS). Guidelines for Use of Vapor Cloud Dispersion Models. 2nd ed. New York: American Institute of Chemical Engineers, 1996.


22. 23.

24.

25. 26.

27.

Center for Chemical Process Safety (CCPS). Guidelines for Consequence Analysis of Chemical Releases. New York: American Institute of Chemical Engineers, 1999. The Netherlands Organization of Applied Scientific Research (TNO). Methods for the Calculation of the Physical Effects of the Escape of Dangerous Materials: Liquids and Gases (The Yellow Book). 3rd ed. The Hague, Netherlands: Sdu Uitgevers, 1997. The Netherlands Organization of Applied Scientific Research (TNO). Methods for the Determination of Possible Damage to People and Objects Resulting from Releases of Hazardous Materials. The Hague, Netherlands: Sdu Uitgevers, 1992. Gowland RT. Applying inherently safer concepts to a phosgene plant acquisition. Process Safety Prog 1996; 15(1):52–57. Edwards DW, Lawrence, D. Assessing the inherent safety of chemical process routes: is there a relation between plant costs and inherent safety? Trans IChemE B, 1993; 71(November):252–258. Heikkila A. Inherent Safety in Process Plant Design. Espoo, Finland: Technical Research Centre of Finland (VTT), 1999.


14 Process Intensification Contributions to Sustainable Development G. Jan Harmsen, Gijsbert Korevaar, and Saul M. Lemkowitz Delft University of Technology, Delft, The Netherlands

This chapter is meant for developers and designers of processes, in particular of intensified processes, who want to ensure that their processes contribute to sustainable development. To this end problems with the present technological solutions in relation to the world-scale environment and society are explained in the first section. Then in the second section sustainable development is explained and the role and criteria for sustainable technology are derived. Finally process intensification is assessed on its contribution to sustainable development. The information presented should be sufficient for a developer or designer of a process to raise awareness of design aspects related to sustainable development, to yield hints to modify the design in the direction of sustainable development, and to find references for further detailed information. 1. 1.1.

PROBLEMS LEADING TO SUSTAINABLE DEVELOPMENT Environmental Problems

The disturbance of several natural ecological cycles became a problem with the start of the Industrial Revolution in the 19th century. The emissions to air, water,


and soil increased exponentionally. In the 20th century these changes in the environment reached global proportions: Increases of concentrations of “greenhouse” gases in the global atmosphere occurring at rates and to levels higher than those ever reached during the previous thousands of years (1). Massive deforestation occurring faster than at any time in human history (2). Species extinction and loss of biodiversity at rates higher than 10,000 per year (3). The publication of the report Limits to Growth (4) by the Club of Rome had a major impact on thinking about the environmental impact of our cultural development. Under the assumption that the five basis elements of this study— population, the production of food, industrialization, pollution, and the use of nonrenewable resources—will keep increasing exponentially, they showed that, if unchanged, this would lead to enormous problems, as soon as the 21st century. The social consciousness of the problems caused by unlimited growth of these elements was greatly increased by this report by the Club of Rome. In the present situation, one can say that natural resources and the environment can no longer be regarded as inexhaustible reservoirs of products and services. On the contrary, in general it is now recognized that natural resources are limited, that the natural environment has a limited capacity, and that this must have consequences for the way we act. 1.2.

Socioeconomic Problems

In addition to adverse changes to the world environment, economic differences between rich and poor are increasing (decreasing “equity”), both within nations and between nations. The richest fifth receives 83% of the total world income (5). Moreover, the gross domestic product (GDP) per capita stays more or less the same for poor countries. This unfairness is also a part of unsustainability, because it blocks possibilities toward worldwide cooperation. It limits the real growth of the intrinsic quality of life, which is wanted by every human being thinking about his or her children’s future. It also leads to extensive illegal immigration of the poor to the richer parts of the world. Also, all parts of society contribute to the problem. For instance, carbon dioxide emissions to the atmosphere due to fossil fuel combustion is caused 50% by industry; the other parts are domestics and transport (the electric power contribution is divided between industry and domestics) (6,7). 1.3.

Present Technology Is Part of the Problem

Since awareness arose of the large-scale environmental, resource, and socioeconomic problems, several technological solutions have been implemented. However,


they did not solve the problems and in some cases caused even bigger problems, as is shown in the cases described later. 1.3.1.

High Chimneys for Acidic Flue Gases

In the 1960s a large number of cities in western Europe suffered from smog formation. One of the major causes was the local emission of acidic flue gases from large industrial areas. Under still weather conditions, these acid components would remain in the atmosphere near the earth’s surface and form smog. The technological solution was to build very high chimneys—over 200 meters high—cutting through the atmospheric inversion layer. The flue gases where dispersed in this way over a larger area. In the decade that followed, it appeared that in more rural areas, notably Sweden, these acid gases caused acidification of forests and lakes, with negative effects on their biological quality. Trees and fish died on a large scale. 1.3.2.

CFC Use

In the 1960s and ’70s, CFCs were used on a large scale as propellant gases in spray cans and as a refrigerant in refrigerators. These gases were considered inert to humans and the environment. Hence, release into the atmosphere was considered acceptable. In 1973, however, Molina discovered that these CFCs broke down the ozone layer high up in the atmosphere (8). This ozone layer itself absorbs the ultraviolet radiation of sunlight and in this way prevents ultraviolet radiation from reaching the earth’s surface and thereby harming humans (skin cancer formation) and other living species (algae in oceans). Due to the slow breakdown of CFCs the effect on the ozone layer is expected to last for centuries (9). 1.3.3.

HFC Replacement of CFC for Refrigeration

Due to the negative effect of CFCs on the ozone layer, a replacement was developed: HFC. This component hardly breaks down the ozone layer. It is now replacing CFCs in domestic refrigerators and air conditioners in cars on a very large scale. However, HFC has a global warming gas potential that is a factor of 1000 higher than carbon dioxide. The present amount of HFC in the atmosphere is estimated to account for a global warming effect of 0.6C (10). 1.3.4. MTBE in Gasoline as a Replacement for Lead Components In gasoline, lead components are used to enhance the octane number. It became clear that these lead components, once emitted from car exhaust pipes, cause brain damage. California was the first to ban these lead components and to stimulate the use of MTBE as a replacement. MTBE both acts as an octane booster and enhances the gasoline’s combustion, by which less volatile organic carbon components are emitted from the car exhaust pipes. However, it appeared that MTBE entered underground water reservoirs by occasional spillage of gasoline. This made the water unfit for drinking due to its bad smell. In California billions


of gallons of drinking water are already contaminated in this way, and legislation is under way to ban the use of MTBE in gasoline. 1.3.5.

Brent Spar Oil Platform in the North Sea

In the 1970s, the awareness of the limited oil resources resulted in a sudden oil price increase by a factor of 4. It became economical to produce crude oil from sea bottoms. One of the first oil platforms in the North Sea was the Brent Spar. In the 1990s, oil production ceased and the platform had to be demolished. In the Brent Spar design, demolition had not been taken into account, and so the demolition appeared to be very difficult. In consultation with the British government it was decided that dumping in a deep trough in the ocean would be the best economical and environmental solution. The environmental activist organization Greenpeace disagreed with this solution, arguing that the problem was not the dumping of the Brent Spar; the real issue, they held, was the principle of dumping in the ocean in general and that the Brent Spar dumping would be used by other companies (nuclear energy companies and others) to claim the same right of dumping. Greenpeace started a campaign to prevent this dumping by asking the general public to boycott Shell gasoline purchases. Especially in Germany this boycott had a noticeable effect, and Shell decided to reverse it decision and to invite any company or organization to propose alternative solutions. In the end, it was decided to bring the platform to a fjord and dismantle it there in such a way that parts could be reused in other applications. 1.3.6.

Preliminary Conclusions on Technology Cases

All the technological solutions mentioned already have in common that the problem to be solved was defined too narrowly in space, time, and lifecycle phases. Wider-ranging and longer-term effects on the local and global scales on humans, nature, and ecology were not taken into account. 2. SUSTAINABLE DEVELOPMENT AND REQUIRED TECHNOLOGY 2.1.

General Concept of Sustainable Development

The general opinion about the aforementioned problems and the ad hoc technical solutions is that they are by no means sufficient. The essential concept when thinking about these subjects nowadays is sustainability. It was greatly helped by the description and definition by the Brundtland report of the World Commission on Environment and Development (WCED), Our Common Future (11): Sustainable development is not a fixed state of harmony, but a process of change in which the exploitation of resources, the direction of investments,


the orientation of technological development and institutional change are made consistent with future as well as present needs. Sustainable development is development that meets the needs of the present generation without compromising the ability of future generations to meet their own needs. This definition contains two key concepts: The concept of needs, in particular the essential needs of the world’s poor, to which overriding priority should be given The idea of limitations imposed by the state of technology and social organization on the environment’s ability to meet present and future needs (12). The most admirable aspect of this report is the attempt to bring the two largest world problems—unlimited exploitation of nature and growing inequity within and among nations—together in one concept that everyone from government to citizen can work on. Two very important remarks have to be made about the definition of the Brundtland Commission. 1. The definition makes clear that development of new technologies, social structures, or whatever has to take into account economic and social issues ( present generations) and long-term and large-scale environmental issues ( future generations). Thus developments that have to lead to sustainability are limited and have to consider the idea that every human being must be able to fulfill his or her needs in a more or less equal way. 2. We have to distinguish between two different kinds of development: one that leads to technological innovation, as used in the term research and development, and one that is about improving the welfare of a society, as used in the term developing countries. This distinction is very important, because it makes clear that sustainable development has different meanings in different countries. In a rich country, sustainable development sets limits to the growth of affluence or the possibilities of new and innovative technologies. In poor countries, sustainable development has to do with helping the population to survive and if possible to thrive. To further understand what the commission meant by this definition it is necessary to read the whole book, because a lot of discussion occurred within the commission about poverty and affluence, distribution of knowledge and information, the rights of the northern part of the world to live more affluently than the southern part, the right to develop, etc. Here we use the summary by Installé (13): Set priorities among the needs. Ever-lasting growth is unsustainable, because large parts of earth’s natural capital are not substitutable by human-made capital.


The actual world economy rules are unsustainable. The existence of various cultures should be encouraged. Intergenerational equity results in a maximum well-being for ourselves under the constraints that the well-being of future generations must be at least equal to ours. 2.1.1.

Ideas Complementary to Brundtland

The definition and vision of the Brundtland Report are described more practically by three other United Nations Commissions, the International Union for Conservation of Nature and Natural Resources (IUCN), the United Nations Environment Program (UNEP), and the World Wide Fund For Nature (WWF), called Caring for Earth: A Strategy for Sustainable Living (14). This report defines sustainable development as: “Improving the quality of human life while living within the carrying capacity of supporting ecosystems.” This report contains nine chapters, with the following titles, that also can be seen as nine issues of sustainable development. 1. 2. 3. 4. 5. 6. 7. 8.

Respecting and caring for the community of life Improving the quality of human life Conserving the earth’s vitality and diversity Minimizing the depletion of nonrenewable resources Keeping within the earth’s carrying capacity Changing personal attitudes and practices Enabling communities to care for their own environments Providing a national framework for integrating development and conservation 9. Creating a global alliance 2.1.2.

The Three Parts of Sustainable Development

So far the ideas of sustainable development have been described mainly in global political terms. The financial and business world also took hold of the concept and reformulated it into three parts: social, ecological, and economic (15,16). This in turn was translated into a catch phrase: People, planet, and profit (17). We will summarize these parts. 2.1.2.1 Ecological Part (Planet). The ecological part has to do with the impact of human action on nature. Mainly this means all the known environmental problems and processes that disrupt the ecosystems (ozone depletion, acidification, greenhouse effect, destruction of species, wastes, etc.). In a sustainable world all those known problems must be minimized or avoided. In addition, for as far as possible, sustainable development must have the power to avoid new problems. The precautionary principle is therefore adopted.


Precaution principle: At the United Nations Conference on Environment and Development (UNCED) in Rio de Janeiro the (nonbinding) Rio Declaration of 1992 was agreed on, with the following text: Principle 15. In order to protect the environment, the precautionary approach shall be widely applied by States according to their capabilities. Where there are threats of serious or irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation. This is a better-safe-than-sorry principle that advocates the reduction of inputs into the environment of substances especially where there is reason to believe that harmful effects are likely to occur (18). Therefore, the precautionary principle has to play an important role in decisions about new technology. 2.1.2.2 Social Issues (People). The sustainable development of society concerns needs. This means that consumer participation is very important and that the creation of new markets has to be done with caution. Further, concerning the lasting factor of societies, it is essential that sustainable development be in agreement with local culture. 2.2.2.3 Economic Issues (Profit). It is obvious that if scarce nonrenewable resources such as fossil fuels and rich metal ores are going to be depleted, prices will become extremely high. Therefore, economics is very important in sustainable development. This is also true for the feasibility of sustainability concepts: If they are not profitable, then they will not be accepted. However, it is very important to think over the meaning of the term profitability. Indeed, if external costs are taken into account, sustainable technology will be more profitable in the long term than conventional technology. External costs is an economic concept, which gives the possibility of setting the prices right. External costs are the cost involved by the rest of the lifecycle of products and by-products and not accounted for in present cost calculations. For example, the price of fossil fuels depends not only on exploration costs but also on the costs involved by the production of CO2 and resultant adverse climatic effects, which will be many trillions of dollars. In an article in Nature the value of the world ecosystem services is estimated to be worth 33 1012 $ per year (19). This means that for economic activities that destroy part of the ecosystem, the ecovalue (service) reduction should be accounted for. These external cost can be considerable compared to the present calculated cost. The present global economic production expressed as GNP is estimated at 18 1012 $ per year; so the world ecosystems produce a factor 1.8 higher annual value (19). This means that the external cost can even exceed the present cost when the ecosystem is seriously affected by human activity.


2.1.3.

Other Views on Sustainable Development

So far the mainstream thoughts about sustainable development have been described. However, there are other voices, which have a different view on sustainable development and critisize (part of) the Brundtland description. Their general comment is that the Brundtland description stems from an anthropocentric worldview that places man too much at the center. Two other views are the ecocentric and the theocentric Judeo-Christian worldviews. A brief description of each follows, together with their main criticism on the Brundtland view. 2.1.3.1. Ecocentric Worldview. In this view nature has a value of its own. It places nature at the center and man as a part of nature. Man should not have a large impact on nature. All biological species should be maintained. Technology should fit into and depend on nature. Nature should be kept in a steady state. If this means that human needs are to be restricted, then that is accepted. To quote Gandhi: “The world provides sufficient for everybody’s need but not for everybody’s greed.” People who have this worldview often criticize the Brundtland description because it sees nature mainly as a source for goods and services for mankind (20). 2.1.3.2. Theocentric Judeo-Christian Worldview. In this view, nature is seen as God’s creation. Man is mandated to be God’s steward for this creation. A typical Bible text in this respect is: “to work it and to take care of it” (Bible, Genesis 2:15). Also, in this worldview people should take care of their neighbors and in particular take care of the weak and the poor. Furthermore, in this view man should restrain himself in his needs. He should not be greedy. A comprehensive description is given by Elsdon (21). Some criticism of people with this worldview on the Brundtland description is that it sees nature as man’s possession and that it does not mention a restriction on man’s greed. 2.2.

Sustainable Technology Positioning

In the last 30 years, environmental impact has been mainly reduced by end-ofpipe solutions. These solutions always mean that more construction material and energy is required; hence they reduce the environmental impact for a certain component but often increase the impact for another component and are always more expensive. Then the next concept arrived: clean production technology. Jackson (1993) defines clean production as “a conceptual and procedural approach to production that demands that all phases of the lifecycle of a product or a process should be addressed with the objective of prevention or minimization of short- and longterm risks to human health and to the environment (18).”


With this type of technology, products and processes are reformulated to manage emissions and waste “upstream.” These solutions can mean both less emission and less cost and energy. In these views of technology, the social part and the economic parts of sustainable development are not taken into account. In the next section we will derive a general view on sustainable technology, first by showing all interactions with the sustainable development aspects and then some specifics for each sustainable development part. 2.2.1. Social, Economic, and Ecological Interactions with Sustainable Technology Providing for the needs of society while staying within ecological constraints is not only a task for technology, but also the result of the interaction of culture, political structure, and technology; (22). Culture determines the size and the nature of the needs for which fulfillment is justified and the conditions that technology and structure have to satisfy. Structure is the way in which fulfilling the needs is organized by means of production and consumption. It has to do with organization, economy, and policy. Technology is not only the total of means available for fulfilling the needs; technology itself also has an influence on the culture. The latter influence is difficult to predict but can be discovered from the past. For instance, the development of lightweight, low-cost plastics led to packaging of small-portion consumer items. This in turn enhanced the development of self-service supermarkets, which led to shopping by car rather then walking to nearby small shops (23). This contributed to a more individual culture. Presently the whole infrastructure of a developed country is based on auto transport. The road is no longer a meeting place of persons but a depersonalized area, again leading to a more individualistic society. Stainer et al. give some guidelines for the different actors in a technologybased culture (24): Engineers: Be aware of the implications for society and environment. Nonscientists: Understand applied science. Companies: Sustainable growth is allied to risk management. Politicians: Develop the skills for sustainable policy. All groups: Recognize the need for communication. A very interesting concept developed by Moser is the principle of invasiveness in relation to embeddedness of technology (25). This concept makes it clear that current technologies have a high impact on the environment (invasiveness) but that their rapid development and the high public concern about their impact are not very well anchored in society; this means a low embeddedness. New, cleaner technology, and especially sustainable technology, has to be developed so that they will not need to change very fast and be more strongly embedded in society.


All this is said to make clear that something has to change drastically in the focus about using technology if we want to speak about sustainable technology. In sustainable technology, technology is still serving humans and their society (otherwise, we need not speak about technology at all!). However, the technology should also be the servant of nature, both in an active and a passive way. 2.2.2. Ecological Impacts as a Function of Cultural, Economic, and Technological Activities The ecological impacts caused by human activities are large. To give more insight into the effects of economy, culture, and technology on ecological impacts, the following semiquantitative expression is often used. It is the so-called “master” equation. This expresses the environmental impact as a product of population size (P), gross domestic product per person {GDP/P} and environmental impact per unit GDP {EI/GDP} EI P * {GDP/P} * {EI/GDP} Often the following substitution is made: GDP/P W (W meaning “wealth”) EI/GDP T (T meaning “technology”) resulting in EI P * W * T This master equation can then be used to calculate the increase in environmental impact for the population, or the increase in wealth with the same technology, or to calculate the required emission reduction per unit GDP by new technology. If, for instance, wealth worldwide reaches the same level as now in the industrialized world, then W increases by a factor of 4. If the population (P) does not increase and technology (T) stays the same, then EI (e.g., carbon dioxide emissions to the atmosphere) would increase by a factor of 4. To keep EI at the same level, the technology should be improved to give an emission reduction of a factor 4 per unit product or service. In reality this use of the expression is too simplistic: Population growth is a function of the level of wealth. In Western society the natural population growth is less than zero (the birth rate is less than 2 per woman). The same could happen in developing countries once a certain level of wealth has been reached. Replacing GDP/P by wealth is also under debate. All kinds of occurrences, such as car accidents leading to medical care consumption, are adding to GDP, while these things do not add to wealth at all. Corrections to the GDP for these aspects have already been proposed by (26). There is a trend that beyond a certain wealth, the environmental impact emission per unit capita is reduced. Jackson shows that on average above


5000 GDP per capita, the hazardous waste per unit GDP is reduced (18). Hence the level of W also has an effect on EI/GDP. The environmental impact per unit GDP (T) is not only a function of technology but also a function of human behavior and societal culture. Jackson shows that countries with the same GDP per capita show large differences in hazardous waste per unit GDP. For instance, Scandinavian countries, compared to Canada and the United States show, a similar GDP per capita but up to a factor 4 difference in hazardous waste per unit capita (18, p. 115). This difference is due to differences in culture. Scandinavian countries value nature more than does the United States. A good source for further reading on this aspect of wealth increase and impact reduction is von Weizsäcker et al. (27). 2.2.3.

Transition to Sustainable Technology System Levels

Now when we talk about the transition to sustainable technology it is helpful to understand that technology has to deal with different system levels, as shown in Table 1. It is obvious that changes in societal infrastructure are more complex and require more disciplines and more stakeholder’s involvement. These changes are on a very long time scale (decades to centuries). Scenario building is an important part of the methods to robust long-term developments (28). This field is too big to elaborate further in this chapter. Changes to new, sustainable products or new processes, fitting into existing infrastructures, are less complex and can be achieved on a shorter time scale. The system levels, from process downward, will be further discussed later.

TABLE 1 System Scale Levels and Technology Disciplines System scale levels World biotic and abiotic Societal infrastructure

Industrial complex Process Unit operation Equipment Catalyst/dispersed entity Nano Molecules

Major disciplines involved Ecology, politics, physics, chemistry Politics, social science, economics, laws, ecology, civil engineering, city and landscape architecture Civil engineering, economics, law Chemical engineering Chemical engineering Chemical engineering, mechanical engineering Chemistry, physics, chemical engineering Chemistry, physics, engineering Chemistry, physics


2.3.

Sustainable Technology Scorecard

To aid the developer or designer of a new technology the author has derived a scorecard to assess the technology on its contribution to sustainable development. It is based on the sustainable development aspects: social, ecology, and economy. The scorecard is shown in Table 2. The basis and use of the scorecard will now be explained.

TABLE 2 Sustainable Technology Scorecard Sustainable development item Social/people Provide for the needs of the poor: water, food, clothing, local energy, etc Fair distribution of wealth, power, and knowledge Social acceptance Safe Noise Smell Occupational health

Plot area impact Skyline impact Ecological/planet Sensitivity to world-scale nature and ecology Depletion of abiotic resources Depletion of biotic resources; biodiversity Dehydration Ozone depletion potential (ODP) Global warming potential Photochemical ozone pollutants


Comment

Prime goal

Adaptable, nondisruptive to society By stakeholder engagement Loss prevention at all conditions Below legal limits No obnoxious emissions Long-term effects of exposure to chemicals should be known and acceptable Low; important in densely populated areas Low aesthetic and bird friendly No emission of components whose ultimate environmental fate is unknown Keep air, surface water, and soil healthy Maintain biodiversity Maintain water reservoirs No ODP gas emissions Green house gas emissions reduced by factor 4 Volatile organic component emissions below expected future legal limits

TABLE 2 (cont.) Sustainable development item Acidification

Human toxicity

Ecotoxicity (terrestrial and aquatic)

Nutrification (eutrophication)

Radiation

Thermal pollution

Waste

Economy/profit Scarce resource depletion Drinking water resource depletion Fossile fuel depletion External (future) cost Capital expenditure and decommisioning cost Operational cost long term Profitable over total lifecycle

2.3.1.

Comment Volatile organic component emissions below expected legal limits Volatile organic component emissions below expected legal limits Volatile organic component emissions below expected legal limits Volatile organic component emissions below expected legal limits Volatile organic component emissions below expected legal limits Volatile organic component emissions below expected legal limits Volatile organic component emissions below expected legal limits

future

future

future

future

future

future

future

No full depletion; replacement by renewables No depletion No full depletion; replacement by renewables Include in present accounting

Social/People

A prime goal of sustainable development on a world scale is to provide for the needs of the poor. Throughout one should keep in mind that developed economies have different challenges than survival economies. For survival countries the main goal is to provide for the needs of the people in an affordable way while not consuming natural capital (29). Hence this is an important item on the


scorecard. One can think of technology providing water, food, clothing, and clean energy for cooking and housing while adapting to the local social, economic (resources), and ecological situation (30). Another item on the scorecard requiring some explanation is social acceptance. In the previous sections it is shown that sustainable technology is to be embedded in society. This means, in our view, that society should be considered in the design stage. A discussion with representatives of stakeholders should be held to clarify the effects of the new technology on the ecology, society, and economy and to seek acceptance. Stakeholders can be environmental activists, civilians, government representatives, manufacturing companies, and consumers. These stakeholders often have different worldviews. In stakeholder discussions this is often not recognized. The previous section on worldviews in relation to sustainable development should help to understand the background of certain arguments used in stakeholder discussions. Scholes reveals that such a stakeholder discussion was carried out by Shell in the design phase and led to changes in the design and to greater acceptance of the new process (31). 2.3.2.

Ecological Assessment

The precautionary principle should be applied. This means that for all emissions over the whole lifecycle, the final environmental fate needs to be known. If this knowledge is not available, the emission of that component should be kept at zero. All other scorecard items stem directly from lifecycle analysis assessment theory (32). For some impact types, quantitative norms can be stated. For stratospheric ozone layer depletion components, the emission norm is zero. This means in practice that components such as CFCs and SF6 should not be used at all, not even in contained technical applications, because containment over the lifecycle cannot be ensured. For other emissions, such as potential global-warming gases, the norm will depend on the time horizon of the application. If the application is foreseen for a long time horizon, say, 50 years, then the global-warming gas emissions should be reduced by a factor of 4–10 compared to the present (27). To be able to assess the environmental impact of a technology over the whole lifecycle, a lifecycle assessment (LCA) should be performed. A brief description of LCA follows. Life cycle assessment is a method for calculating the environmental impact of a product or service over its whole lifecycle. Steps in a lifecycle include: Exploration Refining Manufacturing


Use End of product life (i.e., combustion, reuse, or recycle) Process decommission A process (manufacturing) is therefore part of a product recycle. Often the decommissioning phase of the process plant is not taken into account in the LCA. As, for instance, happened in the Brent Spar case and for most nuclear power plants. An LCA consists of the following stages: 1. 2. 3. 4. 5.

Goal definition and scoping Inventory analysis Impact assessment (classification) Valuation Improvement

1. Goal definition and scoping defines the purpose of the LCA and its scope, the latter meaning depth and broadness. This is important because LCA can be carried out in great detail, requiring great input of time and money. When the goal is comparing alternatives it is sufficient to gather knowledge that is not absolutely correct but good enough to make relative comparisons. Scope involves selecting system boundaries—what and what not to include. For example, infrastructure may include the plant that makes a product but also the whole transportation system (e.g., roads, their construction and maintenance) to distribute it. Limiting scope is essential for practical execution of LCA studies. 2. The inventory analysis determines the material and energy inputs and outputs relating to one functional unit of product or service and links these to environmental impacts. This stage is simply a material and energy balance, albeit a complete one, that is, covering all lifecycles of the product or service being considered and relating to all inputs and outputs of all of these lifecycles, including even micro-outputs, such as (when relevant) micro-emissions of dioxins down to milligrams per year for plants, with total emissions in megatons per year (e.g., power plant). 3. Impact assessment translates the inventories into recognized environmental impacts and calculates a value for each of these impacts. Generally recognized environmental impacts are given in Table 3. Assigning an emission of a given substance a given environmental impact depends on the properties of the given substance. Emitted sulphur dioxide, for example, is converted into sulphuric acid and is therefore classified as an “acidification” impact. Emitted carbon dioxide and nitrous oxide are greenhouse gases and are therefore classified as “global warming” substances. Once emissions have been classified into a given environmental impact, they must be assessed. This means calculating the quantitative value of this impact. Environmental impacts are calculated by means of conversion factors, or


TABLE 3 Generally Recognized Environmental Impacts Depletion Abiotic resources Biotic resources

Pollution Depletion of stratospheric ozone layer Global warming Formation of photochemical oxidants and pollutants Acidification Human toxicity Ecotoxicity (terrestrial and aquatic) Nutrification (eutrophication) Radiation Thermal pollution (dispersion of heat) Noise Smell Occupational health

Disturbances Desiccation (dehydration) Physical ecosystem degradation and landscape degradation Human victims

so-called equivalency factors. We illustrate this by using the impact category “global warming.” Conversion factors for global warming potential (GWP) are expressed in terms of CO2 equivalency (kg CO2 equivalent), as given in Table 4. Using GWP values, emissions to the air of various substances can be converted to an equivalent CO2 global-warming effect by means of the following formula: n

Global - warming effect (kg) ∑ (Emissions to air)i (GWP)i i1

Example: The GWP of a factory that per year emits 1 megaton of CO2, 10 kilotons of CH4, and 1 kiloton of N2O is 3(1 1 megaton 0.01 62 0.001 290) 1.352 megatons CO2 equivalent per year Other environmental impacts can be calculated similarly by using the appropriate equivalency factor. Impacts of human toxicity and ecotoxicity are, for example, determined by factors such as toxicity data (for humans, resp. various nonhuman life forms), persistence, and bioaccumulation. Equivalency factors for


TABLE 4 Global Warming Potential (GWP) for a Number of Substances (32) Substance

Substance formula

Carbon dioxide Methane Nitrous oxide CFC11

CO2 CH4 N2O CFCl3

GWP (kg CO2/kg substance) 1 62 290 5,000

Source: Hauschild, 1998, vol. 1.

a wide range of substances can be found in the literature (32) and for industrial chemicals (33). 4. Valuation. In the valuation step, all impacts are multiplied by norm factors. The resulting figures are added up to yield a single total figure. The norm factors are subjective, for they express the relative importance one gives to totally different environment effects. This subjectivity can be made more acceptable by having the norm factors set by a panel with representatives of various worldviews or political parties. 5. Improvement. One of the great uses of LCA is to pinpoint places in the lifecycle that cause major environmental impacts and thus lead to highly efficient improvement of the impact spectrum for a given product. The LCA results can be used for a sustainability assessment of a new technology by comparing the outcome with the existing technology results. In the conceptual phase of the new technology, a relative assessment can be used to highlight where improvements should be made and where the problem areas are. This is shown in Section 3 on process intensification assessment. Detailed Information on the LCA method is provided by Hauschild (32). 2.3.3.

Economy/Profit

It is obvious that if scarce nonrenewable resources such as fossil fuels and rich metal ores are going to be depleted, prices will become extremely high. Which means that the next generation will experience higher costs for the same goods, which is not fair. If scarce resources are depleted, then developments should also start to provide alternatives at a similar price, so-called “strong sustainability economic development” (2). Obvious scarce resources—water and fossil fuels— are explicitly put on the scorecard. The future high resource cost and the external cost related to emissions should also be taken into account in the total lifecycle cost. Long-term profit is also placed on the scorecard. If a technology is not profitable it will not be accepted by business. However, it is very important to think


over the meaning of the term long-term profitability. It is recommended that the external costs, as far as they are known, be included in the cost calculations, for in the long term these external costs will likely become internalized via taxation or otherwise. Some external cost figures can be obtained from the literature. Table 5 gives the values for carbon dioxide and volatile organic components. The estimation of external cost is a scientific field in development. To find more data, the reader could search the literature on “external cost” or “future cost.” 2.3.4.

Use of the Scorecard in the Conceptual Design Phase

In the conceptual design phase a quantitative assessment requiring lots of information is not desirable. A qualitative assessment highlighting the main improvements and areas of major concern requiring alternative designs is more suitable. The scorecard can be used qualitatively, for instance, by estimating whether the new technology improves individual scorecard items as compared to existing technologies. Note: There is a tendency to arrive at a single sustainability score. In the author’s opinion little is gained and a lot is lost by the use of a single figure. What is lost is the information about where the problems with the new technology are located. This in turn means that suggestions for improvements cannot be found. Also, the danger of balancing poor-scoring areas with very positive areas becomes real. A sustainable technology should score positive on all three legs (social, ecology, economy) and should have no significant negative score in any impact area. 3. THE POTENTIAL FOR PROCESS INTENSIFICATION TO CONTRIBUTE TO SUSTAINABLE DEVELOPMENT Elaborate descriptions of process intensification will be found elsewhere in this book. Here a short description suffices, followed by an assessment on its potential to contribute to sustainable development, based on present industrial cases and general features.

TABLE 5 Estimates of External Cost for Two Components Relevant to Chemical Processes Emission type

External cost

Carbon dioxide net emission to atmosphere Volatile organic carbon emission to atmosphere

$46/ton (34) $8000/ton (34)

Source: SIKA, 1995.


3.1.

Definition of Process Intensification

Process intensification consists of the development of novel apparatuses and techniques that, compared to those commonly used today, are expected to bring dramatic improvements in manufacturing and processing, substantially decreasing the ratio of equipment size to production capacity, energy consumption, or waste production and ultimately resulting in cheaper and more sustainable technologies (35). Stankiewicz and Moulijn further divide PI into two areas: equipment and methods. They give an extensive list of equipment examples. The methods are subdivided into multifunctional reactors, hybrid separations, alternative energy sources, and other methods. For multifunctional equipment, Siirola provides a general description of functions and how to combine them into a piece of equipment (36). Functions operating in the same temperature and pressure range can be combined in one piece of equipment. Multifunctional reactors represent one subclass of solutions with promising features in cost and energy requirements (37). 3.2. Present Status of Process Intensification from Industrial Cases We now apply the criteria for sustainable technologies derived in Section 2, Table 2, to industrial PI cases. Table 6 shows the results. A short description of the cases follows. Italic numbers in the table are derived from the references, as is explained in the following text. 3.2.1. Industrial Cases of Process Intensification: Background for Table 6 3.2.1.1. Case E: Eastman Chemical Methyl Acetate Process (Siirola, 1995, 1998). In the commercial operation of the Eastman methyl acetate process, many reaction and separation functions are combined in one single large column (a few meters in diameter and 80 meters high (38). The number of pieces of major equipment compared to a conventional design is reduced by a factor of 10 (36). The primary energy consumption and capital expenditure is reduced by a factor of 5 (38). The plot area of this process is probably a factor of 10 smaller than the conventional process, because the number of major units has been reduced by a factor of 10. The major contributions to global warming, acidification, thermal pollution, and fossil fuel depletion are probably all directly related to the energy required. Because this is reduced by a factor of 5, it is assumed that these contributions reduce by the same factor, 5.


TABLE 6 Sustainable Technology Scorecard Results for Industrial PI Cases Improvement potential (reduction factor) Sustainable development item

E

S

SG

G

SP

Range

Social/people Provide for the needs of the poor: water, food, clothing Fair distribution of wealth, power, and knowledge Social acceptance Safe: lower reactive, dangerous content Noise Smell Occupational health Size (equipment volume reduction) Smaller plot area impact 10 Smaller skyline impact

Yes

Ecological/planet Sensitivity to world-scale nature and ecology Depletion of abiotic resources (clean air, etc.) Depletion of biotic resources; biodiversity Dehydration Depletion of stratospheric ozone layer Global warming Formation of photochemical pollutants Acidification Human toxicity Ecotoxicity (terrestrial and aquatic) Nutrification (eutrophication) Radiation Thermal pollution Waste

5 10

>1.4

1.4

1.2 1.2

1.2–5 12–10

5

>1.4

1.4

1.2

1.2–5

5

16

1.4 3.5

1.2

1.2–16 3.5

5

>1.4

1.2

1.2–5

1.6 1.6

0.8–5 1.4–5

Economy/profit Scarce resource depletion Drinking water resource depletion Fossil fuel depletion External (future) cost low Capital expenditure Operational cost Profitable over total lifecycle

Yes

— 3 4–10

3 4

5 5

0.8–1.6 1.4

Derived from process data, assuming other lifecycle steps are hardly affected.


It is likely that the number of flanges and valves has been reduced at least by the same number as the pieces of equipment. The VOC reduction in the Eastman case can therefore be a factor of 10 lower than for the conventional design. 3.2.1.2. Case S: Sulzer Hydrogen Peroxide Distillation System (39). The safety in this new process is drastically improved because of low operating temperature, minimal product holdup in the system, reliable safety devices, and proper selection and treatment of the construction materials (39). The number of units has been reduced by a factor of 4; therefore it is assumed that the plot area is reduced by the same factor, 4. The energy consumption has been reduced by 30% (reduction factor is 1.4). We have used this factor for global warming, acidification, and fossil fuel depletion. The thermal pollution from cooling water is reduced by a factor of 16 and the operational cost by a factor of 5 (39). 3.2.1.3. Case SG: Shell Global Solutions Natural Gas Dehydration (40). This case involves membrane separation for natural gas conditioning (dehydration). The mass of equipment is reduced by 70% (factor of 3 reduction) compared to the conventional process. From this we derived that the construction volume is reduced by the same factor, 3. The capital expenditure and operation cost reduction figures are given by Rijkens (40). 3.2.1.4. Case G: GlaxoSmithKline Fine Chemical from Carbonyl Process (41). The fine chemical is produced in a high-heat exchange reactor. The residence time is thereby reduced by a factor of 1800(!) compared to a conventional batch reactor. The reactive content is thereby considerably reduced; hence the process is safer. The CO2 emission related to this process is reduced by a factor of 1.4 (from 18 to 13 g CO2/mol product). The global warming, acidification, and thermal pollution are assumed to be reduced by the same factor. The waste is reduced by a factor of 3.5. 3.2.1.5. Case SP: Shell International Chemicals—PDC Bulk Chemical Process Design (37). This new process design consists of a multifunctional reactor, which largely combines reaction, heat exchange, and component separation in one piece of equipment. Data on energy and cost element reductions are provided. 3.2.2.

Conclusion on the Industrial PI Cases

It is clear that for all cases, considerable improvements are made compared to conventional processes in some criteria in the social, ecological, and economic areas. It is also clear that for most criteria, no assessment can be made because data are not reported.


3.3. Potential Contribution of Process Intensification to Sustainable Development From the definition and the examples given it is clear that process intensification does not belong to end-of-pipe technology. Most example cases belong to the category of clean technology. In the following section we will concentrate on its potential for contributing to sustainable development. Table 7 contains a summary of the potential of process intensification for sustainable development. Some aspects will be explained in the sections that follow. 3.3.1.

Social/People

3.3.1.1. Providing for Needs, Especially of the Poor. Process intensification is mainly concerned with processes and less with consumer products. Moreover it is only located on the system-level process. So a big overall contribution to sustainable development is not expected. However, by miniaturization of certain processes those processes may become part of a higher system level and have a larger impact then initially imagined. Take, for instance, the development of small fuel cells for electricity production in remote villages in developing countries, fed with hydrogen from biomass waste. These fuel cells may be massproduced at low cost, by which they enhance the quality of life in developing countries. 3.3.1.2. Social Acceptance. Social acceptance of chemical plants is still an issue. This is due to the fact that big disasters have occasionally occurred (Bhopal, Seveso) and also because chemical plants smell, due to diffusive emissions of volatile components. Via process intensification the amount of chemically hazardous, reactive material can be reduced considerably, by which the size of an emission in the case of an explosion will be far less and the chance of an explosion itself reduced by the lowered hazardous, reactive content. Often this low reactive content can lead to designing for total containment; i.e., no relief system is required by which further emission reductions are achieved. Moreover, the diffusive emissions (smell) can be reduced by reducing the number of flanges and valves. Because the size of the plant will also be reduced, it is likely that the social status of a chemical plant can be raised by the introduction of process intensification. 3.3.2.

Ecology/Planet

3.3.2.1. Dematerialization. By scale, the reduction of material and energy intensity by process intensification should reduce environmental impact. 3.3.2.2. Global-Warming Impact Reduction. There is a general reason why PI, via by function integration, can result in lower energy requirements and thereby in lower carbon dioxide emissions (given the present way of providing


TABLE 7 Sustainable Technology Scorecard for the Potential Contribution of PI to Sustainable Development Sustainable development item Social/people Provide for the needs of the poor: water, food, clothing Fair distribution of wealth, power, and knowledge Social acceptance Safe: lower reactive, dangerous content Noise Smell Occupational health Construction volume reduction Plot area impact Skyline impact Ecological/planet Sensitivity to world-scale nature and ecology Depletion of abiotic resources (clean air, etc.) Depletion of biotic resources; biodiversity Dehydration Depletion of stratospheric ozone layer Global warming Formation of photochemical pollutants Acidification Human toxicity Ecotoxicity (terrestrial and aquatic) Nutrification (eutrophication) Radiation Thermal pollution Waste Economy/profit Scarce resource depletion Drinking water resource depletion Fossil fuel depletion External (future) cost low Lower capital expenditure Lower operational cost Profitable over total lifecycle


Process intensification improvement potential

Not identified yet Not identified yet Probably yes Yes; factor > 10 possible Yes; fewer pumps and pieces of equipment Yes; fewer flanges, less diffusive emissions Yes; by less diffusive emissions Yes; factor > 4 possible Yes; factor > 4 possible Yes; factor > 4 possible To be assessed for each case Yes; lower emissions Not identified Not identified Yes; far less energy required Yes; less diffusive emissions VOC Yes; less energy required

Yes; less energy required Yes To be assessed for each case Not identified yet Yes; improvement factor > 4 Yes Yes; factor > 4 Yes; factor > 4 Likely

power via fossil fuel combustion) and thereby reducing the impact of the chemical process industry (from CO2 emissions) on global warming. The basic principle behind high-quality energy carriers (fossil fuels, solar energy) follows from thermodynamics. All process steps are carried out in a finite time and are therefore irreversible, by which high-quality energy is converted to lower-quality energy. This is called exergy loss. By reducing the number of highly irreversible process steps, exergy losses are reduced (42). This can be achieved by combining process functions into one process step. It can also be obtained by reducing the number of units just by selecting the best units and their sequence. It can also be achieved by reducing the biggest driving forces by evenly spreading them over the process steps. This will result in general in energy savings and in reductions of carbon dioxide emission (43). For distillations the savings are of the order of 30% (reduction by a factor of 1.4) (43). Individual pieces of PI equipment, as listed by Stankiewicz and Moulijn (5), requiring alternative energy sources such as microwaves and centrifugal fields will often result in higher energy requirements. Once again, the process step is made more irreversible (faster by higher driving forces). By assessing the energy requirements of the whole lifecycle, energy requirement reductions may be achieved, via, for instance, a higher selectivity, i.e., a reduction in feedstock consumption. And feedstock preparation is also associated with energy requirements. 3.3.2.3. Photochemical Pollutant Reduction. Smog is caused by photochemical oxidants. The formation of photochemical oxidants is mainly due to volatile organic compound (VOC) emissions. Seventy to ninety percent of VOC emissions from chemical plants are from leaking flanges, seals, etc. (44). Process intensification can reduce considerably the number of pieces of equipment and thereby the number of flanges and seals. Moreover, the few remaining flanges could be designed such that fugitive emissions are below the limits for smog formation, and the additional cost can be limited. This means fewer leaks and lower fugitive emissions to air, water, and soil. The introduction of process intensification could therefore make a major contribution to VOC emission reduction. 3.3.2.4. Acidification. The acidification and thermal pollution impacts of a chemical plant in general do not come from the plant itself but are related to the conversion of fossil fuel to energy (steam, electricity); hence the same emission reduction factor as for carbon dioxide is assumed for the process intensification potential. 3.3.3.

Economy/Profit

3.3.3.1. Scarce Resource Depletion. For any new technology, the longterm and worldwide effects on scarce-resource consumption should be assessed over its entire lifecycle. For instance, even minute amounts of rare earth metals per


application can, in the long run (and when the technology is applied worldwide), deplete that scarce resource. (This is of course not only an economic issue but also an ethical and ecological one). The present resources of drinking water and fossil carbon components are already seen as limited. Careful use and development of alternatives is now being pursued. Process intensification has the potential of requiring less energy and thereby requiring less fossil carbon as fuel. 3.3.3.2. Economic Cost Reduction Potential. The major part of the cost of manufacturing chemical products is the feedstocks. The capital expenditure per unit product is in general between 10% and 30%. The energy cost is in general between 5% and 10%. With bulk chemicals the feedstock costs are in general close to the theoretical minimum, due to the high selectivity achieved (90% and higher). The introduction of PI in this field can lead to a reduction in capital expenditure and of energy utilities by a factor 5 (38); a variable cost reduction factor of 1.4 is achievable. In the fine chemicals industry the selectivity figures are in general far lower. Here big feedstock cost savings is achievable by more selective processes. This is mainly achieved by changing from stoichiometric acid and alkaline feed-driven processes to truly catalytic processes (45). But via, for instance, high-heat exchange reactors (HEX reactors) that allow a much shorter reaction time and at the optimal temperature, process intensification can also reduce the formation of by-products considerably. A by-product reduction by a factor of 4 is achieved via a HEX reactor (46). This means a considerable reduction in feedstock cost and waste handling. 3.4. Traps for Process Intensification Leading to Unsustainable Technologies Larger emissions or consumption of scarce resource could occur elsewhere in the lifecycle via the large-scale introduction of process intensification. For example, the proposed printed circuit board reactors could in the manufacturing step cause large emissions of VOCs from cleaning chemicals. And the demolition phase could also cause a large waste stream. 4.

EPILOGUE

It is clear from the foregoing that presently known PI methods can contribute considerably to the process industry in meeting the social, ecological, and economic constraints of sustainable development. For each new application it remains a challenge to the engineers to identify ways to provide for the needs of people, especially the poor, while meeting all SD constraints.


REFERENCES 1. Houghton J. Global Warming, the Complete Briefing. 2nd ed. New York: Cambridge University Press, 1997. 2. van Dieren W. Taking Nature into Account; Toward a Sustainable National Income. New York: Springer-Verlag, 1995. 3. Gore A. Earth in the Balance. New York: Plume Books, 1992. 4. Meadows DH, et al. The Limits to Growth. New York: Universe Books, 1972. 5. Reid, D. Sustainable Development, an Introductory Guide. London: Earthscan Publications, 1995. 6. Intergovernmental Panel of Climate Change (IPCC). Second Assessment Report. New York: Cambridge University Press, 1995. 7. Intergovernmental Panel of Climate Change (IPCC). Climate change information sheet no. 25: New energies and policies. 1999. Internet: http://www.preen.org./iucc/ fact25.html. 8. Molina MJ, et al. Stratospheric Sinc for Chlorofluorromethanes: Chlorine Atomic Catalysed Destruction of Ozone. Nature 1974; 249: 810. 9. Meadows DH, et al. Beyond the Limits. London: Earthscan, 1991. 10. Boeker, E. Environmental Physics. New York: Wiley, 1999. 11. Brundtland GH. Our Common Future. World Commission on Environment and Development report (WCED). Oxford, UK: Oxford University Press, 1987. 12. Korevaar G, Harmsen GJ, Lemkowitz, S. Sustainable Technology: Introduction to the Field of Sustainability and Sustainable Design. Course book Faculty Applied Sciences, Delft, Delft University of Technology, 2001. 13. Installé, M. Engineering education for a more sustainable development: some proposals and an experimentation. International Conference on Industrial Ecology and Sustainability, Troyes, France, 1999. 14. IUCN, UNEP, WWF. Caring for the Earth. Gland, Switzerland: IUCN, 1991. 15. Serageldin I. Making development sustainable. Finance Dev 1993; 30(4):6–10. 16. World Business Counsel for Sustainable Development (WBCSD). Exploring Sustainable Development: Global Scenarios 2000–2050. New York: WBCSD, 1997. 17. Shell. People, Planet & Profit. The Shell Report 2000. London: Shell International Ltd. 18. Jackson T. Clean Production Strategies: Developing Preventive Environmental Management in the Industrial Economy. New York: Lewis Publishers, 1993. 19. Costanza R, et al. The value of the world’s ecosystem services and natural capital. Nature 1997; 387(6230). 20. Worster D. The Wealth of Nature. Oxford, UK: Oxford University Press, 1993. 21. Elsdon R. GreenHouse Theology: Biblical Perspectives on Caring for Creation. Tunbridge Wells, England: Monarch, 1992. 22. Jansen JLA, Vergragt PJ. Naar duurzame ontwikkeling met technologie: uitdaging in programmatisch perspectief. Milieu 1993; 8:179–183. 23. Smits, M. De onduurzaamheid van duurzame techniek (the unsustainability of sustainable technology). Wijsgerig Perspectief 1993/1994; 34:193–200. 24. Stainer A, et al. Environment, technology and corporate responsibility. Int J Environ Pollut 1999; 6:149–159.


25. 26. 27. 28. 29. 30.

31.

32. 33. 34. 35. 36. 37.

38. 39.

40.

41.

42.

Moser A. Ecotechnology in industrial practice: implementation using sustainability indices and case studies. Ecological Eng 1996; 7:117–138. Daly HE, Cobb JB. For the Common Good: Redirecting the Economy Towards Community, Environment and a Sustainable Future. 2nd ed. Boston: Beacon Press, 1994. von Weizsäcker E, Lovins AB, Lovins LH. Factor Four: Doubling Wealth, Halving Resource Use. London: Earthscan Publications, 1998. Schwartz P. The Art of the Long View. New York: Doubleday Dell, 1991. Hart S. Beyond greening: strategies for a sustainable world. Harvard Bus Rev 1997; (Jan–Feb):66–76. Harmsen GJ. Sustainable process design. In: Weijnen MPC, Herder, PM, eds. Environmental Performance Indicators in Process Design and Operation. Proceedings of EFCE Event 616. Delft: Delft University Press, 1999:59–68. Scholes G. Integrating sustainable development into the Shell Chemicals business. In: Weijnen MPC, Herder, PM, eds. Environmental Performance Indicators in Process Design and Operation. Proceedings of EFCE Event 616. Delft: Delft University Press, 1999:69–82. Hauschild M, Wenzel H. Environmental Assessment of Products. Vols. 1 and 2. London: Chapman & Hall, 1998. VNCI. Guideline Environmental Performance Indicators for the Chemical Industry. Leidschendam, Netherlands: VNCI, 2001. Swedish Institute for Analysis and Communication (SIKA). Samplan No. 13, 95.12.13. Stockholm: SIKA, 1995. Stankiewicz AJ, Moulijn JA. Process intensification: transforming chemical engineering. Chem Eng Prog 2000; (Jan):22–34. Siirola JJ. An industrial perspective on process synthesis. 4th Int. Conf. Foundations CAPE, AIChE Symp. Series 1995; 91(304):222–233. Harmsen GJ, Hinderink AP. We want less: process intensification by process synthesis methods. In: Green A, ed. 3rd International Conference on Process Intensification for the Chemical Industry, Antwerpen, 25–27 Oct. 1999. London: Professional Engineering Publications, 1999:23–30. Siirola JJ. Synthesis of equipment with integrated functionality. Syllabus First Dutch Process Intensification: Profits for the Chemical Industry Symposium, 7 May 1998. Meili A. Practical process intensification shown with the example of a hydrogen peroxide distillation system. Syllabus First Dutch Process Intensification: Profits for the Chemical Industry Symposium, 7 May 1998. Rijkens HC. Membrane developments for natural gas conditioning. Syllabus Second Dutch Symposium Process Intensification: A Challenge for the Process Industry. November 2000. Jimenez-Gonzales C. How do you select the greenest technology? Development of guidance for the pharmaceutical industry. AIChE Proceedings Topical Conference Implementing Sustainability in the Process Industries. AIChE 2001 Spring National Meeting, Houston TX April 22–26, 2001:188–216. Harmsen GJ, Chewter L. Industrial applications of multi-functional, multi-phase reactors. Int. Symp. Multifunctional Reactors I, Amsterdam, 26–28 April 1999. Chem Eng Sci 1999; 54:1541–1545.


43.

44. 45. 46.

47.

Sauar E. Energy Efficient Process Design by Equipartion of Forces with Applications to Distillation and Chemical Reaction. PhD dissertation. Norwegian TechnicalNatural Science University of Trondheim, 1998. Schaic JR. Estimate fugitive emissions from process equipment. Chem Eng Prog 1991; 31–35. Anastas P, Warner JC. Green Chemistry, Theory and Practice. Oxford, UK: Oxford University Press, 1998. Philips, CH. Development of a novel compact chemical reactor-heat exchanger. In: Green A, ed. 3rd International Conference on Process Intensification for the Chemical Industry, Antwerpen, 25–27 Oct. 1999. London: Professional Engineering Publications, 1999:71–88. Bossel H. Earth at a Crossroads, Paths to a Sustainable Future. New York: Cambridge University Press, 1998.


Re Engineering the Chemical Processing Plant

Recommend Documents