Modern Drying Technology, Product Quality and Formulation, Volume 3 [Volume 3 ed.] 3527315586, 9783527315581

Citation preview

Edited by Evangelos Tsotsas and Arun S. Mujumdar Modern Drying Technology

Modern Drying Technology Edited by E. Tsotsas and A. Mujumdar

Other Volumes Volume 1: Computational Tools at Different Scales ISBN: 978-3-527-31556-7

Volume 2: Experimental Techniques ISBN: 978-3-527-31557-4

Forthcoming Volumes Volume 4: Energy Savings ISBN: 978-3-527-31559-8

Volume 5: Process Intensification ISBN: 978-3-527-31560-4

Forthcoming Volumes Modern Drying Technology Set (Volumes 1 – 5) ISBN: 978-3-527-31554-3

Edited by Evangelos Tsotsas and Arun S. Mujumdar

Modern Drying Technology Volume 3: Product Quality and Formulation

The Editors: Prof. Evangelos Tsotsas Otto von Guericke University Thermal Process Engineering Universitätsplatz 2 39106 Magdeburg Germany

All books published by Wiley-VCH are carefully produced. Nevertheless, authors, editors, and publisher do not warrant the information contained in these books, including this book, to be free of errors. Readers are advised to keep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate. Library of Congress Card No.: applied for

Prof. Arun S. Mujumdar National University of Singapore Mecnical Engineering/Block EA 07-0 9 Engineering Drive 1 Singapore 117576 Singapore

British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.d-nb.de. # 2011 Wiley-VCH Verlag & Co. KGaA, Boschstr. 12, 69469 Weinheim, Germany All rights reserved (including those of translation into other languages). No part of this book may be reproduced in any form – by photoprinting, microfilm, or any other means – nor transmitted or translated into a machine language without written permission from the publishers. Registered names, trademarks, etc. used in this book, even when not specifically marked as such, are not to be considered unprotected by law. Composition

Thomson Digital, Noida, India

Printing and Binding Cover Design

Strauss GmbH, Mörlenbach

Adam Design, Weinheim

Printed in the Federal Republic of Germany Printed on acid-free paper ISBN: 978-3-527-31558-1

V

Contents Series Preface XIII Preface of Volume 3 XVII List of Contributors XXI Recommended Notation XXV EFCE Working Party on Drying: Address List 1

1.1 1.2 1.3 1.4 1.5 1.6

2

2.1 2.2 2.3 2.4 2.4.1 2.4.2 2.5 2.6

XXXI

Quality Changes in Food Materials as Influenced by Drying Processes 1 Catherine Bonazzi and Elisabeth Dumoulin Introduction 1 Biochemical Reactions Induced by Drying 5 Physical Transformations During Drying 9 Mechanical Transformations Induced by Drying Storage and Rehydration of Food Products 16 Conclusion 17 References 18

14

Impact of Drying on the Mechanical Properties and Crack Formation in Rice 21 Catherine Bonazzi and Francis Courtois Introduction 21 Impact of Drying Conditions on Head Rice Yield for Paddy and Parboiled Rice 24 Characterization of Fissures Formation by Image Analysis Techniques 28 Characterization of the Mechanical Properties of the Rice Material 33 Stress–Strain Relationships for Linear Materials 34 Failure Strength in Rice Grains 36 Modeling the Impact of Drying on the Final Quality of Rice Grains 39 Conclusion 45 References 46

VI

Contents

3

3.1 3.2 3.2.1 3.2.2 3.2.3 3.2.4 3.2.5 3.2.6 3.3 3.3.1 3.3.1.1 3.3.1.2 3.3.1.3 3.3.2 3.3.2.1 3.3.2.2 3.3.3 3.3.4 3.3.4.1 3.3.4.2 3.3.5 3.3.6 3.4 3.4.1 3.4.2 3.5

4

4.1 4.2 4.3

Characterization and Control of Physical Quality Factors During Freeze-Drying of Pharmaceuticals in Vials 51 Julien Andrieu and Séverine Vessot Introduction 51 Characterization Methods of the Key Quality Factors During Freeze-Drying of Pharmaceuticals in Vials 52 State Diagram, Melting Curves, Vitreous Transition, Collapse Temperature 54 Characterization Methods: DSC, MDSC, Freeze-Drying Microscopy 55 Ice Structure and Morphology: Cold Chamber Optical Microscopy 55 Heat Flux Heterogeneity in the Sublimation Chamber 57 Permeability of Freeze-Drying Cake: Pressure Rise Tests 59 Estimation of Mean Product Temperature 61 Influence of Freezing and Freeze-Drying Parameters on Physical Quality Factors 63 Influence of Freezing Protocol on Ice Morphology 63 Influence of Freezing Rate 64 Influence of Vial Type and Filling Height 66 Annealing 67 Controlled Nucleation 70 Controlled Nucleation by Ultrasound Sonication 70 Effect of Ultrasound on Structural and Morphological Properties 72 Relationship between Nucleation Temperatures and Sublimation Rates 73 Freeze-Dried Cake Morphology 74 Water Vapor Mass Transfer Resistance 74 Freeze-Dried Layer Permeability 76 Importance of Temperature Control 78 Influence of Operating Conditions on Sublimation Kinetics 79 Product Quality and Stability During Drying and Storage 83 Product Quality and Formulation 83 Product Quality and Polymorphism 84 Conclusions 85 References 87 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes 91 Antonello A. Barresi and Davide Fissore Introduction 91 Control of the Freezing Step 94 Monitoring of the Primary Drying 96

Contents

4.3.1 4.3.2 4.3.3 4.3.3.1 4.3.3.2 4.3.3.3 4.4 4.5 4.6 4.7 4.8

5

5.1 5.2 5.2.1 5.2.1.1 5.2.1.2 5.2.2 5.2.3 5.3 5.3.1 5.3.1.1 5.3.1.2 5.3.2 5.3.2.1 5.3.2.2 5.3.2.3 5.3.3 5.4 5.4.1 5.4.1.1 5.4.1.2 5.4.1.3 5.4.1.4 5.4.2 5.4.3 5.4.3.1 5.4.3.2

Monitoring of Single Vials 99 Monitoring of a Group of Vials 103 Monitoring of the Whole Batch 106 Detection of the Endpoint of the Primary Drying 106 Monitoring the Primary Drying Using the Measurement of the Sublimation Flux 113 Monitoring the Primary Drying Using Methods Based on the PRT 114 Control of the Primary Drying 125 Monitoring and Control of Secondary Drying 135 Quality by Design 139 Continuous Freeze-Drying 142 Conclusion 143 References 146 Understanding and Preventing Structural Changes During Drying of Gels 155 Thomas Metzger, Angélique Léonard, Wahbi Jomaa, and Hajime Tamon Introduction 155 Gels and Their Applications – Quality Aspects 156 Preparation of Wet Gels 156 Silica Gels 156 Resorcinol-Formaldehyde (RF) Gels 157 Properties of Dry Gels 158 Applications of Dry Gels 160 Structural Characterization of Gels – Quality Assessment Characterization of Wet Gels 162 Small Angle X-Ray Scattering (SAXS) 162 Thermoporometry 164 Characterization of Dry Gels 166 Nitrogen Adsorption 166 Mercury Porosimetry 168 Other Methods 171 Characterization of Gels During Drying 172 Drying Methods for Gels – Quality Loss 174 Convective Drying 174 Introduction 174 Shrinkage 175 Differential Shrinkage and Stress 177 Cracking 180 Freeze-drying 182 Supercritical Drying 185 Supercritical Drying of the Initial Solvent 185 Low-Temperature Process with CO2 187

162

VII

VIII

Contents

5.5 5.5.1 5.5.2 5.5.2.1 5.5.2.2 5.5.2.3 5.5.3 5.5.4 5.5.4.1 5.5.4.2 5.5.4.3 5.5.4.4 5.5.4.5 5.5.4.6 5.5.5 5.6 5.6.1 5.6.1.1 5.6.1.2 5.6.1.3 5.6.2 5.7

6 6.1 6.2 6.2.1 6.2.2 6.2.3 6.2.4 6.2.5 6.2.6 6.2.6.1 6.2.6.2 6.2.6.3 6.2.6.4 6.2.7

Advanced Drying Techniques – Preserving Quality 189 Subcritical Drying 189 Freeze-Drying 190 General Remarks 190 RF and Carbon Cryogels 192 Ice Templating (for Silica Gels) 195 Vacuum Drying 197 Convective Drying 198 Preliminary Remarks 198 Preventing Shrinkage and Cracks by Aging (Silica Gels) 199 Making Shrinkage Reversible by Surface Modification (Silica Gels) 201 RF Gels – From First Results to a Systematic Investigation 204 Aging of RF Gels 208 Concluding Remarks 209 Microwave Drying 210 Advanced Modeling of Convective Drying – Understanding Quality 211 Macroscopic Models 211 General Remarks 211 Diffusion Model 211 More Rigorous Modeling 217 Development of A Pore-Scale Model 218 Summary 220 References 223 Morphology and Properties of Spray-Dried Particles 231 Peter Walzel and Takeshi Furuta Introduction 231 Morphology of Spray-Dried Particles 234 Classification of the Morphology of Spray-Dried Powders 234 Solutions of Low Molecular Weight Substances 236 Solutions of Polymers 240 Suspensions Containing Small Solid Particles 240 Suspensions Containing Large Solid Particles 244 Complex Dispersions Such as Emulsions and Other Formulations 244 Hard Shell Particles 244 Gelatinization 245 Microencapsulated Flavor Powders Formed from Emulsions 245 Particles from Proteins, Enzymes and Carrier Materials 247 Particles Obtained in Freeze Spray Drying 251

Contents

6.2.8 6.3 6.3.1 6.3.2 6.3.2.1 6.3.2.2 6.3.2.3 6.3.2.4 6.3.2.5 6.3.3 6.3.3.1 6.3.3.2 6.3.3.3 6.3.3.4 6.4 6.4.1 6.4.2 6.4.2.1 6.4.2.2 6.4.2.3 6.4.3 6.4.4 6.4.4.1 6.4.4.2 6.4.4.3 6.4.5 6.4.5.1 6.4.5.2 6.5 6.5.1 6.5.2 6.5.3 6.5.4 6.5.5 6.6

Particles Formed in Integrated Fluidized Beds 253 Retention of Flavor in Spray-Dried Food Products 253 General Remarks on Microencapsulation 253 Encapsulation of Flavor by Spray Drying 255 Theory and Mechanism 255 Microencapsulation of Hydrophilic Flavors 255 Spray Drying of Emulsified Hydrophobic Flavors 256 Factors Affecting the Retention of Emulsified Hydrophobic Flavors During Spray Drying 257 Stickiness of the Spray-Dried Powder 260 Release and Oxidation of the Encapsulated Flavor During Storage 261 Influence of Glass Temperature on the Storage Stability of the Encapsulated Flavor 261 Release of Flavor from Spray-Dried Powder During Storage 262 Oxidation of Encapsulated Flavor During Storage 267 Relaxation Process Correlation by Glass Transition Temperature 267 Encapsulation and Microencapsulation of Enzymes and Oil by Spray Drying 269 Microencapsulation of Enzymes by Spray Drying 269 Stress on Proteins During the Spray Drying Processes 270 Adsorption Stress 270 Shear Stress 271 Thermal and Dehydration Stress 271 Protein Encapsulation Theory by Spray Drying 272 Spray Drying of Protein Solutions 273 Drying of a Single Suspended Droplet 273 Stabilization of Enzymes During Spray Drying: Effects of Formulation Composition 274 Effect of Process Variables on the Stabilization of Enzymes During Spray Drying 275 Microencapsulation of Oil 278 Spray Drying of Oil Emulsions 278 Oxidation of Lipids Encapsulated in Spray-Dried Particles 279 General Quality Aspects 280 Porosity of Spray-Dried Particles and Species Distribution 280 Strength of Particles and Attrition 281 Bulk Density and Product Flowability 282 Residual Moisture 282 Reconstitution Behavior 283 Concluding Remarks 284 References 286

IX

X

Contents

7

7.1 7.2 7.3 7.3.1 7.3.1.1 7.3.1.2 7.3.1.3 7.3.1.4 7.3.1.5 7.3.2 7.3.2.1 7.3.2.2 7.3.2.3 7.3.2.4 7.3.3 7.4 7.4.1 7.4.1.1 7.4.1.2 7.4.1.3 7.4.2 7.5 7.5.1 7.5.2 7.5.3 7.5.4 7.5.4.1 7.5.4.2 7.5.4.3 7.5.4.4 7.6 7.7 7.7.1 7.7.2 7.7.3

Particle Formulation in Spray Fluidized Beds 295 Mirko Peglow, Sergiy Antonyuk, Michael Jacob, Stefan Palzer, Stefan Heinrich, and Evangelos Tsotsas Introduction 295 General Principles of Particle Formulation in Spray Fluidized Beds 296 Influence of Material Properties 299 Adhesion Mechanisms and Mechanical Strength of Agglomerates 299 Material Structure and Properties 300 Van der Waals Forces 301 Capillary Forces Due to Liquid Bridges Between Particles 303 Viscous Forces in Sinter Bridges Between Amorphous Particles 304 Mechanical Strength of Agglomerates 308 Breakage of Agglomerates and of Granulated Products 315 Elastic–Brittle Breakage Behavior 316 Elastic–Plastic Breakage Behavior 317 Plastic Breakage Behavior 319 Breakage of Granules with Layered Structure 320 Consideration of Primary Particle Properties in Agglomeration 321 Influence of Operating Conditions 324 Mechanical Strength of Granulated Particles 324 Influence of Binder Content in the Sprayed Solution 325 Influence of the Particle Retention Time 326 Influence of Process Temperature 327 Catalyst Impregnation in Fluidized Beds 329 Influence of Apparatus Design 332 Apparatus Design Features with an Influence on Product Quality 332 Residence Time Distribution 338 Dispersive Growth in Batch Granulation 344 Discrete Particle Modeling of a Wurster Coater 349 Principles of the DPM 350 Parameters for the DPM Simulation 351 Influence of the Spout Velocity 352 Influence of the Wurster Gap Distance 355 Neural Networks, Encapsulation 357 Stochastic Discrete Modeling of Agglomeration 363 General Principles 363 Computational Method 364 Results 367

Contents

7.7.3.1 7.7.3.2 7.7.3.3 7.8

Effect of Liquid Flow Rate and Viscosity Thermal Effects 367 Effect of Particle Porosity 369 Summary and Outlook 372 References 374 Index

379

367

XI

XIII

Series Preface The present series is dedicated to drying, i.e. to the process of removing moisture from solids. Drying has been conducted empirically since the dawn of the human race. In traditional scientific terms it is a unit operation in chemical engineering. The reason for the continuing interest in drying and, hence, the motivation for the series concerns the challenges and opportunities. A permanent challenge is connected to the sheer amount and value of products that must be dried – either to attain their functionalities, or because moisture would damage the material during subsequent processing and storage, or simply because customers are not willing to pay for water. This comprises almost every material used in solid form, from foods to pharmaceuticals, from minerals to detergents, from polymers to paper. Raw materials and commodities with a low price per kilogram, but with extremely high production rates, and also highly formulated, rather rare but very expensive specialties have to be dried. This permanent demand is accompanied by the challenge of sustainable development providing welfare, or at least a decent living standard, to a stillgrowing humanity. On the other hand, opportunities emerge for drying, as well as for any other aspect of science or living, from either the incremental or disruptive development of available tools. This duality is reflected in the structure of the book series, which is planned for five volumes in total, namely: Volume 1: Computational tools at different scales Volume 2: Experimental techniques Volume 3: Product quality and formulation Volume 4: Energy savings Volume 5: Process intensification As the titles indicate, we start with the opportunities in terms of modern computational and experimental tools in Volumes 1 and 2, respectively. How these opportunities can be used in fulfilling the challenges, in creating better and new products, in reducing the consumption of energy, in significantly improving existing or introducing new processes will be discussed in Volumes 3, 4 and 5. In this sense, the first two volumes of the series will be driven by science; the last three will try to show how engineering science and technology can be translated into progress.

XIV

Series Preface

In total, the series is designed to have both common aspects with and essential differences from an extended textbook or a handbook. Textbooks and handbooks usually refer to well-established knowledge, prepared and organized either for learning or for application in practice, respectively. On the contrary, the ambition of the present series is to move at the frontier of “modern drying technology”, describing things that have recently emerged, mapping things that are about to emerge, and also anticipating some things that may or should emerge in the near future. Consequently, the series is much closer to research than textbooks or handbooks can be. On the other hand, it was never intended as an anthology of research papers or keynotes – this segment being well covered by periodicals and conference proceedings. Therefore, our continuing effort will be to stay as close as possible to a textbook in terms of understandable presentation and as close as possible to a handbook in terms of applicability. Another feature in common with an extended textbook or a handbook is the rather complete coverage of the topic by the entire series. Certainly, not every volume or chapter will be equally interesting for every reader, but we do hope that several chapters and volumes will be of value for graduate students, for researchers who are young in age or thinking, and for practitioners from industries that are manufacturing or using drying equipment. We also hope that the readers and owners of the entire series will have a comprehensive access not to all, but to many significant recent advances in drying science and technology. Such readers will quickly realize that modern drying technology is quite interdisciplinary, profiting greatly from other branches of engineering and science. In the opposite direction, not only chemical engineers, but also people from food, mechanical, environmental or medical engineering, material science, applied chemistry or physics, computing and mathematics may find one or the other interesting and useful results or ideas in the series. The mentioned interdisciplinary approach implies that drying experts are keen to abandon the traditional chemical engineering concept of unit operations for the sake of a less rigid and more creative canon. However, they have difficulties of identification with just one of the two new major trends in chemical engineering, namely process-systems engineering or product engineering. Efficient drying can be completely valueless in a process system that is not efficiently tuned as a whole, while efficient processing is certainly valueless if it does not fulfil the demands of the market (the customer) regarding the properties of the product. There are few topics more appropriate in order to demonstrate the necessity of simultaneous treatment of product and process quality than drying. The series will try to work out chances that emerge from this crossroads position. One further objective is to motivate readers in putting together modules (chapters from different volumes) relevant to their interests, creating in this manner individual, task-oriented threads trough the series. An example of one such thematic thread set by the editors refers to simultaneous particle formation and drying, with a focus on spray fluidized beds. From the point of view of process-systems engineering, this is process integration – several “unit operations” take place in the same equipment.

Series Preface

On the other hand, it is product engineering, creating structures – in many cases nanostructures – that correlate with the desired application properties. Such properties are distributed over the ensemble (population) of particles, so that it is necessary to discuss mathematical methods (population balances) and numerical tools able to resolve the respective distributions in one chapter of Volume 1. Measuring techniques providing access to properties and states of the particle system will be treated in one chapter of Volume 2. In Volume 3, we will attempt to combine the previously introduced theoretical and experimental tools with the goal of product design. Finally, important issues of energy consumption and process intensification will appear in chapters of Volumes 4 and 5. Our hope is that some thematic combinations we have not even thought about in our choice of contents will arise in a similar way. As the present series is a series of edited books, it can not be as uniform in either writing style or notation as good textbooks are. In the case of notation, a list of symbols has been developed and will be printed in the beginning of every volume. This list is not rigid but foresees options, at least partially accounting for the habits in different parts of the world. It has been recently adopted as a recommendation by the Working Party on Drying of the European Federation of Chemical Engineering (EFCE). However, the opportunity of placing short lists of additional or deviant symbols at the end of every chapter has been given to all authors. The symbols used are also explained in the text of every chapter, so that we do not expect any serious difficulties in reading and understanding. The above indicates that the clear priority in the edited series was not in uniformity of style, but in the quality of contents that are very close to current international research from academia and, where possible, also from industry. Not every potentially interesting topic is included in the series, and not every excellent researcher working on drying contributes to it. However, we are very confident about the excellence of all research groups that we were able to gather together, and we are very grateful for the good cooperation with all chapter authors. The quality of the series as a whole is set mainly by them; the success of the series will primarily be theirs. We would also like to express our acknowledgements to the team of Wiley-VCH who have done a great job in supporting the series from the first idea to realization. Furthermore, our thanks go to Mrs Nicolle Degen for her additional work, and to our families for their tolerance and continuing support. Last but not least, we are grateful to the members of the Working Party on Drying of the EFCE for various reasons. First, the idea about the series came up during the annual technical and business meeting of the working party 2005 in Paris. Secondly, many chapter authors could be recruited among its members. Finally, the Working Party continues to serve as a panel for discussion, checking and readjustment of our conceptions about the series. The list of the members of the working party with their affiliations is included in every volume of the series in the sense of acknowledgement, but also in order to promote networking and to provide access to national working parties, groups and individuals. The present edited books are

XV

XVI

Series Preface

complementary to the regular activities of the EFCE Working Party on Drying, as they are also complementary to various other regular activities of the international drying community, including well-known periodicals, handbooks, and the International Drying Symposia. June 2007

Evangelos Tsotsas Arun S. Mujumdar

XVII

Preface of Volume 3 The first two volumes of this series have treated “Computational tools at different scales” and “Experimental techniques” that can empower “Modern Drying Technology” with the aim of producing superior products with better processes. Now, it is time to turn from the means to the goal, treating “Product quality and formulation” in Volume 3. This emphasis on the product is deliberate, because even the most efficient process is not of real value, if not able to fulfill – if not push – the requirements of the market. The topic is presented in seven chapters: Chapter 1: Quality changes in food materials as influenced by drying processes Chapter 2: Impact of drying on the mechanical properties and crack formation in rice Chapter 3: Characterization and control of physical quality factors during freezedrying of pharmaceuticals in vials Chapter 4: In-line product quality control of pharmaceuticals in freeze-drying processes Chapter 5: Understanding and preventing structural changes during drying of gels Chapter 6: Morphology and properties of spray-dried particles Chapter 7: Particle formulation in spray fluidized beds Chapter 1 refers to a big, utterly important group of products to be dried, namely foods. It summarizes food properties, introduces the glass transition temperature as a humidity dependent landmark between the glassy and the rubbery state of amorphous materials, and discusses biochemical, physical and mechanical transformations that can take place during drying. Furthermore, it connects drying with quality changes during storage and with properties relevant to the final use of the processed food. One good example of what can happen after drying is the fissuring and breakage of rice kernels due to stresses and strains that developed during the process. Therefore, this example is used in Chapter 2 in order to show how the previously discussed general principles can be cast into specific and precise characterization methods and models for the preservation of the quality of a valuable but perishable good.

XVIII

Preface of Volume 3

In Chapter 3 the focus is shifted to pharmaceuticals, specifically to active ingredients with a high molecular weight, such as therapeutic proteins or enzymes. Such compounds are usually produced biotechnologically, so that they often have to be transformed from an aqueous solution to a solid form. This is commonly carried out by freezing and then freeze-drying, in order to protect the complex molecular structure from deterioration. The chapter discusses thoroughly, what kinds of damage can occur during the process, and how they can be avoided. And, it shows impressively, how intimate the interrelation of freeze-drying to the preceding process of freezing is. This interconnection results from the fact that the solid skeleton of freeze-dried cakes is created during freezing by the size and spatial placement of the ice crystals. Such causality offers rich opportunities of beneficial manipulation by changes in the freezing protocol, controlled nucleation or annealing, which are worked out in detail. Though the degradation of pharmaceuticals during freeze-drying is not permissible too conservative an operation also should be avoided, because it is very expensive. The key for resolving this dilemma between product quality and process efficiency is monitoring and control. Consequently, methods that can be used for monitoring and control during freeze-drying of pharmaceuticals are presented in Chapter 4. This is done in a very comprehensive and precise way, distinguishing among methods that refer to single vials, groups of vials, and the entire dryer for the primary or the secondary period of drying. Close reference to process analytical technology (PAT) is given throughout. Gels are a class of materials with high porosity, very small primary particle size, and a plethora of possible applications. However, such applications require that the gels can be dried without destroying the structures which are generic for their properties. This is not an easy task, because very small primary particles imply very high capillary forces during drying, so that the material can crack and break. Chapter 5 points out that convective drying may still be successful if applied in an educated way, and compares with numerous alternatives, such as freeze-drying and supercritical drying. Apart from the detailed discussion of processing options, the preparation and the characterization of gel materials are elucidated. Though the preservation of existing structures is a big goal, structures and the conjugated properties can even be created by drying. This is always the case when the removal of water or some other solute is accompanied by the formation of the solid phase, as in spray drying, which is treated comprehensively in Chapter 6. This chapter refers to solutions of components with a low or high molecular weight, as well as to suspensions of small or large particles, and shows how drying conditions and material properties influence the morphology of the resulting products. Methods of formulation by encapsulation of, for example, flavors or enzymes, are presented in detail, including stability and quality of the obtained products. The idea of formulation by drying is elaborated further in Chapter 7. Here, drying after spraying on fluidized particles with the aim of producing agglomerates, layered granules, or coatings is discussed. It is worked out on many examples, how the processes and the products can be enhanced by manipulation of material properties, operating conditions, and apparatus design. The physical background is explained

Preface of Volume 3

down to the molecular scale in order to derive conditions for adhesion around small particle contacts. Understanding and characterization of properties relevant to the processing or final use of the particles are, again, important issues. Furthermore, modeling tools with different degrees of resolution and sophistication – such as discrete particle modeling, Monte Carlo simulations, and neural networks – which can separately or in combination support process and product development are presented. Readers looking for thematic threads within the Modern Drying Technology series will easily recognize many, including those between the present: – Chapter 1 and Chapter 2 of Vol. 2 (drying of foods) – Chapters 2 and 5 and Chapters 3 and 4 of Vol. 1 (thermo-mechanics) – Chapter 4 and Chapter 1 of Vol. 2 (monitoring) – Chapter 5 and Chapter 3 of Vol. 2 (x-ray tomography) – Chapter 6 and Chapter 5 of Vol. 1 (spray drying) – Chapter 7 and Chapter 6 of Vol. 1, as well as Chapter 5 of Vol. 2 (fluidized bed formulation) Readers interested in transport phenomena at different scales will find molecular, pore-scale, particle-scale and particle system or processing equipment considerations, as in every volume of the series, and those aiming at interdisciplinary approaches will see clear links to food engineering, pharmaceutical technology, biotechnology, mechanics, and material science. People looking for their specific product may not be able to find it in the present volume, but they may learn from methods and approaches successfully applied to other products. For a book without encyclopedic ambitions, which aims at the educated use of modern scientific methods in practice, this would be the biggest success. As to the acknowledgements, for Volume 3 they are identical to those in the series preface. We would like to stress them by reference and not repeat them here. June 2011

Evangelos Tsotsas Arun S. Mujumdar

XIX

XXI

List of Contributors Editors Prof. Evangelos Tsotsas Otto von Guericke University Magdeburg Thermal Process Engineering PSF 4120 39106 Magdeburg Germany Email: [email protected] Prof. Arun S. Mujumdar Dept. of Mechanical Engineering 9 Engineering Drive 1 Singapore 117576 Singapore Email: [email protected] Authors Prof. Julien Andrieu Université C. Bernard - Lyon1/ E.S.C.P.E. Laboratoire d’Automatique et de Génie des Procédés (LAGEP) UMR Q 5007 CNRS UCB Bâtiment 308G 43 boulevard du 11 Novembre 1918 69622 Villeurbanne Cedex France Email: [email protected]

Dr. Sergiy Antonyuk Technical University Hamburg-Harburg Solids Process Engineering and Particle Technology 21073 Hamburg Germany Email: [email protected] Prof. Antonello A. Barresi Politecnico di Torino Dipartimento di Scienza dei Materiali e Ingegneria Chimica Corso Duca degli Abruzzi 24 10129 Torino Italy Email: [email protected] Dr. Catherine Bonazzi INRA/AgroParisTech UMR1145 Ingénierie Procédés Aliments 1 avenue des Olympiades 91300 Massy France Email: catherine.bonazzi@ agroparistech.fr

XXII

List of Contributors

Dr. Francis Courtois INRA/AgroParisTech UMR1145 Ingénierie Procédés Aliments 1 avenue des Olympiades 91300 Massy France Email: francis.courtois@ agroparistech.fr Prof. Elisabeth Dumoulin INRA/AgroParisTech UMR1145 Ingénierie Procédés Aliments 1 avenue des Olympiades 91300 Massy France Email: elisabeth.dumoulin @agroparistech.fr Dr. Davide Fissore Politecnico di Torino Dipartimento di Scienza dei Materiali e Ingegneria Chimica Corso Duca degli Abruzzi 24 10129 Torino Italy Email: davide.fi[email protected] Prof. Takeshi Furuta Tottori University Department of Chemistry and Biotechnology Graduate School of Engineering 4-101, Koyama-minami Tottori, 680-8552 Japan Email: [email protected] Prof. Stefan Heinrich Technical University Hamburg-Harburg Solids Process Engineering and Particle Technology 21073 Hamburg Germany Email: [email protected]

Dip.-Ing. Michael Jacob Glatt Ingenieurtechnik GmbH Nordslrasse 12 99427 Weimar Germany Email: [email protected] Prof. Wahbi Jomaa University Bordeaux 1 Laboratory TREFLE Esplanade des Arts et Métiers 33405 Talence France Email: [email protected] Prof. Angélique Léonard Laboratoire de Génie Chimique Département de Chimie Appliquée Université de Liège Bâtiment B6c - Sart-Tilman 4000 Liège Belgium Email: [email protected] Jun.-Prof. Thomas Metzger Otto von Guericke University Magdeburg Thermal Process Engineering PSF 4120 39106 Magdeburg Germany Email: [email protected] Prof. Stefan Palzer Nestlé Product Technology Centre York Nestec York Ltd P.O. Box 204 Haxby Road York YO 91 1XY United Kingdom Email: [email protected]

List of Contributors

Jun.-Prof. Mirko Peglow Otto von Guericke University Magdeburg Thermal Process Engineering PSF 4120 39106 Magdeburg Germany Email: [email protected]

Dr. Séverine Vessot Laboratoire d’Automatique et de Génie des Precédés (LAGEP) UMR Q 5007 CNRS UCB Lyon1-CPE Bât. 308G, 43 Bd. du 11 Novembre 1918 69622 Villeurbanne Cedex France Email: [email protected]

Prof. Hajime Tamon Department of Chemical Engineering Kyoto University Katsura Kyoto 615-8510 Japan Email: [email protected]

Prof. Peter Walzel Universität Dortmund Fachbereich Bio- und Chemieingenieurwesen Particle Technology Emil Figge Str. 68 44227 Dortmund Germany Email: peter.walzel@ bci.tu-dortmund.de

Prof. Evangelos Tsotsas Otto von Guericke University Magdeburg Thermal Process Engineering PSF 4120 39106 Magdeburg Germany Email: [email protected]

XXIII

XXV

Recommended Notation . .

.

.

Alternative symbols are given in brackets Vectors are denoted by bold symbols, a single bar, an arrow or an index (e.g., index: i) Tensors are denoted by bold symbols, a double bar or a double index (e.g., index: i, j) Multiple subscripts should be separated by colon (e.g., rp;dry : density of dry particle)

A aw B b C(K) c D D(d) d E F _ FðVÞ f f G G g H H H h hðaÞ ~hðhN Þ Dhv I

surface area water activity nucleation rate breakage function constant or coefficient specific heat capacity equipment diameter diffusion coefficient diameter or size of solids energy mass flux function volumetric flow rate relative (normalized) drying rate multidimensional number density shear function or modulus growth rate acceleration due to gravity height enthalpy Heaviside step function specific enthalpy (dry basis) heat-transfer coefficient molar enthalpy specific enthalpy of evaporation total number of intervals

m2 — kg 1 m 1 s m 3 various J kg 1 K 1 m m2 s 1 m J — m3 s 1 — — Pa kg s 1 ms 2 m J — J kg 1 Wm 2 K 1 J mol 1 J kg 1 —

1

XXVI

Recommended Notation

J J _ JÞ jðm; K kðbÞ L MðmÞ ~ MðM; MN Þ _ MðWÞ _ mðJ; jÞ m_ N N _ NÞ NðW n n n _ NÞ nðJ P P p _ QðQÞ _ qðqÞ R R ~ NÞ RðR r r S S s T t u u V _ VðFÞ v v W _ WðMÞ w X

numerical flux function Jacobian matrix mass flux, drying rate dilatation function or bulk modulus mass transfer coefficient length mass molecular mass mass flow rate mass flux, drying rate volumetric rate of evaporation number molar amount molar flow rate molar density, molar concentration number density outward normal unit vector molar flux power total pressure partial pressure/vapor pressure of component heat flow rate heat flux equipment radius individual gas constant universal gas constant radial coordinate pore (throat) radius saturation selection function boundary-layer thickness temperature time velocity, usually in z-direction displacement volume, averaging volume volumetric flow rate specific volume general velocity, velocity in x-direction weight force mass flow rate velocity, usually in y-direction solids moisture content (dry basis)

— various kg m 2 s 1 Pa ms 1 m kg kg kmol 1 kg s 1 kg m 2 s 1 kg m 3 s 1 — mol mol s 1 mol m 3 m 3 mol m W kg m s kg m s

2

s

2 2

W Wm 2 m J kg 1 K 1 J kmol 1 K m m — s 1 m K, 8C s ms 1 m m3 m3 s 1 m3 kg 1 ms 1 N kg s ms —

1 1

1

1

Recommended Notation

x x x x0 ~(x N) x Y y y(v) ~y(y N) z

mass fraction in liquid phase particle volume in population balances general Eulerian coordinate, coordinate (usually lateral) general Lagrangian coordinate molar fraction in liquid phase gas moisture content (dry basis) spatial coordinate (usually lateral) mass fraction in gas phase molar fraction in gas phase spatial coordinate (usually axial)

Operators r r. D

gradient operator divergence operator difference operator

Greek letters aðhÞ bðkÞ b d dðDÞ e e e e h u k l m m n p r P s s s s t

heat-transfer coefficient mass-transfer coefficient aggregation kernel Dirac-delta distribution diffusion coefficient voidage emissivity small-scale parameter for periodic media strain efficiency angle, angular coordinate thermal diffusivity thermal conductivity dynamic viscosity moment of the particle-size distribution kinematic viscosity circular constant density, mass concentration summation operator surface tension Stefan–Boltzmann constant for radiative heat transfer standard deviation (of pore-size distribution) stress dimensionless time

— m3 m m — — m — — m

Wm 2 K ms 1 s 1 m2 s — — —

1

— — rad m2 s 1 Wm 1 K kg m 1 s various m2 s 1 — kg m 3 Nm Wm m Pa —

1

1 1

1 2

K

4

XXVII

XXVIII

Recommended Notation

F w w v v(y) Subscripts a as b bed c c cr D dry dp eff eq f g H i i,1,2,. . . i,j,k in l m max mf min N o out P p pbe ph r rel s S surf V

characteristic moisture content relative humidity phase potential angular velocity mass fraction in gas phase

at ambient conditions at adiabatic saturation conditions bound water bed cross section capillary at critical moisture content drag dry at dewpoint effective equilibrium (moisture content) friction gas (dry) wet (humid) gas inner component index, particle index coordinate index, i; j; k ¼ 1 to 3 inlet value liquid (alternative: as a superscript) mean value maximum at minimum fluidization minimum molar quantity outer outlet value at constant pressure particle population balance equation at the interface radiation relative velocity solid (compact solid phase), alternative: as a superscript at saturation conditions surface based on volume

— — Pa rad s —

1

Recommended Notation

v w w wb wet 1

vapor, evaporation water wall at wet-bulb conditions wet at large distance from interface

Superscripts, special symbols v volumetric strain * rheological strain * at saturation conditions or hi average, phase average a or hia intrinsic phase average ~ spatial deviation variable

XXIX

XXXI

EFCE Working Party on Drying: Address List Dr. Odilio Alves-Filho Norwegian University of Science and Technology Department of Energy and Process Engineering Kolbjørn Hejes vei 1B 7491 Trondheim Norway odilio.fi[email protected] Prof. Julien Andrieu (delegate) UCB Lyon I/ESCPE LAGEP UMR CNRS 5007 batiment 308 G 43 boulevard du 11 novembre 1918 69622 Villeurbanne cedex France [email protected] Dr. Ir. Paul Avontuur Glaxo Smith Kline New Frontiers Science Park H89 Harlow CM19 5AW UK [email protected] Prof. Christopher G. J. Baker Drying Associates Harwell International Business Centre 404/13 Harwell Didcot Oxfordshire OX11 ORA UK [email protected]

Prof. Antonello Barresi (delegate) Politecnico di Torino Dip. Scienza dei Materiali e Ingegneria Chimica Corso Duca degli Abruzzi 24 10129 Torino Italy [email protected] Dr. Rainer Bellinghausen (delegate) Bayer Technology Services GmbH BTS-PT-PT-PDSP Building E 41 51368 Leverkusen Germany rainer.bellinghausen@bayertechnology. com Dr. Carl-Gustav Berg Abo Akademi Process Design Laboratory Biskopsgatan 8 20500 Abo Finland cberg@abo.fi Dr. Catherine Bonazzi (delegate) AgroParisTech – INRA JRU for Food Process Engineering 1, Avenue des Olympiades 91744 Massy cedex France [email protected]

XXXII

EFCE Working Party on Drying: Address List

Paul Deckers M.Sc. (delegate) Bodec, Process Optimization and Development Industrial Area ‘t Zand Bedrijfsweg 1 5683 CM Best The Netherlands [email protected]

Prof. Dr. Istvan Farkas (delegate) Szent Istvan University Department of Physics and Process Control Pater K. u. 1. 2103 Godollo Hungary [email protected]

Prof. Stephan Ditchev University of Food Technology 26 Maritza Blvd 4002 Plovdiv Bulgaria [email protected]

Dr.-Ing. Dietrich Gehrmann Wilhelm-Hastrich-Str. 12 51381 Leverkusen Germany [email protected]

Dr. German I. Efremov Pavla Korchagina 22 129278 Moscow Russia [email protected] Prof. Trygve Eikevik Norwegian University of Science and Technology Department of Energy and Process Engineering Kolbjørn Hejes vei 1B 7491 Trondheim Norway [email protected] Dr.-Ing. Ioannis Evripidis Dow Deutschland GmbH & Co. OHG P.O. Box 1120 21677 Stade Germany [email protected]

Prof. Dr.-Ing. Adrian-Gabriel Ghiaus (delegate) Technical University of Civil Engineering Thermal Engineering Department Bd. P. Protopopescu 66 021414 Bucharest Romania [email protected] Prof. Dr.-Ing. Gheorghita Jinescu University “Politehnica” din Bucuresti Faculty of Industrial Chemistry, Department of Chemical Engineering 1, Polizu street Building F Room F210 78126 Bucharest Romania [email protected] Prof. Dr. Gligor Kanevce St. Kliment Ohridski University Faculty of Technical Sciences ul. Ivo Ribar Lola b.b. Bitola FYR of Macedonia [email protected]

EFCE Working Party on Drying: Address List

Prof. Dr. Markku Karlsson (delegate) UPM-Kymmene Corporation P.O. Box 380 00101 Helsinki Finland [email protected] Ir. Ian C. Kemp (delegate) GMS, GSK Priory Street Ware, SG12 0XA UK [email protected] Prof. Dr. Ir. P.J.A.M. Kerkhof Eindhoven University of Technology Department of Chemical Engineering P.O. Box 513 5600 MB Eindhoven The Netherlands [email protected] Prof. Matthias Kind Universität Karlsruhe (TH) Institut für Thermische Verfahrenstechnik Kaiserstr. 12 76128 Karlsruhe Germany [email protected] Prof. Eli Korin Ben-Gurion University of the Negev Chemical Engineering Department Beer-Sheva 84105 Israel [email protected] Emer. Prof. Ram Lavie Technion – Israel Institute of Technolgy Department of Chemical Engineering Technion City Haifa 32000 Israel [email protected]

Dr. Ir. Angélique Léonard (delegate) Université de Liège, Département de Chimie Appliquée Laboratoire de Génie Chimique Bâtiment B6c - Sart-Tilman 4000 Liège Belgium [email protected] Jean-Claude Masson RHODIA, Recherches et Technologies 85 avenue des Frères Perret BP 62 69192 Saint-Fons Cedex France [email protected] Prof. Natalia Menshutina Mendeleev University of Chemical Technology of Russia (MUCTR) High Technology Department 125047 Muisskaya sq.9 Moscow Russia [email protected] Jun.-Prof. Dr. Thomas Metzger Otto-von-Guericke-University Thermal Process Engineering P.O. Box 4120 39016 Magdeburg Germany [email protected]. de Prof. Antonio Mulet Pons (delegate) Universitat Politecnica de Valencia Departament de Tecnologia d’Aliments Cami de Vera s/n 46071 Valencia Spain [email protected]

XXXIII

XXXIV

EFCE Working Party on Drying: Address List

Prof. Zdzislaw Pakowski (delegate) Technical University of Lodz Faculty of Process and Environmental Engineering ul. Wolczanska 213 93-005 Lodz Poland [email protected] Prof. Patrick Perré (delegate, chairman of WP) AgroParisTech 14 Rue Girardet 54042 Nancy France [email protected] Dr. Roger Renström Karlstad University Department of Environmental and Energy Systems Universitetsgatan 2 65188 Karlstad Sweden [email protected] Prof. Michel Roques Université de Pau et des Pays de l’Adour ENSGTI, 5 rue Jules- Ferry 64000 Pau France [email protected] Dr. Carmen Rosselló (delegate) University of Iles Baleares Dep. Quimica Ctra Valldemossa km 7.5 07122 Palme Mallorca Spain [email protected]

Emer. Prof. G. D. Saravacos (delegate) Nea Tiryntha 21100 Nauplion Greece [email protected] Dr.-Ing. Michael Schönherr BASF, GCT/T - L 540 Research Manager Drying Process Engineering 67056 Ludwigshafen Germany [email protected] Prof. Dr.-Ing. Ernst-Ulrich Schlünder Lindenweg 10 76275 Ettlingen Germany [email protected] Dr. Alberto M. Sereno (delegate) University of Porto Department of Chemical Engineering Rua Dr Roberto Frias 4200-465 Porto Portugal [email protected] Dr. Milan Stakic Vinca Institute for Nuclear Sciences Center NTI P.O. box 522 11001 Belgrade Serbia [email protected] Prof. Stig Stenstrom (delegate) Lund University Institute of Technology Department of Chemical Engineering P.O. Box 124 22100 Lund Sweden [email protected]

EFCE Working Party on Drying: Address List

Prof. Ingvald Strommen (delegate) Norwegian University of Science and Technology Department of Energy and Process Engineering Kolbjørn Hejes vei 1b 7491 Trondheim Norway [email protected] Prof. Czeslaw Strumillo (delegate) Technical University of Lodz Faculty of Process and Environmental Engineering Lodz Technical University ul. Wolczanska 213 93-005 Lodz Poland [email protected] Prof. Radivoje Topic (delegate) University of Belgrade Faculty of Mechanical Engineering 27, marta 80 11000 Beograd Serbia [email protected] Prof. Dr.-Ing. Evangelos Tsotsas (delegate, former chairman of WP) Otto-von-Guericke-University Thermal Process Engineering P.O. Box 4120 39016 Magdeburg Germany [email protected]. de

Dr. Henk C. van Deventer (delegate) TNO Quality of Life P.O. Box 342 7300 AH Apeldoorn The Netherlands [email protected] Michael Wahlberg M.Sc. Niro Gladsaxevej 305 2860 Soeborg Denmark [email protected] Prof. Roland Wimmerstedt Lund University Institute of Technology Department of Chemical Engineering P.O. Box 124 22100 Lund Sweden [email protected] Dr. Bertrand Woinet (delegate) SANOFI-CHIMIE, CDP bât. 8600 31-33 quai armand Barbès 69583 Neuville sur Sa^ one cedex France bertrand.woinet@sanofi-aventis.com Prof. Ireneusz Zbicinski Lodz Technical University Faculty of Process and Environmental Engineering ul. Wolczanska 213 93-005 Lodz Poland [email protected]

XXXV

j1

1 Quality Changes in Food Materials as Influenced by Drying Processes Catherine Bonazzi and Elisabeth Dumoulin 1.1 Introduction

Drying and dewatering plays a major role in food manufacturing or food processing activities worldwide. Often one of the last operations in the food processing, it controls to a large extent the quality of the final product. Drying is applied to a wide variety of food products, from cereals to finished goods, from raw materials to byproducts. The processes used are numerous, according to the type and quantity of product to dry, the amount of water to eliminate, the final desired quality or functionality of the dried product (Tab. 1.1). Drying and dewatering impact the mechanical, sensory and nutritional properties of food products, and can be used to create new functionalities (Bonazzi and Bimbenet, 2003, 2008). Drying is one of the main techniques for preserving agricultural and food products; it takes place in the processing of many products, as the main operation or as a consequence of other processing steps. Heat and mass transfer phenomena which are typical of drying also appear during other processes, as in cooking, baking, roasting, smoking, refrigeration, freezing, during storage, and during pneumatic transportation. The main objective of drying is to decrease the water activity (aw) of various perishable materials to values 75% reference length Length 3 mm, the curve Rp ¼ f(Ldry) was no longer linear. This point corresponds to the limit beyond which the total sublimation area inside the vials certainly decreases (wall effect due to radiation flux). Nevertheless, the same authors observed that, without annealing treatment, the experimental values of Rp could not be predicted so accurately by the same model. This discrepancy possibly results from wider ice crystal size distributions in samples without annealing. A significant homogenization of morphology certainly occurred in the case of cycles with annealing, leading to more uniform pore sizes all through the bulk of the freezedried matrix which is in accordance with the hypothesis of Eq. 3.2. Moreover, for standard formulations used industrially in vaccine freeze-drying, Chouvenc et al. (2004b) observed that the mass transfer resistance values increased more or less monotonically as a function of the dried layer thickness, and that an annealing

j75

j 3 Characterization and Control of Physical Quality Factors During Freeze-Drying

76

Rp (Pa m2 s kg–1)

1000000

800000

600000

400000

200000

0 0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

Dried matter thickness x 102 (m) Fig. 3.19 Water vapor mass transfer resistance versus freeze-dried layer thickness. Continuous line ¼ model prediction, according to Hottot et al. (2005).

treatment after the freezing step was an efficient method to homogenize and improve the permeability of the dried zone, thus reducing the primary drying times. 3.3.4.2 Freeze-Dried Layer Permeability The concept of permeability introduces another overall parameter often used to characterize globally the morphology of porous media with respect to gas flow. Thus, for interpreting the sublimation data during freeze-drying processes, the water vapor permeability, denoted by K, is classically defined by the following relationship (Darcy’s law): _ ¼ m

~w KM gradP ~ RT

ð3:3Þ

Equation 3.3 shows that K is inversely proportional to the water vapor mass transfer resistance, previously defined by Eq. 3.1. From the above relationship, it is possible to estimate the experimental values of K _ over the from experimental values of the primary drying rate, denoted by m, sublimation period corresponding to the freeze-dried layer thickness, Ldry, by the following relationship: Kexp ¼

~ s Ldry RT _
m ~w Ps Pc M

ð3:4Þ

Nakagawa et al. (2007a) interpreted the dependence of the water vapor mass transfer permeability of the dried layer on ice crystal morphologies by using this relationship for different nucleation temperatures. The increase in the nucleation temperature was expected to increase the water vapor mass transfer permeability of freeze-dried materials and to contribute to acceleration of the primary sublimation rates during freeze-drying.

3.3 Influence of Freezing and Freeze-Drying Parameters on Physical Quality Factors 1.5

2.5

Experiment

Experiment

Calculation

Calculation

2.0

dried layer thickness Ldry = 3 mm

K [10–3 m2 s–1]

K [10–3 m2 s–1]

j77

1.0

0.5

dried layer thickness Ldry = 4 mm

1.5

1.0

0.5 Mannitol

0

–14

–12

–10

–8

–6

–4

–2

BSA

0

0

–14

Tn [ ºC]

–12

–10

–8

–6

Tn [ ºC]

Fig. 3.20 Comparison of experimental and estimated dried layer permeability values according to Nakagawa et al. (2007a).

The experimental values of permeability, Kexp, plotted in Fig. 3.20 correspond to mean values of dried layer thickness, Ldry, of about 3 and 4 mm for the mannitol and the BSA systems, respectively. On the other hand, the dried cake permeability values were theoretically estimated by the same authors from values of the mean size of the ice crystals by assuming that the freeze-dried cake texture could be represented by a bundle of capillary tubes (mean diameter, denoted dpore ). Under conditions of molecular flow in the Knudsen regime (Kn ¼ 4) this theoretical freeze-dried layer permeability, denoted by Kmodel, was estimated by the following equations: e Kmodel ¼ Dk V t

ð3:5Þ

where the water vapor Knudsen diffusivity, Dk, was expressed by the following relationship: sffiffiffiffiffiffiffiffiffiffi ~ s 1 8RT
d ð3:6Þ Dk ¼ ~ w pore 3 pM Here, V represents a total flow contribution factor, equal to: V¼

1 1 þ dpore =l

ð3:7Þ

where l represents the mean free path of the gas molecules under selected sublimation conditions. For this estimation, Nakagawa et al. (2007a) assumed that the mean ice crystal size corresponded to the pore diameter, dpore in Eqs. 3.6 and 3.7. The form factor e/t, in Eq. 3.5, has been considered as an empirical constant and was identified from a set of experimental data. The experimental and calculated values, denoted by Kexp and

–4

–2

0

j 3 Characterization and Control of Physical Quality Factors During Freeze-Drying

78

Kmodel, respectively, plotted in Fig. 3.20, show pretty good agreement. These data clearly show that the water vapor mass transfer resistance decreases as the nucleation temperature increases, so that the drying process can be accelerated by increasing the nucleation temperature. This tendency is in logical agreement with the ice crystal size estimation obtained both from image analysis and from the model prediction. Consequently, it is confirmed that, for a given system, the mass transfer phenomena during the sublimation step of the freeze-drying process are significantly dependent on the morphology of the ice crystals and, in consequence, significantly dependent on the nucleation temperatures of the supercooled formulation. 3.3.5 Importance of Temperature Control

Freeze-drying is a mild drying process during which the temperature profile of the sample all through the freezing step and the two subsequent drying steps has great importance, not only for the operating costs of the whole process – costs that are generally pretty high due to very low total gas pressure and the temperatures involved for fragile and thermosensible drugs – but also for the many quality attributes required by the stringent quality standards of such products. As the sample is cooled, supercooling (about 10–15 K) below the equilibrium freezing point takes place before nucleation and crystal growth can occur, accompanied by a rise in the temperature of the solution up to the equilibrium freezing point (cf. Fig. 3.3). Then, the sample product temperature decreases slowly until freezing of most of the free water is completed and, then, it decreases rapidly down to the shelf temperature. The ice crystal growth induces an increase in viscosity and concentration of the cryo-concentrated solution. Other solutes of the formulation, such as sodium chloride, buffer salt components or other low molecular inorganic excipients, may also crystallize. Generally the drug (API) itself, the organic components and the high molecular excipients do not crystallize; they stay amorphous and diluted in a glassy phase containing some amount of unfrozen water. A large part of this unfrozen water is separated during the last process step by means of diffusion inside the amorphous phase and/or by desorption from the pore surfaces of the freeze-dried matrix. As previously indicated, primary and secondary drying steps are conducted at very low total gas pressure (5–60 Pa) and at low sublimation front temperatures ( 10 to 40  C) during the sublimation step, whereas the product temperature is increased above the ambient temperature during the secondary drying step (usually to levels between 20 and 40  C in the case of thermosensible drugs). The main objective of these two drying steps is to generate a form of the labile drug that is stable for many months. The drug is normally a parenteral product that has to be rehydrated before injection. The therapeutic or biological activity of the final product (vaccines, for example) is quite sensitive to the product temperature history from the first step (freezing) up to the last step (desorption) when the temperature is strongly increased. The rate of sublimation increases rapidly with the sublimation front (or product) temperature by about a factor of two for temperature increases of

3.3 Influence of Freezing and Freeze-Drying Parameters on Physical Quality Factors

6 K. Process efficiency (primary and secondary drying times) is directly related to these main operating parameters. It must be borne in mind that too high processing temperatures will compromise several important quality factors (therapeutic or biological activity, mechanical structure resistance, porosity, appearance, crack formation, etc.). Many experimental studies have shown that, for systems which contain one solute that gives a crystalline phase in solid state, the primary drying step must be conducted below the eutectic temperature. Nevertheless, most drug formulations do not crystallize or have high eutectic temperatures so that the eutectic temperature limit is not very often an issue in freeze-drying cycle optimization, if we except the systems containing high concentrations of NaCl, this salt having a eutectic temperature around 20  C. Otherwise, with amorphous or glassy systems, the sublimation step must be carried out below the vitreous transition temperature or below the collapse temperature, which is generally a few degrees above the vitreous transition temperature of the system. If the product is sublimed above this limiting vitreous transition temperature, the solid phase collapses, because it has enough fluidity to induce a viscous flow when the ice is removed from the frozen sample. This collapse will lead to a loss of mechanical resistance of the dried cake structure and will result in higher moisture content and in a loss of biological activity with some protein formulations (vaccines). Some other defects like an increase in the rehydration times or some loss of pharmaceutical performance and elegance of the cake can also occur. 3.3.6 Influence of Operating Conditions on Sublimation Kinetics

For many materials (gels, ceramics, wood, etc.), it is well known that the drying kinetics curves are very helpful for understanding the influence of the main operating parameters (temperature, total gas pressure) on many quality factors like color changes, dried material shrinkage, risk of crack formation, crust formation (solute crystallization, for example in liquid containing sugars), the formation of a vaportight skin of polymers (for example during the reactive drying of paintings), and so on. Thus, examination of the drying curves associated with the temperature profiles inside the product constitute an important tool and a pertinent guide for setting up and optimizing the drying conditions of many types of manufactured products, including pharmaceuticals or drugs during lyophilization cycles. Moreover, the use of drying curves enables detection of the main governing mechanisms that control the heat and mass transfer phenomena and, in this way, enables one to determine the main input physical variables for setting up and validating software for the control and monitoring of the freeze-dryer (shelf temperature or total gas pressure). The influence of shelf temperature and of sublimation chamber total gas pressure  (t)) and on the drying rate curve ( dX ðtÞ) is discussed below. on the drying curve (X dt As observed by many authors during the drying of different types of materials, for example during the convective or infrared reactive drying of thin coating films, the

j79

j 3 Characterization and Control of Physical Quality Factors During Freeze-Drying

80

classical drying curves (moisture content, drying rates, mean product temperature) as a function of time exhibit roughly three drying periods: .

.

.

An acceleration period during which the sublimation rate increases continuously (transient sublimation state). A quasi-stationary period during which the sublimation rate is quite constant. This corresponds to a quasi-steady state with an approximately constant sublimation front temperature. The duration of this plateau period seems to decrease as the sublimation temperature increases. The experimental data presented in Fig. 3.21 show that the sublimation rate increases by about a factor of two for an increase of 20  C in the shelf temperature (from 25 to 5  C). A decreasing drying rate period where the drying rates decrease rapidly with a simultaneous monotonic and significant increase in the mean product temperature.

Sublimation rate (–dX(t)/dt).105 s–1

Furthermore, it is quite surprising to observe that the scientific literature on freezedrying in vial configuration contains little data concerning such drying curves. Nevertheless, some data are available (Genin et al., 1996) and show that a shelf temperature increase from 0 to 25  C increases the drying rate by a factor of two; the same drying rate increase (factor of two) was obtained by increasing the total gas pressure by a factor of 100. Additionally, some experimental studies have been carried out by using a microbalance placed inside the freeze-drying chamber (Roth et al., 2001; Jun et al., 2004). The sublimation data reported by these authors show three desiccation periods. These data also indicate the importance of vial package density which affects the values of the sublimation rate. A coherent set of drying kinetics data concerning the freeze-drying of model pharmaceutical formulations in vials under standard operating conditions, namely at very low total gas pressure (10–26 Pa) and at low sublimation temperature ( 5 to 25  C) was published recently by Hottot et al. (2007). A typical set of sublimation rate data gained by these authors is presented in Fig. 3.21 concerning the influence of

8.0 7.0 6.0 5.0 4.0 3.0

Tshelf = –5°C Tshelf = –15°C Tshelf = –25°C

2.0 1.0 0.0 0

5 10 X(t), kg water (kg dried matter)–1

15

Fig. 3.21 Sublimation rates for different shelf temperatures at P ¼ 18 Pa according to Hottot et al. (2007).

3.3 Influence of Freezing and Freeze-Drying Parameters on Physical Quality Factors

the shelf temperature. These data show that, for a constant total gas pressure (P ¼ 18 Pa), a plateau appears at mean water content values located in the range 5 kg kg 1 < X(t) < 13 kg kg 1. Moreover, the sublimation rates corresponding to the height of this plateau increase with the shelf temperature, which means that they also increase with the sublimation front temperature. Further experimental results by the same authors for the two other pressures investigated (10 and 26 Pa) show similar trends. For the standard sublimation conditions of Hottot et al. (2007), the domain of water content with an approximately constant drying rate seems to correspond to a stationary state of sublimation, as generally assumed in simple front models. Such a model has been implemented by the same authors for the purpose of validation by comparing with measured sublimation kinetics (Hottot et al., 2006b). In total, these experimental data show that, under standard freeze-drying conditions for very thermosensible drugs, the sublimation kinetics are mainly dependent on the sublimation front temperature which is fixed by the shelf temperature. Therefore, shelf temperature is the key input parameter for the control of important textural (or physical) quality features and of therapeutic activity factors. The set of data plotted in Fig. 3.22 shows only a slight influence of the chamber total gas pressure on the sublimation kinetics. These observations are in agreement with the results reported by Kuu et al. (2005) who proposed an interesting approach to transferring laboratory scale freeze-drying data to industrial scale conditions. The same authors observed that the chamber total pressure only has a significant influence on the drying kinetics when the shelf temperature is maintained below 15  C, see also Trappler (2001); Rambhatla and Pikal (2003). Consequently, we conclude that the sublimation process under standard conditions of very low temperature and total gas pressure, as encountered in thermosensible drugs freeze-drying, is mainly governed by the overall heat transfer flux to the sublimation

Sublimation rate, (-dX(t)/dt).105 s–1

6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0

P = 18 Pa P = 10 Pa P = 26 Pa

1.5 1.0 0.5 0.0 0

5

10

15

20

X(t), kg water (kg dried matter)–1 Fig. 3.22 Sublimation rates for different total gas pressures at Tshelf ¼ 15  C, according to Hottot et al. (2007).

j81

j 3 Characterization and Control of Physical Quality Factors During Freeze-Drying

82

front, namely by the heat conduction flux from the plate through the vial bottom and also by the radiation flux coming from the vial surroundings, specifically from the sidewalls of the sublimation chamber and from the upper shelf. The previously discussed experimental data have recently been extended to pharmaceutical formulations with organic co-solvent – active principle ingredient (API) þ water þ tert-butyl alcohol (TBA) – by Daoussi et al. (2009) who used an automatic microbalance placed inside the product chamber to determine the sublimation kinetics. The same shapes of sublimation kinetics curves were observed. The strong influence of the shelf temperature, regardless of the total gas pressure, proves that the sublimation process under conditions of very low total gas pressure and temperature is still mainly controlled by heat transfer, as in the case of aqueous formulations (Fig. 3.23). The sublimation rate was not a monotically increasing function of total gas pressure but reached a maximal value at the intermediate total gas pressure investigated (15 Pa) then decreased. Daoussi et al. (2009) observed that the use of organic co-solvent reduced considerably the sublimation times to around 3 h, these values being 10 to 11 times lower than the values observed with pure aqueous formulations under the same operating conditions of temperature and pressure. Simultaneously, the organic co-solvent allowed preservation of the main quality factors of the freeze-dried product (residual solvent and water contents, visual appearance, color, cake permeability, rehydration times, stability, etc.). This proves that formulations with organic co-solvent lead to sublimation times comparable with drying times observed with agitated vacuum contact dryers; consequently, these organic formulations, if suitable for stabilizing the investigated active principle (API), can significantly reduce the operating costs of freeze-drying compared with those for pure aqueous systems.

Shelf temperature = –10°C Shelf temperature = –20°C Shelf temperature = –30°C Shelf temperature = –40°C

dX/dt (s–1)

0.0015

0.001

0.0005

0 0

5

10

15

20

25

30

X (kg kg–1) Fig. 3.23 Drying kinetics as a function of total solvent (water þ TBA) content at different shelf temperatures; formulation at 20% water þ 80% TBA (by mass). Influence of shelf temperature. According to Daoussi et al., 2009.

3.4 Product Quality and Stability During Drying and Storage

3.4 Product Quality and Stability During Drying and Storage 3.4.1 Product Quality and Formulation

The final product quality attributes are strongly related to the level of optimization of the composition of the liquid formulation, which is a multidisciplinary and challenging problem usually solved by tedious experimental approaches in the development of freeze-drying cycles. The topics of formulation that involve complex knowledge of physical chemistry, biochemistry and biology are outside the scope of this chapter. However, general concepts that are generally applied are summarized hereafter. The mechanisms that govern product stability during storage are the same as those involved during the drying steps, the only difference being the water content and the time scales for the two phenomena. For example, in the case of pharmaceutical proteins, the role of the stabilizer is to slow down the degradation reactions by introducing high activation energy barriers which prevent the formation of significant amounts of denatured protein forms. Two main concepts are generally proposed to explain the role of excipients in promoting stabilization effects during the two drying steps and during the storage step: the vitrification hypothesis and the water substitution hypothesis (Pikal, 1999a). The “vitrification” concept relies on the experimental observation that efficient stabilizers are always convenient glass-former materials with very high viscosity at temperatures below the glass transition temperature. Several experimental studies have pointed out a sharp decrease in mobility (chain fluctuation and chain rotation) and a consequent decrease in reactivity when the glass transition temperature is approached. The “water substitute” concept states that the stabilizer replaces the water and forms hydrogen bonds with the protein, like water also does. These bonds maintain the native protein conformation and, in this way, stabilize the protein. The water substitution hypothesis has been confirmed either by spectroscopy studies (FTIR), or by experimental observations that have shown that the most efficient stabilizers are sugars which form strong hydrogen bonds with the protein. Indeed, freeze-dried systems with loss of activity present notably altered IR spectra. Nevertheless, some exceptions exist and not all additives which are good stabilizers show chemical and bonding properties similar to water. Many experimental data have shown that the stabilization of fragile and thermosensible pharmaceutical proteins requires a lyoprotectant that is a good glass-former, leading to a single amorphous solid phase, with a moderate interaction with the protein surface to avoid phase separation (crystallization, etc.). As already indicated, nucleation temperature, nucleation rates and solvent crystal growth rates are of crucial importance for the morphology and for the structure of the resulting freeze-dried product. Freeze-drying generally leads to an

j83

j 3 Characterization and Control of Physical Quality Factors During Freeze-Drying

84

amorphous phase (e.g., with sugars) that often crystallizes during the secondary drying period (Pikal, 1999b). Moreover, lyophilization can also produce metastable crystalline forms directly from aqueous solutions upon freezing, with formation of an amorphous solid during secondary drying. This crystallization behavior depends on multiple factors, such as the freezing rate, the nature and concentration of the solutes and the nature and concentration of the additives (cryoprotectants, lyoprotectants, etc.). 3.4.2 Product Quality and Polymorphism

The formation of polymorphs and various hydrates can also affect many quality factors and must be carefully considered in stability criteria (Pikal, 1999b; Morris et al., 2001). Many experimental results have shown that crystallization from aqueous solutions is favored by higher concentrations of crystallizable solute. On the contrary, the presence of high concentrations of other solutes that remain amorphous generally prevents the crystallization of the main possible crystallizable solute. For example, the crystallization of mannitol, which is easy if this solute is quite pure, becomes improbable in the presence of other solutes in high concentration. Polymorphism, which is a frequent and well-known phenomenon in drug crystallization, has a direct influence on the physical and chemical properties of the freeze-dried cake and, consequently, can impact drastically the manufacturing aptitude and the performances of dosage forms (tablets or capsules). Polymorphism has a direct influence on the dissolution kinetics variations and, thus, it also influences the bioavailability and the stability of the API. Due to the complexity of phase changes during freeze-drying – they are related to formulation, process and storage conditions – these problems are often ignored in the pharmaceutical R&D and industry. This may have important consequences because the process is usually long and expensive and, thus, undesired phase changes can greatly modify the processing times and the final therapeutic or biological activity of the freeze-dried products (for example, with pharmaceutical proteins). In some cases, drug crystallization significantly increases its stability. Conversely, the crystallization of one excipient – especially the crystallization of a buffer or a stabilizer in protein formulations – leads generally to a significant loss of stability (Pikal, 1999b). On the other hand, crystallization of poorly soluble drugs may increase the reconstitution times or even lead to incomplete dissolution, due to the better solubility of the amorphous phases in comparison to the respective crystalline forms. Crystallization during the storage step is also sensitive to the residual water content and to the storage temperature. Generally, recrystallization is more probable at storage temperatures higher than the vitreous transition temperature, but with some systems, and for long storage times, recrystallization can also occur at temperatures below the glass transition temperature.

3.5 Conclusions

Crystallization can also exert a major impact on the freeze-drying cycle duration, since it modifies the composition of the solute phase and, thereby the value of Tg0 . This influence can be positive or negative, depending on when crystallization occurs and on what component crystallizes: an increase in Tg0 , as would be produced by crystallization of inorganic salts, would allow sublimation at higher temperatures and thus result in shorter sublimation times with a positive effect on operating costs. On the contrary, a decrease in Tg0 would require that the sublimation step be run at lower temperatures, leading to longer sublimation times (negative impact). It is well known that mannitol, which is a commonly used excipient as a bulking agent, can lead to vial breakage problems if its crystallization occurs during the primary drying step, so that the crystallization of mannitol from API formulations has to take place during the freezing step to avoid this problem (Pikal, 1999b, 2002).

3.5 Conclusions

The strong coupling between final quality parameters and transport phenomena poses serious challenges for the optimization of the drying cycles used in modern industry for manufacturing many products and materials, particularly in the pharmaceutical industry. The understanding of these relationships is still more crucial in the case of freeze-drying processes of fragile and very thermosensible materials, such as pharmaceuticals and drugs. Complex and interdependent phenomena of phase transition (ice nucleation, vitreous transition of cryoconcentrated phase, polymorphism) occur during freeze-drying, inducing sharp changes in the morphological and rheological properties of the material. The stresses generated during the freezing and the drying steps modify the native conformational structure of thermosensible active principles like pharmaceutical proteins and, in this way, they also modify their therapeutic activity. Removal of water by two-phase changes (crystallization, sublimation) induces more or less shrinkage, which changes the volume fraction of the phases and thus modifies the values of the main parameters of heat and mass transport that takes place under severe conditions (low temperature and high vacuum). Transport parameters are not generally available in the literature. Furthermore, the final therapeutic activity and stability of the drug during the storage period are strongly dependent on the residual moisture content, on eventual moisture gradients and on the temperature history of the product during the whole freeze-drying process. These problems can be generally solved by tedious experimental approaches. As explained in this chapter, the freezing and drying requirements for the control of the main quality factors are strongly coupled and related to the system formulation, namely to the state diagram data, which represent the key data in setting up the lyophilization cycles at the pilot and then industrial scales. As a consequence, fine and advanced physical modeling of these freezing and drying steps, with enhanced

j85

j 3 Characterization and Control of Physical Quality Factors During Freeze-Drying

86

computer facilities, can, in the future, play an important role and lead to a more scientific and more rational methodology for the control of these physical quality factors involved in the optimization of freeze-drying processes of fragile pharmaceuticals. Consequently, a common effort should be made to establish and complete the data basis concerning the numerous thermodynamic, thermophysical, transport and rheological properties necessary for modeling and simulating the different steps of this complex mild drying process. Moreover, the methods available for the characterization of most of these end-use properties should be improved and additional characterization methods should be adapted from other research fields (material science, applied biochemistry, physical chemistry, etc.). Improved characterization methods promise a better description and a safer control of numerous end-use properties for existing freeze-dryers as well as for new machines with more possibilities to better comply with more and more severe quality requirements in the future. The use of non-invasive sensors or of rapid nonintrusive methods for on-line and in situ estimation of the main parameters of the process could also help to overcome the difficulties observed, for example, the artifacts resulting from invasive sensors inserted inside vials and presently commonly used.

Additional Notation Used in Chapter 3

h Kv K Rp T 0g Tg Tn Ts

solution height inside the vial overall heat transfer coefficient dried layer permeability water (solvent) vapor mass transfer resistance glass transition temperature at maximum concentration of the cryo-concentrated phase glass transition temperature temperature nucleation temperature sublimation front temperature

m Wm 2K 1 m2 s 1 Pa m2 s kg 1 K,  C K,  C K,  C K

Greek letters

l t V

mean free path of gas molecules tortuosity factor total flow contribution factor

Subscripts

c i s p

chamber interface sublimation product

m

References

Abbreviations

API BSA DSC FTIR IR MDSC MTM NMR PAR US SEM

active principle ingredient bovine serum albumin differential scanning microscopy Fourier transform infra-red infra-red modulated scanning microscopy manometric temperature measurement nuclear magnetic resonance pressure rise analysis ultrasound scanning electron microscopy

References Bardat, A., Biguet, J., Chatenet, E., Courteille, F., 1993. Moisture measurement: A new method for monitoring freeze-drying cycles. J. Parenteral Sci. Tech. 47(6): 293–299. Berger, K. G., White, G. W., 1971. An electron microscopical investigation of fat destabilisation in ice cream. J. Food Technol. 6: 285–294. Caillet, A., Cogne, C., Andrieu, J., Laurent, P., Rivoire, A., 2003. Characterization of the structure of ice cream by optical microscopy: Influence of freezing parameter on ice crystal structure. Lebensm. -Wiss. u. -Technol. 36: 743–749. Chouvenc, P., Vessot, S., Andrieu, J., Vacus, P., 2004a. Optimization of the freeze-drying cycle: A new model for pressure rise analysis. Drying Technol. 22(7): 1577–1601. Chouvenc, P., Vessot, S., Andrieu, J., Vacus, P., 2004b. Optimization of the freeze-drying cycle: Characterization of annealing effects by the pressure rise analysis method. Proceedings of 14th International Drying Symposium (IDS 2004), Sao Paulo, Vol. A, 359–365. Chouvenc, P., Vessot, S., Andrieu, J., 2006. Experimental study of the impact of annealing on ice structure and mass transfer parameters during freeze-drying of a pharmaceutical formulation. PDA J. Pharm. Sci. Tech. 60(2): 95–103.

Chow, R., Blindt, R., Chivers, R., Povey, M., 2003. The sonocrystallization of ice in sucrose solutions: Primary and secondary nucleation. Ultrasonics 41 (8): 595–604. Chow, R., Blindt, R., Chivers, R., Povey, M., 2005. A study on the primary and secondary nucleation of ice by power ultrasound. Ultrasonics 43(4): 227–230. Daoussi, R., Vessot, S., Andrieu, J., Monnier, O., 2009. Sublimation kinetics and sublimation end-points during freeze-drying of pharmaceutical active principle with organic co-solvent formulations. Chem. Eng. Res. Des., 87: 899–907. Faydi, E., Andrieu, J., Laurent, P., 2001. Experimental study and modelling of the ice crystal morphology of model standard ice cream, Part I: Direct characterization method and experimental data. J. Food Eng. 48: 283–291. Franks, F., 1990. Freeze-drying: from empiricism to predictability. Cryo-Lett 11: 93–110. Genin, N., Rene, F., Corrieu, G., 1996. A method for on line determination of residual water content and sublimation end point during freeze drying. Chem. Eng. Process. 35: 255–263. Hickling, R., 1965. Nucleation of freezing by cavity collapse and its relation to cavitation damage. Nature 206: 915–917.

j87

j 3 Characterization and Control of Physical Quality Factors During Freeze-Drying

88

Hottot, A., Vessot, S., Andrieu, J., 2004. A direct characterization method of the ice morphology: Relationship between mean crystals size and primary drying times of freeze-drying processes. Drying Technol. 22: 2009–2021. Hottot, A., Vessot, S., Andrieu, J., 2005. Determination of mass and heat transfer parameters during freeze-drying cycle of pharmaceutical products. PDA J. Pharm. Sci. Tech. 59: 138–153. Hottot, A., Vessot, S., Andrieu, J., 2006a. Freeze-drying of pharmaceuticals in vials: Influence of freezing protocol and sample configuration on ice morphology and freezedried cake texture. Chem. Eng. Process. 46: 666–674. Hottot, A., Peczalski, R., Vessot, S., Andrieu, J., 2006b. Freeze-drying of pharmaceutical proteins in vials: Modeling of freezing and sublimation steps. Drying Technol. 24: 561–570. Hottot, A., Andrieu, J., Vessot, S., 2007. Sublimation kinetics during freeze-drying of pharmaceutical protein formulation. Drying Technol. 25: 1–6. Hottot, A., Nakagawa, K., Andrieu, J., 2008. Effect of ultrasound-controlled nucleation on structural and morphological properties of freeze-dried manitol solution. Chem. Eng. Res. Des. 86: 193–200. Hottot, A., Andrieu, J., Vessot, S., Shalaev, E., Gatlin, L. A., Ricketts, S., 2009a. Experimental study and modelling of freeze-drying in syringe configuration, Part I: Freezing step. Drying Technol. 27(1): 40–48. Hottot, A., Andrieu, J., Hoang, V., Shalaev, E. Y., Gatlin, L. A., Rickett, S., 2009b. Experimental study and modelling of freeze-drying in syringe configuration, Part II: Mass and heat transfer parameters and sublimation end-points. Drying Technol. 27(1): 49–58. Inada, T., Zhang, X., Yabe, A., Kozawa, Y., 2001. Active control of phase change from supercooled water to ice by ultrasonic vibration 4, Part 1: I Control of freezing temperature. Int. J. Heat Mass Transfer 44: 4523–4531. Jun, X., Jeffery, H., Volker, L., Wang, D. Q., 2004. Investigation of freeze-drying sublimation rates using a freeze-drying

microbalance technique. Int. J. Pharm. 279 (1–2): 95–105. Kochs, M., Korber, I., Heschel, B., Nunner, B., 1993. The influence of the freezing process on vapour transport during sublimation in vacuum freeze-drying of macroscopic samples. Int. J. Heat Mass Transfer 34: 1727–1738. Kuu, W. Y., Hardwick, L. S., Akers, M. J., 2005. Correlation of laboratory and production freeze drying cycles. Int. J. Pharm. 302: 56–67. Mahdjoub, R., Chouvenc, P., Seurin, M. J.L.,, Andrieu, J., Briguet, A., 2006. Sucrose solution freezing studied by magnetic resonance imaging. Carbohydr. Res. 341: 492–498. Mckenzie, A. P., 1975. Collapse during freezedrying: Qualitative and quantitative aspects, in Freeze-drying and Advanced Food Technology (eds S. A. Golblith, L. Rey, W. W. Rothmayer). Academic Press, London, pp. 277–307. Milton, N., Pikal, M. J., Roy, M. L., Nail, S. L., 1997. Evaluation of manometric temperature measurement as a method of monitoring product temperature during lyophilization. PDA J. Pharm. Sci. Tech. 51(1): 7–16. Morris, K. R., Griesser, U. J., Eckhardt, C. J., Stepwell, J. G., 2001. Theorical approaches to physical transformations of active pharmaceutical ingredientds during manufacturing processes. Adv. Drug Deliv. Rev. 48: 91–114. Morris, J., Morris, G. J., Taylor, R., Zhai, S., Slater, N. K. H., 2004. The effect of controlled nucleation on ice structure, drying rate, and protein recovery in vials in a modified freeze-dryer. Cryobiology 49: 308–309. Nakagawa, K., Hottot, A., Vessot, S., Andrieu, J., 2006. Influence of controlled nucleation by ultrasounds on ice morphology of frozen formulations for pharmaceutical proteins freeze-drying. Chem. Eng. Process. 45: 783–791. Nakagawa, K., Hottot, A., Vessot, S., Andrieu, J., 2007a. Modeling of freezing step during freeze-drying of drugs in vials. AIChE J. 53(5): 1362–1372. Nakagawa, K., Murakami, W., Andrieu, J., 2007b. Influence of freezing conditions on

References crystalline structure of mannitol. Proceedings of 16th International Drying Symposium (IDS 2008), Hyderabad, India, Vol. C, 1794–1800. Obert, J. P., 2001. Mod
elisation, optimisation et suivi en ligne du proc
ed
e de lyophilisation: Application  a l’am
elioration de la productivit
e et de la qualit
e des bact
eries lactiques lyophilis
ees. Diss. INRA Universit
e Paris-Grignon, France. Patapoff, T., Overcashier, D. E., Hsu, C. C., 1999. Lyophilization of proteins formulations in vials: Investigation of the relationship between resistance to water vapour flow during primary drying and small-scale product collapse. J. Pharm. Sci. 88(7): 688–695. Patapoff, T., Overcashier, D. E., 2002. The importance of freezing on lyophilization cycle development. Biopharm. 3: 16–22. Petersen, A., Rau, G., Glasmacher, B., 2006. Reduction of primary freeze-drying time by electric field induced ice nucleus formation. Heat Mass Transfer 42: 929–938. Pikal, M. J., Roy, M. L., Shah, S., 1984. Mass and heat transfer in vial freeze drying of pharmaceuticals: Role of the vial. J. Pharm. Sci. 73(9): 1224–1237. Pikal, M. J., 1985. Use of laboratory data in freeze-drying process design: Heat and mass transfer coefficients and the computer simulation of freeze drying. J. Parenteral Sci. Tech. 39(3): 115–139. Pikal, M. J., 1990. Freeze-drying of proteins, Part I: Process design. Biopharm. 3: 18–27. Pikal, M. J., 1992. Freeze-drying, in Encyclopedia of pharmaceutical technology. Vol. 6, (eds. J. Swarbrick and J.C. Boylan), Marcel Dekker, New York, pp. 275–303. Pikal, M. J., 1999a. Mechanisms of protein stabilization during freeze-drying and storage: The relative importance of thermodynamic stabilization and glassy state relaxation dynamics. Drugs and Pharmaceutical Science, 96: 161–198. Pikal, M. J., 1999b. Impact of polymorphism on the quality of lyophilised products, in Polymorphism in Pharmaceutical solids (ed. H. G. Brittazin). Marcel Dekker, New York, pp. 395–419. Pikal, M. J., 2002. Lyophilization, in Encyclopedia of Pharmaceutical Technology. (eds. J. Swarbrick and J.C. Boylan),Marcel Dekker, New York, pp. 1299–1326.

Rambhatla, S., Pikal, M. J., 2003. Heat and mass transfer scale up issues during freeze drying, I: A typical radiation and the edge vial effect. AAPS Pharm. Sci. Tech. 4–2(14): 1–10. Rambhatla, S., Ramot, R., Bhugra, C., Pikal, M. J., 2004. Heat and mass transfer scale-up issues during freeze-drying, II: Control and characterization of degree of supercooling. AAPS Pharm. Sci. Tech. 5(4): 1–8. Roth, C., Winter, G., Lee, G., 2001. Continuous measurement of drying rate of crystalline and amorphous systems during freeze-drying using an in situ microbalance technique. J. Pharm. Sci. 90 (9): 1345–1355. Roy, M. L., Pikal, M. J., 1989. Process control in freeze-drying: Determination of the end point of sublimation drying by an electronic moisture sensor. J. Parenteral Sci. Tech. 43 (2): 60–66. Saclier, M., Peczalski, R., Andrieu, J., 2008. Mod
elisation de la nucleation de cristaux de glace d
eclench
ee par ultrasons. R
ecents Progres en G
enie des Proc
ed
es, Ed. SFGP 97, (8 pages). Schneid, S., Gieseler, H., 2008. Evaluation of a new wireless temperature remote interrogation system (TEMPRIS) to measure temperature during freeze-drying. AAP Pharm. Sci. Tech. 9 (3): 729–739. Searles, J. A., Carpenter, J. F., Randolph, T., 2001a. The ice nucleation temperature determines the primary drying rate of lyophilization for samples frozen on a temperature-controlled shelf. J. Pharm. Sci. 90(7): 860–869. Searles, J. A., Carpenter, J. F., Randolph, T., 2001b. Annealing to optimize the primary drying rate, reduce freeze-induced drying rate heterogeneity, and determine T0g in pharmaceutical lyophilization. J. Pharm. Sci. 90(7): 872–887. Trappler, E., 2001. Scale-up strategy for a lyophilization process. Am. Pharm. Rev. 4 (3): 55–60. Woinet, B., Andrieu, J., Laurent, M., Min, S. G., 1998. Experimental and theoretical study of model food freezing, Part II: Characterization and modelling of the ice crystal size. J. Food Eng. 35(4): 395–407.

j89

j 3 Characterization and Control of Physical Quality Factors During Freeze-Drying

90

Zhai, Z., Taylor, R., Sanhes, R., Slater, N. K. H., 2003. Measurement of lyophilisation primart drying rates by freeze-drying microscopy. Chem. Eng. Sci. 58: 2313–2323. Zhang, X., Inada, T., Yabe, A., Lu, S. Kozawa, Y., 2001. Active control of phase change from supercooled water

to ice by ultrasonic vibration, II: Generation of ice slurries and effect of bubble nuclei. Int. J. Heat Mass Transfer 44: 4533–4539. Zhang, X., Inada, T., Tezuka, A., 2003. Ultrasonic-induced nucleation of ice in water containing air bubbles. Ultrason. Sonochem. 10(2): 71–76.

j91

4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes Antonello A. Barresi and Davide Fissore 4.1 Introduction

Freeze-drying is a process where water (or another solvent) is removed from a frozen product by sublimation. The process consists of three steps: first the product to be dried is frozen; then the pressure is lowered below the triple point to cause ice sublimation (primary drying) while heat is continuously supplied, for example, by conduction through a heated shelf, as the sublimation process is endothermic. Finally, the residual water, strongly bound to the partially dried product, is reduced to a low level by using high vacuum and moderate temperatures (secondary drying), thus ensuring the long term preservation of the product. The low operating temperatures make freeze-drying suitable for highly heat-sensitive products, like pharmaceuticals, that may be damaged by the higher temperature required by traditional drying processes; moreover, freeze-drying warrants a final product that can be easily re-hydrated. Nevertheless, freeze-drying is employed only for valuable goods because of the slow drying rate and the use of vacuum which result in high investment and operating costs (Mellor, 1978; Liapis, 1987; Jennings, 1999; Oetjen and Haseley, 2004; Rey and May, 2004). Product quality control in a freeze-drying process requires the in-line monitoring of the temperature and of the residual water content of the product. In fact, product temperature has to be maintained below a value corresponding to the eutectic point of the system in the case of solutes that crystallize, in order to avoid melting, that is, the formation of a liquid phase, or to the glass transition temperature (Tg0 ) in the case of solutes, such as proteins, that do not crystallize, in order to avoid shrinkage and/or collapse of the cake structure (Franks, 2007). Shrinkage and collapse can be responsible for a higher residual water level in the final product, an extended reconstitution time and the loss of activity of the pharmaceutical ingredient; besides this, a collapsed product is often rejected because of the unattractive physical appearance (Pikal and Shah, 1990; Wang, 2000; Rambhatla et al., 2005). In addition to the temperature, the residual water content has to be monitored in order to detect the

Modern Drying Technology Volume 3: Product Quality and Formulation, First Edition. Edited by Evangelos Tsotsas and Arun S. Mujumdar. Ó 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

92

endpoint of the primary drying, so that secondary drying is started only when primary drying is complete: in fact, if secondary drying is started before the sublimation endpoint, the product temperature may exceed Tg0 , thus causing collapse in some vials, while if secondary drying is delayed after the sublimation endpoint, the cycle is not optimized and the cost of the operation increases. Finally, the residual water content at the end of secondary drying has to be monitored: for most products the target level of residual water is very low, usually less than 1.0–3.0%, so that the viability, immunologic potency and the stability of the product is not compromised over time. However, for certain products it has been demonstrated that a too low level of residual water should be avoided, as viability, or other characteristics, are compromised by over-drying (Hsu et al., 1992): living cells can lose viability, the tertiary structure of complex proteins can be impaired, with subsequent loss of activity, or, finally, monolayers of water can be removed from active sites on molecules which can then react with traces of oxygen and degrade. Currently, even the most advanced industrial freeze-dryers have control systems that are no more than data acquisition tools for certain key variables (Liapis et al., 1996). Monitored data and information obtained in previous runs carried out with the same formulation are used to manage the process, assuming that if the operating conditions are the same as used in the validation batches, the same results will be obtained. This statement is false, because neither ingredients nor processing conditions can remain exactly the same, for example, there can be changes from batch-to-batch due to stochastic subcooling, leading to different nucleation temperatures; in some cases the working temperature, as well as the temperature ramp required to reach this temperature, is selected by the operator: this, however, may not guarantee repeatable conditions for the freezing and sublimation steps. In some other cases the operating conditions (temperature of the shelf and pressure in the drying chamber) are selected by means of the mathematical simulation of the drying phase, with the goal to optimize the process (Rene et al., 1993; Lombra~ na et al., 1993a, b; Boss et al., 2004; Velardi and Barresi, 2008a). Poor process control is a consequence of the impossibility of measuring in-line the parameters of interest, namely the product temperature and the residual water content. Moreover, regulatory guidance, up to now, has imposed operation of the manufacturing process in an open loop, so that only monitoring was allowed during production. Nevertheless, at least during the phase of process development carried out at laboratory or pilot scale, it would be very useful to have an in-line control system to minimize the drying time: process development can be expensive and time consuming but no regulatory restrictions apply during this phase and, thus, the use of an efficient control system can give significant advantages. Recently, the research in this field has been strongly encouraged by the issue of the Guidance for Industry PAT (Process Analytical Technology) by the US Food and Drug Administration in September 2004. This guidance describes a regulatory framework encouraging the design and implementation of innovative pharmaceutical development, manufacturing and quality assurance to support innovation and efficiency to have safe, effective and affordable medicines. PAT is considered to be a system for designing, analyzing and controlling manufacturing through timely measurements

4.1 Introduction

of critical quality and performance attributes of raw and in-process materials and processes, with the goal of ensuring final product quality: quality cannot be tested into products, but it should be built-in or should be by design. The benefits that can be achieved by an optimal control and monitoring policy have been recently discussed by Sadikoglu et al. (2006). The pressure in the drying chamber and the temperature of the shelf are the two variables that can be used for control purposes, as they affect mass and heat transfer. The pressure can have an opposite effect on the two phenomena: Sandall and Wilke (1967) found experimentally and theoretically that there is an optimal pressure that maximizes the drying rate, while Nail (1980) and Pikal et al. (1984) showed that, even if sometimes mass transfer can be the rate-limiting factor, heat transfer from the heat source to the sublimation front is usually the rate-limiting process, as the low pressure reduces the gas thermal conductivity, thus increasing the heat transfer resistance in the air gap between the vials and the shelf (or, if a tray is used, between the vials and the tray and between the tray and the shelf). Nevertheless, the situation seems to be controversial since Jennings (1986), by measuring the sublimation rate of ice, found that decreasing pressure has a positive effect on the sublimation rate. Livesey and Rowe (1987) considered different case studies and pointed out that, even if from a theoretical point of view the sublimation rate is expected to be negatively affected by an increase in the chamber pressure, the enhancement of the heat transfer has a more significant effect. An optimum in the operating conditions can be found as a function of product characteristics: Franks (1998) showed values of the chamber pressure and of the shelf temperature that allow one to maintain the maximum product temperature at a certain level, that can be set equal to the maximum allowable value in order to minimize the drying time. Similar results were given by Oetjen and Haseley (2004) using a simplified model for the main drying, while Trelea et al. (2007) used a detailed mathematical model to calculate the value of the shelf temperature that minimizes the drying time, besides maintaining the product temperature below the maximum allowable value; Velardi and Barresi (2008a) included the chamber pressure in the optimization of the process. During secondary drying the temperature is higher than for primary drying due to the low value of residual water and, thus, the higher glass transition temperature (Franks, 2007). Also, in this case the shelf temperature and the total pressure in the drying chamber can be manipulated in order to minimize the time required to reach the desired amount of residual water in the product, taking into account the constraints on maximum allowable product temperature (Sadikoglu et al., 1998; Sadikoglu, 2005; Pikal et al., 2005; Tang et al., 2005). The variation of the maximum allowed product temperature with the residual water content has been taken into account in the control algorithm proposed by Trelea et al. (2007), thus resulting in a further reduction of the drying time. This chapter aims to discuss various issues concerning the monitoring and control of the three steps of a freeze-drying process of pharmaceuticals in vials, namely freezing, primary and secondary drying. Various devices used to monitor the process, both in pilot-scale and in industrial-scale equipment, as well as the tools proposed to control the process will be described and discussed.

j93

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

94

4.2 Control of the Freezing Step

Freezing conditions can strongly influence the primary and secondary drying, as well as the characteristics of the final product, as this step determines the shape and the dimensions of the ice crystals that form the structure of the frozen product (see Chapter 3 for more details). As far as primary drying is concerned, the ice crystals should be large in order to result in a highly porous structure, thus allowing the water vapor to flow from the subliming interface to the drying chamber without a significant resistance; with respect to secondary drying, the ice crystals should be small, so that the specific surface area is high, as this is beneficial with respect to water desorption. The optimal ice crystal size has to be determined with respect to the cost of the whole duration of both primary and secondary drying (Jennings, 1999). For a given formulation various factors can affect the freezing process, namely the cooling rate, that affects the supercooling degree and the ice nucleation temperature (Searles et al., 2001a), the type of freezing (Patapoff and Overcashier, 2002), and the use of an annealing step (Searles et al., 2001b); Hottot et al. (2007) extended the analysis of the parameters that can affect the structure of the ice crystals by investigating the influence of the type of the vial and of the filling height. The differences in the ice morphology are generally related to the nucleation rates and to the nucleation temperatures, that is, to the undercooling and to the thermal gradients inside the solution. It is well known that when the aqueous solution in a vial is cooled below the thermodynamic freezing temperature it remains in a subcooled metastable liquid state until nucleation occurs. The nucleation temperature is distributed around a value below the thermodynamic one, that can depend on the freezing rate, and this causes vial-to-vial variations in the ice crystal structure and, thus, in the drying rate and in the physical properties of the freeze-dried product. This can be one of the most important reasons for batch heterogeneity and can seriously impact monitoring and control of the process. Control of the nucleation processes is thus a key factor for the optimization of the morphological properties of the freeze-dried matrix. Various techniques have been proposed in the past to control the freezing step, for example, the “ice fog” technique and the use of vibrations. In the “ice fog” technique crystals are introduced in the vials to act as nucleating agents for ice formation in subcooled aqueous solutions. This procedure involves lowering the shelf temperature and cooling the samples to the desired temperature of nucleation; then a flow of nitrogen gas at a controlled high pressure (e.g., 1 bar) is circulated through copper coils immersed in liquid nitrogen and is introduced into the humid drying chamber: ice crystals form and enter the vials due to the increase in pressure, thus causing the nucleation of the solution at the desired temperature (Rowe, 1990; Rambhatla et al., 2004). However, the nucleation event does not occur concurrently or instantaneously within all vials upon introduction of the cold gas into the freeze-dryer, because the ice crystals take some time to enter each of the vials to initiate nucleation, and transport times are likely to be different for vials in different locations within the freeze-dryer. Internal convection

4.2 Control of the Freezing Step

devices may thus be required in an industrial-scale equipment to assist a more uniform distribution of the “ice fog” in the drying chamber. Vibration has also been used to induce a phase transition in metastable materials. Vibrations sufficient to induce nucleation occur at frequencies above 10 kHz and can be produced using a variety of equipment. Vibrations in this frequency range are often termed “ultrasonic”, although frequencies in the range from 10 to 20 kHz are typically within the audible range of humans. Ultrasonic vibration often produces cavitation, that is, the formation of small gas bubbles, in a subcooled solution: in the transient (or inertial) cavitation regime, the gas bubbles rapidly grow and collapse, causing very high localized fluctuations of pressure and temperature, while in the stable (or non-inertial) regime the gas bubbles exhibit stable volume or shape oscillations without collapse. The ability of ultrasonic vibration to induce nucleation in a metastable material is often attributed to the disturbances caused by transient cavitation. Control of the nucleation process can enable the freezing of all unfrozen solutions in a freeze-dryer to occur within a more narrow temperature and time range, thereby yielding a lyophilized product with greater uniformity from vial to vial. The industrial realization of ultrasound nucleation poses some problems: a modified freeze-dryer technology (Telstar, Barcelona) was recently realized (Morris et al., 2004) and demonstrated to improve significantly the whole batch homogenization and to reduce the variations of the sublimation rates, without reducing the protein activity. A nucleation temperature range near the melting point was used, thus confirming that the closer the nucleation temperature is to the melting point, the larger is the size of the ice crystals (Passot et al., 2007). Nakagawa et al. (2006) showed that a significant increase in the sublimation rate can be obtained by increasing the nucleation temperature from the average spontaneous nucleation temperature to around
2  C with ultrasound control. They used an ultrasound transducer that was tightly attached to an aluminum plate placed in thermal contact with the aluminum heat exchanger by clamper fixing and coupled with an ultrasound generator (MW400GSIP, SODEVA, France). The nucleation of the samples was realized by 1 s of ultrasound propagation at a selected temperature during cooling the system at
1 K min
1. Work on industrial equipment has shown that this technique has a great potential for application in normal production, evidencing, at the same time, the technical problems that must be solved. Figure 4.1 shows an example of the results that can be obtained when using ultrasound controlled nucleation in comparison to the case of spontaneous nucleation. The data refer to a freeze-drying cycle of liquid formulations of Human Recombinant Interferon a2b. Freezing was carried out in the “LyoGamma special” industrial prototype by Telstar mentioned above, that was equipped with both standard and forced-nucleation shelves, for comparison. The shelves were initially cooled to
60  C at approximately 0.3 K min
1. Samples on the nucleating shelf were forced to nucleate at an average sample temperature of
4  C, while samples on the standard shelf were allowed to nucleate spontaneously. This spontaneous nucleation was observed to occur in the range
8 to
12  C. Primary drying was carried out at
25  C and 0.1 mbar and secondary drying was carried out at þ 25  C for 8 h; the weight of five vials was continuously monitored by means of a balance

j95

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

96

14 12 10

Mass, g

8 6 4 2 0

0

2

4

6

8 10 12 14 16 18 20 Time, h

Fig. 4.1 Comparison between the time evolution of the product mass during sublimative drying of a liquid formulation of Human Recombinant Interferon a2b with spontaneous nucleation (&) and with forced nucleation (&).

placed in the drying chamber (see Section 4.3.2). The sharp change in the slope of the curve of the weight of the samples in correspondence to the end of the main drying when forced nucleation is used confirms that the batch is much more homogeneous and the vials terminate the primary drying simultaneously, whereas in the case of spontaneous nucleation some vials end the primary drying before the others, thus resulting in a smooth change of the slope of the curve. The forced nucleation influences not only the duration of the sublimation step, but also the residual moisture at the end of the process: the samples frozen with forced nucleation showed higher residual moisture (0.80% vs. 0.35%) than the samples nucleated spontaneously, while a reduction in the drying time by 43% with respect to the spontaneous nucleation condition was observed. These results are in perfect agreement with the morphology of ice crystals observed for the samples obtained with forced and spontaneous nucleation, respectively. As shown in Fig. 4.2, crystals of larger size are obtained with forced nucleation, so that the mass transfer resistance during the primary drying is lower and the process is faster. During secondary drying a lower active surface is available for the desorption of water and, consequently, a higher residual moisture is obtained than in case of spontaneous nucleation.

4.3 Monitoring of the Primary Drying

After freezing the product, most of the solvent is removed by sublimation during the primary drying. In this stage it is necessary to monitor product temperature and the residual water content in such a manner that the control system can both optimize

4.3 Monitoring of the Primary Drying

Fig. 4.2 Comparison between the cake structure obtained from a liquid formulation of Human Recombinant Interferon a2b for (a) spontaneous nucleation and (b) forced nucleation; scanning electron microscope, metallized samples.

the process and prevent any damage to the product and any loss of activity in labile materials. A limitation of the present technology is the impossibility of obtaining in-line measurement of the parameters of interest without interfering with the process dynamics or impairing the sterile conditions usually required when pharmaceuticals are processed. An example of a widespread, but invasive, monitoring device is, in fact, the measurement of the product temperature obtained by inserting a thin thermocouple, or a resistance thermal detector (RTD), inside the vial. While thermocouples are more frequently used in lab-scale equipment, RTDs are preferred in manufacturing due to their mechanical robustness and to the possibility of being sterilized (Willemer, 1991; Oetjen, 1999). Nevertheless, RTDs are usually much larger than thermocouples and the measurement accuracy can be distorted by the sensor geometry (Presser, 2003). The use of thermocouples and RTDs may alter the elementary phenomena of nucleation and ice crystal growth: it is well known that vials where thermocouples or RTDs are placed tend to show a lower degree of supercooling than the surrounding vials and, therefore, form fewer and larger ice crystals which, finally, results in lower product resistance to mass transfer and shorter drying time in comparison to the rest of the batch. While this difference may be inconsequential in the laboratory, the sterile and particle-free environment in

j97

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

98

manufacturing leads to substantially higher supercooling of the solution, resulting in larger differences between vials with and without temperature sensors. Moreover, the insertion of even thin thermocouples affects the heat transfer to the product and the probe insertion compromises the sterility of the product and is not compatible with automatic loading/unloading systems used in industrial-scale freeze-dryers (Nail and Johnson, 1991; Schneid and Gieseler, 2008). The development of wireless systems can make the use of the product probes compatible with automatic loading systems, even if all the other concerns remain valid. Wireless solutions (“active transponders”) have been available for about two decades, using the ISM-frequency-bands for short-range data transmission from the sensor-units, and start to be suitable for use in freeze-dryers; however, these active transponders show similar problems to RTDs. Moreover, all of them require batteries for sensor-operation and data-transmission, thus resulting in limited operating time dependent on battery capacity and, finally, unpredictable risks when using batteries in a sterile environment. Recently, a new generation of wireless probes was developed (“passive transponders”) which generate the energy required for transmission from an electromagnetic field instead of using batteries (Hammerer, 2007). The temperature remote interrogation system (TEMPRIS) sensors have been recently proposed by Schneid and Gieseler (2008): this system allows real-time temperature measurement at the bottom center of a vial (if placed correctly) and is beneficial for scale-up, as the same sensors can be used in the lab and on a manufacturing scale. The main drawback concerns the placement of the sensor at the bottom center of the vial: this was found to be crucial to obtain reliable temperature profiles and endpoint monitoring, but no “brackets” are available at the moment to assure “bottom center” position. Moreover, the dimension of the sensor is a problem, as well as the fact that the thermocouple is invasive with respect to the product in the vial. Despite the various drawbacks listed above, thermocouples have been proposed to monitor the primary drying and to detect the endpoint of this stage. When the primary drying ends, an increase in the product temperature is measured at the bottom of the vials due to the loss of thermal contact between the sensor and the ice. Moreover, the product temperature increases as there is no longer an endothermic sublimation process that uses the heat supplied by the heating shelf. The product temperature can be used to detect the end of primary drying also using a completely different procedure, by regularly and drastically reducing the pressure in the chamber: if no decrease in the product temperature is observed as a consequence of the pressure reduction, that in the presence of residual ice would cause an increase in water flux by endothermic sublimation, then the primary drying can be considered complete (Thompson, 1988). Finally, by coupling a mathematical model of the process to the measurement of the product temperature in the vial (or of the wall temperature of the vial) it is possible to build a soft-sensor that allows estimation in-line of the whole product temperature profile and the mass/heat transfer coefficients; this has been called the “smart-vial” concept (Barresi et al., 2007, 2009a, b, c) Non-invasive monitoring techniques have been proposed as valuable alternatives to the use of thermocouples. Most of them are based on the results obtained from the

4.3 Monitoring of the Primary Drying

pressure rise test (PRT): they use the in-line measure of the pressure rise occurring when the valve placed between the drying chamber and the condenser is closed for a short time interval (typically from 5 to 30 s) and estimate the temperature of the sublimating interface and, generally, also other parameters, using a mathematical model of the process. This technique was well known already in the early 1960s (see, for example, Rieutord, 1965), but has only recently been refined to obtain more accurate and reliable results. Several algorithms have been proposed in the literature to interpret the PRT, namely the barometric temperature measurement (Willemer, 1991; Oetjen, 1999; Oetjen et al., 2000; Oetjen and Haseley, 2004), the manometric temperature measurement (Milton et al., 1997; Tang et al., 2006a, b, c), the dynamic pressure rise (Liapis and Sadikoglu, 1998), the pressure rise analysis (Chouvenc et al., 2004, 2005; Hottot et al., 2005) and the dynamic parameters estimation (Velardi et al., 2008). The sublimation flux of the solvent can be calculated from the mathematical model used to fit the curve of the pressure rise or from the slope of the curve of the pressure rise at the beginning of the test (Fissore et al., 2011a) the two procedures should of course give similar results and this can be used as a consistency check (Pisano, 2009; Pisano et al., 2009). By integration of the solvent flux it is possible to calculate the residual ice content of the solid, thus detecting the endpoint of primary drying. Moisture sensors, mass spectrometer, windmill sensor, pressure gauges and other devices have been proposed in the past to monitor the primary drying and to detect the end of this stage: a technical comparison of these and other recently proposed devices is given in Mayeresse et al. (2007), Wiggenhorn et al. (2008) and Barresi et al. (2009a). All these sensors do not provide any information about the status of the product during the operation and, thus, they cannot be used in a control loop, but they may be very useful to establish when primary drying is completed and secondary drying can be started without causing quality loss. The various techniques available to monitor the primary drying can be roughly divided into three groups as they can be used to monitor single vials, a group of vials or the whole batch (Barresi et al., 2009a): these techniques will be discussed and compared in the following sections. 4.3.1 Monitoring of Single Vials

Besides the use of thermocouples and RTDs, various other techniques have been proposed to monitor freeze-drying in a vial. A quick mention can be made of the various analytical techniques, such as low-temperature X-ray powder diffractometry (Cavatur and Suryanarayanan, 1998), low-resolution pulse nuclear magnetic resonance (Monteiro Marques et al., 1991), FTIR spectroscopy (Remmele et al., 1997), and visual microscopic observation (Mackenzie, 1964), that have been used for in situ characterization of samples being lyophilized in special lyophilization equipment, connected to the analytical instrument. These techniques are very useful for process understanding and process design, but at the moment it seems very difficult to use them in a conventional freeze-dryer.

j99

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

100

Removal of samples during the process using a manipulator is a technique, proposed and recommended as an independent control procedure, to follow the sublimation of water (and also the desorption of bound water in secondary drying) especially in the case of very expensive products, such as those produced by genetic engineering and biotechnology; but it is generally employable only in scouting tests, where, on the other hand, it can also allow a step-by-step optimization of the entire process (Willemer, 1987). Of course, sampling can be coupled with any analytical technique available at-line to measure the residual ice or residual moisture, such as gravimetry, titration, spectroscopic methods (Skibsted, 2006) and even X-Ray photography (Schelenz et al., 1994). The use of sensors based on spectroscopy methods has been investigated for inline monitoring. Near-infrared (NIR) spectroscopy has been proposed for monitoring the process because the physical changes occurring, that is, freezing, sublimation and desorption, generate significant spectral changes (Presser et al., 2002a; Br€ ulls et al., 2003; Presser, 2003). NIR allows a non-invasive, non-destructive and rapid moisture determination and offers a wide range of capabilities in the process analytical technology of freeze drying pharmaceuticals (Ciruczak, 2002). The end of the primary and secondary drying process can be determined precisely. Moreover, the freezing point, the ice formation process and the transition from frozen solution to dried material can be observed. Good agreement between NIR spectroscopy and product temperature monitoring of the freezing process and of the transition from the frozen product to the ice-free material has been reported (Br€ ulls et al., 2003). The major advantage of this method is the simultaneous and rapid in situ process monitoring for process and product conditions resulting in the direct determination of the moisture during different drying stages (Presser et al., 2002a). Wiggenhorn et al. (2005a, b, 2008) showed results obtained by locating the NIR sensor on the shelf inside the drying chamber and directly fitted with the probe tube to the outer wall of the vial: this arrangement offered the advantage of sterility compared to other set-ups, where the single fiber reflectance probe of the FT-NIR probe is located inside the vial in combination with a thermocouple (Br€ ulls et al., 2003). It is, furthermore, much less disturbing as it does not affect the drying kinetics of the sample vial nor does it restrict the fill volume or the vial type. Recent studies have described a noninvasive, in-line and real-time analysis of the lyophilization process by means of Raman spectroscopy: De Beer et al. (2007) reported some results on the in-line characterization of physical phenomena (i.e., mannitol crystallization) in a lab-scale apparatus, using a fiber-optic non-contact probe placed above the freeze-dried product; they also evidenced the utility of coupling in-line Raman spectroscopy with at-line NIR spectroscopy and X-ray powder diffraction. Besides spectroscopy methods, other techniques have also been proposed for the in-line monitoring of the water content in single vials: among them, dielectric measurements (Suherman et al., 2002) seem the most promising, at least for detection of the endpoint; the electrodes can be placed outside the vial to reduce interference with the process. Monitoring of electric properties, and in particular of the product resistance (or inductance), has also been proposed. The technique seems suitable especially for substances that show a sharp eutectic melting, as in this case

4.3 Monitoring of the Primary Drying

there is a very large variation in the resistance when the product approaches the limit temperature. The sensitivity of the technique is very high, even if applicable easily only to certain crystalline products, but electrodes must be inserted in the vials: thus, it has been employed mainly for detection of the product critical temperature. Notwithstanding the limitations described, and some reliability problems, resistance monitoring was proposed also for the automatic control of the whole operation, and this will be briefly discussed in a later section (Rey, 1961; De Luca and Lachman, 1965; Jennings, 1999). Jennings and Duan (1995) developed a different technique to monitor the primary drying, based on the determination of the total energy necessary to carry out the primary drying and on calorimetric measurement to calculate the heat transfer coefficient in the monitored vial, and thus the rate of heat transport. For this purpose a differential method, called drying process monitoring (DPM), is used: two thermocouples are fixed to the bottom of an empty vial and a vial filled with the product, thus allowing calculation of the heat transfer to the filled vial used for the sublimation of ice; a drop in the heat transfer rate can be observed at the end of the main drying (Jennings, 1999). This method requires the introduction of two vials with thermocouple connections in the production batch; thus, even if more sophisticated, and probably more reliable, than the simple measurement given by one thermocouple, the DPM maintains most of the drawbacks of the thermocouples, including the obvious fact that measuring the situation in the two special vials may not represent an average of all the batch vials. Another device that makes use of the measurement of the temperature of the product (or of the vial) is the soft-sensor (or observer): it provides a real-time estimation of some parameters or state variables, for example, the whole product temperature profile and the mass and heat transfer coefficients, using the temperature measure and a mathematical model of the process. Let us consider a dynamic system defined by the following set of differential equations: x_ ¼ f ðx; uÞ

ð4:1Þ

where x 2 R n is the state of the system, u 2 R m is the vector of the control variables and f is an application from R n  R m to R n giving the derivatives of the state as a function of the state itself and of the control law u applied to the system. In order to build an observer some information concerning the measured variables is required: in this way the equations describing the dynamics of the system become the following:  x_ ¼ f ðx; uÞ ð4:2Þ y ¼ hðx; uÞ where the components of the vector y 2 R q are the measured variables and the equation y ¼ hðx; uÞ involves the description of these quantities in terms of the components of the state vector and of the control law. The observer for the system described by Eq. 4.2 is another system of equations, whose state is denoted by ^x:  ^x_ ¼ f ð^x; uÞ þ Kðy
^yÞ ð4:3Þ ^y ¼ hð^x; uÞ

j101

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

102

Temperature, K

where ^y are the estimations of y obtained from the observer and K is the gain of the observer, that is, a non-linear function of the state (^x) and of the input (u) that ensures asymptotic stability, driving the estimation error e ¼ y
^y to zero. The synthesis of an observer, that is, the calculation of the gain K, is a complex task and a lot of algorithms have been proposed in the literature. The Extended Kalman Filter is one of the most common techniques (Becerra et al., 2001) and it has been used to track the freezedrying process by Barresi et al. (2009b) and Velardi et al. (2009): they used a simplified pseudo-stationary model and the measurement of the product temperature at the bottom of the vial to estimate the temperature and the position of the moving front as well as the heat and mass transfer coefficients. As an alternative, a High Gain observer has been designed and tested as, in this case, the mathematical formulation is simpler and the computational time required for the estimation is generally lower; moreover, the High Gain observer exhibits less sensitivity towards noisy measurements (Velardi et al., 2010). Both soft-sensors have been validated by numerical simulations using a detailed one-dimensional model as a source of experimental data (Barresi et al., 2009a): the quality of the estimations given by both observers has been verified to be roughly the same, but the computational effort required by the Kalman Filter is higher and its tuning is quite tricky, while the estimations obtained using the High Gain observer are provided faster and the tuning is simpler. Preliminary experimental results confirm that in-line estimations can be very good, as is evidenced in Fig. 4.3, which shows the prediction of the interface temperature and the position of the frozen interface obtained using the Kalman Filter. The main drawback of using an observer to monitor

242 240

(a)

238 236 234

Interface position, m

232 230 -3 2.5×10 (b)

-3

2.0×10

-3

1.5×10

-3

1.0×10

-4

5.0×10

0.0 0

2

4

6

8

10

Time, h Fig. 4.3 Measurements and estimations of the product temperature and of the moving front position during the FD of 10% w/w sucrose solution (dv ¼ 14.2  10
3 m, Lp ¼ 7.2  10
3 m, Pc ¼ 10 Pa). (a) Product temperature at the moving front (solid line) and

at the bottom (dotted line) estimated using the Kalman filter; (o) temperature at the bottom of the product measured by a thermocouple. (b) Interface position estimated by the Kalman filter (solid line) from the temperature measurement given in the upper graph (Velardi et al., 2009).

4.3 Monitoring of the Primary Drying Freeze-dryer chamber Door of the freeze-dryer

vial

Vial

thermocouples

Wireless communication system

Processing module

Temperature measurement system

Heating/cooling plate - Temperature profile inside the vials - Position of the sublimation interface

Fig. 4.4 Sketch of the “smart vial” concept, the system using the soft-sensor to monitor the freezedrying process (Barresi et al., 2007).

the process is that the state estimation is limited to a single vial. On the other hand, the temperature estimation concerns the entire temperature profile of the product in the vial and not only the temperature at a particular point, as is obtained using a thermocouple. Moreover, the results obtained for a particular vial can be compared to those obtained for other vials placed in different positions in the drying chamber, thus allowing evaluation of the heterogeneity of the batch. Since the insertion of a probe, although extremely tiny, in contact with the product should be avoided because of the various drawbacks previously discussed, a different observer can be designed, using the measurement of the external temperature at the bottom of the vial (Barresi et al., 2007, 2009b, c) and a mathematical model that takes into account also the heat transfer in the glass wall (Velardi et al., 2005; Velardi and Barresi, 2008a). This is the “smart vial” that has been patented recently by Barresi et al. (2007). A wireless system, using an active or passive transponder, can be used to transmit measured values to a PC. The passive device (Fig. 4.4) is able to perform temperature measurements in several vials and to communicate the results to an external receiver through a central unit. The low-frequency RF-ID technology is employed in order to avoid the use of batteries (Vallan et al., 2005a). A second kind of device has been recently designed and tested: it makes use of a small battery that can be completely embedded in the vials and is able to transmit the measurements within a range of several meters by means of a 2.4 GHz radio (Barresi et al., 2009b; Fissore et al., 2009c; Corbellini et al., 2010). 4.3.2 Monitoring of a Group of Vials

While the sensors described in the previous paragraph allow monitoring of only single vials, a balance placed directly in the vacuum chamber of the freeze-dryer allows monitoring of a group of vials: the direct weight measurement enables the

j103

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

104

tracking of primary drying and detection, with good accuracy, of the endpoint of this stage. Remarkable improvements can be obtained with respect to sample-extractors (Nail and Gatlin, 1993; Tang and Pikal, 2004) as the vacuum conditions are not modified and the weighing procedure is performed in an automatic way. Initially proposed in food freeze-drying processes (Oetjen et al., 1962), were different types of balances to weigh one or more vials in line. Bruttini et al. (1986, 1991) proposed a balance which supported the heating plate and the tray: freezing was carried out as a separate step because when it was carried out in situ, the vibrations induced by fluid pulsation from the cryostat severely disturbed the measurement. A certain drawback of weighing devices is that the measurement cannot be representative of the process because the weighed vials are exposed to different thermal conditions than the rest of the batch. In the capacitive balance proposed by Rovero et al. (2001) the heat transfer to the product is limited by the volumetric gap that acts as an additional resistance when heat is transferred through the shelf by conduction, while it works efficiently in the case of radiative heating. Moreover, some of the balances so far proposed require vials with a specific geometry that does not always correspond to that of the vials of the batch, and the measurement is limited only to that special vial (Christ, 1995; Roth et al., 2001; Gieseler, 2004; Gieseler and Lee, 2008a, b). Vallan et al. (2005b) proposed and patented a weighing device that works inside the vacuum chamber: it is composed of a motorized balance which is able to rise and weigh several vials (up to 15 small vials in the first prototype, but extensible even to a much larger number and adaptable to different sizes of vials with a proper loading cell), and a miniaturized radio-controlled thermometer, connected to the balance tray, that can measure the temperature of these vials without altering the mass measurement because of the force transmitted by the thermocouple wires, as is sketched in Fig. 4.5 (Vallan, 2007). Since these vials are almost always in contact with the shelf and lifted just during the measurement, the thermal exchange between the vials and the heating surface is not significantly affected and, therefore, the measurement is representative of the whole batch. The balance is connected to a PC by means of a serial interface; it has a resolution of about 10 mg and a total uncertainty of about 100 mg from
40 to þ 40  C. A series of tests has shown that the weighing frequency can be chosen in a wide range without affecting the process, but both the monitored vials and the

Reader

RF link Thermocouples Temperature Measurement System

Balance Step Motor

Vial

External power supply

Vial

RS232

PC

Cooling shelf

Fig. 4.5 Sketch of the balance for freeze-dryers (Vallan et al., 2005b; Vallan, 2007).

4.3 Monitoring of the Primary Drying

balance case must be properly shielded to avoid systematic errors due to radiation effects from the walls: if this is not done, the vials lifted by the balance are in a condition similar to that of the vials at the sides of the batch, where radiation effects are much more important, and the balance response can be considered representative of this fraction of vials (Pisano et al., 2008). This device has been tested in various freeze-drying cycles performed in different working conditions: a good agreement between the time evolution of the values of mass and temperature has been evidenced during these runs. Figure 4.6 shows an example of the results that can be obtained; the measure of the temperature obtained by means of thermocouples placed at the bottom of some vials is also shown for comparison, as well as the calculated average sublimation rate. When the mass of the vials weighed by the balance becomes very low and almost constant, it means that the endpoint of primary drying has been reached; of course, the calculation of the sublimation rate from the derivative of the mass measurement increases the sensitivity and allows determination of the end of primary drying with good accuracy. The end of the sublimation drying in Fig. 4.6 is confirmed by the strong increase in the product temperature occurring at the same time.

Mass, kg Temperature, K

(a) 0.140

2×10-4

0.135

1×10-4

0.130 320

0

Sublimation rate, kg m-2s-1

3×10-4

0.145

300 280

(b)

260 240 220

0

2

4 6 Time, h

8

Fig. 4.6 Example of results obtained during a freeze-drying cycle using the special balance with the embedded wireless temperature measurement in order to monitor the primary drying stage of a mannitol-dextran solution (6–14% by weight) in a pilot-scale freeze-dryer. The freezing stage was run at 223 K for about 5 h, while the main drying was carried out setting the fluid temperature at 293 K and the chamber pressure at 10 Pa (Nv ¼ 98 on tray, dv ¼ 14.2  10
3 m, Lp ¼ 7.2  10
3 m).

10

(a) Time evolution of the product mass (line: the gross mass of 15 vials containing 1 cm3 of solution each, including glass vials and tray) and average sublimation rate calculated from this measurement (symbols). (b) Time evolution of the heating fluid temperature (solid line) and of the product temperature measured by thermocouples inserted close to the bottom in three of the vials weighed by the balance (dashed lines).

j105

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

106

4.3.3 Monitoring of the Whole Batch

As stated in Section 4.1, it is necessary to monitor product temperature in order to avoid collapse (or melting) of the product, and also to detect the end of primary drying, in order to start secondary drying when all the vials have completed their main drying: if secondary drying is delayed the total duration of the cycle, and thus its cost, is increased, while if secondary drying is started before the sublimation endpoint, the product temperature is increased too early and may thus exceed Tg0 and cause the product to collapse in some vials. The sensors that can be used to monitor the whole batch can be roughly divided into two groups, those that allow detection only of the endpoint of the primary drying and those that also allow estimation of the product temperature and the residual water content from the results of the pressure rise test (PRT); the latter can be used in a control loop to optimize the primary drying step. 4.3.3.1 Detection of the Endpoint of the Primary Drying A capacitance manometer or a thermal conductivity gauge, like the Pirani gauge, can be used to measure the pressure in the drying chamber: the latter is cheaper than the capacitance manometer, but the accuracy is generally lower and the signal depends on the gas type and, in the case of mixtures, like water and inert gas, on the composition. Thus, the use of Pirani (and generally of all the thermal conductivity gauges) should be discouraged for monitoring freeze-drying because the chamber gas composition changes continuously during each run and is generally different in different cycles as it depends on set-up, loading and product features (Armstrong, 1980). On the other hand, taking into account the known dependence of the Pirani response on the water vapor fraction, it is possible to evaluate the partial pressure of water in the drying chamber using the signals obtained from the capacitance manometer and from the Pirani sensor. Moreover, as suggested by Nail (Armstrong, 1980; Nail and Johnson, 1991), it is possible to detect the end of the primary drying as the concentration of water into the drying chamber becomes very low at that point and, thus, the pressure measured by the Pirani (that is generally calibrated for air) approaches that measured by the capacitive gauge. The use of the ratio of the pressure signals given by the two gauges, that approaches unity at the end of the primary drying, instead of the simple measure given by the Pirani sensor, is more reliable because it eliminates the possible effect of a variation in the total pressure. It must be taken into account that the method is quite sensitive, and senses the last vials to dry; this is generally desirable, but there is the possibility that some vials in an abnormal configuration, such as vials that fall off the shelf and thus dry much more slowly than regular ones, are responsible for a misleading signal (Nail and Johnson, 1991). One possible limitation to the application of this simple method is the restriction in the use of thermal conductivity gauges in equipment where steam sterilization is required, even if producers claim that new models using different materials for the filament (nickel or platinum rather than the standard tungsten) can cope with sterilization. Recently, new sensors based on the same principle as the Pirani gauge have been proposed: a stainless steel shield is used to

4.3 Monitoring of the Primary Drying

protect against condensation and a pulsed mode of operation allows higher signal resolution, an extended range of measurement and higher long-term stability (Ploechinger and Salzberger, 2006). An early method used for measuring the partial vapor pressure in the chamber is that of the vapor sampling condensing (or trap method), described in detail by Kan (1962). It is still used, especially in small apparatus for research (Pikal et al., 1984; Roy and Pikal, 1989). Moisture sensors and mass spectrometers can be used to monitor the time evolution of the water concentration in the chamber. Dew point sensors can detect the gas composition or the relative humidity owing to a change in the dielectric constant of a gold sputtered foil material: they indicate a sharp decrease in the dew point when the water vapor decreases to almost 0% (Bardat et al., 1993) and can have greater sensitivity than a thermal conductivity gauge. Roy and Pikal (1989) used a moisture sensor that exploits the variation of the capacity of a thin film of aluminum oxide due to moisture (Ondyne, by Endress þ Hauser HydroGuard 2250, Greenwood, IN): according to the authors, the sensor has the sensitivity to determine the presence of ice in less than 1% of the vials of the batch. This device, first proposed by Bouldoires (1969), was successively used by Genin et al. (1996) and by Rambhatla et al. (2004) to monitor the process and to detect the sublimation endpoint; Trelea et al. (2007), Chouvenc et al. (2004) and Barresi et al. (2009a) used a similar moisture sensor, developed by Panametrics. Genin et al. (1996) also developed a procedure that led to a patented method (Rene et al., 1995) for the estimation of the residual water content in the product at any time during the process. The use of a radio frequency mass-spectrometer with rapid response (Favitron by E. Leybold’s Nachfolger) to detect the end of primary drying (and also of secondary drying) was described by Kan (1962), evidencing that it was able to work even in the absence of inert bleeding. Later, the use of a quadrupole mass spectrometer (QMS) to monitor primary drying was reported by several authors (Jennings, 1980, 1999; Nail and Johnson, 1991; Willemer, 1991; Connelly and Welch, 1993; Presser et al., 2002b; Wiggenhorn et al., 2005a, b). The working principle of a QMS is simple: the gas is sampled to the instrument where the molecules are fragmented, ionized and accelerated by an electric field; the ions are then driven to the detector which gives a signal proportional to the concentration and to the type of the impacting fragment. It is quite difficult to get quantitative measurements as it is not easy to calibrate the instrument for gaseous mixtures with water; moreover, it is necessary to repeat the calibration before each run as the response factor of the instrument can be variable; for this purpose, Jennings (1980) suggested the use of a capacitance manometer to make the calibration. Nevertheless, significant information can be obtained from the QMS even if only the crude signals, given in terms of ionic currents, are investigated: the time evolution of the ionic current corresponding to the fragment of mass 18 (i.e., water), i18, divided by the total pressure reading made by the QMS, was proposed by Jennings (1980) to detect the end of the primary drying: this signal is almost constant during all the primary drying and decreases at the endpoint. With the implementation of an aseptic sterile filter between the drying chamber and the QMS, this device can also be used in commercial scale production set-ups where full aseptic conditions

j107

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

108

are mandatory (Wiggenhorn et al., 2005a, b). Despite the high sensitivity and the possibility of also monitoring freeze-drying processes with organic solvents, a QMS is very expensive and, in the common case of just water solvent, it does not provide more “information” than the ratio between the pressure measured by the Pirani gauge and by a capacitance manometer. The QMS may be extremely valuable for quality control, as it can also be used as a diagnostic tool for system leak detection, back-streaming of pump oil, out-gassing of elastomeric components of the freeze-dryer and sublimation of low molecular weight formulation components (Leebron and Jennings, 1981; Nail and Johnson, 1991). Figure 4.7 shows an example of a freeze-drying cycle where various previously described sensors have been used to monitor the process; in this case a relatively

Temperature, K

300 280 260

(a)

240

0.6

60 0.4

40 20

0.2

0

0.0

1.50 PPirani/PBaratron

0.8

(b)

80

i18/PQMS, A mbar-1

Water content, %

220 100

(c)

1.25 1.00 0.75 5

10

15

20

25

30

35

Time, h Fig. 4.7 Example of the results obtained during a freeze-drying cycle using various devices to monitor the primary drying stage of 5.5% by weight lactose solution in a pilot-scale freeze-dryer (Barresi et al., 2009a). The freezing stage was run at 223 K for about 5 h, with an initial cooling rate of 1 K min
1, while the primary drying was carried out at 263 K, with Pc ¼ 10 Pa (Nv ¼ 713 on tray, Lp ¼ 7.2  10
3 m, dv ¼ 14.2  10
3 m). (a) Bottom product temperature measured by a thermocouple

(dashed line) and heating fluid temperature (solid line). (b) Moisture content in the chamber measured by Panametrics moisture sensor (Panametrics MMS35, solid line) and ratio between the ionic current of mass 18 and the pressure measured by the QMS (GeneSys 300 by European Spectrometry Systems, dotted line). (c) Ratio between the signals given by a thermal conductivity gauge (Pirani PSG-101-S) and by a capacitive sensor (MKS Type 626A Baratron).

4.3 Monitoring of the Primary Drying

large batch (more than 700 vials) of lactose solution, with no empty vials as radiation shield, is considered. In Fig. 4.7a the temperature at the bottom of the product measured by a thermocouple inserted into a vial is shown. The water vapor concentration in the chamber, measured by a Panametrics moisture sensor, as well as the ratio between the signals given by a thermal conductivity gauge and by a capacitive sensor are shown in Fig. 4.7a and b, respectively; in Fig. 4.7b the measure given by a QMS, that is, the ionic current corresponding to the fragment of mass 18 divided by the total pressure reading made by the QMS, is shown. The response of the three devices is consistent and taking the point where the signal reaches a minimal and constant value as an estimate of the end of primary drying, a time of about 28–29 h results for all sensors. Actually, all the devices measure the concentration, or partial pressure, of water, even if adopting a different principle. As the chamber gas composition is dependent on the sublimation rate, and this is generally strongly reduced at the end of primary drying, a large variation in its value can be easily captured and used as an indication of the end of the sublimation step. However, it must be taken into account that this is an indirect measurement, affected by several variables, and problems can arise when changing the scale of the apparatus, the size of the batch, the method of pressure control in the chamber, or even the nature of the product. Better results can be obtained if the estimation of the sublimation flux is used to monitor the progress of the drying, as will be shown in the following (Pisano, 2009; Pisano et al., 2009). The shape of the curves obtained with the various devices is quite different. The moisture sensor shows quite a typical behavior: after the total pressure has been reduced to the operating value, the signal first sharply increases, and, after reaching a maximum, slowly decreases for almost all the primary drying, to drop towards the end. This behavior has already been described by Genin et al. (1996) who explained the slow decrease with the slight reduction in the drying rate due to increased thickness of the dry layer and, hence, of the mass transfer resistance, in the period of almost linear variation of residual ice thickness with time; the fast drop was explained by a large reduction in the sublimation rate. The moisture sensor was considered to be very sensitive by the first authors who studied it, being able to sense as few as 0.3% of the vials with ice remaining (Roy and Pikal, 1989): variation in the signal is large enough to detect the end of primary drying, but it suffers from low accuracy, that strongly limits the possibility of accurately estimating the amount of sublimed water from the integration of the moisture signal, and low response time. The latter is mainly due to water desorption from alumina, as noted by Genin et al. (1996) who reported that in the third phase the water desorption rate from the porous alumina could be an order of magnitude greater than desorption from the product at the beginning of secondary drying. The QMS requires a relatively long initial time interval for the stabilization of the pressure inside it; after that, the signal remains almost constant, until it drops, similarly to that of the moisture sensor. It can be concluded that, in the case of water solvent, the performance may be comparable to that of the moisture sensor, but it must be remembered that calibration is extremely difficult for QMS, the cost is much higher and operation requires caution. The use of the two pressure

j109

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

110

gauges, if applicable, offers, in our opinion, the best compromise between cost and performance: the signal remains almost constant, until it starts to drop, and the use of the ratio of the two pressure signals, instead of a simple differential measure, allows a good sensitivity. The only inconvenience that has been observed is that the ratio of the signals of two pressure gauges can sometimes show a baseline signal shift at the end of primary drying, mainly due to the relatively low accuracy of the Pirani instrument that can make the determination of the end of sublimation uncertain. However, the decrease is initially very sharp and in very good agreement with the signals of the other devices. The previous considerations may change if a third component is present in addition to inert and water, as in the case of the use of mixed solvents. In that case the response of the Pirani is affected by the presence of the new species, being unable to discriminate, even if the signal gets close to that of the capacitive gauge, when the composition of the chamber consists of only the inert species. The moisture sensor response, in principle, is independent of the presence of hydrocarbons, freon and carbon dioxide, but can be affected by low molecular weight alcohols, even if Roy and Pikal (1989) report having used it to monitor lyophilization of solutions containing up to 10% ethanol without any problems. The QMS in this case can obviously work efficiently and monitor the different species simultaneously. In any case one must bear in mind that the response of all these devices can be biased by the fact that the atmosphere that they measure is not exactly the average one in the chamber and depends on the positioning of the sensor, on the use of controlled leakage and on the fluid dynamics of the vapor inside the chamber: the moisture sensor can be positioned properly in the chamber, but the movements of the shelf for the stoppering generally limit its use to a peripheral position, while the Pirani is connected to the chamber by a short duct and the QMS has to sample the gas from the chamber (Rasetto, 2009; Rasetto et al., 2009). Another sensor recently proposed to measure water concentration in the chamber is the cold plasma ionization device. The inductive coupled plasma/optical emission spectroscopy employs a radio-frequency that creates cold plasma in a quartz tube under vacuum; the light emitted by the plasma is characteristic of the gas present in the plasma. The optical spectrum is analyzed and a very sensitive measure of the humidity is displayed in real-time. The cold-plasma sensor seems particularly promising: it is steam sterilizable, simple to integrate even in an industrial-scale freeze-dryer, reproducible and sensitive; the main drawbacks are the uncertainty on the final point determination, the problem of calibration and the dependence of the response on the probe location (Mayeresse et al., 2007). Figure 4.8 shows an example of the results that can be obtained by using the cold plasma sensor to monitor primary drying. As the vacuum was pulled down, the signal increased rapidly over 40 min to reach vapor saturation inside the freeze-drying chamber; it then remained saturated for about 12 h before decreasing according to a sigmoid-shaped curve. After 18 h of drying, the signal remained almost constant, thus indicating that most of the vapor had been replaced by nitrogen. After about 28 h secondary drying was started, with a shelf temperature rise and pressure drop: the sensor response detected this pressure drop and the baseline signal shifted to 8.5% of relative saturation.

4.3 Monitoring of the Primary Drying

Plasma signal, arbitrary unit

1.0 0.8 0.6 0.4 0.2 0.0 0

10

20

30

40

Time, h Fig. 4.8 Cold plasma sensor response during a freeze-drying cycle of 11000 vials with the sensor positioned on the top of the chamber (figure adapted from Mayeresse et al., 2007).

Figure 4.9 shows a comparison of the results obtained during the freeze-drying of a 5% aqueous sucrose solution in plastic syringes (6 mm inner diameter) carried out at
40  C and 26.6 Pa. Various devices are used to track the process, namely a Panametrics moisture analyzer, a Pirani gauge, a continuous weighing scale (manufactured by CHRIST) that provides a real time weighing under vacuum of a single product container (single syringe located inside the rack supporting the whole set of other containers or single vial placed on the shelf) and, finally, a LYOTRACK sensor (Adixen, France) that measures the humidity of the gas phase using the cold-plasma ionization principle (Hottot et al., 2009). Figure 4.9 shows that the sublimation end-point times derived from the CHRIST scale had always and systematically the lowest values (by about 3 h), in comparison with the other sensors, that presented comparable values (þ /
5%). The Pirani gauge and the LYOTRACK sensor showed similar humidity profiles, but significant noise was observed with the Pirani gauge signal, possibly due to air injection for total gas pressure regulation, whereas the LYOTRACK sensor was not disturbed at all by this air injection. Moreover, the Pirani gauge and the LYOTRACK sensor showed a saturation effect (with a nearly constant signal) about 300 min after the beginning of sublimation. It is also worth noting that the Panametrics hygrometer, which showed a continuously decreasing signal as a function of time, did not allow a clear sublimation end-point determination. Finally, Hottot et al. (2009) concluded that the LYOTRACK and Pirani sensors provided similar and consistent results, with clear and quite precise sublimation end-point determinations. Thus, the LYOTRACK sensor was recommended by these authors as being the more sensitive and the more precise sensor with a variation range between 100% and 1% (arbitrary units), instead of a variation range between 16 and 20 Pa in the case of the Pirani gauge. Moreover, this new sensor has the main advantages that it can be sterilized and it can be used with pure organic co-solvent formulations.

j111

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

signal, a. u. (dry basis) Dew point, °C

Mean humidity

LYOTRACK

PPirani, Pa

112

21 20 19 18 17 16 15 1.0 0.8 0.6 0.4 0.2 0.0 20

(a)

(b)

(c)

15 10 5 0 -25 -30 -35 -40 -45 -50 -55

(d)

0

200

400

600

800

1000

Time, min Fig. 4.9 Example of a freeze-drying cycle monitored using various devices: (a): Pirani gauge (filtered signal), (b) LYOTRACK sensor, (c) CHRIST scale, (d) Panametrics hygrometer. (Courtesy of Prof. J. Andrieu, LAGEP-CPE, Lyon, France).

Concerning the detection of the end of sublimation, as discussed above, it is generally assumed in previous works that it corresponds to the decrease of the monitored signal (e.g., pressure ratio or moisture concentration) to a low constant value, but it has been already pointed out that it may be difficult to define a respective numerical criterion, so that an uncertainty of a few hours can result. A significant difference in the value of the first derivative (end of plateau value), a zero value for the first derivative (end of the sharp signal decrease) or a zero value for the second derivative (inflection point in the middle of the sharp signal decrease) have been proposed as numerical criteria, but without a link to a theoretical background. In the case of the moisture sensor a special function called SEP(t) that uses the values of the total and partial pressure in the chamber and of the partial pressure in the condenser has been proposed on the basis of an inspectional analysis (Genin et al., 1996). The exact determination of the endpoint with the previous type of sensors may be made intrinsically difficult by the fact that the desorption rate from the fraction of dried material and from the chamber walls can have a value comparable to the value of the sublimation rate at the end of the primary drying. One last comment concerns the duration of the period in which the signal drops; this is related to the heterogeneity of

4.3 Monitoring of the Primary Drying

the batch and to the process conditions, and can be quite long, for example, about 4–5 h in the case shown in Fig. 4.7 which refers to a relatively large batch where the side vials are affected by radiation. It must be stressed that the point where the signal starts decreasing with a large slope does not correspond to the point where all the vials have completed drying: in fact, Roy and Pikal (1989) have evidenced that the signal of the moisture sensor does not change slope even when a significant fraction of the vials have completed primary drying; these results have been confirmed by Pisano (2009) using the various sensors in some experiments carried out on a batch in which the heat flow was deliberately different for a fraction of the vials, stopping the cycle and measuring the residual water in the vials: when the Pirani-Baratron pressure ratio starts decreasing, a relevant fraction of vials of the batch are already dried in such a way that the flow rate is reduced under a critical value inducing a variation in the composition of the gas in the chamber. 4.3.3.2 Monitoring the Primary Drying Using the Measurement of the Sublimation Flux A windmill sensor situated within the channel interconnecting the process chamber and the condenser has been proposed to monitor the batch: it can measure, at least qualitatively, the vapor flow, thereby providing information on the rate of freezedrying, as well as on the completion of the freeze-drying processing (Tenedini and Bart, 2001). A different arrangement had been proposed by Couriel (1977), with a cover that fits over the tray and two turbines at either end of the tray: such a device would strongly limit the vapor flow, with negative effects on the product temperature, as discussed in the following. The use of a mass flow controller to measure the gas flow necessary for pressure control and the tunable diode laser absorption spectroscopy sensor have been recently proposed to monitor the whole batch by means of direct or indirect measurement of the sublimation flow. When the pressure in the drying chamber is regulated by means of controlled leakage, the signal of the mass flow controller can be exploited to detect the end of the sublimation (Chase, 1998): in fact, the decrease in the partial pressure of water associated with the end of sublimation is responsible for the increase in the gas flow necessary to maintain a given chamber pressure. As an example, Fig. 4.10 shows the results obtained during the freeze-drying of 400 l of lactose in bulk: the nitrogen flow decreases sharply for approximately 2 h at the beginning of the operation as a consequence of the increase in the sublimation rate and the partial pressure of water as the lactose is heated. Over the following 13 h the nitrogen flow remains essentially constant, indicating relatively steady-state sublimation. The nitrogen flow starts to increase sharply after approximately 16 h of primary drying, as a consequence of the lower sublimation rate and, thus, of the reduction in the partial pressure of water in the chamber. During the next 2–3 h the nitrogen flow rate continues increasing until the slope decreases to zero, thus indicating that all the ice has sublimed and the chamber pressure is essentially made up of nitrogen gas. The tunable diode laser absorption spectroscopy (TDLAS) sensor is a real-time, non-invasive device that is used to measure water vapor concentration and gas flow velocity in the duct connecting the freeze-drying chamber and the condenser using

j113

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

114

1000

800 150 600 100 400 50

200

0

Nitrogen mass flow, sccm

Chamber pressure, µm Hg

200

0 0

10

20 Time, h

30

40

Fig. 4.10 Chamber pressure (solid line) and nitrogen flow (dashed line) during a freeze-drying cycle of 400 l of lactose in bulk (figure adapted from Chase, 1998).

Doppler-shifted near-infrared absorption spectroscopy (Kessler et al., 2004; Gieseler et al., 2007a). The concentration and gas velocity measurements are combined with knowledge of the duct cross-sectional area to determine the water vapor mass flow rate; the rate measurements can be integrated to provide a determination of the total water removed throughout the process. It has been proposed also for the determination of the heat transfer coefficient between the shelf and the product (Kuu et al., 2009) and the temperature of the product (Schneid et al., 2009): in both cases a mathematical model of the process is required to relate the sublimation flux measured by the TDLAS to the heat transfer coefficient (if the product temperature at the bottom of the vial is known) or to the product temperature (if the heat transfer coefficient is known). Moreover, because of its high sensitivity for residual moisture, the TDLAS sensor can be used to monitor secondary drying (Schneid et al., 2007). This device can be installed in lab-scale and in production-scale equipment and can be easily placed in a sterile process environment, even if it could be difficult to retrofit existing units as it should be located in the freeze-dryer duct; the main drawbacks are the cost and the difficulty of calibration: fluid flow modeling can be used to provide acceptable density and velocity determinations (Kessler et al., 2008). The MTM-based methods, that also can provide information about the sublimation flux, will be discussed in the next section. 4.3.3.3 Monitoring the Primary Drying Using Methods Based on the PRT Differently from the approaches previously discussed, methods based on the pressure rise test are able to give information about the state of the whole system, that is, the temperature and the residual water content, and can thus be used in a control loop; in addition, they also allow evaluation of the sublimation rate and, thus, the same considerations discussed above hold, even if the information can be obtained only at discrete times. Early works (Neumann, 1961, 1986; Nail and

4.3 Monitoring of the Primary Drying

Johnson, 1991; Willemer, 1991) investigated the transient pressure response measured during a PRT, using it as a method for determining the end of primary drying, or for an estimate of the product temperature based on the saturating steam pressure of the ice. Oetjen proposed and patented a method, called barometric temperature measurement (BTM), to estimate the temperature of the subliming interface using the value of the pressure at which the first derivative of the pressure rise curve has a maximum (Oetjen, 1999; Oetjen et al., 2000; Oetjen and Haseley, 2004). Mathematical models have been used in the past to compute the temperature of products undergoing lyophilization on the basis of pressure rise data (Milton et al., 1997; Liapis and Sadikoglu, 1998; Obert, 2001; Chouvenc et al., 2004; Velardi et al., 2008). In all these algorithms some parameters of the system are calculated by means of a regression analysis, that is, by fitting the measured pressure rise response to the values calculated using a mathematical model: what differentiates one method from the others is the mathematical model and the parameters estimated. Milton et al. (1997) proposed the manometric temperature measurement (MTM): the transient pressure response is mathematically modeled under the assumption that four mechanisms contribute to the pressure rise, namely the direct sublimation of ice through the dried product layer at a constant temperature, the increase in the ice temperature due to continuous heating of the frozen matrix during the measurement, the increase in the temperature at the sublimation interface when a stationary temperature profile is obtained in the frozen layer and, finally, the leaks in the chamber. The four contributions are considered purely additive; the values of the thickness and of the thermal gradient are needed but they are not known exactly. The values of the vapor pressure over ice, of the product resistance and the heat transfer coefficient at the vial bottom are determined with regression analysis. A modification of the previous model was proposed by Obert (2001) who considered also the desorption of the bound water during the primary drying, which can contribute to the increase in the total pressure, and the thermal inertia of the glass wall of the vial. The temperature at the bottom of the vial and the thickness of the frozen layer should be known in order to use this algorithm, but they are only guessed in the proposed procedure. The overall heat transfer coefficient is expressed adopting the heat and mass transfer steady-state hypothesis, while a non-linear regression analysis is carried out in order to estimate the vapor pressure at the interface, the mass transfer resistance in the dried product and the desorption rate. A more rigorous model, based on heat and mass balances, is used in the pressure rise analysis (PRA) proposed by Chouvenc et al. (2004). The thermal capacity of the portion of the vial glass in contact with the frozen product is taken into account in the heat balance for the frozen product. They assume that the temperature increase at the interface is the same as the mean product temperature rise, which would be exact if the temperature gradient along the ice during the PRT were constant. This assumption can be reasonable towards the end of the primary drying, when the thickness of the ice is small, but not in the first part of the drying cycle, when the frozen layer thickness is higher and accumulation effects prevent the temperature gradient from being constant. Chouvenc et al. (2004) also assume a constant temperature at the vial bottom during the PRT; however, the vial bottom is

j115

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

116

continuously heated during the process, while the heat removal at the interface is reduced due to the increased chamber pressure, which reduces the driving force for sublimation. The thickness of the frozen layer is estimated by considering constant sublimation flow between two subsequent PRTs: the value of the thickness in a generic run is retrieved by subtracting from the total initial mass of product the sum of the mass sublimed up to the time of the current PRT. The subliming interface temperature at the beginning of the PRT and the mass transfer resistance of the porous layer are estimated through regression. A more complex mathematical model (Sadikoglu and Liapis, 1997) has been used by Liapis and Sadikoglu (1998) to estimate the whole temperature profile in the frozen layer of the product and the position of the moving front. Many parameters are needed to perform the analysis, namely the diffusivity and the permeability of the porous layer, the shelf-vial heat transfer coefficient, the temperature and the partial pressure at the top of the vial, thus making its practical in-line application a complex task, even if feasible in theory. The dynamic parameters estimation (DPE) algorithm proposed by Velardi et al. (2008) solves the energy balance for the frozen layer to get the temperature profile in the product taking into account the different dynamics of the temperature at the interface and at the vial bottom. The energy balance in the frozen layer during the PRT can be described by the following equations: qT lfrozen q2 T ¼ rfrozen cp;frozen qz2 qt Tjt¼t0 ¼ Ti;0 þ

lfrozen

lfrozen

for

t > t0 ; 0  z  Lfrozen

~ w pw;i;0
pw;c;0 z KM Dhs ~ i;0 L
Lfrozen lfrozen RT

~ w pw;i
pw;c qT KM ¼ Dhs ~ qz z¼0 RTi L
Lfrozen

for

for t  t0

qT ¼ Kv ðTshelf
TB Þ for qz z¼Lfrozen

t  t0

0  z  Lfrozen

ð4:4Þ

ð4:5Þ

ð4:6Þ

ð4:7Þ

Thermodynamic equilibrium is assumed at the subliming interface; moreover, at the beginning of the PRT the heat fluxes at z ¼ 0 (interface) and at z ¼ Lfrozen (bottom of the vial) are assumed to be equal, thus Kv can be derived by equating the boundary conditions from Eqs. 4.6 and 4.7, both taken at t ¼ t0: 2 3
1 T
T L i;0 shelf frozen 5
Kv ¼ 4 ð4:8Þ ~ p
p Dhs K Mw w;i;0 w;c;0 lfrozen ~ i;0 L
Lfrozen RT

Figure 4.11a shows a sketch of the vial geometry with the coordinate system. Due to the contribution of gas conduction, Kv is dependent on the pressure, which varies during the PRT, but, as has been evidenced by Chouvenc et al. (2004), due to the thermal inertia of the system this dependence is not relevant during the PRT and the constant value given by Eq. 4.8 can be used. The radiation flux from the bottom is accounted for in the overall heat transfer coefficient that is estimated from numerical

4.3 Monitoring of the Primary Drying Pressure rise test and collection of Pc,meas data

(b)

(a)

Guess of Ti,0, K,

L frozen from Eq. 4.11 Kv from Eq. 4.8 T t=t from Eq. 4.5

dried layer

0

z =0

new Ti,0, K,

Integration of ODE system in t0 , tf

frozen layer

Is

z = Lfrozen

min

NO

Ti,0, K,

Pc t j

Pc,meas t j

j

2

? YES

Ti,0 , K,

,

Lfrozen, Kv, T z, t

Fig. 4.11 (a) sketch of the vial geometry with the coordinate system used by the DPE algorithm and (b) steps of the optimization procedure in the DPE algorithm.

regression. Radiation from the upper tray generally has a negligible effect due to the presence of the stopper that, at least partially, shields the product, and of the dried layer. Radiation from the side-walls affects the dynamics of a very low number of vials (only 6–7% of the vials of a batch in an industrial-scale apparatus are affected by radiation, that is, those placed at the side of the shelf) while in a small-scale apparatus, used for R&D purposes, the problem should be avoided by proper shielding; anyway, a small radiative contribution to total heat transfer is not a problem and, as long as the shape of the axial temperature profile is not significantly modified, the interface temperature is still predicted with good accuracy and the radiation has only the effect of increasing the value of the estimated effective heat transfer coefficient. Concerning the role of the vial wall in the thermal balance of the system, it can be relevant during the primary drying (Schelenz et al., 1994; Br€ ulls and Rasmuson, 2002; Hottot et al., 2006; Velardi and Barresi, 2008a). Nevertheless, it has been proven that the effect of the vial wall, with respect to the heat conduction in the axial direction and to the radiative flux from the chamber wall, can be accounted for in a one-dimensional model by using an effective heat transfer coefficient (Velardi and Barresi, 2008a). Moreover, the contribution of the vial wall to the dynamics of the system during the PRT has been shown, by means of numerical simulations, to be negligible and, for this reason, it has not been considered in the DPE algorithm, differently from Chouvenc et al. (2004) who included in the lumped model a fraction of the heat capacity of the glass to be determined by fitting (Fissore et al., 2011a). The total pressure is calculated taking into account a constant leakage in the chamber: Pc ¼ pw þ pin ¼ pw þ Fleak t þ pin;0

for t  t0

ð4:9Þ

j117

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

118

Applying the ideal gas law and rewriting the mass flow rate as function of the pressure driving force between the interface and the chamber, it follows that: ~ w Vc dpw;c ~ w pw;i
pw;c M KM ¼ Nv A ~ ~ RTi L
Lfrozen RTc dt

for

t  t0

ð4:10Þ

Equation 4.10 is valid if the contribution of all the vials is the same, that is, all the vials have the same values of L, Ti, K. Nevertheless, due to water vapor hydrodynamics and radiation effects the batch can be inhomogeneous (Barresi et al., 2008a, b): as an example, vials located at the edge of the plate sublime faster due to radiation from the wall. As a consequence, when in some vials the primary drying is completed, in other vials the main drying is still taking place and thus the number of vials contributing to the pressure rise during the PRT is different from Nv. For this reason it is generally assumed that methods based on the PRT give accurate and coherent data only in the first half of the primary drying (see, for example, Hottot et al., 2005; Tang et al., 2005), as at the end of this step a fraction of the vials can have completed sublimation before the rest of the batch. To take into account this effect, the term NvA in Eq. 4.10 can be multiplied by a parameter, c, that is equal to one at the beginning of primary drying and that decreases during drying (Rasetto et al., 2008): the value of this parameter can be estimated by means of the DPE algorithm or it can be calculated independently (Barresi et al., 2010). If the value of the gas temperature in the chamber, Tc, is not available, it can be substituted with the product temperature at the interface with a negligible error. The actual thickness of the frozen layer is calculated from a mass balance written across the moving interface, which is solved with the previous equations. This balance can be integrated in time, using for example the trapezoidal rule of integration, between the previous PRT and the actual one, thus obtaining:

Lfrozen ¼

ð
1Þ Lfrozen


" # ð
1Þ ð
1Þ ~ w K pw;i;0
pw;c;0 K ð
1Þ pð
1Þ M w;i;0
pw;c;0 t0
t0 þ ð
1Þ ð
1Þ ~ Ti;0 L
Lfrozen 2 RDr L
Lfrozen Ti;0

ð4:11Þ

where Dr ¼ rfrozen
rdried and the superscript “(
1)” refers to quantities calculated or measured in the previous PRT. The steps of the DPE algorithm are summarized in Fig. 4.11b. In previous equations, and in particular in Eqs. 4.5 and 4.6, the gas flow rate in the dried layer was calculated including the dependence on cake thickness, but neglecting the resistance of the stopper. As an alternative the global mass transfer resistance, Rp, can be used: ~w 1 K M ¼ ~ i L
Lfrozen Rp RT

ð4:12Þ

If the stopper resistance is not negligible, then the global resistance to vapor flow in the dried layer is given by:

4.3 Monitoring of the Primary Drying

1 ¼ Rp þ Rs

 
1 ~ i L
Lfrozen RT þ Rs ~w K M

ð4:13Þ

The DPE algorithm has been demonstrated to work efficiently both under heat transfer and mass transfer control (Barresi et al., 2009c) as it estimates both the heat transfer coefficient and the resistance to the mass flow. This is an advantage with respect to other approaches proposed in the literature which require one to assess if the system is under heat or mass transfer control (Liapis and Litchfield, 1979; Litchfield and Liapis, 1982). If the valve used to separate the drying chamber from the condenser during the PRT is fast closing, as is generally the case in medium and small scale equipment, its dynamics can be neglected; anyway, it has been shown that a slow dynamics can be accounted for in the algorithms based on the PRT (Oetjen and Haseley, 2004; Chouvenc et al., 2005). As an example, Chouvenc et al. (2005) assumed that part of the flux generated by sublimation of ice and, eventually, to a lesser extent, by the chamber leak rate of inert gas, is lost, during the transient closing period; this fraction is considered variable with time, in the time interval required to get the valve fully closed, and is a characteristic of the isolation valve and independent of the operating conditions. This valve characteristic function was then taken into account in the PRA algorithm. The sensitivity of the methods based on the PRT depends on the chamber volume and on the sensitivity of the pressure gauge, in addition to the operating conditions and the nature of the product that influences the value of the sublimation rate. According to the literature (Milton et al., 1997), the sensitivity depends on the ratio of the active subliming surface area to the chamber volume and it decreases as the batch size decreases. Milton et al. (1997) do not specify any lower bound needed to get reliable data. In this context, it may be noted that in production apparatus the most sensitive capacitive gauge available is generally used, with a full scale of only 100 Pa, while in lab-scale freeze-dryers often a gauge with larger scale is used. Milton et al. (1997) carried out their measurements successfully with a value of the chamber volume to the subliming surface area corresponding to 1.28 m; Tang et al. (2006b) and Gieseler et al. (2007b) reported that in order to get accurate temperature measurement using the MTM algorithm a value not higher than 3.5 m was required. We were able to carry out PRTs even with a very small number of vials, up to a chamber volume to subliming surface area ratio of 58 m that corresponds, for the cases investigated, to a sublimation flow/chamber volume ratio (the parameter we suggest to consider) equal to 1.4  10
2 kg h
1 m
3 (Pisano, 2009). Figure 4.12 shows a comparison between the temperature of the product at the bottom of the vials estimated by DPE, MTM and PRA (assuming c ¼ 1); the values of the shelf temperature and of the temperatures measured by some thermocouples are also shown in the upper graph, as well as the ratio of the signal of the Baratron and of the Pirani gauges. One point to be evidenced is that, generally, the estimated product temperature decreases near the end-point, but this drop may be only an artifact because a fraction of vials, the edge-vials, has already finished subliming while DPE,

j119

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

120

(a)

Temperature, °C

10 0

PBaratron/PPirani

1.8 1.6 1.4 1.2 1.0 0.8

20

-10 -20 -30 -40 -32

(b)

TB, °C

-34

-36

-38

-40 0

5

10

15

20

Time, h Fig. 4.12 Monitoring of the freeze-drying cycle of 10% by weight sucrose solution (Nv ¼ 175, dv ¼ 14.2  10
3 m, Lp ¼ 7.2  10
3 m, Pc ¼ 10 Pa). (a) Comparison between bottom product temperature estimated by DPE (~) and the values measured by thermocouples in close contact with the bottom of the vial (dotted

lines). The heating fluid temperature (solid line) and the Pirani to Baratron pressure ratio (dashed line) are also shown. (b) Comparison between the predictions of the temperature at the bottom of the vial obtained using various algorithms (&: MTM, *: PRA, ~: DPE with c ¼ 1).

MTM and PRA continue interpreting pressure rise curves assuming batch uniformity, or rather a constant number of subliming vials. Thus, a decrease in pressure rise, corresponding to a lower sublimation rate, may be interpreted by these algorithms as a reduction in the front temperature. From comparison of these methods it is possible to see that the estimations provided by MTM are reliable only in the first half of the primary drying, after which the estimated temperature exhibits a strong decrease, while the estimations of DPE, that uses a much more complex model to describe the pressure rise occurring during a PRT, are consistent for a larger fraction of the primary drying.

4.3 Monitoring of the Primary Drying

The decrease in the interface temperature estimated by these methods based on the PRT was already reported by Oetjen and Haseley (2004), who proposed to use it as an indication of the end of sublimation; according to the results shown this is not correct, or at least should be interpreted as an indication of the end of the sublimation in the first vials, but in our opinion this criterion is not very robust. Moreover, recent preliminary results show that batch heterogeneity is surely responsible, at least partly, for the observed behavior, but this can also be due to the optimization algorithm and caused by problem ill-conditioning (Fissore et al., 2011a). Anyway, the end of primary drying could be reasonably estimated by extrapolating the predictions of the interface position obtained in the initial part of the run. The estimation of the solvent flux can be used as an alternative way to detect the endpoint of the primary drying: when the solvent flux decreases below a specific limit the main drying can be considered finished. As discussed above, various approaches can be used to obtain estimations of the sublimation flux, for example, the slope of the curve of the pressure rise at the beginning of the PRT, a balance inserted in the drying chamber, the DPE, a system like the TDLAS or the mass flow of nitrogen in the case of controlled leakage. Nevertheless, a decrease in the fluid temperature, and thus in the heat flux given to the system, can also decrease the solvent flux, so that this information has to be coupled with other measurements to assess the end of primary drying: if product temperature is approaching the shelf temperature, then primary drying can be considered completed. Recently, an integrated criterion was proposed to estimate the end of the primary drying: DPE is used to estimate the sublimation flux, and the actual mass of sublimed solvent and the frozen layer thickness are calculated after each PRT by integration of the flux over time; the average mass sublimation rate is then calculated and when this is lower than a fixed bound, the primary drying is considered completely finished (Velardi and Barresi, 2008b; Pisano, 2009; Pisano et al., 2009, 2010a). The steps of the algorithm are the following: .

.

.

.

Do a PRT and run the DPE algorithm to obtain a full estimation of the state of the system in terms of product temperature and heat and mass transfer coefficients. Calculate the current solvent flux using either DPE outcomes or the slope of the pressure rise curve at the beginning of the test. Integrate the solvent flow rate over time so as to calculate the current sublimed mass of solvent. Calculate a stop coefficient that is directly related to the average subliming mass rate and is used as a reference for establishing whether or not the main drying is finished: C¼

mðti Þ
mðti-1 Þ 1 ti
ti-1 mtot

ð4:14Þ

where ti is the current time instant and ti
1 is the time instant at which the previous PRT was done. . Compare this coefficient with a lower bound fixed by the user, which consists in the percentage variation of the sublimed solvent mass with respect to the total one

j121

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

122

.

(for example 1% h
1). If C is lower than this limit and the estimated frozen layer thickness is not close to the initial one, confirming that the process is not at the beginning, when the sublimation rate can be low due to the low initial product temperature, the primary drying can be considered finished. If the sublimation step is not yet finished, it could be interesting to estimate the time left to the endpoint. This can be easily done using a mathematical model that describes the dynamics of the process (e.g., the same as in the DPE algorithm) and uses the process parameters estimated by DPE.

Figure 4.13 compares the endpoint time estimated using the previously described algorithm and other devices. The cycle has been run loading in the freeze-dryer only the 15 vials monitored by the balance, thus avoiding any difference between the dynamics of the vials of the balance and that of others of the batch. We can see that there is very good agreement between the flow rate measured by mass measurements and by the pressure rise technique. On the contrary, it can be noticed that the product temperature response detected the end of the two monitored vials 10 h before the real endpoint, which can be explained by a higher drying rate of the vials wherein the thermocouple probe is inserted. The endpoint time, equal to 18.5 h, agrees with the 25

15

250

10 240 5 230

Pressure ratio

20

260

2.0

0

-4

2.5×10

(b)

Water content, g

Sublimation flux, kg m–2s–1

(a)

Sublimed mass, g

Temperature, K

270

-4

2.0×10

-4

1.5×10

-4

1.0×10

-5

5.0×10

0.0 5

10

15

20

Time, h Fig. 4.13 Example of the results obtained during a freeze-drying cycle using various devices to track the primary drying stage of 10% by weight sucrose solution in a pilot-scale freeze-dryer. The freezing phase was run at 223 K for about 5 h, while during the primary drying phase the fluid temperature was set at 253 K and the chamber pressure at 10 Pa (Nv ¼ 15, dv ¼ 14.2  10
3 m, Lp ¼ 7.2  10
3 m, not shielded). (a) Comparison between the endpoint time evaluated by means of the algorithm described

(c) 1.5

1.0

16 14 12 10 8 6 4 2 0

(d)

5

10

15

20

Time, h in the text (vertical dotted line) and the product temperature response (dashed lines); the shelf temperature (solid line) and the sublimed solvent mass evolution calculated by integration of the flux estimated from the slope of the pressure rise curve (symbols) are also reported. (b) Comparison between the ice sublimation flux calculated from the pressure rise curves ( ) and from the mass measurements (*). (c) Pirani–Baratron pressure ratio. (d) Evolution of the water mass left in the 15 monitored vials measured by the balance.

.

4.3 Monitoring of the Primary Drying

value measured by both the Pirani–Baratron pressure ratio and product mass evolution. From the previous discussion it becomes evident that various devices should be used together to monitor primary drying. The LyoMonitor (Barresi et al., 2009a) is an example of a system that can manage various devices and collect their measurements to monitor primary drying. Currently, as shown in Fig. 4.14, it includes various sensors, namely: .

.

.

. .

A multi-point wired thermometer composed of a set of nine copper–constantan thermocouples, a conditioning circuit and a commercial multimeter equipped with a multiplexer. An innovative wireless thermometer (Vallan et al., 2005a) that can manage one or more measurement modules equipped with 14 thermocouples each: it sends the results to a “reader” placed outside the vacuum chamber and connected by means of a serial interface to a PC that schedules, acquires and collects the measurements. The reader powers the thermometer and the modules through the same radio-frequency link that is employed for data communication, so that the modules can work without batteries. A special weighing device (Vallan et al., 2005b; Vallan, 2007) working inside the vacuum chamber and able to measure contemporaneously the weight and temperature of a group of vials during the drying process (see Section 4.3.2). A valve control and an acquisition system for the PRT. Pressure and moisture sensors: the system is able to acquire the output signal of a thermal conductivity gauge, a capacitance manometer and a moisture analyzer by means of an external multimeter that is interfaced to the PC through the IEEE-488 interface.

Moreover, some other process variables that are measured in the freeze-dryer using embedded devices (i.e., shelf and fluid temperatures, controlled-leakage valve opening and inert mass flow rate for pressure control) can be acquired through a dedicated RS485 interface, thus providing a complete evaluation of the status of the system. A monitoring system like LyoMonitor allows one to monitor both single vials, using the soft-sensor, as well as the whole batch, using the PRT and the DPE algorithm. Moreover, the soft-sensor can be coupled to the DPE algorithm: the values of the interface temperature, of the heat transfer coefficient between the fluid and the bottom product and of the mass transfer resistance in the dried layer estimated by the DPE algorithm can be used to initialize the equations of the observer as this strongly improves the convergence of the algorithm of the observer (Barresi et al., 2009c). DPE and soft-sensor give similar information about the state of the product, but it hardly happens that the values they provide are the same: data reconciliation is thus required. As far as the detection of the ending point of primary drying is concerned, LyoMonitor allows one to track various signals, that is, the Pirani–Baratron ratio, the reading of the moisture analyzer, the measurement of the weight loss of a group of vials, the estimation of the sublimation flux obtained by means of DPE and the residual ice content given by the soft-sensor. As previously

j123

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

124

Fig. 4.14 Scheme of the LyoMonitor system: (1) multi-point thermometer equipped with fault diagnosis for product and chamber monitoring; (2) proprietary in-line balance; (3) signal acquisition from embedded devices; (4) capacitive moisture sensor; (5) pressure sensors: capacitance manometer and thermal conductivity gauge; (6) “dynamic parameters

estimation” (DPE) with fast pressure data acquisition through an acquisition board (6a) and valve control for pressure rise test (6b); (7) miniaturized radio-controlled thermometer for the vials weighed by the balance; (8) “smart vial” observer (8a) and wireless additional thermometer (8b); (9) QMS.

4.4 Control of the Primary Drying

discussed, all these signals can be used to detect the end of the primary drying, with some limitations, and thus the user should carefully monitor all of them in order to assess the end of this stage.

4.4 Control of the Primary Drying

As stated in the introduction, the control of a freeze-drying process, with the goal to reduce the time required to reach the desired amount of residual solvent, is a challenging task due to the impossibility of measuring in-line the variables of interest, that is, the product temperature and the residual solvent content. Some papers have appeared in the past about this issue and proposed to use a mathematical model of the process to calculate off-line the optimal operating conditions (i.e., the shelf temperature and the chamber pressure) for the primary drying. A very simple approach consists of carrying out the process using constant values for the chamber pressure and for the temperature of the heating shelf: Fig. 4.15 shows an example of the results that can be obtained using the detailed onedimensional model of Velardi and Barresi (2008a) to calculate the time required to

55

0K

50

24

45

5K

24

Primary drying time, h

40

0K

25

35 30 25 20 15 10 5

260 K

300 K

280 K

0 0

10

20

30

40

50

60

Pressure, Pa Fig. 4.15 Effect of chamber pressure and heating shelf temperature on the primary drying time for constant shelf temperature. The locus corresponding to the minimum of the primary drying time for the various shelf temperatures is also shown (dotted line). The dashed line

corresponds to the values of chamber pressure and of shelf temperature that satisfy the constraint on the maximum product temperature. Case study: 5% solution of Bovine Serum Albumin solution (Fissore et al., 2008a).

j125

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

126

complete the primary drying in a vial placed at the center of the shelf, for which radiation can be neglected. For a given Tshelf it is possible to find an optimum value of the chamber pressure that minimizes the drying time. In fact, when operating at very low vacuum, a pressure increase leads to a diminution of the drying time because heat transfer is improved; however, if the pressure is raised beyond a certain value, the time required to complete the primary drying starts to increase because the driving force for mass transfer becomes too small and sublimation takes place slowly. The optimum condition is visible in Fig. 4.15 (dotted line) for the curves calculated using values of Tshelf lower than 260 K, while it is shifted to pressures higher than 60 Pa if Tshelf is higher than 260 K. The calculated couple of values of Tshelf and of chamber pressure that minimize the primary drying step does not take into account the presence of the constraint given by the maximum temperature that allows safe operation without any denaturation of the product. If the maximum product temperature is taken into account (240 K in this example) the optimal pressure becomes, in this case, the minimum that can be obtained in the apparatus (dashed line). Some advantages can be obtained if the operation is carried out with a shelf temperature and a chamber pressure that vary during the operation. Liapis and Litchfield (1979) used a quasi-steady-state model of the process to optimize the primary drying; constraints are placed on the scorch temperature of the dried product and on the melting point of the frozen product. As the temperature profile in the vial is fully known from the mathematical simulation of the process, they distinguish between a process whose dynamics is controlled by the heat transfer from the shelf and a process whose dynamics is controlled by the mass transfer: in the first case the manipulation of the shelf temperature is effective, while in the second case it is necessary to manipulate the chamber pressure. A similar approach has been used by Lombra~ na and Diaz (1987a, b), while variational calculus involving a detailed multidimensional model of the process has been used by Sadikoglu et al. (2003). Fissore et al. (2008a) proposed to continuously manipulate the shelf temperature so that the maximum product temperature (Tp,max) is equal to the maximum allowable  ) at any instant. The control law is thus expressed as: value (Tp;max Tshelf ðtÞ ¼ f ðTp;max ; tÞ

ð4:15Þ

A mathematical model of the process is required to link the shelf temperature to the maximum product temperature. An example of the results that can be obtained using this control algorithm is given in Fig. 4.16a. At the beginning of the drying phase Tshelf is raised at a 0.5 K min
1 constant rate; after about 2 h of drying, when the shelf temperature reaches a value of 281 K, the product temperature becomes equal to the maximum allowed value (240 K) and the control logic overrides the initial heating program (solid lines): the temperature of the shelf starts being regulated in such a way that Tp,max remains constant and equal to the target value. In this way the primary drying is completed in 13.5 h, as can be seen from the time evolution of the position of the sublimation front (Fig. 4.16b). If the shelf temperature is maintained constant, the product temperature is left free to vary (dashed lines); thus, Tshelf must be fixed at a

26

14

8 0

5

10

15

20

Time, h Fig. 4.16 (a) and (b) Comparison between the model-based control strategy (solid lines) and the constant Tshelf (dashed lines) for the freezedrying of a 5% solution of Bovine Serum Albumin (dv ¼ 14.2  10
3 m, Lp ¼ 8  10
3 m,  ¼ 240 K). (a) Time evolution Pc ¼ 15 Pa, Tp;max of the temperature of the heating shelf and of the maximum temperature of the product; (b) moving front position. (c) Effect of chamber pressure and heating shelf temperature on the

K

12

K

16

0 25

18

10

0.8 1.0

20

0 28

0.6

22

K

0.4

(c)

0 27

(b)

0.2

T

K

Tp, max

24

=

Tshelf

Primary drying time, h

(a)

0 26

280 270 260 250 240 230 0.0

elf sh

Relative front position, 1-(Lfrozen/L) Temperature, K

4.4 Control of the Primary Drying

300 K 320 K 0

5

10

15

20

Pressure, Pa primary drying time when the process is controlled using the model-based control strategy that maintains the maximum product temperature at a value lower than 240 K. The locus corresponding to the minimum of the primary drying time for the various shelf temperatures is also shown (dotted line). Case study: 5% solution of Bovine Serum Albumin (Fissore et al., 2008a).

lower value in order not to exceed the maximum allowed temperature that is approached only at the end of the drying phase: a longer primary drying time is thus obtained (the moving front reaches the bottom of the vial after 22 h). Various simulations of the primary drying can be carried out for different values of the pressure in the chamber and using the previously described control strategy: results are given in Fig. 4.16c. In these simulations Tshelf is initially ramped at 0.5 K min
1 to the set-point value of Tshelf, which parametrizes the curves. Then Tshelf is maintained constant until the maximum product temperature reaches the respective limiting value (set equal to 240 K in this example). At this point the previously described control strategy takes over. It is possible to identify also in this case a curve (dotted line) which represents the locus of the values of chamber pressure and initial setpoint of Tshelf that ensure minimization of the time required to complete the primary drying and to compare this locus with the analogous one obtained in the case of constant Tshelf. For example, if Tshelf is maintained constant at 280 K and the constraint on the maximum product temperature is active, the minimum drying time (11.1 h) is obtained at a pressure of 3.3 Pa, while, when the optimal heat input strategy is adopted, a lower minimum drying time is found (10.25 h) with a higher chamber pressure (6.75 Pa), thus with minor energy consumption. Concerning the case of constant shelf temperature (Fig. 4.15), it can be observed that the best operating conditions are those with Tshelf ffi 300 K and with a value of the chamber pressure as low as possible, resulting in a minimum time of 10.5 h. In the case of a variable heating strategy, the pressure should also be as low as possible, but with an

j127

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

128

initial set-point Tshelf ffi 320 K the main drying ends after 9 h. Thus, it can be concluded that the manipulation of Tshelf allows one to carry out the process faster, under the constraint given by the maximum allowable temperature of the product. Chamber pressure is maintained constant: it is varied in the case of emergency, when, due to errors in the off-line optimization or to malfunctioning of the control system, the maximum temperature of the product increases beyond the limit value. In such a situation pulling the vacuum down immediately works as a thermal switch with an instantaneous result, whereas cooling the shelves requires a longer time (Rey and May, 2004). The main drawback of all these model-based algorithms is that they require that the model describes perfectly the dynamics of the process and that all the parameters and all the variables of the process are known. As this seldom happens, it is necessary to use one of the techniques described in Section 4.3 to monitor the primary drying in order to know the value of the temperature of the product, that is the controlled variable, and, thus, to calculate the control action. The first automatic control systems proposed date back to the early 1960s, and were based either on the barometric temperature measurement of the batch or on the monitoring of the resistivity of one or more sampled vials. Even if they had many limitations, and up to recent times were never really applied in industrial applications, especially those of the second kind, they introduced the concept of modern closed-loop control, and some control strategies that are still valid (Nail and Gatlin, 1985; Jennings, 1999). Thus Rey (1963, 1976) proposed to manipulate the shelf temperature to maintain constant resistivity, measuring eventually more vials to take into account nonhomogeneity of the batch. An improved monitoring method, based on the resistivity measurement of the sample and of a comparison material, together with the product temperature, was proposed by Jennings (1982) to increase product yield. Rieutord (1965) also proposed to use the resistivity measurement as an alternative to temperature measurement, but showed that by acting on the total pressure in the drying enclosure it was possible to supervise and regulate the energy supply to the product: to this end he realized the controlled bleeding system, using an inert gas to control the pressure in the chamber; the shelf temperature could be kept constant, or regulated independently. The possibility of independent temperature control of the different shelves was also considered. The manipulation of shelf temperature and chamber pressure for control of heat transfer have very different performances: the first one is slow and unstable, as a consequence of the large lag compared to the fast response of resistivity, while direct control of inert gas flow is more rapid, but does not inherently balance the heat transfer rates. Jefferis (1981, 1983) elaborated a cascade control algorithm in which dryer pressure was a function of product resistivity and condenser temperature. A simple control system for the shelf temperature based on the BTM output was first proposed by Neumann (1961) and then described by Oetjen et al. (1962). Neumann (1963) later proposed a simple control system where chamber pressure is manipulated in order to maintain the product at the desired temperature; this was obtained either by means of a throttling valve connecting the chamber to the

4.4 Control of the Primary Drying

condenser, or acting on the condenser temperature. Willemer (1987) showed an example of nominal-actual value regulation of the shelf temperature, measurement of the product temperature by BTM and its control by manipulation of the chamber pressure during the sublimation phase. The thermodynamic lyophilization control system was later described by Oetjen (1999) and by Oetjen and Haseley (2004): it uses the results of the BTM algorithm and a set of heuristics for the calculation of the control actions. Tang et al. (2005) and Pikal et al. (2005) proposed, and patented, an expert system, named SMARTÔ Freeze-Dryer, for manipulating the shelf temperature and the chamber pressure using a simple model and the results obtained by means of the MTM algorithm. Gieseler et al. (2007b) validated experimentally the SMART FreezeDryer with different types of excipients, formulations (involving crystalline and amorphous products) and vials, evidencing that the algorithm can be a useful tool for development of a lyophilization cycle during a single freeze-drying run, but the quality of the cycle optimization is dependent upon the accuracy of the parameters which must be provided by the user (e.g., collapse temperature, vial cross-sectional area). Tenedini and Bart (2001) patented a method for monitoring both the primary and the secondary drying by using the measurements provided by a windmill sensor and some “rules of thumb”. Moreover, they proposed to save energy by automatically disabling the vacuum pump when the pressure in the process chamber falls below a predefined set-point; the vacuum source is then reconnected if the process chamber pressure rises above a second predefined set-point. They pointed out that by controlling pressure within the process chamber only through selective connecting/disconnecting of the vacuum pump, a higher level of product purity is achieved compared with the conventional requirement of an inert gas bleed system, while still providing a comparable level of pressure control. The speed of freeze-drying can also increase as a consequence of the higher specific heat of the water vapor (that is the prevalent species in the chamber) with respect to the inert gas. In a patent by Lambert and Wang (2003) it is proposed to monitor in-line the cake resistance by inserting a tube in some vials and measuring the pressure drop of an inert gas; it is suggested that the results from the measuring system are used to provide a control signal to a control system, but it seems difficult to really apply the proposed method, even in a laboratory apparatus. Barresi et al. (2009a, c) proposed to use DPE in a control loop where the heating fluid temperature is manipulated. This control algorithm, named LyoDriver, uses the estimations obtained by means of the DPE algorithm (product temperature, heat transfer coefficient between the heating fluid and the product at the bottom of the vial, and the diffusivity coefficient of the vapor in the dried layer), as well as some process variables (i.e., the temperature of the fluid, the pressure in the chamber, and the cooling rate of the freeze-dryer) and a simplified mathematical model for the primary drying (Velardi and Barresi, 2008a). In order to run LyoDriver the user must set the values of the prediction horizon (hp), i.e. the number of time for which the controller computes a proper heating policy on the basis of the prediction of evolution of product temperature, and the time between a control action and the next one.

j129

j 4 In-Line Product Quality Control of Pharmaceuticals In Freeze-Drying Processes

130

After that, LyoDriver calculates a sequence of suitable set-points for the fluid temperature (Tfluid,sp), one for each control interval all through the prediction horizon, in such a way that the product temperature is as close as possible to its target. At the beginning of primary drying, when the temperature of the product is well below the upper limit, the heating fluid temperature is raised at its maximum rate compatible with the actual machine capacity and, in this way, the product approaches its limit as fast as possible. After this first step, a PRT is performed and LyoDriver, using the DPE algorithm, estimates the time-varying product temperature at the bottom of the vial (where the temperature is higher) over the whole prediction horizon and recalculates the optimal heating policy accordingly to the current system state. This is regularly repeated at each successive DPE run so that potential mismatches between the LyoDriver model predictions and the actual process behavior can be taken into account. If the estimated product temperature approaches its limit, LyoDriver reduces the shelf temperature in such a way that the product is maintained below its target. Another point to be stressed is that the shelf temperature evolution is calculated taking into account the real dynamics of the heating and cooling systems. Two control algorithms have been proposed and compared: the former is a simple feedback controller that calculates the control action as a function of the difference between the product temperature at the bottom of the vial and the maximum allowed value, while the latter relies on a model-based algorithm that calculates the fluid temperature to maintain the temperature of the product at the bottom of the vial equal to the maximum value (Pisano et al., 2010a). Both control laws use an unsteady-state model of the process, made of a set of equations that supplies the evolution of both product temperature and frozen layer thickness. These equations are integrated from the initial time (t0), which is zero for the first run and equal to the time elapsed from the first test in next ones, up to the prediction horizon (tN) set by the user, or up to the estimated end-point of the primary drying phase, that corresponds with the time at which the frozen layer thickness is equal to zero (tN ). If the feedback logic with a simple proportional controller is used, the optimal heating strategy is calculated throughout all the prediction horizon considering the optimal sequence of set-point shelf temperatures as a piecewise-linear function, according to the relationships: t0  t < t1

Y Tshelf ;sp ¼ Tshelf ðt0 Þ þ KP ðTB ðt0 Þ
TB;sp Þ

t1  t < t2

Y Tshelf ;sp ¼ Tshelf ðt1 Þ þ KP ðTB ðt1 Þ
TB;sp Þ ð4:16Þ

.. . tN
1  t < tN

Y Tshelf ;sp ¼ Tshelf ðtN
1 Þ þ KP ðTB ðtN
1 Þ
TB;sp Þ

Here, tj
tj
1 defines the time interval between two control actions, KP is the proportional gain of the controller, TB(tj) – TB,sp is the difference between the product temperature at the bottom of the vial and its set-point, that is the temperature to which the product has to be driven. One aspect to be stressed is that the set-point of TB might be lower than the target temperature set by the user, because the controller iteratively calculates a new target in order to guarantee that possible temperature overshoots in

4.4 Control of the Primary Drying

the prediction horizon are maintained under the maximum temperature allowed by the product. Moreover, LyoDriver calculates TB,sp also taking into account temperature rises caused by regular PRTs. The control interval usually corresponds to the time between two subsequent PRTs and only the first control action of the calculated sequence is actually applied: the heating policy is regularly recalculated after the system state has been updated. As an alternative, more control actions can be applied between a DPE and the next one in order to disturb less the dynamics of the process. The value of the gain of the controller has to be calculated using some performance criteria, for example, the minimization of the integral square error (ISE) between the product temperature and the set-point value from the current time (t0) up to the prediction horizon (tN): ð tN ð4:17Þ minðISEÞ ¼ min ðTB;predicted ðtÞ
TB;sp Þ2 dt KP

KP

t0

The simplified model of Velardi and Barresi (2008a) is used to estimate the time evolution of the product temperature required by Eq. 4.17. If a model-based approach is implemented, the optimal sequence of shelf temperature set-points throughout the prediction horizon is calculated as a piecewise-linear function in such a way that the bottom product temperature is equal to the target value. Again, the simplified mathematical model of Velardi and Barresi (2008a) can be used for this purpose: 2 0 1 3
1 1 Lfrozen ðt0 ÞA 5 4 @
1 þ t0 t 10 nm), before aggregation occurs with relatively narrow necks, as in Fig. 5.1b (Boonamnuayvitaya et al., 2006). In practice, a two-step process with a change of pH is often used, offering more possibilities to engineer the gel structure. Not only the size distribution of particles and pores, but also the porosity can be adjusted, simply by changing the silica concentration in the starter solution. Controlled aging of the wet gel is used to strengthen the network: hydrolysis and condensation reactions are to some extent reversible, and solid can be dissolved from thermodynamically unfavorable regions to condense at more favorable places. During this Ostwald ripening, small pores are filled and solid bridges grow, thereby reducing curvature – and area – of the solid/liquid interface (H€ using and Schubert, 2006). When the gel is aging, new bonds can also be formed between neighboring branches of the network, which are brought into contact by thermal fluctuations. As a consequence, the gel not only stiffens, but also shrinks linearly by a few percent and expels some of the pore liquid; this is called syneresis (Scherer, 1999). The described synthesis and post-processing of silica gels is performed at (or slightly above) room temperature. For commercial silica gels, water glass Na2SiO3 is used as a cheap silica source; then, before the gelation step, sodium is ion-exchanged from the aqueous sodium silicate solution to get a silicic acid (Schwertfeger et al., 1998). 5.2.1.2 Resorcinol-Formaldehyde (RF) Gels RF gels are probably the most intensively studied organic gels with high porosity applications. Dry RF gels are dark red and transparent (H€ using and Schubert, 2006); by pyrolysis in an inert atmosphere, they can be converted into electrically conducting

j157

j 5 Understanding and Preventing Structural Changes During Drying of Gels

158

Fig. 5.2 Synthesis of RF gel: addition and condensation reaction (simplified from Lin and Ritter (1997)).

carbon gels (black and opaque) making them suitable for applications complementary to those of silica gels. The two steps of gel synthesis are sketched in Fig. 5.2: in an aqueous (or organic) solution of resorcinol and formaldehyde (molar ratio 1: 2), an addition reaction is catalyzed by sodium carbonate to form hydroxymethyl derivatives; subsequently, these are interlinked by methylene (–CH2–) or methylene ether (–CH2OCH2–) bridges in endothermic condensation reactions (Al-Muhtaseb and Ritter, 2003). Historically, sodium carbonate is referred to as a catalyst, although its only role is to tune the pH value (Job et al., 2004). Gel structure can be tuned by the molar ratio between resorcinol and catalyst (R/C), which controls the relative rate of the necessary addition reaction; the reasoning is analogous to the silica gel case: low R/C favors the addition reaction so that polymer-like gels are formed (Fig. 5.1a); for high R/C, emerging hydroxymethyl derivatives are immediately consumed in the condensation reaction, leading to colloidal gel networks (Fig. 5.1b). Again, the amount of solvent determines the porosity of the solid network. Common post-processing of the wet gel includes curing at elevated temperature (80  C) and aging in a dilute acid, both for several days. This increases the degree of crosslinking and hence strengthens the gel network, fortunately, without the negative effect of shrinkage. Besides the silica and RF model systems, there is a large variety of further technical gels. Other material classes are metal oxide gels, such as alumina gels, which can be pure or serve as a carrier for metal compounds (catalytic sites); and inorganic/organic hybrid gels, which can combine the positive properties of both species. Doping or modification of gels is mostly done in the wet state, but also by impregnation of the dry gel (H€ using and Schubert, 2006). 5.2.2 Properties of Dry Gels

First, a list of outstanding physical properties of highly porous dry gels will be given, all of which are determined by solid structure. In numerous research projects all over

5.2 Gels and Their Applications – Quality Aspects

the world, a large amount of knowledge has been accumulated on how the structure can be engineered by setting the process parameters for gelation and subsequent aging. The challenge is, therefore, to preserve the structure during liquid removal. Historically, the only successful technique was supercritical drying (see Section 5.4.3) and the resulting dry gels were named aerogels. Note that the term aerogel is often used for highly porous gels, no matter the chosen drying route; in this text, however, we will use it for gels dried under supercritical conditions. Alternative routes are reflected in the names for the dry gels: cryogels are the result of freeze-drying, xerogels are prepared by convective drying or, in a wider sense, evaporative drying. The following properties (Hrubesh, 1998; H€ using and Schubert, 2006) only apply to gels with low solid fraction, that is, to porous media that cannot easily be created by other production routes: .

.

.

.

Extremely high porosity/extremely low density: Dry gels can have the highest porosity and the lowest density of all solid materials but the actual values can vary over a wide range, depending on the choice of the gel system and its synthesis conditions. For silica aerogels, the most extreme values have been reached with bulk densities as low as 0.004 g cm 3 and porosities higher than 99% (Tillotson and Hrubesh, 1992). High specific surface area: Gels are built from primary particles that can be approximated by mono-sized spheres. The specific surface area of such structures scales with the inverse of particle size, which is in the nanometer range (typically 5 nm), so that values around 600 m2 g 1 are typical for most aerogels, organic and inorganic. Extremely low thermal conductivity: In porous media, heat transfer has contributions from solid and gas conduction and from radiation. Gels with low solid volume fraction show little solid heat conduction. Additionally, according to Knudsen, heat transfer by gas molecules in voids is significantly reduced as compared to the unconfined gas if the pore size is less than (or about) the mean free path (70 nm for ambient air); this insulation effect is enhanced in vacuum. Radiation also depends on the pore size if the pore walls are opaque; then, for small pore size, many absorption–emission steps are needed for macro-scale heat transfer, just as in a super-insulator. At room temperatures, the thermal conductivity of silica aerogels is typically below 0.02 W m 1 K 1 (air: 0.026 W m 1 K 1); for RF aerogels, 0.012 W m 1 K 1 have been reported (Lu et al., 1992), and for evacuated silica aerogels, values as low as 0.01 W m 1 K 1 are possible (Scheuerpflug et al., 1985). At high temperatures, that is, in the infrared range, silica does not absorb well, but special opacifiers, such as carbon black or titanium dioxide, can be incorporated in the solid network during gelation (Lu et al., 1992). Optical transparency: If the solid phase itself is transparent, as for silica gels, light can be transmitted through the dry gel with little interaction since the structural units of gels are much smaller than the optical wavelengths. Only structural inhomogeneities at the nanometer scale lead to Rayleigh scattering and the typical blue appearance (against a dark background), and surface inhomogeneities at the micrometer scale blur the view through aerogel plates (Wang et al., 1992), making

j159

j 5 Understanding and Preventing Structural Changes During Drying of Gels

160

.

.

.

.

.

the gel a translucent, rather than transparent material. Typically, 84% of light is transmitted, which is only a small reduction from the 89% transmission for bulk silica glass (Sakka, 2005). Low refractive index: Compared to bulk solid materials such as silica glass, which has a refractive index of around 1.5, typical silica aerogels exhibit values very close to unity (Bellunato et al., 2007), meaning that the speed of light is only slightly reduced. The refractive index can be tuned in the range from 1.007 to 1.25 by adjusting the density of the aerogel (Wang et al., 1992; Kharzheev, 2008). Low dielectric constant: In aerogels, the dielectric constant is well below 2, and the value depends on the gel density so that it can be tuned by synthesis conditions (Geis et al., 2000). Other solid materials have a dielectric constant (usually much) greater than 2, for example, quartz glass has a value of 4. Low sound speed: The velocity of sound in silica aerogels is very low for a solid material; typical values are around 100 m s 1, as compared to 5000 m s 1 in quartz glass (Daughton et al., 2003). Mechanical properties: Even for very porous materials (99%), the stress–strain curve shows perfect elastic behavior, and the conchoidal fracture morphology indicates that the material is brittle, like a conventional glass; however, elastic and rupture moduli can be 10 000 times lower than those of silica glass (Woignier et al., 2005). Under compression, at increasing strain, the network exhibits yield, densification and plastic hardening (Woignier et al., 2005). Typical silica aerogels have a compressive strength of 0.15–0.3 N mm 2 with an elastic compression of using and Schubert, 2006). 2–4%; their tensile strength is about 0.02 N mm 2 (H€ Of course, these values depend on the connectivity in the solid network which is determined by the preparation conditions. Hydrophobicity: Silica aerogels are hydrophilic because of the silanol groups (:Si-OH); in a humid atmosphere, adsorption and capillary condensation occur and capillary forces can destroy the fragile structure. To achieve long-term stability, hydrophobation is necessary, for example, by trimethyl silylation of the wet gel (Shewale et al., 2008).

5.2.3 Applications of Dry Gels

It is clear that highly porous gels have a great number of interesting potential applications (Hrubesh, 1998; Sakka, 2005; H€ using and Schubert, 2006) because of their previously described extraordinary properties. Some specific applications are already reality; many others are still awaiting improvements in the economy and safety of production for their commercial breakthrough. In the following, the most important applications of dry gels are listed: .

Thermal insulation: Large commercial potential lies in applications coupling thermal insulation with transparency, especially in the building sector, where monolithic silica aerogels might be used as transparent windows (Jensen et al., 2004); layers of granular aerogels where diffuse lighting is acceptable or

5.2 Gels and Their Applications – Quality Aspects

.

.

.

.

.

desired (Schmidt and Schwertfeger, 1998); and aerogel coatings for insulation of solar heat collectors, where both transparency to sunlight and the confinement of generated heat are essential. Of course, there are also numerous thermal insulation applications where transparency plays no role, for example, in high or low temperature storage. In space technology, opacified aerogels serve as lightweight thermal insulation, for example, in the exploration of Mars where electronics have to be protected from the 100 K temperature difference between night and day (Jones, 2006). Sound insulation: Sound absorption in buildings or, more specialized, in anechoic chambers, opens another promising field of applications. Unlike for other materials, good absorption is also observed for the low frequency range. While suitable bulk aerogels are not yet available, respective composite materials containing aerogel particles can already be produced (Schmidt and Schwertfeger, 1998). Electrodes for example, for supercapacitors or water treatment: Monolithic carbon aerogels are very suitable electrode materials because of their low electric resistance and high specific surface area. One possible application is in supercapacitors where the voids of the gel are filled with an electrolyte. If voltage is applied, energy is stored by charge separation in the electrochemical double layer. High energy density and high power density make the device suitable for bridging short power failures. In other electrode applications, harmful contaminants are removed from industrial wastewater, or seawater is desalinated by (easily reversible) capacitive deionization during the passage by and through a stack of alternately polarized electrodes (Farmer et al., 1996, 1997). Cherenkov detectors in high energy physics: The velocity and, hence, energy of a charged atomic particle is measured by Cherenkov detectors. They contain a medium with a speed of light lower than the selected range of particle velocities. When traveling through such a medium, the particle produces an electromagnetic shockwave; from the cone angle, the particle velocity is obtained. Here, aerogels with tuned density are able to cover a velocity range for which the previously used compressed gases or liquids were not suitable; additionally, they are much safer to handle. These detectors are large (>1 m3), and significant amounts of aerogel have already been produced for this application (Fricke, 1986). Optical coatings: The losses of optical devices can be reduced by coating them with an aerogel of (matched) low refractive index (Hrubesh and Poco, 1995; Hrubesh, 1998); in this way, more sunlight can reach the active surface of a solar cell and, in fiber optics, light collection at the fiber entrance and signal propagation efficiency can be improved. Catalysis and chromatography: These are obvious applications for aerogels with their tunable microstructure exhibiting high internal surface area and large pore volume. Concerning catalysis (H€ using and Schubert, 2006), the sol–gel process allows good dispersion of the active component and aerogels can be prepared from most relevant oxides. However, some problems arise from the low permeability and the poor thermal conductivity; here, the deposition of aerogels on macroscopic structures offers a solution. In high-performance liquid chromatography, silica monoliths with bimodal pore size distributions

j161

j 5 Understanding and Preventing Structural Changes During Drying of Gels

162

.

are used to combine good permeation and high binding capacity (Lubda et al., 2005; Nunez et al., 2008). Space science: During the Stardust mission (1999–2006), an aerogel array was used to capture cometary and interstellar particles. The particles stayed largely intact when they were smoothly slowed down by the small filaments of the gel network and could be returned to earth for analysis (Jones, 2006).

For all these applications, the pore structure and high porosity of the aerogel are very important, even if the geometrical shape may vary from large monoliths to particulate material, and to films. After gel synthesis, it is therefore crucial to preserve the solid network during the drying step.

5.3 Structural Characterization of Gels – Quality Assessment

Before addressing the drying step itself, a brief overview will be given of the techniques by which the structure of gels can be characterized, and some nonstandard techniques will be discussed in greater detail. This will allow assessment of the quality of gels that have been dried by different routes. And, in some cases, it will allow direct quantification of the quality loss during drying. 5.3.1 Characterization of Wet Gels

The structural features of a gel are typically in the range of a few nanometers; and direct 3D visualization of wet gel structure is not possible: X-ray microtomography (m-CT) has a spatial resolution of about (or just below) one micrometer; electron microscopy (in transmission mode, TEM, or in scanning mode, SEM, for the surface) has a substantially higher resolution, but works only on dry gels. Therefore, indirect methods, such as scattering of X-rays, are used to characterize gel structure and also to investigate the sol–gel transition. 5.3.1.1 Small Angle X-Ray Scattering (SAXS) In the following, small angle X-ray scattering (SAXS) will be briefly introduced (Roe, 2000; Brinker and Scherer, 1990). As depicted in Fig. 5.3a, the X-ray beam hits the sample and the scattered signal is detected at a range of small angles. Structural units that are very large compared to the X-ray wave length (0.154 nm for Ka1 emission from Cu) do not significantly scatter the X-rays and, therefore, cannot be detected; with decreasing size of the structural features, the corresponding scattering angles increase and structural information can be extracted from the spectrum, which is commonly plotted as signal intensity over wave vector. The interpretation of such spectra is a research field of its own (e.g., Sinko et al., 2008). The most common characteristics extracted from SAXS spectra are particle size and fractal dimension, but other properties such as specific surface

j163

5.3 Structural Characterization of Gels – Quality Assessment 1/R

(a)

1/r

(b) fractal regime scattering angle ~ wave vector

intensity

X-ray

Guinier regime

Porod regime

2R 2r

-2

10

-1

10

0

10

1

10

wave vector (nm -1)

Fig. 5.3 Small-angle X-ray scattering (SAXS): (a) measurement principle and (b) typical spectrum for a dilute suspension of fractal aggregates built from dense primary particles.

area may also be accessed (Berthon et al., 2001). Here, only some elements of interpretation will be discussed for an illustrative example, the spectrum of a suspension of aggregates shown in Fig. 5.3b. For very small angles, in the Guinier regime, the signal intensity is highest and, from its decay with increasing wave vector, the aggregate size can be computed. For wave vectors corresponding to smaller sizes, the spectrum contains information about the inner structure of the aggregates. The absolute value of the negative slope in the log–log representation equals the fractal dimension, if it is less than 3; slopes between 3 and 4 indicate surface fractal structures. The Porod regime, with a slope of 4, indicates compact particles with a smooth surface. If these particles are monodisperse, regular oscillations in intensity are observed which are characteristic for particle size (Berthon et al., 2001). The transition between different regimes marks critical length scales of the system. In this example, the radius of primary particles r indicates the transition from dense to fractal geometry; whereas the radius of the fractal aggregates R is determined by the transition to the Guinier regime. It should be stressed that SAXS spectra of real systems may deviate significantly from this example. For instance, in fractal gels, there is no marked upper size limit except of the system size so that the Guinier regime can neither be clearly identified, nor can it be used for a straightforward analysis as above. If larger structural features are investigated, light scattering provides a suitable technique due to the greater wavelengths; by a combination of light scattering with SAXS, it could be proven that gels are fractal over several length scales (Schaefer et al., 1984). In a traditional way, SAXS is used to investigate the correlation between synthesis parameters and the resulting gel structure, for example, the dependence of the fractal dimension of silica gels on TEOS concentration in the starter solution (Vollet et al., 2008). Time-dependent SAXS has been applied to study sol–gel transition, in terms of particle growth and change in fractal properties in the forming structure (Tamon and Ishizaka, 1998). But the method also works on dry gels and, for

j 5 Understanding and Preventing Structural Changes During Drying of Gels

164

Fig. 5.4 SAXS spectra for RF wet gels and aerogels: both types have been prepared in acetone solution with acid catalyst (R/C ¼ 200); the mass fraction of RF in the starter solution was 5% for “AA200/5” and 20% for “AA200/20” (taken from Berthon et al. (2001)).

example, has helped to characterize long-term restructuring of silica–alumina aerogels (Sinko et al., 2008). Berthon et al. (2001) performed SAXS measurements on gels in the wet state and after supercritical drying. They proved that for polymer-like RF gels formed in acetone and with acid catalysis, the structure is not fully preserved by drying with supercritical CO2. The SAXS spectra for wet and dry gels, replicated in Fig. 5.4, show a significant change in slope from 2.6 (indicating fractal geometry) to 4 (indicating compact smooth particles) for large wave vectors. This suggests that fractal structures at small length scales have collapsed into dense particles ( 3 nm in size) during drying. For the high porosity (88%) sample “AA200/5”, the dry gel consists of fractal aggregates of particles; for the “AA200/20” aerogel, no fractal features are visible. The authors attribute the evolution of gel structure rather to solvent exchange (CO2 for acetone) than to drying itself. For base-catalyzed (R/C ¼ 200) gelation in water or acetone, for which colloidal gels are expected, SAXS spectra suggest non-fractal structures that are preserved during drying. 5.3.1.2 Thermoporometry Another suitable method to characterize the pore structure of wet gels is thermoporometry (Brun et al., 1977). It uses the fact that the melting temperature of ice is reduced with decreasing crystal size due to the increasing curvature of the ice/water interface (Gibbs–Thomson effect). In the characterization experiment, the sample is either gradually frozen or slowly heated from the frozen state in a differential scanning calorimeter (DSC). The enthalpy flow released or required for the phase change at a given temperature is recorded. It is proportional to the pore volume that

5.3 Structural Characterization of Gels – Quality Assessment

Fig. 5.5 Ice penetrating into a pore during freezing (ice considered as non-wetting).

freezes/melts at the corresponding undercooling DT. This undercooling can be converted into the pore radius for water as pore liquid by (Brun et al., 1977) rp ½nmŠ ¼ 0:57 þ

64:7 ; for 0 < DT < 40 K DT½KŠ

ð5:3Þ

Hence, the pore volume distribution of the gel can be computed from the DSC thermogram. This method is only suitable for pore radii of 1–30 nm because liquid in smaller pores does not freeze and, for larger pores, the undercooling is too small for reliable measurement. Scherer (1993) interprets the freezing of a gel as the penetration of ice into progressively smaller pores as the undercooling DT increases (see Fig. 5.5) and, therefore, sees thermoporometry as an analogue of mercury porosimetry for saturated porous media. If the material is not rigid, but compliant, liquid may flow out of the pores instead of ice penetrating them, resulting in gel shrinkage (see Section 5.4.2) and a shift of the pore size distribution to smaller values. Figure 5.6 shows the results of recent literature work, in which thermoporometry was applied to silica gels synthesized with ultrasound-assisted hydrolysis (Vollet et al., 2008). For higher dilution of the TEOS starter solution, the gel

Fig. 5.6 DSC thermograms for silica gels with porosities from 78% (R ¼ 4) to 88% (R ¼ 16); R indicates the molar ratio of hydrolysis water to TEOS (taken from Vollet et al. (2008)).

j165

j 5 Understanding and Preventing Structural Changes During Drying of Gels

166

Fig. 5.7 Pore size distribution of wet RF gel (R/C ¼ 200, R/W ¼ 0.5 g cm 3) by thermoporometry, and of corresponding dry gels by nitrogen adsorption (taken from Yamamoto et al. (2005)).

porosity was increased and the mean pore radius shifted, approximately from 10 to 25 nm. Besides the peak resulting from nanoscale pores (0  C), which the authors attribute to macropores. Following Scherer’s argumentation, this signal might also result from water that has flown to the gel surface during freezing. Thermoporometry has also been used to assess structural changes during the drying step: Yamamoto et al. (2005) have prepared RF gels, exchanged the pore liquid for tert-butanol and subsequently dried the gels by convective or freeze-drying. The pore volume distribution of the wet gels was measured by thermoporometry, that of the dry gels by nitrogen adsorption. The results, depicted in Fig. 5.7, clearly indicate that nanoscale pores shrink during convective drying whereas the gel structure can be preserved by freeze-drying. 5.3.2 Characterization of Dry Gels 5.3.2.1 Nitrogen Adsorption On dry gels, standard characterization techniques for porous media are used, several of which have been described in Volume 2 of this series: helium pycnometry for pore volume determination (Section 6.3.1.2) as well as nitrogen adsorption at 77 K for surface area (Section 6.3.2.2, BET method), for microporosity (Section 6.3.3.2, Dubinin–Radushkevich method), for pore size distribution (Section 6.3.3.3, BJH method), and for total pore volume (Section 6.3.3.4). When characterizing gels by nitrogen adsorption, other methods are also used for data interpretation, for example, the t-plot method for microporosity (Lippens and de Boer, 1965) and the Dollimore–Heal method (Dollimore and Heal, 1964) or Broekhoff–de Boer theory for mesoporosity (Lecloux, 1981).

5.3 Structural Characterization of Gels – Quality Assessment

An additional technique to measure pore size distributions, which will be briefly introduced later, is mercury porosimetry. It uses the successive intrusion of mercury (non-wetting) and – due to the high surface tension of mercury and small pore size –involves high compressive pressures. Therefore, special attention has to be given to deformation effects that occur when dry gels with low mechanical stiffness and strengths are characterized. However, shrinkage and/or cracks may not only happen during mercury porosimetry, but also during nitrogen adsorption (Scherer et al., 1995a). In nitrogen adsorption, as the relative pressure rises, pores of increasing size are gradually filled by condensing nitrogen; consequently, the interface tension between nitrogen and vapor (0.00885 N m 1) exerts capillary forces on the aerogel and may lead to significant volumetric shrinkage. As the relative pressure approaches unity, the liquid/vapor interface becomes flat and the gel may expand to its full original volume, if no plastic deformation has occurred. Subsequent desorption is analogous to slow atmospheric drying; it is accompanied by shrinkage and a sudden expansion at low relative pressure since a small liquid volume in the capillary bridges can hold the sample under compression (Scherer et al., 1995a). In Fig. 5.8, part of the adsorption–desorption cycle for a compliant gel is given: instead of showing a horizontal asymptote at saturation, the sorption curves reflect expansion and shrinkage of the (saturated) sample. For gels with different mechanical behavior, Reichenauer and Scherer (2000) have measured the sample dimensions during the sorption experiment. The compliant gel characterized in Fig. 5.8 underwent irreversible shrinkage; another, sintered gel was stiff enough to recover to its original size, both at full saturation and after complete desorption. For both gel types, pore volume information is obscured by shrinkage, resulting in underestimation of pore

Fig. 5.8 Nitrogen adsorption-desorption cycle for a silica aerogel with 93% porosity (taken from Reichenauer and Scherer (2001a)).

j167

j 5 Understanding and Preventing Structural Changes During Drying of Gels

168

size; an advanced characterization method has been proposed by the same authors, which uses dilatation and sorption data to obtain the bulk modulus and pore size distribution of the gel (Reichenauer and Scherer, 2001b). Besides contraction of the sample, insufficient equilibration times may lead to wrong interpretation of nitrogen adsorption data (Reichenauer and Scherer, 2001a): because of incomplete filling of pores, the total pore volume is underestimated. In Fig. 5.8, two identically prepared aerogel samples have been characterized with different total duration of the cycle to prove sufficient equilibration.

5.3.2.2 Mercury Porosimetry Whereas from nitrogen sorption data a size distribution can only be extracted for mesopores (with pore diameter 2 nm < dp < 50 nm), standard mercury porosimetry is used to obtain complete pore size distributions in the pore diameter range from 7.5 nm to 150 mm. During the characterization experiment, the sample is first surrounded and then progressively intruded by mercury, as the pressure is increased. Experimental results are commonly plotted as invaded pore volume versus applied pressure (see Fig. 5.9a). The Washburn equation describes at which (capillary) pressure a cylindrical pore of diameter dp is invaded P¼

4s cos q dp

ð5:4Þ

where s ¼ 0.48 N m 1 is the surface tension and q ¼ 140 the contact angle of mercury, that is, 196 MPa corresponds to dp ¼ 7.5 nm. If the porous structure withstands these high pressures, as for some RF cryogels (Kocklenberg et al., 1998), Eq. 5.4 allows conversion of pressures into pore sizes and easy computation of the pore size distribution (see Fig. 5.9b). More typically for gels, however, the structure collapses during characterization due to its weak compressive strength; or, depending on the conditions of the gel synthesis, drying and optional thermal treatment, some intrusion

Fig. 5.9 (a) Mercury porosimetry data for dried copper hydroxide precipitate, pure intrusion case, (b) pore volume distribution by Eq. 5.4 (taken from Job et al. (2006a)).

5.3 Structural Characterization of Gels – Quality Assessment

Fig. 5.10 (a) Mercury porosimetry data for monolithic silica aerogel, irreversible shrinkage case, (b) pore volume distribution by Eq. 5.5 (taken from Job et al. (2006a)).

may be observed after a first part of densification. Job et al. (2006a) discuss how these three different behaviors – pure intrusion, irreversible shrinkage, shrinkage followed by intrusion – can be distinguished and how mercury porosimetry data can be interpreted in each case (see also Pirard et al. (2005)). Weighing of the sample before and after the porosimetry experiment can quantify the mercury which has intruded pores and is subsequently entrapped within the sample. In the case of pure intrusion, the full detected mercury volume remains within the sample, as revealed by the depressurization branch in Fig. 5.9a. If no mercury is found in the porous solid after porosimetry, the volume change of the sample may be measured (by mercury pycnometry as explained below) to confirm the assumed extent of irreversible shrinkage. An example of irreversible shrinkage during mercury porosimetry is given in Fig. 5.10a. The shape of the curve reflects a progressive increase in mercury volume penetrating into the measurement cell, in contrast to the characteristic sudden pore volume jumps which occur during mercury intrusion at pressures corresponding to prevalent pore sizes. Such densification during mercury porosimetry has been reported for various dry gels, among them silica aerogels (Ali
e et al., 2000; Scherer et al., 1995b), RF xerogels (Job et al., 2004, 2005; L
eonard et al., 2005a, 2008), aerogels and cryogels (Job et al., 2005), and carbon xerogels (Job et al., 2004). Evidently, the Washburn equation must not be used to determine pore size distribution in this case. Pirard et al. (1995) showed that the densification is due to a hierarchical collapse of pores: the largest pores collapse first at the lowest pressure values; then, as the mercury pressure increases, pores of decreasing size are successively and completely eliminated by crushing. Based on that mechanism, he proposed an appropriate relation to analyze mercury porosimetry data, which links pore size, down to which crushing occurs, to applied pressure dp ¼ C
P

0:25

ð5:5Þ

j169

j 5 Understanding and Preventing Structural Changes During Drying of Gels

170

Here, C is a constant for material stiffness that needs to be determined experimentally. Figure 5.10b shows the pore size distribution obtained using this equation. It should be noted that densification curves may also be used to determine the bulk compression modulus of silica aerogels (Scherer et al., 1995b). Figure 5.11a shows mercury porosimetry data for a silica xerogel with mixed behavior: the material first shrinks (full circles) up to a critical pressure Pcr beyond which its small, uncollapsed pores are intruded (open circles). The critical pressure is identified by a sudden change of slope of the volume variation curve. If the sample is depressurized before reaching Pcr (crosses), then indeed no mercury uptake is detected in the sample. The pore volume distribution in Fig. 5.11b can be computed using either Eqs. 5.4 or 5.5 depending on the considered pressure domain. At P ¼ Pcr, both equations are valid so that the mechanical constant C – necessary for the hierarchical collapse equation – can be obtained conveniently (Pirard et al., 1998): C ¼ 4s cos q
Pcr0:75

ð5:6Þ

Such mixed behavior during mercury porosimetry has been observed for a range of dry gels, for example: silica xerogels and aerogels (Ali
e et al., 2001), RF xerogels (L
eonard et al., 2008), and carbon xerogels, aerogels and cryogels (Job et al., 2005). For gels, the mechanisms involved during mercury porosimetry tests depend strongly on the microstructure, which is related to the synthesis (L
eonard et al., 2008) and drying conditions (Job et al., 2005); therefore, one must carefully examine the measurement results to be sure of the mechanisms involved. If data corresponding to densification are analyzed using software based on the Washburn equation (usually provided with porosimeters) this yields an unphysical pore size distribution.

Fig. 5.11 (a) Mercury porosimetry data for silica xerogel, shrinkage followed by intrusion, (b) pore volume distribution by combining Eqs. 5.4 to 5.6 (taken from Job et al. (2006a)).

5.3 Structural Characterization of Gels – Quality Assessment

5.3.2.3 Other Methods Another useful technique is mercury pycnometry which can be used to determine the geometrical volume of the dry gel (pores and solid), hence giving complementary data to helium pycnometry which measures only the pore volume. In this characterization method, a volume-calibrated chamber containing the sample is filled with mercury; the mercury volume (or weight) is measured to compute the sample volume. Unlike mercury porosimetry, no pressure is exerted so that mercury neither enters the pores nor crushes the sample. Besides the determination of the porosity, pore size distribution and surface area, the (surface or mass) fractal dimension of dry gels may be of interest. To this purpose, small-angle X-ray scattering can be used; nitrogen sorption and mercury porosimetry also offer possibilities to extract this structural information (see, for example, Blacher et al. (2000)). All previously described characterization methods yield only statistical or averaged information on gel structure. The natural wish to actually see the structure can be (partially) fulfilled by scanning or transmission electron microscopy (SEM or TEM). In SEM (Goldstein et al., 2003), operating in high vacuum, a focused electron beam scans the sample surface and the local response (emitted secondary electrons) is recorded; therefore, spot size sets the resolution limit above 1 nm. To avoid charging of the sample during measurement, it needs to be made electrically conducting by sputter-coating with a thin gold layer, further decreasing the resolution. An alternative is offered by environmental SEM in which the sample is near or at atmospheric pressure allowing discharge through the gas phase. (In principle, ESEM also allows the investigation of wet samples, but to our knowledge has not yet been applied to wet gels.) In SEM, the surface is imaged in a three-dimensional appearance; however, the contrast in the image can be caused by a plenitude of effects so that careful interpretation is recommended. A clear advantage of SEM is that bulk samples can be characterized. In contrast, TEM (Reimer and Kohl, 2008) is an optical method, in which electrons are interpreted as an electromagnetic wave of very short wavelength; as a result, higher resolutions (well below 1 nm) are possible and sputter-coating is not necessary. The sample is placed in high vacuum and imaged in transmission; consequently, it has to be very thin (typically 10–100 nm) to detect sufficient signal – and to actually see individual structural features. The obtained information is a twodimensional projection of the sample, hence containing information from the whole sample depth. Electron microscopy is mostly used qualitatively (some examples will be given in Section 5.5): for SEM, because of limited resolution and restriction to sample surface; and for TEM, because of the difficulty in separating information in the third dimension. Quantitative analysis of TEM data obtained from dry gels allows determination of the size distributions of primary particles, for example, in silica–alumina and silica aerogels, for which Sinko et al. (2008) have measured particle diameters in the range 3–14 nm.

j171

j 5 Understanding and Preventing Structural Changes During Drying of Gels

172

Fig. 5.12 X-ray transmission images (a,b) and reconstructed cross-sections (c,d) for RF gels showing shrinkage (a,c) and cracks (b,d).

5.3.3 Characterization of Gels During Drying

X-ray tomography is a non-destructive imaging technique giving access to the internal structure of the investigated object. The technique is based on the local variation of the X-ray attenuation coefficient of matter in the X-ray path. Twodimensional cross-sectional images are reconstructed from transmission data collected by irradiating the object with an X-ray beam in many different directions (see Fig. 5.12), so that full volume data of the sample are accessible. (For a more detailed description of the measurement principle and subsequent image analysis, refer to Chapter 4 in Volume 2 of this series.) Nowadays, laboratory microtomographs allow the production of images with voxel size below 1 mm. Of course, this is not enough to see individual particles or pores in gels, but other important geometric and structural features are accessible: the dimensions, shape and volume of the gel sample, internal defects such as cracks, and local averages of gel density, porosity and moisture content. Therefore, the method offers a way of investigating shrinkage and the occurrence of cracks. Since X-ray tomography keeps the sample intact, it is suitable not only for the characterization of the dried product, but also to follow the evolution of gel characteristics during the drying process. Up to now, this requires an interruption of the process to scan the sample being dried, but, in principle, the tomograph can be built around laboratory drying equipment. For such dynamic measurement, the time for a full scan of the sample is a crucial parameter; it depends on the sample size and the desired resolution and may vary from a few minutes, if 5003 voxels are sufficient, to several hours, for 20003 voxels. As most gels shrink during drying, it is essential to determine their surface area evolution to calculate the drying flux from mass loss measurements. With appropriate image analysis, very accurate shrinkage curves (as in Fig. 5.34) can be obtained from X-ray transmission images and reconstructed cross-sections (Fig. 5.12). The

5.3 Structural Characterization of Gels – Quality Assessment

same set of images can also be used to detect and quantify cracks appearing during drying. If crack-free materials are the quality goal, this experimental method can provide essential information for modeling and optimization of the drying process. Indeed, it is well-known (and will be explained in the following section) that too severe drying conditions may lead to sample cracking; and for a theoretical description, simulation of internal stresses must be combined with appropriate cracking criteria (see Section 5.6.1). Finally, the images obtained from microtomography at different stages of drying can be processed to extract internal moisture distributions, which are very difficult to access by any other experimental technique. To this aim, calibration is necessary since only the wet gel density is available from tomography. Different possible drying behaviors are illustrated in Fig. 5.13 for RF gels, for which the attenuation of X-rays may be considered proportional to wet gel density (Escalona et al., 2008). If the gel does not shrink, its density decreases as water is replaced by air; for gels exhibiting ideal shrinkage, that is, remaining fully saturated, the density increases as water “is replaced” by the dried RF resins whose intrinsic density is higher. Real behavior includes an initial period of ideal shrinkage followed by the emptying of pores with no further shrinkage. Therefore, the relationship between measured density and moisture content will generally not be unique for the whole drying process. Since all mentioned measurements are based on the same set of reconstructed images, they can be performed simultaneously. In a more general context, X-ray microtomography can also help to determine the internal structure of the dried product. Unlike SEM, which requires sample preparation and yields only 2D information for a small part of the surface so that many repetitions are necessary for statistically relevant results, it is a non-destructive technique that gives full 3D information. For these reasons, X-ray microtomography can complement classical measurement techniques, if the dry material exhibits large pores which cannot be characterized by nitrogen adsorption (from 2 to 50 nm) or mercury porosimetry (from 7.5 nm to 150 mm).

Fig. 5.13 Evolution of total gel density (as relevant for X-ray tomography) during drying for different material behavior (adapted from Escalona et al. (2008)).

j173

j 5 Understanding and Preventing Structural Changes During Drying of Gels

174

Fig. 5.14 Phase diagram to illustrate different drying methods for gels.

5.4 Drying Methods for Gels – Quality Loss

In this section, the major drying techniques for gels are presented. We stress again, that the term xerogel, despite its literal meaning (“dry gel”), will only be used for gels obtained by convective (or, more generally, evaporative) drying. Xerogels usually exhibit significant shrinkage, and it is difficult to avoid cracks, the reason lying in capillary effects, as will be discussed below. Therefore, other techniques such as freeze-drying and supercritical drying have been applied, where no liquid/gas phase boundary exists. As we will see, producing freeze-dried cryogels also bears a big risk for gel structure – if not in the drying step, then during freezing. The first successful technique to produce aerogels (in the sense that the full volume of pore liquid has been replaced by gas) was supercritical drying, this being the reason why here the term is only used for dry gels obtained by this route. Figure 5.14 summarizes the different drying techniques in the phase diagram for the pore-filling substance. 5.4.1 Convective Drying 5.4.1.1 Introduction If a wet gel with low solid fraction and small pores is convectively dried without taking specific precautions, it will shrink significantly and irreversibly; a typical example is given in Fig. 5.15. Besides shrinkage, cracks may occur during drying, as shown for a rod of alumina gel in Fig. 5.16. The drying rate plays an important role in crack formation, so that different crack patterns are obtained or cracks can even be prevented by slow drying.

5.4 Drying Methods for Gels – Quality Loss

Fig. 5.15 Convective drying of a hydrogel (based on polyvinylpyrrolidone) (taken from Pakowski et al. (2006)).

In the following, we will sketch the theoretical framework for convective drying as developed by Scherer (1986, 1987a, b, c, 1988, 1989, 1990a, b, 1992a) and Brinker and Scherer (1990), which serves as a good basis to understand structural damage and to explore possibilities for preventing it. 5.4.1.2 Shrinkage As liquid is removed from the gel surface by evaporation, the liquid/gas interface becomes curved. The resulting capillary pressure Pc is defined as the pressure difference between the gas and the liquid and is computed from the interface tension s and the radius of curvature rc (see Fig. 5.17a) as Pc ¼ Pg Pw ¼

2 s 2 s cos q ¼ rc rp

ð5:7Þ

The conversion into pore radius rp and contact angle q is only valid for a fully developed meniscus in cylindrical pores under the condition that the adsorbed liquid film can be neglected. This pressure can attain very high values, for example, approximately 30 MPa for water in perfectly wetted pores of radius 5 nm. Before analyzing the response of the gel to this pressure, the stress in the wet gel is defined; it

Fig. 5.16 Crack patterns on alumina gel for (a) slow and (b) fast convective drying (taken from Pourcel et al. (2007b)).

j175

j 5 Understanding and Preventing Structural Changes During Drying of Gels

176

Fig. 5.17 Mechanisms during convective drying: (a) shrinkage of solid matrix as capillary pressure increases and (b) differential shrinkage when liquid flow is limited, leading to solid tension build-up and cracks when surface pores empty (only half of gel is shown).

has contributions from solid and liquid, as the respective forces Fs and Fw per total cross-sectional area A, sx ¼

Fxs þ Fxw ~x yPw ¼s A

ð5:8Þ

~x is the solid network stress and y the porosity, and a negative liquid Here, s pressure is considered as compressive; equivalent equations hold for the y- and z_ v , liquid can easily directions. For a small cube of gel or very low evaporation rates m flow to keep pressure gradients negligible, and uniform Pw can be assumed. Then, ~x ¼ yPw the liquid pressure is entirely balanced by the solid network stress s (negative stress indicates compression), and the wet gel shrinks with no total stress, sx ¼ 0. If the gel has a purely elastic solid network, then volumetric strain, that is, relative volume change, is given by e¼

yPw K0

ð5:9Þ

where K0 is the bulk modulus of the gel. Gels of initial bulk density r0 can shrink reversibly until the density reaches a critical value rcr, roughly determined by P :¼

rcr As m s cos q 1 K0

ð5:10Þ

where As is the specific surface area (given in m2 g 1 and roughly proportional to the inverse of the primary particle radius) and 2.5 < m < 4 a material constant, both dependent on synthesis conditions (Smith et al., 1995a). At the critical point (typically rcr ¼ 1.4r0), the gel enters the plastic range: gel deformation is irreversible

5.4 Drying Methods for Gels – Quality Loss

and the shrinking gel becomes stiffer; the increase in bulk modulus can be approximated by Kp ¼ K0 ðr=rcr Þm

ð5:11Þ

Since shrinkage reduces the pore radius, the resulting compressive pressure also increases. With such a model, the density and pore size of the dry gel may be predicted (Smith et al., 1995a): for P < 1, the density increase during drying is less than 50% and reversible; otherwise, the dry gel density increases p with ffiffiffiffiffi increasing P (irreversible shrinkage), for compliant gels roughly as 1:4r0 P. Shrinkage is hence more pronounced for higher surface tension, smaller contact angle, higher specific surface area (assumed as constant) and lower bulk modulus of the gel. In the above analysis, kinetics plays no role. It is therefore not surprising that shrinkage shows negligible dependence on drying rate (Smith et al., 1995a). 5.4.1.3 Differential Shrinkage and Stress Cracking, however, is strongly influenced by drying conditions: the fragments obtained when drying a wet gel cylinder become smaller as the oven temperature, and accordingly drying rate, is increased (Smith et al., 1995a). To understand this, we must consider that liquid flow in the drying gel is limited by viscous effects. As illustrated in Fig. 5.17b, evaporation occurs at the gel surface so that shrinkage and decrease in liquid pressure start there. As a consequence, liquid is pumped from the core of the gel according to Darcy’s law _w¼ m

K rPw mw =rw

ð5:12Þ

where K is the permeability of the gel, mw the dynamic viscosity and rw the density of the pore liquid. Since gels have very low permeabilities – essentially, K is proportional to the square of the pore radius – even low drying rates can lead to high pressure gradients: for a typical permeability of 10 18 m2 and a drying rate of 1 kg m 2 h 1, one obtains 3 MPa cm 1. The liquid pressure profile is linked to the volumetric strain rate by e_ ¼

K 2 r Pw mw

ð5:13Þ

signifying that the solid network contracts locally when pore liquid flows out of that region. In order to compute stress and strain in the drying gel, first, a constitutive equation for the wet gel must be derived. For an elastic solid network, the linear strain depends on the stress and liquid pressure as ex ¼

 Pw 1 sx nðsy þ sz Þ þ E 3Kp

ð5:14Þ

j177

j 5 Understanding and Preventing Structural Changes During Drying of Gels

178

where E is the Young’s modulus and n the Poisson’s ratio of the solid network. If this is applied to a plate (of thickness 2 L) that is dried from both sides, the local volumetric strain can be computed as e ¼ ðð1 C0 ÞhPw i þ C0 Pw Þ=Kp

ð5:15Þ

where hPwi is the average liquid pressure in the plate and C0 is a function of Poisson’s ratio and approximately 1/3 for highly compressive gels. There is a fundamental difference to Eq. 5.9, which describes free strain and is valid when liquid pressure and shrinkage are uniform in the gel. If pressure gradients develop, local shrinkage depends on both local and average liquid pressure, because regions of high and low pressure are connected and cannot contract freely at their natural rate. Combining Eqs. 5.13 and 5.15 and using a boundary condition of constant _ v , a time-dependent diffusion _ w jz¼L ¼ m evaporation flux at the plate surface m equation for the liquid pressure can be obtained and solved. For the stress in the plane of the plate, one finds sx ¼ sy  hPw i Pw

ð5:16Þ

This expresses the fact that stresses do not depend on the pressure level and only develop if there is a pressure gradient; further, it shows that during drying not only compression, but also tension will occur, the latter being a greater danger for the fragile gel structure. (Scherer nicely illustrates this behavior by a comparison with thermal stresses developing in a plate that is cooled from the surface.) Typical qualitative pressure and stress profiles in a drying elastic gel plate are plotted in Fig. 5.18. One sees that the pressure profile becomes parabolic after a time that is characteristic for liquid flow; then, the profile shape stays unchanged, only the pressure level decreases. When the menisci at the gel surface are fully developed, the minimum pressure value is reached, given by Eq. 5.7, and the surface pores dry out, signifying the end of the first drying period. For the stress this means that a stationary profile is approached with a maximum tensile stress at the plate surface that can be given as sx jz¼L 

_ vL mw m 3rw K

ð5:17Þ

Fig. 5.18 Drying of an elastic gel plate (thickness 2L): qualitative profiles of (a) liquid pressure and (b) in-plane stress as evolving in time.

5.4 Drying Methods for Gels – Quality Loss

(For extremely high drying rates, when the liquid pressure only drops at the surface, the stress assumes the highest possible value, namely capillary pressure in Eq. 5.7.) In the above model, shrinkage is not explicitly accounted for, but – taking a Lagrangian approach – the obtained profiles are still meaningful for a time-dependent plate of thickness L, if shrinkage is (nearly) uniform. Furthermore, constant material properties have been assumed which is certainly not allowed when the gel shrinks significantly. Despite these simplifications, the analysis is very helpful to identify the crucial parameters for preventing structural damage during drying: harmful stress increases with higher drying rate, larger sample thickness and lower permeability. A similar derivation can be done for gels that have a viscous solid network. Then, the constitutive equation for the wet gel is given by the relation between strain rate and stress e_ x ¼

 1 Pw sx Nðsy þ sz Þ þ F 3KG

ð5:18Þ

where F is uniaxial solid viscosity, KG bulk viscosity and N the corresponding Poisson’s ratio. The local volumetric strain rate is a function of average and local liquid pressure e_ ¼ ðð1 C0 ÞhPw i þ C0 Pw Þ=KG

ð5:19Þ

and can be linked to Eq. 5.13 to get a pressure profile that is constant in time. As above, the gel is assumed highly compressive, and local stress is computed from Eq. 5.16. Figure 5.19 shows how profiles of pressure and stress in the saturated gel depend on the two parameters rffiffiffiffiffiffiffiffiffiffiffiffiffi _v mw 3 KG m ð5:20Þ C1 ¼ L and C2 ¼ 3 KKG rw L

Fig. 5.19 Drying of a viscous gel plate (thickness 2L): qualitative static profiles of (a) liquid pressure and (b) in-plane stress.

j179

j 5 Understanding and Preventing Structural Changes During Drying of Gels

180

(C1 is a measure of the resistance that the pore liquid offers to contraction of the solid phase.) The maximum tensile stress, occurring at the plate surface, is given by   C1 cosh C1 1 ð5:21Þ sx jz¼L  C2 sinh C1 Again, the analysis is simplified by taking material properties constant and neglecting shrinkage. Furthermore, and unlike Scherer, we have neglected syneresis, that is, gel shrinkage due to restructuring of the solid network by continuing condensation reactions. Nevertheless, we find that, as for the elastic plate, higher drying rate, larger sample thickness and lower permeability increase the stress. In fact, Eq. 5.21 is approximated by Eq. 5.17 for small C1; for high C1, stress increases with increasing gel viscosity KG. It must be stressed that, in reality, there is a time dependence, since the gel stiffens and becomes less permeable as it shrinks; therefore, during drying, liquid pressure gradients increase and stress reaches a maximum when the menisci are fully developed and the surface pores start to dry out – just as in the elastic case. So far, we have only presented two limiting cases, the purely elastic and the purely viscous gel. Generally, one may say that the viscous model describes the compliant gel during the early stage of drying and that the elastic model is suited for later stages when the gel has become more rigid and the menisci are about to enter the gel. Of course, the mechanical behavior of real gels is more complex, and, for quantitative results, numerical solution with an explicit description of shrinkage is needed, for example, Scherer (1987b), and a combined model seems more appropriate, for example, for the start of drying (Scherer, 1988). Besides a good choice of the mechanical model, material parameters are a difficult issue. One method that proved to be elegant and accurate is a three-point bending test in which viscoelastic properties and permeability are estimated simultaneously (Scherer, 1992b). The present overview was restricted to models that are only valid during the first drying period. However, this is sufficient for our discussion on structural damage since the highest stress level is attained at the critical point, when the solid network can no longer (or not fast enough) contract under capillary pressure and the liquid/ gas interface starts to recede from the gel surface. Beyond this point, the liquid pressure at the liquid/gas interface remains constant and is given by Eq. 5.7 whereas both drying rate and distance for liquid flow decrease as the evaporation front recedes. Consequently, in the wet gel region, pressure differences can be leveled out and stress will disappear during the second drying period. (Due to the small pore size and the narrow size distribution, both not favorable for sustained capillary flow, all of the surface is expected to dry out rapidly so that the constant rate period is not extended significantly beyond the critical point.) 5.4.1.4 Cracking The tensile stresses that have been derived for the first drying period can, in principle, on their own lead to failure of the gel. But mostly, cracks arise from the propagation of small mechanical defects under macroscopic tension. The material on either side of the crack can contract more freely so that the stress is partially relieved (see

5.4 Drying Methods for Gels – Quality Loss

Fig. 5.20 Different explanations for crack initiation: (a) pre-existing surface defect or (b) micro cracks caused by local capillary forces.

Fig. 5.20a). The stress concentrated at the tip of the crack is proportional to the macroscopic tensile stress sx and the square-root of crack length c. If it exceeds the critical stress intensity Icr pffiffiffiffiffi 1:1sx pc > Icr ð5:22Þ

the crack will propagate catastrophically since crack-tip stress continuously increases. For constant macroscopic tension, a critical crack length can be defined. This theoretical description has been nicely confirmed on gels by bending experiments on notched beams and on beams with defined surface roughness (Alaoui et al., 2000). However, as we have seen above, the stress level rises gradually during the first drying period and experiences no dramatic change at the transition to the second drying period. Therefore, propagation of surface defects does not seem an appropriate explanation of why cracks usually occur at that very moment (Simpkins et al., 1989). According to Scherer, another mechanism is responsible: as liquid menisci enter the gel, capillary forces that held the pore walls together during the first drying period are switched off locally; by this, micro-cracks are initiated at the pore scale (see Fig. 5.17b). As the drying front advances, these micro-cracks can link together. The drying front has an irregular shape and its width decreases with increasing drying rate (Shaw, 1987); these effects result from the size distribution of pores and the viscous capillary flow that tries to keep small pores saturated. Pore network models, as presented in Chapter 2 of Volume 1 of this series, can help to study these phenomena (Vorhauer et al., 2010). The interlinked micro-cracks are under macroscopic tension over the width of the drying front, that is, the partially saturated region (see Fig. 5.20b). If this part of the crack exceeds the critical length, it can propagate catastrophically. With increased drying rate, the macroscopic stress increases linearly, and this cannot be compensated by the reduction of drying front width so that faster drying results in material failure whereas slow drying can prevent it. If the macroscopic stress is small, the micro-cracks will not propagate into macro-cracks, but rather heal when drying is complete. Gel size has no influence on the drying front width, but larger gels develop higher stresses so that the proposed hybrid model of macroscopic stress and local failure can also explain why larger gels are more likely to crack.

j181

j 5 Understanding and Preventing Structural Changes During Drying of Gels

182

Scherer further argues that local stresses alone cannot be the reason for macroscopic cracks, because gels that are too large or dried at too high rates generally break into only a few pieces. If the reason were the local pressure differences between pores (or pore regions) of different saturation status, then the gel sample should rather be pulverized – irrespective of drying rate and sample size. So, besides macroscopic tension, the irregular interface of the fractal drying front is thought to play a crucial role in crack formation. There is experimental evidence for the existence of liquid and dry patches at the phase front, which are significantly larger than the pore size: during drying of porous glass, the partially saturated region of the drying front becomes opaque (Scherer, 1992a) because these patches scatter the light (unlike the much smaller pores themselves). It is still not clear whether the irregular shape of the phase front harms by generating defects or beneficially diffuses the stress. Since such an irregular front cannot be modeled by a continuum approach, only pore-scale models, as discussed in Section 5.6.2, are expected to bring more light to this interesting problem. In the above expressions for gel stress, the liquid pressure level does not appear explicitly. Nevertheless, capillary pressure plays a crucial role because it decides to what extent the material will shrink, and thereby determines how much the gel stiffens and by how much its permeability is reduced. Consequently, lowering the capillary pressure – by increasing pore size, reducing surface tension or increasing contact angle – is beneficial for product quality because it reduces the maximum tensile stress. So far, the qualitative recommendations to avoid cracks – besides reducing sample size and drying at a low rate – may be summarized as: . . .

Increase the gel strength (or the critical stress intensity Icr) Increase the pore size for higher permeability and lower capillary pressure Use low-surface tension pore liquid and increase the contact angle for reduced capillary pressure.

We conclude with a remark on phenomena at the end of the convective drying process. As the outer regions of the gel become completely dry, the compressive action of capillary pressure is eliminated and for (partially) elastic material behavior, springback, that is, re-expansion of the gel network may be observed. The interior regions are still under compression so that stress is reversed. This effect is nicely proven by experiment: a plate that is dried from one side is curved towards the open surface during the first drying period; during the second drying period the curvature is away from the open surface (Scherer, 1987c). Such stress reversal may also lead to cracks at the very end of drying when the wet region becomes small (Simpkins et al., 1989). 5.4.2 Freeze-drying

After seeing the problems in convective drying because of capillary forces, one is inclined to test a drying technique that does not involve a liquid/vapor phase boundary. The most evident alternative is freeze-drying, which is widely used for

5.4 Drying Methods for Gels – Quality Loss

thermosensitive products, mainly foods and pharmaceuticals, and when the solid structure is to be preserved, for example, in biological samples. Therefore, it seems quite surprising that this alternative should also face severe problems – producing powder or very coarse pore structure (Scherer, 1992a). One straightforward explanation again involves capillarity, when residual pore liquid exists because of incomplete freezing or a too early temperature rise at the end of the drying process. This has indeed some significance if we recall that the melting temperature is much lower in small pores (see thermoporometry in Section 5.3.1.2). However, Scherer (1993) points out that it is the freezing step itself that can destroy the fragile gel structure. One mechanism which may lead to damage by freezing is the density change during solidification of the pore liquid. For water, the specific volume increases by roughly 9%. Depending on the freezing conditions and pore size distribution, different scenarios can be imagined. For a significant temperature gradient, for example, if the wet gel is immersed into a cold liquid, an ice crust may form at or near the surface of the gel. When the inner regions freeze, and expand, this crust will fracture easily, rather than inhibiting crystallization. If the temperature is lowered uniformly, then, for distributed pore size, liquid starts to freeze in the larger pores, for which the required undercooling is less (see thermoporometry in Section 5.3.1.2), and a percolating ice network gradually builds up, which is broken when the liquid in the smaller pores freezes. Also, if we assume that the liquid solidifies simultaneously in all the pores, because of a uniform pore size, a pressure builds up that must be released. For a volume increase during solidification, the damaging effect seems very obvious, but also, if the solidified solvent has less volume than the liquid, as is normal, a significant tension builds up that is likely to destroy the already frozen pore regions. Interestingly, it is another phenomenon that is likely to destroy the freezing gel, independent of the density change. If an ice layer forms at the surface of a gel, then we have the situation as in Fig. 5.5, that significant undercooling is required for the ice to penetrate the small pores. Alternatively, liquid may be drained from the pores and solidify on the crystal surface, as depicted in Fig. 5.21a. Since the growth of ice crystals is rather fast – Scherer estimates a front velocity u ¼ 0.1 mm s 1 for an undercooling of DT ¼ 0.2 K – the liquid flow out of the wet gel may be orders of magnitude higher than during evaporative drying.

Fig. 5.21 Freezing of gels: (a) drainage of liquid to ice at gel surface with tension build-up; (b) ice crystal growing inside the gel and pushing primary particles apart.

j183

j 5 Understanding and Preventing Structural Changes During Drying of Gels

184

The destructive effects, such as differential shrinkage and tensile stress at the gel surface, are of the same kind in both cases, so that for the freezing process, maximum tensile stress can be computed analogously to Eq. 5.17 as sx 

mw uL 3K

ð5:23Þ

But freezing, as compared to convective drying, takes minutes rather than hours or days, so that stresses higher than the modulus of rupture can develop, and failure even occurs without the mechanism of crack propagation. The resulting fragments themselves will be subjected to the same process of liquid drainage to a surface layer of ice. Breakage into finer fragments will continue until fragment size L is small enough for the gel to withstand the tensile stress sx in Eq. 5.23. If crystals nucleate inside the gel, as can be promoted by incorporating a catalyst into the gel network, then large undercooling is required for the penetration of ice into the small neighboring pore openings. Again, liquid will rather flow towards the ice crystal, which itself pushes the primary particles apart as it grows (see Fig. 5.21b), the phenomenon being analogous to frost heave in soil. Such growth of internal ice crystals may result in large pores of the freeze-dried gel, or cause macroscopic cracks when the gel cannot withstand the developing stress. Scherer also addresses gel shrinkage during freezing and finds that the gel is subjected to the maximum compressive pressure when flow towards a surface ice crystal is stopped by liquid/vapor menisci in other surface regions, and the ice crystal finally penetrates the pores. Then, the pressure in the liquid is Pw ¼

2sc;l rp

ð5:24Þ

where sc,l is the interface tension between crystal and liquid (approx. 0.04 N m 1 for water) and rp the pore radius. So, shrinkage is expected to be less pronounced than during convective drying, for which capillary pressure, Eq. 5.7, sets the condition for compressive pressure. In order to reduce the damage to gels during freezing, we may conclude that the density change in the pore liquid during solidification should be small and that crystal growth from the gel surface should be avoided in favor of nucleation inside the gel. A more detailed analysis of the stress exerted on the solid network by internal ice crystals (Scherer, 1993) reveals that the pore liquid should also have low entropy of fusion. For limited shrinkage, the interface tension between crystal and pore liquid sc,l should be low. And, most evidently, the gel strength plays a crucial role and should be as high as possible. Scherer also points out that very fast freezing could prevent damage because it would make the liquid vitrify rather than crystallize. Once the gel is frozen, it is sublimation-dried under vacuum. Care must be taken that the ice does not melt and that drying is complete before the temperature is raised to ambient in order to avoid further damage to the gel. Concerning the drying step, freeze-drying has the disadvantage of long process times, since the drying rate is limited by the saturation vapor pressure of ice that is generally very low.

5.4 Drying Methods for Gels – Quality Loss

5.4.3 Supercritical Drying 5.4.3.1 Supercritical Drying of the Initial Solvent Having seen that even costly freeze-drying is no evident solution, we turn to the drying technique that first produced highly porous dry gels, namely supercritical drying. The most obvious method is to take the wet gel as it is and transfer the pore liquid into its supercritical state. This is done by heating the gel in an autoclave in order to reach the supercritical state by increasing both temperature and pressure (see Fig. 5.14). For the following reason, it is crucial to place the wet gel in a bath of sufficient excess liquid (Phalippou et al., 1990): during heating, the liquid expands and part of it evaporates until the whole autoclave is either filled by liquid or vapor, and only further heating makes the fluid supercritical. There must be more initial liquid per autoclave volume than the critical density, for example, 0.28 g cm 3 for methanol and ethanol, to keep the gel covered by liquid and avoid evaporative drying of the gel during the heating step – with capillary effects. (Alternatively, the autoclave may be pressurized by an inert gas before heating.) The critical values for some singlecomponent solvents relevant in gel synthesis are given in Tab. 5.1. If the pore liquid consists of several components, the situation is more complicated because, then, the mixture must be supercritical (Phalippou et al., 1990). The heating rate dT/dt > 0 must not be too high to avoid non-uniform thermal expansion of the gel, especially the pore liquid, with the risk of damage. Scherer (1992c) has derived the stress in a heated gel cylinder (of radius R), which arises from the difference between the thermal expansion coefficients of liquid and solid, Da > 0, and low gel permeability that hinders the expanding liquid from flowing out of the gel. He finds a long-time solution for axial stress at the cylinder surface that is valid for compliant gels, both elastic and viscous,   3m R2 dT ð5:25Þ sz jr¼R  w y Da 8K dt

where the term in brackets is the thermal strain rate. If syneresis, and the corresponding shrinkage of the solid network, is accounted for, the stress is increased by an additional strain rate. The analysis further shows, that mechanical constraints of the gel should be avoided, otherwise significantly higher stresses may

Tab. 5.1

Critical constants for different solvents (H€ using and Schubert, 2006).

methanol ethanol isopropanol, acetone water carbon dioxide

Tcr ( C)

Pcr (MPa)

240 243 235 374 31

7.9 6.3 4.7 22.1 7.3

j185

j 5 Understanding and Preventing Structural Changes During Drying of Gels

186

develop. From time-dependent calculations, with silica gel and methanol as the pore liquid, Scherer can explain experimental observations, in which gel cylinders of diameter 6 cm crack at a heating rate of 2 K min 1, but stay intact when heating at 0.3 K min 1. After reaching the supercritical state, the pore fluid is removed by venting the autoclave. This depressurization is usually done under isothermal conditions to avoid the risk of condensation, and it must be slow enough to prevent significant stresses in the gel because the fluid cannot escape instantaneously but expands within the gel and causes the solid network to swell. Scherer (1994) has investigated the limiting conditions for the depressurization rate dP/dt < 0: assuming ideal gas behavior during the expansion, he obtains the following approximate expression for the maximal axial tensile stress at the surface of a gel cylinder (attained at the end of depressurization): sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  ffi   y m w dP  smax ð5:26Þ z jr¼R  R 4pK  dt 

The influence of sample size and depressurization rate has also been studied experimentally for different silica gels (Woignier et al., 1994). For all gels, there was a maximum rate below which the samples stayed monolithic and above which they cracked. The obtained results were in reasonable agreement with a more advanced model than Eq. 5.26 (Scherer, 1994). For cylindrical samples with both diameter and height equal to 3 cm, the maximum depressurization rate is around 0.1 MPa min 1 (see also Sakka and Kozuka (2005)). Further analysis for ethanol shows that the process time can be significantly reduced by decreasing the depressurization rate during the process, since for an identical decrease in pressure, the (non-ideal) gas expands much more at lower pressures (Woignier et al., 1994). Near atmospheric pressure, the autoclave is flushed with an inert gas, for example, nitrogen, to prevent condensation in small pores (Sakka and Kozuka, 2005). Finally, the autoclave is cooled down – again slowly to avoid mechanical damage. The above-described supercritical drying route is simple in theory but, in practice, has several disadvantages which come from the elevated critical parameters of the solvents used in gel synthesis (see Tab. 5.1). High temperatures and pressures, especially in large volumes of an industrial production, cause a considerable safety problem, and organic solvents bear the additional risk of inflammability. But further than this, the gels may undergo undesired structural changes at the elevated temperatures. Parts of the solid network may form crystals, or may be destroyed by the rather corrosive alcohol or water, and the elevated temperatures will enhance the above-described effects of ripening, that is, elimination of small pores by dissolution and re-precipitation, and syneresis, that is, formation of new bonds when distinct branches of the gel network approach each other by thermal fluctuations. During syneresis in silica gels, water is produced, so that the pore liquid changes composition and the initial critical conditions may not be enough to avoid capillary effects. It should be noted

5.4 Drying Methods for Gels – Quality Loss

Fig. 5.22 Pressure and temperature protocols for different drying methods with supercritical CO2 (ambient and supercritical conditions are denoted by indices 0 and 2, respectively).

that significant shrinkage (around 40% in volume) may occur during the described route of supercritical drying (Roig et al., 1998). 5.4.3.2 Low-Temperature Process with CO2 In order to circumvent the problems associated with the high critical temperatures of the pore liquid, a low-temperature supercritical drying route has been proposed (Tewari et al., 1985): it involves carbon dioxide, which has a low critical temperature and is inert. Due to the smaller temperature variation, the supercritical drying step can be several times shorter; however, a washing step is needed to replace the pore liquid by liquid carbon dioxide (see Fig. 5.22a). This is done in the autoclave at or near room temperature (at 20  C, CO2 is liquid for P1 > 5.6 MPa); again, the gel is initially covered by excess pore liquid, which is purged at elevated pressure before the actual washing step can start (liquid CO2 floats on top because of its low density) (Masmoudi, 2006). If the original pore liquid is not miscible in carbon dioxide, as is the case for water, it must first be substituted by an intermediate, for example, alcohol or acetone. Typically, these washing steps require many hours because the transport mechanism is diffusion; and the characteristic time increases with the square of the gel dimensions. Washing with liquid CO2 is done by rinsing the pressure vessel, and the concentration of the initial pore liquid is monitored by online gas chromatography; when it is lower than a critical value, for example, 10 3 of the initial concentration (Heinrich et al., 1995), the liquid is transferred into its supercritical condition. Isothermal depressurization and cooling of the autoclave conclude the drying process (see Fig. 5.22a). Significant improvement of this supercritical drying method can be achieved if the washing step itself is done with supercritical CO2, as illustrated in Fig. 5.22b, because the diffusion process is significantly faster for this fluid (Bisson et al., 2003). It is, however, not sufficient to be in the supercritical state of the single component CO2; instead, the mixture of CO2 and pore liquid must be supercritical. Taking a singlecomponent pore liquid, for example, ethanol, then, for any temperature above the critical temperature of CO2, there is a critical pressure (Schneider, 1978). In Fig. 5.23, these values are connected to a binary critical curve. Above this curve, the two phases are a completely miscible supercritical fluid; below the curve, liquid ethanol coexists with gaseous CO2. The consequences for the washing step are clear: if the system is biphasic, capillary effects are expected to destroy the gel. Van Bommel and

j187

j 5 Understanding and Preventing Structural Changes During Drying of Gels

188

Fig. 5.23 Occurrence of cracks in gels dried with supercritical CO2 depending on washing parameters as defined in Fig. 5.22b (taken from van Bommel and de Haan (1994)). (The shaded region marks the supercritical state of the single component CO2).

de Haan (1994) have confirmed this expectation by the experimental results shown in Fig. 5.23. Monolithic aerogel rods (ø 1.5 cm) could be obtained after supercritical washing with CO2 for 3 h at as moderate conditions as 35  C and 8.5 MPa and depressurization in less than 1 h. The same authors proved experimentally that recycle of the washing fluid is possible by first separating the ethanol (at 20  C), then liquefying CO2 (at 5  C), both at reduced pressure (5.0 MPa), and finally heating it back to the supercritical state. They designed a semi-continuous industrial production plant for aerogel plates of 3 cm thickness, in which five autoclaves operate in turn and water could be used for cooling and heating. Due to longer diffusion times and lower depressurization rates (about 0.1 MPa min 1) for the thick plates, a cycle time of 16 h was estimated. It may be concluded that the supercritical route is most successful in preserving the gel during the drying step. Due to the absence of interfacial forces, both shrinkage and cracks can be successfully avoided. The standard technique uses carbon dioxide for its moderate critical values, and the necessary washing step is mostly done under supercritical conditions. In the frame of two recent European research projects (HILIT and HILIT þ, 1998–2005), large monolithic silica aerogels for windows (55  55  1.5 cm3, see Fig. 5.24) could be successfully produced by this route (Masmoudi, 2006). Recent experimental work aims to optimize synthesis and aging conditions for producing aerogels with increased mechanical strength and good transmission properties in a reasonable processing time (Strøm et al., 2007); and to estimate effective diffusion coefficients for optimizing the washing with supercritical CO2 (Masmoudi et al., 2006). Theoretical work has tried to predict this coefficient from

5.5 Advanced Drying Techniques – Preserving Quality

Fig. 5.24 Highly insulating and light transmitting (HILIT) aerogel glazing; aerogel sheet is placed between two glass panes and evacuated (taken from Jensen et al. (2004)).

molecular and Knudsen diffusivities accounting for the pore size distribution of the gel (Orlovic et al., 2005). In the next section, we will present experimental efforts to look for alternatives to the high-pressure process of supercritical drying. The goal is the production of highly porous dry gels by subcritical and ambient pressure routes or by freeze-drying.

5.5 Advanced Drying Techniques – Preserving Quality

Having seen that supercritical drying can produce large monolithic highly porous dry gels of high quality, we will now show how less expensive and safer alternatives can compete with it. In principle, all other techniques are more harmful to the gel structure, and new approaches had to be developed to limit the damage during drying. The assignment of literature work to subsections on freeze-drying (Section 5.5.2), vacuum drying (Section 5.5.3) and convective drying (Section 5.5.4) accounts for the historical context and is not strict, because different drying techniques have often been compared in one and the same work. 5.5.1 Subcritical Drying

Only recently, the possibility of using subcritical drying techniques to produce xerogels has been explored; most of them will be treated in Sections 5.5.3 and 5.5.4 on vacuum and convective drying, respectively. Here, only one method close to supercritical drying as followed by Kirkbir et al. (1998a, b) will be presented. Wet silica gels were first washed (e.g., with ethanol) and then heated in a pressure

j189

j 5 Understanding and Preventing Structural Changes During Drying of Gels

190

chamber. During heating to about 250  C, the pressure increase first followed the vapor-pressure curve of the solvent until all solvent was evaporated (e.g., at 213  C and 3.5 MPa); further pressure increase was due to vapor expansion up to a maximal pressure (e.g., 4.4 MPa) well below the pressures required in supercritical drying (e.g., >6.4 MPa). Similarly, different maximum pressures were adjusted and several solvents investigated. In this drying method, capillary forces do develop, but are reduced due to solvent exchange and elevated temperatures. Additionally, the solvent exchange may induce structural changes and thereby influence gel behavior. From silica gels with moderate porosities (80–90%), large crack-free monoliths (e.g., cylinders with ø 5.6 cm and length 25 cm) could be prepared over wide ranges of maximal pressures. For ethanol as the pore liquid, no cracks were observed; additionally, shrinkage was negligible above a subcritical threshold whereas, below that pressure, the gel shrank significantly (linear shrinkage of 20% and more). For isobutanol, shrinkage was only observed at atmospheric pressure, but gels cracked below a threshold pressure. These threshold pressures are related to the temperature at which the gel dries out and, hence, to the maximal surface tension that the gel experiences; they depend on synthesis conditions which control the structural and mechanical properties of the gel, such as pore size and strength, respectively. Although process pressures can be significantly reduced (by 50%) by this approach, the previously described problems associated with the elevated temperatures remain. 5.5.2 Freeze-Drying 5.5.2.1 General Remarks Freeze-drying is an alternative to avoid liquid/vapor interfaces, as water is removed by sublimation after the freezing step. As indicated in Section 5.4.2, damage by cracks can be reduced by a pore liquid with low density change and low entropy of fusion; furthermore, a high freezing temperature is desirable for easy and complete freezing, and the saturation vapor pressure should be elevated for short drying times. One prominent substance fulfilling these requirements is tert-butanol (CH3)3COH (see Tab. 5.2); from its low entropy of fusion, it is expected to have a low crystal–liquid interface tension so that gel shrinkage should also be limited (Scherer, 1993). Tab. 5.2

Material properties relevant for freeze-drying. Freezing temperature ( C)

water tert-butanol

0.0 25.8

Density change during freezing (%) 8.5 0.04

Entropy of fusion (J cm 3 K 1)

1.2 0.24

Saturation vapor pressure at 0  C (Pa)

61 821

5.5 Advanced Drying Techniques – Preserving Quality

It seems to be difficult to produce high-porosity monoliths from inorganic gels by freeze-drying. In an early work, Degn Egeberg and Engell (1989) could produce large silica cryogels of elevated porosity (>70%) that cracked into several large translucent pieces. Without solvent exchange, or if not enough washing steps were used, small cryogel flakes were obtained; the authors attribute this to incomplete freezing of residual pore liquid (especially ethanol with a freezing point of 117  C), which subsequently boils during vacuum freeze-drying. Pajonk et al. (1990) report silica cryogels that have been produced as powders with relatively low density ( 0.2 g m 3), but with less pore volume and lower surface area than the corresponding aerogels. Structural properties closer to those of aerogels could be obtained by aging to strengthen the gel (as described below) or by synthesis under basic conditions for larger pore size. In contrast, the method is successfully used for RF gels (Mathieu et al., 1997) probably because these organic gels have a strength and elasticity that can better withstand the stresses during freeze-drying (Kocklenberg et al., 1998). Before presenting literature results on RF cryogels, we will provide some background knowledge on RF gels that has been acquired from supercritical drying and subsequent pyrolysis of these organic gels. Bock et al. (1997) studied RF gels with a target density of 0.35 g cm 3 (with skeletal density rs  1.5 g cm 3, corresponding to porosity y  77%) and systematically varied the molar R/C ratio. By nitrogen adsorption, SAXS and TEM measurements, they could show that for an increase in this synthesis parameter from 100 to 800 and above, the particle and pore radius are increased from below 5 nm to above 50 nm. At the same time, the BET surface area of RF aerogels decreases from above 600 m2 g 1 to below 100 m2 g 1 (for R/C ¼ 1500); after pyrolysis at 1050  C, this variation is reduced to the range 600–400 m2 g 1. The specific volumes of mesopores (2 nm < dp < 50 nm), which correspond to voids between the primary particles of the gel, and micropores (dp < 2 nm), which are voids inside these particles, are suitable criteria to measure the quality of the dry gels. Both are of specific interest for high-surface-area applications, whereas macropores only contribute to porosity, that is, accessibility of the small pores and low overall density. During pyrolysis, mass loss and volumetric shrinkage (both about 50%) are observed so that the solid density is increased, but the porosity remains almost constant; further, the pore size distribution is shifted to values about one third smaller, and micropores are created. Tamon et al. (1998) were interested in RF dry gels as precursors for mesoporous carbon with porosities > 80% and surface areas > 800 m2 g 1 for applications in adsorption, chromatography, catalysis, and as electrodes. Their supercritical drying experiments are complementary to those of Bock et al. (1997): by varying R/C from 12.5 to 100 and using different dilution ratios, they show that the mesopore volume and peak radius of monomodal pore size distribution of RF aerogels can be engineered; they describe the same behavior during pyrolysis (mesopore volume reduction, decrease in peak radius, unchanged shape of pore size distribution and moderate increase in BET surface area). An increase in pyrolysis temperature (from 500 to 1000  C) results in more shrinkage, but leaves the peak radius unchanged.

j191

j 5 Understanding and Preventing Structural Changes During Drying of Gels

192

5.5.2.2 RF and Carbon Cryogels The first RF cryogels were prepared by Mathieu et al. (1997) and Kocklenberg et al. (1998); they varied the R/C ratio between 50 and 250, used different dilution ratios and aging times, and characterized the pore structure of the resulting cryogels by nitrogen sorption, mercury porosimetry and SAXS measurements. They reported monolithic structures with good cohesion, high BET surface areas ( 450 m2 g 1), a mesopore volume of about 0.5 cm3 g 1 and a micropore volume of about 0.2 cm3 g 1. At the same time, Tamon and coworkers started exploring alternatives to the expensive and dangerous supercritical route for producing carbon precursors, with a focus on freeze-drying (Tamon and Ishizaka, 1999; Tamon et al., 1999). The gels were synthesized from resorcinol, formaldehyde (37% in methanol), sodium carbonate and distilled water (W) with two different R/C ratios (25 and 200) and two different dilution ratios R/W (0.125 and 0.25 g cm 3). The solution was poured into glass tubes (ø 4 mm, 4 cm long) and gelled and cured for 1 day at 25  C, 1 day at 50  C and 3 days at 90  C. Then, different drying routes were used to produce RF dry gels: .

.

. .

Supercritical drying: the hydrogel is first washed with acetone (>1 week), then dynamically washed with liquid CO2 (2 h) and with supercritical CO2 (4 h). Freeze-drying: the hydrogel is immersed in 10 times its volume of tert-butanol (>1 d, repeat three times), then frozen (at 30  C for 1 h) and freeze-dried (at 30  C for 1 d, then at 10  C for 1 d, then at 0  C for 1 d). Vacuum drying: at room temperature. Hot air drying: at 50  C.

All dry gels are aged (at 250  C for 8 h) and characterized by nitrogen sorption. The results, reported in Fig. 5.25 and Tab. 5.3, indicate that freeze-drying produces more shrinkage and less mesopore volume than supercritical drying, but is still much better than the evaporative drying routes. Furthermore, this quality loss can be compensated by subsequent pyrolysis (heating rate 250 K h 1; first to 250  C and hold for 2 h; then to 1000  C and hold for 4 h). Both aerogel and cryogel undergo significant volumetric shrinkage (>70%) that is reflected in a loss of macropore volume, but their porosity stays high (>80%) because of mass loss during carbonization.

Fig. 5.25 Size distribution of mesopores in dry RF gels (R/C ¼ 200, R/W ¼ 0.25 g cm 3) for different drying methods (from Tamon et al. (1999)).

5.5 Advanced Drying Techniques – Preserving Quality Tab. 5.3 Structural properties of dry RF gels (R/C ¼ 200, R/W ¼ 0.25 g cm 3) for different drying methods (from Tamon et al. (1999)).

supercritical drying freeze-drying vacuum drying hot air drying

Volumetric shrinkage (%)

Dry gel porosity (%)

Mesopore volume (cm3 g 1)

Macropore volume (cm3 g 1)

12

85

2.71

0.91

25 33 82

79 (no data) 35

1.02 0.62 0.18

1.42 (no data) 0.15

The mesopore volume of the cryogels is increased so that mesopore volumes around 1.5 cm3 g 1 and BET surface areas in the range 900–1200 m2 g 1 can be achieved for both supercritical and freeze-drying. After these encouraging results, the influence of freezing and freeze-drying conditions on gel quality was investigated further (Tamon et al., 2000; Yamamoto et al., 2001a, 2005). For identical synthesis conditions as above, the pore liquid, washing protocol and temperature protocol during freezing and freeze-drying were varied. If the gel was rinsed in distilled water (>1 d), it did not stay monolithic but was crushed into pieces, as expected from the theoretical considerations in Section 5.4.2. For water as the pore liquid, shrinkage is much more pronounced, the mesopores show considerable volume loss and reduction in size (see Fig. 5.26a); and the BET surface area is only 400 m2 g 1. For solvent exchange with tert-butanol, the RF cryogels have a BET surface area around 550 m2 g 1, a mesopore volume around 1.1 cm3 g 1 and a peak pore radius of about 6 nm. High reproducibility of these data has been proven (Yamamoto et al., 2001a). It is important, that the initial pore liquid, which contains methanol and formaldehyde (with freezing temperatures below 90  C), is sufficiently replaced, otherwise capillary forces may develop due to unfrozen residual liquid. The influence of the number of rinses (each 1 d) is shown in Fig. 5.26b: for only one rinse, the observed volumetric shrinkage is 55%, whereas two rinses or more can reduce this to 30%. At a low freezing temperature, 196  C (using liquid nitrogen) instead of 30  C, undercooling is more pronounced and liquid can more easily freeze in the pores instead of flowing out of them; consequently, the pore size (and volume) is better preserved (see Fig. 5.26a). Efforts to decrease the freeze-drying time showed that 5 d are needed at a constant temperature of 30  C. If the process is stopped after 3 d or less, the pore liquid is incompletely removed and, after thawing, causes capillary forces and gel shrinkage to peak: consequently, the cryogels have a smaller peak pore radius ( 70 mol m 3, there are virtually no mesopores (i.e., pores

5.5 Advanced Drying Techniques – Preserving Quality

larger than 2 nm). The differences in the obtained gel structures are explained by the different particle growth: for high C/W (high pH), the particles of the gel are densely distributed and can only grow a little, whereas for low values of C/W, fewer particles are created and they can grow to larger sizes. The limitation to low C/W can be overcome, if gelation is assisted by ultrasonic irradiation, producing so-called “sonogels” (Tonanon et al., 2005). Then, also for C/ W ¼ 80 mol m 3, carbon gels (ø 3 mm, 4 cm long) can be obtained with high BET surface area (700 m2 g 1), significant mesopore volume (0.6 cm3 g 1) and a peak pore radius of 2 nm, which is not possible by the unassisted synthesis route. Additionally, gelation times are reduced by the ultrasonic irradiation from more than 1 d to a few hours. In the effort to replace freeze-drying by cheaper and faster hot air drying, the respective evolutions of pore size distribution were tracked for a gel-rod, by use of thermoporometry on the wet gel and nitrogen sorption on the cryogel and xerogel (see Fig. 5.7): whereas freeze-drying largely preserves the pore size, quality loss is observed during hot air drying due to significant shrinkage of the mesopores (Yamamoto et al., 2005). Nevertheless, for particulate carbon gels, hot air drying seems to be an alternative. Wet gel microspheres of 30–70 mm have been synthesized by dispersing the starter solution (R/C ¼ 400, R/W ¼ 0.25 g cm 3) into cyclohexane just before gelation, creating an inverse emulsion (Yamamoto et al., 2005). Then, hot air drying (at 50  C for 1 d) of these microspheres – in comparison to the above described process of freeze-drying – produces carbon gels with less, but still competitive, mesopore volume (0.52 cm3 g 1, instead of 0.85 cm3 g 1), the same micropore volume (0.22 cm3 g 1) and the same elevated BET surface area (almost 700 m2 g 1). This route may offer an alternative for the production of mesoporous carbon microspheres for use in, for example, adsorption techniques. In order to save costs, not only in drying but also on the expensive raw materials, resorcinol has been replaced by the natural component wattle tannin (Tamon et al., 2006a). For optimized synthesis conditions, pyrolyzed tannin-formaldehyde cryogels attained BET surface areas up to >600 m2 g 1 and mesopore volumes of up to 0.8 cm3 g 1, and are hence competitive. By additionally replacing formaldehyde with furfural, first results are promising, but more research to optimize the synthesis conditions is needed (Kraiwattanawong et al., 2007). 5.5.2.3 Ice Templating (for Silica Gels) The effects during gel freezing that are caused by liquid flow out of the pores, which are usually seen as a disadvantage, may also be used for an elegant templating technique to create engineered macromorphologies. Nishihara et al. (2005) studied silica cryogels obtained from controlled unidirectional freezing of pore water by immersion into a cold bath (immersion rates 6–20 cm h 1; bath temperature 196 or 60  C). After some distance, pseudo-steady-state growth of an array of polygonal ice rods in the fresh gel is observed, which serve as templates for ordered macropores in a microhoneycomb. Following low-temperature aging ( 30  C), these ice templates are removed simply by thawing (50  C). Then, the gel may ripen for several days in a hydrothermal treatment under basic

j195

j 5 Understanding and Preventing Structural Changes During Drying of Gels

196

conditions, before water (in the smaller pores) is substituted by tert-butanol (three one-day washings). By freeze-drying (at 10  C) mesoporous walls are obtained; and heat treatment concludes the production process, by which large monolithic silica gels have been obtained (ø 10 mm). The macropore size can be tuned in the range 4–40 mm since the diameter of the ice crystals depends on the immersion velocity u and the temperature difference (T0 T1) as dice /

1 uðT0 T1 Þ

ð5:27Þ

The wall thickness can be adjusted from 0.2 to 2 mm; it is proportional to the product of the ice crystal diameter dice and the silica concentration in the starter solution. The overall porosity of the micro-honeycomb is about 94%; that of the walls around 40% (corresponding to a random packing of monosized spheres). Meso- and micro-porosity can be adjusted independently: through hydrothermal treatment the initial high BET surface area (up to 900 m2 g 1) decreases and the small pores (radii < 1 nm) grow (up to 45 nm) so that mesopore volume is created. Microhoneycombs are suitable for many applications, since the macrochannels ensure a low pressure drop, and diffusion parameters in the porous walls can be engineered. Depending on the firmness of the hydrogel, which can be tuned by synthesis parameters, several morphologies (see Fig. 5.27) are obtained (Mukai et al., 2008): microhoneycombs for softer gels, and polygonal fibers for harder gels; and if the hydrosol is frozen before gelation, lamella or flat fibers result for higher or lower mobility of silica particles, respectively. A modern use of titania–silica microhoneycombs may be in self-cleaning applications (Tamon et al., 2006b): by the sol–gel process, the titania particles are more finely dispersed in the matrix than by conventional methods, and much higher BETsurface areas can be reached; consequently, pulverized gels proved to have outstanding photocatalytic activity in the decomposition of large organic molecules.

Fig. 5.27 Principle of ice templating and possible structures (adapted from Tamon et al. (2006b).

5.5 Advanced Drying Techniques – Preserving Quality

5.5.3 Vacuum Drying

The use of vacuum drying to obtain RF xerogels and the derived carbon materials has been explored by Job et al. (2004, 2005); they showed that the pore structure can be partially preserved without pretreatment. The motivation was that traditional geldrying methods – supercritical and freeze-drying – are expensive and time-consuming and may, nevertheless, induce structural changes and quality loss; and that vacuum drying is a soft drying method. In a first study (Job et al., 2004), RF gels were prepared for several pH values (tuned by addition of sodium carbonate) and a fixed initial solid density of 0.35 g cm 3 (adjusted by addition of de-ionized water) and aged at 85  C for 3 d. The appearance of the wet gels varied from opaque and light brown at low pH (high R/C ratio) to translucent dark red at high pH (low R/C). These hydrogels were dried in a vacuum oven at 60  C by gradually reducing the pressure from atmospheric to 1000 Pa over 5 d, and then were kept at 150  C for 3 d. The obtained xerogels were pyrolyzed in a nitrogen atmosphere by progressively increasing the temperature to 800  C (at 1.7 K min 1 to 150  C, held for 15 min, at 5 K min 1 first to 400  C where they were held for 1 h, then to 800  C and held for 2 h ) and then slowly cooled. Above this pyrolysis temperature, no further mass loss was observed and infrared spectra remain unchanged (Job et al., 2004). The carbonized materials are black, matt (for low pH) or bright (for high pH). In order to characterize the structure of RF and carbon xerogels, a combination of nitrogen adsorption (for micro- and meso-pores) and mercury porosimetry (for pore diameters from 7.5 to 150 nm) was used to obtain the BET surface area and pore volume (microporous and total); helium and mercury pycnometry were applied to determine the skeletal and bulk density. During drying, the gels showed significant shrinkage but stayed monolithic. An increase in pH (from 5.45 to 7.35) results in smaller pore size – hence shifting the materials from mesoporous to microporous – and lower porosity (from 68% to only 22%). After pyrolysis, carbon gels are obtained with densities from 0.53 to 1.35 g cm 3 (as compared to skeletal density rs  2.2 g cm 3). Their BET surface areas are around 600 m2 g 1 for pH < 6.5, whereas for higher pH (low R/C) no pores can be detected, but the apparent density suggests a (closed) porosity of 40%. The carbon materials possess good mechanical strength. In a second study (Job et al., 2005), vacuum drying was directly compared to supercritical drying (after washing with ethanol and liquid CO2) and vacuum freezedrying (with no solvent exchange and freezing with liquid nitrogen). For this comparison, different R/C ratios and two different initial solid densities (0.15 and 0.35 g cm 3) were chosen. As R/C increased from 50 to 1000, the major structural change was an increase in the size of the primary particles, from about 5–10 to 25–30 nm. Generally speaking, capillary forces are more pronounced as the size of the particles – and hence the pores – is reduced, that is, for low R/C; and the mechanical stability of the RF gels increased with their density.

j197

j 5 Understanding and Preventing Structural Changes During Drying of Gels

198

All aerogels stayed monolithic, but some showed volumetric shrinkage, which was attributed to residual water and ethanol after the washing steps and can reach up to 40% for high R/C and high dilution ratio. By variation of the synthesis conditions, a wide range of apparent densities (as low as 0.19 g cm 3) and pore sizes can be covered. Cryogels could only be obtained as small pieces; according to the authors, this might be due to the thermal shock and solvent expansion during freezing (Job et al., 2005). Two distinct behaviors could be observed: if the solid density of the wet gels was low, ice crystals may grow inside the gel, reducing the size of the meso- and macro-pores and creating megalopores larger than 10 mm. As discussed previously (see Fig. 5.21), this phenomenon is not related to the density change during freezing and not specific for water. (Note that classical macropores have sizes 50 nm < dp < 1 mm.) For wet gels with high solid density, the solid structure was either preserved, or the gel shrank considerably (by 56% in volume) if the pore size was smaller than 40 nm. This was probably caused by melting of the confined solvent since sample temperature was not controlled but increased during the process. Both aerogels and cryogels are very fragile if their porosity is high (>80%), and mechanically resistant if they are denser. Vacuum-dried xerogels underwent more shrinkage than gels dried by the other methods; volumetric shrinkage was more pronounced for smaller particle size and lower initial solid density and reached up to 87%. The final density depended only on R/C, which determined the particle size, and increased from 0.52 to 1.1 g cm 3, as this ratio decreased from 1000 to 50. In fact, the gels are compressed until the capillary force (set by the size of particles or pores) is counterbalanced by the mechanical strength of the network. The xerogels were monolithic for large pore size (at high R/C), whereas for small pore size (at low R/C) they could show cracks or, for low initial density, broke into pieces; they were not friable or fragile, and could be handled without damage. Due to shrinkage, samples with high porosity and small pores (1 d, there is only a little shrinkage and the best xerogels are obtained. There is also some shrinkage during supercritical drying that can be reduced by aging. For aged silica gels of density around 0.21 g cm 3, convective drying can produce gels of the same quality as supercritical drying – without shrinkage. With a skeletal density of rs  2.15 g cm 3 (Hæreid et al., 1995), this corresponds to a porosity of y  90%. By the effects of aging, the shear modulus ( bulk modulus) of wet gels increases by one order of magnitude (to values > 10 MPa), and the surface area of the corresponding xerogels decreases from 1000 to 700 m2 g 1.

j199

j 5 Understanding and Preventing Structural Changes During Drying of Gels

200

Fig. 5.28 Drying of silica gels aged in monomer solution; aging times are indicated for xerogels; gray lines indicate levels of volumetric shrinkage (adapted from Einarsrud (1998)).

The shrinkage of aged gels can also be understood by a model (Scherer et al., 1996) that predicts the solid density of the gel after convective drying, using wet gel data (permeability and bulk modulus) and accounting for the increase in bulk modulus during drying. It should be mentioned that there is no obvious alternative to the aging step: on the one hand, a simple increase in silica concentration in the starter solution only increases the gel density, without the corresponding increase in stiffness; on the other hand, quite stiff gels can be obtained under acidic synthesis conditions, but they have a smaller pore radius so that capillary forces still cause damage (Einarsrud, 1998). A similar approach has been followed by Leventis et al. (2005) who studied polymer crosslinking of silica gels. The hydrogel (from TMOS via a base-catalyzed route) is first aged in mother liquor, then washed with ethanol, acetone and finally an acetone solution of hexamethylene diisocyanate O¼C¼N–(CH2)6–N¼C¼O; during subsequent curing, a conformal polymer coating forms on the gel network (see Fig. 5.29). If such a gel is dried in a 40  C oven at ambient pressure after substituting the pore liquid for pentane, no shrinkage occurs because of the dimensional stabilization by crosslinking. Moreover, the xerogel shows no difference in appearance and structural parameters to supercritically dried gels. The xerogels (ø 10 mm, length 40 mm) have a relatively low BET surface area of 160 m2 g 1 and a relatively high density of 0.56 g cm 3; indeed, crosslinked gels with lower densities were not stable during drying. However, their mechanical strength is increased by two orders of magnitude, and they are less hydrophilic in comparison to the underlying plain silica gel (Leventis et al., 2005). Besides providing a method to produce future lightweight materials, the study is interesting because it addresses the role of surface tension in drying shrinkage

5.5 Advanced Drying Techniques – Preserving Quality

Fig. 5.29 Modification of silica aerogel by polymer-crosslinking and effect of capillary forces during convective drying (rearranged SEM images from Leventis et al. (2005)).

or rather collapse of the pore structure. To this purpose, crosslinked aerogels have been soaked in liquids with different surface tensions and dried convectively. As shown in Fig. 5.29, drying with pentane preserves the structure, whereas the pores collapse for liquids with higher surface tension, such as chloroform. The average pore radius (measured by nitrogen sorption) decreases from 9 nm to below 2 nm as the surface tension of the pore liquid increases from 0.016 to 0.027 N m 1. Strengthening of the gel network for successful convective drying has also been investigated for other classes of gels. Gan et al. (2005) produced highly porous alumina monoliths. An aluminum nitrate/ethanol solution was gelled with propylene oxide as gelation agent and formamide as the drying control chemical additive (DCCA). After aging in mother liquor, the gels were washed and aged in ethanol, TEOS/ethanol and again ethanol, to strengthen them by silica crosslinking, before drying at ambient pressure with a slow (10 K h 1) temperature rise to 70  C. Different molar ratios of gelation agent and DCCA have been selected to produce low-density silica-reinforced alumina gels. An increasing amount of formamide leads to smaller pores (pore size distributions by nitrogen sorption) and to more shrinkage; the role of DCCA during the sol–gel process is not clear, but the xerogel density passes through a minimum as its concentration is increased (see Fig. 5.30). For the optimal composition of the starter solution, a density of 0.26 g cm 3 (i.e., porosity >90%) was reached with particle sizes from 10–30 nm (see Fig. 5.30) and BET surface area around 450 m2 g 1. Other DCCA substances did not yield as low densities, and without any DCCA, the xerogel density was 0.4 g cm 3. 5.5.4.3 Making Shrinkage Reversible by Surface Modification (Silica Gels) Instead of preventing shrinkage, one may also try to make it reversible. To this aim, Smith et al. (1995b) modified the surface of the silica matrix before drying. The underlying idea is as follows: in unmodified silica gels, when pore walls get into close

j201

j 5 Understanding and Preventing Structural Changes During Drying of Gels

202

Fig. 5.30 Optimal ratio of DCCA for producing low-density alumina gel and TEM image (adapted from Gan et al. (2005)).

proximity during drying shrinkage, condensation reactions occur between neighboring hydroxy groups and make the shrinkage permanent. In order to prevent this crosslinking, the surface is methylated in a reaction with trimethylchlorosilane (TMCS): :Si-OH þ ClSiðCH3 Þ3 ! :Si-O-SiðCH3 Þ3 þ HCl

ð5:28Þ

(Such a substitution is also called silylation.) Then, bond formation in the compressed state is prevented and springback is possible when the capillary forces disappear at the end of the drying. In Fig. 5.31, the drying behavior of modified and unmodified gels is shown: both shrink to around 27% of their original volume before springback is observed to merely 30% for the unmodified gel and 97% for the modified one. Drying was performed at 35  C for 2 d, after exchanging the pore liquid for n-heptane. A related study (Smith et al., 1995a) showed that the density of the dry gels did not significantly vary when drying times of a cylinder with ø 10 mm were

Fig. 5.31 Volume of silica gel during convective drying; springback occurs at the very end and its magnitude depends on the surface chemistry (from Smith et al. 1995b).

5.5 Advanced Drying Techniques – Preserving Quality

reduced from two days to several minutes. However, the dried gels broke into fragments of decreasing size (from 8 mm to < 3 mm) when the drying rate was increased, thereby proving that the stresses and damage by cracks depend on the drying rate. More recent work investigates the role of surface modification with TMCS for small silica gel spheres (ø 3 mm) for springback and vapor diffusion during the second drying period (Bisson et al., 2004). Drying took 2 h at 25  C under controlled nitrogen flow. The modification step needs to be long enough (2 d) for complete volume recovery. Then, the dry gel density is almost as low as that obtained by supercritical drying (0.15 g cm 3 instead of 0.13 g cm 3); the slight increase is due to mass uptake by the chemical reaction. The shrinkage before springback is also reduced if the surface is (more completely) modified; the authors argue that lower capillary forces must be the reason, but that the explanation cannot be given in terms of surface tension and contact angle. Instead, they showed by experiment that the wetting energy for the pore liquid (isopropanol) is significantly reduced by surface modification, and they argue that this sets a limit for how much (capillary) force can be transmitted from the pore liquid to the solid. The authors also highlight the low effective diffusivity of vapor that is three orders of magnitude lower than expected from the Knudsen diffusivity. Transport seems to be controlled by adsorption–desorption effects. Finally, surface modification also makes the dry gel less hydrophilic (Bisson et al., 2004). With the purpose of producing hydrophobic silica xerogel monoliths of high porosity, other silylation agents have been systematically explored by Rao et al. (2007). During drying of the modified gels with hexane (temperature protocol: 50  C for 6 h, 120  C for 12 h, 200  C for 6 h), their volume change was recorded. The extent of shrinkage during drying is similar for all investigated agents, but volume recovery during springback is better if the silylation agent contains a higher number of alkyl groups; and hexamethyldisilazane (H3C)6Si2NH is even more suitable than TMCS, yielding a density of 0.06 g cm 3 (instead of 0.105 g cm 3) and a porosity of 96.9% (instead of 94.6%), and having better transmission of light (65% instead of only 30%). The excellent hydrophobicity of the modified gels has been confirmed by contact angle measurement and long-term adsorption test in a water-saturated atmosphere, and their thermal stability up to 325  C has also been proven. In recent efforts towards a cheaper production process for highly porous silica gels, costly raw materials such as TEOS have been replaced by cheap water glass; then, sodium is substituted in an ion exchange (Schwertfeger et al., 1998; Hwang et al., 2007). Furthermore, time-consuming and costly solvent exchanges used for surface modification with TMCS can be limited. In one approach (Schwertfeger et al., 1998), silica hydrogel is put in hexamethyldisiloxane (HMDSO) and TMCS is added; then TMCS reacts with the pore water to give HMDSO and HCl. HMDSO is not miscible in water and forms a layer on the pore walls expelling the aqueous HCl. So, besides silylation with TMCS, the solvent is exchanged automatically, and the byproducts are separated outside the gel. Durably hydrophobic xerogels with porosities > 85% have been produced in this way (Schwertfeger et al., 1998).

j203

j 5 Understanding and Preventing Structural Changes During Drying of Gels

204

Fig. 5.32 Springback during convective drying of surface modified silica gel (from Hwang et al. (2007)). Here, “drygel” indicates the point when pores start to dry out; “aerogel” is the highlyporous xerogel.

In another approach (Hwang et al., 2007), one solvent exchange with isopropanol/ TMCS/n-hexane solutions was performed on the hydrogel with similar reactions and a phase separation as above: the surface is modified and pore water is replaced by low surface tension hexane. In this way, large monoliths (discs of ø 22 mm, 7 mm thick) have been obtained by controlled evaporation under hexane atmosphere at room temperature for 3 d. If the silica content in the starter solution (i.e., before gelation) was too low, the gel collapsed during drying and resulted in a powder; if it was too high, many cracks were observed. In the safe range (4 to 8 wt%), the performance depended on the amount of TMCS in the exchange solution: too little led to irreversible shrinkage, too much produced fracture during springback. Only the right amount (with molar ratio TMCS/pore water of 0.3 to 0.4) resulted in good springback (94% volume recovery) and crack-free xerogels with porosities around 93% (see Fig. 5.32) and surface areas around 675 m2 g 1. Instead of surface modification, one may also directly synthesize organic–-inorganic hybrid gels to obtain springback behavior during drying, and elastic deformability of the dry gel: for example, Kanamori et al. (2008) produced gels with high porosity (> 80%) from methyltrimethoxysilane in a two-step process with urea as base-releasing agent and a surfactant to prevent phase separation. For certain synthesis parameters, supercritical drying can be substituted for convective drying without change of pore structure (confirmed by SEM images). During drying, the hybrid gel networks display reversible volumetric shrinkage of more than 60%; additionally, uniaxial compression of the dry gel up to a linear strain of 80% is reversible. The authors explain this extraordinary behavior by low crosslinking density keeping the gel flexible, low density of silanol groups on the network surface that could make shrinkage irreversible and repulsive forces between methyl groups. 5.5.4.4 RF Gels – From First Results to a Systematic Investigation Encouraged by the results obtained with vacuum drying (see Section 5.4.3), L
eonard and coworkers (2005a, b) explored the possibility of convective drying for RF hydrogels. Using wet gels (ø 22 mm, 13 mm high) prepared with identical synthesis conditions as before, convective drying was directly compared to vacuum drying. The gels were dried for 4–5 h with air of ambient humidity, heated to 70  C and flowing at

5.5 Advanced Drying Techniques – Preserving Quality

2 m s 1; all xerogels stayed monolithic despite the immense reduction in drying time (by a factor of 40). An increase in pH from 6 to 7 leads to an increase in microporosity and in BET surface area from 117 to 355 m2 g 1 and an increase in solid density in the RF xerogels. Vacuum drying yields higher surface area (approximately 500 m2 g 1 for the whole pH range), but always produces more shrinkage than convective drying, suggesting that the lower evaporation rates might be the reason (see also Mayor and Sereno (2004)). For pH 6, convective drying produced xerogels with the highest porosity (77%) and no shrinkage (initial and final solid density were both 0.35 g cm 3). This is explained by two effects: at lower pH, the pore size is large and the gel elastic, in contrast to the viscoelastic behavior and small pore size at higher pH which both favor shrinkage under capillary compression. After pyrolysis, these xerogels are slightly more porous (up to 82% for pH 6), but have slightly increased density due to carbonization (the lowest value being 0.39 g cm 3 for pH 6) – an effect that is also reported for vacuum-dried gels. Additionally, the convectively dried gels show an increase in BET surface area and microporosity for all investigated pH (up to 628 m2 g 1 and 0.27 cm3 g 1, respectively, for pH 6). In contrast, open porosity and BET surface area could not be detected for pyrolyzed vacuum-dried gels at pH 6.5 and 7, probably because less condensation reactions (in lack of H þ ) lead to a gel structure that is weak enough to collapse during the chemical restructuring of pyrolysis. After these promising first results, a systematic investigation of convective drying was started, with a variation in drying air temperature and velocity to study their influence on the drying behavior of RF gels with different R/C ratio (Job et al., 2006b). During drying, the sample is continuously weighed to deduce the drying rate. Additionally, the sample is characterized by X-ray microtomography (with a spatial resolution of 41 mm) at regular intervals to track changes in sample volume and surface, and also the evolution of cracks; care is taken that the intervals for characterization are short enough to not disturb the drying process. Several RF hydrogels (ø 28 mm, 10 mm high), with different R/C ratios, but all having the same initial solid density of 0.35 g cm 3, were gelled and aged at 70  C for 24 h. Then, they were convectively dried in a laboratory drying tunnel with ambient air (Y  0.007); different settings of temperature and velocity were chosen. Drying rate curves (evaporation flux versus moisture content) were reported and the correlation between drying conditions and drying kinetics was found to obey classical theory. The dry gels were characterized by nitrogen sorption and mercury porosimetry to obtain BET surface areas, approximate pore sizes, micropore volumes (pore radii 50

75

115

2.2

160

1.1

70 92.5 115 30 50 70 115

5.6 4.2 1.8 20.1 17.8 6.9 2.8

monolith (cracks) monolith (cracks) monolith (cracks) monolith monolith fragments monolith fragments fragments fragments

320

0.15

35–45

55

308

0.14

10–15

37

500

300

12 13 47 50 — 60 — — —

cracks during drying, which first opened and then closed due to stress reversal at the end of drying, as shown in Fig. 5.33. For R/C ¼ 500, the gels showed no cracks for low enough drying rates, but otherwise broke into pieces; if R/C is further reduced to 300, drying conditions had to be softened to avoid such catastrophic cracks (as expected from the theory of Scherer). For identical drying conditions, the drying time decreases with increasing R/C, probably due to increasing pore size that facilitates vapor diffusion. From height and cross-section measurements, shrinkage is found to be isotropic. (For gels synthesized with a wider range of R/C that were dried slowly, the shrinkage curves are shown in Fig. 5.34.) During the first drying period, the shrinkage is ideal, that is, the volume reduction is only due to evaporation and all the pores are filled. In

Fig. 5.33 X-ray microtomograph cross-sections of RF gel (R/C ¼ 1000, ø 28 mm, 10 mm high, X0 ¼ 2.03) during convective drying at 115  C for (a) X ¼ 0.47, (b) X ¼ 0.2 and (c) X ¼ 0.002 (from Job et al. (2006a)).

5.5 Advanced Drying Techniques – Preserving Quality

Fig. 5.34 Shrinkage of RF gels (; 22 mm, 13 mm high) during slow convective drying with ambient air (solid lines are guides to the eyes; adapted from L
eonard et al. (2008)).

the second drying period, the gel volume remains constant and the pores dry out. With increasing R/C, that is, increasing pore size and stronger gel structure, the shrinkage period becomes shorter, or even negligible. Accordingly, the highest xerogel porosity was obtained for R/C ¼ 1000, for which shrinkage is low and the drying front recedes quickly from the gel surface (as can be estimated from the different gray levels in Fig. 5.33a). The extent of shrinkage is found to be independent of the drying conditions, for the investigated range of air temperatures and velocities. Similarly, the structural properties – BET surface area, micropore volume and average pore size – show no dependence on the drying conditions, but are determined by synthesis parameters. As in vacuum drying, RF xerogels with high porosity and small pores ( 105 s), when the gel stops shrinking and its surface is no longer saturated with pore liquid; then macroscopic stress in combination with local differences in capillary forces results in cracks (as described in Section 5.4.1). In this case, the crack pattern is irregular. If the drying rate exceeds a second critical value, cracks already occur at the very beginning (at times 90%) and pore size in the low-nanometer range, as required for insulation purposes, are to be produced by evaporative drying techniques. New multi-scale modeling is expected to shed more light on the mechanisms causing shrinkage and cracking. This will allow optimization of drying conditions, but might also give the chance to engineer gel structures that are suitable for convective drying.

Additional Notation Used in Chapter 5

As C0 C1 C2 c dp E F F f Icr K KG Kp K0 m N n R R Ri r rp

specific surface area constant ( 1/3) constant defined by Eq. 5.20 constant defined by Eq. 5.20 crack length pore diameter Young’s modulus force uniaxial solid viscosity phase volume fraction critical stress intensity permeability bulk solid viscosity bulk modulus initial bulk modulus exponent in Eq. 5.11 Poisson’s ratio for viscous gel unit normal vector aggregate radius radius of gel cylinder mechanical moduli in Fig. 5.43 primary particle radius pore radius

m2 kg 1 — Pa — m m Pa N Pa s — Pa m0.5 m2 Pa s Pa Pa — — — m m Pa m m

j221

j 5 Understanding and Preventing Structural Changes During Drying of Gels

222

Greek letters

a e_ q k n P ~ s t t y

thermal expansion coefficient strain rate contact angle mean curvature Poisson’s ratio quantity defined by Eq. 5.10 solid network stress shear stress relaxation time porosity (voidage)

K 1 s 1 rad m 1 — — Pa Pa s —

Subscripts and superscripts

c c cr p S T x, y, z 0

(ice) crystal curvature critical pore shrinkage total in coordinate direction initial

Abbreviations

BET C CT DCCA DSC ESEM HMDSO R RF S SAXS SEM TEM TEOS TMCS TMOS W

Brunauer, Emmett and Teller theory catalyst computed tomography drying control chemical additive differential scanning calorimetry environmental scanning electron microscopy hexamethyldisiloxane resorcinol resorcinol-formaldehyde surfactant small angle X-ray scattering scanning electron microscopy transmission electron microscopy tetraethoxysilane trimethylchlorosilane tetramethoxysilane water

References

References Alaoui, A. H., Woignier, T., Pernot, F., Phalippou, J., Hihi, A., 2000. Stress intensity factor in silica alcogels and aerogels. J. Non-Cryst. Solids 265: 29–35. Ali
e, C., Benhaddou, A., Pirard, R., Lecloux, A. J., Pirard, J. P., 2000. Textural properties of low-density xerogels. J. Non-Cryst. Solids 270: 77–90. Ali
e, C., Pirard, R., Pirard, J.-P., 2001. Mercury porosimetry: applicability of the bucklingintrusion mechanism to low-density xerogels. J. Non-Cryst. Solids 292: 138–149. Al-Muhtaseb, S. A., Ritter, J. A., 2003. Preparation and properties of resorcinolformaldehyde organic and carbon gels. Adv. Mater. 15: 101–114. Bellunato, T., Calvi, M., Matteuzzi, C., Musy, M., Perego, D. L., Storaci, B., 2007. Refractive index dispersion law of silica aerogel. Eur. Phys. J. C 52: 759–764. Berthon, S., Barbieri, O., Ehrburger-Dolle, F., Geissler, E., Achard, P., Bley, F., Hecht, A. M., Livet, F., Pajonk, G. M., Pinto, N., Rigacci, A., Rochas, C., 2001. DLS and SAXS investigations of organic gels and aerogels. J. Non-Cryst. Solids 285: 154–161. Bisson, A., Rigacci, A., Lecomte, D., Rodier, E., Achard, P., 2003. Drying of silica gels to obtain aerogels: phenomenology and basic techniques. Drying Technol. 21: 593–628. Bisson, A., Rodier, E., Rigacci, A., Lecomte, D., Achard, P., 2004. Study of evaporative drying of treated silica gels. J. Non-Cryst. Solids 350: 230–237. Blacher, S., Heinrichs, B., Sahouli, B., Pirard, R., Pirard, J.-P., 2000. Fractal characterization of wide pore range catalysts: application to Pd-Ag/SiO2 xerogels. J. Colloid Interface Sci. 226: 123–130. Bock, V., Emmerling, A., Saliger, R., Fricke, J., 1997. Structural investigation of resorcinol formaldehyde and carbon aerogels using SAXS and BET. J. Porous Mater. 4: 287–294. Bogdanis, E., 2001. Modelling of heat and mass transport during drying of an elastic-viscoelastic medium and resolution by the finite element methods. Diss. University Pau, France. Boonamnuayvitaya, V., Tayamanon, C., Sae-ung, S., Tanthapanichakoon, W., 2006.

j223

Synthesis and characterization of porous media produced by a sol-gel method. Chem. Eng. Sci. 61: 1686–1691. Brinker, C. J., Scherer, G. W., 1990. Sol-gel science. Academic Press, Boston, USA. Brun, M., Lallemand, A., Quinson, J.-F., Eyraud, C., 1977. A new method for the simultaneous determination of the size and the shape of pores: the thermoporometry. Thermochim. Acta 21: 59–88. C
aceres, G., Bruneau, D., Jomaa, W., 2007. Two-phase shrinking porous media drying: a modeling approach including liquid pressure gradient effects. Drying Technol. 25: 1927–1934. Daughton, D. R., MacDonald, J., Mulders, N., 2003. Acoustic properties of silica aerogels between 400 mK and 400 K. J. Non-Cryst. Solids 319: 297–303. Degn Egeberg, E., Engell, J., 1989. Freeze-drying of silica gels prepared from siliciumethoxid. Rev. Phys. Appl. 24: C4–C23. Dollimore, D., Heal, G. R., 1964. An improved method for the calculation of pore size distribution from adsorption data. J. Appl. Chem. 14: 109–114. Einarsrud, M.-A., 1998. Light gels by conventional drying. J. Non-Cryst. Solids 225: 1–7. Escalona, I., Jomaa, W., Olivera-Fuentes, C., Crine, M., Blacher, S., L
eonard, A., 2008. Convective drying of gels: comparison between simulated and experimental moisture profiles obtained by X-ray microtomography. Proceedings of 16th International Drying Symposium (IDS2008), Hyderabad, India, Vol. C, 1209–1215. Farmer, J. C., Fix, D. V., Mack, G. V., Pekala, R. W., Poco, J. F., 1996. Capacitive deionization of NaCl and NaNO3 solutions with carbon aerogel electrodes. J. Electrochem. Soc. 143: 159–169. Farmer, J. C., Bahowick, S. M., Harrar, J. E., Fix, D. V., Martinelli, R. E., Vu, A. K., Caroll, K. L., 1997. Electrosorption of chromium ions on carbon aerogel electrodes as a means of remediating ground water. Energ. Fuel 11: 337–347.

j 5 Understanding and Preventing Structural Changes During Drying of Gels

224

Fricke, J., 1986. Aerogele. Phys. unserer Zeit 17(4): 101–106. Gan, L., Xu, Z., Feng, Y., Chen, L., 2005. Synthesis of alumina aerogels by ambient drying method and control of their structures. J. Porous Mater. 12: 317–321. Geis, S., M€ uller, B., Fricke, J., 2000. Dielectric constants of SiO2 and RF aerogels measured by a response function method. J. Porous Mater. 7: 423–433. Goldstein, J. I., Newbury, D. E., Echlin, P., Joy, D. C., Lyman, C. E., Lifshin, E., Sawyer, L., Michael, J. R., 2003. Scanning electron microscopy and X-ray microanalysis. 3rd edn, Kluwer Academic/Plenum Publishers, New York, USA. Hæreid, S., Dahle, M., Lima, S., Einarsrud, M.-A., 1995. Preparation and properties of monolithic silica aerogels from TEOS-based alcogels aged in silane solutions. J. Non-Cryst. Solids 186: 96–103. Heinrich, T., Klett, U., Fricke, J., 1995. Aerogels—Nanoporous materials, Part I: Sol-gel process and drying of gels. J. Porous Mater. 1: 7–17. Hrubesh, L. W., Poco, J. F., 1995. Thin aerogel films for optical, thermal, acoustic and electronic applications. J. Non-Cryst. Solids 188: 46–53. Hrubesh, L. W., 1998. Aerogel applications. J. Non-Cryst. Solids 225: 335–342. H€ using, N., Schubert, U., 2006. Aerogels, in Ullmann’s encyclopedia of industrial chemistry. Wiley-VCH, Weinheim, Germany. Hwang, S.-W., Jung, H.-H., Hyun, S.-H., Ahn, Y.-S., 2007. Effective preparation of crack-free silica aerogels via ambient drying. J. Sol-Gel Sci. Techn. 41: 139–146. Jensen, K. I., Schultz, J. M., Kristiansen, F. H., 2004. Development of windows based on highly insulating aerogel glazings. J. Non-Cryst. Solids 350: 351–357. Job, N., Pirard, R., Marien, J., Pirard, J.-P., 2004. Porous carbon xerogels with texture tailored by pH control using sol-gel process. Carbon 44: 619–628. Job, N., Thery, A., Pirard, R., Marien, J., Kocon, L., Rouzard, J.-N., Beguin, F., Pirard, J.-P., 2005. Carbon aerogels, cryogels and xerogels: influence of the drying method on the textural properties

of porous carbon materials. Carbon 43: 2481–2494. Job, N., Pirard, R., Pirard, J.-P., Ali
e, C., 2006a. Non intrusive mercury porosimetry: pyrolysis of resorcinol-formaldehyde xerogels. Part. Part. Syst. Charact. 23: 72–81. Job, N., Sabatier, F., Pirard, J.-P., Crine, M., L
eonard, A., 2006b. Towards the production of carbon xerogel monoliths by optimizing convective drying conditions. Carbon 44: 2534–2542. Job, N., Panariello, F., Marien, J., Crine, M., Pirard, J.-P., L
eonard, A., 2006c. Synthesis optimization of organic xerogels produced from convective air-drying of resorcinol-formaldehyde gels. J. Non-Cryst. Solids 352: 24–34. Jomaa, W., Puiggali, J.-R., 1991. Drying of shrinking materials: modellings with shrinkage velocity. Drying Technol. 9: 1271–1293. Jones, S. M., 2006. Aerogel: space exploration applications. J. Sol-Gel Sci. Techn. 40: 351–357. Kanamori, K., Aizawa, M., Nakanishi, K., Hanada, T., 2008. Elastic organic-inorganic hybrid aerogels and xerogels. J. Sol-Gel Sci. Techn. 48: 172–181. Kharzheev, Yu. N., 2008. Use of silica aerogels in Cherenkov counters. Phys. Part. Nuclei. 39: 107–135. Kharaghani, A., Metzger, T., Tsotsas, E., 2011. Discrete modeling approach to mechanical effects during isothermal drying. AIChE J. 51: 872–885. Kirkbir, F., Murata, H., Meyers, D., Chaudhuri, S. R., 1998a. Drying of aerogels in different solvents between atmospheric and supercritical pressures. J. Non-Cryst. Solids 225: 14–18. Kirkbir, F., Murata, H., Meyers, D., Chaudhuri, S. R., 1998b. Drying of large monolithic aerogels between atmospheric and supercritical pressures. J. Sol-Gel Sci. Techn. 13: 311–316. Kirsch, C., Metzger, T., Tsotsas, E., 2009. Towards a micro-scale simulation of convective gel drying. Proceedings of XII Polish Drying Symposium, Lodz, Poland, 223–234.

References Kistler, S. S., 1931. Coherent expanded aerogels and jellies. Nature 127: 741. Kocklenberg, R., Mathieu, B., Blacher, S., Pirard, R., Pirard, J.-P., Sobry, R., Van den Bossche, G., 1998. Texture control of freeze-dried resorcinolformaldehyde gels. J. Non-Cryst. Solids 225: 8–13. Kraiwattanawong, K., Mukai, S. R., Tamon, H., Lothongkum, A. W., 2007. Preparation of carbon cryogels from wattle tannin and furfural. Microporous and Mesoporous Materials 98: 258–266. Lecloux, A. J., 1981. Texture of catalysts, in Catalysis: Science and technology. Vol. 2 (eds J. R. Anderson, M. Boudart). Springer, Berlin, Germany, pp. 171–230. L
eonard, A., Job, N., Blacher, S., Pirard, J.-P., Crine, M., Jomaa, W., 2005a. Suitability of convective air drying for the production of resorcinol-formaldehyde and carbon xerogels. Carbon 43: 1808–1811. L
eonard, A., Job, N., Jomaa, W., Pirard, J.-P., Crine, M., Puigalli, J.-R., 2005b. Textural comparison of resorcinol-formaldehyde xerogels obtained from vacuum drying and convective drying. Proceedings of 7th World Congress of Chemical Engineering (WCCE7), Glasgow, Scotland, P4-009. L
eonard, A., Crine, M., Jomaa, W., 2006. Modelling of the convective drying of resorcinol-formaldehyde resins: influence of the drying conditions on the induced stress tensor. Proceedings of 15th International Drying Symposium (IDS2006), Budapest, Vol. A, 273–278. L
eonard, A., Crine, M., Jomaa, W., 2007. Production of cylindrical carbon xerogel monoliths: effect of sample diameter on drying-induced cracks. Extended Abstracts of 6th European Congress of Chemical Engineering (ECCE6), Copenhagen, Denmark, II (T2-7b). L
eonard, A., Blacher, S., Crine, M., Jomaa, W., 2008. Evolution of mechanical properties and final textural properties of resorcinolformaldehyde xerogels during ambient air drying. J. Non-Cryst. Solids 354: 831–838. Leventis, N., Palczer, A., McCorkle, L., Zhang, G., Sotiriou-Leventis, C., 2005. Nanoengineered silica-polymer composite aerogels with no need for supercritical fluid drying. J. Sol-Gel Sci. Techn. 35: 99–105.

j225

Lin, C., Ritter, J. A., 1997. Effect of synthesis pH on the structure of carbon xerogels. Carbon 35: 1271–1278. Lippens, B. C., de Boer, J. H., 1965. Studies on pore systems in catalysis, Part 4: The tmethod. J. Catal. 4: 319–323. Lu, X., Wang, P., Arduini-Schuster, M.-C., Kuhn, J., B€ uttner, D., Nilsson, O., Heinemann, U., Fricke, J., 1992. Thermal transport in organic and opacified silica monolithic aerogels. Proceedings of 3rd International Symposium on Aerogels, J. Non-Cryst. Solids 145: 207–210. Lubda, D., Lindner, W., Quaglia, M., du Fresne von Hohenesche, C., Unger, K. K., 2005. Comprehensive pore structure characterization of silica monoliths with controlled mesopore size and macropore size by nitrogen sorption, mercury porosimetry, transmission electron microscopy and inverse size exclusion chromatography. J. Chromatography A 1083: 14–22. Masmoudi, Y., 2006. Etude du s
echage au CO2 supercritique pour l’
elaboration de mat
eriaux nanostructur
es: application aux aerogels de silice monolithiques. Diss, Ecole des Mines de Paris, France. Masmoudi, Y., Rigacci, A., Ilbizian, P., Cauneau, F., Achard, P., 2006. Diffusion during the supercritical drying of silica gels. Drying Technol. 24: 1121–1125. Mathieu, B., Blacher, S., Pirard, R., Pirard, J.-P., Sahouli, B., Brouers, F., 1997. Freeze-dried resorcinol-formaldehyde gels. J. Non-Cryst. Solids 212: 250–261. Mayor, L., Sereno, A. M., 2004. Modelling shrinkage during convective drying of food materials: a review. J. Food Eng. 61: 373–386. Mukai, S. R., Nishihara, H., Tamon, H., 2008. Morphology maps of ice-templated silica gels derived from silica hydrogels and hydrosols. Micropor. Mesopor. Mater. 116: 166–170. Nishihara, H., Mukai, S. R., Yamashita, D., Tamon, H., 2005. Ordered macroporous silica by ice templating. Chem. Mater. 17: 683–689. Nunez, O., Nakanishi, K., Tanaka, N., 2008. Preparation of monolithic silica

j 5 Understanding and Preventing Structural Changes During Drying of Gels

226

columns for high-performance liquid chromatography. J. Chromatogr. A 1191: 231–252. Orlovic, A., Petrovic, S., Skala, D., 2005. Mathematical modeling and simulation of gel drying with supercritical carbon dioxide. J. Serb. Chem. Soc. 70: 125–136. Pajonk, G. M., Repellin-Lacroix, M., Abouarnadasse, S., Chaouki, J., 1990. From sol-gel to aerogels and cryogels. J. Non-Cryst. Solids 121: 66–67. Pakowski, Z., Glebowski, M., Adamski, R., 2006. Modeling of drying of highly shrinking materials using hydrogels as an example. Drying Technol. 24: 1075–1081. Pekala, R. W., 1989. Organic aerogels from the polycondensation of resorcinol with formaldehyde. J. Mater. Sci. 24: 3221–3227. Phalippou, J., Woignier, T., Prassas, M., 1990. Glasses from aerogels, Part I: The synthesis of monolithic silica aerogels. J. Mater. Sci. 25: 3111–3117. Pirard, R., Blacher, S., Brouers, F., Pirard, J.-P., 1995. Interpretation of mercury porosimetry applied to aerogels. J. Mater. Res. 10: 2114–2119. Pirard, R., Heinrichs, B., van Cantfort, O., Pirard, J.-P., 1998. Mercury porosimetry applied to low density xerogels: relation between structure and mechanical properties. J. Sol-Gel Sci. Techn. 13: 335–339. Pirard, R., Ali
e, C., Pirard, J.-P., 2005. Specific behavior of sol-gel materials in mercury porosimetry: collapse and intrusion, in Handbook of sol-gel science and technology, Vol. 2: Characterization of sol-gel materials and products (eds S. Sakka, R. M. Almeida). Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 211–233. Pourcel, F., Jomaa, W., Puiggali, J.-R., Rouleau, L., 2007a. Criterion for crack initiation during drying: alumina porous ceramic strength improvement. Powder Technol. 172: 120–127. Pourcel, F., Jomaa, W., Puiggali, J.-R., Rouleau, L., 2007b. Crack appearance during drying of an alumina gel: thermo-hydromechanical properties. Drying Technol. 25: 759–766.

Rao, A. P., Rao, A. V., Pajonk, G. M., 2007. Hydrophobic and physical properties of the ambient pressure dried silica aerogels with sodium silicate precursor using various surface modification agents. Appl. Surf. Sci. 253: 6032–6040. Reichenauer, G., Scherer, G. W., 2000. Nitrogen adsorption in compliant materials. J. Non-Cryst. Solids 277: 162–172. Reichenauer, G., Scherer, G. W., 2001a. Effects upon nitrogen sorption analysis in aerogels. J. Colloid Interface Sci. 236: 385–386. Reichenauer, G., Scherer, G. W., 2001b. Extracting the pore size distribution of compliant materials from nitrogen adsorption. Colloid. Surf. A 187–188: 41–50. Reichenauer, G., Pfrang, T., Hofmann, M., 2007. Drying of meso- and macroporous gels—length change and drying dynamics. Colloid. Surf. A 300: 211–215. Reimer, L., Kohl, H., 2008. Transmission electron microscopy. 5th edn, Springer, New York,USA. Roe, R. J., 2000. Methods of x-ray and neutron scattering in polymer science. Oxford University Press, Oxford, UK. Roig, A., Mata, I., Molins, E., Miravitlles, C., Torras, J., Llibre, J., 1998. Silica aerogels by supercritical extraction. J. Eur. Ceram. Soc. 18: 1141–1143. Rottereau, M., Gimel, J.-C., Nicolai, T., Durand, D., 2004. Monte Carlo simulation of particle aggregation and gelation, Part 1: Growth, structure and sized distribution of the clusters. Eur. Phys. J. E 15: 133–140. Sakka, S. (ed.) 2005. Handbook of sol-gel science and technology, Vol. 3: Applications of sol-gel technology. Kluwer Academic Publishers, Dordrecht, The Netherlands. Sakka, S., Kozuka, H. (eds) 2005. Handbook of sol-gel science and technology, Vol. 1: Sol-gel processing. Kluwer Academic Publishers, Dordrecht, The Netherlands. Schaefer, D. W., Martin, J. E., Wiltzius, P., Cannell, D. S., 1984. Fractal geometry of colloidal aggregates. Phys. Rev. Lett. 52: 2371–2374. Scherer, G. W., 1986. Drying gels, Part I: General theory. J. Non-Cryst. Solids 87: 199–225. Scherer, G. W., 1987a. Correction of “Drying gels, Part I: General theory”. J. Non-Cryst. Solids 92: 375–382.

References Scherer, G. W., 1987b. Drying gels, Part 2: Film and flat plate. J. Non-Crystalline Solids 89: 217–238. Scherer, G. W., 1987c. Drying gels, Part 3: Warping plate. J. Non-Cryst. Solids 91: 83–100. Scherer, G. W., 1987d. Drying gels, Part 4: Rigid Gels. J. Non-Cryst. Solids 92: 122–144. Scherer, G. W., 1988. Drying gels, Part 5: . Viscoelastic plate. J. Non-Cryst. Solids 99: 324–358. Scherer, G. W., 1989. Drying gels, Part 8: Revision and review. J. Non-Cryst. Solids 109: 171–182. Scherer, G. W., 1990a. Stress and fracture during drying of gels. J. Non-Cryst. Solids 121: 104–109. Scherer, G. W., 1990b. Theory of drying. J. Am. Ceram. Soc. 73: 3–14. Scherer, G. W., 1992a. Recent progress in drying of gels. J. Non-Cryst. Solids 147–148: 363–374. Scherer, G. W., 1992b. Bending of gel beams: method of characterizing mechanical properties and permeability. J. Non-Cryst. Solids 142: 18–35. Scherer, G. W., 1992c. Stress development during supercritical drying. J. Non-Cryst. Solids 145: 33–44. Scherer, G. W., 1993. Freezing gels. J. NonCryst. Solids 155: 1–25. Scherer, G. W., 1994. Stress in aerogel during depressurization of autoclave, Part 1: Theory. J. Sol-Gel Sci. Techn. 3: 127–139. Scherer, G. W., Smith, D. M., Stein, D., 1995a. Deformation of aerogels during characterization. J. Non-Cryst. Solids 186: 309–315. Scherer, G. W., Smith, D. M., Qiu, X., Anderson, J. M., 1995b. Compression of aerogels. J. Non-Cryst. Solids 186: 316–320. Scherer, G. W., Haereid, S., Nilsen, E., Einarsrud, M.-A., 1996. Shrinkage of silica gels aged in TEOS. J. Non-Cryst. Solids 202: 42–52. Scherer, G. W., 1999. Structure and properties of gels. Cement Concrete Res. 29: 1149–1157. Scheuerpflug, P., Caps, R., B€ uttner, D., Fricke, J., 1985. Apparent thermal conductivity of evacuated SiO2-aerogel tiles under variation of radiative boundary conditions. Int. J. Heat Mass Transfer 28: 2299–2306.

j227

Schmidt, M., Schwertfeger, R., 1998. Applications of silica aerogel products. J. Non-Cryst. Solids 225: 364–368. Schneider, G. M., 1978. Physicochemical principles of extraction with supercritical gases. Angew. Chem. Int. Ed. Engl. 17: 716–727. Schwertfeger, R., Frank, D., Schmidt, M., 1998. Hydrophobic waterglass based aerogels without solvent exchange or supercritical drying. J. Non-Cryst. Solids 225: 24–29. Shaw, T. M., 1987. Drying as an immiscible displacement process with fluid counterflow. Phys. Rev. Lett. 59: 1671–1675. Shewale, P. M., Rao, A. V., Rao, A. P., 2008. Effect of different trimethyl silylating agents on the hydrophobic and physical properties of silica aerogels. Appl. Surf. Sci. 254: 6902–6907. Simpkins, P. G., Johnson, D. W., Fleming, D. A., 1989. Drying behavior of colloidal silica gels. J. Am. Ceram. Soc. 72: 1816–1821. Sinko, K., Torma, V., Kovacs, A., 2008. SAXS investigation of porous nanostructures. J. Non-Cryst. Solids 354: 5466–5474. Smith, D. M., Scherer, G. W., Anderson, J. M., 1995a. Shrinkage during drying of silica gel. J. Non-Cryst. Solids 188: 191–206. Smith, D. M., Stein, D., Anderson, J. M., Ackerman, W., 1995b. Preparation of lowdensity xerogels at ambient pressure. J. NonCryst. Solids 186: 104–112. Stepanek, F., Marek, M., Adler, P. M., 1999. Modeling capillary condensation hysteresis cycles in reconstructed porous media. AIChE J. 45: 1901–1912. Strøm, R. A., Masmoudi, Y., Rigacci, A., Petermann, G., Gullberg, L., Chevalier, B., Einarsrud, M.-A., 2007. Strengthening and aging of wet silica gels for up-scaling of aerogel preparation. J. Sol-Gel Sci. Techn. 41: 291–298. Tamon, H., Ishizaka, H., 1998. SAXS study on gelation process in preparation of resorcinol-formaldehyde aerogel. J. Colloid Interface Sci. 206: 577–582. Tamon, H., Ishizaka, H., Araki, T., Okazaki, M., 1998. Control of mesoporous structure of organic and carbon aerogels. Carbon 36: 1257–1262.

j 5 Understanding and Preventing Structural Changes During Drying of Gels

228

Tamon, H., Ishizaka, H., 1999. Preparation of organic mesoporous gel by supercritical/ freeze-drying. Drying Technol. 17: 1653–1665. Tamon, H., Ishizaka, H., Yamamoto, T., Suzuki, T., 1999. Preparation of mesoporous carbon by freeze-drying. Carbon 37: 2049–2055. Tamon, H., Ishizaka, H., Yamamoto, T., Suzuki, T., 2000. Influence of freeze-drying conditions on the mesoporosity of organic gels as carbon precursors. Carbon 38: 1099–1105. Tamon, H., Mukai, S. R., Nishihara, H., Yoshida, T., Yamamoto, T., Tonanon, N., Tanthapanichakoon, W., 2003. Microwave drying for synthesis of mesoporous carbon gels. Proceedings of 3rd Asia-Pacific Drying Conference (ADC2003), Bankok, Thailand, 123–132. Tamon, H., Kawahara, M., Kraiwattanawong, K., Nishihara, H., Mukai, S. R., 2006a. Preparation of carbon cryogel by sol-gel polycondensation of wattle tannin with formaldehyde. Proceedings of 15th International Drying Symposium (IDS2006), Budapest, Hungary, Vol. C, 1859–1866. Tamon, H., Nishihara, H., Mukai, S. R., 2006b. Synthesis of titania-silica microhoneycomb by unidirectional freezing and freeze-drying of hydrogel. Proceedings of 15th International Drying Symposium (IDS2006), Budapest, Hungary, Vol. C, 1817–1824. Tewari, P. H., Hunt, A. J., Lofftus, K. D., 1985. Ambient-temperature supercritical drying of transparent silica aerogels. Mater. Lett. 3: 363–367. Tillotson, T. M., Hrubesh, L. W., 1992. Transparent ultralow-density silica aerogels prepared by a two-step sol-gel process. J. Non-Cryst. Solids 145: 44–50. Tonanon, N., Siyasukh, A., Tanthapanichakoon, W., Nishihara, H., Mukai, S. R., Tamon, H., 2005. Improvement of mesoporosity of carbon cryogels by ultrasonic irradiation. Carbon 43: 525–531. Tonanon, N., Wareenin, Y., Siyasukh, A., Tanthapanichakoon, W., Nishihara, H., Mukai, S. R., Tamon, H., 2006. Preparation of resorcinol formaldehyde (RF) carbon gels:

use of ultrasonic irradiation followed by microwave drying. J. Non-Cryst. Solids 352: 5683–5686. van Bommel, M. J., de Haan, A. B., 1994. Drying of silica gels with supercritical carbon dioxide. J. Mater. Sci. 29: 943–948. Vollet, D. R., Scalari, J. P., Donatti, D. A., Ruiz, A. I., 2008. A thermoporometry and small-angle x-ray scattering study of wet silica sonogels as the pore volume fraction is varied. J. Phys. Condens. Matter 20: 025225. Vorhauer, N., Metzger, T., Tsotsas, E., 2009. Empirical macroscopic model for drying of porous media based on pore networks and scaling theory. Drying Technol. 28: 991–1000. Wang, P., K€orner, W., Emmerling, A., Beck, A., Kuhn, J., Fricke, J., 1992. Optical investigations of silica aerogels. J. Non-Cryst. Solids 145: 141–145. Wiener, M., Reichenauer, G., Scherb, T., Fricke, J., 2004. Accelerating the synthesis of carbon aerogel precursors. J. Non-Cryst. Solids 350: 126–130. Woignier, T., Scherer, G. W., Alaoui, A., 1994. Stress in aerogel during depressurization of autoclave, Part 2: Silica gels. J. Sol-Gel Sci. Techn. 3: 141–150. Woignier, T., Phalippou, J., Despetis, F., Etienne, P., Alaoui, A., Duffours, L., 2005. Mechanical properties of gels: from alcogel and aerogels to glasses. Ch. 2, in Handbook of sol-gel science and technology, Vol. 2: Characterization and properties of sol-gel materials and products (eds S. Sakka, R. M. Almeida). Kluwer Academic Publishers, Dordrecht, The Netherlands. Wu, D., Fu, R., 2008. Requirements of organic gels for a successful ambient pressure drying preparation of carbon aerogels. J. Porous Mater. 15: 29–34. Yamamoto, T., Nishimura, T., Suzuki, T., Tamon, H., 2001a. Control of mesoporosity of carbon gels prepared by sol-gel polycondensation and freeze-drying. J Non-Cryst. Solids 288: 46–55. Yamamoto, T., Nishimura, T., Suzuki, T., Tamon, H., 2001b. Effect of drying method on mesoporosity of resorcinol-formaldehyde drygel and carbon gel. Drying Technol. 19: 1319–1333.

References Yamamoto, T., Endo, A., Ohmori, T., Nakaiwa, M., Mukai, S. R., Tamon, H., 2005. Effect of drying method on gas adsorption characteristics of carbon gel microspheres. Drying Technol. 23: 2119–2129.

j229

Zubizarreta, L., Arenillas, A., Menendez, J. A., Pis, J. J., Pirard, J.-P., Job, N., 2008. Microwave drying as an effective method to obtain porous carbon xerogels. J. Non-Cryst. Solids 354: 4024–4026.

j231

6 Morphology and Properties of Spray-Dried Particles Peter Walzel and Takeshi Furuta 6.1 Introduction

Spray drying is, by definition, the transformation of a feed of liquid (solution or dispersion) or paste material into a dried particulate powder. The feed liquid is atomized into small droplets of several dozens to a few hundreds mm, which are dried to the powder by contacting with a hot drying medium (Masters, 1991). The volume mean particle size of the powder is typically in the range 10 mm < dV,50 < 1 mm, with mean particle sizes of 50 mm < dV,50 < 300 mm being more frequent in practice. One advantage of spray drying over other drying methods is that the process leads directly and in a fairly simple way from a liquid feed to solid particles. Another crucial advantage is that drying is very rapid and can be completed in a very short time. Drying times between 5 and 30 s are typical for sprays with a droplet diameter of 10–200 mm according to Furuta et al. (1994). Furthermore, it is possible to form spherical powder particles by spray drying, and it is also possible to transport the product powder from the drying vessel to the storage bins pneumatically, that means by using the drying gas (typically air). Spray drying plants commonly consist of a feed pump, an atomizer, an air heater, an air distributor, a drying chamber, and systems for exhaust air cleaning and powder recovery, as shown in Fig. 6.1. Atomization is the key technology in spray drying, because it determines the droplet size distribution, which – in its turn – has a great influence on both the size of the final powder and the kinetics of the process. Atomization is commonly performed by means of spray nozzles or centrifugal disk atomizers. The centrifugal rotary disk atomizer is favorably applied to highly viscous liquids and slurry feeds. It has the advantage of enabling easy control of the droplet diameter by varying the rotational speed of the disk. Spray nozzles may operate with one fluid (the pressurized liquid feed) or two fluids (the liquid feed and air). Two-fluid nozzles are an alternative for the atomization of highly viscous fluids, and a very good candidate when aiming at the production of very small (even submicron) particles. An overview of recent developments is given, for example, in Walzel (2001). Very small particles can also be produced by means of ultrasonic atomizers. In any case,

Modern Drying Technology Volume 3: Product Quality and Formulation, First Edition. Edited by Evangelos Tsotsas and Arun S. Mujumdar. Ó 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.

j 6 Morphology and Properties of Spray-Dried Particles

232

Fig. 6.1 Typical spray-drying system consisting of a feed pump, an atomizer, an air heater, an air dispenser, a drying chamber, and equipment for exhaust air cleaning and powder recovery.

atomization of the liquid feed to more or less fine droplets is the main reason for the mentioned fast kinetics of spray drying, due to a dramatic increase in surface area; (transformation of one large drop of 1 cm in diameter to many droplets of 100 mm increases the surface area by a factor of 100, and transformation to fine droplets of 1 mm increases it by a factor of 10 000). The spray drying process is frequently operated concurrently regarding the gas and the spray flow, as depicted in Fig. 6.1. Under these conditions the thermal stress on the particles is low and even sensitive materials such as food, pharmaceuticals, cosmetics and other high value products can be dried. In other cases the thermal stress may be of little relevance, for example, when drying thermally resistant detergents or ceramics. Then, a partly countercurrent flow arrangement or a fountain-like spray pattern or even completely countercurrent arrangements can also be applied. Countercurrent flow between the gas and the droplets usually leads to significant particle collisions and agglomeration, and thus to larger particles. During the spray drying process, the solution droplet may shrink due to the evaporation of solvent, which is most often water. By decreasing the water content and water activity, spray drying is generally used in the food industry to ensure the microbiological stability of products, avoid the risk of chemical and/or biological degradations, reduce the storage and transport costs, and, finally, obtain a product with specific properties such as instantaneous solubility. Typical spray-dried food products are listed in Tab. 6.1. It should be mentioned that over 80% of the

6.1 Introduction Tab. 6.1

Typical spray-dried food products.

Type of material

Ingredients

Flavoring agents Vitamins Minerals Oils and fats Herbs and bioactives Other

Oil, spices, seasonings, sweeteners Vitamin E, b-carotene, ascorbic acids Calcium, magnesium and phosphorus Fish oils (DHA, EPA), sea buckthorn oil, lycopene Creatine, probiotic bacteria Enzymes, leavening agents, psyllium, yeast

manufacture of food flavor powders involves spray drying. When the water content of the droplet reaches a critical value, a dry crust is frequently formed at the droplet surface. The morphology of the particles depends on the inlet and outlet air temperature. Commonly, if the droplet is dried very slowly, it is transformed into a fully dried particle. At high drying temperature, however, cycles of repeated expansion and collapse of the particle occur, due to the formation of an internal air bubble. The internal and surface morphologies of spray-dried powders affect the powder flowability, re-dispersability, density, and the stability of the encapsulated active ingredients. The vacuole and surface roughness of spray-dried powders are important morphological characteristics with an influence on powder flowability. The quality factors of the particles are a direct result of the design and operating conditions of the spray dryer, such as the type of atomizer, drying air temperature, feed rate, and viscosity or nature of the solid matrix. The dissolved or dispersed solid concentration and the additives in the feed liquid, such as sugar, lipids, polymers, and proteins, also affect the morphology (i.e., size and degree of hollowness) of spraydried powders (Hadinoto et al., 2007). The drying process of liquid droplets containing solid matter generates very different shapes and structures of the finally dried solid particles, as already described in Masters (1991). It is obvious that the structure of the solids depends on both the solid material, and the operational parameters such as air and liquid temperature and droplet size. Singe droplets in a large air volume may dry at a different rate compared to droplets incorporated within a dense spray, even at the same ambient air temperature. This is due to the decrease in the driving forces when many particles a short distance apart release their vapor and thus reduce the overall driving force. If this is the case, the shape of the spray jet may also influence the drying process, even at the same air inlet temperature (see Chapter 5 in Volume 1 of this series). The outlet gas temperature and outlet air humidity are also good indicators for the driving potential of the total drying process as well as the inlet air temperature. Furthermore, the product is usually discharged with a temperature fairly close to the outlet air or gas temperature. Only in the case of countercurrent operations do significant differences appear between air outlet temperature and the discharge temperature of the product. Chapter 5 of Volume 1 described in detail how the flow patterns of hot air and the trajectories of the drying droplets can be analyzed by computational fluid dynamics

j233

j 6 Morphology and Properties of Spray-Dried Particles

234

(CFD) to improve and optimize the drying system. In the present chapter, the morphology of spray-dried particles is examined as a function of the outlet air temperature and additives, such as surfactants and proteins. Scanning electron microscopy (SEM) and confocal laser scanning microscopy (CLSM) are used for the investigation of the outside and inside morphology of the spray-dried particles.

6.2 Morphology of Spray-Dried Particles 6.2.1 Classification of the Morphology of Spray-Dried Powders

Extensive studies of the particle morphology during drying have been performed by Walton and Mumford (1999a, b) and Walton (2000) by means of a single suspended droplet drying method. They classified the typical particle formation processes of a droplet containing dissolved solids, as illustrated schematically in Fig. 6.2. In the early stages of drying, the droplet has a free liquid surface, where

Fig. 6.2 Scheme for the particle morphology of skin-forming materials (modified from Walton (2000)).

6.2 Morphology of Spray-Dried Particles

water evaporates rapidly. The depletion of water content at the surface will cause the solute to be more concentrated – this will depend on the speed of evaporation and the rate at which the water can be replenished from the interior of the droplet. Because of the increase in concentration, solids may precipitate out of the solution at the surface of the droplet first, leading to the formation of a crust or skin around a hollow particle. The thickness of the crust will depend on the drying rate – high initial drying rates will lead to larger particles with thin shells and low density, whereas low initial drying rates will lead to smaller particles with thick shells and high density. Another possible reason for the formation of hollow, high-porosity particles is the presence of occluded or absorbed gases in the feed liquid, which may lead to bubble formation, even below the boiling point. Once the particle has formed, it may remain intact or may fracture due to internal pressure. If the skin is pliable, then, depending on the drying conditions, it may puff up or collapse and shrivel. In the case of low air temperature, drying is slower during the surface evaporation period, so that the particle decreases in size and forms a thicker skin. In the falling rate period, bubbles may appear by nucleation due to temperature increase. Subsequent expansion and collapse of bubbles result in hollow particles or shriveled particles. However, if the nucleation is suppressed, the droplet will be dried to a dense solid particle. On the other hand, if the process takes place at high ambient or air temperature, repeated cycles of expansion and collapse of internal bubbles may occur after the skin formation, resulting in the various types of particles shown in Figs. 6.6 and 6.11 to 6.16. Mechanisms of bubble formation and expansion were discussed by Greenwald (1980) who concludes that a two step mechanism is operative: .

.

First, an air bubble is created by desorption of air which is either present in the feed liquid or absorbs shortly below the atomizer (Greenwald and King, 1981, 1982). Alternatively, such an air bubble may be incorporated during atomization (Verhey, 1972a, b, 1973). The second step needed to bring about expansion is that of the drop temperature approaching or exceeding the boiling point, whereupon a large amount of water vapor will be formed in the bubble, causing it to grow substantially in size.

The largest effect on particle morphology is, however, given by the composition and solid materials contained in the liquid to be sprayed (Walton, 2000; Jørgensen, 2005). Several different options exist concerning the composition of the slurry: . . . . .

solution of low molecular weight substances (crystallization and crusts) solution of high molecular weight substances such as polymers (skin forming) suspension of (insoluble) small solid particles, d < 2 mm (crust forming) suspension containing large solid particles d > 5 mm complex dispersions such as, for example, emulsions and other formulations.

The formation of different morphologies can either be studied in large scale spray dryers or in specially designed spray dryer pilot plants, see for example Zbicinski

j235

j 6 Morphology and Properties of Spray-Dried Particles

236

et al. (2002a, b), Jørgensen (2005). A further option is to observe the drying process within a levitator (Toei et al., 1978; Kastner et al., 2001; Brenn et al., 2001; Groenewold et al., 2002). This device allows the suspending of particles at a stable position within the gas environment. These particles have a size of typically d > 300 mm and can be dried on a lab scale under distinct and defined conditions. Another option is the drying of droplets suspended on a wire which also gives some insight into the structure-forming process (Hecht and King, 2000a; Dicoi and Walzel, 1998). The particle size, d, can be calculated from the diameter of the spray droplets, dd, and the solid mass fraction of the solution or suspension, xs, according to the relationship 

1=3 1
xs rs d ¼ dd ð1
eÞ þ1 xs rl

ð6:1Þ

Here, rs is the solid density and rl the liquid density. Due to the small magnitude of the exponent, only strong variations in concentration, densities and porosity lead to significant differences between particle size and droplet size. However, the overall porosity, e, of the final spray-borne particle may indeed be strongly affected by the formation of solid structures during the drying process. It is not known in advance and so far can only be determined experimentally, which also holds for particle density. Particle porosity lies in the relatively narrow range of 0.5 < e < 0.6 only for relatively large and rather compact particles that exhibit approximately uniform distribution of voids throughout their whole cross-section. Typical examples of particle structures as formed by the drying process and depending on the solid matter contained in the liquid feed will be given in the following. 6.2.2 Solutions of Low Molecular Weight Substances

When spraying solutions of substances of low molecular weight, a first drying step occurs above the solubility limit of the solute, and the drop reduces in size until the solubility limit is reached. For this first step, which is usually called the first or constant rate period (CRP), the drying velocity, that is, the solvent mass evaporated per unit time, is given by the vapor pressure of the solution at the drop surface, pv,S, and the vapor pressure in the vicinity of the particle, pv,1. The vapor pressure at the surface depends – assuming a well-mixed state within the droplet – on the drop temperature and on the water activity within the solution. The surface temperature remains low in the CRP as the solvent, due to its heat of evaporation Dhv, uses up the sensible heat (expressed by the specific heat capacity cP,v of the air–vapor mixture) transferred to the particle by the gas in a hot atmosphere. The particle surface temperature TS is more or less close to the wet bulb temperature of pure water, depending on the water activity in the case of dissolved matter (see Eq. 5.44 in Volume 1 of this series). The dependence of the vapor pressure of the solvent on the surface temperature TS may be expressed by the Antoine or Clausius–Clapeyron equation, as

6.2 Morphology of Spray-Dried Particles

in Volume 1 of this series, Eq. 5.36. The surface temperature TS is also linked to the mass transfer coefficient k, the heat transfer coefficient a, the total gas pressure P, and the gas temperature T1 at a sufficiently large distance from the surface according to the relationship a

  ~ v P P
pv;1 Dhv cP;v Dhv M ln ln 1 þ ðT1
TS Þ ¼ k ~ cP;v Dhv P
pv;S RTm

ð6:2Þ

(see Bird et al. (2002), Jørgensen (2005), Gnielinski et al. (1993), Abramzon and Sirignano (1989), and Volume 1 of this series). Here, Tm is the mean temperature between the particle and the air. As soon as the solubility limit is reached at the drop temperature, crystallization of the solute may set in. Supersaturation is due to evaporation, and, therefore, highest close to the surface of the droplet, where the evaporation takes place. Hence, crystallization will start from there. Depending on the nucleation properties of the solute, more or less nuclei are formed at the surface until a solid layer of crystals builds a crust. Depending on the permeability of the crust, solution can be transported through small capillaries between the crystals whereas the surrounding gas can enter through large capillaries to compensate the low internal pressure. As a result, hollow beads are formed consisting of a polycrystalline shell. Some basic considerations help us to understand the behavior of solid phase formation and the morphology of particles. High solute and air temperature usually mean higher solubility and higher mass content before the onset of crystallization. (However, some materials do not show increased solubility with increased temperature, as e.g., NaCl). High solubility also means a late onset of crystallization and low solvent content. Due to this fact, particles should form a shell with a thick wall. Materials with low solubility, but still sprayed into the dryer as a solution, are expected to form a thinner shell. However, the nucleation properties, such as nucleation rates and the growth rates of crystals, also play a major role. Some materials do not form crystals but solidify in an amorphous or glassy state. A typical example is lactose, for example, sprayed as a 20 wt% solution into air at an air exit temperature of 90  C, as shown in Fig. 6.3. Some materials exhibit a rubbery state in the case of too high temperatures or water content. Such materials are difficult to dry due to their sticky surface. For this reason it is necessary to know the glass transition temperature, Tg, below which the surface becomes hard enough to avoid lumping. The glass transition also depends strongly on the water content (Gordon and Taylor, 1952). A typical example is maltodextrine, see, for example, Br€ockel et al. (2007). Single drop measurements of drying kinetics and stickiness have been presented for different hydrocarbon solutions by Adhikari et al. (2003). Acid-rich foods such as fruits, juices and vegetable juices, and sugar-rich components belong to the group of sticky products with low Tg. These materials are very difficult to dry in normal spray drying equipment due to the requirement of low temperature and the high residence time resulting from the low driving force. In foods, starch, such as maltodextrine, is very often added as an anti-sticking agent and drying aid; (for the glass transition temperature of maltodextrine see also Kilburn et al. (2005)). Freeze spray drying, which will be briefly discussed in Section 6.2.7, is

j237

j 6 Morphology and Properties of Spray-Dried Particles

238

Fig. 6.3 Glassy lactose particles obtained from a 20 wt% aqueous solution at an outlet air temperature of 90  C; Mean particle size: d ¼ 65 mm. The transparency of the particles is a clear indicator for the absence of larger crystals.

an option when no additives are allowed to modify the surface properties of the particles. However, freeze spray drying is more costly. When comparing small and large drops under the same driving force conditions, the drying of smaller drops is faster due to the higher mass and heat transfer. This affects the nucleation and growth of crystals, as the supersaturation increases with decreasing drop size. There are several competing timescales, the most important being those for mass transfer, nucleation and the growth rate of the crystals. Nucleation and growth rate are different for different materials (Gnielinski et al., 1993). This is one reason for the different morphologies of small and large particles. Another reason is the different trajectories of large and small particles within the dryer. Small particles may recirculate in a hot air stream while large particles escape more easily from turbulent eddies and follow the influence of gravity to a much larger extent. In Brenn et al. (2001) a proposal is presented for distinguishing between the formation of hollow beads and compact particles obtained from crystallizing substances by means of the parameter G¼

Dl rl 1 Dg rg Sh* lnð1 þ BM Þ

ð6:3Þ

The model was validated on NaCl solutions. In Eq. 6.3, Dl is the diffusion coefficient of the solute in the liquid phase, Dg is the diffusion coefficient of the solvent vapor in the gas phase, rl and rg are the liquid and gas densities, respectively. Sh is the dimensionless mass transfer rate in the vapor phase. This modified Sherwood number, that accounts for the film thinning effect of Stefan flow, lies typically in the range 2 < Sh < 5. The quantity BM is the Spalding transfer number according to Abramzon and Sirignano (1989) (Sirignano (1999), compare with Eq. 1.66 in Volume

6.2 Morphology of Spray-Dried Particles

Fig. 6.4 Example of a simple crystallizing material (NaCl particles after spray drying from 20 wt% solution). No significant difference in the particle shape was found at air inlet

temperatures of 140 up to 220  C. Cubic and hemispherical-shaped particles are visible. Some particles show a void space in their center.

2 of this series). When G > 3.3 solid particles are formed, whereas hollow particles are expected for G < 3.3. The latter takes place when the gas-side mass transfer rate is high and when the diffusion rate of the solvent in the droplet is comparatively small. Figure 6.4 shows particles formed from 25 wt% NaCl solution in water. Only a few particles exhibit hollow structures as G was in the range 5 < G < 6. Many materials such as salts, for example, Na2SO4, show different stable modifications for given temperatures with different content of crystal water. Usually, a stable modification is formed first, at the given drop or particle temperature. The hygroscopic property of the product may then be the result of a subsequent transformation into a different modification stable at ambient temperature and humidity. As the lattice of the crystals is restructured during uptake of water from the ambient air, the strength of the particle may diminish and dust can be formed. The surface structure of spray-dried particles can be regarded as a key for fine particle adhesion in customer applications. This effect is needed for example, for inhalation therapies when fine active particles with d < 5 mm are added to spray-dried carrier particles within the size range 60–100 mm in order to obtain a dose unit with reasonable flowability and with proper detachment properties of the fines. Experiments with mannitol (Maas et al., 2009) gave rise to quite different surface morphologies when sprayed into air at different temperatures. In the case of low air exhaust temperatures, T ¼ 80  C, the carrier particles exhibit a fairly smooth surface. At an exhaust air temperature of T ¼ 130  C, the particles show rough surfaces formed by crystals covering the particle surface, see Fig. 6.5. In this case, the air temperature is a vehicle to adjust the desired adhesion properties.

j239

j 6 Morphology and Properties of Spray-Dried Particles

240

Fig. 6.5 SEM of mannitol particles obtained from 15 wt% solution at a dryer exhaust temperature of (a) 80  C and (b) 130  C.

6.2.3 Solutions of Polymers

When spraying solutions of polymers, the spraying process is usually limited to viscosities less than 0.5 Pa s under spraying conditions. Higher viscosities give rise to the formation of fiber-particle mixtures and deformed spheres. Finally, at m > 10 Pa s only fibers are formed (Lohner et al., 2005; Eggers and Villermaux, 2008). The same behavior can be observed when highly viscous melts are sprayed, see for example, Walz and Mayer (1966). The vapor pressure of the solvents in the presence of large solute molecules is reduced during the drying process. The solvent activity can then be described by the Flory–Huggins equation (Flory, 1942; Shinoda, 1978; Vrentas and Vrentas, 1996). Besides phase equilibria, the diffusion of the solvent is usually significantly reduced, decreasing strongly with increasing solids content, especially close to the particle surface. The reduced penetration of the solvent through the resulting shell leads to significantly reduced evaporation, and to a considerable rise in the particle temperature. While a thin skin-like shell is formed on the surface, the inner part of the droplet remains in a liquid state. As the cooling effect of the evaporating solvent ceases, the droplet reaches the air temperature, mostly higher than the boiling point of the solvent. Bubbles appear in the droplet and the vapor finally penetrates the shell, forming a blowhole. As a result, particles show indentations and sometimes an irregular structure. This process may be repeated several times, so that the shape of the particles becomes more and more irregular, as shown in Figs. 6.6 and 6.7 with styrene–butadiene–styrene (SBS) latex. 6.2.4 Suspensions Containing Small Solid Particles

In the case of small insoluble primary particles with a diameter of, typically, dv,50 < 2 mm within the droplet, the drying process takes place from the outer shell,

6.2 Morphology of Spray-Dried Particles

Fig. 6.6 SBS-latex particles dried at moderate air inlet temperatures, such as 120  C. The particles are hollow and show indentations due to collapse of the skin.

reducing steadily the size of the drop. Such conditions appear when spraying suspensions. A typical application is the preparation of ceramics. In the production of ceramics, spray drying is usually followed by a further compaction of the bulk material by external pressure and, finally, by a sintering process (Nebelung et al., 2008). At a high speed of shrinkage, the diffusion rate of solids accumulated with a high concentration in the outer regions of the suspension droplet is not fast enough to equalize the particle concentration difference within the drop. The diffusion coefficient Dsp of suspension particles with diameter dp is usually about 1 to 4 orders of magnitude smaller than, for example, low molecular solutes. It can be estimated from the relationship Dsp ¼ kT=ð3pmdp Þ

ð6:4Þ

where k is the Boltzmann constant, T the temperature in K and m is the viscosity of the dispersing agent. A bridge or shell-like structure is formed by the primary particles

Fig. 6.7 SBS-latex particles dried at higher air temperatures, such as 160  C. The particles are hollow and have bubbles in their interior as well as indentations due to collapse of the skin.

j241

j 6 Morphology and Properties of Spray-Dried Particles

242

when they come into contact. As there are only fairly small pores between small particles, the “filtration” resistance of the particle system is high and the resulting outward liquid flow is small. Whereas liquid flows mainly through the small pores, air penetrates in the reverse, inward direction through larger pores (Koch and Walzel, 2001). Due to the capillary effect, this leads to a considerable under-pressure within the particles, causing the shell to bend inwards forming indentations or dimples or even a donut-like shape. For suspensions, Minoshima et al. (2001, 2002) proposed a model based on the buckling behavior of the crust. According to this model, a particle is allowed to shrink as long as the buckling pressure is larger than the permeation pressure of the evaporating fluid. The buckling pressure depends on the shell thickness as PB ¼ f1=½2ðdd =dshell Þ
1g2 f8E=½3ð1
n2 Þ1=2 g

ð6:5Þ

Here, E is the Young’s modulus of the primary particles, n is the Poisson number of the material, dd is the drop diameter and dshell is the inner shell diameter. The mobility and softness of the shell consisting of primary particles also depends to a large extent on their electrostatic charge or f-potential. High f-potentials mean high charges of equal sign on the particles and repellent forces, whereas low or zero f-potential leads to immediate agglomeration and fixation of the position at which particles first come into contact. Therefore, structures with large pore volumes appear in the case of low f-potentials and structures of higher density arise when the particles can move, even when in contact with neighboring particles, and find a more stable position within the bulk. Highly stabilized suspensions, that is, those with high f-potentials, have a low viscosity and are easier to spray; they give rise to hollow particles with a fairly dense shell, quite similar to the case of crystallizing solutes. An example of these structures formed from 0.8 mm Si3N4 primary particles is given in Fig. 6.8. The stabilization usually is obtained by a pH-shift or by dispersants (polyelectrolytes). In the case of suspensions with low stability, no large central cavities are formed within the particles, so that particles of low but uniform density are formed. This particle structure is desired if a high density is to be achieved by subsequent pressing of the powder. Flash spray drying of suspensions was examined by Monse et al. (2001). Suspensions of BaSO4 particles with a primary particle size of dV,50 ¼ 1.7 mm were sprayed. The suspension was heated to temperatures within the range 140  C < T < 160  C and flashed through a swirl nozzle into the tower with an outlet air temperature of about 160  C. Figure 6.9 shows the morphology of the particles obtained from superheated and flashed (Fig. 6.8b), as well as from non-flashed (60  C, Fig. 6.8a) suspensions. The particles generated during the flash process have much higher density and only a small cavity in the center. This effect can be explained by the sudden immobilization of the primary particles throughout the agglomerate when a considerable part of the suspending liquid suddenly evaporates. In contrast, the nonsuperheated spray operation at the same drying air temperature gives rise to donuts

6.2 Morphology of Spray-Dried Particles

Fig. 6.8 SEM pictures of spray-dried particles generated from Si3N4-suspensions with different degrees of stabilization (Fries, 2008).

and to particle shapes with a large central cavity. The stabilization state of the suspension was reasonable in all cases with f ¼ þ 40 mV at a pH of 7.2. CMC (CRT 2000 GA from Wolff Cellulosics) was added as a binder (0.2% by weight related to the mass of BaSO4). The BaSO4 mass fraction of the slurry was 60%. Smaller primary particles lead to a higher strength of the agglomerate spray particle (Rumpf, 1975). However, the strength may still be insufficient for the subsequent treatment of the product. It is, therefore, often necessary to add some binder, which helps to provide the necessary adhesion between the primary particles, even in the dry state. Different binders are known such as carbon hydrates

Fig. 6.9 SEM photos of spray-dried BaSO4suspension particles: Upper part of figure shows the outer appearance, lower part of the figure shows the cut through particles. (a) Donut-like particles produced at a suspension

feed temperature of 60  C; (b) particles obtained from superheated suspension at a feed temperature of 160  C. The particles are less porous and show only small holes or cracks in the particle center. (Monce et al., 2001).

j243

j 6 Morphology and Properties of Spray-Dried Particles

244

(for example CMC), sugar, starches, molasses, gelatin, alginates, carrageenes or polyvinylalcohol, polyvinylpyrrolidone, polyacrylates and polyethyleneglycol, as well as silicon dioxide in different precipitation states. The addition of binder often further lowers the permeability of the crust and gives rise to stronger indentations and to the formation of hollow particles or donut-shaped particles (Jørgensen, 2005). For an investigation of the spray drying of TiO2-, SiO2- and Al2O3-slurries with different kinds of additives see also Jørgensen (2005). In general, high drying temperatures (T > 150  C) and small primary particles of the slurry (d < 5 mm) give rise to the formation of hollow particles or particles with indentations; low temperature drying conditions (T < 150  C) lead to uniform and compact particles. Highly concentrated slurries also lead to compact particles, as there is no space for the primary particles to rearrange their configuration during the drying step. 6.2.5 Suspensions Containing Large Solid Particles

Large primary suspension particles, that is, usually particles with dV,50 > 5 mm, provide a higher permeability of the pore system emerging during spray drying. Consequently, the liquid is transported through the pores outwards much more easily than in the small primary particle system and the shell remains wet for a longer time. The larger pores also lead to a smaller capillary, that is, indentation pressure. Due to that fact, drops from these suspensions have been observed to dry most likely into a compact non-hollow particle (Jørgensen, 2005). However, in this case binders are mandatory to avoid fracturing of the agglomerated spray particles even with cautious handling. Furthermore, the viscosities of suspensions consisting of coarse particles are lower, so that such suspensions can usually be sprayed with higher solids content, which again contributes to a more uniform distribution of the voids within the particle. 6.2.6 Complex Dispersions Such as Emulsions and Other Formulations 6.2.6.1 Hard Shell Particles Sometimes particles with a hard but permeable shell are desired. This is, for example, the case for catalyst particles in FCC processes like VPO precursors (Contractor et al., 1987). These agglomerated particles containing the active catalyst are used within a circulating fluidized bed to perform redox reactions at high temperatures. Hard thermal-resistant and abrasion-resistant shells can be obtained by adding precipitated silicon oxide to the precursor suspension before spray drying (Uihlein, 1993; Koch and Walzel, 2001). During the drying process, the nanosized silicon oxide colloids migrate, together with the suspending liquid (water), through the small pores towards the surface of the spray particles forming a gel and glueing the primary particles there. In compensation, air penetrates inwards through the large pores to equalize the negative pressure. The large pores are then emptied from

6.2 Morphology of Spray-Dried Particles

Fig. 6.10 Drying of a hemispherical suspension particle on a cooled glass sheet. Left to right: Intrusion of air fingers, formation of wet isles, migration of colored (dark) liquid towards the surface and deposition of the dissolved dark color at the outer rim of the particle.

the liquid. These pores later provide the necessary access for the gas to reach the active sites of the catalyst and the silicon oxide shell persists even during the tempering of the catalyst. Experiments with suspension droplets deposited as half spheres on a cooled glass surface allow observation of the drying and migration process on a microscope from below when these droplets are subjected to a hot gas flow (Monse et al., 2006). The initial drop in Fig. 6.10 consisted of a 40 wt% suspension with 4.5 mm BaSO4 particles in water also containing 5% dissolved black color. First, the intrusion of air fingers into the large pores in the outer rim of the droplet can be observed. Later, there are some isles of liquid (dark) left within the internal part of the droplet. As the drying proceeds, liquid is transported through the small channels or pores between the beads towards the surface. Simultaneously, the dissolved color is carried along with the liquid and is finally deposited close to the surface, where it remains as a dried layer. The same occurs when a gel such as silicon oxide is dissolved in the liquid. However, the viscosity of the gel must be low enough for the migration process. The large pores stay open and allow gas penetration into the core of the particles, as this is required for catalysts. 6.2.6.2 Gelatinization In the case of starch, one has to consider gelatinization effects taking place at high drying temperatures. This has a strong influence on the morphology. In Jørgensen (2005), experiments were carried out with maize and rice starch suspensions at different ambient air temperatures. At drying air temperatures of T > 200  C hollow particles were formed with high contents of gelatinized material in the case of rice starch. This is an indication that temperatures within the particle were beyond the gelatinization temperature of rice starch (70  C). At air temperatures of T < 150  C the primary starch particles remain separate and non-hollow particles are formed with uniform and compact structures throughout the cross-sections. However, these particles have low mechanical strength. 6.2.6.3 Microencapsulated Flavor Powders Formed from Emulsions In studies of the morphology of encapsulated flavor powders, scanning electron microscopy (SEM) has been commonly used to observe the porosity and surface integrity (Rosenberg et al., 1985; Kim and Morr, 1996; R
e and Liu, 1996;

j245

j 6 Morphology and Properties of Spray-Dried Particles

246

Fig. 6.11 SEM picture of (a) outside and (b) inside structure of spray-dried particles. The active core material (D-limonene emulsion) is in the form of small droplets embedded in the

shell region of the carrier matrix (b). In the center of the capsules, a large void can be observed which occupies most of the capsule volume.

Soottitantawat et al., 2003, 2004b, 2007). Several techniques have also been developed to provide information about the inner microstructure of the microcapsules (Moreau and Rosenberg, 1993). Figure 6.11 shows typical SEM pictures of the outside and inside structure of spray-dried particles (Soottitantawat et al., 2003). The inner structure of spray-dried particles obtained from emulsions is expected to contain the active core material (here D-limonene emulsion) in the form of small droplets embedded in the shell region of the carrier matrix. In the center of the capsules, a large void can be observed which occupies most of the capsule volume. The formation of the central void can be related to the expansion of the capsule (Chang et al., 1988; El-Sayed et al., 1990; Maa et al., 1997; Hecht and King, 2000a, b). These quality factors are a direct result of the design and operating conditions of spray drying, such as the type of atomizer, drying air temperature, feed rate and feed concentration, and the viscosity or nature of the carrier materials (El-Sayed et al., 1990; Onwulata et al., 1996; Finney et al., 2002). Hollow particles are produced at higher air temperatures due to the expansion of air bubbles in the droplets during the drying (Verhey, 1972a, b, 1973; Greenwald and King, 1981; Alamilla-Beltr
an et al., 2005). In order to observe the encapsulated flavor droplet (emulsion) inside the spraydried particle, ethyl-n-butyrate was used as a model flavor. Nile Red was dissolved in the ethyl-n-butyrate and served as a fluorescein probe of the oil phase (ethyl-nbutyrate emulsion) of the solution. Flavor labeled in this way was added to the carrier solution and emulsified. Figure 6.12 presents the results of the application of CLSM to investigate the morphology and the arrangement of encapsulated flavor droplets in the spray-dried powder. The images shown in Fig. 6.12a were produced by merging three kinds of CLSM pictures made with an Argon laser (for sodium fluorescence in a green color), with a He–Ne laser (for Nile Red in a red color), and transmission pictures for a particle with (right-hand side) and without (left-hand side) a vacuole. Figure 6.12b depicts merged images from CLSM with the He–Ne laser and transmitted images, in order to more clearly observe the – labeled – droplets of flavor in the powder (Soottitantawat et al., 2004b).

6.2 Morphology of Spray-Dried Particles

Fig. 6.12 CLSM pictures of the morphology and the arrangement of encapsulated flavor droplets in the spray-dried powder; merges of transmitted images with (a) CLSM images for sodium fluorescence in green color and for Nile Red in red color, (b) He–Ne laser CLSM images

in red color (Soottitantawat et al., 2004b). Droplets of flavor in the particles become visible especially in part (b) of the figure. Compact particles (left-hand side) can clearly be distinguished from particles with vacuole (righthand side).

The photos of Fig. 6.12 visualize in a non-destructive manner the distribution of emulsion flavor droplets (red color) over the cross-section of the spray-dried powder and the internal morphology of the product (i.e., the presence, or not, of a vacuole). Therefore, they show that the CLSM is a new valuable tool for studying and observing the encapsulated flavor. The technique can even be applied to study the release characteristics of flavor from the powder in real time (Yoshii et al., 2007). 6.2.6.4 Particles from Proteins, Enzymes and Carrier Materials In the last decade confocal laser scanning microscopy (CLSM) was shown to be a helpful tool for various further tasks of microparticle characterization (Lamprecht et al., 2000a, b, c). It minimizes the light scattered from out-of-focus structures, and permits the identification of several compounds through use of different fluorescence labels. Therefore, CLSM can be applied as a non-destructive visualization technique for microparticles. Moreover, CLSM allows visualization and characterization of structures not only on the surface, but also inside the particles, provided the carrier matrices are sufficiently transparent and can be fluorescently labeled; by collecting several coplanar cross-sections, a three-dimensional reconstruction of the inspected objects is possible. Figure 6.13 shows the application of CLSM to investigatation of the cross-sectional structures of spray-dried powders of maltodextrin (MD) with a dextrose equivalent value of DE ¼ 2 and 20. Florescein sodium salt was dissolved in the feed solution as a fluorescent probe of the carrier

j247

j 6 Morphology and Properties of Spray-Dried Particles

248

Fig. 6.13 The cross-sectional structures of spray-dried powders, observed by CLSM (MD, DE ¼ 2 and 20). Florescein sodium salt was dissolved in the feed solution to mark the carrier

solid. As only fluorescent material can be detected by CLSM, spray-dried particles containing a vacuole appear as green fluorescent rings.

solid. The laser supplied a 485 nm excitation wavelength allowing emission of the fluorescein sodium at excitation/emission wavelengths of 485/530 nm. Since only fluorescent material can be detected by CLSM, the shells of spray-dried particles containing vacuoles are observed in the form of green fluorescent rings (Soottitantawat et al., 2007). Figure 6.14 shows the SEM and CLSM pictures of spray-dried powder morphology for blends of GA-MD (GA: gum Arabic) as the carrier solid. Dextrose equivalent values of MD vary between 2 and 25; trehalose is also shown to represent the limiting case of a carrier material with a very high DE value (Soottitantawat, 2005). According to the SEM pictures of the surface structure of the powder, the powder shows a deeper surface grooving at low DE values. The grooving of the surface was observed to be less deep at higher DE values, so that the particles are smoother. Particularly for trehalose, no vacuoles inside the particles could be observed. The volume ratios of the vacuole to the powder observed by using CLSM for 300 randomly selected particles are shown in Fig. 6.15 as a function of the DE value of the carrier material of the powder. The volume ratio is nearly constant till a DE value of 11, and then decreases sharply with increasing DE value, reaching zero for trehalose. The behavior of trehalose can be easily anticipated from the results of Fig. 6.14. The outer surface morphology of spray-dried trehalose/alcohol dehydrogenase (ADH) powder, with the addition of different amounts of bovine serum albumin (BSA) or b-lacto-globulin (Lg) is illustrated in Fig. 6.16 (Yoshii et al., 2008). The particle surface appears smooth in the case of ADH alone, no grooves could be found. However, the addition of proteins (BSA or Lg) in the feed solution affects the appearance of the particles. Fine grooves appear when the mass ratios of b-Lg or BSA to ADH are higher than 0.5. The increase in protein content deforms the surface of

6.2 Morphology of Spray-Dried Particles

Fig. 6.14 (a) SEM and (b) and (c) CLSM pictures of spray-dried powder morphology for blends of GA-MD as the carrier solid. The dextrose equivalent value (DE) of MD was varied between 2 and 25; trehalose represents a very high DE value.

j249

j 6 Morphology and Properties of Spray-Dried Particles

250

Fig. 6.15 Volume ratio of vacuole to total particle as a function of DE values of MD of the carrier material. Spray-drying conditions: Inlet air temperature 180  C, atomizer rotational speed 30 000 rpm. The error bars show the standard error of the mean of 40 samples.

Fig. 6.16 Effect of addition of BSA and b-lactoglobulin on the spray-dried particle morphology; (trehalose mass fraction 30%, ADH content: 0.23 mg ADH/g trehalose, standard drying conditions).

6.2 Morphology of Spray-Dried Particles

Fig. 6.17 The effect of outlet temperature and additives on the formation of hollow particles in spray-dried powders. Control: &: blend of GA/MD; Additives: : gelatin (1 wt%), ^: ethanol (5 wt %), ~: decaglycerine monolaurate (0.14 wt%).

.

the particles resulting in rougher surfaces. However, there were no observable differences in the appearance of particles containing one or the other of the two proteins used. The particles with only ADH were very sticky due to quick adsorption of water that destroyed the granule structures. These different external structures could be attributed to the nature of the surface crust. For example, addition of proteins may increase the porosity but may also change the flexibility of the surface, mechanical strength, evaporation of water linked to the proteins, and so on (Maa et al., 1997; Landstr€om et al., 2000). Another factor potentially influencing the surface morphology could be the hydrodynamic behavior of the liquid droplet, since, upon atomization, the liquid droplets are subjected to surface turbulence and internal motion (Maa et al., 1997). The effect of outlet temperature on the formation of hollow particles in spray-dried powders is illustrated in Fig. 6.17 for various additives involved in the feed liquid (Yoshii et al., 2006). The percentage of hollow particles increases linearly with the increase in the outlet air temperature. When 0.14 wt% of surfactant (HLB 10) is added to the GA/MD solution, the percentage of hollow particles is the highest, reaching about 87% at an outlet temperature of 135  C. This phenomenon might be related to bubble formation in the sprayed droplets. High percentages of hollow particles are also obtained by the addition of small amounts of ethanol or gelatin. Gelatin is an enhancer of film formation. The film-forming property might promote the expansion of the spray droplets. 6.2.7 Particles Obtained in Freeze Spray Drying

Freeze spray drying is applied to substances with otherwise very sticky surfaces, as in the food industry, or to enable the formation of very porous structures with excellent

j251

j 6 Morphology and Properties of Spray-Dried Particles

252

instant properties. In any case, the process is extremely protective and allows the maintenance of biological activities, even of very sensitive proteins like vaccines (Garmise, 2007) and bacteria, enzymes or peptides. Further protection of sensitive biopharmaceuticals is provided by additives, such as for example, mannite, trehalose or glycine (Constantino et al., 2001), The respective excipients must be added to hygroscopic materials also for various applications in the food industry. The process is well suited to the formation of particles which are porous and relatively large in comparison to the original droplets, but, nevertheless, still have a small aerodynamic diameter ( 1, an induction period followed by a burst of flavor emulsion can be observed. Note, however, that the described classification of release mechanisms is valid for a given single microcapsule. A mixture of microcapsules usually includes a distribution of capsules varying in size and wall thickness. Since any spray-dried powder produced from an emulsion is essentially such a mixture of microcapsules with variations in their properties, the parameter n in Eq. 6.6 varies depending on the properties of the powder. Equation 6.6 is essentially analogous to the equation of Kohlraush–Williams–Watts (KWW). This relationship can be expressed as (Williams and Watts, 1970) M ¼ M0 exp½ð
t=tÞb ;

R ¼ M=M0

ð6:7a; bÞ

where M(t) corresponds to the residual amount of flavor in the powder, M0 to the initial mass of the flavor, t is the relaxation time which corresponds to the inverse of the release rate of the flavor, and b is the relaxation constant. Equation 6.7a, was originally proposed to express the relaxation phenomena in a polymer. The relaxation constant b represents the breadth of the energy distribution in polymer relaxation phenomena; b ¼ 1 means simple relaxation, whereas smaller values of b mean a larger width of the energy distribution. As already mentioned, spray-dried powder consists of various particles having different release characteristics. Consequently, the total release behavior may be considered as the sum of many KWW relaxation equations valid for individual particles, i, or classes of particles: h i X M¼ M0;i exp ð
t=tÞbi ð6:8Þ

The release of flavor from the spray-dried powder during storage is recognized as a kind of relaxation phenomenon in an amorphous glass, inside which emulsion droplets of different sizes are distributed. Therefore, it should be possible to develop alternative correlation equations of flavor release from a statistical perspective. Considering the distribution of activation energy for the rate constants (Kawamura et al., 1981), the following equation was developed for the correlation of the complicated time-dependent phenomena:

with

C RT ¼ pffiffiffiffiffiffiffiffi R¼ C0 2ps k1 ¼

þð1


1

 2 ! R2 T 2 ln k1
ln k1;0 exp
expð
k1 tÞdðln k1 Þ 2s2


 kT DG exp
h RT

ð6:9Þ

ð6:10Þ

j263

j 6 Morphology and Properties of Spray-Dried Particles

264

Here C is the flavor concentration in the stored spray-dried powder, C0 is the initial flavor concentration, k1,0 is the initial release rate constant, DG is the activation energy of flavor release, k is Boltzmann’s constant and h is Planck’s constant. Equation 6.9 was originally developed to express the inactivation kinetics of a-chymotrypsin and glucoamylase covalently bound to a water-insoluble support in an aqueous system (Kawamura et al., 1981). Equation 6.9 was applied to express the oxidation kinetics of fish oil (EPA) in linoleic acid powder, based on the assumption that the free energy of activation DG follows a Gaussian distribution with the standard deviation s (Yoshii et al., 2003; Ishido et al., 2002, 2003). The abovementioned three equations are equivalent to the perspective of simulating the flavor release from spray-dried powder. All of the parameters – namely n in Avrami’s equation, b in the KWW-equation, and a Gaussian distribution of DG with the standard deviation s in Eq. 6.9 – can be understood as a consequence of the activation energy distribution of the release rate.

Release Rate of Flavor At Various Relative Humidities and Temperatures The release behavior of flavors not only varies with the relative humidity, but also with the composition of the carrier materials. The combination of moisture and temperature strongly influences the degree of plasticization of the carrier material during storage, thereby governing the rates of flavor loss. A typical example of the release kinetics of encapsulated flavor in spray-dried powder is illustrated in Fig. 6.26. The plot shows the residual mass fraction of Dlimonene in powder stored at 50  C at 23%, 51%, 75%, and 96% relative humidity (RH) according to measurements by Soottitantawat et al. (2004a). In all cases, a blend of GA and MD was the capsule matrix. The relative humidity greatly affects the

D-limonene

retention (–)

1.0

0.8

0.6

Humidity RH 96% RH 75% RH 51% RH 23%

0.4

0.2

0 0

5

10

15

20

Time (d) Fig. 6.26 Release kinetics of D-limonene encapsulated in spray-dried powders stored at various relative humidities and at 50  C. Carrier material is a blend of GA and MD. Relative humidity (RH): & 23%, ~ 51%, ! 75%, * 96%.

6.3 Retention of Flavor in Spray-Dried Food Products

release rate of D-limonene, but the relationship is not simple. Considering only 23% and 96% RH the release of D-limonene increases with increasing RH. However, when comparing the results of 51% and 75% RH, one can see that the release of D-limonene at 51% RH is higher than that observed at 75% RH. These results prove that the release of D-limonene is closely related at least to the water activity of the powder (Yoshii et al., 2001; Levi and Karel, 1995). The loss of D-limonene during storage may be caused by two mechanisms: diffusion through the matrix of the carrier material and oxidation. However, the loss by oxidation was at most 5–6% of the initial Dlimonene content. Therefore, the observed loss of D-limonene may be considered to result mainly from release by diffusion (Whorton and Reineccius, 1995). To evaluate the release rate constant of D-limonene, Eq. 6.6 was applied to the measured release kinetics, resulting in the solid curves in Fig. 6.26. These solid curves show that Eq. 6.6 can correlate successfully the release curves of D-limonene and the release rate constant kR can be estimated as a function of RH (Soottitantawat et al., 2004a). The same correlation was also applied to various other carrier materials. Figure 6.27 shows the resulting release rate constant kR in dependence on the water activity aw of the carrier matrices. In brief, the release rate constant kR first increases with increasing aw, followed by a decrease at around aw ¼ 0.70. At a still higher aw, the release rate constant tends to increase again because the powder matrices are destroyed. The variation pattern of kR against aw is very similar irrespective of the kind of carrier material. A similar phenomenon was observed in the study of the equilibrium head space volatile concentration of roasted coffees as a function of water

Release rate constant, kR × 104 (day–1)

1000

800

600

400

200

0 0.0

0.2

0.4

0.6

0.8

1.0

Water activity (–) Fig. 6.27 The relationship between the release rate constants kR and the water activity of the carrier matrices aw at 50  C; Carrier material: * blend of GA-MD, & blend of SSPS-MD, ! blend of HI-CAPÒ 100-MD, ~ HI-CAPÒ 100. The error bars indicate 95% confidence levels.

j265

j 6 Morphology and Properties of Spray-Dried Particles

266

Fig. 6.28 Outer structural changes during storage at 50  C of spray-dried powder produced in the form of wall capsules from the blend GA-MD; (a) storage at 23% RH for 1 w, (b) storage at 51% RH for 1 w, (c) storage at 75% RH for 1 d.

activity. The volatile concentrations were found to be low at very low and at high water activities. They were highest at intermediate water activity in the range 0.25–0.35 for the light and medium-roasted coffees and at about 0.7 for the dark-roasted coffee (Anese et al., 2005). The decrease in kR between 51% and 75% RH in Fig. 6.27 corresponds to the mentioned reversal of the release curves of D-limonene in Fig. 6.26. It may be closely related to the particle matrix structure changes from amorphous glass to the rubbery state by the plasticizing effect of moisture adsorption. In order to explain the structural changes that happen in the powder at different aw, scanning electron microscopy (SEM) was used to observe the outer structure of the powder during storage, as shown in Fig. 6.28 (Soottitantawat et al., 2004a). At a low RH (0.23) the spray-dried particles remain in their original shape and the carrier solid seems to be still in the glassy state (Fig. 6.28a). As compared to the release rate at a water activity of 0.23, the release rate at aw  0.51 is higher, though the structural changes cannot be observed clearly (Fig. 6.28b). This suggests that the higher mobility of D-limonene is due to the start of plasticizing of the capsule matrices. When the water activity increased further to a value of around 0.7, the powders began to be re-hydrated (Fig. 6.28c). At this stage, it may be assumed that a decrease in effective surface area resulted in a decrease in D-limonene evaporation from the surface of the powder particles. Most particles are observed to be clogged and adhering together into a paste-like mass, which indicates the rubbery state of the carrier matrices. The regime corresponds to the minimal value of kR. At a very high aw, the release rate increased again (Fig. 6.27). The most likely explanation is that the emulsion droplets of D-limonene in the powder are opened due to the destruction of the capsule matrices (Whorton and Reineccius, 1995). PTR-MS as a New Methodology for Analyzing Flavor Release Recently, thanks to the pioneering work by Dronen and Reineccius (2003), proton transfer reaction mass spectrometry (PTR-MS) has been used as a rapid analysis method to measure the release time-course of flavors from spray-dried powders. The traditional equilibrium method for flavor release mentioned above is extremely time-intensive, and commonly several weeks are necessary to obtain a full release profile of the flavor from the powder. The PTR-MS method has been applied extensively to analyze the release

6.3 Retention of Flavor in Spray-Dried Food Products

kinetics of volatile organic compounds from roasted and ground coffee beans. The release profiles were mathematically analyzed by means of Eq. 6.7a, to obtain the release kinetic parameters (Mateus et al., 2007). The method is quite sophisticated and quickly leads to results. However, as suggested by Dronen and Reineccius (2003), comparing release profiles with those obtained by the traditional method, the disadvantage of PTR-MS is its inability to capture time-dependent phenomena (relaxation phenomena) such as the collapse and glass transition of the carrier matrices. As mentioned above, the relaxation phenomena related to the state change in the carrier matrices have a significant effect on the flavor release characteristics. However, these time-dependent phenomena cannot be detected and properly accounted for with the PTR-MS method. 6.3.3.3 Oxidation of Encapsulated Flavor During Storage In addition to the release of flavors through the wall of the spray-dried particles, oxidation of the encapsulated flavors is also an important index of stability. Many different products form in oxidation reactions of flavor. In the case of the encapsulation of the model flavor D-limonene, limonene oxide (limonene-1, 2-epoxide) and carvone are commonly chosen as indicators of the oxidation. These products are generated together in the oxidation reaction of D-limonene (Anandaraman and Reineccius, 1986). The formation of the oxides also markedly depends on RH and increases initially during storage. During the initial period, the formation of oxides increases linearly with time, so that the apparent oxidation rate constant can be calculated on the basis of zero order kinetic reaction schemes (Anandaraman and Reineccius, 1986). However, over a longer storage time the formation rate of the mentioned oxides tends to decrease, particularly at higher RH. This might be explained by the accelerated degradation towards other oxide compounds and release of the oxides into the surroundings (Soottitantawat et al., 2003). The initial oxidation rate constants (zeroth order reaction rate) kX are plotted in Fig. 6.29 for carrier matrices consisting of GA-MD during storage at 50  C. The changes in kX are very similar to those of the release rate constants kR shown in Fig. 6.27. In fact, the data for the GA-MD blend are the same (empty cycles in Fig. 6.27, full cycles in Fig. 6.29), so that they provide a basis for assessing the influence of water activity on kX from Fig. 6.29. In this way, one can see that the oxidation rate constant kX reaches a maximum at aw  0.5 where the powder matrix structure begins to change from glassy to a rubbery state, and comes to a minimum at aw  0.7. 6.3.3.4 Relaxation Process Correlation by Glass Transition Temperature The release and the oxidation processes of the encapsulated D-limonene are closely related to the structural changes in the capsule matrices. Physico-chemical changes caused by the phase transition of carbohydrate from amorphous glass to rubbery are commonly expressed with the temperature difference between the storage temperature, T, and the glass transition temperature, Tg, of the carrier matrices, T
Tg. The idea is based on the fact that the viscosity (or relaxation time) of the carrier matrices follows the Williams–Landel–Ferry (WLF) equation expressed as a function of T
Tg (Williams et al., 1955). Therefore, the release rate constants kR and the oxidation rate

j267

j 6 Morphology and Properties of Spray-Dried Particles

268

400

300 1.0 200 0.5 100

0

0

–50

0

50

100

0 150

Release rate constant, kR × 104 (d–1)

mg oxide (g limonene)–1 d–1

Oxidation rate constant, kX

1.5

T–Tg(ºC) Fig. 6.29 Correlation of kR and kX with T
Tg at T ¼ 50  C: oxide, ~ carvone. Carrier material is a blend of GA and MD.

.

D-limonene

released, * limonene

constants kX are correlated with T
Tg, as shown in Fig. 6.29. The rate constants kR and kX increase with an increase in T
Tg up to the point where T
Tg becomes roughly equal to zero. This is followed by a decrease and then an increase again with increasing T
Tg. A high release rate near the glass transition temperature (this means at T-Tg ffi 0) can be explained by the increasing mobility of D-limonene and oxygen molecules. Furthermore, low release and oxidation rates are observed in the range 0 < T
Tg < 50 K, when the powders are transformed from amorphous to a rubbery state. In this region, higher mobility of D-limonene and oxygen should occur, but at the same time, the collapse of the powders occurs, so that the particles clump and adhere together. This results in closing the pore spaces between the particles and decreasing the surface area for D-limonene evaporation and oxygen uptake. Whorton and Reineccius (1995) observed a similar behavior and defined the state of stickiness and aggregation above the glass transition temperature as the “re-encapsulation of flavor.” These results imply that the release rate and the oxidation reaction rate have a mutually close connection to Tg. The rubbery state and collapse of carrier matrices are associated with greater stability as compared to the glassy state because the decrease in surface area which occurs with the aggregation of carrier matrices can restrict both the release of flavor and the uptake of oxygen from the surface to the interior of the particle (Nelson and Labuza, 1992; Labrousse et al., 1992; Levi and Karel, 1995). Figure 6.30 is a schematic picture of a spray-dried particle in a humid air environment, in which the particle can absorb water vapor, followed by a state change of the carrier matrix from amorphous into rubbery. The encapsulated flavor can easily move in the rubbery carrier matrix. At the same time, oxygen uptake into the matrix wall becomes stronger, and oxidation of the encapsulated flavor progresses. Most interestingly, around the glass transition temperature both the release

6.4 Encapsulation and Microencapsulation of Enzymes and Oil by Spray Drying

Phase transformation of carrier solid Oxygen Amorphous absorption wall materials Flavor encapsulated

Case hardening

O2

Flavor migration and evaporation

Water adsorption

Fig. 6.30 Schematic illustration of a spray-dried particle in a humid air environment. The particle can adsorb water vapor, so that the state of the carrier matrix can change from amorphous into rubbery.

and the oxidation rates change according to nearly the same trends with T
Tg, as shown in Fig. 6.29, indicating that the flavor diffusion and the oxygen uptake can be treated as similar processes.

6.4 Encapsulation and Microencapsulation of Enzymes and Oil by Spray Drying 6.4.1 Microencapsulation of Enzymes by Spray Drying

A biotechnologically produced molecule with therapeutic activity is commonly a protein consisting of a chain of several hundred amino acids with a complex threedimensional structure. Protein-based drugs should remain stable for several years, maintaining the active conformation even under unfavorable conditions during transportation or storage. Freeze drying is one of the most common formulation methods of protein drugs, as discussed in detail in Chapters 3 and 4. However, freeze drying does not lead to well-defined microparticles and is a comparatively expensive and time-consuming process. Spray drying is an attractive alternative for the preparation of solid pharmaceutical substances compared to freeze-drying. However, spray drying bears risks of product damage due to the surface-induced denaturation of biomolecules during atomization and to the subsequent drying at a high temperature of the drying gas, which can be potentially detrimental to heat-sensitive biological materials such as enzymes. Therefore, it is important to develop new formulation concepts for encapsulation and stabilization of proteins during spray drying. Several factors may cause substantial inactivation of enzyme proteins during spray drying. Proteins are presumably exposed to high shear forces during the atomization

j269

j 6 Morphology and Properties of Spray-Dried Particles

270

process. Though proper selection of atomizers is important for the reduction of the inactivation of enzymes, few studies of this subject have been made. After atomization, droplets of enzyme solution come into contact with the high temperature air, suffering from thermal stress, which can result in an irreversible structural change of the enzyme (thermal denaturation). This stress is decreased by the addition of excipients such as disaccharides. Their role during spray drying is to replace water and to form an amorphous glass, thus providing a stabilizing effect. The air temperature is obviously a significant factor in thermal stress during spray drying. Some researchers studied the thermal inactivation of several enzymes during drying by the single suspended droplet method. They suggested that most enzymes start to degrade when the droplet enters the falling rate period of drying. In industrial spray dryers, the major part of the water in the droplet evaporates almost instantaneously in the vicinity of the atomizer. The dried droplets are then surrounded by air with the still high outlet temperature until being entrapped in the cyclone. This period during the spray drying process actually corresponds to the falling rate period in drying of single droplets, where enzymes suffer from severe thermal stress, resulting in extensive enzyme denaturation. Therefore, low air temperatures must often be used, and the resulting decrease in product yield must be accepted. Besides lowering the air temperature, the addition of some proteins or surfactants offers another way to prevent direct contact of protein materials with the high temperature air. These substances tend to distribute preferentially on the surface of the particle, thus protecting the enzyme from thermal damage. However, the effect of additives in relation to the spray drying conditions has not yet been studied extensively. In industrial spray dryers the size of the sprayed droplets is a few hundreds of micrometers. Therefore, the addition of proteins or surfactants may be one of the key techniques for the production of pharmaceutical enzymes by spray drying (Broadhead et al., 1992; Giunchedi and Conte, 1995; Wendel and Celik, ¸ 1997; Lee, 2002; R
e, 2006; Vehring, 2007). 6.4.2 Stress on Proteins During the Spray Drying Processes

Spray drying of protein solutions can result in substantial inactivation. Proteins have the tendency to denature and undergo irreversible aggregation during various steps of spray drying for various stress reasons. These reasons are shown schematically in Fig. 6.31 (Lee, 2002) and will be discussed in the following. 6.4.2.1 Adsorption Stress The loss of native protein by adsorption to surfaces is problematic in biopharmaceutical applications. Proteins are able to adsorb at liquid/gas, liquid/solid and liquid/liquid interfaces due to their amphiphilic property. Adsorption stress can occur in the vessel of the feed protein solution, and at the liquid/gas interface between the sprayed droplets and the hot air. Adsorption to the vessel wall is assumed to be practically negligible, whereas the adsorption at the liquid/gas interface may lead to substantial degradation because of the strong increase in the liquid/gas

6.4 Encapsulation and Microencapsulation of Enzymes and Oil by Spray Drying

Shear stress Nozzle, Atomizer Feed line

Shear stress Liquid/air interfacial stress

Tank Adsorption stress

Thermal stress Drying chamber

Fig. 6.31 Schematic diagram of a spray-drying system illustrating possible stresses experienced by protein solution and droplets.

interface area during atomization. Spray drying of adsorption sensitive proteins such as the recombinant human growth hormone (rhGH) has already evidenced the detrimental effect of adsorption stress (Maa et al., 1997). Addition of an appropriate amount of surfactant can reduce the inactivation of protein induced by atomization (Maa et al., 1997; Adler and Lee, 1999; Yoshii et al., 2008). The investigation of the surface element composition of spray-dried particles by ESCA (electron spectroscopy for chemical analysis) revealed the decrease in protein content at the particle surface by addition of surfactant (Adler and Lee, 1999). 6.4.2.2 Shear Stress Proteins are presumably exposed to high shear forces in the liquid feed duct, in centrifugal atomizers and, especially, in pressure nozzles during the atomization process. Soottitantawat et al. (2003) showed that emulsions containing large droplets are broken during atomization with a centrifugal atomizer due to the high shear force induced in the centrifugal outward flow on the rotating vane. Maa and Hsu (1996, 1997) extensively examined the effect of shear stress on the denaturation of rhGH. They concluded that the combination of shear and liquid/air interface triggered the formation of non-covalent aggregates of the protein, while shear alone did not induce significant denaturation of the protein. Addition of an appropriate amount of a surfactant to the protein solution can reduce the amount of aggregation induced by atomization. 6.4.2.3 Thermal and Dehydration Stress After atomization, the protein comes into contact with the high temperature air. Proteins may denature with structural changes under thermal stress and lose their activity. This thermal stress can generate an irreversible structural change during spray drying, resulting in thermal denaturation of proteins. Therefore, excipients are often added to the feed protein solution to improve the stability in the manufacturing process and storage. Numerous studies have documented the protective effects of

j271

j 6 Morphology and Properties of Spray-Dried Particles

272

excipients on the thermal degradation by means of analytical methods such as FTIR and SE-HPLC. The mechanisms of stabilization by the excipients during dehydration are supposed to be the direct interaction between excipients with the protein to protect the biostructure through hydrogen bonds, the trapping of water molecules close to the protein surface, and the entrapment of a particular protein conformation in a highly viscous amorphous glass of the excipients. These mechanisms will be further discussed in the subsequent section. 6.4.3 Protein Encapsulation Theory by Spray Drying

Encapsulation involves the incorporation of active ingredients such as flavors, enzymes, cells or other materials in small capsules. The choice of excipients for encapsulation is very important for the encapsulation efficiency and protein stability within the matrix. Applications of this technique have increased in the food and pharmaceutical industries since the encapsulated materials can be protected from moisture, heat or other extreme conditions. Thus their stability is improved and their viability maintained. Powder formation can lower the water activity of the material, the reactivity and the diffusivity of encapsulated compounds, and the diffusivity of residual water. In the food industry microencapsulation is often associated with the already discussed retention of flavor compounds during drying and storage. In pharmaceutical applications, the purpose of microencapsulation is to control the release and improve the bioavailability of active ingredients. The exclusion of water from the environment of a protein may provide resistance to chemical modification of the protein structure during processing and storage. Processes such as freeze-drying and spray drying have been employed to prepare solid-state dosage forms of pharmaceutical proteins with vastly reduced water content. Several theories of protein encapsulation in glassy materials have been proposed based on the three considerations mentioned in Section 6.4.2.3: .

.

The first is the “water replacement hypothesis” proposed by Crowe et al. (1984) for the preservation of proteins in the dried state. According to this hypothesis, biostructures are protected by means of hydrogen bonds resulting from their direct interaction with stabilizer molecules. Certain physiological solutes “replace” the water lost around polar residues of proteins from the primary hydration shell during dehydration. As shown in Fig. 6.32, the wall material, in this case trehalose, is acting as a substitute for hydration water molecules. Carpenter and Crowe (1984) suggested that some of the hydroxy groups of sugars form hydrogen bonds with the polar residues of proteins. Sugars such as trehalose are often used in pharmaceutical, food and biomedical applications to prepare glassy matrices for long-term storage of biological materials. Secondly, the “water layer hypothesis” proposes that the stabilization of proteins is achieved by the trapping of water molecules close to the biomolecular surface (Belton and Gill, 1994). The protein stabilization conferred by the excipients during dehydration is brought about primarily by these excipients substituting for water molecules on the surface of the protein.

6.4 Encapsulation and Microencapsulation of Enzymes and Oil by Spray Drying

Fig. 6.32 Concept of the “water replacement hypothesis.”.

.

The third hypothesis is called the “glass entrapment hypothesis” which proposes stabilization through entrapment of a particular biomolecular conformation in a high viscosity glass (Ansari et al., 1992). Franks et al. (1994) suggested the similar “glass state theory” explaining the glass formation to be responsible for the stabilization of proteins. The molecular mobility of the protein-containing system is greatly limited by trapping protein molecules in a glassy matrix. As a consequence the rates of diffusion-controlled reactions, including protein unfolding, protein aggregation and chemical degradation, are reduced.

6.4.4 Spray Drying of Protein Solutions 6.4.4.1 Drying of a Single Suspended Droplet Yamamoto and Sano (1992, 1994) studied the thermal inactivation of several enzymes during drying of a single suspended droplet. The droplet had a diameter of 1.5 to 2 mm and contained b-galactosidase, glucose oxidase, or alkali phosphatase. It was suspended on a fine glass filament and dried in an air stream. The time course of enzyme inactivation was obtained during drying at constant drying conditions (temperature, humidity, velocity, etc.). The same technique was previously applied to investigate the time profile of lipoxygenase inactivation during drying (Liou, 1982). The kinetic model for the enzymatic inactivation reaction was obtained from separate experiments conducted at various temperatures and relative humidities. In this way, it was possible to successfully estimate enzyme retention during drying by means of numerical analysis, solving the nonlinear dehydration equation coupled with the enzymatic inactivation reaction. Yoshii et al. (2005) studied the deactivation of alcohol dehydrogenase (ADH) in a drying suspended droplet containing mixed carbohydrates, namely trehalose and randomly methylated b-cyclodextrin (RM-b-CD). They suggested that trehalose was the best excipient and RM-b-CD had a stabilizing effect for alcohol dehydrogenase during drying. Meerdink and van’t Riet (1991), and Meerdink (1993) investigated the inactivation of thermo-stable a-amylase in maltodextrin solution droplets, generated by electrostatic atomization and dried during the free fall vertically through a drying tower. The residual activity of thermo-stable

j273

j 6 Morphology and Properties of Spray-Dried Particles

274

a-amylase during drying was predicted with a first-order Arrhenius-type mathematical model. For this purpose, Meerdink and van’t Riet (1991) described the kinetic constant of the deactivation reaction, kd, as a water-dependent function of the deactivation energy Ed and the pre-exponential factor k0 as follows:
 a kd d Ed;0 þ bX ¼ exp cX
ð6:11Þ RT k0;1 Here, X is the water content on a dry basis. The parameters a, b, c, d, k0,1 and Ed,0 were estimated with separate experiments, each of them conducted at constant water content and constant temperature. 6.4.4.2 Stabilization of Enzymes During Spray Drying: Effects of Formulation Composition Effect of Enzyme Content on the Stabilization of Enzyme Activity in Spray-Dried Powder The initial enzyme concentration in the feed liquid was found to affect the stability of enzyme activity during spray drying (Yoshii et al., 2008). Trehalose solution of 30 wt% containing alcohol dehydrogenase (ADH) at five different ADH contents was spraydried at 120  C inlet air temperature with a rotary disk atomizer. The retention of ADH activity is presented in Fig. 6.33 as a function of the initial amount of ADH in the formulation solution. The retention of ADH increases linearly in the low concentration region of ADH, reaching a plateau value of around 2.8 times higher activity than at lower initial ADH content. ADH molecules appear to mutually protect themselves during spray drying.

Retention of ADH activity (–)

1.0

0.8

0.6

0.4

0.2

0 0

0.2

0.4

0.6

0.8

1.0

ADH mass content mg (g

1.2

1.4

trehalose)–1

Fig. 6.33 Effect of ADH mass content on the retention of ADH activity. Trehalose mass fraction: 30%; Spray drying conditions: rotary disk atomizer at 30 000 rpm, inlet temperature 120  C, outlet temperature 70  C, air mass flow rate 100 kg h
1, liquid feed mass flow rate 40 ml min
1.

6.4 Encapsulation and Microencapsulation of Enzymes and Oil by Spray Drying

Retention of ADH activity (–)

1.0 0.8 0.6 0.4 0.2

0

0

0.5 1.0 1.5 2.0 β -Lg or BSA / ADH (mass ratio)

Fig. 6.34 Retention of ADH activity as a function of the mass ratio of additive proteins to ADH for (~) BSA and (&) lactoglobulin. Trehalose mass fraction: 30%, initial amount of ADH: 0.23 mg ADH (g trehalose)
1. The spray-drying conditions are the same as in Fig. 6.33.

Enzyme Stabilization by the Addition of Other Proteins and Additives To overcome the thermal damage of ADH during spray drying, Yoshii et al. (2008) added a small amount of other proteins such as bovine serum albumin (BSA) or b-lactoglobulin (b-Lg) to the formulation solution. The preservation of ADH activity increased with increasing amount of proteins added, see Fig. 6.34. As mentioned previously, proteins tend to accumulate at the air/droplet interface. Several studies confirmed that the composition of the droplet surface was preserved during spray drying, the protein accumulating on the surface of the spray-dried particle. One of the possible reasons for the protection of ADH by the addition of proteins is that BSA and b-Lg may have covered the surface of the sprayed droplet thus preventing the exposure of ADH to the hot air stream. Effect of Surfactant and Other Additives on the Retention of Enzyme Activity A common method to improve the preservation of the native structure of an enzyme is to add surfactants to the enzyme solution. Non-ionic surfactants can reduce the aggregation of enzyme proteins during spray drying at or above the critical micelle concentration of the surfactants in the solution (Maa et al., 1997, 1998a; Yoshii et al., 2008). Moreover, increasing the surfactant concentration in a spray solution progressively reduces the inactivation of enzyme during the spray drying. Figure 6.35 illustrates the protective effect of Tween 80 on the inactivation of ADH during spray drying. The addition of 0.047% Tween 80 improved the enzyme activity retention by 9.2%, and the addition of 0.093% by 15.5%. Comparing with BSA and b-Lg, the addition of Tween 80 on ADH has a similar effect. 6.4.4.3 Effect of Process Variables on the Stabilization of Enzymes During Spray Drying The effect of process conditions on the stability of enzymes during spray drying has been investigated with both a pilot plant spray dryer (Yoshii et al., 2008) and bench-top

j275

j 6 Morphology and Properties of Spray-Dried Particles

276

Retention of ADH activity (–)

0.60

0.55

0.50

0.45

0.40

0.35 A

B

C

Fig. 6.35 Effect of surfactant (Tween 80) on the retention of ADH activity. A: ADH alone, B: 0.467 mg Tween 80 (g trehalose)
1, C: 0.933 mg Tween 80 (g trehalose)
1. Trehalose mass fraction: 30%. The spray drying conditions are the same as in Fig. 6.33.

spray dryers (Maa et al., 1998b; Etzel et al., 1996; Meerdink and van’t Riet, 1991). The inlet and outlet temperatures of the air are important factors for the preservation of enzymatic activity during drying (Yoshii et al., 2008). The retention of activity at different inlet air temperatures is illustrated in Fig. 6.36 against the outlet air temperature for formulations with and without addition of b-Lg. The outlet air temperature markedly affects the retention of ADH activity, particularly between 70 and 83  C in the case of the formulation containing b-Lg. At higher outlet temperature, the activity becomes impracticably low and converges to the same value for both formulations, with and without b-Lg. These results infer that it is advisable to keep the outlet air temperature as low as possible, provided that a dried powder can still be produced. Since the inlet and outlet temperature could not be controlled independently in this work, it could not be concluded which temperature was more important in retaining enzymatic activity. Similar results have been obtained experimentally by Etzel et al. (1996) for the residual activity of alkaline phosphatase in dependence on the outlet air temperature. The same authors predicted successfully the retention of activity by coupling a mathematical model of droplet drying with a model for the inactivation reaction. Maa et al. (1998b) examined a bench-top spray dryer with the goal of improving the efficiency in both production yield and throughput for the preparation of protein spray-dried powder. They found that a significant loss of particles occurred in the cyclone system. System modifications such as different cyclone designs and replacement of the bag-filter unit by a vacuum system were performed, allowing the protein to be dried at lower inlet/outlet air temperatures.

6.4 Encapsulation and Microencapsulation of Enzymes and Oil by Spray Drying

Retention of ADH activity (–)

1.0

0.8

0.6 (120°° C)

0.4 (140) (160)

0.2

(180)

0 70

80

90

100

110

Outlet air temperature (°C) Fig. 6.36 Effect of outlet temperature (abscissa) and inlet air temperature (parameter) on ADH retention during spray drying; * ADH alone, & ADH with b-lactoglobuline. Trehalose mass fraction: 30%, initial ADH content: 0.23 mg ADH (g trehalose)
1.

As previously mentioned, the liquid feed undergoes a certain shear stress during the atomization process. To assess the influence of shear stress on enzymatic activity, both a rotary disk atomizer and a two-fluid nozzle were used to spray the liquid feed. As shown in Fig. 6.37, the retention of ADH activity was found to be nearly the same with the examined methods of atomization. These results indicate that the shear

Retention of ADH activity (–)

1.0

0.8

0.6

0.4

0.2

0 0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

ADH concentration mg (g trehalose)–1 Fig. 6.37 Comparison of the retention of ADH activity between a rotary disk atomizer (*) and a two-fluid nozzle ( ); Trehalose mass fraction: 30%. Other conditions are the same as in Fig. 6.33.

.

j277

j 6 Morphology and Properties of Spray-Dried Particles

278

stress associated with atomization was either almost equal in both cases or very low for a significant difference to be obtained. 6.4.5 Microencapsulation of Oil

Powdery oils are classified into two categories: formulated microcapsules of oil and encapsulated natural foods. The former are the encapsulated powders of flavors, fish oil and some functional oils such as MCT oil (middle chain triglyceride oil). They are applied to milk, soup and bakery products, because of their beneficial physiological functions. The latter are dehydrated milk powders, dried eggs and powdered soup. The encapsulated oil powder is commonly produced by spray drying of the emulsified oil (O/W emulsion), which is analogous to the technique used for the production of flavor powders. The storage stability of encapsulated oil (avoidance of oxidation of lipids) is extremely important because lipid oxidation leads to loss of nutritional value and undesirable off-flavor formation. It is essential to control the oxidative reaction when highly polyunsaturated fatty acids are involved. Oxidation mechanisms of the bulk oil are well documented, but the oxidation of the encapsulated lipid dispersed in the carbohydrate matrices is fairly complex. Influencing factors for encapsulated lipid oxidation are the type and amount of the wall matrices (carbohydrate and protein), emulsifier, emulsion drop size, storage humidity, and the storage temperature. 6.4.5.1 Spray Drying of Oil Emulsions The most interesting point concerning microencapsulated oils is the influence of the characteristics of the matrices on encapsulation efficiency, which is defined as the fraction (or percentage) of the encapsulated oil in the total amount of oil (sum of surface oil and encapsulated oil). The amount of surface oil is obtained by simply washing the powder with an organic solvent, such as hexane, prior to extraction. Then, the remaining amount of oil, which is the encapsulated oil, is determined by extraction from the interior of the powdery product. The amount of lipids washed by hexane from the powder (surface oil) is, therefore, a meaningful indicator for the quality of encapsulation; the smaller the amount of surface oil, the better the encapsulation. Based on this viewpoint, Imagi et al. (1990) evaluated the encapsulation ability of wall materials by drying single suspended droplets in equipment of the above-mentioned kind. Both gum arabic and gelatin were confirmed to be excellent wall materials with the highest encapsulation efficiency. However, maltodextrin and pullulane had no emulsifying activity, not being suitable for lipid encapsulation when used alone. Egg albumin and sodium caseinate had a high emulsifying performance, but they also had a poor ability for crust formation on the particle surface, so that the resulting particles were susceptible to oxidation during storage. It can be concluded that it is essential to find an optimal combination of emulsifier and wall material (Matsuno and Adachi, 1993). Numerous studies have been executed for the microencapsulation of seed oil (flaxseed oil, sea buckthorn seed oil, etc.), fish oil, polyunsaturated fatty acids (PUFAs) and milk fat (Partanen et al., 2008, 2005, 2002; Tuomasjukka et al., 2006; Minemoto  et al., 2002a, b; Ishido et al., 2002; F€aldt and Bergenstahl, 1995; Millqvist-Fureby, 2003;

6.4 Encapsulation and Microencapsulation of Enzymes and Oil by Spray Drying

Kim et al., 2002). The cross-sectional structure of the particles from SEM images revealed that small spherical enclosures of oil emulsion appeared in the shell region of the particle. Spray-dried particles were often of hollow structure. The size distribution of the reconstituted emulsion was changed and shifted to larger emulsion droplet sizes, suggesting that partial denaturation of the protein in drying decreases the solubility of proteins (Partanen et al., 2008). The surface composition of the spraydried particles was studied by ESCA (Partanen et al., 2008; Millqvist-Fureby, 2003; Kim  et al., 2002; F€aldt and Bergenstahl, 1995). The oil content near the surface of the particle can be calculated on the basis of the elemental ratio determined by ESCA. The oil/fat coverage on the particle surface depended on the melting temperature of the oil or fat  (F€aldt and Bergenstahl, 1995) and the particle size (Partanen et al., 2008). 6.4.5.2 Oxidation of Lipids Encapsulated in Spray-Dried Particles During the last two decades, numerous studies have focused on the impact of the relative humidity of the environmental air on the oxidative stability of encapsulated lipids (Partanen et al., 2008, 2005, 2002; Tuomasjukka et al., 2006; Minemoto et al., 2002a, b; Ishido et al., 2002). An excellent review on the oxidation of microencapsulated oils has been recently provided by Velasco et al. (2003). In previous sections, the release rate and the oxidation rate of encapsulated flavor were shown to be closely related to the relative humidity – the relationship being very complicated and depending on the glass transition temperature of the wall materials. The effects of relative humidity on the oxidation of oil described in the literature are in agreement with a stability map proposed by Labuza (1971), which suggests a high rate of oxidation at both low and high values of relative humidity (Minemoto et al., 2001; Partanen et al., 2008). This behavior is illustrated in Fig. 6.38. Partanen et al. (2008) explained the high rate of oxidation at very dry conditions by cracking of the wall matrix that enhances oxygen diffusion. At high relative humidity, cavities promoting

POV of flaxseed oil, m-eq (kg oil)–1

300 250 200 150 100 50 0 0

20

40

60

80

100

Water activity (%) Fig. 6.38 Influence of water activity (relative humidity) on the peroxide value (POV) of bulk flaxseed oil (*) and of flaxseed oil encapsulated in spray-dried WPI ( ) after 9 w of storage at 37  C.

.

j279

j 6 Morphology and Properties of Spray-Dried Particles Unoxidized fraction of linoleic acid (–)

280

1.0 0.8 0.6

0.4 0.2 0

0

5

10

15

20

25

30

35

Storage time (d) Fig. 6.39 Effect of the emulsion size on the oxidation process of linoleic acid encapsulated in spraydried powders; Storage at 37  C and 12% of relative humidity. Emulsion droplet size: * 1.04 mm, ~ 1.44 mm, & 2.13 mm.

the oxygen transfer may be formed between aggregations of oil globules. The size of the oil emulsion affects the stability of linoleic acid encapsulated in spray-dried powder (Minemoto et al., 2002a). The smaller the emulsion drop size, the slower the oxidation, as shown in Fig. 6.39, because finer emulsion droplets can more easily be embedded inside dried particles. The oxidation process of linoleic acid encapsulated in spray-dried particles was simulated numerically, based on the assumption that the bulk linoleic acid can be oxidized following the autocatalytic reaction scheme. The oxidation rate constant was related to the free energy of activation according to Eyring’s theory. The numerical results correlated well with the experimental oxidation data (Ishido et al., 2002).

6.5 General Quality Aspects

Experimental techniques for the characterization of drying processes and dried products have been presented in great detail in Volume 2 of this series. Additionally, several specific quality aspects of spray-dried powders have been discussed extensively in the foregoing sections of this chapter. This discussion will be concluded by brief reference to some less specific, general quality features of various spray-dried products. 6.5.1 Porosity of Spray-Dried Particles and Species Distribution

The bulk density of a powder, which often is an important product quality aspect, depends, among others, on the inter-particulate porosity. The porosity of particle

6.5 General Quality Aspects

layers and the pore size distribution are frequently obtained by mercury porosimetry (Washburn, 1921; Dullien and Batra, 1970). There is a distinct increase in the penetration pressure when the liquid metal enters from inter-particulate voids (void spaces between the spray-dried particles) into the intra-particulate openings (pores within the particles) as these are smaller by usually at least one order of magnitude. Besides this overall assessment of porosity, local examinations and visualizations are also possible by means of X-ray tomography, as shown in Lim and Barigou (2004) and Br€ockel et al. (2007, Vol. 2). The local composition of individual elements within an agglomerate system is only partly visible in SEM, but can be verified by EDX (energy dispersive X-ray method) and EDXRF (energy dispersive X-Ray fluorescence specroscopy), see for example, van Grieken and Markowicz (2002) and Beckhoff et al. (2006). 6.5.2 Strength of Particles and Attrition

The strength and hardness of particles can be obtained by single particle measurement of the yield stress with an indentometer, see for example, McGlinchey (2005), or with a simple pressure test device (Antonyuk, 2009; Grauherr, 1993). The particle strength is of special importance in the ceramics industry, where particles have to be subsequently pressed in molds to form larger and compact specimens which later undergo thermal treatment such as sintering (Br€ ockel et al., 2007, Vol. 1; Sheng et al., 2004; Samimi et al., 2005; Rahaman, 1995; Salmang and Scholze, 2007). The lower the porosity of the specimen after the compression, the lower the residual porosity after the sintering process (Agniel, 1992). To obtain a dense structure during the compression step, the spray-dried particles must considerably yield and deform at the contact points. A similar situation can be found when tablets of pharmaceutical materials are formed from spray-dried particles (Alderborn and Nystr€ om, 1996; Huntington, 2004; Yaginuma et al., 2007). Also here, a certain compressibility of the spray-dried particles is required, together with a good flowability of the feed powder. The deformation of the feed particles at contact points is necessary in order to obtain tablets with uniform strength and porosity. Other applications such as spray-dried catalysts require very high strength and abrasion resistance of the particles, even at high temperatures. Spray-dried catalysts have, typically, a particle size in the range 60 to 80 mm, corresponding to their frequent use in fluidized beds or in pneumatic transportation systems designed to serve as reactors. High strength can be achieved by layers of abrasion-resistant additives in the spray-dried slurry, such as silicon dioxide, in molecular or submicron size, that migrates to the shell of the particles during the spray-drying step. However, a certain porosity of the outer layer is also important for the mass transfer of reactants to the active sites within the pore system of the spray-dried particles. At least large pores must be kept open to provide access to deeper layers inside the particles. For the design of catalyst particles see for example, Br€ ockel et al. (2007, Vol. 2), Uihlein (1993), Kastner et al. (2001). Different tests are available to measure the attrition behavior of particles. In the pharmaceutical industry, the friability of powders or granules is tested by tumbling

j281

j 6 Morphology and Properties of Spray-Dried Particles

282

the particles in a drum and checking for the resulting fines (ISO Tumble Drum Test 3271, 1975 and 1995). An alternative option is to operate an air jet sieve for a given time and air differential pressure, and to determine the loss of fines during that period. Another method applies pneumatic transportation systems and impact plates or cyclones to trigger abrasion in a well-defined manner (Antonyouk, 2006; Pitchumani et al., 2003; Schultz and Kleinebudde, 1995). 6.5.3 Bulk Density and Product Flowability

The bulk density of powders can be measured for example, with the Hosokawa powder characterization tester (www.hmicronpowder.com, McGlinchey (2005)). The flowability of powders can be quantified either in shear cells at considerably low normal pressure (Schwedes and Schulze, 1990) or with small standardized storage bins taking the mass flow over time. The angle of repose, when the sprayed powder leaves a standardized laboratory bin, is also an indicator of the flowability of powders (see e.g., DIN ISO 4490, McGlinchey (2005), Wouters and Geldard (1996)). More recent methods are the agitated cell with fluidization (“Powder Flow Analyzer” by Stable Microsystems: www.stablemicrosystems.com) and the already mentioned versatile powder characterization tester provided by Hosokawa. High flowability is needed to feed the particles into fast-running tableting machines. The dosing system of tableting machines relies fully on the volumetric particle feeding system to obtain uniform tablet masses as superficial feed material is removed from the molds before compression. Even so, additives may help to improve the flowability of the particles (Yaginuma et al., 2007). The bulk density of the particles, therefore, must be maintained within narrow limits. The bulk density can be determined as “poured bulk density,” just by pouring the particles from above into a cylinder (DIN EN 725-9, 2006; DIN EN 1236, 1995), or as “tapped density,” when this cylinder is tapped several times before measurement (DIN EN 725-8, 2006). To obtain the right volume the surplus particles are best removed by leveling the surface with a straight edge. Sometimes also the “aerated density” is needed for the design of fluidized beds. It may be determined with, for example, the “Powder Flow Analyzer.” Spray dried particles with a grain size of 60 to 100 mm lie within the optimal operation range of fluidized beds (McGlinchey, 2005). 6.5.4 Residual Moisture

Usually the residual moisture is obtained by keeping the particles for a given time in a heating chamber at specified air temperature and humidity, and recording the weight loss by removal of volatiles. A faster measurement is obtained with infrared irradiation directly on a drying balance. However, frequently the equilibrium moisture depends on the temperature of the stove and on the air humidity according to a product-specific sorption isotherm. This is true for most organic materials. Moreover, the transition between different crystal hydrate states and crystal water release has to be considered, as they depend on temperature.

6.5 General Quality Aspects

In some cases the chemical method given by Karl–Fischer titration can also be applied for residual water determination. This method is based on oxidizing methyl sulfite with iodine into colorless iodide and methyl sulfate. The reaction only takes place when free water is available in the system. Additional, recently developed methods for the determination of residual water content are the near infrared (NIR) absoption method and nuclear magnetic resonance (NMR). The NMR method has been successfully used for the quick determination of residual water with high spatial and temporal resolution in bulk materials such as granules, applying a magnetic field gradient approach (Bl€ umich, 2005). 6.5.5 Reconstitution Behavior

The reconstitution behavior of porous particles (their so-called “instant properties”) can be explained by the fact that the solvent has to penetrate first into the pores of the particles in order to loosen contact forces by, for example, dissolving contact bridges formed by crystals or amorphous material between primary particles (Pfalzer et al., 1973; Schubert, 1975) or by reducing inter-particulate van der Waals forces. The penetration time of the liquid into a porous spray-dried particle with diameter d (wetting time) is given by the well-known Washburn equation (Washburn, 1921; Schubert, 1990): twet ¼

8d2 ð1
eÞm edpr s cosqeff

ð6:12Þ

Here, dpr is the Sauter mean diameter of the primary particles originally dispersed in the feed slurry of the spray drier, s is the surface tension of the liquid, e is the porosity of the spray dried particle, m is the viscosity of the penetrating liquid, and qeff is the mean wetting angle within the pore system. The latter is always larger than the local contact angle q. One can expect fast penetration and high dissolving velocities when the contact angle is small (good wetting properties). Larger pores lead to faster penetration. However the porosity must not exceed a certain limit, typically e  0.5, otherwise the capillary transport may be interrupted within the pore system. The wetting process of powder beds, which as a precondition for the dispersion of the powder particles in a liquid, is, in a similar way as before, determined by the course of liquid penetration into the powdery material. The kinetics of liquid penetration can be measured either by monitoring the weight of the powder layer during wetting or by observing the movement of the liquid front (Teipel, 2005). The liquid front can be more easily observed when the powder layers are prepared with an inclined surface and brought into contact with the liquid from below. The propagation of the wetting boundary is visible in the form of a moving dark-color border and can be documented by photographic means (Wollny and Schubert, 1990). For practical assessment of the reconstitution behavior of spray-dried particles or powders an immersion test has been developed. The particle layer is suddenly

j283

j 6 Morphology and Properties of Spray-Dried Particles

284

contacted from below with the liquid by removing a sheet at its bottom. The height of the powder layer surface is continuously measured giving information about the simultaneous wetting and sinking behavior of the particles due to capillary effects and gravity (Hogekamp and Pohl, 2004). Re-agglomeration of spray-dried particles often provides improved reconstitution behavior. This can be achieved by use of counter-currently operated spray dryers, as in the detergent industry (R€ahse, 2007), with integrated fluidized beds (Herbener, 1987; Masters, 1991) or even with a subsequent external agglomeration step. The dissolving and dispersing behavior of complexly formulated particles has to be measured in laboratory devices either by optical particle analysis, or by chemical analysis of dissolved matter within the solution or suspension. For that purpose, the spray particles are positioned in baskets within stirred containers, and the light extinction is measured in the suspension or the concentration of dissolved chemical species is recorded versus time (Schubert, 1990; Hogekamp and Pohl, 2004). Similar systems are applied to assess the release functions of pharmaceuticals (Martin and Leuenberger, 2002; Schultz and Kleinebudde, 1997).

6.6 Concluding Remarks

It has been shown that the particle formation and the final shape of particles are strongly related to the materials contained in the liquid feed to the spraying and drying system. Besides the materials introduced as a suspension, solution or as a complex dispersion, the drying conditions, and the preceding drop size dominating the evolving particle size play dominant roles in the formation of the structure and in providing a high final biochemical or chemical product quality. This is obvious, when drying complex molecules such as proteins, or when high retention is desired, when drying flavors. Protective agents may help in maintaining the high quality of the product by protecting the structure of sensitive species. However, the temperature of the drying air often also has to be kept within limits in order to avoid unwanted oxidation despite increased heat losses. The spraying process as well as the flow within industrial drying towers is usually subject to stochastic fluctuations, mainly caused by turbulence. This leads to non-uniform trajectories and drying histories of individual particles with different sizes and may also cause agglomeration. It will be an incentive in future to overcome unwanted fluctuations as well as to provide sprays with narrow drop size distribution as a prerequisite for high quality uniform products with largely identical particle morphologies.

Additional Notation Used in Chapter 6

BM C E

Spalding transfer number concentration Young’s modulus

– kg m
3 N m
2

6.6 Concluding Remarks

G k kd kR kx k1 R Tg

free energy Boltzmann constant deactivation rate constant release rate oxidation rate constant relaxation rate constant retention glass transition temperature

J kg
1 J K
1 s
1 various s
1 s
1 –  C

Greek Letters

f n t

zeta potential Poisson’s ratio relaxation time

V – s

Subscripts

B d d pr S shell sp

buckling droplet deactivation primary particle surface shell suspension particle

Abbreviations

ADH BSA CD CFD CLSM CMC CRP DE DHA EDX EDXRF EPA ESCA FCC FT-IR GA KWW

alcohol dehydrogenase bovine serum albumin cyclodextrin computational fluid dynamics confocal laser scanning microscopy carboxymethylcellulose constant rate period dextrose equivalent docosahexaenoic acid energy dispersive X-ray spectroscopy energy dispersive X-ray fluorescence spectroscopy eicosapentaenoic acid electron spectroscopy for chemical analysis fluid catalytic cracking Fourier transform infrared spectroscopy gum Arabic Kohlrausch–Williams–Watts

j285

j 6 Morphology and Properties of Spray-Dried Particles

286

Lg MCT MD NIR NMR O/W POV PTR-MS PUFA RH RM rhGH SBS SE-HPLC SEM SSPS VPO WLF WPI

lactoglobulin middle chain triglyceride maltodextrin near-infrared nuclear magnetic resonance oil in water peroxide value proton transfer reaction mass spectrometry polyunsaturated fatty acid relative humidity randomly methylated human growth hormone styrene-butadiene-styrene size-exclusion high-performance liquid chromatography scanning electron microscopy soluble soybean polysaccharide vanadium phosphate oxide William–Landel–Ferry whey protein isolate

References Abramzon, B., Sirignano, W. A., 1989. Droplet vaporization model for spray combustion calculations. Int. J. Heat Mass Transfer 32: 1605–1618. Adhikari, B., Howes, T., Bhandari, B. R., Troung, V., 2003. Surface stickiness of drops of carbohydrate and organic acid solutions during convective drying: Experiments and modeling. Drying Technol. 21(5): 839–873. Adhikari, B., Howes, T., Leocomte, D., Bhandari, B. R., 2005. A glass transition temperature approach for the prediction of the surface stickiness of a drying droplet during spray drying. Powder Technol. 149: 168–179. Adler, M., Lee, G., 1999. Stability and surface activity of lactate dehydrogenase in spray-dried trehalose. J. Pharm. Sci. 88: 199–208. Agniel, Y., 1992. Bedeutung der Einzelgranalieneigenschaften zur Defektvermeidung in trockengepressten keramischen Modellpulvern. Diss., TH Karlsruhe, Germany.

Alamilla-Beltr
an, L., Chanona-P
erez, J. J., Jim
enez-Aparicio, A. R., Guti
errez-L
opez, G. F., 2005. Description of morphological changes of particles along spray drying. J. Food. Eng. 67: 179–184. Alderborn, G., Nystr€om C., 1996. Pharmaceutical powder compaction. Marcel Dekker, New York, USA. Anandaraman, S., Reineccius, G. A., 1986. Stability of encapsulated orange peel oil. Food Technol. 40(11): 88–94. Anese, M., Manzocco, L., Maltini, E., 2005. Effect of coffee physical structure on volatile release. Eur. Food Res. Technol. 221: 434–438. Ansari, A., Jones, C. M., Henry, E. R., Hofrichter, J., Eaton, W. A., 1992. The role of solvent viscosity in the dynamics of protein conformational-changes. Science 256(5065): 1796–1798. Antonyouk, S., 2006. Deformations- und Bruchverhalten von kugelf€ormigen Granulaten bei Druck- und Stossbeanspruchung. Diss., Universit€at Magdeburg, Germany.

References Antonyuk, S., Heinrich, S., van Buijtenen, M., Deen, N. G., Kuipers, J., 2009. Feuchteeinfluss auf das Partikelstoßverhatten in Wirbelschichten. Chem Ing. Tech. 81: 1177– 1178. Barbosa-C
anovas, G. V., Ortega-Rivas, E., Juliano, P., Yan, H., 2005. Food powders: physical properties, processing, and functionality. Springer, New York, USA, pp. 199–220. Beckhoff, B., Kannengiesser, B., Langhoff, N., Wedell, R., Wolff, H., 2006. Handbook of practical X-ray fluorescence analysis. Springer, Berlin, Germany. Belton, P. S., Gill, A. M., 1994. IR and Raman spectroscopic studies of the inter-action of trehalose with hen egg white lysozyme. Biopolymers 34: 957–961. Beristain, C. I., Azuara, E., Vernon-Carter, E. J., 2002. Effect of water activity on the stability to oxidation of spray-dried encapsulated orange peel oil using mesquite gum (Prosopis Juliflora) as wall material. J. Food Sci. 67: 206–211. Bhandari, B. R., Dumoulin, E. D., Richard, H. M. J., Noleau, I., Lebert, A. M., 1992. Flavor encapsulation by spray drying: Application to citral and linalyl acetate. J. Food Sci. 57(1): 217–221. Bhandari, B. R., Howes, T., 1999. Implication of glass transition for the drying and stability of dried foods. J. Food Eng. 40: 71–79. Bird, R. B., Steward, W. E., Lightfoot, E. N., 2002. Transport phenomena. 2nd edn, John Wiley & Sons, New York, USA. Bl€ umich, B., 2005. Essential NMR for scientists and engineers. Springer, Berlin, Germany. Bomben, J. L., Bruin, S., Thijssen, H. A. C., Merson, R. L., 1973. Aroma recovery and retention in concentration and drying of foods, in Advances in Food Research (eds G. S. Stewart, E. M. Mrak, C. O. Chichester). Academic Press, New York, USA, Vol. 20: pp. 1–111. Brenn, G., Wiedemann, T., Rensink, D., Kastner, O., Yarin, A. L., 2001. Modeling and experimental investigation of the morphology of spray dried particles. Chem. Eng. Technol. 24(11): 1113–1116. Broadhead, J., Rouan, S. K. E., Rhodes, C. T., 1992. The spray drying of

pharmaceuticals. Drug Dev. Ind. Pharm. 18: 1169–1206. Br€ockel, U., Meier, W., Wagner, G., 2007. Product Design and Engineering. Vol. 1 and 2, Wiley-VCH. Weinheim, Germany. Carpenter, J., Crowe, J. H., 1984. An infrared spectroscopy study of the interactions of carbohydrates with dried proteins. Biochemistry 28: 3916–3922. Chang, Y. I., Scire, J., Jacobs, B., 1988. Effect of particle size and microstructure properties on encapsulated orange oil, in Flavor encapsulation (eds S. J. Risch, G. A. Reineccius). American Chemical Society, Washington DC, USA, pp. 87–102. Constantino, H. R., Firouzabadian, L., Wu, C., Carrasquillo, K. G., Griebenow, K., Zale, S. E., Tracy, M. A., 2001. Protein spray freeze drying, Part 2: Effect of formulation variables on particle size and stability. J. Pharm. Sci. 91: 388–395. Contractor, R. M., Bergna, H. E., Horowitz, H. S., Blackstone, C. M., 1987. Butane oxidation over vanadium phosphate catalysts. Catal. Today 1: 49–58. Coumans, W. J., Kerkhof, P., Bruin, S., 1994. Theoretical and practical aspects of aroma retention in spray drying and freeze drying. Drying Technol. 12: 99–149. Crowe, J. H., Crowe, L. M., Chapman, D., 1984. Preservation of membranes in anhydrobiotic organisms: The role of trehalose. Science 223 (4637): 701–703. Dicoi, O., Walzel, P., 1998. Vergleich verschiedener Modellans€atze zur Tropfenverdunstung von Einzeltropfen. Proceedings of Spray ’98, Essen, Germany. DIN EN 735-8, 2006. Bestimmung der geklopften Sch€ uttdichte. DIN EN 725-9, 2006. Bestimmung der Sch€ uttdichte. DIN EN 1236, 1995. Bestimmung der Sch€ uttdichte. DIN ISO 4490, Ermittlung des Fließverhaltens mit Hilfe eines kalibrierten Trichters. Dronen, D. M., Reineccius, G. A., 2003. Rapid analysis of volatile release from powders using dynamic vapour sorption atmospheric pressure chemical ionization

j287

j 6 Morphology and Properties of Spray-Dried Particles

288

mass spectrometry. J. Food Sci. 68: 2158–2162. Dullien, F. A. L., Batra, V. K., 1970. Determination of the structure of porous media. Ind. Eng. Chem. 62(10): 25–53. Eggers, J., Villermaux, E., 2008. Physics of liquid jets. Rep. Progr. Phys. 71: 1–79. El-Sayed, T. M., Wallack, D. A., King, C. J., 1990. Changes in particle morphology during drying of drops, Part 1: Effects of composition and drying conditions. Ind. Eng. Chem. Res. 29: 2346–2354. Etzel, M. R., Shen, S.-Y., Halverson, S. L., Budijono, S., 1996. Enzyme inactivation in a droplet forming a bubble during drying. J. Food Eng. 27: 17–34.  F€aldt, P., Bergenstahl, B., 1995. Fat encapsulation in spray-dried food powders. J. Am. Oil Chem. Soc. 72: 171–176. Finney, J., Buffo, R., Erinaceous, G. A., 2002. Effect of type of atomization and processing temperatures on the physical properties and stability of spray-dried flavors. J. Food Sci. 67 (3): 1108–1114. Flory, P. J., 1942. Thermodynamics of high polymer solutions. J. Chem. Phys. 10(51): 51–61. Franks, F., Hately, R. H. M., Mathias, S. F., 1994. Material science and the production of shelf-stable biologicals. Bio. Pharm. 4: 253–256. Fries, M., 2008. Produktgestaltung keramischer Spr€ uhgranulate f€ ur die uniaxiale Pressverdichtung zu großformatigen Bauteilen. Diss., TU Bergakademie Freiberg, Germany. Furuta, T., Tsujimoto, S., Okazaki, M., Toei, R., 1984. Effect of drying on retention of ethanol in maltodextrin solution during drying of a single droplet. Drying Technol. 2: 311–327. Furuta, T., Hayashi, H., Ohashi, T., 1994. Some criteria of spray dryer design for food liquid. Drying Technol. 12: 151–177. Gacular, M. C., Jr., Kubala, J. J., 1975. Statistical models for shelf life failures. J. Food Sci. 40: 404–409. Garmise, R. J., 2007. Novel dry powder preparations of whole inactivated influenza virus for nasal vaccination. APS Pharm. Sci. Tech. 8(4): Article 81, E1–E9. Gibbs, B. F., Kermasha, S., Alli, I., Mulligan, C. N., 1999. Encapsulation in the

food industry: A review. Int. J. Food Sci. Nutr. 50: 213–224. Giunchedi, P., Conte, U., 1995. Spray-drying as a preparation method of micro-particulate drug delivery systems: An overview. STP Pharma Sci. 5: 276–290. Gnielinski, V., Mersmann, A., Thurner, F., 1993. Verdampfung, Kristallisation, Trocknung. Vieweg Verlag, Braunschweig, Germany. Gordon, M., Taylor, J. S., 1952. Ideal copolymers and the second-order transitions of synthetic rubbers, Part 1: Non-crystalline copolymers. J. Appl. Chem. 2: 493–500. Grauherr, B., 1993. Herstellung und Charakterisierung keramischer Wirbelschichtgranulate und ihr Einfluss auf das Gef€ uge und die Eigenschaften plasmagespritzter Schichten. Diss., RWTH Aachen, Germany. Greenwald, G. C., 1980. Particle morphology in the spray drying of foods. Diss., University of California, Berkeley, USA. Greenwald, G. C., King, C. J., 1981. The effects of design and operating conditions on particle morphology for spray-dried foods. J. Food Process Eng. 4: 171–187. Greenwald, G. C., King, C. J., 1982. The mechanism of particle expansion in spray drying of foods. AICHE Symp. Ser. Food Process Engineering 78(218): 101–110. Groenewold, C., M€oser, C., Groenewold, H., Tsotsas, E., 2002. Determination of single particle drying kinetics in an acoustic levitator. Chem. Eng. J. 86: 217–222. Gunning, Y. M., Gunning, P. A., Kemsley, E. K., Parker, R., Ring, S. G., Wilson, R. H., Blake, A., 1999. Factors affecting the release of flavor encapsulated in carbohydrate matrixes. J. Agric. Food Chem. 47: 5198–5205. Hadinoto, K., Kewu, Z., Tan, R. B. H., 2007. Drug release study of large hollow nanoparticulate aggregates carrier particles for pulmonary delivery. Int. J. Pharm. 341 (1–2): 195–206. Hecht, H. P., King, C. J., 2000a. Spray drying: Influence of developing drop morpholog on drying rates and retention of volatile substances, Part 1: Single drop experiments. Ind. Eng. Chem. Res. 39: 1756–1765.

References Hecht, J. P., King, C. J., 2000b. Spray drying: Influence of developing drop morphology on drying rates and retention of volatile substances, Part 2: Modeling. Ind. Eng. Chem. Res. 39: 1766–1774. Heldmann, D. R., 1974. An analysis of atmospheric freeze drying. J. Food Sci. 39: 147–155. Herbener, R., 1987. Staubarme Pulver durch Spr€ uhtrocknen – Erfahrungen mit neueren Techniken. Chem.-Ing.-Tech. 59: 112–117. Hogekamp, S., Pohl, M., 2004. Methoden zur Beurteilung des Befeuchtungs- und Dispergierverhaltens von Pulven. Chem.-Ing.-Tech. 76: 385–390. Huntington, D. H., 2004. The influence of the spray drying process on product properties. Drying Technol. 22: 1261–1288. Imagi, J., Kako, N., Nakanishi, K., Matsuno, R., 1990. Entrapment of liquid lipids in matrices of saccharides. J. Food Eng. 12: 207–222. Ishido, E., Hakamata, K., Minemoto, Y., Adachi, S., Matsuno, R., 2002. Oxidation process of linoleic acid encapsulated with a polysaccharide by spray-drying. Food Sci. Technol. Res. 8(1): 85–88. Ishido, E., Minemoto, Y., Adachi, S., Matsuno, R., 2003. Heterogeneity during autoxidation of linoleic acid encapsulated with a polysaccharide. J. Food Eng. 59: 237–243. Jørgensen, K., 2005. Drying rate and morphology of slurry droplets. Diss., Technical University of Denmark, Lyngby, Denmark. Karel, M., Langer, R., 1988. Controlled release of food additives, in Flavor encapsulation (eds S. J. Risch, G. A. Reineccius). American Chemical Society, Washington DC, USA, pp. 67–77. Kastner, O., Rensink, G., Brenn, G., Tropea, C., 2001. The acoustic tube levitator – a novel device for determining the drying kinetics of single droplets. Chem. Eng. Technol. 24: 335–339. Kawamura, Y., Nakanishi, K., Matsuno, R., Kamikubo, T., 1981. Stability of immobilized a-chymotrypsin. Biotechnol. Bioeng. 23: 1219–1236. Kilburn, D., Claude, J., Schweizer, T., Alam, A., Ubbink, J., 2005. Carbohydrate polymers

in amorphous states: An integrated thermodynamic and nanostructural investigation. Biomacromolecules 6: 864–879. Kim, Y. D., Morr, C. V., 1996. Microencapsulation properties of gum arabic and several food proteins: Spray-dried orange oil emulsion particles. J. Agric. Food Chem. 44: 1314–1320. Kim, E. H.-J., Chen, X. D., Pearce, D., 2002. Surface characterization for four industrial spray-dried dairy powders in relation to chemical composition, structure and wetting property. Colloids Surf. B 26: 197–212. King, C. J., 1995a. Spray drying: Retention of volatile compounds revisited. Drying Technol. 13: 1221–1240. King, A. H., 1995b. Encapsulation of food ingredients: A review of available technology, focusing on hydrocolloids, in Encapsulation and controlled release of food ingredients (eds S. J. Risch, G. A. Reineccius). American Chemical Society, Washington DC, USA, pp. 26–39. Koch, M., Walzel, P., 2001. Migration of solutes in drying suspension droplets. Proceedings of the Conference on Spray Drying and Related Processes, Dortmund, Shaker Verlag, Aachen, Germany. Lamprecht, A., Schafer, U. F., Lehr, C. M., 2000a. Characterization of microcapsules by confocal laser scanning microscopy: Structure, capsule wall composition and encapsulation rate. Eur. J. Pharm. Biopharm. 49: 1–9. Lamprecht, A., Schafer, U. F., Lehr, C. M., 2000b. Structural analysis of microparticles by confocal laser scanning microscopy. AAPS Pharm. Sci. Tech. 1(3): 10–19. Lamprecht, A., Schafer, U. F., Lehr, C. M., 2000c. Visualization and quantification of polymer distribution in microcapsules by confocal laser scanning microscopy (CLSM). Int. J. Pharm. 196: 223–226.  Landstr€om, K., Alsins, J., Bergenstahl, B., 2000. Competitive protein adsorption between bovine serum albumin and b-lactoglobulin during spray-drying. Food Hydrocolloids 14(1): 75–82. Labuza, T., 1971. Kinetics of lipid oxidation in foods. CRC Rev. Food Technol. 2: 335–405.

j289

j 6 Morphology and Properties of Spray-Dried Particles

290

Labrousse, S., Roos, Y., Karel, M., 1992. Collapse and crystallization in amorphous matrices with encapsulated compounds. Sci. Aliment. 12: 757–769. Leahy, M. M., Anandaraman, S., Bangs, W. E., Reineccius, G. A., 1983. Spray drying of food flavors, Part 2: A comparison of encapsulating agents for the drying of artificial flavors. Perfum. Flavorist 8: 49.52. Lee, G., 2002. Spray drying of proteins, in Rational design of stable protein formulations (eds J. F. Caprenter, M. C. Manning). Plenum Press, New York, USA, pp. 135–158. Levine, H., Slade, L., 1986. A polymer physico-chemical approach to the study of commercial starch hydrolysis products (SHPs). Carbohydr. Polym. 6: 213–244. Levi, G., Karel, M., 1995. The effect of phase transitions on release of n-propanol entrapped in carbohydrate glasses. J. Food Eng. 24: 1–13. Lim, K. S., Barigou, M., 2004. X-ray tomography of cellular food products. Food Research Int. 37(10): 1001–1012. Liou, J. K., 1982. An approximate method for nonlinear diffusion applied to enzyme inactivation during drying. Diss., Agricultural University Wageningen, The Netherlands. Liu, X. D., Atarashi, T., Furuta, T., Yoshii, H., Ohkawara, S., 2001. Microencapsulation of emulsified hydrophobic flavors by spray drying. Drying Technol. 19: 1361–1374. Lohner, H., Czisch, C., Schreckenberg, P., Fritsching, U., Bauckhage, K., 2005. Atomization of viscous melts. Atomization Spray 15(2): 169–180. Maa, Y.-F., Hsu, C. C., 1996. Effect of high shear on proteins. Biotech. Bioeng. 51: 458–465. Maa, Y.-F., Costantino, H. R., Nguyen, P.-A., Hsu, C. C., 1997. The effect of operation and formulation variables on the morphology of spray-dried protein particles. Pharm. Dev. Tech. 2: 213–223. Maa, Y.-F., Hsu, C. C., 1997. Protein denaturation by combined effect of shear and air-liquid interface. Biotech. Bioeng. 54: 503–512. Maa, Y.-F., Nguyen, P.-A., Hsu, S. W., 1998a. Spray-drying of air-liquid interface sensitive

recombinant human growth hormone. J. Pharm. Sci. 87: 152–159. Maa, Y.-F., Nguyen, P.-A., Sit, K., Hsu, S. W., 1998b. Spray-drying performance of a bench-top spray dryer for protein aerosol powder preparation. Biotech. Bioeng. 60: 503–512. Maas, H. G., Schaldach, G., Walzel, P., Urbanetz, N. A., 2009. Manufacturing of tailor made carrier particles for inhalation therapy by spray drying. Proceedings of the 11th Intern. Conference on Liquid Atomization and Spray Systems (ICLASS), Vail Colorado, USA. Martin, A. N., Leuenberger, H., 2002. Physikalische Pharmazie. 4th edn, Wissenschaftliche Verlagsges., Stuttgart, Germany. Masters, K., 1991. Spray drying handbook. 5th edn, Longman, Burnt Mills/Essex, UK. Mateus, M.-L., Lindinger, C., Gumy, J.-C., Liardon, R., 2007. Release kinetics of volatile organic compounds from roasted and ground coffee: Online measurements by PTR-MS and mathematical modelling. J. Agric. Food Chem. 55: 10117–10128. Matsuno, R., Adachi, S., 1993. Lipid encapsulation technology – techniques and applications to food. Trends Food Sci. Technol. 4: 256–261. McGlinchey, D., 2005. The characterization of bulk solids. Blackwell Publishing, Oxford, UK. McNamee, B. F., O’Riordan, E. D., O’Sullivan, M., 1998. Emulsification and microencapsulation properties of gum arabic. J. Agric. Food Chem. 46: 4551–4555. McNamee, B. F., O’Riordan, E. D., O’Sullivan, M., 2001. Effect of partial replacemnet of gum arabic with carbohydrates on its microencapsulation properties. J. Agric. Food Chem. 49: 3385–3388. Meerdink, G., van’t Riet, K., 1991. Inactivation of a thermostable-amylase during drying. J. Food Eng. 14: 83–102. Meerdink, G., 1993. Drying of liquid food droplets: Enzyme inactivation and multicomponent diffusion. Diss., Agricultural University Wageningen, The Netherlands. Menting, L. C., Hoogstad, B., 1967. Volatiles retention during the drying of aqueous

References carbohydrate solutions. J. Food Sci. 32: 87–90. Menting, L. C., Hoogstad, B., Thijssen, H. A. C., 1970. Diffusion coefficients of water and organic volatiles in carbohydrate-water systems. J. Food Technol. 5: 111–126. Millqvist-Fureby, A., 2003. Characterisation of spray-dried emulsions with mixed fat phases. Colloids Surf. B 31, 65–79. Minemoto, Y., Adachi, S., Matsuno, R., 2001. Effect of relative humidity during storage on the autoxidation of linoleic acid encapsulated with a polysaccharide by hotair-drying and freeze-drying. Food Sci. Technol. Res. 7: 91–93. Minemoto, Y., Hakamata, K., Adachi, S., Matsuno, R., 2002a. Oxidation of linoleic acid encapsulated with gum arabic or maltodextrin by spray-drying. J. Microencapsulation 19: 181–189. Minemoto, Y., Fang, X., Hakamata, K., Watanabe, Y., Adachi, S., Kometani, T., Matsuno, R., 2002b. Oxidation of linoleic acid encapsulated with soluble soybean polysaccharide by spray-drying. Biosci. Biotechnol. Biochem. 66: 1829–1834. Minoshima, H., Matsushima, K., Liang, H., Shinohara, K., 2001. Basic model of spray drying granulation. J. Chem. Eng. Japan. 34 (4): 472–478. Minoshima, H., Matsushima, K., Liang, H., Shinohara, K., 2002. Estimation of diameter of granule prepared by spray drying with fast and easy evaporation. J. Chem. Eng. Japan 35: 880–885. Monse, K., Linnepe, T., Groom, S., Walzel, P., 2001. Influence of nozzle type on particle formation in spray drying. Proceedings of the Conference on Spray Drying and Related Processes, Dortmund, Shaker Verlag, Aachen, Germany. Monse, K., Koch, M., Walzel, P., 2006. Migration der kontinuierlichen Phase beim Trocknen von Suspensionstropfen in einer Trocknungskammer. Chem.-Ing.-Tech. 78: 78–82. Moreau, D. L., Rosenberg, M., 1993. Microstructure and fat extractability in microcapsules based on whey proteins or mixtures of whey proteins and lactose. Food Struct. 12: 457–468. Mumenthaler, M., 2000. Atmospheric sprayfreeze drying: A suitable alternative in

freeze-drying technology. Int. J. Pharm. 72: 97–110. Nebelung, M., Fries, M., Kraft, T., 2008. Produktdesign von Pressgranulaten – Anforderungen und Realit€at, in Pulvermetallurgie: Neue Anforderungen, neue Produkte, neue Verfahren (ed. H. Kolaska). Heimdall Verlag, Witten, Germany, pp. 107–118. Nelson, K. A., Labuza, T. P., 1992. Relationship between water and lipid oxidation rates, in Water activity and glass transition theory (ed. A. J. St. Angelo). American Chemical Society, Washington DC, pp. 93–103. Onwulata, C. I., Smith, P. W., Cooke, P. H., Holsinger, V. H., 1996. Particle structures of encapsulated milk fat powders. Lebebsm.Wiss. u. Technol. 29: 163–172. Palzer, S., 2005. The effect of glass transition on the desired and undesired agglomeration of amorphous food powders. Chem. Eng. Sci. 60: 3959–3968. Partanen, R., Raula, J., Sepp€anen, R., Buchert, J., Kauppinene, E., Forssell, P., 2008. Effect of relative humidity on oxidation of flaxseed oil in spray dried whey protein emulsions. J. Agric. Food Chem. 56: 5717–5722. Partanen, R., Hakala, P., Si€ovall, O., Kallio, H., Forssell, P., 2005. Effect of relative humidity on the oxidative stability of microencapsulated sea buckthorn seed oil. J. Food Sci. 70: E37–E43. Partanen, R., Ahro, M., Hakala, M., Kallio, H., Forssell, P., 2002. Microencapsulation of caraway extract in b-cyclodextrin and modified starches. Eur. Food Res. Technol. 214: 242–247. Pfalzer, L., Bartusch, W., Heiss, R., 1973. € ber die physikalischen Untersuchungen u Eigenschaften agglomerierter Pulver. Chem.-Ing.-Tech. 45(8): 510–516. Pitchumani, R., Meesters, G. M. H., Scarlett, B., 2003. Breakage behavior of enzyme granules in a repeated impact test. Powder Technol. 130: 138–146. Rahaman, M. N., 1995. Ceramic processing and sintering. Marcel Dekker, New York, USA. R€ahse, W., 2007. Produktdesign in der chemischen Industrie. Springer, Berlin, Germany. Reineccius, G. A., Coulter S. T., 1969. Flavor retention during drying. J. Dairy Sci. 52(8): 1219–1223.

j291

j 6 Morphology and Properties of Spray-Dried Particles

292

Reineccius, G. A., Bangs, W. E., 1985. Spray drying of food flavors, Part 3: Optimum infeed concentrations for the retention of artificial flavors. Perfum. Flavorist 10: 27–29. Reineccius, G. A., 1988. Spray-drying of food flavors, in Flavor encapsulation (eds S. J. Risch, G. A. Reineccius). American Chemical Society, Washington DC, USA, pp. 55–66. Reineccius, G. A., 2005. Flavor chemistry and technology. 2nd edn, CRC Press, New York, USA, pp. 351–384. R
e, M. I., Liu, Y. J., 1996. Microencapsulation by spray drying: Influence of wall systems on the retention of the volatiles compounds. Proceedings of the 10th International Drying Symposium, Krak
ow, 541–549. R
e, M. I., 1998. Microencapsulation by spray drying. Drying Technol. 16: 1195–1236. R
e, M. I., 2006. Formulating drug delivery systems by spray drying. Drying Technol. 24: 433–446. Risch, S. J., 1995. Encapsulation: Overview of uses and techniques, in Encapsulation and controlled release of food ingredients (eds S. J. Risch, G. A. Reineccius). American Chemical Society, Washington DC, USA, pp. 2–7. Risch, S. J., Reineccius, G. A., 1988. Spraydried orange oil: Effect of emulsion size on flavor retention and shelf stability, in Flavor encapsulation (eds S. J. Risch, G. A. Reineccius). American Chemical Society, Washington DC, USA, pp. 67–77. Roos, Y., 1995. Phase transitions in foods, in Food science and technology – International series (No. 360), Academic Press, San Diego/ California, USA. Roos, Y., 2003. Phase and state transitions in food dehydration. Proceedings of the Symposium EUDrying 03, Crete, 153–168. Rosenberg, M., Kopelman, I. J., Talmon, Y., 1985. A scanning electron microscopy study of microencapsulation. J. Food Sci. 50: 139–144. Rosenberg, M., Kopelman, I. J., Talmon, Y., 1990. Factors affecting retention in spraydrying microencapsulation of volatile materials. J. Agric. Food Chem. 38: 1288–1294. Rulkens, W. H., Thijssen, H. A. C., 1972. The retention of organic volatiles in spraydrying aqueous carbohydrate solutions. J. Food Technol. 7(1): 95–105.

Rumpf, H., 1975. Mechanische Verfahrenstechnik. Carl Hanser Verlag, M€ unchen, Germany. Salmang, H., Scholze, H., 2007. Keramische Verfahrenstechnik. 7th edn, Springer, Berlin, Germany. Samimi, A., Hassanpour, A., Ghodiri, M., 2005. Single bulk compression of soft granules: Experimental study and DEM evaluation. Chem. Eng. Sci. 60 (14): 3993–4004. Sankarikutty, B., Sreekumar, M. M., Narayanan, C. S., Mathew, A. G., 1988. Studies on encapsulation of cardamon oil by spray-drying technique. J. Food Sci. Technol. 25: 325–355. € Schubert, H., 1975. Uber Grenzfl€achenvorg€ange in der Agglomerationstechnik. Chem.-Ing.-Tech. 47(3): 86. Schubert, H., 1990. Instantisieren pulverf€ormiger Lebensmittel. Chem.-Ing.Tech. 62(11): 892–906. Schultz, P., Kleinebudde, P., 1995. Determination of pellet friability by use of an air stream apparatus. Pharm. Ind. 57: 323–328. Schultz, P., Kleinebudde, P., 1997. A new multiparticulate delayed release system, Part 1: Dissolution properties and release mechanism. J. Contr. Rel. 47: 181–189. Schwedes, J., Schulze, D., 1990. Measurement of flow properties of bulk solids. Powder Technol. 61: 59–68. Sheng, Y., Briscoe, B. J., Maung, R., Rovea, M. J., 2004. Compression of polymer bound alumina agglomerates at the micro deformation scale. Powder Technol. 140: 228–239. Shinoda, K., 1978. Principles of solution and solubility. Marcel Dekker, New York, USA. Sirignano, W. A., 1999. Fluid dynamics and transport of droplets and sprays. Cambridge University Press, Cambridge, UK. Sivetz, M., Foote, H. E., 1963. Coffee processing technology. AVI Publishing Co., Westport/ Connecticut, USA. Shahidi, F., Han, X. Q., 1993. Encapsulation of food ingredients. Crit. Rev. Food Sci. Nutr. 33(6): 501–547. Soottitantawat, A., Yoshii, H., Furuta, T., Ohgawara, M., Forssell, P., Partanen, R., Poutanen, K., Linko, P., 2004a. Effect of

References water activity on the release characteristics and oxidative stability of d-limonene encapsulated by spray drying. J. Agric. Food Chem. 52: 1269–1276. Soottitantawat, A., Peigney, J., Uekaji, Y., Yoshii, H., Furuta, T., Ohkawara, M., Linko, P., 2004b. Structural analysis of spray-dried powders by confocal laser scanning microscopy. Proceedings of the Annual Meeting of IFT 2004, Las Vegas, 17-G-22. Soottitantawat, A., Yoshii, H., Furuta, T., Ohgawara, M., Linko, P., 2003. Microencapsulation by spray drying: Influence of emulsion size on the retention of volatile compounds. J. Food Sci. 68: 2256–2262. Soottitantawat, A., 2005. Influence of emulsion size and powder morphology on the microencapsulation and stability of spray-dried flavors. Diss., Tottori University, Japan. Soottitantawat, A., Peigney, J., Uekaji, U., Yoshii, H., Furuta, T., Ohgawara, M., Linko, P., 2007. Structural analysis of spray-dried powders by confocal laser scanning microscopy. Asia Pac. J. Chem. Eng. 2: 41–46. Teipel, U., 2005. Energetic materials. Wiley-VCH, Weinheim, Germany. Thijssen, H. A. C., Rulkens, W. H., 1968. Retention of aromas in drying food liquids. De Ingenieur 80: 45–56. Toei, R., Okazaki, M., Furuta, T., 1978. Drying mechanism of a non-supported state, in Proceedings of 1st International Drying Symposium (ed. A. S. Mujumdar). Science Press, Princeton, USA, pp. 53–59. Tuomasjukka, S., Kallio, H., Forssell, P., 2006. Effect of microencapsulation of dietary oil on postprandial lipemia. J. Food Sci. 71: S225–S230. Ubbink, J., Schoonman, A., 2003. Flavor delivery systems, in Kirk-Othmer encyclopedia of chemical technology. John Wiley & Sons, New York. Uihlein, K., 1993. Butanoxidation an VPOWirbelschichtkatalysatoren. Diss., Univ. Karlsruhe, Germany. van Grieken, R., Markowicz, A. A., 2002. Handbook of X-ray spectroscopy. Marcel Dekker, New York, USA. Vehring, R., 2007. Pharmaceutical particle engineering via spray drying. Pharm. Res. 25: 999–1022.

Velasco, J., Dobarganes, C., M
arquez-Ruiz, G., 2003. Variables affecting lipid oxidation in dried microencapsulated oils. Grasas y Aceites 54: 304–314. Verhey, J. G. P., 1972a. Vacuole formation in spray-dried powder particles, Part 1: Air incorporation and bubble expansion. Neth. Milk Dairy J. 26: 86–202. Verhey, J. G. P., 1972b. Vacuole formation in spray-dried powder particles, Part 2: Location and prevention of air incorporation. Neth. Milk Dairy J. 26: 203–224. Verhey, J. G. P., 1973. Vacuole formation in spray-dried powder particles, Part 3: Atomization and droplet drying. Neth. Milk Dairy J. 27: 3–18. Vrentas, J. S., Vrentas, C. M., 1996. Hysteresis effects for sorption in glassy polymers. Macromolecules 29(12): 4391–4396. Walton, D. E., Mumford, C. J., 1999a. Spray dried products – Characterization of particle morphology. Chem. Eng. Res. Des. 77(1): 21–28. Walton, D. E., Mumford, C. J., 1999b. The morphology of spray-dried products: The effect of process variables upon the morphology of spray dried particles. Chem. Eng. Res. Des. 77(5): 442–490. Walton, D. E., 2000. The morphology of spraydried particles – a qualitative view. Drying Technol. 18: 1943–1986. Walz, A., Mayer, M., 1966. Theoretische und experimentelle Untersuchung zur Herstellung von Stein- und Glaswolle. Glastechn. Berichte 8: 359–370. Walzel, P., 2001. Proceedings of the 1st Conference on Spray Drying. TU Dortmund, Shaker Verlag, Aachen, Germany. Wang, Z. L., Finlay, W. H. et al. (2006) Powder formation by atmospheric sprayfreeze-drying. Powder Technol. 170: 45–52. Washburn, W. W., 1921. The dynamics of capillary flow. Proc. Phys. Rev. 17: 273–283. Wendel, S., Celik, ¸ M., 1997. An overview of spray-drying applications. Pharm. Technol. 21: 124–156. Whorton, C., Reineccius, G. A., 1995. Evaluation of the mechanisms associated with the release of encapsulated flavor materials from maltodextrin matrices, in

j293

j 6 Morphology and Properties of Spray-Dried Particles

294

Encapsulation and controlled release of food ingredients (eds S. J. Risch, G. A. Reineccius). American Chemical Society, Washington DC, USA, pp. 143–160. Williams, G., Watts, D. C., 1970. Nonsymmetrical dielectric relaxation behavior arising from a simple empirical decay function. Trans. Faraday Soc. 66: 80–85. Williams, M. L., Landel, R. F., Ferry, J. D., 1955. The temperature dependence of relaxation mechanisms in amorphous polymers and other glass-forming liquids. J. Am. Chem. Soc. 77: 3701–3707. Wollny, M., Schubert, H., 1990. Eine neue Methode zur Bestimmung des zeitlichen Benetzungsverhaltens von Partikelsch€ uttungen. Proceedings of DECHEMA Jahrestagung, Wiesbaden, Vol. 2, 369. Wouters, I. M. F., Geldard, D., 1996. Characterizing semi-cohesive powders using angle of repose. Part. Syst. Charact. 13: 254–259. Yamamoto, S., Sano, Y., 1992. Drying of enzymes: Enzyme retention during drying of a single droplet. Chem. Eng. Sci. 47: 177–183. Yamamoto, S., Sano, Y., 1994. Drying of carbohydrate and protein solutions. Drying Technol. 12: 1069–1080. Yaginuma, Y., Ozeki, Y., Kakizawa, M., Gomi, S. I., Watanabe, Y., 2007. Effects of powder flowability on die-fill properties in rotary compression. J. Drug Delivery Sci. Technol. 17: 205–210. Yoshii, H., Soottitantawat, A., Liu, X.-D., Atarashi, T., Furuta, T., Aishima, S., Ohgawara, M., Linko, P., 2001. Flavor

release from spray-dried maltodextrin/gum arabic or soy matrices as a function of storage relative humidity. Int. Food Sci. Emerg. Technol. 2: 55–61. Yoshii, H., Furuta, T., Fujiwara, M., Linko, P., 2003. Oxidation stability of powdery ethyl eicosapentaenoate included in cyclodextrins and polysaccharide/ cyclodextrin mixtures. Jpn. J. Food Eng. 4: 25–30. Yoshii, H., Ohashi, T., Kugimoto, Y., Furuta, T., Linko, P., 2005. Stability of alcohol dehydrogenase (ADH) during a suspended droplet drying and storage. Drying Technol. 23: 1217–1227. Yoshii, H., Soottitantawat, A., Iida, K., Ohgawara, M., Furuta, T., 2006. Effect of additives on the formation of hollow particles in spray-dried powder. Proceedings of 15th International Drying Symposium (IDS2006), Budapest, Volume C, 1339–1344. Yoshii, H., Kawamura, D., Neoh, T.-L., Furuta, T., 2007. Proceedings of 5th Asia-Pacific Drying Conference, Hong Kong, 317–322. Yoshii, H., Buche, F., Takeuchi, N., Terrol, C., Ohgawara, M., Furuta, T., 2008. Effects of protein on retention of ADH enzyme activity encapsulated in trehalose matrices by spray drying. J. Food Eng. 87: 34–39. Zbicinski, I., Delag, A., Strumillo, C., Adamiec, J., 2002a. Advanced experimental analysis of drying kinetics in spray drying. Chem. Eng. J. 86: 207–216. Zbicinski, I., Strumillo, C., Delag, A., 2002b. Drying kinetics and particle residence time in spray drying. Drying Technol. 20: 1751–1768.

j295

7 Particle Formulation in Spray Fluidized Beds Mirko Peglow, Sergiy Antonyuk, Michael Jacob, Stefan Palzer, Stefan Heinrich, and Evangelos Tsotsas 7.1 Introduction

Chapter 6 has shown that particles with desired properties can be produced by spray drying. In spray drying, the formation of the particles is a result of evaporation of the solvent from a solution or suspension. Product quality depends on the properties of the materials used, the operating conditions, and – to a certain extent – also on the equipment design. Instead of spraying the droplets into an otherwise empty device, one can also spray them onto fluidized particles. In this way, a large family of spray fluidized bed formulation processes is obtained. Since these are wet processes, they are intimately coupled to, and strongly dependent on drying. Similarly to spray drying, product quality depends on material properties, operating conditions and equipment design. This creates, on the one hand, numerous opportunities for manipulating product properties in the desired direction. On the other hand, due to the large number of interacting involved parameters, it makes process design difficult. Even the observation and characterization of the particle system pose serious challenges. Recent progress in experimental methods for observation and characterization of the particle system were presented in Chapter 5 of Volume 2 of this series. As for design methods, it has been pointed out, in Chapter 6 of Volume 1 of this series, that population balance equations can be used. However, such equations contain kinetic terms for birth, growth and death of particles, which must be reliably known for practical application. Moreover, they are continuous, macro-scale representations that – by definition – cannot directly account for micro-scale physics. It is, therefore, easy to understand that the goal of the present chapter is not to offer a complete theory of spray fluidized bed processes – which does not yet exist – but to provide an overview of the processes and the opportunities that they offer, highlighting some recent advances. First, general principles will be discussed and terminology concerning agglomeration, granulation and coating will be defined in Section 7.2. Then, the influences of material properties, operating conditions and apparatus

Modern Drying Technology Volume 3: Product Quality and Formulation, First Edition. Edited by Evangelos Tsotsas and Arun S. Mujumdar.
2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.

j 7 Particle Formulation in Spray Fluidized Beds

296

design will be treated with the help of selected examples in Sections 7.3, 7.4 and 7.5, respectively. The discussion of the influence of material properties (Section 7.3) will focus on why particles adhere to each other, how strong they become, how they break, and how primary particle properties can be considered in population balance modeling. Section 7.4 will exemplify the influence of operating parameters on the properties of particles produced by granulation. Section 7.5 will first identify aspects of apparatus design that can have a serious influence on product quality. This general discussion will then be exemplified by analyzing the influence of apparatus design on the residence time distribution of the solids, which, in turn, has an influence on, for example, deactivation of ingredients or particle size distribution (PSD). In a second example, we will show that different PSDs can be obtained even in devices that – at first glance – look like perfectly mixed vessels. Finally, discrete particle modeling (DPM) will be introduced as a promising method for connecting apparatus design to the process and the product. Concerning this connection, DPM is not the only possible method. Apart from the already mentioned population dynamics, empirical approaches and stochastic discrete models are also possible. The former will be treated in Section 7.6 with an example of aroma encapsulation, the latter in Section 7.7, focusing on the unique potential of discrete Monte Carlo simulations to explain the influence of drying on the kinetics of spray fluidized bed agglomeration. Finally, Section 7.8 will give a short summary and an outlook.

7.2 General Principles of Particle Formulation in Spray Fluidized Beds

Various definitions and names are used, in the scientific literature and in industrial applications, for spray fluidized bed processes and for the products which are manufactured using these processes. To clarify the terminology, Fig. 7.1 is used here. Figure 7.1 distinguishes between agglomeration, granulation and coating by means of schemes for the underlying principles and the resulting product structures. Agglomeration is, in terms of population balances, an aggregation that combines small primary particles to give larger structures (Fig. 7.1a). A binding mechanism is necessary for agglomeration to happen. In the simplest case, sprayed water (or solvent) dissolves the surface of the primary particles. Collisions between the particles create liquid bridges, which can – due to the dissolved material – be transformed to solid bridges by drying. Weak solidified bridges are broken by further collisions, whereas those which are strong enough survive and provide the final product. If the primary particles are not soluble, binder must be added to the sprayed liquid. Further adhesion and bridge formation mechanisms are possible and will be discussed in Section 7.3.1. In granulation (Fig. 7.1b), the atomized liquid spreads on the surface of fluidized particles and creates, depending on the thermal conditions and material properties, a more or less uniform liquid layer, which dries and solidifies. Repetition leads to successive layering of solidified shells and, finally, to an onion-like structure. The process, which is a growth process, starts with small particles (seeds) produced

7.2 General Principles of Particle Formulation in Spray Fluidized Beds

(a)

spraying

binder droplet

(b)

spraying

spray droplet

(c)

powder

seeds

spraying

coating droplet

wetting

drying

agglomerate

liquid bridges

solid bridges

„snowball“ structure

spreading

drying

grain

liquid layer

solidified shell

„onion“ structure

spreading, drying

coated particle

wetting

particle, carrier

Fig. 7.1 Basic principles of spray fluidized bed processes: (a) agglomeration, (b) granulation (layering), (c) coating.

internally (by particle breakage or attrition, or by droplets dried during their flight from the atomization nozzle to the bed – the so-called overspray) or externally (by sieving-milling-circuits that recycle undersized or milled oversized fractions). Consequently, both seeds and granulated particles consist of the material dissolved in the spray. The properties of particles produced by layering granulation are compared to the properties of agglomerates in Tab. 7.1. Coating (Fig. 7.1c) is essentially the same as layering granulation, with the difference being the placing of a more or less thin layer of another material on fluidized cores (carrier particles). The coating can increase the stability of the core during storage, prevent the uptake of moisture or the loss of volatiles, mask taste or

j297

j 7 Particle Formulation in Spray Fluidized Beds

298

Tab. 7.1

Properties of particles produced by agglomeration or granulation.

Property

Agglomeration

Granulation

absence of dust free flow dosing re-dispersion instant behavior solubility tabletting porosity bulk density surface hygroscopicity sphericity PSD abrasivity

good good good good excellent

good excellent good good good

easy high low open

low to very low high rather closed reduced high rather narrow low

medium rather wide rather high

odor, or adapt the surface structure, appearance and solubility. Typically, the spray droplets are smaller, the viscosity of the spray liquid is lower, and the fluidization conditions are more turbulent in coating than in layering granulation. Both processes compete with agglomeration, because droplets can dry after their deposition on particles, or they can form liquid bridges. The transition between agglomeration and layering is usually described by the so-called Stokes criterion, which will be discussed in Section 7.7. A possibility to deposit large amounts of coating material in a short period of time is provided by the so-called powder layering or dry coating process. Here, the coating material is fed in powder form and distributed on the (much larger) carrier particles, as shown in Fig. 7.2. Layer uniformity and the avoidance of agglomeration of powder particles with each other are important product quality issues that require special apparatus design. spraying

spray droplet

powder

rolling-up

liquid bridges

solid bridges

drying, solidifying

solidified shell

particle, carrier Fig. 7.2 Basic principle of powder layering (dry coating).

pellet

7.3 Influence of Material Properties Tab. 7.2

Grains vs. powder vs. liquids.

Grains compared to Powders

Powdery Solids compared to Liquids

less dust improved flowing properties easy to dose easy to transport decreased handling hazards

less volume less weight no sedimentation or settlement easy to ship

All the discussed processes can, in principle, also be operated by spraying a melt instead of a solution or suspension. Hot melt coating, granulation or agglomeration has, if possible in terms of the materials to be used, energetic advantages. The reason is that solid phase formation takes place by cooling the melt, without the need to evaporate water or another solvent. In the case of coating, higher amounts of material can be deposited, and carrier particles sensitive to water or solvent can be treated. Irrespective of how they have been produced, granular solids have some typical advantages in comparison to powders, and powdery solids have typical advantages in comparison to liquids. These advantages are summarized in Tab. 7.2. Despite Tab. 7.2, powders can be the preferred product form for various applications, due, for example, to their high volume-specific surface area and the resulting high reactivity. Even ultra-fine powders can be produced in spray fluidized beds. For this purpose, one follows Fig. 7.1c. However, the coating is created with the sole purpose of its subsequent in situ destruction by particle–particle collisions. The material created by attrition is now the real product of the process, which is carried out of the fluidized bed by the gas and can be collected in a cyclone or in filter bags. The core particles are carriers for the coating and, simultaneously, the promoters of its destruction, so that they have to be heavy and rigid. Metallic carrier particles can be heated by wall contact or by induction, so that the creation of ultra-fine powder can be combined with its thermal treatment.

7.3 Influence of Material Properties 7.3.1 Adhesion Mechanisms and Mechanical Strength of Agglomerates

Composition and process conditions determine the nature of the adhesion forces holding the primary particles of agglomerates together. The following adhesion forces are relevant for agglomeration during drying: . . .

Van der Waals forces increasing after deformation of plastic particle surfaces, Capillary forces generated by liquid bridges between particles, Viscous forces in sinter bridges between amorphous particles.

j299

j 7 Particle Formulation in Spray Fluidized Beds

300

The different adhesion forces are highly dependent on the material properties which are a function of the material structure. Thus differences in material structure and properties are explained before the three mentioned adhesion principles are discussed.

7.3.1.1 Material Structure and Properties In view of adhesion principles and their dependence on the material properties one can distinguish mainly two different groups of water-soluble materials: amorphous and crystalline substances (Palzer, 2005, 2009). In an amorphous matrix the molecules are arranged statistically (as in a liquid), whereas the molecules in crystalline structures are highly ordered in repeated three-dimensional geometrical patterns. The two substance groups are characterized by different material properties such as hygrocapacity, hygrosensitivity, and mechanical properties. The hygrocapacity quantifies the ability to bind water by absorption into the molecular matrix and adsorption on the free surface area. The hygrosensitivity of a material describes to what extent its viscosity is affected by the absorbed water. In terms of hydrocapacity and hydrosensitivity both material groups exhibit completely different behavior. While increasing the relative humidity of the surrounding air, crystalline materials do not absorb significant quantities of water. Only a limited amount of crystal water is embedded in the molecular matrix. Once a substance-specific relative air humidity (or water activity) is reached, the crystals dissolve layer by layer. In contrast, amorphous water-soluble substances absorb significant quantities of water when exposed to an increasing relative humidity of the surrounding air. The amount of absorbed water is a function of the water activity that can be described by various types of isotherm equations (Guggenheim, Anderson and de Boer (GAB), Brunnauer, Emmett and Teller (BET), see Weisser (1986)). Water included in crystalline structures in the form of crystal water will not change significantly the mechanical properties of the crystals. Water stored in the amorphous matrix has a plasticizing effect on the amorphous structure. The viscosity and elasticity of the material decrease with increasing water content. In parallel, the glass transition temperature decreases, due to the action of absorbed water as a plasticizer. This decrease in the glass transition temperature can be described using the Gordon and Taylor (1952) equation: Tg ¼

ð1
w ÞTg;dry þ wkTg;w ð1
w Þ þ wk

ð7:1Þ

Tg,dry represents the glass transition temperature of the (dry) solid, Tg,w the glass transition temperature of water, w the water content (wet-based) and k the so-called Gordon–Taylor constant. In parallel with the decrease in glass transition temperature, a decrease in viscosity is also observed. This decrease in viscosity can be described by applying the superposition principle known from polymer physics. Following the time–temperature superposition principle, all relaxation mechanisms have the same temperature dependence (Williams et al., 1955; Markovitz, 1975; Ferry, 1980). Thus,

7.3 Influence of Material Properties

viscoelastic data (like viscosity or modulus) obtained for one temperature and time or frequency can be extrapolated to a different temperature by multiplying the logarithmic time or frequency values with a temperature-dependent shift factor aT. Williams et al. (1955) suggested Eq. 7.2 for estimating the viscosity shift factor: aT ¼

mT0 r0 m0 Tr

ð7:2Þ

The viscosity shift factor relates the viscosity m of the solid material to the density r and the temperature T. By applying the assumption that the product of absolute temperature and density is roughly constant, the Williams, Landel and Ferry (WLF) equation (Williams et al., 1955) is obtained:

C T
Tg m
 log aT ¼ log ¼ m0 B þ T
Tg

ð7:3Þ

T0 is a reference temperature, which is commonly chosen as the glass transition temperature of the solid, C and B are constants. Williams, Landel and Ferry found that the parameters C ¼ 17.4 and B ¼ 51.6 K were suitable for most polymers they investigated. Changes in viscosity due to an increase in temperature and/or moisture above Tg can be estimated using the WLF equation, if the glass transition temperature is known. The mechanical properties (viscosity, elasticity) change drastically if the glass transition temperature is exceeded. In the amorphous glassy state substances react more elastically while exposed to any stress. Due to glass transition, the viscosity decreases from, for example, 1012 Pa s to 108
109 Pa s (Sperling, 1986). Thus, amorphous solids become first rubbery and, finally, more plastically deformable (Roos and Karel, 1991; Slade and Levine, 1991). Agglomeration during drying does not involve high compression forces. Particle collisions or application of moderate stress in a particle bulk are sufficient for a lasting deformation of the surface enabling adhesion of particles. Thus at least a part of the particle surface has to be well above the glass transition temperature in order to provide adhesion points on the particle surface. 7.3.1.2 Van der Waals Forces Van der Waals forces are only relevant if the surface of the primary particles in contact with each other is significantly deformed and can be the forces responsible for initiating the contact prior to sintering. Van der Waals forces are based on temporary load shifts in neighboring surfaces. The van der Waals forces between particles can be estimated by calculating the electrostatic forces between two parallel circular plates with a diameter x according to Lifshitz (1956) or Hamaker (1937) using Eq. 7.4: FvdW ¼ A

x2 a3

with

A¼

hv H ¼ 8p 6

ð7:4Þ

j301

j 7 Particle Formulation in Spray Fluidized Beds

302

A is a constant valid for parallel circular contact areas and a is the distance between the contacting surfaces. A can be calculated using the Lifshitz-van der Waals constant hv (¼ 10
20
10
18 J) or the Hamaker constant H (¼ 10
19
10
18 J). The van der Waals force between a sphere and a plate is significantly smaller than a typical capillary force. For spherical particles smaller than 5 mm, the van der Waals forces exceed gravity. For significant van der Waals forces particles have to be close (force proportional to a
3) and the contact area has to be large enough (force proportional to x2). Thus a deformation of drying particles is a prerequisite for generation of significant van der Waals forces. Since the van der Waals forces depend strongly on the distance a between the particles and on the interparticle contact area, they can be increased by decreasing the former and/or increasing the latter (Eq. 7.4). This can be achieved by plastic or viscoelastic deformation of particle surfaces which depends on the mechanical material properties. The adhesion force between two plastically deformed spheres can be calculated according to Rumpf et al. (1976): FvdW 

PvdW Ft Ppl

with

PvdW ¼

FvdW hv ¼ 2 3; 8p a px 2

a ffi 0:4 mm; h  v ffi 5 eV ð7:5Þ

Here, Ft is the force with which the particles are pressed together and Ppl is the plastic yield pressure. The distance a is assumed to be approximately 0.4 nm when the particles are in contact (Rumpf et al., 1976). However, amorphous water-soluble materials, such as food materials, deform viscoelastically. The deformation and relaxation behavior of such materials can be described by means of various viscoelastic models. Depending on the nature of the stress/strain applied, either the storage and loss modulus or the elasticity and the viscosity are included as material parameters in these models. These rheological material parameters depend on the temperature and the water content as well as on the applied strain rate. The viscoelastic deformation enlarges the contact area and decreases the distance between the particles (see Fig. 7.3). If the stress decreases once again, the achieved deformation is partially reversed (structural relaxation).

Fig. 7.3 Viscoelastic deformation and relaxation of food particles.

7.3 Influence of Material Properties

According to Rumpf et al. (1976) the contact area between two spherical particles which are viscoelastically deformed can be calculated using Eq. 7.6: x 2 d

¼



3 Ft 32 d2

2=3 

t 1 þ m E

2=3

ð7:6Þ

Ft is the force with which the particles are pressed together, t the compression time, m and E the viscosity and elasticity, respectively; x the diameter of the circular contact area and d the particle diameter. Combining Eq. 7.5 and 7.6 yields a relationship that enables calculation of the van der Waals forces between two viscoelastically deformed spheres: FvdW ¼

    hvd2 3 Ft 2=3 t 1 2=3 þ 8pa3 32 d2 m E

ð7:7Þ

As Eqs. 7.6 and 7.7 show, sufficient time and/or significant forces are necessary for agglomeration by plastic or viscoplastic deformation. During drying the forces are either a result of particle impact or of the weight of a particle bed. In many drying processes, such as spray drying or fluidized bed drying, the time that colliding particles are in contact with each other is relatively small. Thus the viscosity at the particle surface has to be low enough to allow yielding of the material in the short contact time. For instance, this is the case if amorphous material is not entirely dry. In spray drying or drying in fluidized beds, high temperatures and moisture contents may lead to low viscosities and thus agglomeration can occur even at short contact time. Furthermore, agglomeration due to van der Waals forces will occur even if the viscosity of the material is in the medium range and the force applied is low, provided that the contact time between two neighboring particles is sufficient to develop significant adhesion forces. This is typically the case for drying on belts or shelves or during storage of particles with high residual moisture content. The shorter the contact time and the higher the applied pressure, the more important is the viscoelastic deformation for developing adhesion forces between particles. After initial contact and adhesion a growing viscous bridge might be formed between the particles due to sintering. At long interparticle contact time and moderate pressure, sintering dominates the strength of the final agglomerate (Rumpf et al., 1976). Sintering will be further discussed in Section 7.3.1.4. 7.3.1.3 Capillary Forces Due to Liquid Bridges Between Particles Liquid bridges are the most important adhesion mechanism in spray or fluidized bed drying. In the case of spray granulation of food liquids, an aqueous concentrated solution is sprayed onto the fluidized particles and simultaneously dried. Droplets are deposited on the particle surface and can build liquid bridges between colliding particles, as will be discussed in more detail later. During spray drying one starts with droplets. Collisions of semidried droplets with dry particles lead to coating of particles. Once these coated particles collide with each other or with dry particles liquid bridges are established between the collision partners. The collision of semidried particles will also lead to the formation of agglomerates.

j303

j 7 Particle Formulation in Spray Fluidized Beds

304

In the case of water-soluble crystalline substances with a low molecular weight the viscosity and stability of the liquid bridges will be low. Thus, the strength of the liquid bridge depends mainly on capillary forces. The capillary pressure Pc in a liquid bridge can be calculated using Eq. 7.8; c is the surface tension, q the wetting angle, s the radius of the meniscus of the liquid bridge and x the diameter of the cross-section area of the bridge: Pc ¼ c cos q



 2 1
x s

ð7:8Þ

The negative capillary pressure between the particles causes an adhesion force Fc between the particles. This adhesion force can be calculated according to Eq. 7.9 (Fisher, 1926; Willett et al., 2007): Fc ¼

pdc 1 þ tanðb =2 Þ

with

Fc;max ¼ pdc

ð7:9Þ

The parameter b represents the angle from the center of the particle to the segment wetted by the liquid bridge. The capillary forces between two particles may also be calculated based on the energy of the system (Rabinovich et al., 2005). With time, the concentration of solids in a liquid bridge may increase because of drying (evaporation of water), migration of water into the particles (by capillarity or, in the case of amorphous solids, by diffusion) or dissolution of solids from the particles in the liquid of the bridge. Consequently, the stability of the bridge, initially dominated by capillary forces, is increasingly governed by viscous forces. Gradually, the relatively weak liquid bridge is transformed into a stable solid bridge. If the liquid bridge contains substances of lower molecular weight crystallization might occur. If the liquid contains a high amount of dissolved molecules with a medium or high molecular weight a viscoelastic amorphous bridge is obtained. The mechanical properties of such a bridge can be described through a complex shear or elongation modulus. 7.3.1.4 Viscous Forces in Sinter Bridges Between Amorphous Particles Particles composed of amorphous substances might sinter together if their viscosity is low enough. Sintering is, therefore, for many undesired agglomeration processes occurring during drying, the main adhesion principle. Amorphous substances have a liquid-like supramolecular structure. Like liquid droplets, two amorphous particles in contact with each other tend to adopt a spherical shape. Accordingly, molecules are transported to the contact point between the two particles. The capillary and vapor pressure gradients between the particle volumes and the contact point between the two particles are the driving force for these transport processes. Linked to these local differences in capillary and vapor pressure different molecular transport mechanisms are observed. The molecules can be transported via the surrounding gas phase (evaporation and sublimation), diffusion on the surface (surface diffusion/grain boundary diffusion) or diffusion

7.3 Influence of Material Properties

Fig. 7.4 Sintering mechanisms.

of the bulk amorphous material (volume diffusion/viscous flow) (Kuczynski, 1949; Schatt, 1992; Wagner, 1997). Substances having a medium or high vapor pressure can be transported via the gas phase surrounding the particles while substances with low vapor pressure (e.g., composed of larger molecules) exhibit volume diffusion. Knowing the surface tension c of the material, the primary particles of diameter d, the sinter bridge diameter x and the radius of meniscus curvature s, the capillary pressure gradient can be calculated through Eq. 7.10:   4c 2 1 DPc ¼ Pcl
Pc2 ¼
c
ð7:10Þ d x s Here, Pc1 represents the capillary pressure in the volume of the particles and Pc2 is the capillary pressure in the sinter bridge. Figure 7.4 illustrates the different sintering mechanisms. However, as mentioned earlier, volume diffusion remains for most amorphous substances (e.g., food substances) the most relevant sintering mechanism. The kinetics of transport by volume diffusion depends on the diffusion coefficient, which is a function of viscosity. The viscosity of water-soluble amorphous substances, and thus also the diffusion coefficient in amorphous matrices, depends on the temperature as well as on the water content of the material. In Fig. 7.5 the surface of a

Fig. 7.5 Plasticized surface with sinter bridges of a maltodextrin (DE 17-20) agglomerate.

j305

j 7 Particle Formulation in Spray Fluidized Beds

306

maltodextrin (DE 17–20) particle is shown. Areas which were humidified with water, and thus exhibit a significantly lower viscosity, are visible as smooth surfaces. While these areas are still wet they act as potential adhesion points during agglomeration. Sintering of organic particles by viscous flow can be modeled using the equations of Frenkel (1945) (see Eq. 7.11) or Rumpf et al. (1976) (see Eq. 7.12): x 2

¼

x 2

  4c 2Ft t þ ¼ 5 d 5pd2 m

d

d

1c t 6dm

ð7:11Þ

ð7:12Þ

Rumpf et al. (1976), who assumed a punctual contact between the particles, neglected changes in particle geometry during the process, and applied the Navier–Stokes equations for viscous flow. In the two equations d is the particle diameter, x the diameter of the sinter bridge, t the contact time, c the surface tension, m the viscosity of the organic substance, and Ft represents the force with which the particles are pressed together. Due to the assumption of non-deformed particles/unchanged geometry these equations are valid only for the initial period of the sinter process. For non-spherical particles the diameter d has to be estimated based on the curvature radius at the contact point between the particles. A prerequisite for the sinter process is molecular contact between the particles. Such contact can be established through particle collisions or by the pressure caused by the weight of particles in a bed. However, the impact of the force Ft on sinter kinetics can often be neglected. While variations in surface tension are commonly small, the viscosity often varies by orders of magnitude during drying due to large changes in temperature and moisture content. According to Palzer (2009) the growth of the sinter bridge during the time tmax can be calculated through Eq. 7.13. x 2 d

¼

tmax ð

t¼0



  ðT
Tg ðtÞÞ 4c 2Ft 1
C B þ ðT
Tg ðtÞÞ þ 10 dt 5 d 5pd2 m

ð7:13Þ

Figure 7.6 shows the calculated and measured bridge diameters for sintering of two dextrose sirup particles stored at 30  C and 70% RH. Despite various simplifications, Eq. 7.13 allows a satisfactory estimation of the sinter kinetics. The accuracy might be further improved by taking into account in the solution of the Navier–Stokes equations viscosity gradients that arise from moisture gradients in the agglomerating particles, and a more precise estimation of viscosity which is the main influencing variable. As one can see in Eq. 7.13, sintering kinetics depends on the difference between the process temperature T and the glass transition temperature Tg. Approaching the glass transition temperature sintering is drastically accelerated. The importance of glass transition for, for example, food or pharmaceutical products has been highlighted in several previous chapters of this volume. It arises

7.3 Influence of Material Properties

Ratio sinter bridge / particle diameter (x/d)

1 0.9 0.8 0.7 0.6 0.5 Calculated x/d for d = 255 µm

0.4

Calculated x/d for d = 355 µm

0.3

Calculated x/d for d = 410 µm Measured x/d for d = 260-280 µm

0.2 Contact radius due to manually initiated collision

0.1

Measured x/d for d = 350-360 µm Measured x/d for d = 400-420 µm

0 0

20

40

60

80

100

120

140

160

180

time t / min Fig. 7.6 Comparison between calculated and measured sinter bridge/particle diameter ratio for pairs of spray dried dextrose sirup particles (DE 21) stored at 30  C and 75% RH (Tg ¼
4  C; c ¼ 70 mN m
1, m ¼ 106 Pa s).

from a shift in the mechanical properties of amorphous solids from a rigid glass-like texture to first rubbery and later viscoplastic textures. This shift in texture, which is a result of a dramatic increase in molecular mobility, enables deformation of particle surfaces and facilitates viscous flow of the material during sintering. Palzer (2009) investigated the influence of moisture and temperature on fluidized bed agglomeration of dextrose sirup (DE 21). The glass transition temperature needs to be exceeded locally during granulation, fluidized bed agglomeration or spray drying in order to create sticky surfaces and to promote agglomeration. Impact of particles on surfaces where the glass transition temperature is exceeded leads to a deformation which increases the van der Waals forces. Following diffusion and capillary condensation a thin material bridge between the particles will be established. Successive rapid sintering will further strengthen the generated interparticle bridge. However, if the glass transition temperature in the bed is exceeded macroscopically by 25–30 K (humidification with moist air), a collapse of the entire bed caused by the formation of large clumps is observed (Palzer, 2009). At this temperature, the viscosity of the dextrose sirup is in the range 107 to 108 Pa s which is in good agreement with the results published by Wallack and King (1988), Downtown et al. (1982) and Aguilera et al. (1993). This reveals one of the main difficulties in controlling the agglomeration of amorphous water-soluble food powders in a

j307

j 7 Particle Formulation in Spray Fluidized Beds

308

fluidized bed or during any drying process. The moisture distribution must not be too homogeneous, as this will cause the whole to collapse once the sticky point of the particle bulk is exceeded. On the other hand, wet zones inside the bed, as often occur in the vicinity of the spray nozzle, will also produce agglomerates with a wide particle size distribution and undesired oversized particle structures. To avoid temperature/moisture combinations leading to a collapse of the bed or an increasing number of oversize particles during fluid bed granulation and drying, the aqueous binder has to be added with a low spray rate onto the fluidized particles, allowing the current of air to convey the evaporated moisture out of the fluid bed. Alternatively, the total added binder quantity can be reduced. In both cases the relative humidity of the air inside the fluid bed decreases and, hence, the glass transition temperature Tg and the viscosity of the particle surface increase. For spray drying such non-homogeneity of moisture can be achieved by adding back dry particles to the semi-dried droplets. 7.3.1.5 Mechanical Strength of Agglomerates The strength of agglomerates generated by the different adhesion processes discussed in the previous sections depends on the strength of the respective adhesion forces acting between primary particles. It can be measured by different methods (see Section 6.3.6 in Volume 2 of this series). According to Rumpf (1970), the tensile strength of agglomerates composed of spheres with diameter d can be estimated through st ¼

1
e KF; pd2

K

p e

ð7:14Þ

Here, F is the adhesion force between two spheres, e the agglomerate porosity, and K the coordination number of the primary particles building the agglomerate. For spherical primary particles K is approximately equal to p/e. On this basis, the tensile strength of agglomerates which are stabilized by viscoelastically enhanced van der Waals forces can be estimated according to st ¼

   2=3 ð1
eÞhv 3 Ft 2=3 t 1 þ 8ea3 32 d2 mðT; w; c_ Þ E ðT; w; c_ Þ

ð7:15Þ

The viscosity and the elasticity both depend on the shear or strain rate c, the temperature T and the plasticizer content w of the amorphous substance. The tensile strength of agglomerates stabilized by low-viscosity liquid bridges can be estimated according to Rumpf (1958) using Eq. 7.16: st ¼ SC

1
e 2c cos q e d

ð7:16Þ

Here S is the saturation of the porous agglomerate with liquid (ratio of liquid volume versus total void volume within the agglomerate), C a constant parameter (C ¼ 6 for monosized spheres), c the surface tension of the liquid, and q the liquid–solid contact angle (cf. Schubert (1975, 1979)).

7.3 Influence of Material Properties

Liquid bridges containing dissolved solids might dry out and thus the tensile strength of the agglomerates further increase. The tensile strength of such dry agglomerates can be calculated according to Rumpf (1958) by applying Eq. 7.17: st ¼

Vdiss ð1
eÞss Vagg

ð7:17Þ

Vdiss is the volume of solid dissolved within the entire agglomerate, Vagg the volume of the entire agglomerate, and ss the tensile strength of the solid substance building the bridge after drying. Combining Eqs. 7.13 and 7.14 the tensile strength of sintered agglomerates can be estimated according to Eq. 7.18: ð1
eÞp st ¼ ss ðT; w; c_ Þ e

tmax ð

t¼0

  g ðtÞÞ 4c 2Ft 1 B
CþððT
T T
Tg ðtÞÞ þ 10 5 d 5pd2 m

ð7:18Þ

However, in practice it remains difficult to predict the tensile strength of sintered agglomerates because the primary particles are normally non-spherical and exhibit various geometries. Furthermore, the tensile strength of the viscoelastic bridges also depends on deformation speed, which makes it difficult to define and determine a stability parameter. Apart from experimental results (Schubert, 1975), numerical simulations by Yang et al. (2008) that use the discrete element method (DEM) are also in good agreement with the model of Rumpf (Eq. 7.14). Note that, according to Eq. 7.14 the coordination number of primary particles decreases with increasing porosity (Smith et al., 1929; Bika et al., 2005), so that the strength of the agglomerates also decreases. A decrease in elasticity modulus and Poisson ratio with increasing porosity has been observed for both porous materials (ceramics, glass, polymers, cement clinkers) and composites (Boccaccini, 1994; Avar et al., 2003). An alternative to Eq. 7.14 has been proposed by Kendall et al. (1987) who considered a balance between the strain energy of compressed elastic primary particles of spherical shape with the potential energy of the applied load and the interface energy, assumed an empirical correlation for the coordination number, and introduced the parameter a as the maximal size of defects limiting the strength of agglomerates to obtain Ft st ¼ 3:7ð1
eÞ4 pffiffiffiffiffiffi d da

ð7:19Þ

Bika et al. (2005) have extended the models of Rumpf and Kendall to agglomerates with solid bridge bonds between primary particles. The solid bridge bonds are formed by evaporation of a liquid bridge and precipitation of dissolved solids. It is assumed that the liquid bridge conserves its shape as it shrinks and solidifies.

j309

j 7 Particle Formulation in Spray Fluidized Beds

310

Fig. 7.7 (a) c-Al2O3 primary particles used to produce the agglomerates, (b) scheme of agglomerate preparation.

The tensile strength of the agglomerate can be obtained from the strength of the solid bridge neck sbr,s by the relationship: " #2c 2 1
e 8xb Vbr st ¼ pb sbr;s ð7:20Þ rp d3 e where xb is the mass fraction of solids in the liquid bridge, that is, the binder content, and Vbr is the liquid bridge volume. The quantities rp and d are the primary particle density and diameter; b and c are numerical coefficients. A simpler equation for agglomerate strength, namely st ¼ ss ð1
eÞðxb
xb;min Þ

ð7:21Þ

was proposed by Tomas (1983). Equation 7.21 is similar to Eq. 7.17. It describes the strength of the agglomerate as a linear function of porosity and binder strength ss. A minimal mass fraction of binder xb,min in the solution is considered to be necessary in order to stick together the primary particles with solid bridge bonds after drying. This model showed good agreement with experiments performed with dry cylindrical agglomerates (Antonyuk and Tomas, 2008; Antonyuk et al., 2011). The agglomerates were prepared from spherical c-Al2O3 particles (d50 ¼ 1.0 mm) with a water solution of hydroxypropyl methylcellulose (HPMC) by compaction in a cylindrical die (Fig. 7.7) and drying for 1 h at 80  C. The binding agent was heated to 40  C and added to the primary particles in a mixer. For all agglomerates produced, the mass ratio of the solid particles to the water solution of the binder was maintained at a constant value of 1.42. However, solutions with different HPMC concentrations (6, 10, 13 and 16 wt%) were used in order to investigate the influence of binder ratio on the mechanical properties of the agglomerates. After compaction and drying the mass fraction of HPMC in the agglomerates, that is, the binder content xb, was 4.0, 6.5, 8.3 and 10.1 wt%, respectively. Compaction was done in the direction of the principal axis. In order to vary the agglomerate height, different cylindrical dies with heights of 8, 12 and 15 mm and constant inner diameter of 15 mm were made. Compression tests were performed on single primary particles and agglomerates by using the Granule Strength Measuring Device (Etewe GmbH, Germany) at a constant stressing velocity of 0.04 mm s
1. Fifty agglomerates of each binder content

7.3 Influence of Material Properties

Fig. 7.8 Scheme of compression tests with diametral loading condition.

and size were tested in order to increase the statistical significance of the measurement. The compression test was carried out laterally to the principal axis of the cylinder (Fig. 7.8). The typical force–displacement curve of a primary spherical particle (c-Al2O3) in Fig. 7.9a shows clearly both elastic and plastic displacement ranges. The elastic displacement of the spherical particle up to the yield point F was described using the Hertz contact theory of spheres (Hertz, 1882; Antonyuk et al., 2011) according to 1 pffiffiffiffiffiffiffiffiffi FN;el ¼ E  d  s3 6

ð7:22Þ

 
1 1
u2 1
u2w þ E ¼ 2 Ew E

ð7:23Þ

The effective modulus of elasticity E in Eq. 7.22 refers to both the particle (without index) and the punch (index w for wall), and is given by:

Fig. 7.9 Typical force–displacement curves during compression of (a) spherical primary particles (c-Al2O3, d ¼ 1.03 mm) and (b) cylindrical agglomerates (dcyl ¼ 15 mm, hcyl ¼ 15 mm) prepared from these c-Al2O3 particles and 4% of HPMC binder.

j311

j 7 Particle Formulation in Spray Fluidized Beds

312

where u and uw are the Poisson ratios of the particle and the punch (see also Chapter 6 in Volume 2 of this series). The effective contact stiffness of elastic deformation depends on the deformation and particle diameter as: 1 pffiffiffiffi kel ¼ E  ds 4

ð7:24Þ

With known properties of the punch, the modulus of elasticity and the elastic contact stiffness of the primary particles were determined to be the values given in Tab. 7.3. Figure 7.9b shows the typical force–displacement curve of a cylindrical c-Al2O3 agglomerate. At the beginning of loading the contacts are elastically deformed. The elastic force–displacement line up to the yield point F can be described using an elastic contact model (Antonyuk and Tomas, 2008): Fel ¼ 0:39

Ecyl hcyl s 1
u2cyl

ð7:25Þ

where Ecyl is the modulus of elasticity of the cylindrical agglomerate, hcyl the height of the cylinder, ucyl the Poisson ratio, and s the displacement during the deformation. The modulus of elasticity of the agglomerates is much lower than that of the primary particles (Tab. 7.3). However, the elastic stiffness of agglomerates and primary particles is of the same order of magnitude. The force–displacement curve changes beyond the yield point from elastic to ideally plastic behavior. The displacement sF and the force FF obtained at the yield point of force–displacement curves and the yield pressure PF were used as model parameters to approximate the plastic force (Antonyuk and Tomas, 2008): qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Fpl ¼ FF þ 2PF hcyl Rcyl ðs
sF Þ ð7:26Þ

where Rcyl is the radius of the cylinder. Compared to the rapid force drop during the breakage of a primary particle (point B, Fig. 7.9a), the force decrease after the primary breakage of agglomerates (Fig. 7.9b) occurs not so clearly because of deformation and secondary breakage of the fragments which were constrained between the two plates. Figure 7.10a shows the influence of binder mass traction xb in the agglomerate on its compression strength s and elastic stiffness kcyl,el. The increase in binder content increases the strength of agglomerates. Point A indicates the minimum binder content which is necessary to stick together the primary particles by solid bridge bonds (here, xb,min ¼ 1.9 wt%). Agglomerate strength was described as a linear function of this minimum necessary binder content and binder strength sb using Eq. 7.21. With increasing binder content the elastic stiffness also increases. Equivalent to Eq. 7.21, a linear relationship between the stiffness and the binder content can be used, kcyl,el ¼ kbr,s(1
e)(xb
xb,min), where kbr,s is the stiffness of the solid bridge bond. The fracture plane of the examined agglomerates coincides with the plane of loading, as indicated in Fig. 7.11. Due to maximum tensile stresses, the cracks

Primary particles Agglomerate

Material

10.5 75.5

FF (N) 9.4 200

sF (mm)

yield point

elastic stiffness kel (F ¼ FF) (kN mm
1) 2.6  0.21 1.1  0.27

modulus of elasticity E (kN mm
2)

25  0.7 0.08  0.02

885  31 144.4  25

plastic stiffness kpl (s ¼ 0.98sB)(N mm
1)

32.5 93

FB (N)

0.033 0.33

sB (mm)

breakage point

Tab. 7.3 Mechanical properties of cylindrical agglomerates (dcyl ¼ 15 mm, hcyl ¼ 15 mm, binder content xb ¼ 4 wt%) and their primary particles (c-Al2O3, d ¼ 1.03  0.06 mm).

7.3 Influence of Material Properties

j313

j 7 Particle Formulation in Spray Fluidized Beds

314

Fig. 7.10 Effects of (a) binder content xb and (b) height hcyl of cylindrical c-Al2O3 agglomerates on their elastic stiffness kcyl,el and strength s during compression.

propagate in the loading direction and separate the agglomerate into two approximately equal fragments. On the micro level, interparticle breakage occurs, that is, only the solid bridge bonds break but not primary particles because of their higher strength in comparison with the binder. The breakage probability at different binder contents is plotted in Fig. 7.12 versus the mass related breakage energy. As expected, the curve shifts to the right with increasing mass concentration of binder in the agglomerate. Compared to the agglomerates, the breakage energy of the primary particles is much larger (Wm,10
Wm,90 ¼ 730–1230 J kg
1). Agglomerates formed with the help of highly viscous liquids and agglomerates with high saturation levels show a dependence of the tensile strength on capillary pressure (Schubert, 1975). Porosity and pore size distribution have a large effect on the breakage behavior – similar to the effect of the size distribution of defects on the breakage of crystalline materials. The pores in agglomerates can be seen as crack release zones with high stress concentration. As a consequence of wetting, chemical reactions can occur within agglomerates, whereby the original mechanical properties can be changed. To avoid such changes, many experiments were carried out with model binder solutions and primary

Fig. 7.11 Photos of the cylindrical agglomerate after breakage during compression and of the corresponding crack zone; 1: unbroken primary particles, 2: broken solid bridge bonds.

7.3 Influence of Material Properties

Fig. 7.12 Breakage probability P of c-Al2O3 agglomerates (dcyl ¼ 15 mm, hcyl ¼ 15 mm, with different binder content xb) versus mass related breakage energy Wm during compression.

particles, for example, glass beads (Pierrat and Caram, 1997; Iveson and Page, 2001). Briscoe et al. (1998), Pepin et al. (2001) and Samimi et al. (2003) performed compression tests on spherical agglomerates and determined the force–displacement curves. An overview of previous work on the strength of wet agglomerates was given by Simon et al. (2001). Compared to crystalline solids, agglomerates are particle compounds that tend to plastic force–displacement behavior. Depending on the formulation process, internal adhesion is influenced by the superposition of van der Waals interactions between fine primary particles, capillary or solid bridges, high-viscosity binder, organic macromolecules, sintering or interlocking of particles. The mechanical breakage of agglomerates is, therefore, rather determined by these micro-binding mechanisms than correlated with the stress state – as even the earliest systematic investigations of the strength of agglomerates show (Rumpf, 1958; Schubert, 1975; Kendall et al., 1987). 7.3.2 Breakage of Agglomerates and of Granulated Products

After having addressed the breakage of cylindrical agglomerates consisting of large primary particles, the breakage of spherical agglomerates made of much smaller primary particles will be treated in the present section to show that typical patterns of elastic–brittle or elastic–plastic breakage behavior can occur. Additionally, plastic breakage and the breakage behavior of layered granules will be discussed. The contents of the present section may be compared with results from the literature gained with structured materials such as concrete (Tomas et al., 1999; Khanal et al., 2008) or fertilizer granules (Salman et al., 2003). Breakage depends on the loading energy, the material, particle shape and particle roughness. Material behavior, whether elastic or plastic, depends on the stressing intensity and agglomerate size. Fracture can hardly be understood without consideration of particle microstructure.

j315

j 7 Particle Formulation in Spray Fluidized Beds

316

Fig. 7.13 Crack propagation during impact of an elastic c-Al2O3-granule.

7.3.2.1 Elastic–Brittle Breakage Behavior Elastic–brittle breakage behavior is characterized by a failure which begins during elastic deformation and is attended by non-stationary cracks, where no external energy input is needed to grow the cracks with a rapid propagation velocity. Such behavior was observed by Antonyuk et al. (2006) on commercially produced, spherical c-Al2O3 particles. These particles had a thin shell around a porous core. The average distance between two pores was about 0.5 mm, corresponding roughly to the respective primary particle size. The granules were tested by impact with a target (steel wall) using an “air gun” (cf. Chapter 6 in Volume 2 of this series). Breakage during impact was captured with a high-speed camera. The breakage mechanism is represented schematically in Fig. 7.13. The meridian cracks occur after the contact with the target wall. The cracks initiate from the perimeter of a circular contact area, where a maximum tension stress appears. Small shell defects that arise during production and transportation are the starting point. With a rapid propagation of cracks (divergent to the impact axis), the grains are separated into several meridian fragments (Fig. 7.14). The smooth area of the meridian cracks through the porous c-Al2O3-granules (“1” in Fig. 7.14) clearly refers to brittle fracture, without plastic deformation. In addition, many small cracks occur within the conical contact area (“2” in Fig. 7.14), where the energy density is very high at the moment of impact. The crack propagates from one pore to another, as shown schematically in Fig. 7.13. As a result, many fine particles are formed within the range 0.5 mm (the already mentioned average distance between

Fig. 7.14 SEM of the fracture surface of a c-Al2O3-granule after impact at 31 m s
1 (1: meridian fracture surface, 2: fracture surface in contact area).

7.3 Influence of Material Properties

Fig. 7.15 SEM of the fracture surface of a c-Al2O3-granule in the contact area (“2” in Fig. 7.14), resulting from crack branching.

pores, Fig. 7.15) to 100 mm. At high impact velocity, secondary cracks are formed and they are perpendicular to the direction of impact. 7.3.2.2 Elastic–Plastic Breakage Behavior Elastic–plastic breakage is characterized by an amount of plastic deformation before the failure. The plastic deformation usually occurs at the contact point, where the energy concentration and stress are much higher than in the whole particle volume. A typical breakage pattern of elastic–plastic spherical particles is shown in Fig. 7.16 for zeolite agglomerates undergoing impact loading. The flattened surface develops under compressive stress around the contact point and deforms plastically. Then, a sharp, non-fragmented cone penetrates into the agglomerate. The resulting tension leads to the meridian cracks, which separate fragments from the remaining core. The core (or nucleus) depicted in Fig. 7.16 has to do with the formulation process of the13X synthetic zeolite, which was an aggregate of fine crystals in the size range 1–8 mm. The crystals were agglomerated with clay as the binder and water as the solvent, which form solid bridges between the primary particles after drying. Zeolite primary particles bonded by the binder can be seen on the SEM of the agglomerate surface in Fig. 7.17. Primary particles do not contact each other directly; they are connected by three to six solid bridge bonds for each primary particle, due to inhomogeneous distribution of the binder. Rolling agglomeration in a drum was the

Fig. 7.16 SEM of the fragment of a zeolite agglomerate after impact at 18 m s
1 (1: nucleus, 2: shell) and schematic representation of crack propagation at different impact forces.

j317

j 7 Particle Formulation in Spray Fluidized Beds

318

Fig. 7.17 SEM of the zeolite agglomerate before impact: (a) agglomerate surface, (b) solid bridge bonds between primary particles.

formulation process. In this process, nuclei are mixed in a drum with much finer zeolite particles, grow in size by accumulating primary particles and form spherical agglomerates. The layers developing first on the nucleus surface have many defects in solid bridges and large porosity, and hence they are weak. Compared to these layers the nucleus of the agglomerate is closely packed, has a better distribution of binder, and a firmer structure. During the impact, the cracks formed at the plastically deformed contact area propagate through the weakest agglomerate layers towards the nucleus, as shown in Fig. 7.16. The nucleus remains unbroken. The nucleus can be fractured at impact velocities larger than 30–35 m s
1 owing to the high inertial forces (right-hand side of Fig. 7.16). At the micro-level one can distinguish between trans-particle and interparticle breakage (similar to trans- and intercrystalline breakage in the damage mechanics of materials). In the case of trans-particle breakage, primary particles and binder are destroyed. On the other hand, in interparticle breakage the failure appears at the contacts of primary particles. This means that only the solid bridge bonds break. Interparticle breakage was observed in the zeolite agglomerate. The unbroken primary particles with broken solid bridges can be seen in the SEM of the fractured surface in Fig. 7.17 and compared with the structure before the impact in Fig. 7.18.

Fig. 7.18 SEM of fractured surface of elastic–plastic zeolite agglomerate after impact at 18 m s
1: (a) fractured surface, (b) broken solid bridge bonds between primary particles.

7.3 Influence of Material Properties

7.3.2.3 Plastic Breakage Behavior Plastic breakage is characterized by a large amount of plastic deformation before failure and by stationary crack growth, which means that external energy is permanently necessary for the cracks to propagate. Measurements of the yield point and the breakage strength of dominantly plastic agglomerates are difficult because of the small yield limit. After reaching this limit, irreversible deformation takes place. In many cases, during this deformation no instantaneous drop of the force, which would indicate the breakage point, can be obtained. Moreover, in the case of stationary crack growth the force does not significantly decrease, so that the breakage point can be detected only as a change of momentary stiffness (Fig. 7.19). Therefore, an acceptable limit of shape change, that is, quality loss, can more adequately characterize the breakage behavior than a value of breakage strength. The force–displacement curve of spherical plastic granules is a straight line along the whole deformation region up to the breakage. Figure 7.20 shows the force–displacement curves of sodium benzoate granules obtained during the uniaxial compression test (Antonyuk et al., 2010). In the case of bigger plastic granules, both the breakage force and the contact stiffness increase. Therefore, the material becomes stiffer with increasing granule diameter. This effect is also valid for elastic granules, see Eq. 7.24. The increase in the breakage force does not influence the material strength of plastic granules. As an example, for different size fractions of sodium benzoate the average compressive strength is about 10.6 MPa. The yield pressure PF in the contact is also independent of particle size (about 113 MPa). Assuming ideal plastic deformation with uniform pressure PF in the contact area, the compressive strength of predominantly plastic granules can be described by the equation

force

sF ¼ 0:1 PF

ð7:27Þ

breakage point

displacement

dA

Fig. 7.19 Typical shape of the force–displacement curve for predominantly plastic behavior.

j319

j 7 Particle Formulation in Spray Fluidized Beds

320

20

plastic deformation

Force F [N]

15

B-primary breakage

granule diameter in mm: 1.24-1.60 0.80-0.96 secondary breakages

10

5

0 0

0.05

0.1 0.15 Displacement s in mm

0.2

Fig. 7.20 Typical force–displacement curves of sodium benzoate granules during compression (stressing velocity: 0.02 mm s
1).

(Antonyuk et al., 2011). At the same deformation value, the formed contact area increases with increasing particle size. 7.3.2.4 Breakage of Granules with Layered Structure As already discussed in Section 7.2, fluidized bed spray granulation leads to particles with a layered structure. By spraying on cores, particles with a firm nucleus and a soft onion-like shell can be formed. Such a particle made of sodium benzoate is depicted in Fig. 7.21. During compression testing of the layered structure, the high tensile stress, which is perpendicular to the direction of loading, leads – as before – to the formation of meridian cracks. According to Fig. 7.22, the shell first deforms around the contact point. Then, a crack (1) is released and reaches the surface that separates the stiff nucleus (2) from the shell (3). The force–displacement curve of Fig. 7.23 shows that, after the formation of a meridian crack in the shell (Point B1), deformation and then

Fig. 7.21 SEM of the cross-section of a sodium benzoate particle produced by fluidized bed spray granulation (1: nucleus, 2: layered shell).

7.3 Influence of Material Properties

Fig. 7.22 SEM of the fracture surface of granulated sodium benzoate after compression test (1: crack, 2: nucleus, 3: shell).

full breakage of the nucleus take place. The latter leads to a secondary increase in the force (point B2). In the depicted case the cohesive forces of the first shell layer with the nucleus are substantially weaker than the cohesive forces between adjoining layers in the shell. 7.3.3 Consideration of Primary Particle Properties in Agglomeration

The rate of agglomeration in spray fluidized beds is proportional to the frequency of collisions and the probability that colliding particles will stick together rather than 15 B1 - primary breakage of the shell

B 2 - breakage of nucleus

Force F [N]

10

5

0 0

0.1

0.2

0.3

Deformation s in mm Fig. 7.23 Force–displacement curve of granulated sodium benzoate during compression (stressing velocity: 0.02 mm s
1, d ¼ 0.87 mm).

j321

j 7 Particle Formulation in Spray Fluidized Beds

322

rebound (see Section 7.7). Particles stick together when colliding at wet positions, provided that the liquid film covering such positions can dissipate the energy of the collision impact (Stokes criterion). It is, thus, evident, that the agglomeration rate depends on the fraction of outer particle surface covered with binder. This fraction does not immediately result from how much liquid is sprayed into the bed, because sprayed liquid may be located both on the surface or in the interior of porous agglomerates. Liquid on the surface can lead to further aggregation, whereas liquid trapped in agglomerate voids is not accessible and, therefore, inactive for agglomeration (steric hindrance). Fractional surface coverage and accessible binder fraction depend on the properties of primary particles, so that these properties are expected to also influence the kinetics of the agglomeration process. An interesting analysis of the mentioned effects has been provided by Stepanek and coworkers (Stepanek and Rajniak, 2006; Stepanek et al., 2009). The goal of these authors was to quantify the influence of the primary particle properties – especially of primary particle shape, surface morphology and roughness – of four typical powders used as excipients in the pharmaceutical industry on their agglomeration behavior. The four excipients were (Stepanek et al., 2009) Avicel (which is rather smooth and round), lactose and mannitol (of intermediate roughness), and A-Tab (rough and rather irregular primary particles). First, primary particles were created computationally in the form of so-called Gaussian blobs (Stepanek and Rajniak, 2006), as illustrated in Fig. 7.24. This method distorts more or less strongly the surface of a sphere towards different primary

Fig. 7.24 Reconstruction of primary particles of different morphology as “Gaussian blobs” according to Stepanek and Rajniak (2006): (a) field of independent random variables, (b)

Gaussian-correlated random field, (c) underlying spherical particle, (d) particle surface modulated by the correlated random field.

7.3 Influence of Material Properties

particles that can closely mimic the different surface morphologies of the four mentioned real powdery materials. Then, agglomerates, each containing 120 primary particles, were formed and loaded with different amounts of liquid by means of a ballistic deposition algorithm (Coelho et al., 1997). The results of the evaluation of the obtained wet agglomerates with respect to the fractional surface coverage and the spatial distribution of the binder are presented in Fig. 7.25. The fractional surface coverage is depicted in Fig. 7.25a. It is an increasing function of the volumetric binder/solid ratio in the granule. The curve shows two distinct regions. At lower ratios the fractional coverage is independent of the primary particle shape (wet granule in pendular state). By adding more liquid, a transition to the capillary state occurs, which is highly dependent on the primary particle morphology. The accessible binder fraction (Fig. 7.25b) follows a sigmoidal dependence on granule composition, which can

Fig. 7.25 Effect of primary particle shape on (a) the relative fractional coverage with binder and (b) the accessible binder fraction, as a function of granule composition.

j323

j 7 Particle Formulation in Spray Fluidized Beds

324

be explained as follows: At relatively low binder fractions most of the binder is bound in capillary bridges between primary particles and thus sterically hindered, that is, shielded from other particles colliding with the considered agglomerate. Once a certain, shape-dependent critical value is exceeded, more and more liquid binder volume becomes exposed on the accessible outer surface of the agglomerate. The accessible binder fraction reaches 100% well before the fractional surface coverage. One manner of using the presented results is to incorporate them in the traditional way of tackling fluid bed granulation theoretically, namely population balance modeling. This can be achieved by expanding the population balance to more internal coordinates than just particle size (see Volume 1 of this series, Chapter 6, Section 6.9.1). The additional property in the case of the present example would be wet agglomerate composition, defined either by the mass fraction of solids within one particle or the binder/solid ratio. The latter can be further split up to account for the spatial distribution – and, thus, accessibility – of the liquid binder and for the thickness of the binder layer on the outer surface of the agglomerate. Alternatively, discrete models of agglomeration (see Section 7.7) could be expanded to account for non-spherical primary particles.

7.4 Influence of Operating Conditions

The previous discussion pointed out that various material properties have a more or less strong influence on particle formulation in spray fluidized beds. Process conditions, such as the spraying rate, gas temperature, or the mass flow rate of the fluidization gas, can also very significantly affect the properties of the resulting particles. This aspect will be illustrated in the present section by means of selected examples. 7.4.1 Mechanical Strength of Granulated Particles

A major quality feature of particles is their mechanical strength. This parameter depends mainly on the material properties but also, as discussed in Section 7.3.2, on the microstructure of the particles. Microscopic changes in the structure, porosity or surface properties of the granules occur under the influence of different process parameters and can lead to major changes in properties on a macroscopic scale, for example breakage and attrition resistance. To demonstrate the effect of different process parameters on the mechanical strength, granulation experiments were performed in a cylindrical fluidized bed unit with an inner diameter of 250 mm and a fluidization chamber height of about 500 mm. The pilot plant (Fig. 7.26) can be operated with superheated steam or air in batch or continuous mode. The experiments presented here were batch, with 5 kg of synthetic zeolite particles with a Sauter diameter of 2.18 mm as the starting material.

7.4 Influence of Operating Conditions

Fig. 7.26 Scheme of pilot plant.

An aqueous solution of sodium benzoate (30 wt%) was sprayed on these particles. The solution also contained different amounts of binder (HPMC, Pharmacoat 606 from ShinEtsu Co.). During the granulation experiments samples were taken at constant time intervals. For each sample the particle size distribution was analyzed by means of a Camsizer (see Section 5.2.2 in Volume 2 of this series). Additionally, the mechanical strength was determined per uniaxial compression (see Chapter 6 in Volume 2 of this series). The results are visualized by plotting the breakage probability versus the breakage energy, similarly to the plot of Fig. 7.12 of Section 7.3.1. 7.4.1.1 Influence of Binder Content in the Sprayed Solution Figure 7.27 shows the evolution of particle size distribution for a trial with 4 wt% of binder. The hold up material grows with time. Moreover, a fraction of very small particles appears after approximately 2.5 h. This peak is caused by breakage or attrition processes producing internal nuclei. Figure 7.28 depicts the breakage probability of the granules after 3 h granulation time for different binder mass fractions. The remaining process parameters, such as mass flow rates and temperatures, were kept constant. The measurements show that higher mass fractions of binder lead to lower specific breakage energy. This result seems to correlate with the different shapes of the granules. It is known, that spherical objects have a high breakage resistance. As Fig. 7.29 shows, such particles are obtained for low binder concentrations or without binder. In contrast, high binder content results in a very irregular surface structure of particles which are less strong and easier to break.

j325

j 7 Particle Formulation in Spray Fluidized Beds

326

Fig. 7.27 Evolution of the number density distribution of bed material with 4 wt% of binder.

7.4.1.2 Influence of the Particle Retention Time To analyze the influence of process time on particle strength, tests were performed under dry conditions without liquid injection. In Fig. 7.30 the breakage probability of different samples taken at 5, 10 and 15 min retention time is depicted.

Fig. 7.28 Influence of binder content on the breakage probability after 3 h of granulation (d32 ¼ 2.9 mm, Tbed ¼ 60  C; symbols: experiment, lines: model described in Chapter 6 of Volume 2 of this series).

7.4 Influence of Operating Conditions

Fig. 7.29 Shape of granules after 3 h granulation: (a) no binder, (b) 4 wt% binder.

The measurements show that the mean specific breakage energy increases with increasing process time. Longer tests reveal that the increase in particle strength stops after about 15 min. These results can be explained by the selection of weak particles, which easily break, out of the particle population. 7.4.1.3 Influence of Process Temperature Finally, the influence of bed temperature was analyzed by varying this process parameter from 60 to 120  C at a constant granulation time of 60 min. Figure 7.31 shows that the highest values of granule strength and breakage energy were obtained

Fig. 7.30 Influence of retention time on the breakage probability (d32 ¼ 2.18 mm, Tbed ¼ 60  C, dry runs; symbols: experiment, lines: model described in Chapter 6 of Volume 2 in this series).

j327

j 7 Particle Formulation in Spray Fluidized Beds

328

Fig. 7.31 Influence of the process temperature on the mean specific breakage energy and compressive strength of granulated particles (core material: zeolite 4A, shell material: sodium benzoate).

at 60  C, whereas the lowest values were measured for a process temperature of 100  C. A slight increase in particle strength was observed for temperatures higher than 100  C. A direct comparison of the surface morphology of the granules is presented in Fig. 7.32. Different temperatures result in a different microstructure. At lower gas temperatures a moderate drying rate can be expected, which provides more time for a crystalline, dense structure to evolve. This results in more elastic behavior and the higher particle strength that was observed at 60  C. Particles granulated at higher temperatures have a smoother surface and seem to be more brittle.

Fig. 7.32 SEM images of particles granulated at different temperatures: (a) 60  C, (b) 120  C.

7.4 Influence of Operating Conditions

In general, different process parameters have an influence on the solid phase formation around core particles. The structure of this shell influences, in turn, the properties of the resulting product, for example, the mechanical strength. 7.4.2 Catalyst Impregnation in Fluidized Beds

The loading of porous carrier particles with catalytically active ingredients usually requires several steps which have to be realized in several pieces of equipment, namely impregnation in a bath, separation of the solids by filtration, drying, and calcination. Here, the fluidized bed technology provides an alternative that combines all these stages in one apparatus. The studies of Hemati et al. (2001, 2003) and Desportes et al. (2005) prove the feasibility of manufacturing catalysts by spraying metallic precursor solutions on a porous support in a hot fluidized bed. The authors show that, according to the operation conditions, two different scenarios are possible: .

.

Soft drying under mild temperature conditions leads to homogeneous impregnation and catalyst deposition over the entire internal pore volume of the carrier. Intensive drying under high temperatures results in deposition only on the external surface of the support due to fast evaporation of the liquid.

The process is affected by various apparatus and process variables, such as nozzle configuration, gas inlet temperature, and flow rate, as well as material properties such as contact angle and surface tension, liquid viscosity and the texture of the solid particles. However, as the mentioned authors state, the solute distribution in the porous particles is mainly controlled by the following two time constants: .

The wetting time twet, which describes the transient penetration of liquid from the surface into the pores towards the particle center; the wetting time can be estimated from twet ¼

2 m L2 ; c cos q rpore

ð7:28Þ

where m is the liquid viscosity, c the surface tension, q the contact angle, L the effective pore length (particle radius multiplied by the tortuosity factor), and rpore is the radius of the capillaries inside the particle. . The drying time tdry of a particle saturated with pure solvent under fluidized bed conditions. Desportes et al. (2005) conducted experiments with coarse alumina particles (mean diameter 2.4 mm) to show that the homogeneity of solute distribution in the particles depends on the ratio between the two characteristic times tdry/twet. Two different experimental conditions that correspond to fast drying and to slow drying (low and high values of tdry/twet, respectively) were realized, as displayed in Tab. 7.4. Manganese nitrate was the precursor material. Material samples were removed after

j329

j 7 Particle Formulation in Spray Fluidized Beds

330

Tab. 7.4 Experimental conditions for the investigation of solute distribution in porous fluidized particles according to Desportes et al. (2005).

Process parameter

Unit

Fast drying

Slow drying

Fluidization gas flow rate Gas inlet temperature Bed temperature Liquid flow rate Solute mass fraction at the end of impregnation tdry/twet Relative humidity

m3 h
1  C  C g h
1 %

28 60 45 187 12

31 78 27 740 40

— %

4 8

11 70

different processing times (Tab. 7.5) and analyzed by means of particle cross-section micrographs. Figure 7.33 shows that, in the case of fast drying, the deposition takes place only at the surface of the particles. In the case of soft, slow drying, the solution penetrates into the particles in the form of a front that moves towards the center with increasing processing time. After sufficient time an even, homogeneous distribution of solute is obtained over the entire particle. Concerning the ratio between tdry and twet, Desportes et al. (2005) conclude that for values below 10 a coating layer is deposited around the carrier, while for larger values the whole particle structure is penetrated by functional components of the sprayed solution. Additional measurements by Desportes et al. (2005) indicate that the pore size distribution pattern of the initial support is conserved during increasing catalyst penetration (transition from S1 to S4 in Fig. 7.34). However, pore sizes are reduced, so that the specific surface area and the specific pore volume of the material gradually decrease (Tab. 7.5). It should be noted that the quality of the products discussed in this section is mainly defined by catalytic activity and catalyst pellet efficiency, the latter being a measure of mass transport limitations within the particle. Depending on the

Sampling time and sample characteristics under different drying conditions (F: fast drying, S: slow drying) according to Desportes et al. (2005). Tab. 7.5

Parameter

Unit

F1

S1

S2

S3

S4

Spraying solution time tdry/twet Relative humidity Specific pore volume

h

3

0.5

1

1.5

3

— % cm3 g
1

4 8 0.39

11 70 0.29

11 70 —

11 70 0.25

11 70 0.22

7.4 Influence of Operating Conditions

Fig. 7.33 Microscope pictures of alumina carrier particles with a metal precursor obtained under different drying conditions (F: fast drying, S: slow drying) (Desportes et al., 2005).

reaction to be conducted and the type of reactor to be used, a shell catalyst or a full catalyst may be the right solution. As shown, both types of supported catalysts can be produced in spray fluidized beds by adjusting the intensity of drying, which can be easily done by setting relevant operation parameters. Therefore, there is an immediate, direct connection between drying and the quality of the formulated product. Drying and liquid penetration are also important for the process already discussed in Section 7.3.3, namely spray fluidized bed agglomeration. The reason for this is that agglomeration takes place with the help of droplets sprayed on the particles, so that it slows down when such droplets are lost either by evaporation (drying) or by liquid penetration into the porous substrate. Influences of this kind can be captured very well with the help of respective micro-scale models integrated into discrete simulations, as we will see in Section 7.7.

Fig. 7.34 Pore size distribution of alumina support particles under slow drying conditions for different spraying times (Desportes et al., 2005).

j331

j 7 Particle Formulation in Spray Fluidized Beds

332

7.5 Influence of Apparatus Design

7.5.1 Apparatus Design Features with an Influence on Product Quality

The design of central as well as peripheral elements of spray fluidized bed equipment has been discussed in considerable detail by Jacob (2007), so that it is sufficient to give here a brief summary before focusing on some special aspects of apparatus design with high relevance for product quality. Important elements of the periphery of spray fluidized beds are: .

.

.

Equipment for processing gas handling. Here, one can distinguish between fresh air (single pass) and gas recycling (closed loop) systems. Such systems always include ventilators, pre-filtration and direct or indirect gas heating units. Frequent additional components are de-humidifiers and humidifiers, coolers and final filter units. The number and arrangement of ventilators can be used to adjust the pressure inside the process chamber relative to the ambient in order to, for example, avoid product contamination and guarantee product hygiene or, vice versa, to protect operators from toxic materials. Closed-loop set-ups are mandatory when inert gas (nitrogen) has to be used instead of air to preserve the quality of products sensitive to oxygen, to avoid explosion risks or to fulfill emission limits, especially in regard to odor. Gas recycling is always combined with heat recovery to reduce the total energy demand and provides opportunities for weather independent operation in the case of sensitive processes. The split ratio between fresh and recycled gas is an important parameter for the operation of such systems. Avoidance of undesired condensation, fouling and hygienic risks are crucial issues. Outlet gas handling. The most important aspect of outlet gas handling is the removal of dust, which can be achieved by internal filters, external filters or wet scrubbers, or a combination of these. Internal filters are textile bags or cartridges placed in the exhaust chamber of the fluidizing unit. They can be cleaned by mechanical shaking or by reverse gas flow (back-purge), whereby the dust falls down into the fluidized bed. Subdivisions of the exhaust chamber in combination with flaps allow cleaning of some filters while simultaneously operating some others, so that the process need not be interrupted. Solids handling facilities. Depending on the mode of operation (batch or continuous), various pieces of equipment can be used for charging or discharging the particulate material, such as flaps, rotary valves, dosing screws, overflows (weirs), discharge pipes, gas distributors that can be folded down or turned aside, exchangeable material containers, chutes, and pneumatic conveying units. However, solids handling is more than simple charging and discharging – especially in continuous spray fluidized bed processes. To guarantee the right particle size of the outlet product classifying discharge systems – such as tubes with countercurrent, upwards oriented gas flow or zig-zag-sifters – may be used. Alternatively

7.5 Influence of Apparatus Design

or additionally, the outlet solids can be fed to a two-deck sieve that separates oversized and undersized particles from the final product, which is the fraction remaining between the two sieve decks. Undersized grains can be recycled to the process, and the same can be done with the oversized fraction after milling or crushing. This solids recycle is extremely important for granulation processes, because the fine-grained recycle particles serve as seeds (nuclei). Such nuclei are necessary for restarting particle growth and stabilizing the process towards a steady-state particle size distribution at the outlet. Feeding seed material from external sources (usually spray dryers) is the alternative to solids recycle. In both cases, seeds generated in the process chamber by overspray (droplets that dry out before reaching the surface of fluidized particles), attrition or breakage must also be considered. The main central elements of spray fluidized bed equipment, which have also been discussed by Jacob (2007), are: .

.

.

The gas distributor. The gas distributor separates the plenum from the processing chamber of the fluidized bed. It must prevent the passage of particles and have enough mechanical strength to avoid excessive vibrations or deformation. Pressure drop in the distributor should be low, but sufficient for good fluidization of the particles. Common forms of the gas distributor are porous – usually sintered – plates, simple perforated plates or wedge-wire plates, which consist of profiles welded on supporting metallic elements. Distributors can direct the gas flow vertically to their surface (i.e., upwards) or impose another angle, depending on the orientation of their perforation or gaps between profiles. Distributors with a gas flow angle other than 90 blow the particles in a certain direction, providing a conveying or transport effect. The process chamber. The chamber where particle fluidization and droplet spraying take place can have different footprint cross-sections, from circular to lengthy rectangular. The cross-sectional area can remain constant or increase in a vertical direction, leading to various conical or prismatic geometries. Furthermore, the process chamber can be uniformly open or divided by partitions in different ways. The gas-inlet chamber and exhaust chamber usually follow the segmentation of the process chamber. The spray system, consisting of nozzles and liquid feed lines. Nozzles are usually specified by the average size and size distribution of the droplets that they produce, by their spray pattern (full cone, hollow cone, flat jet), the spray angle, the average and the distribution of droplet velocity, and the range of possible liquid feed rates. Atomization of the liquid takes place by contact with high-velocity gas (pneumatic or binary nozzles) or by expansion from high pressure (hydraulic nozzles). Alternatively, rotating discs or piezoelectric elements that vibrate with a high frequency can be used to disperse the liquid into droplets. Pneumatic nozzles are the commonest type for applications of the fine chemical, pharmaceutical or food industries. The nozzles can be installed to spray from above on, or from below in the fluidized bed (denoted by “top spray” and “bottom spray,” respectively). Additional options are provided by horizontal or tangential spray.

j333

j 7 Particle Formulation in Spray Fluidized Beds

334

Placement and orientation of the nozzles have an influence on the extent of possible overspray as well as on the tendency of – undesired – deposit formation on the atomizing devices and their supporting structure. Moreover, there is a more or less strong interaction between nozzle and fluidization flow. As to the number of nozzles, it can vary from one in small units to dozens in large industrial plants, including the possibility of multi-head nozzles. When many nozzles are used, their arrangement in, for example, arrays, and the distribution of the liquid feed become critical issues. Modern liquid feed systems not only achieve a controllable and safe metering of liquid, but also provide uniform conditions among multiple nozzles and enable the detection of malfunction by, for example, fouling of individual atomizers.

Each of the mentioned main constructive elements of spray fluidized bed equipment can be used to manipulate product properties in the desired direction. For example, droplet size and spray pattern have an influence on the particle wetting and on the local liquid distribution in the fluidized bed – thus also on particle growth kinetics, the type of particle size enlargement (agglomeration in comparison to granulation and coating) and product properties (e.g., particle porosity and density, and surface morphology). However, even more opportunities for product design result from combining manipulations of the gas distributor, the process chamber and the spray system with each other. In the following, three examples of such combined manipulations will be briefly discussed, which result in Wurster equipment, in so-called ProCell units, and in lengthy rectangular (“horizontal”) fluidized beds. Wurster equipment is named after Dale Wurster – a professor at the MadisonWisconsin University who worked on particle and tablet coating in the 1950s (see for example, Wurster (1959)). A typical configuration of such equipment is depicted in Fig. 7.35. As the figure shows, it combines the following aspects: .

A slightly conical process chamber with a tubular partition placed in the center of it, at some distance from the distributor plate;

Fig. 7.35 Schematic representation of Wurster coater.

7.5 Influence of Apparatus Design . .

Bottom spray in the tubular partition; A segmented gas distributor plate with higher open cross-section below the Wurster partition than in the outer region of the equipment.

The higher open cross-section in the center of the distributor plate results in a high gas velocity in the Wurster partition. Particles are transported pneumatically in a vertical direction inside the tube, while being sprayed with liquid. The uniform, onedimensional character of the flow field reduces the number of interparticle collisions and provides similar residence times for all particles in the tube. Consequently, every particle is covered with a liquid film of approximately the same thickness, whereas agglomeration between the particles is suppressed. As soon as the particles leave the tube, they fall in the outer bed region, where they have time to dry under slowly bubbling fluidization conditions due to a much lower gas velocity. Then, the particles return to the Wurster tube through the gap between this tube and the distributor plate, and the cycle is repeated. The process is quite adequate for the application of uniform coatings on particles, so that it finds ample use in, for example, the pharmaceutical industry. The circulation of solids between the two compartments of the equipment (Wurster tube and outer bed) can be controlled by variation of the geometrical parameters. An even more radical deviation from the classical concept of fluidized beds is realized in units such as those depicted in Figs. 7.36 and 7.37. The design is based on a patent by M€orl et al. (2002), which was successfully put on the market by the Glatt Co., Weimar, Germany, under the name ProCell (Jacob, 2009). As the figures show, ProCell does not have any distributor plate. Instead, the air comes in by means of two gaps created between two throttle cylinders (drum or valve rollers) placed symmetrically at both sides of a central profile (partition). Each drum has a flat cut, so that the size of the air inlet gap and the air inlet velocity can be adjusted by simply rotating the drum. The chamber of the equipment is triangularly prismatic. Combination of this

Fig. 7.36 Scheme of ProCell spouted bed apparatus.

j335

j 7 Particle Formulation in Spray Fluidized Beds

336

Fig. 7.37 Lab-scale ProCell model, as used by Hoffmann et al. (2011) for mass transfer measurements.

specific chamber geometry with the air inlet and the central partition leads to spatially correlated spouted bed flow patterns. The central partition unit combines the gas flow coming from the right with the gas flow coming from the left into a flow field that is symmetrical to the central plane and vertically oriented in the vicinity of this plane, in the typical form of a spout or jet. Spraying is performed in the spout, either from above or – more frequently – from below, at the tip of the partition (Fig. 7.36). The recirculation of solids to the spout is provided by the side contour (planar sides of the triangular prism). A great advantage of ProCell is that it enables the processing of otherwise untreatable solids: materials with a broad particle size distribution, very small or very large size or density, irregular shape or sticky surface (Jacob, 2009). The apparatus has self-cleaning features. Moreover, it is easy to implement indirect heating at the central partition, or even cooling during the return of solids along the side contour. Due to the triangular shape of this contour, the apparatus can be operated with less hold-up and – in the continuous mode – with less residence time than conventional fluidized bed units, which means less deactivation of

7.5 Influence of Apparatus Design

sensitive ingredients. Finally, experiments conducted by Hoffmann et al. (2011) show that heat and mass transfer between the particles and the gas in ProCell equipment are – in terms of overall efficiency – much more intensive than in conventional fluidized beds or in conventional spouted beds. This is attributed to the very high air velocities in the gas inlet. Since the apparatus is prismatic, it can be scaled up by prolongation in the third (horizontal) direction. Additionally, the described design can be replicated in the lateral direction by use of more than one partition profile in parallel. To facilitate the construction of lengthy horizontal units one can refrain from the adjustability of the air inlet, using constant width gaps instead of the roller valves. Different zones can be defined in such lengthy equipment by subdivision of the plenum and by schemes of nozzle placement along the device. Different zones can also be defined in combination with distributor plates, as depicted in Fig. 7.38. Such a horizontal fluidized bed unit has an extended rectangular design of the process chamber. The gas distributor usually has a transport effect, moving the solids from the inlet to the outlet in continuous operation. Segmentation is performed by subdivision of the plenum and by the placement, or not, of nozzles. Since each of the different compartments of the plenum can be equipped with a separate gas supply, it is possible to conduct various sequences of processing steps. For example, a product can be granulated in the first and second segments, dried in the third, and cooled in the final fourth section (Fig. 7.38). The fluidization chamber may be open (as in Fig. 7.38), or it can follow the segmentation of the plenum by means of partitions (weirs). Solids transportation from one segment to the next takes place as overflow or underflow. In the case of underflow, openings that cover all the width of the equipment or just a part of it can be used. Smaller underflow openings (so-called “rat holes”) can be placed to the right

Fig. 7.38 Horizontal fluidized bed unit with segmentation.

j337

j 7 Particle Formulation in Spray Fluidized Beds

338

and to the left in alternating sequence, in order to create – in combination with appropriately aligned transport distributor pieces – a meander-like path of solids motion. The nozzles can be arranged in top-spray or in bottom-spray, with constant or variable spray rate along the equipment. An important aspect of chamber design and segmentation is that the residence time distribution of the solids can be varied within a broad range – from the well-mixed behavior, which is typical for circular apparatuses, to nearly plug flow conditions. The aspect of residence time distribution will be discussed further in the following section.

7.5.2 Residence Time Distribution

The residence time of particles in fluidized bed processes has a huge influence on the product characteristics. A prominent example is fluidized bed drying, where the residence time determines the product moisture, which is an important factor for storage stability (see Chapter 1 in Volume 2 of this series). Another example is fluidized bed coating. The goal of fluidized bed coating may be to place an active ingredient, for example, an enzyme, on inactive carrier material (Hede et al., 2009). If the residence time of the carrier cores is too short, then only an insufficient amount of active ingredient is deposited, and the resulting product is ineffective. On the other hand, if the residence time is too large, an overdose of active ingredient is applied. Additionally, partial inactivation of the enzyme by thermal exposure or agglomeration can take place. In both cases, the product may not fit the specifications for its application. Appropriate equipment design is one way to guarantee the desired product properties by manipulation of the residence time of particles in fluidized bed processes. Moreover, it is necessary to identify suitable process inputs with an influence on residence time for use in feedback control loops. An appropriate design may be that of horizontal equipment, as the one depicted in Fig. 7.38. Due to the fluidization air, particles move along the chamber of such equipment, establishing a complex velocity field. Additionally, rising bubbles of gas mix the solids. The air velocity distribution in a horizontal direction has, therefore, great influence on both the transport and the mixing of particles along the bed. To quantify particle motion, simulations with computational fluid dynamics (CFD) were conducted for a three-dimensional model of an industrial pilot plant within the ANSYS FLUENT environment (version 6.2). No partitions were present in the chamber – in close correspondence to the geometry shown in Fig. 7.38. A Eulerian approach was adopted for the granular phase in the simulations. In Fig. 7.39 the resulting volume fractions of the disperse phase are shown for different total volume flow rates of air. An increase in the bed height and more uniform bubbling can be observed in the process chamber at higher air velocity. Moreover, the higher intensity of particle movement leads to a significant change in the dynamic behavior of the apparatus: At higher particle velocities more particles are ejected into the freeboard region and are distributed along the whole process

7.5 Influence of Apparatus Design

Fig. 7.39 Contours of solids volume fraction in horizontal equipment for volume flow rates of fluidization air at (a) 500 m3 h
1, (b) 700 m3 h
1.

chamber. This will have a direct influence on the residence time of particles in the bed and thus on the product quality at the outlet of the apparatus. In order to quantify the influence of the gas flow rate on the residence time of the particles a simple model can be used that represents the horizontal apparatus by a series of continuously operated stirred tank reactors (CSTRs). The principle of this model is illustrated in Fig. 7.40. The size (length) and number of the tanks express the intensity of back-mixing (mixing in the direction of solids transport). They are fictitious for an open process chamber (as in Fig. 7.38), but may correspond to

j339

j 7 Particle Formulation in Spray Fluidized Beds

340

exhaust feed

L1

discharge L1 L1

inlet air Fig. 7.40 Series-of-tanks model for horizontal equipment.

physically existing individual compartments in the case of built-in partitions (weirs). The average residence time under continuous, steady-state conditions results from the mass of particles in the fluidized bed Mbed and the mass flow rate of _ s: solids M
t ¼

_s M Mbed

ð7:29Þ

Using a correspondingly defined dimensionless residence time tr ¼

t
t

ð7:30Þ

the normalized residence time distribution E(tr) in a series consisting of N stirred tanks can be derived analytically. It is: E ðtr Þ ¼

NN tN
1 expð
N tr Þ ðN
1Þ! r

ð7:31Þ

Figure 7.41 shows that very different residence time distributions are obtained for different numbers of tanks in series. In order to obtain a narrow residence time distribution with limited variance a high number of effective or real compartments may be necessary. To find out the effective number of compartments that corresponds to the horizontal equipment of Fig. 7.38, tracer experiments were conducted. After a prescribed time at steady operating conditions a granular tracer substance was added to the solid feed stream. Samples were taken at the discharge of the plant at defined time intervals and analyzed by means of pH measurement after dissolution in de-mineralized water. The pH value of the solution corresponds to the concentration of tracer, and can be used to calculate residence time curves. Measurements were carried out for the two already mentioned total gas flow rates of 500 and 700 m3 h
1 under otherwise the same conditions (mass flow rate of solids: 20 kg h
1, bed mass: 25 kg, tracer mass: 5 kg, test duration: 140–150 min).

7.5 Influence of Apparatus Design

Fig. 7.41 Residence time distributions of perfectly stirred tanks in series, depending on the number of tanks.

The results are plotted in Fig. 7.42 and make clear that the air flow rate has a direct influence on the residence time distribution. With increasing air flow rate (Fig. 7.42b) the fluidization intensity increases and the distribution widens. This is in qualitative agreement with the previously discussed CFD results. Normalization and comparison with the tanks-in-series model in Fig. 7.43 shows a difference in the shape between measured curves and standard residence time distributions. Specifically, there is a time delay at the beginning of the experiments which is independent of fluidization velocity. This behavior is probably related to the fact that the outlet rotary valve of the apparatus was operated at low speed, producing some jam of material in the connection between the bed outlet and the valve. Consequently, tracer particles reached their sampling point immediately after the valve with some delay. Additionally, the shape of the measured curves is modified in comparison to the theoretical curves by a shift of their maxima to lower values of time. A possible reason for this shift is the use of an air distribution plate with transport effect in the experiments. This distributor type is designed to create a horizontal air velocity component pushing the solids to the outlet side of the process chamber. Therefore, distributions of hold-up along the equipment, which have not been accounted for in the evaluation, may have been present. If one attempts – despite the mentioned qualitative deviations – a guess of the number of effective compartments on Fig. 7.43, something like N ¼ 2 would be the outcome. Consequently, the equipment of Fig. 7.38 would perform in the sense of a narrow residence time distribution better than a cylindrical fluidized bed (which corresponds to just one perfectly stirred tank), but would still give quite broad distributions of product properties at the outlet. To squeeze such distributions towards uniformity, partitions (weirs) can be placed in the fluidization chamber. Additionally, the equipment can be further elongated. The combination of both

j341

j 7 Particle Formulation in Spray Fluidized Beds

342

–3

10

(a)

–1

residence time distribution [min ]

25

20

15

10

5

0 0

10

20

30

40

50

60

70

80

90

100 110 120 130 140 150

time [min]

–3

(b)

10

7

–1

residence time distribution [min ]

8

6 5 4 3 2 1 0 0

10

20

30

40

50

60

70

80

90

100 110 120 130 140 150

time [min] Fig. 7.42 Residence time distributions measured in horizontal equipment for volume flow rates of fluidization air at (a) 500 m3 h
1, (b) 700 m3 h
1; (two measurements for each gas flow rate under essentially the same conditions).

measures leads to horizontal fluidized beds with the residence time behavior of many (10, 20, or more) tanks in series. In fact, Baker and Lababidi (2010) point out that large industrial equipment of this type comes close to the limit of N ! ¥, which would mean plug-flow of the solids. Fyhr et al. (1999) report the value of N ¼ 10 for a horizontal fluidized bed dryer. Nilsson and Wimmerstedt (1988) use the dispersion model instead of the tanks-in-series model to express the mixing of solids in a longitudinal direction, and present – based on measured data – a correlation for the respective dispersion coefficient. Since the dispersion model and the tanks-in-series model are equivalent to each other, this correlation can be used to estimate the effective number of compartments for technical applications.

7.5 Influence of Apparatus Design

(a) 1.8 N=2

1.6

N=4 N=6

normalized RTD

1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

reduced time τr (b) 1.0 0.9

N=2

0.8

N=4

normalized RTD

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

reduced time τr Fig. 7.43 Normalized residence time distributions derived from the measured data of Fig. 7.41 in comparison with the tanks-in-series model at (a) 500 m3 h
1, (b) 700 m3 h
1 of fluidization air.

In addition to partition and elongation, the gas flow rate in the different compartments of horizontal fluidized bed equipment can be used to manipulate the residence time distribution of the particles and, thus, the quality of the outlet product. The use of CFD can support the design of compartments, partitions and weirs. Not only the here briefly presented two-fluid (Euler–Euler) approach, but also Euler–Lagrange approaches (so-called discrete particle models, DPM) can serve this goal, as will be discussed in Section 7.5.4.

j343

j 7 Particle Formulation in Spray Fluidized Beds

344

7.5.3 Dispersive Growth in Batch Granulation

In many processes, such as coating or layering granulation, the width (or variance) of the particle property distribution is an important measure of product quality. The variance in particle size is directly related to the variance in thickness of the coating layers that has great influence on the product characteristics, for example the release time of active ingredients in pharmaceuticals. Coating, as well as layering granulation, has a well-established basis in mathematical modeling. Using a population balance approach for the particle property distribution, for example the distribution of size d, yields for the number density n the following type of population balance equation qn qðGnÞ þ ¼ N_ in
N_ out þ P qt qx

ð7:32Þ

with corresponding initial and boundary conditions. Here, G denotes the growth rates of the particles, N_ in is the flow density of particles entering the balance volume, and N_ out is the flow density of particles leaving the system, for example by an outlet. Particle production processes, for example agglomeration and breakage, are collected in the net production density P. In the case of coating (or layering) only, this term is not present in the balance equation. In modeling, the growth rate G is commonly considered to be independent of the particle property. For example, M€orl et al. (2007) assume that the growth velocity of particles is directly related to their surface area in the fluidized bed, to obtain for spherical particles G¼2

_ sus M rs p m2

ð7:33Þ

_ sus denotes the mass flow rate of solid material entering the system with the Here, M spray, rs is the density of the solidified layers, and m2 is the second moment of the particle size distribution, so that p m2 equals the total particle surface area. The population balance equation then reads as qn qn þ G ¼ N_ in
N_ out þ P qt qd

ð7:34Þ

This fundamental assumption simplifies the structure of the mathematical problem significantly. Additionally, it is also backed by various experimental results – see for instance the references in M€orl et al. (2007). Under the assumption of property-independent growth of particles, Eq. 7.34 represents a quasi-linear transport equation. Although the transport velocity can vary with process time, it is assumed to be the same for each particle at every point in time. This ultimately results in the effect that the initial distribution (at t ¼ t0) is shifted along the property coordinate, see Fig. 7.44. The shape of the distribution is

7.5 Influence of Apparatus Design

Fig. 7.44 Evolution of particle size distribution under a property independent growth law; the shape of the distribution is conserved.

preserved during this transport, due to the assumption of the property independence of the growth rate. However, in various coating and granulation experiments a change in the shape of the distribution is observed, for example a dispersion of the distribution. In Fig. 7.45 the results of five different coating experiments are shown, which were conducted in: Wurster equipment, a conventional fluidized bed apparatus (FB) in top and bottom spray configuration, and a spouted bed apparatus (SB) in top and bottom spray configuration. Although identical initial conditions were used and the process conditions are comparable (see Tab. 7.6), different final distributions are achieved. This effect cannot be explained by the common model of Eq. 7.34, but it underlines the influence that different types of equipment with different flow patterns may have on the process result. In order to capture deviations from property independent growth in terms of the population balance equation, two possibilities are imminent. First is the introduction

Fig. 7.45 Results of coating processes in apparatuses of different design: (a) spouted bed (SB), (b) Wurster equipment and conventional fluidized bed (FB); Starting from an identical initial distribution, different final distributions are achieved under comparable process conditions.

j345

j 7 Particle Formulation in Spray Fluidized Beds

346

Tab. 7.6

Process conditions for the fluidized bed coating experiments depicted in Fig. 7.45.

Bed material Solution Spraying rate Fluidization gas flow rate

Cellets Na-benzoate, 30 wt% 18 g min
1 70 kg h
1 for Wurster and FB, 100 kg h
1 for SB 90  C 2.5 bar 1 kg

Gas temperature Nozzle pressure Initial bed mass

of new growth models – of the same structure as Eq. 7.33, but not propertyindependent – using higher moments of the distribution (Hoffmann et al., 2010). Introducing generalized growth rates Gi ¼ 2

_ sus di
2 M rs p mi

ð7:35Þ

which are based on an arbitrary moment mi of the particle size distribution, the total growth rate of particles G can be defined as: G¼

N X

lk Gk

with

N X k¼0

k¼0

lk ¼ 1

ð7:36Þ

The constraint on the choice of the parameters lk is due to the required mass conservation in the model. A selection of l2 ¼ 1 leads to the well known surfaceproportional growth law (Eq. 7.33). This approach can come close to experimental results, as Fig. 7.46 shows. The data used for this comparison were measured for batch granulation in a conventional

Sim., t = 5290 s (G3)

0.14

25

Sim., t = 5290 s (G5)

0.12

20

Sim., t = 0 s Exp., t = 0 s Exp., t = 5290 s

15 10

0.08 0.06 0.04

5 0 0.4

Experiment Simulation (G5)

0.1

σ [mm]

q [mm–1] 0

30

0.02 0.6

0.8

1

1.2

1.4

x [mm] Fig. 7.46 Comparison of experimental and simulation results for higher order growth models. Plotted are the approximations for volume-proportional growth ðl3 ¼ 1Þ and the

0 0

30

60

t [min] fifth-order model ðl5 ¼ 1Þ, which can be interpreted as mass flux of particles during sedimentation from a suspension.

90

7.5 Influence of Apparatus Design Process conditions for the experiment depicted in Figs. 7.46 and 7.48, and parameters of the two-compartment model. Tab. 7.7

Bed material Solution Spraying rate Fluidization gas flow rate Bed temperature Atomizing gas flow rate Initial bed mass Volume fraction spray zone Mean residence time spray zone

c-Al2O3 Na-benzoate, 30 wt% 100 ml min
1 500 kg h
1 80  C 50 kg h
1 4 kg 2% 4s

top-spray fluidized bed with the process conditions summarized in Tab. 7.7. There is a striking difference between these data and the behavior illustrated in Fig. 7.44, because the size distribution broadens very significantly with time (the standard deviation s increases). This observation is analogous to the broadening of concentration profiles while moving along a chromatographic column. It can, thus, be denoted by dispersive growth, whereby the dispersion does not take place in space, but along the property coordinate. However, the higher order growth rates used to approach the measured results in Fig. 7.46 are hard to interpret in terms of physical relations. The parameters needed have to be identified and tuned in a cumbersome plant-specific procedure, which complicates statements on the generality and scalability of the results. The second approach favors the use of compartment models to describe the evolution of the property distribution. It has the virtues of needing only a limited number of new parameters to be applied and using the property-independent growth law to achieve good results. In the following, a compartment model for the layering granulation of particles in a fluidized bed is presented, based on the recent work of Silva et al. (2010). In Fig. 7.47 the principal sectioning of the fluidized bed according to this model into a spraying zone and a drying zone is depicted. The fraction of the total volume occupied by the spraying zone is denoted by a, so that the volume fraction of the

Msus n1−α τ1−α α

1-α

∂n1−α ∂t

nα τα

∂nα ∂t

Fig. 7.47 Two-compartment model of a fluidized bed with a spraying zone and a drying zone.

j347

j 7 Particle Formulation in Spray Fluidized Beds

348

drying zone is 1
a. Only particles that are inside the spraying zone are sprayed with liquid. Due to the fluidization, these particles are then transported (after a certain mean residence time ta ) into the drying zone, where the sprayed liquid dries and new solid layers are built up. In mathematical terms the system of Fig. 7.47 is described by two population balance equations, one for each of the two compartments: qna qðGna Þ na n1
a þ ¼
þ qt ta t1
a qx

ð7:37Þ

qn1
a na n1
a
¼ þ qt ta t1
a

ð7:38Þ

where na is the particle size density distribution in the spraying zone and n1
a is the distribution in the drying zone of the fluidized bed. The parameters ta and t1
a are the mean residence times in the spray and drying zones, respectively. They incorporate information on the flow regime (particle exchange rates), and are also directly dependent on apparatus design, for example on nozzle placement and the dimensions of the spraying cone. As was shown by Silva et al. (2010), these model equations can be solved analytically for special initial conditions to gain information on the evolution of the mean and the variance of the initial distribution during the layering process. The same authors also provide analytical solutions for the particle size distributions and experimental results. Particle size distributions obtained by the two-compartment model are depicted in Fig. 7.48, along with the same experimental results as in Fig. 7.46 (Tab. 7.7). The simulated standard deviation is also compared to the experimental standard deviation in this figure, and a good agreement between the results can be observed.

(a)

(b) 30 25

0.1

20 15 10

0.08 0.06 0.04

5 0 0.4

Experiment Simulation

0.12

σ0 [mm]

q0 [mm–1]

0.14

Sim. t = 0 s Sim. t = 5290 s Exp. t = 0 s Exp. t = 5290 s

0.02 0 0.6

0.8

1

x [mm]

1.2

1.4

0

30

60

90

t [min]

Fig. 7.48 Comparison of experimental and simulation results for the two-compartment model; (a) particle size distribution and (b) standard deviation are described well by the model with parameters as listed in Tab. 7.7.

7.5 Influence of Apparatus Design

The parameters to be estimated in this model are a, ta , and t1
a . Here, it is sufficient to know two out of three parameters, as the following equality holds from mass conservation arguments: t1
a 1
a ¼ a ta

ð7:39Þ

The mean residence times ta and t1
a can either be calculated by CFD in the apparatus or determined by measurements, for example via particle image velocimetry (PIV, for a detailed description see Chapter 5 in Volume 2 of this series). The volume fraction of the spray zone a can be estimated from the apparatus dimensions, bed porosity, flow conditions, and the spraying cone of the nozzle. Summarizing, conventional coating and granulation models that utilize propertyindependent growth laws are often not able to describe the experimentally observed temporal change in the variance of a given initial distribution. One way to extend the existing mathematical models is the introduction of new growth models which use higher moments of the distribution to describe the growth. However, these growth laws are often hard to explain on a physical basis, and the parameters necessary in order to apply the model are difficult to identify. Another way to derive dispersive growth behavior from population balances is to divide the fluidized bed into two interacting zones – a spraying zone and a drying zone – while still relying on the wellestablished property independent growth rate for the spraying zone. As was shown by experimental results, this approach is able to describe the observed change in the particle size distribution satisfactorily. Also, the necessary parameters can be inferred from the geometry of the apparatus and the hydrodynamic conditions. This leads to the conclusion that compartment models are suitable to describe a large class of coating and granulation processes in which a change in the variance of the property distribution is observed. 7.5.4 Discrete Particle Modeling of a Wurster Coater

As shown in the previous sections, the influence of residence time on the product quality – in particular on the variance of the PSD – can be significant. Moreover, it was shown that the residence time can be controlled by special design of the apparatus in the form of a spouted bed, Wurster coater or horizontal fluidized bed. In Section 7.5.2 a Euler–Euler approach was presented to study the velocity distribution of the solids in a continuously operated, horizontal fluidized bed. Another popular tool for the simulation of disperse systems is the Discrete Particle Model (DPM), which resolves the equation of motion for each individual particle considering the momentum exchange within the disperse phase due to particle collisions and with the gas phase. The DPM has been applied to fluidized bed processes by several authors (Tsuji et al., 1993; Hoomans et al., 1996), including some attempts at discrete modeling of fluidized bed spray granulation processes (Goldschmidt and Kuipers, 2003; Kafui and Thornton, 2008). The DPM method is still limited by the number of particles (order of magnitude: 106) that can be considered and process time (order of

j349

j 7 Particle Formulation in Spray Fluidized Beds

350

magnitude: seconds) that can be simulated. Nevertheless, due to the rapid development of computer speed and memory, the size of the simulated systems has recently been significantly increased, so that now the DPM is a promising method.

7.5.4.1 Principles of the DPM The fundamental equation of the DPM is the momentum equation of each individual particle with mass Mi and volume Vi, that can be described by Newton’s second law Mi

 dvi d2 r i Vi b
ug
vi þ Mi g þ Fcontact;i þ Fpp;i ¼ Mi 2 ¼
Vi rP þ 1
e dt dt

ð7:40Þ

where vi is the velocity and ri the position of the particle. The forces on the right-hand side of Eq. 7.40 are, respectively, due to the pressure gradient, drag, gravity, contact forces (i.e., forces during collisions) and relatively distant particle–particle interactions (for instance van der Waals forces). The inter-phase momentum transfer coefficient b is frequently modeled by combining the Ergun equation for dense regimes (e < 0.8) ! m ð1
eÞ2 b ¼ 2 150 þ 1:75ð1
eÞRe ð7:41Þ e d and the correlation proposed by Wen and Yu (1966) for the more dilute regimes (e  0.8)   m 3 ð1
eÞ cd Re 2:65 b¼ 2 ð7:42Þ d 4 e with 8 > >
> : 0:44

if Re < 1000

ð7:43Þ

if Re > 1000

where Re is the particle Reynolds number and e is the gas volume fraction (porosity). Hill et al. (2001) and Beetstra et al. (2007) have developed own drag correlations based on Lattice–Boltzmann simulations, which should give a more accurate representation of the fluid–particle interactions in dense gas–solid flows for the price of higher computational effort. All correlations developed in this way until now are, however, limited to monodisperse or bidisperse (Beetstra et al., 2007) arrays of spherical particles. In the case of a collision between two particles, the contact forces are calculated according to a contact model based on the theory developed by Hertz (1882) for the normal impact and a non-slip approximation of the model by Mindlin and Deresiewicz (1953) for the tangential part of the contact force, as proposed by Tsuji et al. (1992). The normal contact force is 3=

Fcontact;N ¼
kN sN2 n
mN vN

ð7:44Þ

7.5 Influence of Apparatus Design

Here, n is the normal unit vector and vN is the relative velocity at the contact point. The elastic part of the contact force is represented by a non-linear spring, assumed proportional to the spring stiffness kN and to sN= (sN: displacement). Additionally, to account for viscoelastic material properties that cause energy dissipation, a damping factor mN related to the coefficient of restitution is included in the model: pffiffiffiffiffiffiffiffiffiffi 1= mN ¼ 2a MkN sN4 ð7:45Þ 3

8 ln eN > < pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi if eN 6¼ 0 a¼ p2 þ ln2 eN > : 1 if eN ¼ 0

2

ð7:46Þ

The normal coefficient of restitution eN is defined as the ratio of the rebound velocity to the impact velocity: sffiffiffiffiffiffiffiffiffiffiffi Ekin;R jvR j eN ¼ ð7:47Þ ¼ v Ekin Usually this value can be obtained from experiments. The symbol M in Eq. 7.45 denotes a combined mass. The tangential component of the contact force Fcontact,T is calculated analogously, as presented by Tsuji et al. (1992). The gas phase is considered as a continuum. Usually, the geometry of the apparatus is discretized in mesh cells and the motion of the gas phase is calculated using volume-averaged Navier–Stokes equations. The influence of particles on the velocity profile of the gas phase is accounted for by adding a sink term to the momentum balance for the gas. To reduce the numerical effort and to avoid local extrema of the solids concentration, the gas flow field is resolved on a relatively coarse grid compared to the particle diameter. The side length of an average Eulerian grid element should be in the range of 10 times the particle diameter (Deen et al., 2007). 7.5.4.2 Parameters for the DPM Simulation A Wurster coater similar to the equipment of Fig. 7.35 was simulated by DPM. The original geometry and the meshed model are shown in Fig. 7.49. To investigate the influence of design and process parameters on the fluid dynamics in the Wurster coater, DPM simulations were performed in 3D with 150 000 spherical particles. At a particle diameter of 2 mm and an average particle density of 1500 kg m
3, this corresponds to a batch size of 0.94 kg. According to experimental results for c-Al2O3, the coefficient of restitution for particle–particle and particle–wall collisions was set to 0.8 and kept constant during all simulations. The operating variables are summarized in Tab. 7.8. Notice that air comes through the distributor plate in the Wurster tube and in the annulus around this tube. The respective velocities can be different, corresponding to distributor segments with different porosity; they are given in Tab. 7.8 as multiples of the minimal fluidization velocity umf. Additionally, air is blown in the Wurster tube by a nozzle. Air velocity at the tip of this nozzle is denoted by the atomizer or spout velocity. In a series of case

j351

j 7 Particle Formulation in Spray Fluidized Beds

352

Fig. 7.49 Simplified Wurster geometry and mesh for fluid dynamics simulation (10 000 tetrahedral cells).

studies, the velocity of the nozzle air, the distribution of air between the Wurster and the surrounding annulus, and the height of the Wurster gap were varied. The spraying of droplets in the system has not been accounted for. 7.5.4.3 Influence of the Spout Velocity The simulation results show that the velocity of the air injected via the nozzle has a strong influence on the fluid dynamics of the whole granulator. In Fig. 7.50, snapshots of the particle positions and their velocities after the same process time are displayed. The granulator is cut vertically in the middle, and the central plane and the parts behind that plane are shown. The color indicates the particle velocity: blue particles move slowly (v < 0.5 m s
1) and red particles are fast (v > 1.5 m s
1). It can be seen from Fig. 7.50 that for all spout velocities the particles are concentrated at the center in the lower part of the Wurster tube. For high spout

Tab. 7.8

Operating variables.

Variable

unit

Case 1

Case 2

Case 3

Case 4

Case 5

Case 6

Particle mass Fluidization air flow Atomizer air flow rate Velocity atomizer Velocity Wurster Velocity annulus Wurster gap distance

g m3 h
1

940 560

940 560

940 560

940 440

940 440

940 440

m3 h
1 m s
1 multiples of umf multiples of umf Multiples of d

0.9 20 10 5 5

4.5 100 10 5 5

7.2 160 10 5 5

7.2

7.2

7.2

160 10 3.8 5

160 10 3.8 10

160 10 3.8 15

7.5 Influence of Apparatus Design

Fig. 7.50 Instantaneous particle positions and velocity distributions inside the Wurster tube at t ¼ 1.4 s; colors indicate the velocity magnitude.

velocities (case 3) the particles are accelerated above the nozzle and transported upwards at high speed, which is indicated by red color. The fluidization regime is stable. At lower spout velocities (cases 1 and 2) the particle trajectories do not follow such a clear regular pattern. Slowly moving particles tend to block the upper end of the tube, inducing a slightly pulsating fluidization regime. Contrary to case 3, in cases 1 and 2 the particles already decelerate while they are transported to the upper end of the Wurster tube. The height of the particle fountain increases with higher spout velocity from 430 mm above the nozzle tip in case 1 to 490 mm in case 3. To assess the radial distribution of particles in the granulator, horizontal slices were cut out of the simulated geometry. This was done at two different heights, as shown in Fig. 7.51. The thickness of each of the slices is 10 mm. The first slice is situated in the lower part of the Wurster tube, just around the tip of the injection nozzle. The second

Fig. 7.51 Horizontal slices.

j353

j 7 Particle Formulation in Spray Fluidized Beds

354

Fig. 7.52 Instantaneous horizontal position and vertical velocity component of the particles in slice 1 at time t ¼ 1.4 s.

slice is at the upper border of the tube and allows visualization of the flow conditions of particles entering the expansion zone. Figures 7.52 and 7.53 show instantaneous particle positions in the two slices, seen from the top. The colors indicate the vertical component of the particle velocity. Red particles are transported upwards, blue particles fall downwards. A comparison can be drawn between the first three case studies to evaluate the influence of the spout velocity on the horizontal distribution of the particles and their vertical velocity component. It can be seen in Fig. 7.52 (slice 1) that the high spout velocity of case 3 tears particles towards the center of the Wurster tube. Particles coming out of the tube slip downwards along the apparatus walls of the annular part of the equipment. The low spout velocity of case 1 reduces the bed expansion in the annulus. A dense bed is observed, whereas high porosities due to rising bubbles occur in cases 2 and 3. In the middle of the ring the movement of the particles does not follow a clear pattern, as the flow in this zone is dominated by rising bubbles. High spout velocities seem to promote the formation of larger bubbles. In slice 2, only a few particles are present (Fig. 7.53). Their movement is directed upwards inside the tube and downwards in the annulus in all three cases,

Fig. 7.53 Instantaneous horizontal position and vertical velocity component of the particles in slice 2 at time t ¼ 1.4 s.

7.5 Influence of Apparatus Design

Fig. 7.54 Time-averaged fluid velocity distribution in the coater; colors indicate the velocity magnitude.

indicating that the circulating regime is intact over a wide fluidization range. Along the border of the Wurster tube, a deceleration of the particles can be observed in case 1, whereas at high spout velocity all particles move upwards at a velocity faster than 1 m s
1 at this level (case 3). For low spout velocities as in case 1, some particles are moving downwards in the Wurster tube, which indicates unwanted back-mixing. Figure 7.54 shows the time-averaged velocity field of the fluid for the three simulation cases. Jet velocities higher than 14 m s
1 are cut off and displayed in white color in the graphic. It can be seen that the nozzle jet has a low injection depth. Even at the highest injection velocity of 160 m s
1 at the nozzle tip (case 3), the fluid velocity decays to 14 m s
1 within less than 30 mm from the tip. The momentum introduced by the jet is immediately transferred to the particles in a relatively small zone above the nozzle tip. In case 1, the penetration depth of the jet injected at 20 m s
1 is almost invisible. In contrast to the particle velocity distribution, the gas flow field inside the Wurster tube is hardly influenced by the jet velocity. This is because even in case 3 less than 2% of the total gas flow is injected via the nozzle. More than 98% of the gas enters the system via the distributor plate at the bottom of the granulator. In Fig. 7.55, snapshots of the volume fraction of the particle phase are shown. High particle concentrations are observed along the apparatus walls, where the particles slip downwards, and at the center of the Wurster tube around the nozzle shaft. 7.5.4.4 Influence of the Wurster Gap Distance The gap distance between the distributor plate and the Wurster tube is a parameter that strongly influences the fluid and particle dynamics inside the granulator. Gap heights of 10, 20 and 30 mm were compared. It can be seen in Fig. 7.56 that a larger gap below the Wurster tube increases the number of particles that are transported into the tube. In case 4, particles are only present at the center of the lower half of the tube. All particles rise at a velocity above 1 m s
1 which is indicated by green and red color. In contrast, in cases 5 and 6

j355

j 7 Particle Formulation in Spray Fluidized Beds

356

Fig. 7.55 Instantaneous particle volume fraction in the Wurster coater at t ¼ 1.4 s.

particles can be found spread over the whole diameter of the Wurster tube. Near the wall of the tube they move at low velocity, indicated by blue color. For the functionality of a Wurster coater it is important that the particles are coated homogeneously. This can be achieved by a regular fluidization regime with a well defined circulating movement of the particles. The thickness of the coating layer on a particle’s surface is controlled by the residence time of the particle inside the spray cone at the nozzle tip. With this perspective, the regime of case 3 seems optimal, as all particles move on similar trajectories and at similar velocity inside the Wurster tube. This will yield a narrow residence time distribution of the particles in the spray zone. Contrary to that, a larger Wurster gap distance, as used in cases 5 and 6, induces a broad residence time distribution of the particles in the Wurster tube and, hence, an irregular coating layer thickness. Summarizing, it can be stated that a homogenous velocity distribution of the particles inside the Wurster tube was achieved at a high spout velocity of 160 m s
1 and at low gap distances between the tube and distributor of 5 times the particle

Fig. 7.56 Instantaneous position and velocity distribution of the particles in the Wurster zone for different gap distances at time t ¼ 1.4 s.

7.6 Neural Networks, Encapsulation

diameter. This set-up will provide a homogeneous coating quality, as the residence time of the particles inside the spray zone is similar for all particles and relatively short. Internal recirculation in the tube and back-mixing of particles falling down from the expansion zone is avoided. The results show that the DPM method offers great potential to describe fluidized bed processes on the scale of individual particles using only physically based input parameters. Especially, the exact description of particle–particle interactions using a sophisticated contact model is a strong feature of this method. However, limitations due to the high numerical effort are still important. Process scale granulators cannot be described within a reasonable simulation time. Therefore, based on the example of fluidized bed granulation, the DPM method should be seen as a tool relevant for obtaining on the micro-scale kinetic rate constants which define the dynamics of the process. For example, the wetting kinetics could be described by approximating the particle moisture content with the residence time of individual particles in the spray zone. Together with this information, results on the collision frequency and collision velocity distributions can enable the definition of physically based aggregation rate terms. Instead of using an empirical kernel, such DPM-based growth kinetics could be implemented in a population balance model to enforce the predictive description of granulation or agglomeration processes on the macro-scale.

7.6 Neural Networks, Encapsulation

Encapsulation of liquids in granular products is an important process used mainly in the food and feed industry to protect valuable ingredients such as flavors, aromatic oils or fragrances against ambient oxygen, to improve the storage stability, and to reduce contamination risks. As discussed in Chapter 6, encapsulation can be conducted by spray drying of an emulsion, where the active ingredient to be retained is in the droplet phase. The same goal can also be achieved by spray fluidized bed granulation, with the additional advantage of a relatively coarse-grained, easy-tohandle product. In the simplest case, the sprayed emulsion is, again, an oil-in-water system, where the water-insoluble active ingredient is homogeneously suspended in an aqueous matrix solution. Spreading of the spray on the surface of the fluidized particles and drying lead to layered growth, so that at the end of the process the remaining solid material (the matrix) contains the dispersed active ingredient. A typical process development task is to assess the influence of operating conditions and formulation parameters on product quality. Since this can hardly be done on first principles, experiments are necessary. However, the number of such experiments should be kept as low as possible, and their results should be organized in a way that enables reliable interpolation. This can be achieved by the use of artificial neural networks, as will be outlined in the following with the help of an example. This specific example refers to the encapsulation of aromatic oils with different volatility (namely orange, mint or pergamot oil) by spray fluidized bed granulation.

j357

j 7 Particle Formulation in Spray Fluidized Beds

358

Tab. 7.9

Parameters of influence for spray fluidized bed encapsulation of aromatic oils.

Parameter

Unit

Min.

Median

Max.



47.5 34.5 19.4

82.0 49.0 47.6

Process conditions Product temperature Spray rate Residence time

C g min
1 min

41.0 16.0 7.5

Formulation of emulsion Active ingredient Oil Water Maltodextrin type 1 Maltodextrin type 2 Starch type 1 Starch type 2 Film forming agent Emulsifier

— % % % % % % % %

orange, mint, bergamot oil 10.0 19.0 47.7 48.0 0.0 29.0 0.0 1.0 0.0 8.0 0.0 3.5 0.0 2.8 0.0 0.1

20.5 50.0 40.7 29.0 30.0 40.0 3.0 3.0

The considered parameters of influence are summarized in Tab. 7.9. They are in total 12, including three operating parameters (product temperature in the process chamber, emulsion spray rate, and average residence time of the solids) and nine quantities that describe possible variations of the emulsion recipe, namely the kind of active ingredient (aromatic oil) considered, the mass fraction of this active ingredient and of water in the spray liquid as well as the kind and mass fraction of matrix materials (two maltodextrins and two starches) and additives (film former, emulsifier) used. Ranges of the variation of these parameters in the experiments are also given in Tab. 7.9. All trials (36 individual experiments) were conducted in a continuously operated, lab-scale ProCell unit (see Section 7.5.1 for a description of this special type of spouted bed equipment) under steady-state conditions. The inlet air temperature was adjusted depending on the target product temperature and the actual spray rate. Samples taken from the continuous product discharge were subject to manifold physical and chemical analysis, including the determination of particle size distribution, untapped bulk density, water content (by infrared), amount of active ingredient (by water steam extraction), and the observation of sphericity, surface morphology and internal structure by scanning electron microscopy (Fig. 7.57). However, the main focus was on two quantities: .

.

yield of active ingredient in the product (encapsulation efficiency or retention efficiency), which is the ratio of encapsulated to sprayed mass of active ingredient, and average particle size.

This focus results from the desire to find liquid formulations and process conditions that give both a high yield of active ingredient in the product and particle sizes in the right range.

7.6 Neural Networks, Encapsulation

Fig. 7.57 Cross-sections of granules containing encapsulated orange oil.

To correlate the mentioned target quantities with the various parameters of influence the concept of artificial neural networks (for more information on mathematical background and applications see, for example B€armann and Biegler-K€onig (1992), Braspenning et al. (1995)) has been applied. Specifically, a simple three-layer feed-forward network was established using the commercial software “NN-tool 2000” (B€armann Software, Germany). The scheme of such a network is shown in Fig. 7.58. The mentioned 12 parameters that describe process conditions and the emulsion recipe constitute the input layer, whereas the target indices of product and process quality are placed in the output layer. Hidden layer and connectivities are obtained by training the network with the help of the experimental data, including a cross-validation procedure. The trained network can first be applied to predict the used experimental results. Comparison of predicted values with this data in, for example, scatter plots is a measure of the accuracy with which the method can reproduce the available experimental information. This accuracy was found to be satisfactory for both target parameters – yield and average particle size. After training and the described control, the artificial neural network can be used to . .

evaluate the influence of process and product parameters on product properties, optimize yield and formulation costs.

Fig. 7.58 Scheme of three-layer artificial neural network.

j359

j 7 Particle Formulation in Spray Fluidized Beds

360

Fig. 7.59 Influence of maltodextrin type 1 and starch type 1 on yield and particle size for orange oil.

Selected results of the parametric study are illustrated in Figs. 7.59–7.61. The surface response plots refer to the influence that changes in the matrix composition (mixtures of maltodextrin type 1 and starch type 1 with different mass fractions of these components) have on the yield and the average particle diameter for the three investigated oils. All other input parameters were set to their median values according to Tab. 7.9. The diagrams show that different encapsulation efficiencies, and also different particle sizes, are obtained for each of the three aromatic oils, which is expected because of differences in the chemical structure and properties of the oils. For instance, the yield of mint (Fig. 7.60) was significantly higher than those of orange (Fig. 7.59) and pergamot (Fig. 7.61). The yield of mint also shows the highest sensitivity to variations in the composition of matrix material in the emulsion. Concerning the influence of other parameters, the yield was found to increase with increasing spray rate, decreasing product temperature and decreasing residence time. Bigger particles were produced at high spray rates and low temperatures, with a

Fig. 7.60 Influence of maltodextrin type 1 and starch type 1 on yield and particle size for mint oil.

7.6 Neural Networks, Encapsulation

Fig. 7.61 Influence of maltodextrin type 1 and starch type 1 on yield and particle size for pergamot oil.

less clear trend with respect to residence time. All this depends on the choice of active ingredient and the recipe in a complex way that is difficult to express verbally, pictorially or by equations, but can be captured by the artificial neural network. Consequently, the artificial neural network can be used to optimize the encapsulation process. This capability is demonstrated in Figs. 7.62 and 7.63. Figure 7.62 shows the dependence of retention yield for orange oil on spray rate and residence time. This diagram was calculated for a simplified composition of emulsion containing only maltodextrin type 1, starch type 1, orange oil and water. The mass fractions of water and oil were kept constant at 50% and 20%, respectively. For the case shown in

0.60

Yield

spray rate –1 g min 16 26 33 39 49

0.50

0.40 0

10

20

30

40

residence time [min] Fig. 7.62 Low yields of encapsulated orange oil for parameter set no. 1.

50

j361

j 7 Particle Formulation in Spray Fluidized Beds

362

0.70

0.60

Yield

spray
rate –1 g min 16 26 33 39 49

0.50

0.40 0

10

20

30

40

50

residence time [min] Fig. 7.63 Significantly improved yields of encapsulated orange oil for parameter set no. 2.

Fig. 7.62 the amount of maltodextrin was higher than the amount of starch (parameter set no. 1). The retention of oil according to Fig. 7.62 is so low that an industrial application would be too expensive due to high raw material losses. However, higher yields of oil can be attained by improving the formulation, guided by the artificial neural network. Specifically, the amount of maltodextrin can be decreased and the amount of starch increased. This leads to the results illustrated in Fig. 7.63 (parameter set no. 2). The yield of active ingredient is now significantly higher than for the initial set of parameters. This is because matrices based mainly on starch have, after drying, a lower porosity than matrices based mainly on maltodextrin. The process depicted in Fig. 7.63 would be industrially viable, though for the price of a more expensive raw material. As both Figs. 7.62 and 7.63 show, low residence times are an advantage when volatile substances have to be encapsulated into a granular structure. Therefore, the ProCell apparatus used in the described investigation was a good choice. As already pointed out in Section 7.5.1, this apparatus allows one to spray on a relatively low bed mass, keeping the residence time of the solids short. Simultaneously, high shear forces in the spout support uniform film formation and minimize the tendency to agglomeration. The use of a conventional fluidized bed for the same application would mean a higher hold-up, longer residence time and, hence, more loss of active ingredient. In general, it should be borne in mind that artificial neural networks do not consider the physics of processes in any way; they simply reflect the experimental information used for their training. However, if properly applied, they can significantly support the design of processes and formulations.

7.7 Stochastic Discrete Modeling of Agglomeration

7.7 Stochastic Discrete Modeling of Agglomeration 7.7.1 General Principles

Stochastic, so-called Monte Carlo (MC), methods use probabilistic tools to simulate the evolution of a finite sample of the particle population during, for example, spray fluidized bed agglomeration. They refrain from a computation of the flow field, so that parameters related to this flow, such as the frequency of collisions between particles, must be treated as known input quantities. In the case of large differences in the flow field between different regions within the apparatus, different particle samples must be considered by combination of a number of interacting MC simulations that run in parallel. Hence, stochastic methods create less information than the DPM (compare with Section 7.5.4), but they are easier to implement and much faster, so that they can be applied over the entire duration of real processes. On the other hand, MC simulations are still discrete. This means that all micro-scale mechanisms with an influence on the overall process behavior– wetting of particles by droplets, the removal of such droplets from the surface of particles by drying or liquid penetration, particle coalescence or agglomerate breakage – can be captured and described on a physical basis. Distributed properties such as the particle size can be continuously monitored in this way without having to – more or less arbitrarily – choose coalescence or breakage kernels. Consequently, MC simulations provide much more profound insight into particle formulation processes than macroscopic population balance equations (PBE), for the price of more computational effort. In this sense, MC is placed somewhere between DPM and PBE. The ability of MC to provide stochastic numerical solutions of the PBE has been discussed by Ramkrishna (1981). Monte Carlo methods are classified into two main groups according to the treatment of the time step: time driven and event driven methods. In the time driven approach, first a time step is specified and then all the possible events within this time step are implemented (Haibo et al., 2005). In the event-driven approach, an event is designed to happen and then time runs by a previously designated amount (Shah et al., 1977). The MC methods can be further distinguished according to the kind of particle number regulation in methods with continuous or periodical regulation. Continuous regulation means that once a particle appears or disappears from the simulation box, another is randomly copied or erased to maintain the number of particles Np constant and equal to a value Np,0 specified at the beginning of the simulation. Under periodic regulation, once the total number of particles reaches 1 /2Np,0, the number of particles is doubled by taking an additional identical control volume. The choice of the method should be based on the mechanism that governs during the process. It has been found that the event-driven method with periodical particle regulation (constant volume Monte Carlo method, CVMC) is more suitable for coalescence dominated processes such as agglomeration (Zhao et al., 2007). The connection between CVMC and the real particulate system arises from the eventdriven nature of the method.

j363

j 7 Particle Formulation in Spray Fluidized Beds

364

In the following, an implementation of the CVMC method for spray fluidized bed agglomeration and respective results will be discussed based on work by Terrazas (Terrazas-Velarde et al., 2009, 2011); Terrazas-Velarde, 2010). In this work an “event” is defined as a collision among the particles within the fluidized bed. The number of events k is given at the beginning of the simulation and the time step is calculated on the basis of collision frequency.

7.7.2 Computational Method

As already pointed out, the frequency of collisions among particles in the fluidized bed, fcoll, must be known in order to apply the MC method. To this purpose the empirical equation      1
e 1
e 2 u0 fcoll ¼ Fcoll 1
1
emf 1
emf

ð7:48Þ

proposed by Buffiere and Moletta (2000), was used. This equation provides reasonable trends regarding the influence of fluidization velocity u0 and expanded bed porosity e, whereby emf is the porosity at minimal fluidization. For a given material, the collision frequency pre-factor Fcoll was fitted to the results of one spray fluidized bed agglomeration experiment conducted under mild thermal conditions, as will be discussed later. With known collision frequency, the length of the time step of the MC method – distance in time between one and the next event – can be calculated to be tstep ¼

1 fcoll

ð7:49Þ

so that the real time elapsed after k events sums to treal ¼

k X

tstep

k¼1

ð7:50Þ

Continuous droplet addition in the model is accomplished by calculating the number of droplets per primary particle per second that are introduced to the real system c¼

  _ l rp;0 d0 3 M Mbed rl dd

ð7:51Þ

_ l is the mass flow rate of sprayed liquid, rl the liquid density, dd the droplet Here, M diameter and Mbed the mass of the bed; rp,0 and d0 are primary particle density and diameter, respectively. Notice that the droplet addition rate c is equal for both the real process and the simulation. From the value of c and the initial number of particles (primary particles) within the simulation box Np,0 the droplet addition

7.7 Stochastic Discrete Modeling of Agglomeration

time tad, which is the time necessary to add one single droplet to the simulation box, is calculated to be tad ¼

1 2N cNp;0

ð7:52Þ

where N is the number of doublings that the simulation system undergoes. The particles to be wetted are chosen randomly among the population. Once a droplet is ready to deposit on a particle it is assumed to take the shape of a spherical cap with volume Vcap, base radius a, and height h that fulfills the relationships  1=3 3Vcap sin3 q a¼ p 2
3 cos q þ cos3 q h¼a

1
cos q sin q

ð7:53Þ

ð7:54Þ

where q is the solid–liquid equilibrium contact angle. Immediately after droplet capture and particle wetting, the deposited droplet is available to potentially form a liquid bridge and produce coalescence with another particle or agglomerate. However, the deposited droplet is going to age until a successful coalescence has taken place due to two mechanisms, namely drying and penetration (“imbibition”). Drying is necessary for solid particle formation and takes place anyway, on compact or on porous particles, whereas imbibition by capillary suction takes place only on a porous substrate. Consequently, droplets age only by drying on compact particles, and by a combination of drying and imbibition on a porous substrate. The drying mechanism reduces the height and increases the viscosity of deposited droplets that consist of a volatile solvent (usually water) and a binder. As the agglomeration process proceeds, existing droplets which have not experienced a successful coalescence are becoming older until they solidify. The kinetics of the height reduction of droplets deposited on compact particles by drying can be estimated easily by neglecting the influence of the binder on the drying rate, and assuming that the contact angle remains constant during the decrease in height and radius (Erbil et al., 2002). In this way, the following equation is obtained (TerrazasVelarde et al., 2011): hdry

   ~w 2 rg M b Pv 1 1
1

~yv t ¼ h0
~ g 1
cos q P 3 rw M 1
cos q 3

ð7:55Þ

Here, Pv is the saturation pressure of the solvent (water) at the surface of the particles, ~yv is the respective molar fraction in the gas phase, and b is the gas-side mass transfer coefficient. In analogous manner (Terrazas-Velarde, 2010), the penetration of solvent into the pores of the primary particle can be described by the relationship

j365

j 7 Particle Formulation in Spray Fluidized Beds

366

himb

   sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 sin q 2 1 1
1 c cos q rpore 1=2 ¼ h0
ep
t 8 ml 3 1
cos q 1
cos q 3

ð7:56Þ

Equation 7.55 can be applied with reasonable estimates of gas-side state variables in a fluidized bed and combined with Eq. 7.56 to capture the simultaneous influence of drying and imbibition. Notice that a drying time and a wetting time can be derived from Eqs. 7.55 and 7.56, respectively. These time constants have a similar meaning to that of the time constants used in Section 7.4.2, but they are more specific, because they do not refer to the whole particle, but to one deposited droplet. Another difference to Section 7.4.2 is that drying and imbibition are not antagonistic in the case of agglomeration: both contribute to a decrease in the availability of surface liquid for coalescence, so that both tend to decrease the experimentally observable agglomeration rate. Successful coalescence presupposes the collision of two particles at, at least, one wet – or still wet – spot. Additionally, the liquid must be able to dissipate, by viscous forces, the kinetic energy of the collision – otherwise the particles will not stick together, but rebound. This second condition can be quantified by application of the so-called Stokes criterion, which was derived by Ennis et al. (1991). According to this criterion a collision is assumed to be successful in the sense of agglomeration when the inequality Stcoal < Stcoal is fulfilled, with Stcoal ¼

Stcoal

2 M ucoll 3p ml d2

    1 h ln ¼ 1þ e ha

ð7:57Þ

ð7:58Þ

In the framework of the Stokes criterion the influence of several process parameters can be quantified. A reduction in the coalescence Stokes number Stcoal and, thus, a higher probability of agglomeration, is achieved by smaller collision velocities ucoll, which depend directly on the fluidization gas velocity. Higher liquid viscosities ml – either by more binder in the sprayed liquid or by droplet aging due to drying – have a similar effect. An increase in the thickness h of the liquid layer that can be attained by using larger droplets will shift the agglomeration limit towards bigger particles (larger value of the critical Stokes number Stcoal ). In contrast, smaller droplets inhibit agglomeration and favor layering granulation or coating. Smaller droplets can be realized by the use of appropriate nozzles and spraying systems, but they can also result from strong drying and fast imbibition in the process. Additional quantities with an influence on the Stokes criterion are the height of asperities on the surface of particles, ha, and the restitution coefficient, e. Averages considering both colliding particles are used for the mass, M, and the diameter, d, in Eq. 7.57. The stochastic approach treats the size enlargement process as an increase in the number of primary particles per agglomerate. In order to be able to compare the model results with a real process, this number is correlated with the agglomerate diameter by assuming a constant agglomerate porosity of 60%. The structure of the primary particles within the agglomerate is based on a concept of fractional

7.7 Stochastic Discrete Modeling of Agglomeration

surface coverage and maximum coordination number Kmax ¼ 6. It should be noted, that successful coalescence can block some droplets according to this concept, making them inaccessible to further collision partners. Such droplets cannot produce a liquid bridge, so that they will age to ineffective solidification by drying. In this way, the steric hindrance of some droplets that has already been mentioned in Section 7.3.3 is taken into account in the model. Further details are given in TerrazasVelarde (2010). 7.7.3 Results

To evaluate the performance of the model, batch experiments were carried out in a lab-scale fluidized bed equipped with a two-fluid nozzle in top-spray configuration. The liquid solution was HPMC in water. Compact glass beads and porous alumina (c-Al2O3) particles were used as the solids, with an initial average diameter d0 (primary particle diameter) equal to 0.4 and 0.36 mm, respectively. A single selected experiment with glass and a single experiment with alumina, both conducted at mild thermal conditions, were used to adjust the collision frequency pre-factor Fcoll. In this way, the value of Fcoll ¼ 10 m
1 (fcoll ¼ 1.6 s
1) was obtained for glass, and Fcoll ¼ 45 m
1 (fcoll ¼ 4.1 s
1) for alumina. Once this parameter was fitted, it was used without further change for comparison with all other data gained with the respective material. A deeper discussion on the effect of the number of collisions on model response, more details about equipment and material properties, and a full documentation of the experimental results can be found in Terrazas-Velarde (2010). Here, just a few comparisons with measured data are presented to show that the model can reliably describe the influence of process parameters. 7.7.3.1 Effect of Liquid Flow Rate and Viscosity _ l on the agglomeration kinetics Figure 7.64 shows the effect of liquid addition rate, M of glass particles, expressed by the ratio of current to initial particle diameter during the process. As can be seen, agglomeration is faster at high spraying rates. A higher liquid flow rate means more liquid droplets introduced to the system, wetter particle surfaces, more wet collisions and, thus, an increase in agglomeration rate. The number of droplets introduced to the system is of essential importance for the rate of the process. Regarding the properties of the sprayed solution, it is observed in Fig. 7.65 that faster agglomeration is obtained as the binder mass fraction and, therefore, the viscosity of the liquid increases. This is due to the higher ability of deposited liquid layers to dissipate the kinetic energy of collisions, so that the particles involved tend to stick together rather than to rebound. 7.7.3.2 Thermal Effects An important advantage of the micro-level approach is that it enables one to analyze mechanisms that cannot be isolated experimentally. As an example, Fig. 7.66 shows

j367

j 7 Particle Formulation in Spray Fluidized Beds

368

Fig. 7.64 Effect of binder addition rate on the growth of glass particles by agglomeration; (Run numbers refer in this and in the following figures to the documentation of data according to Terrazas-Velarde (2010)).

Fig. 7.65 Effect of binder mass fraction in the sprayed droplets (droplet viscosity) on the agglomeration rate of glass particles.

7.7 Stochastic Discrete Modeling of Agglomeration

model results computed at different binder mass fractions with and without implementation of the drying mechanism. Two completely opposing tendencies can be recognized in this plot: Drying is predicted to accelerate the agglomeration in the case of low-viscosity binders, but to inhibit the process significantly when the liquid viscosity is relatively high. When the drying mechanism is not included (denoted by “ND” on the plot) and the liquid viscosity is low (xb ¼ 0.02), the model predicts a total absence of agglomeration (the particle diameter remains constant at its initial value). This is because the liquid deposited on the particles is not able to absorb the kinetic energy of collisions. In reality, granulation by layering would take place under such conditions – with a growth rate which would be much smaller than in the case of agglomeration. However, when the droplets are allowed to dry, the agglomeration process is switched on, under otherwise exactly the same conditions. This is because the viscosity of the liquid is increased by drying to an extent that is sufficient for dissipation of the collision energy and, thus, for coalescence. At high binder mass fractions in the spray (xb ¼ 0.08 and 0.10), the initial viscosity of the liquid is already large enough for the dissipation of collision energy, so that a further increase in viscosity during drying does not really matter. However, drying also means a reduction in the droplet height and availability – which matters, overriding the influence of the viscosity increase, and reducing the agglomeration rate significantly in these cases. The results of Fig. 7.66 point out the complex and intimate coupling between drying and agglomeration in spray fluidized beds. The fact that the droplets dry during the process does not necessarily mean less agglomeration, as drying can also be favorable in the case of diluted binder solutions. In practice, the intensity of drying can be manipulated mainly by changing the gas inlet temperature. The effect of gas inlet temperature on the agglomeration behavior is illustrated in Fig. 7.67. The plot shows a decrease in agglomeration rate with increasing gas inlet temperature, which means that the already discussed effect of reduced droplet height and availability in the system prevails for the depicted data. In total, the application of the MC model enabled one to explain and quantitatively describe the influence of thermal effects on wet agglomeration (Terrazas-Velarde et al., 2011) in a never before achieved way, which is not possible by application of conventional PBE approaches. Drying is the key to this explanation – in combination with the access to micro-scale physical interactions that the model provides. 7.7.3.3 Effect of Particle Porosity The features of the model also allow the simulation of solids which only differ in their porous or non-porous character by accounting, or not, for imbibition (Eq. 7.56). Unfortunately, porous and compact solids with otherwise exactly the same properties (e.g., particle density) do not exist in reality. Therefore, the experimental investigation has stayed with glass as the compact material and alumina as the porous material. However, two fictitious substances, namely porous glass and compact alumina, both with otherwise the same properties as their real counterparts, were added to model evaluation.

j369

j 7 Particle Formulation in Spray Fluidized Beds

370

Fig. 7.66 Drying of deposited droplets can trigger agglomeration at low binder mass fractions, but decreases the agglomeration rate when enough binder is contained in the sprayed solution; (simulations conducted without implementation of the drying mechanism are denoted by “ND”).

Fig. 7.67 Influence of gas inlet temperature on the agglomeration of glass particles.

7.7 Stochastic Discrete Modeling of Agglomeration

Fig. 7.68 Effect of imbibition at different binder mass fractions; comparison between simulations and experimentally obtained agglomeration rates for real and fictitious materials (NP: non-porous, PO: porous).

The results are summarized in Fig. 7.68. The ordinate of this plot shows average agglomeration rates R, made dimensionless by reference to the maximal agglomeration rate measured for the considered material. The first data column of the plot refers to experiments conducted at different binder mass fractions with glass. The highest agglomeration rate was measured at xb ¼ 0.10, so that the respective value of R/Rmax,exp is, by definition, unity. On decreasing the binder mass fraction the agglomeration rate decreases, for the previously discussed reasons, so that the two other points for real glass lie lower. The second column refers to simulations for real glass particles. Comparison with the first column shows that the model does not predict the experimental data perfectly, but reflects the right trend and performs reasonably well. The third column makes the glass particles porous in the simulation. Here we see a dramatic decrease in agglomeration rates, because droplets do not get lost only by drying, but also by imbibition in the porous substrate. The fourth column summarizes the results of experiments with c-Al2O3. Concerning the influence of binder mass fraction, these experiments show the same dependence as the experiments with glass. The alumina particles were rendered compact in the simulations of the fifth column. This manipulation increases the agglomeration rates to approximately twice the measured values, because droplets are no longer lost by capillary suction. Finally, comparison of the sixth with the fourth column shows the relatively good agreement between model and experiment for real alumina. The conclusion of the study is that the imbibition mechanism may cause high droplet losses that reduce the agglomeration rate significantly, so that this

j371

j 7 Particle Formulation in Spray Fluidized Beds

372

mechanism needs to be accounted for – in combination with drying – in the model when dealing with porous substrates.

7.8 Summary and Outlook

Spray fluidized beds offer unique opportunities to formulate particles with specific properties by agglomeration, granulation or coating. Such opportunities stem from the huge variety of changes in material properties, process parameters and apparatus design with an influence on product structure and quality. The same variety is a challenge, because it makes conventional engineering approaches (for instance the use of correlations of dimensionless numbers) virtually impossible. Several scales that range from the particle system (fluidized bed, apparatus) over the single particle, the porous interior of the particle to the molecular level (e.g., around particle contacts) are present and potentially significant. This requires the use of various computational methods – some of them continuous, such as population balance equations, and some discrete, such as discrete particle modeling and Monte Carlo. None of these methods is a panacea, so that the art of engineering is not to choose the best one, but to combine many of them in an educated way – corresponding to their potentials and limitations. Such combinations and scale transitions have already a remarkable record of success, demonstrating, for example, the strong interconnection that exists between drying and product formulation in the considered class of processes. It is expected that the development described in this chapter will continue in the years to come, towards better products and more efficient processes.

Additional Notation Used in Chapter 7

a a a E e F F f G h K k N_ n P

maximal size of defects separation distance base radius of deposited droplet modulus of elasticity restitution coefficient force frequency pre-factor frequency growth rate height coordination number stiffness particle flow density number density probability

m m m N m
2 – N m
1 s
1 m s
1 m
N m
1 m
1 s
1 m
1


7.8 Summary and Outlook

P R s s Tg Wm w x

production density agglomeration rate meniscus radius of liquid bridge displacement glass transition temperature mass-related energy consumption water mass fraction (wet-based water content) diameter of contact area, liquid bridge

m
1 s
1 m s
1 m m K,  C J kg
1
m

Greek Letters

a b b b c c c_ q l u s t

volume fraction of spraying zone mass transfer coefficient momentum transfer coefficient angle covered by liquid bridge surface tension droplet addition rate shear rate contact angle superposition weight factors Poisson’s ratio normal stress residence time

Subscripts and Superscripts

a ad agg B b br cap coal coll cyl d diss dry el F gap kin

asperity droplet addition agglomerate breakage point binder bridge spherical cap (deposited droplet) coalescence collision cylindrical droplet dissolved drying elastic yield point gap kinetic


m s
1 kg m
3 s
1 rad N m
1 s
1 s
1 rad

Pa


j373

j 7 Particle Formulation in Spray Fluidized Beds

374

imb N pore pl pp R r real step sus T t vdW 0 0

imbibition normal pore plastic particle–particle rebound reduced real time time step solids in suspension tangential tensile van der Waals initial superficial

Abbreviations

CFD CSTR CVMC DE DEM DPM FB HPMC MC PBE PSD RH SB

computational fluid dynamics continuous stirred tank reactor contact volume Monto Carlo dextrose equivalent discrete element method discrete particle model fluidized bed hydroxypropyl-methylcellulose Monte Carlo population balance equation particle size distribution relative humidity spouted bed

References Aguilera, J., Levi, G., Karel, M., 1993. Effect of water content on the glass transition and caking of fish protein hydrolysates. Biotechnol. Prog. 9: 651–654. Antonyuk, S., Tomas, J., 2008. Deformation and breakage behaviour of dry cylindrical agglomerates. Proceedings of 4th International Symposium on Reliable Flow of Particulate Solids, Tromsø, Norway, 507–512.

Antonyuk, S., Heinrich, S., Tomas, J., Deen, N. G., van Buijtenen, M. S., Kuipers, J. A. M., 2010. Energy absorption during compression and impact of dry elastic-plastic spherical granules. Granular Matter 12(1): 15–47. Antonyuk, S., Palis, S., Heinrich, S., 2011. Breakage behavior of agglomerates and crystals by static loading and impact. Powder Technol. 206: 88–98.

References Antonyuk, S., Khanal, M., Tomas, J., Heinrich, S., M€ orl, L., 2006. Impact breakage of spherical granules: Experimental study and DEM simulation. Chem. Eng. Process. 45: 838–856. Avar, B. B., Hudyma, N., Karakouzian, M., 2003. Porosity dependence of the elastic modulus of lithophysae-rich tuff: Numerical and experimental investigations. Int. J. Rock Mech. Min. 40: 919–928. B€armann, F., Biegler-K€onig, F., 1992. On a class of efficient learning algorithms for neural networks. Neural Networks 5: 139–144. Baker, C. G. J., Lababidi, H. M. S., 2010. An improved model of plug-flow fluidized bed dryers with an emphasis on energy conservation. Drying Technol. 28: 730–741. Beetstra, R., van der Hoef, M. A., Kuipers, J. A. M., 2007. Drag force from lattice Boltzmann simulations of intermediate Reynolds number flow past mono- and bidisperse arrays of spheres. AIChE J. 53(2): 489–501. Bika, D., Tardos, G. I., Panmai, S., Farber, L., Michaels, J., 2005. Strength and morphology of solid bridges in dry granules of pharmaceutical powders. Powder Technol. 150: 104–116. Boccaccini, A. R., 1994. Zur Abh€angigkeit der mechanischen Eigenschaften zweiphasiger und por€oser Werkstoffe von der Gef€ uge-bzw. Porosit€ atsstruktur. Diss., RWTH Aachen, Germany. Braspenning, P. J., Thuijsman, F., Weijters, A. J. M. P. (eds), 1995. Artificial neural networks., Springer, Berlin, Germany. Briscoe, B. J., Luckham, P. F., Maung, R., Aydin, I., 1998. Nano and micro-mechanics of agglomerates. Proceedings of 3rd World Congress on Particle Technology, Brighton, UK. Buffiere, P., Moletta, R., 2000. Collision frequency and collisional particle pressure in three-phase fluidized beds. Chem. Eng. Sci. 55: 5555–5563. Coelho, D., Thovert, J.-F., Adler, P. M., 1997. Geometrical and transport properties of random packings of spheres and spherical particles. Phys. Rev. E 55: 1959–1977. Deen, N. G., van Sint Annaland, M., van der Hoef, M. A., Kuipers, J. A. M., 2007. Review of discrete particle modelling of fluidized beds. Chem. Eng. Sci. 62: 28–44.

Desportes, S., Steinmetz, D., Hemati, M., Philippot, K., Chaudret, B., 2005. Production of supported asymmetric catalysts in a fluidised bed. Powder Technol. 157: 12–19. Downtown, G., Flores-Luna, J., King, C., 1982. Mechanism of stickiness in hygroscopic, amorphous powders. Ind. Eng. Chem. Fundam. 21: 447–451. Ennis, B. J., Tardos, G., Pfeffer, R., 1991. A microlevel-based characterization of granulation phenomena. Powder Technol. 65: 257–272. Erbil, H. Y., McHale, G., Newton, M. I., 2002. Drop evaporation on solid surfaces: Constant contact angle mode. Langmuir 18: 2636–2641. Ferry, J., 1980. Visco-elastic properties of polymers. 3rd edn, Wiley & Sons, New York, USA. Fisher, R. A., 1926. On the capillary forces in an ideal soil. J. Agric. Sci. 16: 492–505. Frenkel, J., 1945. Viscous flow of crystalline bodies under action of surface tension. J. Phys. (USSR) 9(5): 385–391. Fyhr, C., Kemp, I. C., Wimmerstedt, R., 1999. Mathematical modeling of fluidized bed dryers with horizontal dispersion. Chem. Eng. Process. 38: 89–94. Goldschmidt, M. J. V., Kuipers, J. A. M., 2003. Discrete element modelling of fluidised bed spray granulation. Powder Technol. 138: 39–45. Gordon, M., Taylor, J. S., 1952. Ideal copolymers and the second order transitions of synthetic rubbers, Part 1: Non-crystalline copolymers. J. Appl. Chem. 2: 493–500. Haibo, Z., Chuguang, Z., Minghou, X., 2005. Multi-Monte Carlo approach for general dynamic equation considering simultaneous particle coagulation and breakage. Powder Technol. 154: 164–178. Hamaker, H., 1937. The London-van der Waals attraction between spherical particles. Physica 4(10): 1058–1072. Hede, P. D., Bach, P., Jensen, A. D., 2009. Batch top-spray fluid bed coating: Scale-up insight using dynamic heat- and masstransfer modeling. Chem. Eng. Sci. 64: 1293–1317. Hemati, M., Cherif, R., Saleh, K., Pont, V., 2003. Fluidized bed coating and granulation: Influence of process-related variables and

j375

j 7 Particle Formulation in Spray Fluidized Beds

376

physicochemical properties on the growth kinetics. Powder Technol. 130: 18–34. Hemati, M., Steinmetz, D., Chaudret, B., Philippot, K., 2001. Fabrication de catalyseur d’un nouveau type par impregnation en lit fluidise. Brevet EPI-PCT 2001, FR 02, 01795. € Hertz, H., 1882. Uber die Ber€ uhrung fester elastischer K€orper. J. f€ ur Reine und Angew. Mathematik 92: 156–171. Hill, R. J., Koch, D. L., Ladd, J. C., 2001. Moderate-Reynolds-numbers flows in ordered and random arrays of spheres. J. Fluid Mechanics 448: 243–278. Hoffmann, T., Hailu Bedane, A., Peglow, M., Tsotsas, E., Jacob, M., 2011. Particle-gas mass transfer in a spouted bed with adjustable air inlet. Drying Technol. 29: 257–265. Hoffmann, T., Peglow, M., Tsotsas, E., 2010. Dynamic behaviour of particle-size distribution in fluidized bed granulation. Proceedings of the 6th World Congress on Particle Technology, N€ urnberg. Hoomans, B. P. B., Kuipers, J. A. M., Briels, W. J., van Swaaij, W. P. M., 1996. Discrete particle simulation of bubble and slug formation in a two-dimensional gasfluidised bed: A hard-sphere approach. Chem. Eng. Sci. 51(1): 99–118. Iveson, S. M., Page, N. W., 2001. Tensile bond strength development between liquid-bound pellets during compression. Powder Technol. 117: 113–122. Jacob, M., 2007. Granulation equipment, in Handbook of powder technology: Granulation (eds A. D. Salman, M. J. Hounslow, J. P. K. Seville). Elsevier, Amsterdam, Holland, Vol. 11, Chapter 9, 417–476. Jacob, M., 2009. ProCell technology: Modelling and application. Powder Technol. 189: 332–342. Kafui, D. K., Thornton, C., 2008. Fully-3D DEM simulation of fluidised bed spray granulation using an exploratory surfaceenergy based spray zone concept. Powder Technol. 184: 177–188. Kendall, K., McAlford, N., Birchall, J. D., 1987. Elasticity of particle assemblies as a measure of the surface energy of solids. Proc. R. Soc. London, Ser. A 412(1843): 269–283. Khanal, M., Schubert, W., Tomas, J., 2008. Compression and impact loading

experiments of high strength spherical composites. Int. J. Miner. Process. 86 (1–4): 104–113. Kuczynski, G. C., 1949. Self-diffusion in sintering of metallic particles. Trans. AIME 185: 169–178. Lifshitz, E. 1956. The theory of molecular attraction forces between solids. Sov. Phys. (JETP) 2(1): 73–83. Markovitz, H., 1975. Superposition in rheology. J. Polym. Sci., Polym. Symp. 50: 431–456. Mindlin, R. D., Deresiewicz, H., 1953. Elastic spheres in contact under varying oblique forces. Trans. ASME, Ser. E, J. Appl. Mechanics 20: 327–344. M€orl, L., Heinrich, S., Kr€ uger, G., Ihlow, M., Jordanova, E., 2002. Steuerbare Gasanstr€omeinrichtung f€ ur Strahlschichtapparate. German Patent No. 100 04 939.7-41. M€orl, L., Heinrich, S., Peglow, M., 2007. Fluidized bed spray granulation, in Handbook of powder technology: Granulation (eds A. D. Salman, M. J. Hounslow, J. P. K. Seville). Elsevier, Amsterdam, Holland, Vol. 11, pp. 21–188. Nilsson, L., Wimmerstedt, R., 1988. Residence time distribution and particle dispersion in a longitudinal-flow fluidized bed. Chem. Eng. Sci. 43: 1153–1160. Palzer, S., 2005. The effect of glass transition on the desired and undesired agglomeration of amorphous food powders. Chem. Eng. Sci. 60: 3959–3968. Palzer, S., 2009. Influence of material properties on the agglomeration of watersoluble amorphous particles. Powder Technol. 189: 318–326. Pepin, X., Simons, S. J. R., Blachon, S., Rossetti, D., Couarraze, G., 2001. Hardness of moist agglomerates in relation to interparticle friction, granule liquid content and nature. Powder Technol. 117: 123–138. Pierrat, P., Caram, H. S., 1997. Tensile strength of wet granular materials. Powder Technol. 91: 83–93. Rabinovich, Y., Esayanur, M., Moudgil, B., 2005. Capillary forces between two spheres with a fixed volume liquid bridge: Theory and experiment. Langmuir 21 (24): 10992–10997.

References Ramkrishna, D., 1981. Analysis of population balance, Part 4: The precise connection between Monte Carlo simulation and population balances. Chem. Eng. Sci. 36: 1203–1209. Roos, Y., Karel, M., 1991. Plasticising effect of water on thermal behaviour and crystallization of amorphous food models. J. Sci. 56(1): 38–43. Rumpf, H., 1958. Grundlagen und Methoden des Granulierens. Chem.-Ing.-Tech. 30 (3): 144–158. Rumpf, H., 1970. Zur Theorie der Zugfestigkeit von Agglomeraten bei Kraft€ ubertragung an Kontaktpunkten. Chem.-Ing.-Tech. 42(8): 538–540. Rumpf, H., Sommer, K., Steier, K., 1976. Mechanismen der Haftkraftverst€arkung bei der Partikelhaftung durch plastisches Verformen, Sintern und viskoelastisches Fliessen. Chem.-Ing.-Tech. 48(4): 300–307. Salman, A. D., Fua, J., Gorham, D. A., Hounslow, M. J., 2003. Impact breakage of fertiliser granules. Powder Technol. 130: 359–366. Samimi, A., Ghadiri, M., Boerefijn, R., Groot, A., Kohlus, R., 2003. Effect of structural characteristics on impact breakage of agglomerates. Powder Technol. 130: 428–435. Schatt, W., 1992. Sintervorg€ange: Grundlagen. VDI Verlag, D€ usseldorf, Germany. Schubert, H., 1975. Tensile strength of agglomerates. Powder Technol. 11: 107–119. Schubert, H., 1979. Grundlagen des Agglomerierens. Chem.-Ing.-Tech. 151(4): 266–277. Shah, B. H., Ramkrishna, D., Browanker, J. D., 1977. Simulation of particulate system using the concept of the interval of quiescence. AIChE J. 23: 897–904. Silva, D. O., Tamiozzo, L. M., Duarte, C. R., Murata, V. V., Barrozo, M. A. S., 2011. Modeling of seed coating in spouted bed. Drying Technol. 29(3) 286–294. Simon, M., Iveson, S. M., Litster, J. D., Hapgood, K., Ennis, B. J., 2001. Nucleation, growth and breakage phenomena in agitated wet granulation processes: A review. Powder Technol. 117: 3–39. Slade, L., Levine, H., 1991. Beyond water activity: Recent advances based on an

alternative approach to the assessment of food quality and safety. Crit. Rev. Food Sci. Nutr. 30(2–3): 115–360. Smith, W. O., Foote, P. D., Busang, P. F., 1929. Packing of homogeneous spheres. Phys. Rev. 34: 1271–1274. Sperling, L., 1986. Introduction to physical polymer science. Wiley & Sons, New York, USA, pp. 224–295. Stepanek, F., Rajniak, P., 2006. Droplet morphologies on particles with macroscopic surface roughness. Langmuir 22: 917–923. Stepanek, F., Rajniak, P., Mancinelli, C., Chern, R., Ramachandran, R., 2009. Distribution and accessibility of binder in wet granules. Powder Technol. 189: 376–384. Terrazas-Velarde, K., 2010. Monte Carlo simulation of fluidized bed spray agglomeration. Diss., Otto von Guericke University Magdeburg, Germany. Terrazas-Velarde, K., Peglow, M., Tsotsas, E., 2009. Stochastic simulation of agglomerate formation in fluidized bed spray drying: A micro-level approach. Chem. Eng. Sci. 64: 2631–2643. Terrazas-Velarde, K., Peglow, M., Tsotsas, E., 2011. Investigation of the kinetics of fluidized bed spray agglomeration based on stochastic methods. AIChE J., doi: 10.1002/ aic.12506 Tomas, J., 1983. Untersuchungen zum Fließverhalten von feuchten und leichtl€oslichen Sch€ uttg€ utern. Diss., Freiberger Forschungsheft A677, Germany. Tomas, J., Schreier, M., Gr€oger, T., Ehlers, S., 1999. Impact crushing of concrete for liberation and recycling. Powder Technol. 105: 39–51. Tsuji, Y., Tanaka, T., Ishida, T., 1992. Lagrangian numerical simulation of plug flow of cohesionless particles in a horizontal pipe. Powder Technol. 71: 239–250. Tsuji, Y., Kawaguchi, T., Tanaka, T., 1993. Discrete particle simulation of twodimensional fluidized bed. Powder Technol. 77(1): 79–87. Wagner, T., 1997. Kaltspr€ uhen ein- und mehrphasiger Fl€ ussigkeiten. Diss., ETH Z€ urich, Switzerland. Wallack, D., King, C., 1988. Sticking and agglomeration of hygroscopic, amorphous carbohydrates and food powders. Biotechnol Prog. 4: 31–35.

j377

j 7 Particle Formulation in Spray Fluidized Beds

378

Weisser, H., 1986. Influence of temperature on sorption isotherms, in Transport Phenomena (eds M. Maguer, P. Jelen). Elsevier, London, UK. Wen, Y. C., Yu, Y. H., 1966. Mechanics of fluidization. Chem. Eng. Progr. Symp. Series. 62: 100–111. Willett, C. D., Johnson, S. A., Adams, M. J., Seville, J. P. K., 2007. Pendular capillary bridges, in Handbook of powder technology: Granulation (eds A. D. Salman, M. J. Hounslow, J. P. K. Seville). Elsevier, Amsterdam, Holland, Vol. 11, pp. 1317–1351. Williams, M. L., Landel, R. F., Ferry, J. D., 1955. The temperature dependence of relaxation mechanisms in

amorphous polymers and other glassforming liquids. J. Am. Chem. Soc. 77 (14): 3701–3707. Wurster, D. E., 1959. Air suspension technique of coating drug particles. J. Am. Pharm. Assoc. 48: 451–454. Yang, R. Y., Yu, A. B., Choi, S. K., Coates, M. S., Chan, H. K., 2008. Agglomeration of fine particles subjected to centripetal compaction. Powder Technol. 184 (1): 122–129. Zhao, H., Maisels, A., Matsoukas, T., Zheng, C. G., 2007. Analysis of four Monte Carlo methods for the solution of population balances in dispersed systems. Powder Technol. 173: 38–50.

j379

Index a abrasion resistance, particle strength 281 active transponders, in primary-drying monitoring 98 additives, enzyme stabilization 275 ADH. see alcohol dehydrogenase adhesion – low molecular weight 239 – mechanisms 299–315 – powders 260–261 adhesion bridges 299 adhesion force, spray fluidized beds 296, 302–305 adsorption – nitrogen 166–168 – stress 270–271 aerogels – SAXS spectra 164 – supercritically dried 159 – vacuum drying 198 aged gels, shrinkage 200 agglomerates – adhesion mechanisms and mechanical strength 299–315 – breakage 314, 315–321 – material structure 300–301 – mechanical properties 312 – mechanical strength 299–315, 308–315 – preparation scheme 310 agglomeration – dextrose sirup 307 – discrete modeling 363–372 – glass particles 370 – particle formulation 296 – particles properties 298 – primary particle properties 321–324 – spray fluidized bed processes 297 – stochastic discrete modeling 363–372

– tensile strength 308–310 – triggering 370 aging – cracks 199–201 – effects on shrinkage 209 – RF gels 208–209 – shrinkage 200 agricultural products, textured by drying 3 alcohol dehydrogenase (ADH), activity retention 274–277 alcohol dehydrogenase (ADH) powder, outer surface morphology 248 alumina carrier particles 331 alumina gels – crack patterns 175 – diffusion models 213 alumina monoliths 201 amorphous glass state 261 – vibrational motions 261 amorphous particles – surface tension 305 – viscous forces in sinter bridges 304–308 – volume diffusion 305 amorphous water-soluble materials 302 anhydrous sugars, glass transition temperature 13 annealing, influence on ice morphology 67–69 anthocyans, changes in drying process 7 apparatus design, influence on product quality 332–338 aroma compounds, retention of 9–10 aromatic oils, spray fluidized bed encapsulation 358 artificial neural network 359 ascorbic acid, as a quality index in drying process 6 attrition, particles strength 281–282

Modern Drying Technology Volume 3: Product Quality and Formulation, First Edition. Edited by Evangelos Tsotsas and Arun S. Mujumdar. Ó 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.

j Index

380

b balances, for freeze-dryers 104–105 band dryer, food industry 2 barometric temperature measurement (BTM) – primary-drying monitoring 115 – shelf-temperature control 128–129 BaSO4-suspension 243 batch granulation – dispersive growth 344–349 – dispersive growth in 344–349 – growth 344–349 batch monitoring – endpoint detection of primary drying 106–113 – freeze-drying 106–125 – using sublimation-flux measurement 113–114 bed material, number density distribution 326 belt dryer, food industry 2 binder content, sprayed solution 325–326 biochemical reactions, induced by drying 5–9 bovine serum albumin (BSA) – primary-drying process 127 – spray-dried particle morphology 250 breakage behavior – agglomerates 315–321 – cylindrical agglomerates 314 – elastic-brittle 316–317 – elastic-plastic 317–318 – granules with layered structure 320–321 – plastic 318–320 breakage probability – binder contents 314 – c-Al2O3 agglomerates 315 – granulation time dependence 325 – retention effect 327 breakage ratios, comparison of visual inspection and image anlaysis 32 bridges – adhesion 299 – liquid 303–304 – sinter 304–308 brown rice, obtainment of 22 browning reaction, in drying process 8 BSA. see bovine serum albumin BTM. see barometric temperature measurement buckling pressure, suspensions 242 bulk modulus, drying methods 176

c calorimetric measurements, single-vial monitoring 101

capillary forces – adhesion 299 – liquid bridges between particles 303–304 – pore sizes 167 capillary pressure 175, 304 capsule wall materials 257 caramelization, in drying process 8 carbohydrate polymers, glass transition temperature 13 carbon aerogels, dry gels 161 carbon cryogels, quality preservation 192 cargo rice, obtainment of 22 carotenoids, changes in drying process 6–7 carrier materials, particle creation 247–251 carrier matrices collapse 268 carrier particles, loading with catalytic active components 329 CCD camera, fissure formation in rice 28–29 centrifugal rotary disk atomizer, spray drying 231 CFD. see computational fluid dynamics chamber pressure. see also sublimation chamber – calculated by CFD 140–142 – primary-drying control 93, 125–135 chamber temperature. see also temperature – vial batch monitoring 118 cherenkov detectors 161 chlorophylls, changes in drying process 6 chromatography 161 Clausius-Clapeyron equation, low molecular weight substances 236 CLSM. see confocal laser scanning microscopy CO2, low-temperature gel drying 187–189 coating 296–297 cold chamber optical microscopy, freezedrying 55–57 cold plasma ionization, vial batch monitoring 110–111 collapse temperature, freeze-drying 54–55 colloidal gel networks 157 color, as a quality index in drying process 6 complex dispersions 244–251 compression tests 310–311 computational fluid dynamics (CFD) – calculation of local moisture content 12 – design parameters 139–142 confocal laser scanning microscopy (CLSM) 234, 247 constant rate period (CRP) 236 consumer products, gained by drying 3–4

Index contact stiffness 312 continuous freeze-drying 142–143 continuously operated stirred tank reactors (CSTR) 339 control algorithms, primary freeze drying 125–135 controlled nucleation – and physical quality 70–73 – by ultrasound sonication 70–72 convective drying – advanced modeling 211–220 – contact angle 199 – diffusion model 211–217 – gels 174–182 – hydrogel 175 – quality preservation 198–210 – RF gels 206 convective hot air drying, and mechanical transformations 15 cooling rate. see freezing rate – influence on dried layer permeability 66 crack formation. see also fissured rice – convective gel drying 174–175, 180–182 – critical drying rate 214 – density change 183 – during drying 16 – initiation 181 – in rice 26–27 – shells 320 – supercritical drying 188 – surface and internal 25 crack-free monoliths 190 crack patterns 175 crack propagation 316 cracking, video acquisition 29 cracks. see also fissured rice – from aging 199–201 – surface and internal 25 critical drying rate, crack formation 214 cross-sectional structures, spray-dried powders 248 CRP. see constant rate period crust formation, during annealing process 69 cryogel flakes 191 cryogels 159 crystal nucleation. see nucleation crystalline substances, water-soluble 304 crystallization – in drying process 10 – low molecular weight substances 238 – during storage 16–17 CSTR. see continuously operated stirred tank reactors

d D-limonene – flavor solubility 259 – oxidation reaction 267 – release kinetics 264 – retention 258 Darcy’s law – differential shrinkage 177 – freeze-dried layer permeability 76 – PRT 60 deep bed dryer, food industry 2 dehydration stress 271–272 depressurization, supercritical drying 186 dewatering. see also drying dextrose sirup, fluidized bed agglomeration 307 diametral compression test, rice grains 37 dielectric measurements, single-vial monitoring 100–101 differential scanning calorimetry (DSC) – freeze-drying 55 – gel drying 164 differential shrinkage – diffusion equation 178 – and stress 177–180 dihedral dryer, food industry 2 discrete particle modeling (DPM) – agglomeration 363–372 – principles 350–351 – simulation parameters 351–352 – Wurster coater 349–357 distributor plates, apparatus design 337 DPE. see dynamic parameters estimation DPM. see discrete particle modeling; drying process monitoring dried layer. see freeze-dried layer dried particles – lipids oxidation 279–280 – porosity 280–281 dried powder – flavor release 262–267 – stickiness 260–261 droplet drying 273–274 droplet shrinkage 232 droplet size, feed emulsion 258 drugs. see pharmaceuticals drum dryer, food industry 2 dry coating process 298 dry gels – applications 160–162 – catalysis 161 – characterization 166–172, 166–174 – conductivity 159 – density 159

j381

j Index

382

– dielectric constant 160 – elastic behavior 160 – hydrophobicity 160 – insulation 160 – optical coatings 161 – optical transparency 159 – other methods 171–172 – properties 158–160 – refractive index 160 – sound insulation 161 – sound speed 160 – surface area 159 – thermal conductivity 159 – thermal insulation 160 – transparency 159 – water treatment 161 drying. see also dewatering; freeze-drying; gel drying – advanced modeling 211–220 – convective 174–182, 198–210 – encapsulation and microencapsulation of enzymes and oil by 269–280 – gel characterization 172–174 – gels 155–230 – microencapsulation 269–270 – microwave 210–211 – oil emulsions 278–279 – particle creation 251–253 – preserving quality 189–211 – process variables effect on the stabilization of enzymes 275–278 – protein encapsulation theory 272–273 – protein solutions 273 – quality loss 174–189 – retention of emulsified hydrophobic flavors 257–260 – single suspended droplet 273–274 – stress on proteins 270–272 – subcritical 189–190 – vacuum 197–198 drying chamber. see also sublimation chamber – fluid dynamics in 139–142 – water vapor pressure 62 drying equipment, food industry 2 drying modes – combined 17–18 – in foods 14 drying process – as a controlled texturing operation 3 – impact on mechanical properties and crack formation in rice 21–45 – quality changes in food materials 1–18 drying process monitoring (DPM), single vials 101

drying process severity, and food quality 5 Drying3000 simulator 39 DSC. see differential scanning calorimetry dynamic parameters estimation (DPE) algorithm – primary-drying control 129, 132 – vial batch monitoring 115–116

e easy-to-use products, gained by drying 3 ebullition, as drying mode 14 effective contact stiffness 312 elastic-brittle breakage behavior 316–317 elastic-plastic breakage behavior 317–318 elastoplastic material, stress-strain relationship 36 empirical curve fitting, modeling of rice quality 39 emulsified hydrophobic flavors – retention 257–260 – spray drying 256–257 emulsions – complex dispersions 244–251 – drop size 258 – microencapsulated flavor powders 245–247 – spray drying 278–279 encapsulated flavor – glass temperature influence 261–262 – oxidation 267 – release and oxidation during storage 261–269 encapsulated flavor droplets, CLSM pictures 246–247 encapsulated lipids, oxidation 279–280 encapsulation – enzymes and oil 269–280 – neural networks 357–363 endpoint detection, vial batch monitoring 106–113, 121–122 enzymatic activity, and water activity 8 enzyme activity retention 275 enzyme stabilization – effect of process variables 275–278 – effects of formulation composition 274–275 enzymes – encapsulation and microencapsulation 269–280 – particle creation 247–251 – spray drying microencapsulation 269–270 – thermal stress 271–272 ethanol retention 256 ethyl-n-butyrate powder

Index – flavor powders 246 – flavor release 262 – flavor solubility 259 explosion puffing, combined with drying 18 extended Kalman filter, single-vial monitoring 102

f failure strength, rice grains 36–39 feed liquid, spray drying 231 feedback controlling, primary-drying 134–135 filling height, freeze-drying 66–67 film thinning effect, low molecular weight substances 238 fine glass filament suspension 272 finite element modeling, modeling of rice quality 39 finite strain tensor 218 fish oil, oxidation kinetics 264 fissure formation – characterization by image analysis techniques 28–33 – count algorithm 31 – segmentation method for characterization 30–31 fissure ratios, comparison of visual inspection and image anlaysis 33 fissured rice – definition 23–24 – and relative humidity 24–28 flash spray drying, suspensions 242 flavor compounds, retention of 9–10 flavor droplets, encapsulated 247 flavor encapsulation, theory and mechanism 255 flavor powders, microencapsulated 245–247 flavor release – analysis by PTR-MS 266 – humidities and temperatures 264, 266 – mathematical modeling 262 – and oxidation 261–269 flavor retention, spray-dried food products 253–269 flavor solubility 259 flavor solution, spray-drying scheme 257 flavors – emulsified 257–260 – glass temperature influence 261–262 – microencapsulation 254–256 – oxidation 267 – spray drying 256–257 flaxseed, water activity effect 279 florescein sodium salt, protein particles 248 fluid dynamics, as quality parameter 139–142

fluid temperature, primary-drying control 130 fluidized bed agglomeration, dextrose sirup 307 fluidized bed coating, process conditions 346 fluidized bed dryer, food industry 2 fluidized beds – catalyst impregnation 329–332 – particle formulation 253, 295–378 food industry, drying equipment 2 food materials – biochemical reactions induced by drying 5–9 – drying-process-influenced quality changes in 1–18 – mechanical transformations induced by drying 14–16 – physical transformations during drying 9–14 – storage and rehydration of 16–17 food particle bridges, capillary forces 303–304 food particles – relaxation 302 – viscoelastic deformation 302 – viscous forces in sinter bridges 304–308 food products – flavor retention 253–269 – spray-dried 233 food quality also quality – and drying process severity 5 – gained by drying 4 – and nutritional and sensory properties 4 formulation. see also liquid formulation; particle formulation – complex dispersions 244–251 – enzyme stabilization 274–275 fractal drying front, crack formation 182 fracture morphology, dry gels 160 fracture surface – c-Al2O3 agglomerates 316–317 – sodium benzoate granules 321 – zeolite agglomerate 319 freeze-dried cake morphology – and physical quality 74–78 – and water vapor mass transfer resistance 74–76 freeze-dried cake permeabilty – PRT 59–61 – theoretical 77 freeze-dried layer permeability – experimental 77 – influence of cooling rate 66 – and water vapor mass transfer resistance 76–78

j383

j Index

384

freeze-dried matrix, moisture gradients in 52 freeze-dryer, food industry 2 freeze-dryer balances 104–105 freeze-drying. see also drying; primary-drying control; primary-drying monitoring – chamber pressure 93 – cold chamber optical microscopy 55–57 – collapse temperature 54–55 – continuous 142–143 – control of freezing step 94–96 – control of primary drying 125–135 – DSC 55 – estimation of mean product temperature 61–63 – gels 182–185 – and glass transition 11 – heat flux heterogeneity 57–59 – ice structure and morphology 55–57 – in-line product quality control 91–144 – key quality factors 52–63 – and mechanical transformations 14 – melting curves 54–55 – monitoring and control of secondary drying 135–138 – MTM 59 – of pharmaceuticals 51–86 – PRA 59–61 – principal basic phenomena 51–52 – product quality during drying and storage 83–85 – product-temperature maintanance 91 – quality parameters 139–142 – quality preservation 190 – residual water content 91–92 – RF and carbon cryogels 192–193 – shelf temperature 93 – state diagram 54–55 – vitreous transition 54–55 freeze-drying microscopy 55 freeze-drying parameters, influence on physical quality factors 63–82 freeze-drying process, different steps 52, 91 freeze spray drying, particle creation 251–253 freezing process, and tensile stress 184 freezing protocol, influence on ice morphology 63–69 freezing rate, influence on ice morphology 55–56, 64–66 freezing step. see also nucleation – control of 94–96 full milk particle, agglomerated 253 functional oils 278 functionalities, of food materials 1

g c-Al2O3 agglomerates – breakage probability 315 – fracture surface 316–317 c-Al2O3 particles – elastic-brittle breakage behavior 316 – used to produce agglomerates 310 gap distance, Wurster coater 355–357 gas distributor, apparatus design 333 gas recycling, apparatus design 332 gas temperature, in chamber. see chamber temperature; temperature Gaussian blobs 322 gel applications, quality aspects 156–162 gel drying 155–230 – cracking 180–182 – differential shrinkage 176–180 – freezing 182–185 – low-temperature process 187–189 – methods 174–189 – phase diagram 174 – supercritical 159, 185–189 – X-ray tomography 172 gel networks 157 gel structure – changes 155–230 – characterization 162 – destruction 183 – during drying 155 – resorcinol-formaldehyde gels 158 gel synthesis, optimization 194 gelatinization 245 gelation – quality aspects 156 – ultrasonic irradiation 195 gelinization, starch 245 gels – aging 208–209 – applications 160–162 – characterization during drying 172–174 – characterization of dry 166–172 – characterization of wet 162–166 – crack patterns 175 – diffusion models 213 – ice templating 195–197 – polymer crosslinking 201 – preparation 156–157 – properties 158–160 – quality aspects 156–162 – resorcinol-formaldehyde 157–158, 204–208 – RF aging 208–209 – RF convective drying 206 – RF freeze drying 191 – RF linear shrinkage 209

Index – RF quality preservation 204–208 – RF SAXS spectra 164 – RF synthesis 157–158 – shrinkage 200 – shrinkage prevention and cracks by aging 199–201 – shrinkage reversion 201–204 – silica 156–157 – structural characterization 162 – technical 158 – transmission electron microscopy 171 – wet 156–158 glass encapsulation 254 glass particles – agglomeration 370 – growth 368 glass transition curve, in drying process 10–11 glass transition temperature – of anhydrous sugars and carbohydrate polymers 13 – low molecular weight substances 238 – relaxation process correlation 267–269 – spray-dried powder stickiness 260 – and storage stability of encapsulated flavor 261–262 glassy particles – lactose 238 – surface of 261 Gordon-Taylor constant 300 grains, rice. see rice grains granulated particles, mechanical strength 324–329 granulated products, breakage 315–321 granulation – dispersive growth 344–349 – particle formulation 296 – particles properties 298 – spray fluidized bed processes 297 granulator, radial particle distribution 353 granule shapes 327 granules, breakage behavior 320–321 gray level histograms, rice grains 29 growth rates – low molecular weight substances 238 – total 346 Guidance for Industry PAT (Process Analytical Technology) 92, 143 Guinier regime 163

h hard shell particles 244–245 head rice yield (HRY) – definition 23

– kinetics 43–44 heat flux heterogeneity, freeze-drying 57–59 heat transfer coefficient, for tubing vials 58–59 hexamethyldisiloxane (HMDSO) 203 hierarchical pore collapse 169 high gain observers, single-vial monitoring 102 high-porosity particles, morphology 235 highly hydrated agricultural products, textured by drying 3 highly insulating and light transmitting (HILIT) aerogel 189 HMDSO. see hexamethyldisiloxane hollow particles – morphology 235 – outlet temperature 251, 252 – SBS-latex 241 Hooke’s law, rice grains 34 horizontal fluidized bed unit 337 hot air drying – freeze drying replacement 195 – RF and carbon cryogels 192 hot melt coating 299 HRY. see head rice yield human recombinant interferon, ultrasound triggered nucleation 96–97 hybrid gels 204 hydrogels – convective drying 175 – vaccuum drying 197 hydrolysis, silica gelation 156 hydrophilic flavors, microencapsulation 255–256 hydrophobic flavors – retention of emulsified 257–260 – spray drying 256–257 hydrophobic silica xerogel 203 hydrophobicity, dry gels 160 hygrocapacity, material structure 300 hygrosensitivity 300

i ice crystal size, distribution of 65, 67–68 ice crystal structure – observation methods 57 – on vertical cross-sections 73 ice fog method, controlled nucleation 70 ice morphology – influence by annealing 67–69 – influence by freezing protocol 63–69 – influence by freezing rate 55–56, 64–66 – influence by supercooling 55, 63

j385

j Index

386

– influence by vial type and filling height 66–67 – and physical quality factors 63–69 ice penetration, pores 165 ice structure, freeze-drying material 55–57 ice sublimation front temperature 62 ice templating 195–197 ICP-AES. see inductively coupled plasma/ atomic emission spectroscopy image analysis techniques – compared to visual inspection 32–33 – fissure formation in rice 28–33 IMC. see internal model control impregnation, catalyst 329–332 in-line product quality control, pharmaceuticals 91–144 inductively coupled plasma/atomic emission (ICP-AES) spectroscopy, vial batch monitoring 110–111 industrial products, textured by drying 3 integral square error (ISE), primary-drying control 131, 134 integrated fluidized beds, particle creation 253 intermediate industrial products, textured by drying 3 internal cracks, rice 25 internal model control (IMC), primary-drying control 133 ISE. see integral square error

k Kalman filter, single-vial monitoring 102 kernel structure, rice grain 22 Knudsen regime, molecular diffusion in 75 Kohlraush-Williams-Watts equation 263

l lactose-based materials, spray-drying 11–12 lactose particles, low molecular weight substances 238 large primary suspension particles 244 large solid particles, suspensions 244 layered structured granules, breakage behavior 320–321 layering – solidified shells 296 – spray fluidized bed processes 297 linear materials, stress-strain relationships 34–36 linoleic acid, emulsion size 280 lipid amount, oil emulsions 278 lipid oxidation, in drying process 7 lipids oxidation 279–280

liquid bridges 303–304 – capillary forces 303–304 – forces 303–304 – particle formulation 296 – tensile strength 309 liquid distribution, open-pore particle network 220 liquid drainage 183 liquid encapsulation, neural networks 357 liquid flow rate, stochastic discrete modeling 367 liquid formulation – composition of 83–84 – of pharmaceuticals 53 liquid/gas interface 175 liquid penetration time 283 liquid pressure, drying methods 176 liquid transport models 212 local moisture content, calculated by CFD 12 low hydrated agricultural products, textured by drying 3 low molecular weight substances – solutions 236–240 – transfer coefficients 237 – vapor pressure 236 low-temperature process, CO2 187–189 LyoDriver, primary-drying control algorithm 129–132 LyoMonitor system, vial batch monitoring 123–124 lyophilization. see freeze-drying LYOTRACK sensor, vial batch monitoring 111–112

m macropore size, tuning 196 macroscopic models, convective drying 211–218 Maillard reactions, in drying process 8 maltodextrin (MD) – investigatations by CLSM 247–248 – mint oil particle size 360 – orange oil particle size 360 – pergamot oil particle size 361 – plasticized surface 305 mannitol particles 240 manometric temperature measurement (MTM) – and PRA 59 – primary-drying control 129, 133 – vial batch monitoring 115, 119–120 mass balance equation 218 mass flow rate of water, secondary drying 135–136

Index mass spectrometers, vial batch monitoring 107–110 material structure, agglomerates 300–301 maximum product temperature 126–127, 134 Maxwell model, viscoelastic gels 217 MC. see Monte Carlo methods MDSC. see modulated DSC mean product temperature, freezedrying 61–63 melting curves, freeze-drying 54–55 mercury porosimetry 168–171 mercury pycnometry 171 meridian cracks – agglomerates 316 – shells 320 mesopore sizes, dry RF gels 192, 194 micro-cracks 181 microencapsulated flavor powders 245–247 microencapsulation – enzymes and oil 269–280 – general remarks on 253–255 – hydrophilic flavors 255–256 – oils 278–280 – by spray drying 269–270 microspheres, wet gels 195 microtomography 172–173 microwave drying, quality preservation 210–211 milled rice, obtainment of 22 mint oil particle size 360 model predictive control (MPC) algorithm, primary-drying control 133 modeling – agglomeration 363–372 – convective drying 211–220 – diffusion 211–217 – of final quality of rice grains 39–45 – flavor release 262 – fluid dynamics in drying chamber 141–142 – macroscopic 211–218 – pore-scale 218–220 – of primary-drying process 125–135 – rigorous 217–218 – Wurster coater 349–357 modulated DSC (MDSC), freeze-drying 55 moisture content. see also water content – residual 137, 282–283 moisture gradients, in freeze-dried matrix 52 moisture profiles 208 moisture sensors, vial batch monitoring 107–113 molecular diffusion, in Knudsen regime 75

momentum equation, discrete particle modeling 350 monitoring. see batch monitoring; primarydrying monitoring; single-vial monitoring; vial monitoring monolithic carbon aerogels, dry gels 161 monomer solution, shrinkage prevention 199 Monte Carlo (MC) methods – agglomeration 363 – coalescence 366 morphology. see also ice morphology – alcohol dehydrogenase (ADH) powder 248 – bovine serum albumin (BSA) 250 – fracture 160 – high-porosity particles 235 – hollow particles 235 – spray-dried particles 231–294 – spray-dried powders 234–236 MPC. see model predictive control MTM. see manometric temperature measurement

n NaCl particles 239 near-infrared (NIR) spectroscopy – residual moisture 283 – single-vial monitoring 100 neural networks – artificial 359 – encapsulation 357–363 nitrogen adsorption 166–168 NMR. see nuclear magnetic resonance non-invasive monitoring techniques, primary drying 98–99 non-invasive sensors, freeze drying of pharmaceuticlas 86 nozzles, spray drying 231 nuclear magnetic resonance (NMR) 283 nucleation – control of 70–73, 94–96 – freezing process 184 nucleation temperature – pharmaceuticals 56 – spontaneous 71 – and sublimation rates 73–74 number density distribution, bed material 326 nutritional properties, and food quality 4

o observation methods, of ice crystal structure 57 oil powders 278

j387

j Index

388

oils – encapsulation and microencapsulation 269–280 – microencapsulation 278–280 – orange 359 – particle size 360–361 – spray drying 278–279 – thermal stress 271–272 – yields 361–362 open-pore particle network, liquid distribution 220 operating conditions, and sublimation kinetics 79–82 orange oil – in granules 359 – particle size 360 – yields 361–362 organic-inorganic hybrid gels 204 organic particle sintering 306 outlet gas handling, apparatus design 332 oxidation – encapsulated flavors 261–269 – encapsulated lipids 279–280 oxidation reaction, D-limonene 267

p paddy. see also rice grains – HRY 23–28 – quality kinetics 40–44 parboiled rice – HRY 23–28 – obtainment of 21 particle. see also specific types of particles particle collisions, in DPM 351 particle formulation – carrier materials 247–251 – material properties 299–324 – operating conditions 324–332 – spray fluidized beds 295–378 particle growth rate, total 346 particle modeling, Wurster coater 349–357 particle morphology, skin-forming materials 234 particle porosity, and agglomeration 369–372 particle retention time 326–327 particle size – distribution evolution 345 – spray-dried powders 236 – two-compartment model 348 particle strength, spray-dried particles 281–282 passive transponders, in primary-drying monitoring 98

PAT. see Guidance for Industry Process Analytical Technology pharmaceuticals – freeze-drying 51–86, 91–144 – key quality factors of freeze-drying 52–63 – liquid formulation 53 – nucleation temperature 56 – polymorphism 84–85 phase transitions, dependence on drying speed 12–13 physical quality factors – and controlled nucleation 70–73 – freeze-dried cake morphology 74–78 – ice morphology 63–69 – importance of temperature control 78–79 – influenced by freeze-drying parameters 63–82 – nucleation temperatures and sublimation rates 73–74 – operating conditions and sublimation kinetics 79–82 PI. see proportional-integral compensator Pirani gauges, vial batch monitoring 106 plastic breakage behavior 318–320 plastic range, drying methods 176 pneumatic dryer, food industry 2 polymer crosslinking, silica gels 200 polymer-like gel networks 157 polymer solutions, vapor pressure 240 polymers, solutions 240 polymorphism, and product quality during freeze-drying 84–85 population balance equation, dispersive growth 344 population balance modeling 324 pore-scale model, convective drying 218–220 pore sizes – distribution 331 – mercury porosimetry 168 – tomography 172 – wet RF gel 166 porod regime 163 porosimetry, mercury 168–171 porosity – dry gels 159 – particles 369–372 – spray dried particles 280–281 – xerogels after pyrolysis 205 porous carrier particles, loading with catalytic active components 329 porous media, standard characterization techniques 155 powdered milk products, rubbery state 11 powders

Index – flavor release 262–267 – layering 298 – microencapsulated flavor 245–247 – particle formation 231 – silica aerogels 191 – spray drying 234–236 – stickiness 260–261 PRA. see pressure rise analysis pressure gradient, differential shrinkage 178 pressure rise analysis (PRA) – key quality factors 59–61 – vial batch monitoring 115, 119–120 pressure rise test (PRT) – primary-drying control 131–132 – secondary drying 135–136 – vial batch monitoring 114–125 pressure sensors, vial batch monitoring 106 primary-drying control – chamber pressure 125–135 – DPE algorithm 129, 132 – feedback logic 134–135 – IMC 133 – in-line 125–135 – ISE 131, 134 – LyoDriver 129–132 – MPC 133 – MTM 129, 133 – PI 134 – PRT 131–132 – shelf temperature 125–135 primary-drying monitoring. see also single-vial monitoring – active transponders 98 – BTM 115 – detection of endpoint 106–113 – DPE algorithm 115 – group of vials 103–105 – in-line 96–125 – MTM 115, 119–120 – non-invasive techniques 98–99 – passive transponders 98 – RTD 97–99 – single vials 99–103 – thermocouples 97–99 – using measurement of sublimation flux 113–114 – using methods based on PRT 114–125 – vial batch 106–125 primary particle properties, agglomeration 321–324 ProCell units, apparatus design 334–335 process analytical technology (PAT), guidance for, in industry 92, 143 process chamber, apparatus design 333

process temperature, spray fluidized beds 327–329 process variables 275–278 product flowability, spray dried particles 282 product quality. see also quality – apparatus design 332–338 – during drying and storage 83–85 – and formulation 83–84 – gained by drying 4 – and polymorphism 84–85 product quality control – continuous freeze-drying 142–143 – control of freezing step 94–96 – control of primary drying 125–135 – in-line 91–144 – monitoring and control of secondary drying 135–138 – monitoring of primary drying 96–125 – quality by design 139–142 product stability, during drying and storage 83–85 proportional-integral (PI) compensator, primary-drying control 134 protein addition, enzyme stabilization 275 protein encapsulation theory 272–273 protein loss, surface adsorption 270 protein solutions – aqueous 51 – spray drying 273 proteins – particle creation 247–251 – stress during the spray drying processes 270–272 – stresses 271 proton transfer reaction mass spectrometry (PTR-MS), flavor release 266 PRT. see pressure rise test PTR-MS. see proton transfer reaction mass spectrometry

q QMS. see quadrupole mass spectrometer quadrupole mass spectrometer (QMS), vial batch monitoring 107–110 quality also food quality; product quality – modeling of convective drying 211–220 quality assessment, gels 162 quality by design 139–142 quality considerations, drying food materials 1–18 quality control, in-line 91–144 quality factors. see also physical quality factors – interactions with transport phenomena 53 quality loss, gel drying methods 174–189

j389

j Index

390

quality preservation – advanced drying techniques 189–211 – carbon cryogels 192 – convective drying 198–210 – cracks from aging 199–201 – ice templating 195–197 – microwave drying 210–211 – RF gels 192, 204–209 – shrinkage reversion 201–204 – silica gels 195–197 – vacuum drying 197–198

r radiation from surrounding, as quality parameter 139 re-agglomeration 284 reconstitution behavior, spray dried particles 283–284 rehydration, during storage 16–17 relative humidity (RH) – flavor release rate 264 – lipid oxidation 279 – and rice fissuring 24–25 relaxation function 216 relaxation process correlation – glass transition temperature 267–269 – temperatures 267–269 residence time distribution 338–344 residual moisture content – infrared irradiation 282 – spray dried particles 282–283 residual water content, monitoring of 91–92, 137–139 resistance thermal detector (RTD), in primarydrying monitoring 97–99 resorcinol-formaldehyde (RF) gels – aging 208–209 – convective drying 206 – freeze drying 191 – linear shrinkage 209 – preparation 156–158 – quality preservation 192, 204–208 – saxs spectra 164 – synthesis 158 restitution coefficient 351 retention – emulsified hydrophobic flavors during spray drying 257–260 – enzyme activity 275 retention phenomenon, at microscopic level 10 retention time, particles 326–327 RF gels. see resorcinol-formaldehyde gels RH. see relative humidity

rice – characterization of mechanical properties 33–39 – HRY 23–28 – image analysis techniques 28–33 – mechanical properties and crack formation 21–45 – tempering time 27–28 rice bran, obtainment of 22 rice grains. see also paddy – dehulling 22 – diametral compression test 37 – failure strength 36–39 – glass transition 34 – gray level histograms 29 – harvesting 21 – Hooke”s law 34 – kernel structure 22 – moisture content 21–22 – stress-strain relationships 34–36 – Young”s modulus 35 rice kernels – cracks in 23 – fissured 23–24 – possible states for 24 – shrinkage and cracking 29 – stress cracks 39 – structure 22 – tension tests 37–38 rice processing yield, definition 23 rice quality – kinetics 40–44 – modeling of 39–45 rolling agglomeration 318 rotary dryer, food industry 2 rotational motions, amorphous glass state 261 rough rice. see rice grains rubbery state, of freeze-dried materials 11

s safety, and food quality 4 Sauter mean diameter, reconstitution behavior 283 SAXS. see small angle X-ray scattering SBS. see styrene-butadiene-styrene scanning electron microscopy (SEM), spray-dried particles 234 secondary drying, monitoring and control 93, 135–138 segmentation method, image analysis techniques 30–31 selective diffusion, in drying process 10

Index selective diffusion theory, hydrophilic flavors 255 self-assembly techniques, gelation 156 SEM. see Scanning electron microscopy sensors – LYOTRACK 111–112 – for mean-product-temperature measurements 61–63 – moisture 107–113 – non-invasive 86 – pressure 106 – soft 101–102 sensory properties, and food quality 4 SEP funtion, sublimation endpoint detection 112 series-of-tanks model 340 shear stress 271–272 shelf temperature. see also temperature – BTM control 128–129 – influence on drying curve 79–80, 82 – primary-drying control 93, 125–135 – as quality parameter 139 shrinkage – aged gels 200 – aging effects 209 – by convective hot air drying 15 – differential 177–180 – diffusion models 213 – drying methods 177 – freezing process 184 – gels 175–177 – irreversible 169 – isotropic 206 – linear 209 – pore sizes 168 – prevention 199–201 – reversion 201–204 – video acquisition 29 silica gelation, condensation 156 silica gels – ice templating 195–197 – polymer crosslinking 201 – preparation 156–157 – shrinkage prevention and cracks by aging 199–201 – shrinkage reversion 201–204 silylation agents 203 Si3N4-suspensions, spray-dried particles 243 single suspended droplet, drying 273–274 single-vial monitoring – extended Kalman filter 102 – high gain observers 102 – in-line 99–103 – soft-sensors 101–102

sinter bridges 304–308 – forces 304–308 – viscous forces 304–308 sintering – mechanisms 305 – organic particle 306 skeletal density – RF gels 191 – shrinkage prevention 199 skin-forming materials, particle morphology 234 small angle X-ray scattering (SAXS), drying of gels 162–164 small solid particles, suspensions 240–244 smart-vial concept 98, 103 SMARTTM Freeze-Dryer 129 sodium benzoate granules – breakage 320–321 – force-displacement curves 320 soft-sensors, single-vial monitoring 101–102 solid network stress, drying methods 176 solid particles – suspensions of large 244 – suspensions of small 240–244 solid pharmaceutical substances, preparation 269 solid phase, diffusion models 212 solids, diffusion rate 241 solids handling, apparatus design 332 solutions – binder content 325–326 – low molecular weight substances 236–240 – polymers 240 – spray drying 273 solvent exchanges – RF and carbon cryogels 193 – TMCS surface modification 203 solvents, supercritical drying 185–187 sound insulation, dry gels 161 sound speed, dry gels 160 space science, dry gels 162 specific surface aera, dry gels 159 spectroscopy methods, single-vial monitoring 100 spontaneous nucleation temperatures 71 spout velocity 352–355 spray-dried food products – flavor retention 253–269 – ingredients 233 spray-dried particles – b-lactoglobulin effects 250 – BSA effects 250 – bulk density 282 – compression 311

j391

j Index

392

– emulsions 246 – freeze spray drying 251–253 – hard shell 244–245 – integrated fluidized beds 253 – lipids oxidation 279–280 – morphology and properties 231–294 – porosity 280–281 – proteins, enzymes and carrier materials 247–251 – quality aspects 280 – schematic view 269 – spray-dried 231–294 – structures 246 – surface structure 239 – suspensions of large solid 244 – suspensions of small solid 240–244 spray-dried powders – cross-sectional structures 248 – flavor release 262–267 – morphological characteristics 233 – morphology 249 – morphology classification 234–236 – outer structural changes 266 – stickiness 260–261 spray dryer, food industry 2 spray drying – emulsified hydrophobic flavors 256–257 – encapsulation and microencapsulation of enzymes and oil by 269–280 – enzyme stabilization 274–275 – flavor encapsulation 255 – of lactose-based materials 11–12 – microencapsulation 269–270 – oil emulsions 278–279 – particle creation 251–253 – process variables effect on the stabilization of enzymes 275–278 – protein encapsulation theory 272–273 – protein solutions 273 – retention 257–260 – stress on proteins 270–272 spray-drying system – scheme of 232 – stresses 271 spray fluidized bed encapsulation, aromatic oils 358 spray fluidized bed processes 297 spray fluidized beds – apparatus design 332–357 – particle formulation 295–378 – periphery 332 spray system, apparatus design 333 sprayed solutions, binder content 325–326 springback, convective drying 204

stability, during drying and storage 83–85 stability diagram, of foods 2 stabilizer, role of 83 starch – gelinization 245 – mint oil particle size 360 – orange oil particle size 360 – pergamot oil particle size 361 state diagram, freeze-drying 54–55 stickiness, spray-dried powder 260–261 stochastic discrete modeling 364–367 – agglomeration 363–372 storage – of food materials 16–17 – product quality and stability during 83–85 – release and oxidation of encapsulated flavor 261–269 storage stability, glass temperature influence 261–262 strain difference, diffusion models 213 strength – agglomerates 299–315 – particles 281–282 stress – and differential shrinkage 177–180 – diffusion models 213 – on proteins during drying 270–272 – simulations 215 stress cracks, in rice kernels 39 stress-strain relationships, rice grains 34–36 styrene-butadiene-styrene (SBS) latex 241 subcritical drying, quality preservation 189–190 sublimation chamber. see also drying chamber – gas pressure and drying curve 79–81 – heat flux heterogeneity in 57–59 – total gas pressure 117–118 sublimation endpoint detection, vial batch monitoring 106–113, 121–122 sublimation flux measurement, vial batch monitoring 113–114 sublimation front temperature 62 sublimation kinetics, and operating conditions 79–82 sublimation rates, and nucleation temperatures 73–74 sudden expansion, combined with drying 18 sun-cracks, rice 24 supercapacitors, dry gels 161 supercooling, and ice morphology 55, 63 supercritical drying – gels 159, 185–189 – heating rate 185 – initial solvent 185–187

Index – RF and carbon cryogels 192 – washing step 187 surface cracking, during drying 16 surface cracks, rice 25 surface modification – quality preservation 201–204 – TMCS 203 surfactants, enzyme activity retention 275 suspension droplets – drying 273–274 – glass deposition 245 suspensions – fine glass filament 272 – flash spray drying 242 – large solid particles 244 – small solid particles 240–244 syneresis, silica gelation 157

total gas pressure, sublimation chamber 79–81, 117–118 transmission electron microscopy, characterization of gels 171 transport phenomena, interactions with quality factors 53 trimethylchlorosilane (TMCS) – shrinkage reversion 202 – solvent exchanges 203 tunable diode laser absorption spectroscopy (TDLAS), vial batch monitoring 113–114 tunnel conveyor dryer, food industry 2 two-compartment model – fluidized bed 347 – particle size distributions 348 two population balance equations 348

u t TDLAS. see tunable diode laser absorption spectroscopy temperature. see also chamber temperature; shelf temperature – influence on crack formation in rice 26–27 temperature control, and physical quality 78–79 temperature increase, and biochemical reactions in foods 5–9 temperature remote interrogation system (TEMPRIS) 98 tempering time, rice 27–28 tension tests, rice kernels 37–38 TEOS. see tetraethoxysilane tert-butanol, microwave drying 210 tert-butanol (CH3)3COH, freeze drying 191 tetraalkoxysilane Si(OR)4 156 tetraethoxysilane (TEOS) 156 tetramethoxysilane (TMOS) 156 thermal conductivity gauges, vial batch monitoring 106 thermal effects, stochastic discrete modeling 367–369 thermal stress, enzymes and oil 271–272 thermocouples – insertion in vials 61 – in primary-drying monitoring 97–99 thermograms, with ultrasound triggered nucleation 71 thermoporometry 164–166 three-layer artificial neural network 359 time step length, MC methods 364 TMCS. see trimethylchlorosilane TMOS. see tetramethoxysilane

ultrasonic atomizers, spray drying 231 ultrasonic irradiation, gelation 195 ultrasound, effect on structural and morphological properties 72–73 ultrasound triggered nucleation – controlled 70–72, 95–96 – human recombinant interferon 96–97 – thermograms 71 undercooling 165

v vacuum drying – gels 184 – quality preservation 197–198 – RF and carbon cryogels 192 van der Waals forces – adhesion 299 – agglomerates 301–303 ventilated cabinets drying, food industry 2 vial monitoring – single vials 99–103 – vial batches 106–125 – vial groups 103–105 vial type, influence on ice morphology 66–67 vials – for freeze-drying of pharmaceuticals 51–86 – heat transfer coefficient 58–59 video acquisition, of shrinkage and cracking of rice kernels 29 viscoelastic gels, Maxwell model 217 viscosity – shift factor 301 – stochastic discrete modeling 367 viscous forces – adhesion 299 – between amorphous particles 304–308

j393

j Index

394

viscous solid network, differential shrinkage 179 visual inspection, compared to image analysis 32–33 vitamin C. see also ascorbic acid – as a quality index in drying process 6 vitreous transition, freeze-drying 54–55 vitrification concept, product stabilization 83 volatile flavors 255 volume-averaged liquid density, transport models 212 VPO precursors 244

w Washburn equation 168 water activity – decreasing 1 – and enzymatic activity 8 – and stability diagram of foods 2 water concentration, time evolution measurements 107–113, 136–137 water content. see also moisture content – residual 91–92, 137–139 water flow rate, secondary drying 135–136 water layer hypothesis, protein encapsulation 271 water replacement hypothesis, protein encapsulation 271–273 water-soluble crystalline substances 304 water substitute concept, product stabilization 83 water vapor mass transfer resistance – and freeze-dried cake morphology 74–76 – and freeze-dried layer thickness 69, 76

water vapor pressure, in drying chamber 62 wet gels – characterization 162–166 – preparation 156–158 white rice, obtainment of 22 whole-batch monitoring, freeze-drying 106–125 Williams, Landel and Ferry (WLF) equation 267, 301 wireless probes, in primary-drying monitoring 98 Wurster coater – discrete particle modeling 349–357 – gap distance 355–357 – geometry 352 – particle positions 353–356 – schematic representation of 334 – velocity distributions 353–356

x X-ray tomography, drying gels 172 xerogels – definition 159 – mercury porosimetry 170 – vaccuum drying 198

y Young’s modulus – rice grains 35 – time-dependent 36

z zeolite agglomerates 317–319